Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Transcriptome analysis of clock disrupted cancer cells reveals differential alternative splicing of cancer hallmarks genes

Transcriptome analysis of clock disrupted cancer cells reveals differential alternative splicing... www.nature.com/npjsba ARTICLE OPEN Transcriptome analysis of clock disrupted cancer cells reveals differential alternative splicing of cancer hallmarks genes 1,2,3 1,2,3 1,2,3✉ Deeksha Malhan , Alireza Basti and Angela Relógio Emerging evidence points towards a regulatory role of the circadian clock in alternative splicing (AS). Whether alterations in core- clock components may contribute to differential AS events is largely unknown. To address this, we carried out a computational analysis on recently generated time-series RNA-seq datasets from three core-clock knockout (KO) genes (ARNTL, NR1D1, PER2) and WT of a colorectal cancer (CRC) cell line, and time-series RNA-seq datasets for additional CRC and Hodgkin’s lymphoma (HL) cells, murine WT, Arntl KO, and Nr1d1/2 KO, and murine SCN WT tissue. The deletion of individual core-clock genes resulted in the loss of KO KO KO circadian expression in crucial spliceosome components such as SF3A1 (in ARNTL ), SNW1 (in NR1D1 ), and HNRNPC (in PER2 ), KO which led to a differential pattern of KO-specific AS events. All HCT116 cells showed a rhythmicity loss of a crucial spliceosome gene U2AF1, which was also not rhythmic in higher progression stage CRC and HL cancer cells. AS analysis revealed an increase in alternative first exon events specificto PER2 and NR1D1 KO in HCT116 cells, and a KO-specific change in expression and rhythmicity KO KO pattern of AS transcripts related to cancer hallmarks genes including FGFR2 in HCT116_ARNTL , CD44 in HCT116_NR1D1 , and KO MET in HCT116_PER2 . KO-specific changes in rhythmic properties of known spliced variants of these genes (e.g. FGFR2 IIIb/FGFR2 IIIc) correlated with epithelial-mesenchymal-transition signalling. Altogether, our bioinformatic analysis highlights a role for the circadian clock in the regulation of AS, and reveals a potential impact of clock disruption in aberrant splicing in cancer hallmark genes. npj Systems Biology and Applications (2022) 8:17 ; https://doi.org/10.1038/s41540-022-00225-w INTRODUCTION Discs Large MAGUK Scaffold Protein 1 (DLG1) leading to the production of oncogenic isoforms, or to the reduction of the Alternative splicing (AS) is a central event in transcriptional tumour-suppressing isoforms of those genes . Also, aberrant AS regulation leading to distinct mRNAs being generated from a 1,2 in other members of the SR splicing factor and SR-like families, single gene, and is thus a major contributor for protein diversity . including SRSF4 and Transformer 2 Beta Homolog (TRA2β) have Approximately 95% of human genes undergo AS in a tissue type- been shown to regulate cancer proliferation and metastasis by specific manner , and AS resulting products lead to the affecting the splicing of genes involved in epithelial-mesenchymal production of proteins involved in numerous biological processes 4 5 transition (EMT, e.g., CD44) in breast cancer . Mounting evidence including cell cycle and metabolism . Splicing factors (SFs) mark suggests as well an association between AS regulation and the the splice sites, facilitate spliceosome assembly during the splicing 18,19 circadian clock in various organisms in particular in mammals . process, and can regulate AS decisions and thus the resulting Circadian rhythms are driven by an endogenous oscillator with splicing product(s) . Dysregulations in splicing account for the an approximate period of 24 h and regulate the timing of cellular production of non-functional, or with the incorrect function, processes in most organisms . These rhythms are generated by a protein products with an impact on the onset and progression of 7–12 Transcriptional-Translational Feedback Loop (TTFL) in which Brain several diseases among which cancer . A systematic evaluation and muscle ARNT-Like 1 (BMAL1), also known as ARNTL forms a of 11 different tumour types revealed a subset of transcript heterodimer with Circadian Locomoter Output Cycles Kaput isoform switches associated with gain and/or loss of protein (CLOCK) and regulates the transcriptional activation of Period 1/ domains, which are associated with the hallmarks of cancer . 2/3 (PER 1/2/3), and Cryptochrome 1/2 (CRY 1/2), which inhibit Known examples of AS failures related to cancer include the BMAL1-CLOCK-mediated transcription . In addition, Nuclear aberrant splicing in the Vascular Epithelial Growth Factor (VEGF), Receptor Subfamily 1 Group D Member 1, 2 (NR1D1/2) and RAR essential for angiogenesis, resulting in one transcript isoform with anti-angiogenic properties in cancer . Another example is CD44, Related Orphan Receptor A, B, C (RORA/B/C) regulate BMAL1 which encodes for a cell-surface protein and can act as both transcription and contribute to the fine-tuning of its expression. tumour-suppressing, as well as invasiveness-promoting molecule, The circadian clock regulates the expression of genes and proteins depending on its different splice variants . In addition, alterations related to several cancer hallmarks like cell proliferation, in the expression of several SFs can impact the splicing of their metabolism, and DNA damage . Disruption of circadian rhythms target genes, for e.g., the differential expression of the splicing is associated with increased risk of cancer, including prostrate, factor Serine and Arginine Rich Splicing Factor 6 (SRSF6) affects breast, and colon cancer . the splicing of several oncogenes including the Insulin Receptor Recent work points to a complex cross-talk between the circadian 24–31 (INSR), MAPK Interacting Serine/Threonine Kinase 2 (MKNK2) and clock and AS events underlying different cancer types .Changes Institute for Theoretical Biology (ITB), Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt – Universität zu Berlin, and Berlin Institute of Health, Berlin 10117, Germany. Molecular Cancer Research Center (MKFZ), Medical Department of Hematology, Oncology, and Tumor Immunology, Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin Humboldt – Universität zu Berlin, and Berlin Institute of Health, Berlin 10117, Germany. Institute for Systems Medicine, Faculty of Human Medicine, MSH Medical School Hamburg, Hamburg 20457, Germany. email: angela.relogio@charite.de Published in partnership with the Systems Biology Institute 1234567890():,; D. Malhan et al. in amplitude and peak phase of SFs including hnRNPs and RBM RESULTS proteins in colorectal cancer (CRC) cells were correlated with altered Rhythmicity analysis of mammalian RNA-seq datasets showed splicing outcome of target genes like VEGFA and CD44 .Alater a KO-specific effect in the circadian phenotype both at the study in the same cell lines reported rhythmic alternative splicing gene and at the transcript levels events of cancer-related genes like Cold inducible RNA binding In mammals, many different biological processes like metabolism protein (CIRBP) and Poly(RC) Binding Protein 2 (PCBP2) . Clock- or the immune system are regulated by the endogenous circadian mediated alternative splicing events in the cell differentiation 32 clock . To investigate the effect of clock disruption on the marker Interferon Regulatory Factor 4 (IRF4) were reported in an circadian phenotype (circa 24 h rhythmicity) and the subsequent in vitro model of Hodgkin lymphoma (HL) .Inthe same HL cell effects in biological processes, we analysed recent datasets from lines, clock-mediated changes in the expression of transcript our group for three core-clock knockout (KO) cell lines isoforms were reported in elements associated with the tumour KO KO KO (HCT116_ARNTL , HCT116_NR1D1 , and HCT116_PER2 ) and necrosis factor (TNF) pathway, which controls cell proliferation, compared their gene expression pattern to the corresponding migration, and apoptosis . Rhythmic alternative splicing of the HCT116_WT (Fig. 1a, Supplementary Fig. 1d). Previous in vitro and splicing factor U2AF26 controls the stability of Per1 in mice .A in vivo studies showed an effect of ARNTL, PER2 and NR1D1 KO on splice variant of human BMAL1 (hBMAL1a) is known to act opposite the rhythmic expression of core-clock genes (mostly via biolumi- to its canonical isoform (hBMAL1b) and functions as a negative 33,34 nescence recordings, or mouse studies) , which are in regulator of the circadian clock . However, the impact in alternative agreement with our findings regarding the loss (ARNTL KO) and splicing events resulting from clock alterations in cancer remains period alterations (PER2 and NR1D1 KO) in the respective unexplored. oscillation profiles. Although, it is important to notice that the In this study, we analysed time-course RNA-seq datasets assessment of rhythmicity via in vitro bioluminescence measure- generated from three HCT116 core-core knockouts (ARNTL, PER2, ments of ARNTL promoter activity, or via in vivo analysis of NR1D1) and wild type (WT) cells, to investigate the impact of clock movement in mice, is not directly comparable with RNA-seq time disruption on AS events in CRC cell lines. In addition, we retrieved course measurements. publicly available time-course RNA-seq datasets from human CRC To pinpoint specific changes due to the KO of core-clock genes, cell lines (SW480, SW620), HL cell lines (HDMYZ, L1236), and we additionally gathered publicly available time-course RNA-seq mouse Arntl KO, Nr1d1/2 KO, and SCN (Suprachiasmatic nucleus) datasets for human CRC cell lines (SW480: derived from the WT tissue. Our study revealed that deletion of core-clock genes primary tumour, SW620: derived from a metastasis site, from the (ARNTL, PER2 or NR1D1) altered the circadian rhythmic pattern of same patient), HL cell lines (HDMYZ: stage III, L1236: stage IV), SFs, as well as the resulting AS events in a KO-specific manner. We murine WT and Arntl KO, murine WT and Nr1d1/2 KO, murine SCN observed a common rhythmicity loss of crucial SFs such as U2 tissue, murine WT and Cry1/2 KO, and murine WT and Per1/2 KO, KO Small Nuclear RNA Auxiliary Factor 1(U2AF1) in all three HCT116 , resulting in a total of 222 samples (Fig. 1a). To examine if the as well as in SW620, and L1236 cells. Moreover, we found a phase rhythmicity pattern varies when analysed at the transcript level vs shift in circadian transcripts of SFs such as Receptor of Activated C gene level, we extracted 24 h rhythmic sets both at gene and Kinase 1 (RACK1), Heterogeneous nuclear ribonucleoprotein D transcript levels in six-mammalian time course RNA-seq datasets (HNRNPD) involved in pre-mRNA processing and Serine and (Fig. 1b–d, Supplementary Figs. 1–3). A higher number of 24 h Arginine rich splicing factor 5 (SRSF5), component of the rhythmic features were found at transcript level vs gene level in all KO KO spliceosome machinery, in HCT116_ARNTL , HCT116_NR1D1 , datasets. The number of circadian (~24 h period) expressed genes KO KO and HCT116_PER2 vs. WT, respectively. This led to a distinct and transcripts decreased in all three HCT116 vs. WT (Fig. 1b–d). At the gene level, we identified 1261, 443, 866 and 525 circadian pattern of AS events (specifically in gain and loss of splicing KO KO KO gene sets in HCT116_WT, HCT116_ARNTL , HCT116_NR1D1 ,and events) in HCT116 datasets vs. HCT116_WT. Differential KO HCT116_PER2 , respectively (Fig. 1b; left panel). We observed KO- rhythmicity analysis revealed that the majority of phase shifted KO specific loss in circadian expression of crucial genes such as Keratin transcript pairs consisted of protein-coding biotypes. In HCT116 8(KRT8), which maintains cellular structural integrity, Transmem- cells, genes with phase-shifted transcript pairs were related to KO brane and Coiled-Coil Domain Family 1 (TMCC1), which promotes apoptotic signalling (HCT116_ARNTL ), macrophage proliferation KO endoplasmic reticulum-associated fission, and HECT and RLD (HCT116_NR1D1 ), and negative regulation of chromatin organi- KO Domain Containing E3 Ubiquitin Protein Ligase 3 (HERC3), which zation (HCT116_PER2 ). Moreover, we identified differentially KO enables protein-transferase activity in HCT116_ARNTL , rhythmic transcript isoform pairs of target genes associated with KO KO HCT116_NR1D1 , and in HCT116_PER2 , respectively. At tran- cancer hallmarks. These included Casein Kinase 2 Alpha 1 script level, we observed 5499 (corresponding to 4063 genes), 3039 (CSNK2A1), related to invasion, metastasis, tumour proliferation, (2485 genes), 3519 (2899 genes), 2850 (2345 genes) circadian KO in HCT116_WT and HCT116 . We further identified phase shifted KO KO transcripts in HCT116_WT, HCT116_ARNTL , HCT116_NR1D1 , transcript pairs from cancer hallmarks-related genes e.g., tran- KO and HCT116_PER2 , respectively (Fig. 1c; left panel). Among KO- scripts pairs from MET Proto Oncogene (MET), which controls specific loss of circadian expressed transcripts, we saw a loss of KO several cancer hallmarks, in HCT116_PER2 . Finally, we observed circadian transcript expression of C-Terminal Binding Protein 1 KO-specific altered expression and rhythmicity pattern of tran- (CTBP1), which is a transcriptional repressor, HRAS an oncogene scripts associated with cancer hallmarks related genes such as in involved in signal transduction pathways, and Calmodulin 1 Fibroblast Growth Factor Receptor 2 (FGFR2), which promotes (CALM1), which controls several proteins including enzymes and KO tumour progression seen in HCT116_ARNTL , Caspase 8 (CASP8) ion channels, through calcium-binding. All these genes modulate KO KO which regulates extrinsic apoptotic pathway seen in cellular proliferation in HCT116_ARNTL , HCT116_NR1D1 ,and in KO KO HCT116_NR1D1 , and HRas proto-oncogene, GTPase (HRAS) HCT116_PER2 , respectively. involved in cell proliferation, growth and apoptosis detected in Moreover, differences in the percentage of biotypes (e.g., KO HCT116_PER2 . The evaluation of coding exon sequences for protein-coding biotype) were also observed between circadian selected cancer hallmarks revealed the altered rhythmicity in gene and transcript sets (Fig. 1b, c; right panel). For example, 1089 spliced variants associated with cancer progression such as genes (~86%) and 3509 transcripts (~63%) with protein-coding skipped exon 14 variant of MET. Our findings point to a direct biotype were circadian expressed in HCT116_WT. The comparison impact of circadian clock disruption in AS events related to cancer of rhythmicity at gene and transcript levels in HCT116 datasets hallmarks, which may support cancer onset or progression. showed a roughly linear relationship between the number of npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute 1234567890():,; D. Malhan et al. Fig. 1 RNA-seq data analysis showed differences in the circadian phenotype between gene- and transcript expression values. a Schematic representation of the pipeline for RNA-seq data analysis. The input data consisted of 222 samples from eight mammalian time- course RNA-seq data. Murine datasets from Per1/2 KO (GSE171975) and Cry1/2 KO (GSE135898) were only analysed at the gene level. b– d Classification of 24 h rhythmic features in HCT116 datasets at (b) gene level and (c) transcript level showed a decline in the total number of circadian expressed genes/transcript-sets and changes in their biotypes due to core-clock disruption. d Scatter plot represents each of the HCT116 WT and KO cells according to the total number of 24 h rhythmic genes (x-axis) and the total number of 24 h rhythmic transcripts (y- axis). KO KO circadian genes and circadian transcripts (Fig. 1d). Similar HCT116_ARNTL , 521 genes (~60%) in HCT116_NR1D1 , and KO differences in gene- and transcript-level rhythmicity phenotype 274 genes (~52%) in HCT116_PER2 were rhythmic both at gene were also seen in other human CRC cell lines (SW480; SW620), HL and transcript levels (Supplementary Fig. 1a). We further analysed cell lines (HDMYZ; L1236), murine WT and Arntl KO, murine WT circadian discrepancies between the expression pattern of some and Nr1d1/2 KO, and murine SCN WT tissue datasets (Supple- genes and their corresponding transcripts, for e.g., HIF1A was mentary Figs. 2, 3). Within the set of circadian rhythmic genes, 846 found to be circadian expressed whereas its transcripts were genes (~67%) in HCT116_WT, 192 genes (~43%) in arrhythmic in HCT116_WT. Here, the summarization of arrhythmic Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. Fig. 2 Circadian expression of splicing factors and spliceosome machinery related transcripts in mammalian RNA-seq datasets. Median KO normalized phase-sorted heatmaps (left) and acrophase bins (right) of SFs circadian transcripts (tr.) in (a) HCT116_WT and HCT116 cells (upper panel), SW480 and SW620 cell lines (lower left panel), HDMYZ and L1236 cell lines (lower right panel), b murine WT and Arntl KO from distal bronchiolar epithelium tissue (upper left panel) and from primary tracheal epithelial cells (upper right panel), murine WT and Nr1d1/2 KO from epithelial cells (lower left panel), and murine SCN WT tissue (lower right panel). transcripts resulted in a rhythmic gene. Minichromosome Main- Moreover, our analysis also shows that the KO of core-clock genes tenance Complex Component 2 (MCM2) was rhythmic both at altered the ratio, as well as the circadian rhythmicity expression pattern of genes and transcripts in a KO-specific manner. Hence, gene and transcript levels in HCT116_WT. In contrast, Interleukin 18 (IL18) was identified as arrhythmic at the gene level whereas its we focused on transcript level in our subsequent analysis. transcripts were circadian expressed in HCT116_WT. In this case, the summarization of rhythmic transcripts resulted in an Alterations in core-clock elements influenced the rhythmic arrhythmic gene (Supplementary Fig. 1a; lower panel). Further, expression of splicing factors we analysed the changes in the number of transcripts per To investigate possible alteration in SFs rhythmicity resulting from expressed gene and the number of ~24 h transcripts per gene in clock disruption, we examined the expression of SF transcripts and KO HCT116 vs. WT (Supplementary Fig. 1b, c). The number of their rhythmic properties (Fig. 2). We compiled a total of 534 24–26,35,36 expressed transcripts per gene showed variation in all three human SF genes from previously published studies , KO 37 38 HCT116 vs. WT. For instance, 28 transcripts of Heterogeneous Spliceosome database , and SpliceAid 2 database (Supplemen- Nuclear Ribonucleoprotein H1 (HNRNPH1), an RNA binding tary file 1), and generated orthologous sets of human SFs for protein, were expressed in HCT116_WT. Whereas, 25 transcripts mouse (521 genes) using Ensembl Biomart (Supplementary file KO from HNRNPH1 were expressed in HCT116_ARNTL , 21 transcripts 2), and mapped the curated list of SFs against the circadian KO KO in HCT116_NR1D1 , and 24 transcripts in HCT116_PER2 . Out of transcripts extracted from the RNA-seq datasets as depicted in the the expressed transcripts of HNRNPH1, 5 transcripts in HCT116_WT phase sorted heatmaps and acrophase bin plots (Fig. 2). We saw a KO and 1 transcript in HCT116_PER2 were circadian expressed. decrease in the number of circadian expressed SF transcripts and KO Altogether, our analysis confirmed previous findings from our also changes in their acrophase distribution in HCT116 group , and showed that indeed a transcript-level analysis can compared to HCT116_WT cells (Fig. 2a; upper panel). KO provide new results that an be masked by gene-level analysis. HCT116_PER2 showed the lowest number of circadian rhythmic npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. KO SF transcripts (143 transcripts; 104 genes) followed by spliceosome formation (Fig. 3a). For HCT116_NR1D1 , differential KO KO HCT116_NR1D1 (188 transcripts; 135 genes), HCT116_ARNTL rhythmic SFs transcripts were involved in pre-mRNA processing (204 transcripts; 150 genes), and HCT116_WT (341 transcripts; 216 like HNRNPD and translational repressor like Cold Inducible RNA KO KO KO genes) datasets. Both HCT116_NR1D1 and HCT116_PER2 Binding Protein (CIRBP) (Fig. 3b). Whereas in HCT116_PER2 ,we showed a shift in the SFs acrophase when compared with found differentially rhythmic transcripts of SFs like SRSF5, which is HCT116_WT cells. a part of the spliceosome machinery (Fig. 3c). To find a possible association between the decrease in the To further understand the discrepancies in SFs rhythmicity, we number of rhythmic SF transcripts and cancer progression, we extracted a list of SFs with at least one circadian transcript in analysed other CRC cell lines (SW480, SW620) and HL cell lines HCT116_WT, SW480, HDMYZ, murine WT (from Arntl KO), murine (HDMYZ, L1236). Indeed, we observed a decrease in the number WT (from Nr1d1/2 KO), and mapped them with their correspond- of rhythmic SF transcripts in the metastasis-derived colon cancer ing knockout/altered condition (e.g. primary tumour- vs. (SW620: 216 transcripts; 153 genes) and in stage IV HL (L1236: 215 metastasis-derived cells). We produced categorical heatmaps to transcripts; 156 genes) as compared with the primary colon cancer visualize the loss of rhythmically expressed SFs in each dataset (SW480: 378 transcripts; 212 genes) and stage III HL (HDMYZ: 325 compared with controls (Fig. 3d–h). 216 SFs with at least one transcripts; 201 genes) cell lines, respectively (Fig. 2a; lower panel). rhythmic transcript were observed in HCT116_WT, and less than KO Also, the shift in acrophase distribution of circadian rhythmic SF 50% were rhythmic in HCT116 (Fig. 3d; left panel). Out of 216 transcripts was observed in both SW620 and L1236 cell lines SFs with at least one circadian transcript in HCT116_WT, we KO KO compared with their controls. To further examine the impact of observed 85 SFs in HCT116_ARNTL , 79 SFs in HCT116_NR1D1 , KO each KO on SFs rhythmicity, we examined rhythmic changes in SFs and 66 SFs in HCT116_PER2 with at least one circadian KO transcripts within murine RNA-seq datasets (Fig. 2b). In contrast, transcript. Among the three HCT116 cells, there were 69 slightly higher number of rhythmic SF transcripts were observed common SFs with no rhythmic transcript (Fig. 3d; right panel). in Arntl KO model of distal bronchiolar epithelium (208 transcripts; Similarly, the human CRC cell line (SW620), HL cell line (L1236), 151 genes) and primary tracheal epithelial cells (48 transcripts; 44 murine Arntl KO, and murine Nr1d1/2 KO also showed a loss of genes) compared with their WT (175 transcripts; 138 genes and 46 rhythmic SFs compared with their controls (Fig. 3e–h). All three KO transcripts; 45 genes, respectively) (Fig. 2b; upper panel). Whereas, HCT116 cells showed loss of rhythmicity for U2AF1, an important Nr1d1/2 KO mouse model showed lower number of rhythmic SF RNA splicing mediator gene, which also lost rhythmicity in SW620 transcripts (101 transcripts; 87 genes) compared with WT (107 and L1236 cells, as compared to their corresponding lower grade transcripts; 87 genes) (Fig. 2b). SCN tissue, the master clock, cancer cell lines. We observed KO-specific loss of rhythmicity in showed the highest number of rhythmic SF transcripts (767 different SF transcripts such as RNA Binding Motif Protein 5 transcripts; 384 genes) in mouse WT datasets (Fig. 2b; lower panel, (RBM5), Splicing factor 3a subunit 1 (SF3A1), and Serine and for the complete list of circadian transcripts and their acrophases Arginine Rich Splicing Factor 3 (SRSF3), involved in the spliceo- KO see Supplementary Fig. 4). In addition, we also observed loss in some machinery, uniquely in HCT116_ARNTL . While, KO 24 h rhythmic SFs due to Cry1/2 deletion (Supplementary Fig. 5b; HCT116_NR1D1 resulted in rhythmicity loss of SNW Domain left panel) and Per1/2 deletion (Supplementary Fig. 5b; right panel) Containing 1 (SNW1), another spliceosome component, and KO vs. their controls in mouse datasets. HCT116_PER2 resulted in the loss of rhythmic transcripts of The KOs resulted in a shift in the acrophase bin distribution of HNRNPC, involved in pre-mRNA processing. Taken together, our circadian rhythmic SF transcripts in particular for NR1D1 and PER2 results point to alterations in rhythmic properties of SFs in HCT116 KO. We further carried out pairwise comparisons of circadian SF cell lines, which were KO-specific and were also present in KO transcripts between HCT116_WT and HCT116 to identify datasets derived from other cells and tissues, upon perturbation of commonly circadian rhythmic SF transcripts, and to obtain the core-clock. differentially rhythmic SF transcripts (phase shift ≥3 h) in the KO HCT116 compared with HCT116_WT (Fig. 3a–c). A total of 25 Disruption of circadian clock elements influenced alternative circadian SFs transcripts (23 genes) were found rhythmic in splicing events across human and murine datasets KO HCT116_WT and HCT116_ARNTL , out of those, 7 transcripts Aberrant AS events have been reported to be associated with showed differential rhythmicity (Fig. 3a). Besides alterations in the different types of cancer . We investigated the impact of clock expression profile of SF transcripts, we observed phase shift in disruption in our datasets on AS events with a potential impact in other transcripts, as well as loss of rhythmic oscillations in PER1, cancer onset and progression. We used SUPPA to calculate the PER2 and DBP gene with ARNTL disruption in HCT116 cells proportion spliced-in (PSI) of seven basic modes of AS: (1) (Supplementary Fig. 6a). Alternative 3′ splice site (A3-event), (2) Alternative 5′ splice site Differential rhythmicity analysis between HCT116_WT and KO (A5-event), (3) Alternative first exon (AF-event), (4) Alternative last HCT116_ARNTL cells resulted in a higher number of transcripts exon (AL-event), (5) Mutually exclusive event (MX-event), (6) with phase difference greater than 9 h whereas for most Retention intron event (RI-event), and (7) Skipping exon event (SE- differentially rhythmic splicing factor transcripts a phase differ- event). ence between 6 and 9 h was observed (Supplementary Fig. 6b–f). KO We determined the overall number of genes showing For HCT116_WT vs. HCT116_NR1D1 , we found 33 circadian significant AS events (0.1< PSI < 0.9), and found that most genes rhythmic transcripts (28 genes) and 11 circadian rhythmic had a commonly predominant AS event (SE-event) in all human transcripts to be differentially rhythmic (Fig. 3b). The comparison KO RNA-seq datasets (Supplementary Fig. 7). To further evaluate the of HCT116_WT and HCT116_PER2 resulted in a total of 28 changes in AS events caused by the KO of core-clock genes, we circadian expressed transcripts (26 genes) and out of those 4 classified the set of genes that were alternatively spliced in the transcripts showed differential rhythmicity (Fig. 3c). Only a single WT, but not in the KOs (following the PSI criteria), as “event loss” in transcript from the gene survival of motor neuron 1 (SMN1), which its corresponding KO group. Similarly, uniquely identified sets of plays a catalyst role in the assembly of small nuclear ribonucleo- genes alternatively spliced in a KO cell line were labelled as “event proteins (snRNPs) was found to be rhythmic in all HCT116 WT and gain” (see “Methods” for details). Bar plots were used to depict the KO datasets and it showed the same phase, but different KO total number of genes in the categories loss and gain for each amplitude within HCT116 . Out of seven differentially rhythmic KO transcripts in HCT116_ARNTL vs. WT, were SFs like RACK1, which type of splicing event (Fig. 4). We mapped the list of AS genes with is a component of the ribosomal subunit and Splicing factor the genes containing at least one circadian transcript and their proline and glutamine rich (SFPQ), essential at the early stages of biotypes. For better visualization, we separated the genes Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. Fig. 3 Clock disruption affected rhythmicity of splicing factors in human and murine datasets. a–c Circular plots depict the distribution of KO peak phases of overlapping differentially rhythmic SF transcripts between (a) HCT116_WT and HCT116_ARNTL ,(b) HCT116_WT and KO KO KO HCT116_NR1D1 , and (c) HCT116_WT and HCT116_PER2 . d–h Categorial heatmaps represent the loss of rhythmic SFs in (d) HCT116 vs. WT, (e) SW620 vs. SW480, (f) L1236 vs. HDMYZ, (g) mouse Arntl KO vs. WT in distal bronchiolar epithelium (left) and primary epithelial cells (right), and (h) Nr1d1/2 KO vs. WT. Green colour indicates SFs with at least one ~24 h transcript whereas grey colour indicates SFs with no rhythmic transcript. The numbers above the categorial heatmaps indicate the total number of SFs (genes) with rhythmic transcript(s). according to whether or not they contained transcripts with carried out a functional enrichment analysis, and found biological KO protein-coding biotype. All AS events in HCT116 cells except for processes such as protein acetylation enriched in common AF- KO KO the AF-event in HCT116_NR1D1 and HCT116_PER2 showed gain candidates, positive regulation of cellular response to TGF- KO higher number of genes present in the loss of splicing events vs. beta stimulus enriched in HCT116_NR1D1 unique candidates, gain of splicing events (Fig. 4a–c). However, a large increase in the and endothelial cell proliferation unique AF-event in KO number of genes with AF-gain was observed in both HCT116_PER2 candidates (Supplementary Fig. 8). KO HCT116_NR1D1 (gain vs. loss: 75.8% vs. 24.4%) and A similar number of genes were observed in loss and gain of KO HCT116_PER2 (gain vs. loss: 74.2% vs. 25.7%) (Fig. 4b–c). splicing events within the mouse Arntl KO dataset (Fig. 4f). In the Interestingly, higher number of genes with gain in splicing events case of Nr1d1/2 KO model, a slightly higher number of genes were were observed in the metastasis CRC cell line (SW620) and in stage associated with a loss of splicing events than to a gain of splicing IV HL cell line (L1236) (Fig. 4d–e). We further investigated if the events (Fig. 4g). However, an increase in gain of AF-event as seen same genes were alternatively spliced through AF-event (gain) in KO in HCT116_NR1D1 was not observed in the mouse Nr1d1/2 KO. KO KO both HCT116_NR1D1 and HCT116_PER2 . Subsequently, we Taken together, we observed a distinct pattern in the number of filtered for genes with at least one protein-coding transcript, npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. Fig. 4 Local alternative splicing analysis revealed changes in the pattern of loss and gain of all seven splice modes in human and murine KO KO datasets. Bar plots depict the total number of genes (x-axis) with loss and gain of splicing (a–c) in the HCT116_ARNTL , HCT116_NR1D1 , KO and HCT116_PER2 vs. HCT116_WT cells, (d) in the SW620 vs. SW480 cells, (e) in the L1236 vs. HDMYZ cells, (f)in Arntl KO from distal bronchiolar epithelium (left) and from primary tracheal epithelial cells (right) vs. their WT, and (g)in Nr1d1/2 KO vs. WT. In each barplot, the number of genes containing transcripts with circadian expression and protein-coding biotype (dark blue), protein-coding biotype (blue), and different biotypes (light blue) is indicated. KO genes being alternatively spliced in HCT116 vs. WT pointing to the SE-event for all cancer cell lines (Supplementary Fig. 9a–e). In clock knockout specific alterations in alternative splicing. murine KO models, a smaller number of differentially spliced Further, we carried out pairwise differential splicing analysis candidates were observed (Supplementary Fig. 9f–g). We then between KO and WT (Supplementary Fig. 9). Each differentially searched for common differentially spliced candidates between spliced candidate showed significantly (p < 0.05) higher change in the human and mouse KO datasets. Common differentially spliced transcript abundance in the KO than in the corresponding WT candidates were observed only within SE-event. Sorting Nexin 3 KO KO condition. Both HCT116_NR1D1 and HCT116_PER2 showed (SNX3), which is involved in intracellular trafficking was found to KO higher number of differentially spliced genes than in be differentially spliced in both HCT116_ARNTL cells and mouse KO HCT116_ARNTL cells (Supplementary Fig. 9a–c). Besides, the Arntl KO from distal bronchiolar epithelium compared to their highest number of differentially spliced genes were observed in controls. SNX3-201 and SNX3-204 isoforms were present in Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. KO KO HCT116_ARNTL . Among the common differentially spliced specifictoHCT116_ARNTL . Both the transcript pairs of PCGF2 KO KO candidates between HCT116_NR1D1 and mouse Nr1d1/2 KO, consisted of protein-coding biotypes. In HCT116_NR1D1 ,wealso were Prolyl 4-Hydroxylase Subunit Alpha 1 (P4HA1), associated saw KO-specific phase-shifted transcript pairs with retained intron with oxidoreductase activity and Leucine Rich Repeat Containing 8 and nonsense-mediated decay biotypes from ATP Binding Cassette KO VRAC Subunit A (LRRC8A), which is an essential component of the Subfamily A Member 2 (ABCA2), a lipid transporter. HCT116_PER2 volume-regulated anion channel, and plays a role in cell adhesion showed two unique phase-shifted transcript pairs with protein- and cellular trafficking. LRRC8A-202 and LRRC8A-203 transcripts coding and retained intron biotypes from Chromodomain Helicase contributed to the observed differential splicing of LRRC8A and DNA Binding Protein 8 (CHD8), a transcription suppressor gene. P4HA1-202 and P4HA1-203 transcripts contributed to differential A functional enrichment analysis for genes with phase-shifted splicing of P4HA1. While in the mouse Nr1d1/2 KO, Lrrc8a-201 and isoform pairs was carried out for HCT116 datasets (Supplementary Fig. 10). Genes with phase-shifted isoform pairs in HCT116_WT Lrrc8a-202 isoforms contributed to the observed differential were enriched in processes like DNA damage response and G1/S splicing of Lrrc8a and P4ha1-202 transcript contributed to cell cycle transition (Supplementary Fig. 10). Whereas, differential splicing of P4ha1. Taken together, our results showed KO HCT116_ARNTL phase-shifted candidates were uniquely a core-clock KO-specific impact on AS events within the datasets enriched in processes like regulation of apoptotic signalling and analysed. KO EMT. HCT116_NR1D1 phase-shift candidates were uniquely enriched in processes like macrophage proliferation and intracel- Core-clock KO in HCT116 cells resulted in the occurrence of KO lular cholesterol transport. HCT116_PER2 candidates were differentially rhythmic transcript pairs associated with cancer uniquely enriched in processes like regulation of intracellular hallmarks transport and negative regulation of chromatin organization Based on the differences in SFs rhythmicity and local AS events, as (Supplementary Fig. 10). KO seen in HCT116 , we would expect clock dependent changes in the Subsequently, we mapped the genes with phase-shifted rhythmicity of alternatively spliced transcripts. We used data from isoform pairs (phase-diff ≥3 h) with a cancer hallmark list of genes circadian transcripts to examine whether transcript isoforms from obtained from the cancer hallmark genes database (Fig. 5e). A the same gene in each condition show phase-shifted rhythmic KO total of 76 genes in HCT116_WT, 31 genes in HCT116_ARNTL ,38 KO KO expression. For that, we extracted rhythmic and phase-shifted genes in HCT116_NR1D1 , and 30 genes in HCT116_PER2 with transcript pairs (Fig. 5). In short, we compared each rhythmic differentially rhythmic isoform pairs, were related to cancer transcript with all other rhythmic transcripts of the same gene, in a hallmarks (Fig. 5e). Transcript pairs from CSNK2A1 were observed pairwise manner, within the same cell line datasets. Differentially in all HCT116 datasets. However, different phase-shifted transcript KO rhythmic transcript pairs (q <0.05) were obtained for phase pairs were seen in HCT116_WT and HCT116 . In HCT116_WT, we differences larger than 3 h. The scatter plot depicts the distribution observed phase-shift transcript pairs from SMAD2, a cancer of differentially rhythmic transcript pairs according to their phase hallmark gene. Also, transcript pairs from AKT2, involved in all KO difference and amplitude ratio (Fig. 5a). All HCT116 cells showed a cancer hallmark processes except genome instability were phase- decrease in the number of differentially rhythmic transcript pairs vs. shifted in HCT116_WT. However, differentially rhythmic transcript HCT116_WT (Fig. 5b). 1021 transcript isoform pairs (592 unique pairs of both SMAD2 and AKT2 were not found in the KO cells. genes) showed differential rhythmicity in HCT116_WT and out of Phase-shifted transcripts pairs from MET, which controls several KO those, 750 transcript pairs (469 unique genes) showed a phase shift cancer hallmarks were seen in HCT116_WT and HCT116_PER2 . >3 h. The lowest number of differentially rhythmic transcript pairs In particular, MET-202 and MET-206 in HCT116_WT and MET-203 KO KO were seen in HCT116_ARNTL (336 isoform pairs; 255 unique and MET-206 in HCT116_PER2 cells were differentially rhythmic. KO genes) and out of those, 272 transcript isoform pairs (220 unique In HCT116_ARNTL , we observed a phase-shift in differentially genes) showed the pre-defined phase shift (Fig. 5b). 398 transcript rhythmic transcript pairs (from cancer hallmark genes) including isoform pairs (297 unique genes) were differentially rhythmic in HIF1A and FGFR2 that were neither differentially rhythmic in KO KO HCT116_NR1D1 and out of those, 295 pairs (228 genes) had a HCT116_WT nor in other HCT116 (Fig. 5e). Similarly, we KO phase shift larger than 3 h. HCT116_PER2 cells showed 342 observed phase-shifted transcript pairs from cancer hallmarks KO differentially rhythmic isoform pairs (241 unique genes), and 275 genes like HRAS and CD63 unique to HCT116_PER2 . Moreover, pairs (196 unique genes had a phase shift larger than 3 h (Fig. 5b). we observed phase-shifted transcript pairs related to cancer To further explore the existence of different phase shifts within hallmark genes like CASP8 and Mitogen Activated Protein Kinase 3 KO differentially rhythmic isoform pairs, we compared the peak phases (MAPK3) that were unique to HCT116_NR1D1 (Fig. 5e). of isoform pairs (Fig. 5c). A higher number of differentially rhythmic The association of specific cancer hallmarks genes with transcript pairs showed phase-difference ≥9h (Fig. 5c; right panel) as differentially rhythmic isoform pairs due to different clock compared with phase-difference ≥3h and <6h (Fig. 5c; left panel) alterations in HCT116 motivated the subsequent analysis of and phase-difference ≥6h and <9h (Fig. 5c; middle panel) in all alternatively spliced candidates and uniquely spliced candidates KO HCT116 datasets. To further understand the role of phase-shifted in all HCT116 cells. Therefore, we intersected the list of isoform pairs, we explored their biotypes (Fig. 5d). Transcript isoform candidates that were differentially spliced, and found in the pairswith protein-coding biotypeswereobservedasthe largestpair subsets of loss or gain of splicing event (rhythmic at least in in all HCT116 datasets. Among phase-shifted transcript pairs with HCT116_WT), and further shortlisted the candidates related to protein-coding biotypes, we observed 328 pairs in HCT116_WT, 122 cancer hallmarks. For better visualization, we filtered cancer KO KO pairs in HCT116_ARNTL , 140 pairs in HCT116_NR1D1 ,and 123 hallmarks genes for which at least two transcripts were rhythmic KO pairs in HCT116_PER2 . Also, transcript pairs with different biotypes in HCT116_WT for loss of splicing event (Fig. 6a). Whereas, such as nonsense-mediated decay or retained intron were phase- candidates with at least two rhythmic transcripts in KO were shifted in HCT116 datasets (Fig. 5d). For instance, three transcript shortlisted for gain of splicing event. The complete list of pairs from Baculoviral IAP Repeat Containing (BIRC5), which candidates categorized based on cancer hallmarks is provided in promotes cell proliferation and prevents apoptosis, were differen- Supplementary file 3. tially rhythmic only in HCT116_WT cells. Out of the three transcript HNRNPM, a component of spliceosome machinery and involved pairs, two pairs consisted of protein-coding biotypes and one pair in pre-mRNA processing showed either a change in the rhythmic KO consisted of protein-coding and nonsense-mediated decay biotype. properties or a loss of rhythmicity in HCT116 vs. We observed two phase-shifted transcript pairs from Polycomb HCT116_WT cells (Fig. 6b). Moreover, we observed differences in Group Ring Finger 2 (PCGF2), which controls cell proliferation the mean expression level of HNRNPM transcripts among HCT116 npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. Fig. 5 Phase-shifted spliced isoforms in HCT116 datasets are associated with hallmarks of cancer. Differential rhythmicity analysis between rhythmic transcripts within the same cell line was carried out using DODR. a The scatter plot depicts the distribution of differentially rhythmic transcript pairs according to their phase difference and amplitude ratio in HCT116. b Decline in differentially rhythmic isoform pairs observed KO in all HCT116 . c The chord diagram represents the peak phases of transcript pairs across HCT116 under different phase shift cut-offs. d The chord diagram depicts the biotypes of phase shifted transcript pairs. e The circular plot shows the association of genes containing phase shifted transcript pairs with hallmarks of cancer. Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. Fig. 6 Clock disruption in HCT116 resulted in alterations in AS in genes involved in the hallmarks of cancer. Commonly spliced and KO uniquely spliced candidates in HCT116 were mapped to the cancer hallmarks gene list. a Circular plots depict the association between KO spliced candidates and different cancer hallmarks (b) HNRNPM, a spliceosome machinery component lost its circadian expression in HCT116 KO cells. HCT116 cells also showed discrepancies in the mean expression level of HNRNPM transcripts. Expression of uniquely alternatively KO KO KO spliced candidate transcripts in (c) HCT116_PER2 ,(d) HCT116_ARNTL , and (e) HCT116_NR1D1 were plotted. Genomic region plots of MET, FGFR2 and LRRC8A transcripts represent differences in their exon composition (marked in red) compared to canonical forms. Circadian rhythmic transcripts were plotted using harmonic regression fit and arrhythmic transcripts were plotted using Loess fitinR. npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. KO KO datasets. Both HCT116_NR1D1 and HCT116_PER2 resulted in SW620 cell line (29 genes), the metastatic counterpart of SW480 the loss of rhythmicity of transcripts from HNRNPM. On the from the same patient with a less robust circadian rhythm .In KO contrary, HCT116_ARNTL resulted in the gain of one circadian addition, many of the affected alternatively spliced target genes HNRNPM transcript (HNRNPM-215) vs. HCT116_WT. When analys- with circadian AS patterns were involved in various key cancer ing knockout-specific spliced candidates, we observed cancer pathways including VEGFA (involved in angiogenesis) and CD44 KO hallmarks genes like MET as uniquely spliced in HCT116_PER2 (involved in EMT and metastasis), indicating a relevant role for (Fig. 6c). The genomic region plots of MET transcripts (all protein- temporal AS in mediating cancer progression. coding biotypes) represent the difference in coding exon In the current study, we evaluated the effect of circadian composition compared to the canonical form of MET (MET-202). regulation of AS in the colon cancer cell line HCT116 with a robust Moreover, the comparison of the coding exon sequence between circadian clock, as well as in core-clock knockout mutants of the canonical form and other transcripts showed missing/skipping HCT116, and investigated the potential clock-AS interplay in of exon 1- exon 21 (in MET-206), missing of exon 10 (in MET-201), cancer promoting properties. We also investigated our results in and missing/skipping of all exons between 1 to 21 except exon different colon cancer cell lines (SW480 and SW620), HL cells and 10,11 (in MET-203). The AS transcripts of cancer hallmarks genes healthy mouse tissues for a more general understanding of the like FGFR2 showed differential rhythmicity unique to interplay between alterations of the circadian clock and resulting KO KO HCT116_ARNTL (Fig. 6d). Specifically, HCT116_ARNTL showed AS events. changes in the expression of two of the FGFR2 transcripts (FGFR2- Our analysis showed that the alterations in core-clock 202 and FGFR2-211; both protein-coding). The genomic plots for components directly influenced the rhythmicity pattern of splicing FGFR2 transcripts show the loss of coding exon region in FGFR2- factor transcripts, which subsequently impacted AS events. The 202 and FGFR2-211 vs. FGFR2-206 (canonical form). The compar- SFs whose transcripts showed altered rhythmicity pattern in KO KO ison of coding exons sequence between FGFR2-206 and FGFR2- HCT116 included RACK1, SFPQ (altered in HCT116_ARNTL ), KO 202/-211 showed exclusion of exon 9 in FGFR2-202 while exclusion CIRBP, HNRNPD (altered in HCT116_NR1D1 ), and SRSF5 (altered KO of exon 8 in FGFR2-211 (Fig. 6d). LRRC8A was differentially spliced in HCT116_PER2 ). Both RACK1 and SFPQ regulate numerous KO in both HCT116_NR1D1 vs. HCT116_WT and mouse Nr1d1/2 KO cancer-related cellular processes. vs. WT. LRRC8A-201 was not circadian in HCT116_WT, but it was RACK1 is critical for cell proliferation and transcription , while KO circadian expressed in HCT116_NR1D1 (Fig. 6e). CD44 was SFPQ is a multi-functional protein that regulates several processes KO 45 46 differentially spliced only in HCT116_NR1D1 vs. HCT116_WT, such as RNA post-transcriptional activity , splicing regulation , however, the resulting transcripts were rhythmic only in and DNA repair . Aberrant expression of SFPQ is associated with HCT116_WT. We also observed gain in rhythmic transcripts of aetiology of colorectal cancer . A previous study by Pellarin et al KO cancer hallmark genes unique to specific HCT116 vs. reported the reduction in platinum chemotherapy induced HCT116_WT. For instance, rhythmic transcripts from Succinate apoptosis due to SFPQ via alternative splicing of CASP9 in human Dehydrogenase Complex Flavoprotein Subunit A (SDHA), a epithelial ovarian cancer samples . Our analysis showed that complex of mitochondrial respiratory chain were unique to ARNTL deletion in HCT116 cells resulted in a phase shift (>6 h) KO HCT116_ARNTL vs. HCT116_WT. SDHA-206 (retained intron) within SFPQ transcripts (Fig. 3, Supplementary Fig. 6), pointing was rhythmic in HCT116_WT whereas SDHA-202 (retained intron) towards its regulation via a core-clock component. Similarly, phase KO and SDHA-211 (protein-coding) were rhythmic only in shift in RACK1 transcript in HCT116_ARNTL vs. WT suggests that KO KO HCT116_ARNTL . In case of HCT116_NR1D1 , we observed a its circadian variation is linked uniquely with ARNTL. A previous new rhythmic transcript of Diacylglycerol Kinase Zeta (DGKZ), a study using mouse fibroblasts reported the recruitment of RACK1 regulator of intracellular signalling. DGKZ-215 and DGKZ-218 (both receptor in a circadian manner into the nuclear BMAL1 complex, with a retained intron event) were rhythmic in HCT116_WT and the overexpression of RACK1 suppressed CLOCK-BMAL1 whereas DGKZ-201 (protein-coding) was rhythmic only in transcriptional activity . Similarly, BMAL1 was also shown to KO 51 HCT116_NR1D1 . Similarly, we saw that a new rhythmic transcript regulate the expression of Sfpq in rat pituitary cells , thereby from Platelet Derived Growth Factor Subunit (PDGFA), which is corroborating our findings. KO essential for cell survival was unique to HCT116_PER2 . PDGFA- In addition to the observed phase shift in SFs due to the KO of 201 (protein-coding) was rhythmic in HCT116_WT and PDGFA-203 ARNTL, we also observed phase shifts among other transcripts. KO (protein-coding) was only rhythmic in HCT116_PER2 . Altogether, CIRBP, an RNA binding protein regulates several processes like cell our findings showed that each of three core-clock knockouts proliferation and circadian gene expression while HNRNPD resulted in aberrant alternative splicing of different cancer regulates mRNA stability of genes involved in the cell cycle . hallmarks related genes, pointing to a role for the circadian clock The phase shift in rhythmic expression of CIRBP and HNRNPD KO in the regulation of alternative splicing with potential conse- transcripts seen in HCT116_NR1D1 suggests an effect of NR1D1 quences in tumorigenesis. depletion on the spliceosome machinery. Indeed, Cirbp was reported to regulate sleep and circadian clock via Nr1d1 in mice and Hnrnpd was rhythmic in mice cells . Yet, to our knowledge, DISCUSSION no studies were found reporting NR1D1-mediated HNRNPD Pre-mRNA splicing contributes to generate diversity in the expression modulation. Likewise, the phase shift detected in KO products of more than 95% human genes and leads to SRSF5 transcript in HCT116_PER2 pointed towards its regulation alternatively spliced transcripts that encode distinct proteins. This via PER2. Aberrant expression of SRSF5 is associated with different mechanism is often used to maintain cellular homeostasis and to cancer types and their severity , however, no direct role of PER2- 42,43 regulate cell differentiation and development . Previous mediated SRSF5 rhythmicity was reported before. Moreover, we studies have pointed to a regulation of AS via the circadian clock observed loss of rhythmicity in several relevant components of the in cancer, suggesting a temporal pattern of AS, which affected the spliceosome complex including U2AF1 and MBNL2 (required for expression of alternatively spliced target genes in a time- spliceosome binding to the pre-mRNA branch site and a 24–27 dependent manner . Interestingly, the distinct temporal AS modulator of AS, respectively) with deletion of clock components KO pattern and the time-dependent expression of target genes were in all three HCT116 cells. Mutations in U2AF1 contribute to shown to correlate with the circadian phenotype of the cancer progression and have been reported in several different investigated cancer cell lines. For example, SW480 colon cancer cancer types including CRC , while MBNL2 was reported as cells with a robust circadian clock showed twice as many genes tumour suppressor in hepatocarcinogenesis . Accordingly, the with circadian alternative exons (59 genes) compared to the loss of U2AF1 and MBNL2 rhythmic transcripts, as seen in Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. KO HCT116 , point towards their possible contribution to metastasis macrophage proliferation and intracellular cholesterol transport KO formation. Indeed, our results also showed the loss of U2AF1 and (HCT116_NR1D1 ), intracellular transport and chromatin organi- KO MBNL2 rhythmicity in SW620 (metastasis CRC) and L1236 (stage zation (HCT116_PER2 ). These results highlight the individual IV HL). impact of each of the core-clock KOs in cellular functioning. For Several studies reported that alterations in the spliceosome instance, we identified phase shifted transcript pairs from different 12,58–64 machinery lead to aberrant AS patterns . In agreement with cancer hallmark genes such as BIRC5 in HCT116_WT, PCGF2 in KO KO these data and our findings for HCT116 cells, we also found loss of HCT116_ARNTL , ABCA2 in HCT116_NR1D1 , and CHD8 in KO rhythmicity in SFs genes in recently published RNA-seq datasets of HCT116_PER2 . BIRC5, an immune-related gene was found to murine Per1/2 KO (GSE171975 ) and murine Cry1/2 KO be highly expressed in different tumour types and promotes cell 66 76 (GSE135898 ) (Supplementary Fig. 5). These results highlight proliferation . Moreover, Birc5 was reported rhythmic in colon the relevance of core-clock elements in the regulation of genes mucosa cells and its silencing resulted in an increased sensitivity involved in splicing. Indeed, aberrant pre-mRNA splicing and of HCT116 cells to CDK inhibitors . PCGF2, a ring finger protein, alterations in splicing factors are known to act as oncogenic was reported to negatively regulate granulocyte differentiation in 67 78 drivers and contribute to tumour progression . Our results show human HL-60 cells . While, increased expression of ABCA2,a the existence of alterations in rhythmic properties of SFs member of ATP transporters, was reported in different cancer transcripts due to the KO of core-clock elements, which was types . A previous report on CHD8, a negative regulator of Wnt reflected on the resulting AS transcripts. For instance, the pre- signalling, showed that loss of its expression might be an indicator spliceosomal complex element RBM5 regulates AS of CASP9,an of aggressiveness in gastric cancer . However, the interdepen- 4,68 apoptosis-related gene and showed altered rhythmic proper- dence of PCGF2, ABCA2, and CHD8 on specific core-clock elements KO ties in HCT116_ARNTL vs. HCT116_WT cells. RBM5, a tumour was not reported. We also observed phase-shifted isoforms of suppressor gene and splicing factor, improves the production of some gene such as CSNK2A1 in all HCT116 datasets, however, mRNAs by recognizing incorrect 3′splice sites of epidermal growth different transcript pairs were phase-shifted in each KO, pointing factor receptor (EGFR) pre-mRNA, thereby inhibiting the prolifera- to a KO-specific effect. Moreover, we saw the loss of differentially tion of tumour cells . Previous studies in different cancer types transcript pairs of cancer hallmark genes including SMAD2 and KO suggested RBM5 as a potential target to prevent tumorigen- AKT1 in all HCT116 . We also observed differential rhythmicity in 70,71 esis . In our study, we observed a regulation of RBM5 certain cancer hallmark genes, that were not circadian in expression via the circadian clock, which led to a loss of its HCT116_WT. For instance, HIF1A and FGFR2 showed differentially KO rhythmicity due to ARNTL disruption. rhythmic transcript pairs unique to HCT116_ARNTL . HIF1A is the The impact of circadian dysregulation in the splicing machinery main element of the HIF1-pathway, which plays a crucial role in KO was also seen through the loss/gain of AS events in all HCT116 adaptive responses of tumour cells to hypoxia and promotes cells. Among the different splicing events, exon skipping is the tumour progression and metastasis via oncogenic growth factors most common splicing event in human tissues , as also observed such as TGFß (transforming growth factor beta) and EGF in our study. The unique AF-event gain as seen in (epidermal growth factor) . FGFR2 is an EMT-associated gene KO KO HCT116_NR1D1 and HCT116_PER2 indicates that the deletion which encoded for two isoforms, IIIb and IIIc, characteristic to of either PER2 or NR1D1 in HCT116 alters AS by preferring AF- epithelial and to mesenchymal cells, respectively . The discrepant event to the other six modes of AS. expression pattern, as well as rhythmic properties of these two KO By comparing our results to datasets from mouse models, we FGFR2 spliced variants in HCT116_ARNTL vs WT suggests that KO found SNX3 (common between HCT116_ARNTL and murine ARNTL deletion in HCT116 might promote metastasis via EMT. Arntl KO from distal bronchiolar epithelium tissue) and P4HA1, Similarly, we observed differentially rhythmic transcript pairs of KO LRRC8A (common between HCT116_NR1D1 and murine Nr1d1/2 HRAS (proto-oncogene and member of the RAS-pathway involved KO KO) among the common differentially spliced candidates between in cell proliferation) and CD63 unique to HCT116_PER2 . CD63 is a KO HCT116 and murine KO models, however resulting in different member of the tetraspanin family involved in cell differentiation sets of alternatively spliced transcripts. These differences suggest and migration, which was found to be a prognostic marker in 83,84 KO that the deletion of ARNTL or NR1D1 in HCT116 is likely to have a colorectal and esophageal cancer . In HCT116_NR1D1 ,we different biological impact (protein product) compared to the observed unique differentially rhythmic transcript pairs of CASP8 same KOs in murine models. Differential splicing of SNX3 in and MAPK3. CASP8 is an apoptosis-related cysteine peptidase and KO KO HCT116_ARNTL and P4HA1, LRRC8A in HCT116_NR1D1 point an essential part of the death-inducing signalling complex towards splicing alterations, which are KO specific. SNX3, a unique (DISC) . In cancer, it promotes proliferation and angiogenesis mediator of WNT protein secretion, was reported to mediate EMT through the activation of NF-kB in glioblastoma and breast 73 87 and metastasis in CRC cells including HCT116 . P4HA1, a catalytic cancer . MAPK3 encodes for a member of the MAP kinase family enzyme, was reported to regulate cell proliferation in CRC cells via (aka extracellular signal-regulated kinases (ERK)) and regulates HIF1A and WNT signaling . While, LRRC8A, a main regulatory various cellular processes such as proliferation, differentiation, and subunit of VRAC (volume-regulated anion channel), was found to cell cycle progression in response to a variety of extracellular be upregulated in colon cancer patients and might contribute to signals. These results point towards a cross-talk between aberrant 74 KO metastasis . However, the direct correlation between these genes AS events and cancer hallmarks related to each of the HCT116 . and core-clock components remains unexplored. We also observed cancer hallmark genes like MET, whose Following the temporal regulation of different AS events in different transcript pairs were phase shifted in HCT116_WT and KO HCT116 datasets, our analysis revealed several transcript pairs that HCT116_PER2 . The MET oncogene encodes for a receptor showed differential rhythmicity. Moreover, we also showed the tyrosine kinase with pleiotropic functions in initiating and occurrence of new KO-specific alterations in transcripts rhythmi- sustaining neoplastic transformation, as well as in cancer cell city vs. WT. In HCT116_WT, phase-shifted isoforms were found to survival and tumour dissemination . MET is known to undergo AS be enriched in DNA damage response and G1/S cell cycle and its AS isoform (lacking exon 14) is known to inhibit HGF- transition. This suggests a temporal regulation of DNA repair and induced tyrosine phosphorylation of Met, as well as cell cell cycle pathways that might be regulated by circadian AS proliferation and migration in skeletal muscle myoblasts . events. Indeed, a previous study reported splicing as an emerging We further compared our list of AS spliced candidates to a list of 75 41 pathway contributing to DNA damage response . In the KO cells, known cancer hallmark genes . Of these, ACTB was differentially KO we observed different enriched processes among phase-shifted spliced in all HCT116 vs WT, however rhythmicity of ACTB KO isoforms such as apoptotic signalling and EMT (HCT116_ARNTL ), transcripts were lost in all three KOs. This suggests that the loss in npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. circadian profile of ACTB transcripts may contribute towards demonstrated for numerous anticancer agents in experimental metastasis due to clock alterations. A previous study reported that models, as well as in clinical settings including clinical studies in 102–104 the abnormal expression and polymerization of ACTB contributes patients with metastatic colorectal cancer . Aberrant alter- to invasiveness and metastasis of different cancer types . Besides native splicing events and circadian clock disruption are reported different cancer hallmarks related to AS candidates, we also in several different cancer types, adding an additional complexity 67,105 observed discrepant expression of HNRNPM SF transcripts in all to the genetic landscape of cancer . The results of our study KO three HCT116 vs. WT. Among different KO-specific AS candi- highlight the importance of timed AS, especially in genes KO KO dates, FGFR2 (in HCT116_ARNTL ), MET (in HCT116_PER2 ), CD44 regulating cancer hallmarks that are shown to be suitable drug KO (in HCT116_NR1D1 ) lost/gained rhythmicity in the KOs. In targets. Our data point to the regulation of AS patterns in cancer KO HCT116_ARNTL , FGFR2 IIIb isoform (exon 9 exclusion and exon via the circadian clock. It would be important to further explore 8 inclusion; FGFR2-202) was circadian while in HCT116_WT FGFR2 these findings and to consider drug timing, at least for drugs IIIc (exon 8 exclusion and exon 9 inclusion; FGFR2-211) was targeting such genes, in future clinical studies. circadian. These discrepancies were better seen when we analysed Altogether, the results of our study suggest an interplay the average expression of both isoforms in WT and between circadian clock elements and AS in cancer, with distinct KO HCT116_ARNTL cells. The lower average expression of FGFR2 and unique roles for the core-clock genes ARNTL, PER2 and NR1D1 KO IIIb in HCT116_ARNTL while higher expression of FGFR2 IIIc point in regulating SF rhythmicity and AS events in cancer hallmark to possible alterations in EMT signalling. Matsuda Y et al. reported genes, with relevance in cancer onset and progression. increased FGFR2 IIIc expression among colorectal carcinomas samples and a human anti-FGFR2 IIIc monoclonal antibody was METHODS reported to inhibit growth in colorectal carcinoma cells . These findings suggest that FGFR2 IIIc could be a promising therapeutic Cell culture target for colorectal carcinoma. Human colorectal carcinoma cell line HCT116 (ATCC® CCL-247™, Gaithers- CD44 encodes for a family of cell adhesion molecules involved burg, MD, USA) was cultured in Dulbecco’s Modified Eagle Medium (Gibco, in homotypic and heterotypic interactions with extracellular Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% Fetal Bovine Serum (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) and 1% matrix components, as well as EMT . It promotes proliferation Penicillin–Streptomycin (Gibco, Thermo Fisher Scientific, Waltham, MA, and invasiveness of cells via recruiting ERM proteins (Ezrin, Radixin USA) in an incubator with 5% CO at 37 °C. LUNA™ Automated Cell Counter und Moesin) under certain conditions. However, CD44 can also act (Logos Biosystems, Anyang, South Korea) was used for cell counting and as a tumour suppressor, for example when cells reach confluent morphology analysis. Cell lines were tested for mycoplasma by using the growth conditions, and thus inhibit cell growth. The prognostic Mycoplasma check service of Eurofins Genomics (Eurofins Genomics, value of the CD44 variant 6 (CD44v6) in CRC was debated for years Ebersberg, Germany). 93–95 due to contradictory results . The function of CD44 in a cell is determined by the CD44 isoform pattern expressed . Several CRISPR-Cas9 Knockout generation studies showed the relevance of CD44v6 as an independent A CRISPR-Cas9 mediated approach was applied to generate the core-clock negative prognostic factor and a promising therapeutic target in 96,97 knockout cell lines in HCT116. Briefly, HCT116_WT cells were seeded in CRC . In our study, we showed that NR1D1 disruption in 6-well plates at a density of 4 × 10 cells/well and transfected with CRISPR- HCT116 cells results in aberrant alternative splicing of CD44, which Cas9 plasmids containing GFP marker and guided RNAs targeting multiple points to a clock specific regulation of CD44 splicing, and could exons of ARNTL, PER2 or NR1D1 genes, respectively. Cell transfection was potentially play a role in CD44-targeted therapy regimens. performed using FuGENE HD Transfection Reagent (Promega Corporation, Interestingly, several CD44-targeted drugs have been approved Fitchburg, WI, USA) according to the manufacturer’s instructions. 48 h post- for clinical trials, which highlights the importance of timing transfection, CRISPR/Cas9 GFP-positive cells were single-cell sorted using treatment in cancer therapy targeting CD44 (reviewed in ). Our an S3e cell sorter (Bio-Rad laboratories, Hercules, CA, USA) into 96-well results suggest that the perturbations in core-clock elements plates. Colonies were expanded for subsequent testing and successful knockout colonies were used for the time-course RNA-seq experiment. result in different AS outcomes for this gene. For each knockout condition, several single clones were investigated on Furthermore, certain genes showed a gain of new rhythmic RNA gene expression level to characterize and confirm the knockout. All transcripts in the KO vs. WT. These included SDHA KO cell lines displayed significantly reduced target gene expression KO KO (HCT116_ARNTL vs. WT), DGKZ (HCT116_NR1D1 vs. WT), and compared to WT (Supplementary Fig. 1d). The off-target activity was KO PDGFA (HCT116_PER2 vs. WT). SDHA is responsible for transfer- investigated using Off-Spotter and the Welcome Trust Sanger Institute ring electrons from succinate to ubiquinone (coenzyme Q), and 107 Genome Editing database (WGE) to search for the most likely potential acts as a tumour suppressor and inhibitor of angiogenesis in off-target sites using gRNA sequences. Off-target sites with up to three paraganglioma . DGKZ attenuates protein kinase C activity by mismatches within protein-coding regions were Sanger-sequenced and regulating diacylglycerol levels in intracellular signalling cascade compared to WT. All investigated potential off-target sites in KO cells showed 100% sequence similarity to the WT, indicating absent off-target and signal transduction. It has proven to be associated with modifications. For regions where it was not possible to design specific various signalling pathways, including ERK and MYC and acts as a primers amplifying the off-target site, we compared band sizes on gel potential oncogene in osteosarcoma . PDGFA is a member of the electrophoresis. PDGF as well as VEGF signalling. Paracrine PDGF signalling is commonly observed in epithelial cancers, where it triggers stromal Sample preparation and RNA extraction recruitment and may be involved in EMT affecting tumour growth, angiogenesis, invasion, and metastasis. Overexpression of PDGF HCT116 cells were seeded in triplicates in 12-well plates with a density of 2×10 cells per well. On the next days, cells were synchronized by signalling was shown to drive tumour cell growth and to promote changing the media. Sampling was started 9 h after synchronization and tumorigenesis in colorectal cancer, breast cancer, lung cancer and samples were taken every 3 h for a time-series of 45 h. To prepare the cells sarcomas (reviewed in ref. ). The gained rhythmicity of a coding for RNA extraction, media was discarded and cells were washed with transcript of SDHA due to ARNTL deletion, DGKZ due to NR1D1 phosphate buffer saline (PBS) and lysed using RLT Plus buffer (Qiagen, deletion, and PDGFA due to PER2 deletion in HCT116 may lead to Hilden, Germany) directly on the plate. A total of 64 samples were obtained aberrant protein functions resulting from clock-KO-specific spli- from the HCT116 datasets (WT and three KOs). cing. However, the biological role of these transcripts remains to Total RNA was isolated using the RNeasy Plus Mini kit (Qiagen, Hilden, be elucidated. Germany) according to the manufacturer’s guidelines. Genomic DNA was The relevance of timing treatment for increasing tolerability and digested from the cells using gDNA eliminator columns provided with the efficacy while minimizing toxic side-effects has been largely kit (Qiagen, Hilden, Germany). RNA was eluted in 30 µL RNase-free water. Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. RNA concentration was measured using a Nanodrop 1000 (Thermo Fisher Curation of splicing factor list Scientific, Waltham, MA, USA). RNA was then stored at −80 °C until In total 534 human spliceosome and splicing-related genes were compiled 24,35,36 further use. from the literature , as well as from online databases (Spliceosome 37 38 database and SpliceAid 2 ). A list of human splicing factors was mapped using Ensembl Biomart to obtain their orthologous sets for mouse. A Data generation total of 521 SFs were obtained for mouse. The complete list of splicing High-quality RNA was used to generate mRNA libraries and prepared using factors is provided in Supplementary files 1–2. the TruSeq Stranded mRNA Sample Preparation Kit (Illumina, San Diego, CA, USA) according to the sample protocol guidelines and sequenced on an Illumina NextSeq 500 platform to an average depth of 100 M 75-bp RNA-sequencing data processing paired-end reads at the European Molecular Biology Laboratory (EMBL) All RNA-seq datasets were processed and analysed according to the GeneCore Facility (Heidelberg, Germany). following protocol unless otherwise stated. Quality assessment of raw reads was carried out using fastqcr (version 0.1.2) R package which easily parse, aggregate, and analyse FastQC for large number of samples. Acquisition of additional RNA-seq datasets Based on FastQC reports, over-represented sequences and residual Using a strict criterion of “paired-end RNA-seq”, “less than or equal to 4 h adapter sequences were trimmed from raw reads using Trimmomatic interval”, “mammalian tissue”, “circadian rhythm”, we obtained a total of (version 0.39 ) using default settings. After trimming, only paired-end 134 samples from six different datasets. Published paired-end RNA-seq for reads were used for further downstream analysis. Paired-end reads from Per1/2 was not available, therefore we included for this dataset recently human datasets were aligned to human genome (Homo sapiens GRCh38, published single-end Per1/2 KO (24 samples). Out of seven different Ensembl release 92) and from mouse datasets were aligned to mouse datasets, two datasets were retrieved from human model system and five genome (Mus musculus GRCm38, Ensembl release 102) using STAR aligner datasets were retrieved from the mouse model system. (version 2.6.0 ). Based on STAR transcriptome alignment, transcript-level Based on human model system, RNA-seq datasets were collected abundances were quantified in transcripts per million (TPM) using Salmon derived from: (1) human CRC cell lines (SW480: primary tumour, SW620: 115 116 (version 0.10.2 ). Afterwards, tximport R package was used to import metastatic tumour; Accession number: E-MTAB-7779) from the same transcript level abundance, estimate counts and transcript lengths. patient, and (2) Hodgkin’s lymphoma (HL) cell lines of cancer progression Tximport summarizes the quantification results into matrices for down- stages (HD-MY-Z: stage IIIB, L-1236: stage IV; Accession number: GSE16206) stream analyses. Tximport can be used to obtain both transcript level from different patients. CRC cells (SW480, SW620) and HL cells (HDMYZ, (txOut= TRUE) and gene-level (txOut= FALSE) summarization. Here, we L1236), which represent different progression grades of CRC and HL, used transcript-level counts for detailed alternative splicing analysis. respectively, display various clock phenotypes (Supplementary Fig. 2e). Nevertheless, gene-level summarization was also used for its comparison These datasets were chosen to find a possible correlation between clock with transcript-level analysis. Trimmed mean of M-values (TMM) method of disruption and cancer progression stage. averaging was used as the normalization factor to scale up the raw library Based on mouse model system, we retrieved RNA-seq datasets derived size using edge R package (version 4.0 ). Counts per million (CPM) from (1) mouse embryonic stem cells in WT condition and the Nr1d1/2 KO function was used as a descriptive measure of transcript expression. All (Accession number: GSE125696) generated using CRISPR/Cas9, (2) laser transcript feature with at least 0.5 CPM on average over all time points micro-dissected SCN tissue of male C3H/HeH wild type mice (age: (specific to datasets) were retained and renormalized using selected 10–12 weeks; Accession number: GSE72095), (3) distal bronchiolar features. The complete pipeline for the analysis is depicted in Fig. 1. epithelium tissue isolated using laser capture microdissection from WT Raw unlogged gene-level count data from Per1/2 KO (GSE171975; single- mice and Arntl KO mice bearing the targeted deletion in mouse club cells end RNA-seq) and Cry1/2 KO (GSE135898; paired-end RNA-seq) were (age: 10–20 weeks; Accession number: E-MTAB-6384), and primary tracheal downloaded from NCBI-GEO and used for rhythmicity analysis. epithelial cells isolated from WT mice and Arntl global KO (age: 10–20 weeks; Accession number: E-MTAB-6384). In addition, we retrieved raw gene-level count data from two murine core-clock KO RNA-seq Rhythmicity analysis datasets from NCBI-GEO, liver tissue from WT and Per1/2 KO mice (age: Unlogged CPM values were used to detect rhythmic signals from time- 3 months old, Accession number: GSE171975; single-end RNA-seq), and series datasets. 24 h rhythmicity was evaluated through a non-parametric liver tissue from WT and Cry1/2 KO mice under normal feeding (age: method using RAIN R package (version 1.24.0 ). Rhythmic gene/ 9–14 weeks old, Accession number: GSE135898). All datasets listed in this transcript-sets were obtained using a cut-off of q < 0.05. The acrophase section were downloaded from NCBI-GEO or ArrayExpress. and relative amplitude were estimated for rhythmic gene-/transcript-sets 119 120 using Cosinor within Discorhythm R package (version 1.6.0 ). The Synchronization and sampling protocols of used RNA-seq rhythmic gene/transcript-sets were further filtered using an additional cut- datasets off of relative amplitude (rAMP) ≥ 0.1. E-MTAB-6384 (mouse Arntl-KO lung samples pulmonary airway epithelial cells): For circadian sampling, mice were maintained in constant darkness Alternative splicing analysis using SUPPA and samples collected 1 cycle after transfer to darkness (1 day later) at To calculate the local alternative splicing events based on the expression of circadian time (CT), which by convention anchors expected time of lights transcripts in each dataset, we used SUPPA2 . SUPPA2 is helpful in off and activity onset to CT12. Samples were taken every 4 h for 48 h in studying splicing at the local alternative splicing level or at the transcript constant dark conditions . isoform level. We analyzed seven alternative splicing events types; (1) GSE72095 (mouse SCN samples): After 7 days of acclimatisation, mice Alternative 3′ splice site (A3-event) where 3′ site acts as an acceptor, (2) were singly housed for 7 days prior to tissue harvesting. At each sampling Alternative 5′ splice site (A5-event) where 5′ site acts as a donor, (3) time-point, pooled dissected tissue from five adult male mice were used. Alternative first exon (AF-event) where the first exon is retained after Samples were taken for a total of six time-points over a 12:12 LD cycle at splicing, (4) Alternative last exon (AL-event) where the last exon is retained 4 h intervals (ZT2, 6, 10, 14, 18 and 22), where ZT0 denotes the time of after splicing, (5) Mutually exclusive event (MX-event) where one of two lights on . exons is retained, (6) Retention intron event (RI-event) where the intron is GSE125696 (Nr1d1/2 KO mouse ESCs): On differentiation day 28, cells confined within mRNA, and (7) Skipping exon event (SE-event) where an were treated with 100 nM dexamethasone and frozen at the indicated time exon may be spliced out or retained. points. Time course RNA-Seq was performed using RNA samples at 4 h For each gene in a given tissue or condition, the average percent intervals over 2 days . spliced-in (PSI) value was calculated. Alternative splicing of a gene in a Synchronization of HCT116 RNA-Seq samples: HCT116 cells were particular dataset is considered only if it fulfils the criteria of PSI < 0.9 and > synchronized via medium exchange (which serves as a strong entrainment 0.1. Using the PSI cut-off, we detected specific local events which only signal aka Zeitgeber in these cells). Previous reports on the efficacy of happened either in the WT or in the KO datasets. To classify these events, synchronization agents (e.g. dexamethasone, forskolin, serum shock, we grouped the local events as gain or loss in the KO with respect to its medium change) have shown, that medium exchange is as effective as other mentioned agents in synchronizing the circadian clock, as shown for control. If a KO contained a gene which is spliced by a particular local different CRC cell lines via bioluminescence measurements of ARNTL- event and the WT did not show the gene in that particular event, the event promoter activity . would match to gain in the KO. Similarly, if the control group contained a npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. gene which is alternatively spliced by a particular local event and the Received: 8 November 2021; Accepted: 4 April 2022; corresponding KO did not show the gene in that particular event, the event would correspond to loss in the KO. Besides, we also used SUPPA2 to carry out the differential splicing analysis where a pairwise comparison between the KO and WT was made. The differentially spliced candidates were obtained using p-val < 0.05. REFERENCES 1. Lee, Y. & Rio, D. C. Mechanisms and regulation of alternative Pre-mRNA splicing. Differential rhythmicity analysis Annu. Rev. Biochem. 84, 291–323 (2015). The detection of differentially rhythmic transcripts isoform within each 2. Chen, M. & Manley, J. L. Mechanisms of alternative splicing regulation: insights dataset was carried out using DODR R package (version 0.99.2 ). All 24 h from molecular and genomics approaches. Nat. Rev. Mol. Cell Biol. 10, 741–754 rhythmic transcript sets were used for this analysis. 24 h differentially (2009). 3. Wang, E. T. et al. Alternative isoform regulation in human tissue transcriptomes. rhythmic transcript isoform pairs per single tissue/condition type were Nature 456, 470–476 (2008). obtained using FDR corrected p-value cut-off meta. p < 0.05. To further 4. Paronetto, M. P., Passacantilli, I. & Sette, C. Alternative splicing and cell survival: examine the transcript isoform with more prominent changes in HCT116, from tissue homeostasis to disease. Cell Death Differ. 23, 1919–1929 (2016). phase difference cut-off ≥3 h was applied. 5. Ravi, S., Schilder, R. J. & Kimball, S. R. Role of precursor mRNA splicing in nutrient- induced alterations in gene expression and metabolism. J. Nutr. 145, 841–846 Acquisition of list of cancer hallmark genes (2015). 6. Du, J. X. et al. Splicing factors: Insights into their regulatory network in alter- To understand the direct influence of clock alterations on cancer hallmarks, native splicing in cancer. Cancer Lett. 501,83–104 (2021). we retrieved the complete gene list categorized according to processes 7. Scotti, M. M. & Swanson, M. S. RNA mis-splicing in disease. Nat. Rev. Genet. 17, from Cancer Hallmark Genes database . 19–32 (2016). The CHG database was established by systematically integrating 8. Venables, J. P. Aberrant and alternative splicing in cancer. Cancer Res. 64, manually curated hallmarks genes from the confirmed literature and KEGG 123 7647–7654 (2004). pathways . In parallel, based on pre-defined keywords related to 9. Wang, E. & Aifantis, I. RNA splicing and cancer. Trends Cancer 6, 631–644 (2020). hallmarks, a text mining software Lucene (https://lucene.apache.org/) 10. Ben-Dov, C., Hartmann, B., Lundgren, J. & Valcárcel, J. Genome-wide analysis of was used to identify hallmark specific pathways in the literature and in the alternative pre-mRNA splicing. J. Biol. Chem. 283, 1229–1233 (2008). KEGG database. To establish the CHG database Zhang et al. used: (1) 301 11. Oltean, S. & Bates, D. O. Hallmarks of alternative splicing in cancer. Oncogene 33, KEGG pathways for Lucene search and the extraction of pathways genes, 5311–5318 (2014). (2) Gene variant data (7075 mutation samples, 6177 methylation samples, 12. Bonnal, S. C., López-Oreja, I. & Valcárcel, J. Roles and mechanisms of alternative 9445 copy number variation samples) from TCGA (The Cancer Genome splicing in cancer—implications for care. Nat. Rev. Clin. Oncol. 17, 457–474 Atlas https://www.cancer.gov/tcga) to calculate the frequency of gene (2020). variation, (3) Human protein-protein interaction data from HPRD (Human 13. Climente-González, H., Porta-Pardo, E., Godzik, A. & Eyras, E. The functional 124 125 126 127 Protein Reference Database ), STRING , BioGRID , HTRIdb to impact of alternative splicing in cancer. Cell Rep. 20, 2215–2226 (2017). integrate a gene interaction network. CHG provides the corresponding 14. Harper, S. J. & Bates, D. O. VEGF-A splicing: the key to anti-angiogenic ther- gene sets for ten cancer hallmarks: (1) Activating invasion and metastasis apeutics? Nat. Rev. Cancer 8, 880–887 (2008). 15. Herrlich, P. et al. CD44 acts both as a growth- and invasiveness-promoting (1,149 genes), (2) Enabling replicative immortality (308 genes), (3) Evading molecule and as a tumor-suppressing cofactor. Ann. N. Y. Acad. Sci. 910, growth suppressors (709 genes), (4) Evading immune destruction (609 106–120 (2000). genes), (5) Genome instability and mutation (231 genes), (6) Inducing 16. Cohen-Eliav, M. et al. The splicing factor SRSF6 is amplified and is an onco- angiogenesis (507 genes), (7) Reprogramming energy metabolism (461 protein in lung and colon cancers. J. Pathol. 229, 630–639 (2013). genes), (8) Resisting cell death (1192 genes), (9) Sustaining proliferative 17. Park, S. et al. Differential functions of splicing factors in mammary transforma- signaling (1314 genes), and (10) Tumor promoting inflammation (648 tion and breast cancer metastasis. Cell Rep. 29, 2672–2688.e2677 (2019). genes). 18. Shakhmantsir, I. & Sehgal, A. Splicing the clock to maintain and entrain circadian rhythms. J. Biol. Rhythms 34, 584–595 (2019). 19. Patke, A. et al. Mutation of the human circadian clock gene CRY1 in familial Functional enrichment analysis delayed sleep phase disorder. Cell 169, 203–215.e213 (2017). Gene ontology analysis for selected gene sets was carried out using an R 20. Hastings, M. H., Reddy, A. B. & Maywood, E. S. A clockwork web: circadian timing based package ClusterProfiler (version 4.0 ). Enriched biological pro- in brain and periphery, in health and disease. Nat. Rev. Neurosci. 4, 649–661 cesses were obtained for each of the gene sets. Dot plots were used to (2003). visualize the top 10 enriched biological processes in the case of 21. Andreani, T. S., Itoh, T. Q., Yildirim, E., Hwangbo, D. S. & Allada, R. Genetics of differentially rhythmic isoform pairs. The enriched biological processes Circadian rhythms. Sleep. Med. Clin. 10, 413–421 (2015). for the candidate cancer hallmarks markers were manually curated after 22. Innominato, P. F., Lévi, F. A. & Bjarnason, G. A. Chronotherapy and the molecular removing the redundant biological processes. The visualization of cancer clock: clinical implications in oncology. Adv. Drug Deliv. Rev. 62, 979–1001 (2010). hallmarks connected with selected candidate markers was carried out 23. Shafi, A. A. & Knudsen, K. E. Cancer and the Circadian clock. Cancer Res. 79, using an R based package GO plot (version 1.0.2 ). 3806–3814 (2019). 24. El-Athman, R., Fuhr, L. & Relógio, A. A systems-level analysis reveals circadian regulation of splicing in colorectal cancer. EBioMedicine 33,68–81 (2018). DATA AVAILABILITY 25. El-Athman, R., Knezevic, D., Fuhr, L. & Relógio, A. A computational analysis of alternative splicing across mammalian tissues reveals circadian and ultradian The datasets analysed during the current study are available as follows: recently rhythms in splicing events. Int. J. Mol. Sci. 20, https://doi.org/10.3390/ generated RNA-sequencing data for HCT116 WT cells and KOs in this study have ijms20163977 (2019). been deposited in the ArrayExpress repository (Accession number: E-MTAB-9701). 26. Genov, N., Basti, A., Abreu, M., Astaburuaga, R. & Relógio, A. A bioinformatic Other datasets used in this manuscript are available as follows: SW480 and SW620 analysis identifies circadian expression of splicing factors and time-dependent cell lines, ArrayExpress, E-MTAB-7779; HD-MY-Z and L-1236 cell lines, Gene Expression alternative splicing events in the HD-MY-Z cell line. Sci. Rep. 9, 11062 (2019). Omnibus (GEO) GSE16206; mouse Arntl-KO lung samples Pulmonary airway epithelial 27. Genov, N., Basti, A., Abreu, M. & Relógio, A. Temporal splicing switches in ele- cells, ArrayExpress, E-MTAB-6384; mouse SCN samples, GEO, GSE72095; Nr1d1/2 KO ments of the tnf-pathway identified by computational analysis of transcriptome mouse ESCs, GEO, GSE125696; murine Per1/2 KO liver samples, GEO, GSE171975; data for human cell lines. Int. J. Mol. Sci. 20, https://doi.org/10.3390/ murine Cry1/2 KO liver samples, GEO, GSE135898. ijms20051182 (2019). 28. Pluquet, O. et al. Posttranscriptional regulation of PER1 underlies the oncogenic function of IREα. Cancer Res. 73, 4732–4743 (2013). CODE AVAILABILITY 29. Koyanagi, S. et al. A molecular mechanism regulating circadian expression of vascular endothelial growth factor in tumor cells. Cancer Res. 63, 7277–7283 All programs, packages and functions used in this manuscript are freely available and (2003). are referred to in the Method section. Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. 30. Preußner, M. et al. Rhythmic U2af26 alternative splicing controls 61. Martín, E., Vivori, C., Rogalska, M., Herrero-Vicente, J. & Valcárcel, J. Alternative PERIOD1 stability and the circadian clock in mice. Mol. Cell 54, 651–662 (2014). splicing regulation of cell-cycle genes by SPF45/SR140/CHERP complex controls 31. Lee, J.,Park, E.,Kim, G.H., Kwon,I.&Kim, K. Asplice variant of humanBmal1 acts as a cell proliferation. RNA 27, 1557–1576 (2021). negative regulator of the molecular circadian clock. Exp. Mol. Med. 50,1–10 (2018). 62. Venables, J. P. Unbalanced alternative splicing and its significance in cancer. 32. Duffield, G. E. DNA microarray analyses of circadian timing: the genomic basis of Bioessays 28, 378–386 (2006). biological time. J. Neuroendocrinol. 15, 991–1002 (2003). 63. Zhang, J. & Manley, J. L. Misregulation of pre-mRNA alternative splicing in 33. Wu, Y. et al. Systematic studies of the circadian clock genes impact on tem- cancer. Cancer Discov. 3, 1228–1237 (2013). perature compensation and cell proliferation using CRISPR tools. Biology (Basel) 64. Gebauer, F., Schwarzl, T., Valcárcel, J. & Hentze, M. W. RNA-binding proteins in 10, https://doi.org/10.3390/biology10111204 (2021). human genetic disease. Nat. Rev. Genet. 22, 185–198 (2021). 34. Bunger, M. K. et al. Mop3 is an essential component of the master circadian 65. Aviram, R., Dandavate, V., Manella, G., Golik, M. & Asher, G. Ultradian rhythms of pacemaker in mammals. Cell 103, 1009–1017 (2000). AKT phosphorylation and gene expression emerge in the absence of the cir- 35. Relógio, A. et al. Alternative splicing microarrays reveal functional expression of cadian clock components Per1 and Per2. PLoS Biol. 19, e3001492 (2021). neuron-specific regulators in Hodgkin lymphoma cells. J. Biol. Chem. 280, 66. Weger, B. D. et al. Systematic analysis of differential rhythmic liver gene 4779–4784 (2005). expression mediated by the circadian clock and feeding rhythms. Proc. Natl 36. Seiler, M. et al. Somatic mutational landscape of splicing factor genes and their Acad. Sci. USA 118, https://doi.org/10.1073/pnas.2015803118 (2021). functional consequences across 33 cancer types. Cell Rep. 23,282–296.e284 (2018). 67. Zhang, Y., Qian, J., Gu, C. & Yang, Y. Alternative splicing and cancer: a systematic 37. Cvitkovic, I. & Jurica, M. S. Spliceosome database: a tool for tracking components review. Signal Transduct. Target Ther. 6, 78 (2021). of the spliceosome. Nucleic acids Res. 41, D132–141 (2013). 68. Bechara, E. G., Sebestyen, E., Bernardis, I., Eyras, E. & Valcarcel, J. RBM5, 6, and 10 38. Piva, F., Giulietti, M., Burini, A. B. & Principato, G. SpliceAid 2: a database of differentially regulate NUMB alternative splicing to control cancer cell pro- human splicing factors expression data and RNA target motifs. Hum. Mutat. 33, liferation. Mol. Cell 52, 720–733 (2013). 81–85 (2012). 69. Su, Z. et al. Lentiviral vector-mediated RBM5 overexpression downregulates 39. Kinsella, R. J. et al. Ensembl BioMarts: a hub for data retrieval across taxonomic EGFR expression in human non-small cell lung cancer cells. World J. Surg. Oncol. space. Database (Oxf.) 2011, bar030 (2011). 12, 367 (2014). 40. Wojtuszkiewicz, A. et al. Pre-mRNA splicing in cancer: the relevance in onco- 70. Jiang, Y. et al. RBM5 inhibits tumorigenesis of gliomas through inhibition of Wnt/ genesis, treatment and drug resistance. Expert Opin. Drug Metab. Toxicol. 11, β-catenin signaling and induction of apoptosis. World J. Surg. Oncol. 15, 9 (2017). 673–689 (2015). 71. Prabhu, V. V. & Devaraj, N. Regulating RNA binding motif 5 gene expression—a 41. Zhang, D. et al. CHG: a systematically integrated database of cancer hallmark novel therapeutic target for lung cancer. J. Environ. Pathol. Toxicol. Oncol. 36, genes. Front. Genet. 11, https://doi.org/10.3389/fgene.2020.00029 (2020). 99–105 (2017). 42. Gallego-Paez, L. M. et al. Alternative splicing: the pledge, the turn, and the 72. Florea, L., Song, L. & Salzberg, S. L. Thousands of exon skipping events differentiate prestige: the key role of alternative splicing in human biological systems. Hum. among splicing patterns in sixteen human tissues. F1000Res 2, 188 (2013). Genet. 136, 1015–1042 (2017). 73. Pan, B. et al. SNX3 suppresses the migration and invasion of colorectal cancer 43. Baralle, F. E. & Giudice, J. Alternative splicing as a regulator of development and cells by reversing epithelial-to-mesenchymal transition via the β-catenin path- tissue identity. Nat. Rev. Mol. Cell Biol. 18, 437–451 (2017). way. Oncol. Lett. 18, 5332–5340 (2019). 44. McCahill, A., Warwicker, J., Bolger, G. B., Houslay, M. D. & Yarwood, S. J. The 74. Zhang, H. et al. High expression of leucine‑rich repeat‑containing 8A is indica- RACK1 scaffold protein: a dynamic cog in cell response mechanisms. Mol. tive of a worse outcome of colon cancer patients by enhancing cancer cell Pharm. 62, 1261–1273 (2002). growth and metastasis. Oncol. Rep. 40, 1275–1286 (2018). 45. Bottini, S. et al. Post-transcriptional gene silencing mediated by microRNAs is 75. Lenzken, S. C., Loffreda, A. & Barabino, S. M. L. RNA splicing: a new player in the controlled by nucleoplasmic Sfpq. Nat. Commun. 8, 1189 (2017). DNA damage response. Int. J. Cell Biol. 2013, 153634 (2013). 46. Mora Gallardo, C. et al. Dido3-dependent SFPQ recruitment maintains efficiency 76. Xu, L., Yu, W., Xiao, H. & Lin, K. BIRC5 is a prognostic biomarker associated with in mammalian alternative splicing. Nucleic Acids Res. 47, 5381–5394 (2019). tumor immune cell infiltration. Sci. Rep. 11, 390 (2021). 47. Jaafar, L., Li, Z., Li, S. & Dynan, W. S. SFPQ•NONO and XLF function separately and 77. Siffroi-Fernandez, S. et al. Functional genomics identify Birc5/survivin as a together to promote DNA double-strand break repair via canonical non- candidate gene involved in the chronotoxicity of cyclin-dependent kinase homologous end joining. Nucleic Acids Res. 45, 1848–1859 (2017). inhibitors. Cell Cycle 13, 984–991 (2014). 48. Klotz-Noack, K. et al. SFPQ depletion is synthetically lethal with BRAF(V600E) in 78. Jo, S. et al. Inhibition of PCGF2 enhances granulocytic differentiation of acute colorectal cancer cells. Cell Rep. 32, 108184 (2020). promyelocytic leukemia cell line HL-60 via induction of HOXA7. Biochem. Bio- 49. Pellarin, I. et al. Splicing factor proline- and glutamine-rich (SFPQ) protein reg- phys. Res. Commun. 416,86–91 (2011). ulates platinum response in ovarian cancer-modulating SRSF2 activity. Onco- 79. Davis, W. & Tew, K. D. ATP-binding cassette transporter-2 (ABCA2) as a ther- gene 39, 4390–4403 (2020). apeutic target. Biochem. Pharmacol. 151, 188–200 (2018). 50. Robles, M. S., Boyault, C., Knutti, D., Padmanabhan, K. & Weitz, C. J. Identification 80. Sawada, G. et al. CHD8 is an independent prognostic indicator that regulates of RACK1 and protein kinase Calpha as integral components of the mammalian Wnt/β-catenin signaling and the cell cycle in gastric cancer. Oncol. Rep. 30, circadian clock. Science 327, 463–466 (2010). 1137–1142 (2013). 51. Becquet, D. et al. Evidence for an internal and functional circadian clock in rat 81. Masoud, G. N. & Li, W. HIF-1α pathway: role, regulation and intervention for pituitary cells. Mol. Cell Endocrinol. 382, 888–898 (2014). cancer therapy. Acta Pharm. Sin. B 5, 378–389 (2015). 52. Liao, Y., Tong, L., Tang, L. & Wu, S. The role of cold-inducible RNA binding 82. Warzecha, C. C., Sato, T. K., Nabet, B., Hogenesch, J. B. & Carstens, R. P. ESRP1 and protein in cell stress response. Int J. Cancer 141, 2164–2173 (2017). ESRP2 are epithelial cell-type-specific regulators of FGFR2 splicing. Mol. Cell 33, 53. White, E. J., Brewer, G. & Wilson, G. M. Post-transcriptional control of gene 591–601 (2009). expression by AUF1: mechanisms, physiological targets, and regulation. Biochim. 83. Kaprio, T., Hagström, J., Andersson, L. C. & Haglund, C. Tetraspanin CD63 inde- Biophys. Acta 1829, 680–688 (2013). pendently predicts poor prognosis in colorectal cancer. Histol. Histopathol. 35, 54. Hoekstra, M. M., Emmenegger, Y., Hubbard, J. & Franken, P. Cold-inducible RNA- 887–892 (2020). binding protein (CIRBP) adjusts clock-gene expression and REM-sleep recovery 84. Lai, X. et al. Decreased expression of CD63 tetraspanin protein predicts elevated following sleep deprivation. Elife 8, https://doi.org/10.7554/eLife.43400 (2019). malignant potential in human esophageal cancer. Oncol. Lett. 13,4245–4251 (2017). 55. Kim, H. R. et al. SRSF5: a novel marker for small-cell lung cancer and pleural 85. Tummers, B. & Green, D. R. Caspase-8: regulating life and death. Immunol. Rev. metastatic cancer. Lung Cancer 99,57–65 (2016). 277,76–89 (2017). 56. Kandoth, C. et al. Mutational landscape and significance across 12 major cancer 86. Fianco, G. et al. Caspase-8 contributes to angiogenesis and chemotherapy types. Nature 502, 333–339 (2013). resistance in glioblastoma. Elife 6, https://doi.org/10.7554/eLife.22593 (2017). 57. Lee, Y. H., Jhuang, Y. L., Chen, Y. L., Jeng, Y. M. & Yuan, R. H. Paradoxical 87. De Blasio, A. et al. Unusual roles of caspase-8 in triple-negative breast cancer cell overexpression of MBNL2 in hepatocellular carcinoma inhibits tumor growth line MDA-MB-231. Int J. Oncol. 48, 2339–2348 (2016). and invasion. Oncotarget 7, 65589–65601 (2016). 88. Comoglio, P. M., Trusolino, L. & Boccaccio, C. Known and novel roles of the MET 58. Anczuków, O. & Krainer, A. R. Splicing-factor alterations in cancers. RNA 22, oncogene in cancer: a coherent approach to targeted therapy. Nat. Rev. Cancer 1285–1301 (2016). 18, 341–358 (2018). 59. Blijlevens, M., Li, J. & van Beusechem, V. W. Biology of the mRNA splicing 89. Park, M. et al. A novel isoform of met receptor tyrosine kinase blocks hepatocyte machinery and its dysregulation in cancer providing therapeutic opportunities. growth factor/Met signaling and stimulates skeletal muscle cell differentiation. J. Int. J. Mol. Sci. 22, https://doi.org/10.3390/ijms22105110 (2021). Biol. Chem. 290, 1804–1817 (2015). 60. da Silva, M. R. et al. Splicing regulators and their roles in cancer biology and 90. Guo, C., Liu, S., Wang, J., Sun, M. Z. & Greenaway, F. T. ACTB in cancer. Clin. Chim. therapy. Biomed. Res. Int. 2015, 150514 (2015). Acta 417,39–44 (2013). npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. 91. Matsuda, Y., Hagio, M., Seya, T. & Ishiwata, T. Fibroblast growth factor receptor 2 123. Kanehisa, M., Sato, Y. & Kawashima, M. KEGG mapping tools for uncovering IIIc as a therapeutic target for colorectal cancer cells. Mol. Cancer Therapeutics hidden features in biological data. Protein Sci. 31,47–53 (2022). 11, 2010–2020 (2012). 124. Keshava Prasad, T. S. et al. Human Protein Reference Database-2009 update. 92. Isacke, C. M. & Yarwood, H. The hyaluronan receptor, CD44. Int J. Biochem. Cell Nucleic Acids Res. 37, D767–772 (2009). Biol. 34, 718–721 (2002). 125. Snel, B., Lehmann, G., Bork, P. & Huynen, M. A. STRING: a web-server to retrieve 93. Coppola, D. et al. CD44V6 expression in human colorectal carcinoma. Hum. and display the repeatedly occurring neighbourhood of a gene. Nucleic Acids Pathol. 29, 627–635 (1998). Res. 28, 3442–3444 (2000). 94. Mikami, T. et al. Decreased expression of CD44, alpha-catenin, and deleted colon 126. Stark, C. et al. BioGRID: a general repository for interaction datasets. Nucleic Acids carcinoma and altered expression of beta-catenin in ulcerative colitis-associated Res. 34, D535–539 (2006). dysplasia and carcinoma, as compared with sporadic colon neoplasms. Cancer 127. Bovolenta, L. A., Acencio, M. L. & Lemke, N. HTRIdb: an open-access database for 89, 733–740 (2000). experimentally verified human transcriptional regulation interactions. BMC 95. Gotley, D. C. et al. Alternatively spliced variants of the cell adhesion molecule CD44 Genomics 13, 405 (2012). and tumour progression in colorectal cancer. Br. J. Cancer 74,342–351 (1996). 128. Yu, G., Wang, L. G., Han, Y. & He, Q. Y. clusterProfiler: an R package for comparing 96. Todaro, M. et al. CD44v6 is a marker of constitutive and reprogrammed cancer biological themes among gene clusters. Omics 16, 284–287 (2012). stem cells driving colon cancer metastasis. Cell Stem Cell 14, 342–356 (2014). 129. Walter, W., Sánchez-Cabo, F. & Ricote, M. GOplot: an R package for visually 97. Saito, S. et al. CD44v6 expression is related to mesenchymal phenotype and poor combining expression data with functional analysis. Bioinformatics 31, prognosis in patients with colorectal cancer. Oncol. Rep. 29, 1570–1578 (2013). 2912–2914 (2015). 98. Li, M. et al. Drug delivery systems based on CD44-targeted glycosaminoglycans for cancer therapy. Carbohydr. Polym. 251, 117103 (2021). 99. Burnichon, N. et al. SDHA is a tumor suppressor gene causing paraganglioma. ACKNOWLEDGEMENTS Hum. Mol. Genet. 19, 3011–3020 (2010). We are gratefull to all members of Relógio group for their critical remarks and 100. Yu, W., Tang, L., Lin, F., Yao, Y. & Shen, Z. DGKZ acts as a potential oncogene in feedback. Vector icons used in this manuscript were obtained from Flaticon (https:// osteosarcoma proliferation through its possible interaction With ERK1/2 and www.flaticon.com/) and modified accordingly. MYC pathway. Front Oncol. 8, 655 (2018). 101. Heldin, C. H. Targeting the PDGF signaling pathway in tumor treatment. Cell Commun. Signal 11, 97 (2013). AUTHOR CONTRIBUTIONS 102. Dallmann, R., Okyar, A. & Levi, F. Dosing-time makes the poison: Circadian A.R. conceptualized the study, attained the funding, carried out the investigation, regulation and pharmacotherapy. Trends Mol. Med 22, 430–445 (2016). provided supervision, prepared the initial draft, reviewed and revised the manuscript. 103. Lévi, F., Okyar, A., Dulong, S., Innominato, P. F. & Clairambault, J. Circadian timing D.M. carried out the transcriptome data analysis, visualization, prepared the initial in cancer treatments. Annu. Rev. Pharm. Toxicol. 50, 377–421 (2010). draft, reviewed and revised the manuscript. A.B. carried out in vitro experiments, 104. Giacchetti, S. et al. Phase III trial comparing 4-day chronomodulated therapy reviewed and revised the manuscript. All authors have read and agreed to the final versus 2-day conventional delivery of fluorouracil, leucovorin, and oxaliplatin as version of the manuscript. first-line chemotherapy of metastatic colorectal cancer: the European Organi- sation for Research and Treatment of Cancer Chronotherapy Group. J. Clin. Oncol. 24, 3562–3569 (2006). FUNDING 105. Alamoudi, A. A. Why do cancer cells break from host circadian rhythm? Insights from unicellular organisms. Bioessays 43, e2000205 (2021). The work in the group of A.R. was funded by the Dr. Rolf M. Schwiete Stiftung. A.B. 106. Pliatsika, V. & Rigoutsos, I. “Off-Spotter”: very fast and exhaustive enumeration of was additionally funded by the Berlin School of Integrative Oncology (BSIO) graduate genomic lookalikes for designing CRISPR/Cas guide RNAs. Biol. Direct 10, 4 (2015). program funded within the German Excellence Initiative, and the German Cancer 107. Hodgkins, A. et al. WGE: a CRISPR database for genome engineering. Bioinfor- Consortium (GCC). Open Access funding enabled and organized by Projekt DEAL. matics 31, 3078–3080 (2015). 108. Zhang, Z. et al. Genome-wide effect of pulmonary airway epithelial cell-specific Bmal1 deletion. FASEB J. 33, 6226–6238 (2019). COMPETING INTERESTS 109. Pembroke, W. G., Babbs, A., Davies, K. E., Ponting, C. P. & Oliver, P. L. Temporal The authors declare no competing interests. transcriptomics suggest that twin-peaking genes reset the clock. Elife 4, https:// doi.org/10.7554/eLife.10518 (2015). 110. Ikeda, R. et al. REV-ERBα and REV-ERBβ function as key factors regulating ADDITIONAL INFORMATION mammalian Circadian output. Sci. Rep. 9, 10171 (2019). Supplementary information The online version contains supplementary material 111. Fuhr, L. A. The Circadian Clock Modulates Tumour Progression and Drug Response available at https://doi.org/10.1038/s41540-022-00225-w. in Colorectal Cancer Cells through Metabolic Phenotype Rewiring (Humboldt- Universität zu Berlin, 2019). Correspondence and requests for materials should be addressed to Angela Relógio. 112. Simon, A. FastQC: a quality control tool for high throughput sequence data. Avail- able online: https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ (2010). Reprints and permission information is available at http://www.nature.com/ 113. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina reprints sequence data. Bioinformatics 30, 2114–2120 (2014). 114. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29,15–21 Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims (2013). in published maps and institutional affiliations. 115. Patro, R., Duggal, G., Love, M. I., Irizarry, R. A. & Kingsford, C. Salmon provides fast and bias-aware quantification of transcript expression. Nat. Methods 14, 417–419 (2017). 116. Soneson, C., Love, M. I. & Robinson, M. D. Differential analyses for RNA-seq: Open Access This article is licensed under a Creative Commons transcript-level estimates improve gene-level inferences. F1000Res 4, 1521 (2015). Attribution 4.0 International License, which permits use, sharing, 117. Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package adaptation, distribution and reproduction in any medium or format, as long as you give for differential expression analysis of digital gene expression data. Bioinformatics appropriate credit to the original author(s) and the source, provide a link to the Creative 26, 139–140 (2009). Commons license, and indicate if changes were made. The images or other third party 118. Thaben, P. F. & Westermark, P. O. Detecting rhythms in time series with RAIN. J. material in this article are included in the article’s Creative Commons license, unless Biol. Rhythms 29, 391–400 (2014). indicated otherwise in a credit line to the material. If material is not included in the 119. Cornelissen, G. Cosinor-based rhythmometry. Theor. Biol. Med. Model. 11,16(2014). article’s Creative Commons license and your intended use is not permitted by statutory 120. Carlucci, M. et al. DiscoRhythm: an easy-to-use web application and R package regulation or exceeds the permitted use, you will need to obtain permission directly for discovering rhythmicity. Bioinformatics 36, 1952–1954 (2019). from the copyright holder. To view a copy of this license, visit http://creativecommons. 121. Trincado, J. L. et al. SUPPA2: fast, accurate, and uncertainty-aware differential org/licenses/by/4.0/. splicing analysis across multiple conditions. Genome Biol. 19, 40 (2018). 122. Thaben, P. F. & Westermark, P. O. Differential rhythmicity: detecting altered © The Author(s) 2022 rhythmicity in biological data. Bioinformatics 32, 2800–2808 (2016). Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png npj Systems Biology and Applications Springer Journals

Transcriptome analysis of clock disrupted cancer cells reveals differential alternative splicing of cancer hallmarks genes

Loading next page...
 
/lp/springer-journals/transcriptome-analysis-of-clock-disrupted-cancer-cells-reveals-c7FI3r7umv

References (135)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2022
eISSN
2056-7189
DOI
10.1038/s41540-022-00225-w
Publisher site
See Article on Publisher Site

Abstract

www.nature.com/npjsba ARTICLE OPEN Transcriptome analysis of clock disrupted cancer cells reveals differential alternative splicing of cancer hallmarks genes 1,2,3 1,2,3 1,2,3✉ Deeksha Malhan , Alireza Basti and Angela Relógio Emerging evidence points towards a regulatory role of the circadian clock in alternative splicing (AS). Whether alterations in core- clock components may contribute to differential AS events is largely unknown. To address this, we carried out a computational analysis on recently generated time-series RNA-seq datasets from three core-clock knockout (KO) genes (ARNTL, NR1D1, PER2) and WT of a colorectal cancer (CRC) cell line, and time-series RNA-seq datasets for additional CRC and Hodgkin’s lymphoma (HL) cells, murine WT, Arntl KO, and Nr1d1/2 KO, and murine SCN WT tissue. The deletion of individual core-clock genes resulted in the loss of KO KO KO circadian expression in crucial spliceosome components such as SF3A1 (in ARNTL ), SNW1 (in NR1D1 ), and HNRNPC (in PER2 ), KO which led to a differential pattern of KO-specific AS events. All HCT116 cells showed a rhythmicity loss of a crucial spliceosome gene U2AF1, which was also not rhythmic in higher progression stage CRC and HL cancer cells. AS analysis revealed an increase in alternative first exon events specificto PER2 and NR1D1 KO in HCT116 cells, and a KO-specific change in expression and rhythmicity KO KO pattern of AS transcripts related to cancer hallmarks genes including FGFR2 in HCT116_ARNTL , CD44 in HCT116_NR1D1 , and KO MET in HCT116_PER2 . KO-specific changes in rhythmic properties of known spliced variants of these genes (e.g. FGFR2 IIIb/FGFR2 IIIc) correlated with epithelial-mesenchymal-transition signalling. Altogether, our bioinformatic analysis highlights a role for the circadian clock in the regulation of AS, and reveals a potential impact of clock disruption in aberrant splicing in cancer hallmark genes. npj Systems Biology and Applications (2022) 8:17 ; https://doi.org/10.1038/s41540-022-00225-w INTRODUCTION Discs Large MAGUK Scaffold Protein 1 (DLG1) leading to the production of oncogenic isoforms, or to the reduction of the Alternative splicing (AS) is a central event in transcriptional tumour-suppressing isoforms of those genes . Also, aberrant AS regulation leading to distinct mRNAs being generated from a 1,2 in other members of the SR splicing factor and SR-like families, single gene, and is thus a major contributor for protein diversity . including SRSF4 and Transformer 2 Beta Homolog (TRA2β) have Approximately 95% of human genes undergo AS in a tissue type- been shown to regulate cancer proliferation and metastasis by specific manner , and AS resulting products lead to the affecting the splicing of genes involved in epithelial-mesenchymal production of proteins involved in numerous biological processes 4 5 transition (EMT, e.g., CD44) in breast cancer . Mounting evidence including cell cycle and metabolism . Splicing factors (SFs) mark suggests as well an association between AS regulation and the the splice sites, facilitate spliceosome assembly during the splicing 18,19 circadian clock in various organisms in particular in mammals . process, and can regulate AS decisions and thus the resulting Circadian rhythms are driven by an endogenous oscillator with splicing product(s) . Dysregulations in splicing account for the an approximate period of 24 h and regulate the timing of cellular production of non-functional, or with the incorrect function, processes in most organisms . These rhythms are generated by a protein products with an impact on the onset and progression of 7–12 Transcriptional-Translational Feedback Loop (TTFL) in which Brain several diseases among which cancer . A systematic evaluation and muscle ARNT-Like 1 (BMAL1), also known as ARNTL forms a of 11 different tumour types revealed a subset of transcript heterodimer with Circadian Locomoter Output Cycles Kaput isoform switches associated with gain and/or loss of protein (CLOCK) and regulates the transcriptional activation of Period 1/ domains, which are associated with the hallmarks of cancer . 2/3 (PER 1/2/3), and Cryptochrome 1/2 (CRY 1/2), which inhibit Known examples of AS failures related to cancer include the BMAL1-CLOCK-mediated transcription . In addition, Nuclear aberrant splicing in the Vascular Epithelial Growth Factor (VEGF), Receptor Subfamily 1 Group D Member 1, 2 (NR1D1/2) and RAR essential for angiogenesis, resulting in one transcript isoform with anti-angiogenic properties in cancer . Another example is CD44, Related Orphan Receptor A, B, C (RORA/B/C) regulate BMAL1 which encodes for a cell-surface protein and can act as both transcription and contribute to the fine-tuning of its expression. tumour-suppressing, as well as invasiveness-promoting molecule, The circadian clock regulates the expression of genes and proteins depending on its different splice variants . In addition, alterations related to several cancer hallmarks like cell proliferation, in the expression of several SFs can impact the splicing of their metabolism, and DNA damage . Disruption of circadian rhythms target genes, for e.g., the differential expression of the splicing is associated with increased risk of cancer, including prostrate, factor Serine and Arginine Rich Splicing Factor 6 (SRSF6) affects breast, and colon cancer . the splicing of several oncogenes including the Insulin Receptor Recent work points to a complex cross-talk between the circadian 24–31 (INSR), MAPK Interacting Serine/Threonine Kinase 2 (MKNK2) and clock and AS events underlying different cancer types .Changes Institute for Theoretical Biology (ITB), Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin, Humboldt – Universität zu Berlin, and Berlin Institute of Health, Berlin 10117, Germany. Molecular Cancer Research Center (MKFZ), Medical Department of Hematology, Oncology, and Tumor Immunology, Charité – Universitätsmedizin Berlin, corporate member of Freie Universität Berlin Humboldt – Universität zu Berlin, and Berlin Institute of Health, Berlin 10117, Germany. Institute for Systems Medicine, Faculty of Human Medicine, MSH Medical School Hamburg, Hamburg 20457, Germany. email: angela.relogio@charite.de Published in partnership with the Systems Biology Institute 1234567890():,; D. Malhan et al. in amplitude and peak phase of SFs including hnRNPs and RBM RESULTS proteins in colorectal cancer (CRC) cells were correlated with altered Rhythmicity analysis of mammalian RNA-seq datasets showed splicing outcome of target genes like VEGFA and CD44 .Alater a KO-specific effect in the circadian phenotype both at the study in the same cell lines reported rhythmic alternative splicing gene and at the transcript levels events of cancer-related genes like Cold inducible RNA binding In mammals, many different biological processes like metabolism protein (CIRBP) and Poly(RC) Binding Protein 2 (PCBP2) . Clock- or the immune system are regulated by the endogenous circadian mediated alternative splicing events in the cell differentiation 32 clock . To investigate the effect of clock disruption on the marker Interferon Regulatory Factor 4 (IRF4) were reported in an circadian phenotype (circa 24 h rhythmicity) and the subsequent in vitro model of Hodgkin lymphoma (HL) .Inthe same HL cell effects in biological processes, we analysed recent datasets from lines, clock-mediated changes in the expression of transcript our group for three core-clock knockout (KO) cell lines isoforms were reported in elements associated with the tumour KO KO KO (HCT116_ARNTL , HCT116_NR1D1 , and HCT116_PER2 ) and necrosis factor (TNF) pathway, which controls cell proliferation, compared their gene expression pattern to the corresponding migration, and apoptosis . Rhythmic alternative splicing of the HCT116_WT (Fig. 1a, Supplementary Fig. 1d). Previous in vitro and splicing factor U2AF26 controls the stability of Per1 in mice .A in vivo studies showed an effect of ARNTL, PER2 and NR1D1 KO on splice variant of human BMAL1 (hBMAL1a) is known to act opposite the rhythmic expression of core-clock genes (mostly via biolumi- to its canonical isoform (hBMAL1b) and functions as a negative 33,34 nescence recordings, or mouse studies) , which are in regulator of the circadian clock . However, the impact in alternative agreement with our findings regarding the loss (ARNTL KO) and splicing events resulting from clock alterations in cancer remains period alterations (PER2 and NR1D1 KO) in the respective unexplored. oscillation profiles. Although, it is important to notice that the In this study, we analysed time-course RNA-seq datasets assessment of rhythmicity via in vitro bioluminescence measure- generated from three HCT116 core-core knockouts (ARNTL, PER2, ments of ARNTL promoter activity, or via in vivo analysis of NR1D1) and wild type (WT) cells, to investigate the impact of clock movement in mice, is not directly comparable with RNA-seq time disruption on AS events in CRC cell lines. In addition, we retrieved course measurements. publicly available time-course RNA-seq datasets from human CRC To pinpoint specific changes due to the KO of core-clock genes, cell lines (SW480, SW620), HL cell lines (HDMYZ, L1236), and we additionally gathered publicly available time-course RNA-seq mouse Arntl KO, Nr1d1/2 KO, and SCN (Suprachiasmatic nucleus) datasets for human CRC cell lines (SW480: derived from the WT tissue. Our study revealed that deletion of core-clock genes primary tumour, SW620: derived from a metastasis site, from the (ARNTL, PER2 or NR1D1) altered the circadian rhythmic pattern of same patient), HL cell lines (HDMYZ: stage III, L1236: stage IV), SFs, as well as the resulting AS events in a KO-specific manner. We murine WT and Arntl KO, murine WT and Nr1d1/2 KO, murine SCN observed a common rhythmicity loss of crucial SFs such as U2 tissue, murine WT and Cry1/2 KO, and murine WT and Per1/2 KO, KO Small Nuclear RNA Auxiliary Factor 1(U2AF1) in all three HCT116 , resulting in a total of 222 samples (Fig. 1a). To examine if the as well as in SW620, and L1236 cells. Moreover, we found a phase rhythmicity pattern varies when analysed at the transcript level vs shift in circadian transcripts of SFs such as Receptor of Activated C gene level, we extracted 24 h rhythmic sets both at gene and Kinase 1 (RACK1), Heterogeneous nuclear ribonucleoprotein D transcript levels in six-mammalian time course RNA-seq datasets (HNRNPD) involved in pre-mRNA processing and Serine and (Fig. 1b–d, Supplementary Figs. 1–3). A higher number of 24 h Arginine rich splicing factor 5 (SRSF5), component of the rhythmic features were found at transcript level vs gene level in all KO KO spliceosome machinery, in HCT116_ARNTL , HCT116_NR1D1 , datasets. The number of circadian (~24 h period) expressed genes KO KO and HCT116_PER2 vs. WT, respectively. This led to a distinct and transcripts decreased in all three HCT116 vs. WT (Fig. 1b–d). At the gene level, we identified 1261, 443, 866 and 525 circadian pattern of AS events (specifically in gain and loss of splicing KO KO KO gene sets in HCT116_WT, HCT116_ARNTL , HCT116_NR1D1 ,and events) in HCT116 datasets vs. HCT116_WT. Differential KO HCT116_PER2 , respectively (Fig. 1b; left panel). We observed KO- rhythmicity analysis revealed that the majority of phase shifted KO specific loss in circadian expression of crucial genes such as Keratin transcript pairs consisted of protein-coding biotypes. In HCT116 8(KRT8), which maintains cellular structural integrity, Transmem- cells, genes with phase-shifted transcript pairs were related to KO brane and Coiled-Coil Domain Family 1 (TMCC1), which promotes apoptotic signalling (HCT116_ARNTL ), macrophage proliferation KO endoplasmic reticulum-associated fission, and HECT and RLD (HCT116_NR1D1 ), and negative regulation of chromatin organi- KO Domain Containing E3 Ubiquitin Protein Ligase 3 (HERC3), which zation (HCT116_PER2 ). Moreover, we identified differentially KO enables protein-transferase activity in HCT116_ARNTL , rhythmic transcript isoform pairs of target genes associated with KO KO HCT116_NR1D1 , and in HCT116_PER2 , respectively. At tran- cancer hallmarks. These included Casein Kinase 2 Alpha 1 script level, we observed 5499 (corresponding to 4063 genes), 3039 (CSNK2A1), related to invasion, metastasis, tumour proliferation, (2485 genes), 3519 (2899 genes), 2850 (2345 genes) circadian KO in HCT116_WT and HCT116 . We further identified phase shifted KO KO transcripts in HCT116_WT, HCT116_ARNTL , HCT116_NR1D1 , transcript pairs from cancer hallmarks-related genes e.g., tran- KO and HCT116_PER2 , respectively (Fig. 1c; left panel). Among KO- scripts pairs from MET Proto Oncogene (MET), which controls specific loss of circadian expressed transcripts, we saw a loss of KO several cancer hallmarks, in HCT116_PER2 . Finally, we observed circadian transcript expression of C-Terminal Binding Protein 1 KO-specific altered expression and rhythmicity pattern of tran- (CTBP1), which is a transcriptional repressor, HRAS an oncogene scripts associated with cancer hallmarks related genes such as in involved in signal transduction pathways, and Calmodulin 1 Fibroblast Growth Factor Receptor 2 (FGFR2), which promotes (CALM1), which controls several proteins including enzymes and KO tumour progression seen in HCT116_ARNTL , Caspase 8 (CASP8) ion channels, through calcium-binding. All these genes modulate KO KO which regulates extrinsic apoptotic pathway seen in cellular proliferation in HCT116_ARNTL , HCT116_NR1D1 ,and in KO KO HCT116_NR1D1 , and HRas proto-oncogene, GTPase (HRAS) HCT116_PER2 , respectively. involved in cell proliferation, growth and apoptosis detected in Moreover, differences in the percentage of biotypes (e.g., KO HCT116_PER2 . The evaluation of coding exon sequences for protein-coding biotype) were also observed between circadian selected cancer hallmarks revealed the altered rhythmicity in gene and transcript sets (Fig. 1b, c; right panel). For example, 1089 spliced variants associated with cancer progression such as genes (~86%) and 3509 transcripts (~63%) with protein-coding skipped exon 14 variant of MET. Our findings point to a direct biotype were circadian expressed in HCT116_WT. The comparison impact of circadian clock disruption in AS events related to cancer of rhythmicity at gene and transcript levels in HCT116 datasets hallmarks, which may support cancer onset or progression. showed a roughly linear relationship between the number of npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute 1234567890():,; D. Malhan et al. Fig. 1 RNA-seq data analysis showed differences in the circadian phenotype between gene- and transcript expression values. a Schematic representation of the pipeline for RNA-seq data analysis. The input data consisted of 222 samples from eight mammalian time- course RNA-seq data. Murine datasets from Per1/2 KO (GSE171975) and Cry1/2 KO (GSE135898) were only analysed at the gene level. b– d Classification of 24 h rhythmic features in HCT116 datasets at (b) gene level and (c) transcript level showed a decline in the total number of circadian expressed genes/transcript-sets and changes in their biotypes due to core-clock disruption. d Scatter plot represents each of the HCT116 WT and KO cells according to the total number of 24 h rhythmic genes (x-axis) and the total number of 24 h rhythmic transcripts (y- axis). KO KO circadian genes and circadian transcripts (Fig. 1d). Similar HCT116_ARNTL , 521 genes (~60%) in HCT116_NR1D1 , and KO differences in gene- and transcript-level rhythmicity phenotype 274 genes (~52%) in HCT116_PER2 were rhythmic both at gene were also seen in other human CRC cell lines (SW480; SW620), HL and transcript levels (Supplementary Fig. 1a). We further analysed cell lines (HDMYZ; L1236), murine WT and Arntl KO, murine WT circadian discrepancies between the expression pattern of some and Nr1d1/2 KO, and murine SCN WT tissue datasets (Supple- genes and their corresponding transcripts, for e.g., HIF1A was mentary Figs. 2, 3). Within the set of circadian rhythmic genes, 846 found to be circadian expressed whereas its transcripts were genes (~67%) in HCT116_WT, 192 genes (~43%) in arrhythmic in HCT116_WT. Here, the summarization of arrhythmic Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. Fig. 2 Circadian expression of splicing factors and spliceosome machinery related transcripts in mammalian RNA-seq datasets. Median KO normalized phase-sorted heatmaps (left) and acrophase bins (right) of SFs circadian transcripts (tr.) in (a) HCT116_WT and HCT116 cells (upper panel), SW480 and SW620 cell lines (lower left panel), HDMYZ and L1236 cell lines (lower right panel), b murine WT and Arntl KO from distal bronchiolar epithelium tissue (upper left panel) and from primary tracheal epithelial cells (upper right panel), murine WT and Nr1d1/2 KO from epithelial cells (lower left panel), and murine SCN WT tissue (lower right panel). transcripts resulted in a rhythmic gene. Minichromosome Main- Moreover, our analysis also shows that the KO of core-clock genes tenance Complex Component 2 (MCM2) was rhythmic both at altered the ratio, as well as the circadian rhythmicity expression pattern of genes and transcripts in a KO-specific manner. Hence, gene and transcript levels in HCT116_WT. In contrast, Interleukin 18 (IL18) was identified as arrhythmic at the gene level whereas its we focused on transcript level in our subsequent analysis. transcripts were circadian expressed in HCT116_WT. In this case, the summarization of rhythmic transcripts resulted in an Alterations in core-clock elements influenced the rhythmic arrhythmic gene (Supplementary Fig. 1a; lower panel). Further, expression of splicing factors we analysed the changes in the number of transcripts per To investigate possible alteration in SFs rhythmicity resulting from expressed gene and the number of ~24 h transcripts per gene in clock disruption, we examined the expression of SF transcripts and KO HCT116 vs. WT (Supplementary Fig. 1b, c). The number of their rhythmic properties (Fig. 2). We compiled a total of 534 24–26,35,36 expressed transcripts per gene showed variation in all three human SF genes from previously published studies , KO 37 38 HCT116 vs. WT. For instance, 28 transcripts of Heterogeneous Spliceosome database , and SpliceAid 2 database (Supplemen- Nuclear Ribonucleoprotein H1 (HNRNPH1), an RNA binding tary file 1), and generated orthologous sets of human SFs for protein, were expressed in HCT116_WT. Whereas, 25 transcripts mouse (521 genes) using Ensembl Biomart (Supplementary file KO from HNRNPH1 were expressed in HCT116_ARNTL , 21 transcripts 2), and mapped the curated list of SFs against the circadian KO KO in HCT116_NR1D1 , and 24 transcripts in HCT116_PER2 . Out of transcripts extracted from the RNA-seq datasets as depicted in the the expressed transcripts of HNRNPH1, 5 transcripts in HCT116_WT phase sorted heatmaps and acrophase bin plots (Fig. 2). We saw a KO and 1 transcript in HCT116_PER2 were circadian expressed. decrease in the number of circadian expressed SF transcripts and KO Altogether, our analysis confirmed previous findings from our also changes in their acrophase distribution in HCT116 group , and showed that indeed a transcript-level analysis can compared to HCT116_WT cells (Fig. 2a; upper panel). KO provide new results that an be masked by gene-level analysis. HCT116_PER2 showed the lowest number of circadian rhythmic npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. KO SF transcripts (143 transcripts; 104 genes) followed by spliceosome formation (Fig. 3a). For HCT116_NR1D1 , differential KO KO HCT116_NR1D1 (188 transcripts; 135 genes), HCT116_ARNTL rhythmic SFs transcripts were involved in pre-mRNA processing (204 transcripts; 150 genes), and HCT116_WT (341 transcripts; 216 like HNRNPD and translational repressor like Cold Inducible RNA KO KO KO genes) datasets. Both HCT116_NR1D1 and HCT116_PER2 Binding Protein (CIRBP) (Fig. 3b). Whereas in HCT116_PER2 ,we showed a shift in the SFs acrophase when compared with found differentially rhythmic transcripts of SFs like SRSF5, which is HCT116_WT cells. a part of the spliceosome machinery (Fig. 3c). To find a possible association between the decrease in the To further understand the discrepancies in SFs rhythmicity, we number of rhythmic SF transcripts and cancer progression, we extracted a list of SFs with at least one circadian transcript in analysed other CRC cell lines (SW480, SW620) and HL cell lines HCT116_WT, SW480, HDMYZ, murine WT (from Arntl KO), murine (HDMYZ, L1236). Indeed, we observed a decrease in the number WT (from Nr1d1/2 KO), and mapped them with their correspond- of rhythmic SF transcripts in the metastasis-derived colon cancer ing knockout/altered condition (e.g. primary tumour- vs. (SW620: 216 transcripts; 153 genes) and in stage IV HL (L1236: 215 metastasis-derived cells). We produced categorical heatmaps to transcripts; 156 genes) as compared with the primary colon cancer visualize the loss of rhythmically expressed SFs in each dataset (SW480: 378 transcripts; 212 genes) and stage III HL (HDMYZ: 325 compared with controls (Fig. 3d–h). 216 SFs with at least one transcripts; 201 genes) cell lines, respectively (Fig. 2a; lower panel). rhythmic transcript were observed in HCT116_WT, and less than KO Also, the shift in acrophase distribution of circadian rhythmic SF 50% were rhythmic in HCT116 (Fig. 3d; left panel). Out of 216 transcripts was observed in both SW620 and L1236 cell lines SFs with at least one circadian transcript in HCT116_WT, we KO KO compared with their controls. To further examine the impact of observed 85 SFs in HCT116_ARNTL , 79 SFs in HCT116_NR1D1 , KO each KO on SFs rhythmicity, we examined rhythmic changes in SFs and 66 SFs in HCT116_PER2 with at least one circadian KO transcripts within murine RNA-seq datasets (Fig. 2b). In contrast, transcript. Among the three HCT116 cells, there were 69 slightly higher number of rhythmic SF transcripts were observed common SFs with no rhythmic transcript (Fig. 3d; right panel). in Arntl KO model of distal bronchiolar epithelium (208 transcripts; Similarly, the human CRC cell line (SW620), HL cell line (L1236), 151 genes) and primary tracheal epithelial cells (48 transcripts; 44 murine Arntl KO, and murine Nr1d1/2 KO also showed a loss of genes) compared with their WT (175 transcripts; 138 genes and 46 rhythmic SFs compared with their controls (Fig. 3e–h). All three KO transcripts; 45 genes, respectively) (Fig. 2b; upper panel). Whereas, HCT116 cells showed loss of rhythmicity for U2AF1, an important Nr1d1/2 KO mouse model showed lower number of rhythmic SF RNA splicing mediator gene, which also lost rhythmicity in SW620 transcripts (101 transcripts; 87 genes) compared with WT (107 and L1236 cells, as compared to their corresponding lower grade transcripts; 87 genes) (Fig. 2b). SCN tissue, the master clock, cancer cell lines. We observed KO-specific loss of rhythmicity in showed the highest number of rhythmic SF transcripts (767 different SF transcripts such as RNA Binding Motif Protein 5 transcripts; 384 genes) in mouse WT datasets (Fig. 2b; lower panel, (RBM5), Splicing factor 3a subunit 1 (SF3A1), and Serine and for the complete list of circadian transcripts and their acrophases Arginine Rich Splicing Factor 3 (SRSF3), involved in the spliceo- KO see Supplementary Fig. 4). In addition, we also observed loss in some machinery, uniquely in HCT116_ARNTL . While, KO 24 h rhythmic SFs due to Cry1/2 deletion (Supplementary Fig. 5b; HCT116_NR1D1 resulted in rhythmicity loss of SNW Domain left panel) and Per1/2 deletion (Supplementary Fig. 5b; right panel) Containing 1 (SNW1), another spliceosome component, and KO vs. their controls in mouse datasets. HCT116_PER2 resulted in the loss of rhythmic transcripts of The KOs resulted in a shift in the acrophase bin distribution of HNRNPC, involved in pre-mRNA processing. Taken together, our circadian rhythmic SF transcripts in particular for NR1D1 and PER2 results point to alterations in rhythmic properties of SFs in HCT116 KO. We further carried out pairwise comparisons of circadian SF cell lines, which were KO-specific and were also present in KO transcripts between HCT116_WT and HCT116 to identify datasets derived from other cells and tissues, upon perturbation of commonly circadian rhythmic SF transcripts, and to obtain the core-clock. differentially rhythmic SF transcripts (phase shift ≥3 h) in the KO HCT116 compared with HCT116_WT (Fig. 3a–c). A total of 25 Disruption of circadian clock elements influenced alternative circadian SFs transcripts (23 genes) were found rhythmic in splicing events across human and murine datasets KO HCT116_WT and HCT116_ARNTL , out of those, 7 transcripts Aberrant AS events have been reported to be associated with showed differential rhythmicity (Fig. 3a). Besides alterations in the different types of cancer . We investigated the impact of clock expression profile of SF transcripts, we observed phase shift in disruption in our datasets on AS events with a potential impact in other transcripts, as well as loss of rhythmic oscillations in PER1, cancer onset and progression. We used SUPPA to calculate the PER2 and DBP gene with ARNTL disruption in HCT116 cells proportion spliced-in (PSI) of seven basic modes of AS: (1) (Supplementary Fig. 6a). Alternative 3′ splice site (A3-event), (2) Alternative 5′ splice site Differential rhythmicity analysis between HCT116_WT and KO (A5-event), (3) Alternative first exon (AF-event), (4) Alternative last HCT116_ARNTL cells resulted in a higher number of transcripts exon (AL-event), (5) Mutually exclusive event (MX-event), (6) with phase difference greater than 9 h whereas for most Retention intron event (RI-event), and (7) Skipping exon event (SE- differentially rhythmic splicing factor transcripts a phase differ- event). ence between 6 and 9 h was observed (Supplementary Fig. 6b–f). KO We determined the overall number of genes showing For HCT116_WT vs. HCT116_NR1D1 , we found 33 circadian significant AS events (0.1< PSI < 0.9), and found that most genes rhythmic transcripts (28 genes) and 11 circadian rhythmic had a commonly predominant AS event (SE-event) in all human transcripts to be differentially rhythmic (Fig. 3b). The comparison KO RNA-seq datasets (Supplementary Fig. 7). To further evaluate the of HCT116_WT and HCT116_PER2 resulted in a total of 28 changes in AS events caused by the KO of core-clock genes, we circadian expressed transcripts (26 genes) and out of those 4 classified the set of genes that were alternatively spliced in the transcripts showed differential rhythmicity (Fig. 3c). Only a single WT, but not in the KOs (following the PSI criteria), as “event loss” in transcript from the gene survival of motor neuron 1 (SMN1), which its corresponding KO group. Similarly, uniquely identified sets of plays a catalyst role in the assembly of small nuclear ribonucleo- genes alternatively spliced in a KO cell line were labelled as “event proteins (snRNPs) was found to be rhythmic in all HCT116 WT and gain” (see “Methods” for details). Bar plots were used to depict the KO datasets and it showed the same phase, but different KO total number of genes in the categories loss and gain for each amplitude within HCT116 . Out of seven differentially rhythmic KO transcripts in HCT116_ARNTL vs. WT, were SFs like RACK1, which type of splicing event (Fig. 4). We mapped the list of AS genes with is a component of the ribosomal subunit and Splicing factor the genes containing at least one circadian transcript and their proline and glutamine rich (SFPQ), essential at the early stages of biotypes. For better visualization, we separated the genes Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. Fig. 3 Clock disruption affected rhythmicity of splicing factors in human and murine datasets. a–c Circular plots depict the distribution of KO peak phases of overlapping differentially rhythmic SF transcripts between (a) HCT116_WT and HCT116_ARNTL ,(b) HCT116_WT and KO KO KO HCT116_NR1D1 , and (c) HCT116_WT and HCT116_PER2 . d–h Categorial heatmaps represent the loss of rhythmic SFs in (d) HCT116 vs. WT, (e) SW620 vs. SW480, (f) L1236 vs. HDMYZ, (g) mouse Arntl KO vs. WT in distal bronchiolar epithelium (left) and primary epithelial cells (right), and (h) Nr1d1/2 KO vs. WT. Green colour indicates SFs with at least one ~24 h transcript whereas grey colour indicates SFs with no rhythmic transcript. The numbers above the categorial heatmaps indicate the total number of SFs (genes) with rhythmic transcript(s). according to whether or not they contained transcripts with carried out a functional enrichment analysis, and found biological KO protein-coding biotype. All AS events in HCT116 cells except for processes such as protein acetylation enriched in common AF- KO KO the AF-event in HCT116_NR1D1 and HCT116_PER2 showed gain candidates, positive regulation of cellular response to TGF- KO higher number of genes present in the loss of splicing events vs. beta stimulus enriched in HCT116_NR1D1 unique candidates, gain of splicing events (Fig. 4a–c). However, a large increase in the and endothelial cell proliferation unique AF-event in KO number of genes with AF-gain was observed in both HCT116_PER2 candidates (Supplementary Fig. 8). KO HCT116_NR1D1 (gain vs. loss: 75.8% vs. 24.4%) and A similar number of genes were observed in loss and gain of KO HCT116_PER2 (gain vs. loss: 74.2% vs. 25.7%) (Fig. 4b–c). splicing events within the mouse Arntl KO dataset (Fig. 4f). In the Interestingly, higher number of genes with gain in splicing events case of Nr1d1/2 KO model, a slightly higher number of genes were were observed in the metastasis CRC cell line (SW620) and in stage associated with a loss of splicing events than to a gain of splicing IV HL cell line (L1236) (Fig. 4d–e). We further investigated if the events (Fig. 4g). However, an increase in gain of AF-event as seen same genes were alternatively spliced through AF-event (gain) in KO in HCT116_NR1D1 was not observed in the mouse Nr1d1/2 KO. KO KO both HCT116_NR1D1 and HCT116_PER2 . Subsequently, we Taken together, we observed a distinct pattern in the number of filtered for genes with at least one protein-coding transcript, npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. Fig. 4 Local alternative splicing analysis revealed changes in the pattern of loss and gain of all seven splice modes in human and murine KO KO datasets. Bar plots depict the total number of genes (x-axis) with loss and gain of splicing (a–c) in the HCT116_ARNTL , HCT116_NR1D1 , KO and HCT116_PER2 vs. HCT116_WT cells, (d) in the SW620 vs. SW480 cells, (e) in the L1236 vs. HDMYZ cells, (f)in Arntl KO from distal bronchiolar epithelium (left) and from primary tracheal epithelial cells (right) vs. their WT, and (g)in Nr1d1/2 KO vs. WT. In each barplot, the number of genes containing transcripts with circadian expression and protein-coding biotype (dark blue), protein-coding biotype (blue), and different biotypes (light blue) is indicated. KO genes being alternatively spliced in HCT116 vs. WT pointing to the SE-event for all cancer cell lines (Supplementary Fig. 9a–e). In clock knockout specific alterations in alternative splicing. murine KO models, a smaller number of differentially spliced Further, we carried out pairwise differential splicing analysis candidates were observed (Supplementary Fig. 9f–g). We then between KO and WT (Supplementary Fig. 9). Each differentially searched for common differentially spliced candidates between spliced candidate showed significantly (p < 0.05) higher change in the human and mouse KO datasets. Common differentially spliced transcript abundance in the KO than in the corresponding WT candidates were observed only within SE-event. Sorting Nexin 3 KO KO condition. Both HCT116_NR1D1 and HCT116_PER2 showed (SNX3), which is involved in intracellular trafficking was found to KO higher number of differentially spliced genes than in be differentially spliced in both HCT116_ARNTL cells and mouse KO HCT116_ARNTL cells (Supplementary Fig. 9a–c). Besides, the Arntl KO from distal bronchiolar epithelium compared to their highest number of differentially spliced genes were observed in controls. SNX3-201 and SNX3-204 isoforms were present in Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. KO KO HCT116_ARNTL . Among the common differentially spliced specifictoHCT116_ARNTL . Both the transcript pairs of PCGF2 KO KO candidates between HCT116_NR1D1 and mouse Nr1d1/2 KO, consisted of protein-coding biotypes. In HCT116_NR1D1 ,wealso were Prolyl 4-Hydroxylase Subunit Alpha 1 (P4HA1), associated saw KO-specific phase-shifted transcript pairs with retained intron with oxidoreductase activity and Leucine Rich Repeat Containing 8 and nonsense-mediated decay biotypes from ATP Binding Cassette KO VRAC Subunit A (LRRC8A), which is an essential component of the Subfamily A Member 2 (ABCA2), a lipid transporter. HCT116_PER2 volume-regulated anion channel, and plays a role in cell adhesion showed two unique phase-shifted transcript pairs with protein- and cellular trafficking. LRRC8A-202 and LRRC8A-203 transcripts coding and retained intron biotypes from Chromodomain Helicase contributed to the observed differential splicing of LRRC8A and DNA Binding Protein 8 (CHD8), a transcription suppressor gene. P4HA1-202 and P4HA1-203 transcripts contributed to differential A functional enrichment analysis for genes with phase-shifted splicing of P4HA1. While in the mouse Nr1d1/2 KO, Lrrc8a-201 and isoform pairs was carried out for HCT116 datasets (Supplementary Fig. 10). Genes with phase-shifted isoform pairs in HCT116_WT Lrrc8a-202 isoforms contributed to the observed differential were enriched in processes like DNA damage response and G1/S splicing of Lrrc8a and P4ha1-202 transcript contributed to cell cycle transition (Supplementary Fig. 10). Whereas, differential splicing of P4ha1. Taken together, our results showed KO HCT116_ARNTL phase-shifted candidates were uniquely a core-clock KO-specific impact on AS events within the datasets enriched in processes like regulation of apoptotic signalling and analysed. KO EMT. HCT116_NR1D1 phase-shift candidates were uniquely enriched in processes like macrophage proliferation and intracel- Core-clock KO in HCT116 cells resulted in the occurrence of KO lular cholesterol transport. HCT116_PER2 candidates were differentially rhythmic transcript pairs associated with cancer uniquely enriched in processes like regulation of intracellular hallmarks transport and negative regulation of chromatin organization Based on the differences in SFs rhythmicity and local AS events, as (Supplementary Fig. 10). KO seen in HCT116 , we would expect clock dependent changes in the Subsequently, we mapped the genes with phase-shifted rhythmicity of alternatively spliced transcripts. We used data from isoform pairs (phase-diff ≥3 h) with a cancer hallmark list of genes circadian transcripts to examine whether transcript isoforms from obtained from the cancer hallmark genes database (Fig. 5e). A the same gene in each condition show phase-shifted rhythmic KO total of 76 genes in HCT116_WT, 31 genes in HCT116_ARNTL ,38 KO KO expression. For that, we extracted rhythmic and phase-shifted genes in HCT116_NR1D1 , and 30 genes in HCT116_PER2 with transcript pairs (Fig. 5). In short, we compared each rhythmic differentially rhythmic isoform pairs, were related to cancer transcript with all other rhythmic transcripts of the same gene, in a hallmarks (Fig. 5e). Transcript pairs from CSNK2A1 were observed pairwise manner, within the same cell line datasets. Differentially in all HCT116 datasets. However, different phase-shifted transcript KO rhythmic transcript pairs (q <0.05) were obtained for phase pairs were seen in HCT116_WT and HCT116 . In HCT116_WT, we differences larger than 3 h. The scatter plot depicts the distribution observed phase-shift transcript pairs from SMAD2, a cancer of differentially rhythmic transcript pairs according to their phase hallmark gene. Also, transcript pairs from AKT2, involved in all KO difference and amplitude ratio (Fig. 5a). All HCT116 cells showed a cancer hallmark processes except genome instability were phase- decrease in the number of differentially rhythmic transcript pairs vs. shifted in HCT116_WT. However, differentially rhythmic transcript HCT116_WT (Fig. 5b). 1021 transcript isoform pairs (592 unique pairs of both SMAD2 and AKT2 were not found in the KO cells. genes) showed differential rhythmicity in HCT116_WT and out of Phase-shifted transcripts pairs from MET, which controls several KO those, 750 transcript pairs (469 unique genes) showed a phase shift cancer hallmarks were seen in HCT116_WT and HCT116_PER2 . >3 h. The lowest number of differentially rhythmic transcript pairs In particular, MET-202 and MET-206 in HCT116_WT and MET-203 KO KO were seen in HCT116_ARNTL (336 isoform pairs; 255 unique and MET-206 in HCT116_PER2 cells were differentially rhythmic. KO genes) and out of those, 272 transcript isoform pairs (220 unique In HCT116_ARNTL , we observed a phase-shift in differentially genes) showed the pre-defined phase shift (Fig. 5b). 398 transcript rhythmic transcript pairs (from cancer hallmark genes) including isoform pairs (297 unique genes) were differentially rhythmic in HIF1A and FGFR2 that were neither differentially rhythmic in KO KO HCT116_NR1D1 and out of those, 295 pairs (228 genes) had a HCT116_WT nor in other HCT116 (Fig. 5e). Similarly, we KO phase shift larger than 3 h. HCT116_PER2 cells showed 342 observed phase-shifted transcript pairs from cancer hallmarks KO differentially rhythmic isoform pairs (241 unique genes), and 275 genes like HRAS and CD63 unique to HCT116_PER2 . Moreover, pairs (196 unique genes had a phase shift larger than 3 h (Fig. 5b). we observed phase-shifted transcript pairs related to cancer To further explore the existence of different phase shifts within hallmark genes like CASP8 and Mitogen Activated Protein Kinase 3 KO differentially rhythmic isoform pairs, we compared the peak phases (MAPK3) that were unique to HCT116_NR1D1 (Fig. 5e). of isoform pairs (Fig. 5c). A higher number of differentially rhythmic The association of specific cancer hallmarks genes with transcript pairs showed phase-difference ≥9h (Fig. 5c; right panel) as differentially rhythmic isoform pairs due to different clock compared with phase-difference ≥3h and <6h (Fig. 5c; left panel) alterations in HCT116 motivated the subsequent analysis of and phase-difference ≥6h and <9h (Fig. 5c; middle panel) in all alternatively spliced candidates and uniquely spliced candidates KO HCT116 datasets. To further understand the role of phase-shifted in all HCT116 cells. Therefore, we intersected the list of isoform pairs, we explored their biotypes (Fig. 5d). Transcript isoform candidates that were differentially spliced, and found in the pairswith protein-coding biotypeswereobservedasthe largestpair subsets of loss or gain of splicing event (rhythmic at least in in all HCT116 datasets. Among phase-shifted transcript pairs with HCT116_WT), and further shortlisted the candidates related to protein-coding biotypes, we observed 328 pairs in HCT116_WT, 122 cancer hallmarks. For better visualization, we filtered cancer KO KO pairs in HCT116_ARNTL , 140 pairs in HCT116_NR1D1 ,and 123 hallmarks genes for which at least two transcripts were rhythmic KO pairs in HCT116_PER2 . Also, transcript pairs with different biotypes in HCT116_WT for loss of splicing event (Fig. 6a). Whereas, such as nonsense-mediated decay or retained intron were phase- candidates with at least two rhythmic transcripts in KO were shifted in HCT116 datasets (Fig. 5d). For instance, three transcript shortlisted for gain of splicing event. The complete list of pairs from Baculoviral IAP Repeat Containing (BIRC5), which candidates categorized based on cancer hallmarks is provided in promotes cell proliferation and prevents apoptosis, were differen- Supplementary file 3. tially rhythmic only in HCT116_WT cells. Out of the three transcript HNRNPM, a component of spliceosome machinery and involved pairs, two pairs consisted of protein-coding biotypes and one pair in pre-mRNA processing showed either a change in the rhythmic KO consisted of protein-coding and nonsense-mediated decay biotype. properties or a loss of rhythmicity in HCT116 vs. We observed two phase-shifted transcript pairs from Polycomb HCT116_WT cells (Fig. 6b). Moreover, we observed differences in Group Ring Finger 2 (PCGF2), which controls cell proliferation the mean expression level of HNRNPM transcripts among HCT116 npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. Fig. 5 Phase-shifted spliced isoforms in HCT116 datasets are associated with hallmarks of cancer. Differential rhythmicity analysis between rhythmic transcripts within the same cell line was carried out using DODR. a The scatter plot depicts the distribution of differentially rhythmic transcript pairs according to their phase difference and amplitude ratio in HCT116. b Decline in differentially rhythmic isoform pairs observed KO in all HCT116 . c The chord diagram represents the peak phases of transcript pairs across HCT116 under different phase shift cut-offs. d The chord diagram depicts the biotypes of phase shifted transcript pairs. e The circular plot shows the association of genes containing phase shifted transcript pairs with hallmarks of cancer. Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. Fig. 6 Clock disruption in HCT116 resulted in alterations in AS in genes involved in the hallmarks of cancer. Commonly spliced and KO uniquely spliced candidates in HCT116 were mapped to the cancer hallmarks gene list. a Circular plots depict the association between KO spliced candidates and different cancer hallmarks (b) HNRNPM, a spliceosome machinery component lost its circadian expression in HCT116 KO cells. HCT116 cells also showed discrepancies in the mean expression level of HNRNPM transcripts. Expression of uniquely alternatively KO KO KO spliced candidate transcripts in (c) HCT116_PER2 ,(d) HCT116_ARNTL , and (e) HCT116_NR1D1 were plotted. Genomic region plots of MET, FGFR2 and LRRC8A transcripts represent differences in their exon composition (marked in red) compared to canonical forms. Circadian rhythmic transcripts were plotted using harmonic regression fit and arrhythmic transcripts were plotted using Loess fitinR. npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. KO KO datasets. Both HCT116_NR1D1 and HCT116_PER2 resulted in SW620 cell line (29 genes), the metastatic counterpart of SW480 the loss of rhythmicity of transcripts from HNRNPM. On the from the same patient with a less robust circadian rhythm .In KO contrary, HCT116_ARNTL resulted in the gain of one circadian addition, many of the affected alternatively spliced target genes HNRNPM transcript (HNRNPM-215) vs. HCT116_WT. When analys- with circadian AS patterns were involved in various key cancer ing knockout-specific spliced candidates, we observed cancer pathways including VEGFA (involved in angiogenesis) and CD44 KO hallmarks genes like MET as uniquely spliced in HCT116_PER2 (involved in EMT and metastasis), indicating a relevant role for (Fig. 6c). The genomic region plots of MET transcripts (all protein- temporal AS in mediating cancer progression. coding biotypes) represent the difference in coding exon In the current study, we evaluated the effect of circadian composition compared to the canonical form of MET (MET-202). regulation of AS in the colon cancer cell line HCT116 with a robust Moreover, the comparison of the coding exon sequence between circadian clock, as well as in core-clock knockout mutants of the canonical form and other transcripts showed missing/skipping HCT116, and investigated the potential clock-AS interplay in of exon 1- exon 21 (in MET-206), missing of exon 10 (in MET-201), cancer promoting properties. We also investigated our results in and missing/skipping of all exons between 1 to 21 except exon different colon cancer cell lines (SW480 and SW620), HL cells and 10,11 (in MET-203). The AS transcripts of cancer hallmarks genes healthy mouse tissues for a more general understanding of the like FGFR2 showed differential rhythmicity unique to interplay between alterations of the circadian clock and resulting KO KO HCT116_ARNTL (Fig. 6d). Specifically, HCT116_ARNTL showed AS events. changes in the expression of two of the FGFR2 transcripts (FGFR2- Our analysis showed that the alterations in core-clock 202 and FGFR2-211; both protein-coding). The genomic plots for components directly influenced the rhythmicity pattern of splicing FGFR2 transcripts show the loss of coding exon region in FGFR2- factor transcripts, which subsequently impacted AS events. The 202 and FGFR2-211 vs. FGFR2-206 (canonical form). The compar- SFs whose transcripts showed altered rhythmicity pattern in KO KO ison of coding exons sequence between FGFR2-206 and FGFR2- HCT116 included RACK1, SFPQ (altered in HCT116_ARNTL ), KO 202/-211 showed exclusion of exon 9 in FGFR2-202 while exclusion CIRBP, HNRNPD (altered in HCT116_NR1D1 ), and SRSF5 (altered KO of exon 8 in FGFR2-211 (Fig. 6d). LRRC8A was differentially spliced in HCT116_PER2 ). Both RACK1 and SFPQ regulate numerous KO in both HCT116_NR1D1 vs. HCT116_WT and mouse Nr1d1/2 KO cancer-related cellular processes. vs. WT. LRRC8A-201 was not circadian in HCT116_WT, but it was RACK1 is critical for cell proliferation and transcription , while KO circadian expressed in HCT116_NR1D1 (Fig. 6e). CD44 was SFPQ is a multi-functional protein that regulates several processes KO 45 46 differentially spliced only in HCT116_NR1D1 vs. HCT116_WT, such as RNA post-transcriptional activity , splicing regulation , however, the resulting transcripts were rhythmic only in and DNA repair . Aberrant expression of SFPQ is associated with HCT116_WT. We also observed gain in rhythmic transcripts of aetiology of colorectal cancer . A previous study by Pellarin et al KO cancer hallmark genes unique to specific HCT116 vs. reported the reduction in platinum chemotherapy induced HCT116_WT. For instance, rhythmic transcripts from Succinate apoptosis due to SFPQ via alternative splicing of CASP9 in human Dehydrogenase Complex Flavoprotein Subunit A (SDHA), a epithelial ovarian cancer samples . Our analysis showed that complex of mitochondrial respiratory chain were unique to ARNTL deletion in HCT116 cells resulted in a phase shift (>6 h) KO HCT116_ARNTL vs. HCT116_WT. SDHA-206 (retained intron) within SFPQ transcripts (Fig. 3, Supplementary Fig. 6), pointing was rhythmic in HCT116_WT whereas SDHA-202 (retained intron) towards its regulation via a core-clock component. Similarly, phase KO and SDHA-211 (protein-coding) were rhythmic only in shift in RACK1 transcript in HCT116_ARNTL vs. WT suggests that KO KO HCT116_ARNTL . In case of HCT116_NR1D1 , we observed a its circadian variation is linked uniquely with ARNTL. A previous new rhythmic transcript of Diacylglycerol Kinase Zeta (DGKZ), a study using mouse fibroblasts reported the recruitment of RACK1 regulator of intracellular signalling. DGKZ-215 and DGKZ-218 (both receptor in a circadian manner into the nuclear BMAL1 complex, with a retained intron event) were rhythmic in HCT116_WT and the overexpression of RACK1 suppressed CLOCK-BMAL1 whereas DGKZ-201 (protein-coding) was rhythmic only in transcriptional activity . Similarly, BMAL1 was also shown to KO 51 HCT116_NR1D1 . Similarly, we saw that a new rhythmic transcript regulate the expression of Sfpq in rat pituitary cells , thereby from Platelet Derived Growth Factor Subunit (PDGFA), which is corroborating our findings. KO essential for cell survival was unique to HCT116_PER2 . PDGFA- In addition to the observed phase shift in SFs due to the KO of 201 (protein-coding) was rhythmic in HCT116_WT and PDGFA-203 ARNTL, we also observed phase shifts among other transcripts. KO (protein-coding) was only rhythmic in HCT116_PER2 . Altogether, CIRBP, an RNA binding protein regulates several processes like cell our findings showed that each of three core-clock knockouts proliferation and circadian gene expression while HNRNPD resulted in aberrant alternative splicing of different cancer regulates mRNA stability of genes involved in the cell cycle . hallmarks related genes, pointing to a role for the circadian clock The phase shift in rhythmic expression of CIRBP and HNRNPD KO in the regulation of alternative splicing with potential conse- transcripts seen in HCT116_NR1D1 suggests an effect of NR1D1 quences in tumorigenesis. depletion on the spliceosome machinery. Indeed, Cirbp was reported to regulate sleep and circadian clock via Nr1d1 in mice and Hnrnpd was rhythmic in mice cells . Yet, to our knowledge, DISCUSSION no studies were found reporting NR1D1-mediated HNRNPD Pre-mRNA splicing contributes to generate diversity in the expression modulation. Likewise, the phase shift detected in KO products of more than 95% human genes and leads to SRSF5 transcript in HCT116_PER2 pointed towards its regulation alternatively spliced transcripts that encode distinct proteins. This via PER2. Aberrant expression of SRSF5 is associated with different mechanism is often used to maintain cellular homeostasis and to cancer types and their severity , however, no direct role of PER2- 42,43 regulate cell differentiation and development . Previous mediated SRSF5 rhythmicity was reported before. Moreover, we studies have pointed to a regulation of AS via the circadian clock observed loss of rhythmicity in several relevant components of the in cancer, suggesting a temporal pattern of AS, which affected the spliceosome complex including U2AF1 and MBNL2 (required for expression of alternatively spliced target genes in a time- spliceosome binding to the pre-mRNA branch site and a 24–27 dependent manner . Interestingly, the distinct temporal AS modulator of AS, respectively) with deletion of clock components KO pattern and the time-dependent expression of target genes were in all three HCT116 cells. Mutations in U2AF1 contribute to shown to correlate with the circadian phenotype of the cancer progression and have been reported in several different investigated cancer cell lines. For example, SW480 colon cancer cancer types including CRC , while MBNL2 was reported as cells with a robust circadian clock showed twice as many genes tumour suppressor in hepatocarcinogenesis . Accordingly, the with circadian alternative exons (59 genes) compared to the loss of U2AF1 and MBNL2 rhythmic transcripts, as seen in Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. KO HCT116 , point towards their possible contribution to metastasis macrophage proliferation and intracellular cholesterol transport KO formation. Indeed, our results also showed the loss of U2AF1 and (HCT116_NR1D1 ), intracellular transport and chromatin organi- KO MBNL2 rhythmicity in SW620 (metastasis CRC) and L1236 (stage zation (HCT116_PER2 ). These results highlight the individual IV HL). impact of each of the core-clock KOs in cellular functioning. For Several studies reported that alterations in the spliceosome instance, we identified phase shifted transcript pairs from different 12,58–64 machinery lead to aberrant AS patterns . In agreement with cancer hallmark genes such as BIRC5 in HCT116_WT, PCGF2 in KO KO these data and our findings for HCT116 cells, we also found loss of HCT116_ARNTL , ABCA2 in HCT116_NR1D1 , and CHD8 in KO rhythmicity in SFs genes in recently published RNA-seq datasets of HCT116_PER2 . BIRC5, an immune-related gene was found to murine Per1/2 KO (GSE171975 ) and murine Cry1/2 KO be highly expressed in different tumour types and promotes cell 66 76 (GSE135898 ) (Supplementary Fig. 5). These results highlight proliferation . Moreover, Birc5 was reported rhythmic in colon the relevance of core-clock elements in the regulation of genes mucosa cells and its silencing resulted in an increased sensitivity involved in splicing. Indeed, aberrant pre-mRNA splicing and of HCT116 cells to CDK inhibitors . PCGF2, a ring finger protein, alterations in splicing factors are known to act as oncogenic was reported to negatively regulate granulocyte differentiation in 67 78 drivers and contribute to tumour progression . Our results show human HL-60 cells . While, increased expression of ABCA2,a the existence of alterations in rhythmic properties of SFs member of ATP transporters, was reported in different cancer transcripts due to the KO of core-clock elements, which was types . A previous report on CHD8, a negative regulator of Wnt reflected on the resulting AS transcripts. For instance, the pre- signalling, showed that loss of its expression might be an indicator spliceosomal complex element RBM5 regulates AS of CASP9,an of aggressiveness in gastric cancer . However, the interdepen- 4,68 apoptosis-related gene and showed altered rhythmic proper- dence of PCGF2, ABCA2, and CHD8 on specific core-clock elements KO ties in HCT116_ARNTL vs. HCT116_WT cells. RBM5, a tumour was not reported. We also observed phase-shifted isoforms of suppressor gene and splicing factor, improves the production of some gene such as CSNK2A1 in all HCT116 datasets, however, mRNAs by recognizing incorrect 3′splice sites of epidermal growth different transcript pairs were phase-shifted in each KO, pointing factor receptor (EGFR) pre-mRNA, thereby inhibiting the prolifera- to a KO-specific effect. Moreover, we saw the loss of differentially tion of tumour cells . Previous studies in different cancer types transcript pairs of cancer hallmark genes including SMAD2 and KO suggested RBM5 as a potential target to prevent tumorigen- AKT1 in all HCT116 . We also observed differential rhythmicity in 70,71 esis . In our study, we observed a regulation of RBM5 certain cancer hallmark genes, that were not circadian in expression via the circadian clock, which led to a loss of its HCT116_WT. For instance, HIF1A and FGFR2 showed differentially KO rhythmicity due to ARNTL disruption. rhythmic transcript pairs unique to HCT116_ARNTL . HIF1A is the The impact of circadian dysregulation in the splicing machinery main element of the HIF1-pathway, which plays a crucial role in KO was also seen through the loss/gain of AS events in all HCT116 adaptive responses of tumour cells to hypoxia and promotes cells. Among the different splicing events, exon skipping is the tumour progression and metastasis via oncogenic growth factors most common splicing event in human tissues , as also observed such as TGFß (transforming growth factor beta) and EGF in our study. The unique AF-event gain as seen in (epidermal growth factor) . FGFR2 is an EMT-associated gene KO KO HCT116_NR1D1 and HCT116_PER2 indicates that the deletion which encoded for two isoforms, IIIb and IIIc, characteristic to of either PER2 or NR1D1 in HCT116 alters AS by preferring AF- epithelial and to mesenchymal cells, respectively . The discrepant event to the other six modes of AS. expression pattern, as well as rhythmic properties of these two KO By comparing our results to datasets from mouse models, we FGFR2 spliced variants in HCT116_ARNTL vs WT suggests that KO found SNX3 (common between HCT116_ARNTL and murine ARNTL deletion in HCT116 might promote metastasis via EMT. Arntl KO from distal bronchiolar epithelium tissue) and P4HA1, Similarly, we observed differentially rhythmic transcript pairs of KO LRRC8A (common between HCT116_NR1D1 and murine Nr1d1/2 HRAS (proto-oncogene and member of the RAS-pathway involved KO KO) among the common differentially spliced candidates between in cell proliferation) and CD63 unique to HCT116_PER2 . CD63 is a KO HCT116 and murine KO models, however resulting in different member of the tetraspanin family involved in cell differentiation sets of alternatively spliced transcripts. These differences suggest and migration, which was found to be a prognostic marker in 83,84 KO that the deletion of ARNTL or NR1D1 in HCT116 is likely to have a colorectal and esophageal cancer . In HCT116_NR1D1 ,we different biological impact (protein product) compared to the observed unique differentially rhythmic transcript pairs of CASP8 same KOs in murine models. Differential splicing of SNX3 in and MAPK3. CASP8 is an apoptosis-related cysteine peptidase and KO KO HCT116_ARNTL and P4HA1, LRRC8A in HCT116_NR1D1 point an essential part of the death-inducing signalling complex towards splicing alterations, which are KO specific. SNX3, a unique (DISC) . In cancer, it promotes proliferation and angiogenesis mediator of WNT protein secretion, was reported to mediate EMT through the activation of NF-kB in glioblastoma and breast 73 87 and metastasis in CRC cells including HCT116 . P4HA1, a catalytic cancer . MAPK3 encodes for a member of the MAP kinase family enzyme, was reported to regulate cell proliferation in CRC cells via (aka extracellular signal-regulated kinases (ERK)) and regulates HIF1A and WNT signaling . While, LRRC8A, a main regulatory various cellular processes such as proliferation, differentiation, and subunit of VRAC (volume-regulated anion channel), was found to cell cycle progression in response to a variety of extracellular be upregulated in colon cancer patients and might contribute to signals. These results point towards a cross-talk between aberrant 74 KO metastasis . However, the direct correlation between these genes AS events and cancer hallmarks related to each of the HCT116 . and core-clock components remains unexplored. We also observed cancer hallmark genes like MET, whose Following the temporal regulation of different AS events in different transcript pairs were phase shifted in HCT116_WT and KO HCT116 datasets, our analysis revealed several transcript pairs that HCT116_PER2 . The MET oncogene encodes for a receptor showed differential rhythmicity. Moreover, we also showed the tyrosine kinase with pleiotropic functions in initiating and occurrence of new KO-specific alterations in transcripts rhythmi- sustaining neoplastic transformation, as well as in cancer cell city vs. WT. In HCT116_WT, phase-shifted isoforms were found to survival and tumour dissemination . MET is known to undergo AS be enriched in DNA damage response and G1/S cell cycle and its AS isoform (lacking exon 14) is known to inhibit HGF- transition. This suggests a temporal regulation of DNA repair and induced tyrosine phosphorylation of Met, as well as cell cell cycle pathways that might be regulated by circadian AS proliferation and migration in skeletal muscle myoblasts . events. Indeed, a previous study reported splicing as an emerging We further compared our list of AS spliced candidates to a list of 75 41 pathway contributing to DNA damage response . In the KO cells, known cancer hallmark genes . Of these, ACTB was differentially KO we observed different enriched processes among phase-shifted spliced in all HCT116 vs WT, however rhythmicity of ACTB KO isoforms such as apoptotic signalling and EMT (HCT116_ARNTL ), transcripts were lost in all three KOs. This suggests that the loss in npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. circadian profile of ACTB transcripts may contribute towards demonstrated for numerous anticancer agents in experimental metastasis due to clock alterations. A previous study reported that models, as well as in clinical settings including clinical studies in 102–104 the abnormal expression and polymerization of ACTB contributes patients with metastatic colorectal cancer . Aberrant alter- to invasiveness and metastasis of different cancer types . Besides native splicing events and circadian clock disruption are reported different cancer hallmarks related to AS candidates, we also in several different cancer types, adding an additional complexity 67,105 observed discrepant expression of HNRNPM SF transcripts in all to the genetic landscape of cancer . The results of our study KO three HCT116 vs. WT. Among different KO-specific AS candi- highlight the importance of timed AS, especially in genes KO KO dates, FGFR2 (in HCT116_ARNTL ), MET (in HCT116_PER2 ), CD44 regulating cancer hallmarks that are shown to be suitable drug KO (in HCT116_NR1D1 ) lost/gained rhythmicity in the KOs. In targets. Our data point to the regulation of AS patterns in cancer KO HCT116_ARNTL , FGFR2 IIIb isoform (exon 9 exclusion and exon via the circadian clock. It would be important to further explore 8 inclusion; FGFR2-202) was circadian while in HCT116_WT FGFR2 these findings and to consider drug timing, at least for drugs IIIc (exon 8 exclusion and exon 9 inclusion; FGFR2-211) was targeting such genes, in future clinical studies. circadian. These discrepancies were better seen when we analysed Altogether, the results of our study suggest an interplay the average expression of both isoforms in WT and between circadian clock elements and AS in cancer, with distinct KO HCT116_ARNTL cells. The lower average expression of FGFR2 and unique roles for the core-clock genes ARNTL, PER2 and NR1D1 KO IIIb in HCT116_ARNTL while higher expression of FGFR2 IIIc point in regulating SF rhythmicity and AS events in cancer hallmark to possible alterations in EMT signalling. Matsuda Y et al. reported genes, with relevance in cancer onset and progression. increased FGFR2 IIIc expression among colorectal carcinomas samples and a human anti-FGFR2 IIIc monoclonal antibody was METHODS reported to inhibit growth in colorectal carcinoma cells . These findings suggest that FGFR2 IIIc could be a promising therapeutic Cell culture target for colorectal carcinoma. Human colorectal carcinoma cell line HCT116 (ATCC® CCL-247™, Gaithers- CD44 encodes for a family of cell adhesion molecules involved burg, MD, USA) was cultured in Dulbecco’s Modified Eagle Medium (Gibco, in homotypic and heterotypic interactions with extracellular Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10% Fetal Bovine Serum (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) and 1% matrix components, as well as EMT . It promotes proliferation Penicillin–Streptomycin (Gibco, Thermo Fisher Scientific, Waltham, MA, and invasiveness of cells via recruiting ERM proteins (Ezrin, Radixin USA) in an incubator with 5% CO at 37 °C. LUNA™ Automated Cell Counter und Moesin) under certain conditions. However, CD44 can also act (Logos Biosystems, Anyang, South Korea) was used for cell counting and as a tumour suppressor, for example when cells reach confluent morphology analysis. Cell lines were tested for mycoplasma by using the growth conditions, and thus inhibit cell growth. The prognostic Mycoplasma check service of Eurofins Genomics (Eurofins Genomics, value of the CD44 variant 6 (CD44v6) in CRC was debated for years Ebersberg, Germany). 93–95 due to contradictory results . The function of CD44 in a cell is determined by the CD44 isoform pattern expressed . Several CRISPR-Cas9 Knockout generation studies showed the relevance of CD44v6 as an independent A CRISPR-Cas9 mediated approach was applied to generate the core-clock negative prognostic factor and a promising therapeutic target in 96,97 knockout cell lines in HCT116. Briefly, HCT116_WT cells were seeded in CRC . In our study, we showed that NR1D1 disruption in 6-well plates at a density of 4 × 10 cells/well and transfected with CRISPR- HCT116 cells results in aberrant alternative splicing of CD44, which Cas9 plasmids containing GFP marker and guided RNAs targeting multiple points to a clock specific regulation of CD44 splicing, and could exons of ARNTL, PER2 or NR1D1 genes, respectively. Cell transfection was potentially play a role in CD44-targeted therapy regimens. performed using FuGENE HD Transfection Reagent (Promega Corporation, Interestingly, several CD44-targeted drugs have been approved Fitchburg, WI, USA) according to the manufacturer’s instructions. 48 h post- for clinical trials, which highlights the importance of timing transfection, CRISPR/Cas9 GFP-positive cells were single-cell sorted using treatment in cancer therapy targeting CD44 (reviewed in ). Our an S3e cell sorter (Bio-Rad laboratories, Hercules, CA, USA) into 96-well results suggest that the perturbations in core-clock elements plates. Colonies were expanded for subsequent testing and successful knockout colonies were used for the time-course RNA-seq experiment. result in different AS outcomes for this gene. For each knockout condition, several single clones were investigated on Furthermore, certain genes showed a gain of new rhythmic RNA gene expression level to characterize and confirm the knockout. All transcripts in the KO vs. WT. These included SDHA KO cell lines displayed significantly reduced target gene expression KO KO (HCT116_ARNTL vs. WT), DGKZ (HCT116_NR1D1 vs. WT), and compared to WT (Supplementary Fig. 1d). The off-target activity was KO PDGFA (HCT116_PER2 vs. WT). SDHA is responsible for transfer- investigated using Off-Spotter and the Welcome Trust Sanger Institute ring electrons from succinate to ubiquinone (coenzyme Q), and 107 Genome Editing database (WGE) to search for the most likely potential acts as a tumour suppressor and inhibitor of angiogenesis in off-target sites using gRNA sequences. Off-target sites with up to three paraganglioma . DGKZ attenuates protein kinase C activity by mismatches within protein-coding regions were Sanger-sequenced and regulating diacylglycerol levels in intracellular signalling cascade compared to WT. All investigated potential off-target sites in KO cells showed 100% sequence similarity to the WT, indicating absent off-target and signal transduction. It has proven to be associated with modifications. For regions where it was not possible to design specific various signalling pathways, including ERK and MYC and acts as a primers amplifying the off-target site, we compared band sizes on gel potential oncogene in osteosarcoma . PDGFA is a member of the electrophoresis. PDGF as well as VEGF signalling. Paracrine PDGF signalling is commonly observed in epithelial cancers, where it triggers stromal Sample preparation and RNA extraction recruitment and may be involved in EMT affecting tumour growth, angiogenesis, invasion, and metastasis. Overexpression of PDGF HCT116 cells were seeded in triplicates in 12-well plates with a density of 2×10 cells per well. On the next days, cells were synchronized by signalling was shown to drive tumour cell growth and to promote changing the media. Sampling was started 9 h after synchronization and tumorigenesis in colorectal cancer, breast cancer, lung cancer and samples were taken every 3 h for a time-series of 45 h. To prepare the cells sarcomas (reviewed in ref. ). The gained rhythmicity of a coding for RNA extraction, media was discarded and cells were washed with transcript of SDHA due to ARNTL deletion, DGKZ due to NR1D1 phosphate buffer saline (PBS) and lysed using RLT Plus buffer (Qiagen, deletion, and PDGFA due to PER2 deletion in HCT116 may lead to Hilden, Germany) directly on the plate. A total of 64 samples were obtained aberrant protein functions resulting from clock-KO-specific spli- from the HCT116 datasets (WT and three KOs). cing. However, the biological role of these transcripts remains to Total RNA was isolated using the RNeasy Plus Mini kit (Qiagen, Hilden, be elucidated. Germany) according to the manufacturer’s guidelines. Genomic DNA was The relevance of timing treatment for increasing tolerability and digested from the cells using gDNA eliminator columns provided with the efficacy while minimizing toxic side-effects has been largely kit (Qiagen, Hilden, Germany). RNA was eluted in 30 µL RNase-free water. Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. RNA concentration was measured using a Nanodrop 1000 (Thermo Fisher Curation of splicing factor list Scientific, Waltham, MA, USA). RNA was then stored at −80 °C until In total 534 human spliceosome and splicing-related genes were compiled 24,35,36 further use. from the literature , as well as from online databases (Spliceosome 37 38 database and SpliceAid 2 ). A list of human splicing factors was mapped using Ensembl Biomart to obtain their orthologous sets for mouse. A Data generation total of 521 SFs were obtained for mouse. The complete list of splicing High-quality RNA was used to generate mRNA libraries and prepared using factors is provided in Supplementary files 1–2. the TruSeq Stranded mRNA Sample Preparation Kit (Illumina, San Diego, CA, USA) according to the sample protocol guidelines and sequenced on an Illumina NextSeq 500 platform to an average depth of 100 M 75-bp RNA-sequencing data processing paired-end reads at the European Molecular Biology Laboratory (EMBL) All RNA-seq datasets were processed and analysed according to the GeneCore Facility (Heidelberg, Germany). following protocol unless otherwise stated. Quality assessment of raw reads was carried out using fastqcr (version 0.1.2) R package which easily parse, aggregate, and analyse FastQC for large number of samples. Acquisition of additional RNA-seq datasets Based on FastQC reports, over-represented sequences and residual Using a strict criterion of “paired-end RNA-seq”, “less than or equal to 4 h adapter sequences were trimmed from raw reads using Trimmomatic interval”, “mammalian tissue”, “circadian rhythm”, we obtained a total of (version 0.39 ) using default settings. After trimming, only paired-end 134 samples from six different datasets. Published paired-end RNA-seq for reads were used for further downstream analysis. Paired-end reads from Per1/2 was not available, therefore we included for this dataset recently human datasets were aligned to human genome (Homo sapiens GRCh38, published single-end Per1/2 KO (24 samples). Out of seven different Ensembl release 92) and from mouse datasets were aligned to mouse datasets, two datasets were retrieved from human model system and five genome (Mus musculus GRCm38, Ensembl release 102) using STAR aligner datasets were retrieved from the mouse model system. (version 2.6.0 ). Based on STAR transcriptome alignment, transcript-level Based on human model system, RNA-seq datasets were collected abundances were quantified in transcripts per million (TPM) using Salmon derived from: (1) human CRC cell lines (SW480: primary tumour, SW620: 115 116 (version 0.10.2 ). Afterwards, tximport R package was used to import metastatic tumour; Accession number: E-MTAB-7779) from the same transcript level abundance, estimate counts and transcript lengths. patient, and (2) Hodgkin’s lymphoma (HL) cell lines of cancer progression Tximport summarizes the quantification results into matrices for down- stages (HD-MY-Z: stage IIIB, L-1236: stage IV; Accession number: GSE16206) stream analyses. Tximport can be used to obtain both transcript level from different patients. CRC cells (SW480, SW620) and HL cells (HDMYZ, (txOut= TRUE) and gene-level (txOut= FALSE) summarization. Here, we L1236), which represent different progression grades of CRC and HL, used transcript-level counts for detailed alternative splicing analysis. respectively, display various clock phenotypes (Supplementary Fig. 2e). Nevertheless, gene-level summarization was also used for its comparison These datasets were chosen to find a possible correlation between clock with transcript-level analysis. Trimmed mean of M-values (TMM) method of disruption and cancer progression stage. averaging was used as the normalization factor to scale up the raw library Based on mouse model system, we retrieved RNA-seq datasets derived size using edge R package (version 4.0 ). Counts per million (CPM) from (1) mouse embryonic stem cells in WT condition and the Nr1d1/2 KO function was used as a descriptive measure of transcript expression. All (Accession number: GSE125696) generated using CRISPR/Cas9, (2) laser transcript feature with at least 0.5 CPM on average over all time points micro-dissected SCN tissue of male C3H/HeH wild type mice (age: (specific to datasets) were retained and renormalized using selected 10–12 weeks; Accession number: GSE72095), (3) distal bronchiolar features. The complete pipeline for the analysis is depicted in Fig. 1. epithelium tissue isolated using laser capture microdissection from WT Raw unlogged gene-level count data from Per1/2 KO (GSE171975; single- mice and Arntl KO mice bearing the targeted deletion in mouse club cells end RNA-seq) and Cry1/2 KO (GSE135898; paired-end RNA-seq) were (age: 10–20 weeks; Accession number: E-MTAB-6384), and primary tracheal downloaded from NCBI-GEO and used for rhythmicity analysis. epithelial cells isolated from WT mice and Arntl global KO (age: 10–20 weeks; Accession number: E-MTAB-6384). In addition, we retrieved raw gene-level count data from two murine core-clock KO RNA-seq Rhythmicity analysis datasets from NCBI-GEO, liver tissue from WT and Per1/2 KO mice (age: Unlogged CPM values were used to detect rhythmic signals from time- 3 months old, Accession number: GSE171975; single-end RNA-seq), and series datasets. 24 h rhythmicity was evaluated through a non-parametric liver tissue from WT and Cry1/2 KO mice under normal feeding (age: method using RAIN R package (version 1.24.0 ). Rhythmic gene/ 9–14 weeks old, Accession number: GSE135898). All datasets listed in this transcript-sets were obtained using a cut-off of q < 0.05. The acrophase section were downloaded from NCBI-GEO or ArrayExpress. and relative amplitude were estimated for rhythmic gene-/transcript-sets 119 120 using Cosinor within Discorhythm R package (version 1.6.0 ). The Synchronization and sampling protocols of used RNA-seq rhythmic gene/transcript-sets were further filtered using an additional cut- datasets off of relative amplitude (rAMP) ≥ 0.1. E-MTAB-6384 (mouse Arntl-KO lung samples pulmonary airway epithelial cells): For circadian sampling, mice were maintained in constant darkness Alternative splicing analysis using SUPPA and samples collected 1 cycle after transfer to darkness (1 day later) at To calculate the local alternative splicing events based on the expression of circadian time (CT), which by convention anchors expected time of lights transcripts in each dataset, we used SUPPA2 . SUPPA2 is helpful in off and activity onset to CT12. Samples were taken every 4 h for 48 h in studying splicing at the local alternative splicing level or at the transcript constant dark conditions . isoform level. We analyzed seven alternative splicing events types; (1) GSE72095 (mouse SCN samples): After 7 days of acclimatisation, mice Alternative 3′ splice site (A3-event) where 3′ site acts as an acceptor, (2) were singly housed for 7 days prior to tissue harvesting. At each sampling Alternative 5′ splice site (A5-event) where 5′ site acts as a donor, (3) time-point, pooled dissected tissue from five adult male mice were used. Alternative first exon (AF-event) where the first exon is retained after Samples were taken for a total of six time-points over a 12:12 LD cycle at splicing, (4) Alternative last exon (AL-event) where the last exon is retained 4 h intervals (ZT2, 6, 10, 14, 18 and 22), where ZT0 denotes the time of after splicing, (5) Mutually exclusive event (MX-event) where one of two lights on . exons is retained, (6) Retention intron event (RI-event) where the intron is GSE125696 (Nr1d1/2 KO mouse ESCs): On differentiation day 28, cells confined within mRNA, and (7) Skipping exon event (SE-event) where an were treated with 100 nM dexamethasone and frozen at the indicated time exon may be spliced out or retained. points. Time course RNA-Seq was performed using RNA samples at 4 h For each gene in a given tissue or condition, the average percent intervals over 2 days . spliced-in (PSI) value was calculated. Alternative splicing of a gene in a Synchronization of HCT116 RNA-Seq samples: HCT116 cells were particular dataset is considered only if it fulfils the criteria of PSI < 0.9 and > synchronized via medium exchange (which serves as a strong entrainment 0.1. Using the PSI cut-off, we detected specific local events which only signal aka Zeitgeber in these cells). Previous reports on the efficacy of happened either in the WT or in the KO datasets. To classify these events, synchronization agents (e.g. dexamethasone, forskolin, serum shock, we grouped the local events as gain or loss in the KO with respect to its medium change) have shown, that medium exchange is as effective as other mentioned agents in synchronizing the circadian clock, as shown for control. If a KO contained a gene which is spliced by a particular local different CRC cell lines via bioluminescence measurements of ARNTL- event and the WT did not show the gene in that particular event, the event promoter activity . would match to gain in the KO. Similarly, if the control group contained a npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. gene which is alternatively spliced by a particular local event and the Received: 8 November 2021; Accepted: 4 April 2022; corresponding KO did not show the gene in that particular event, the event would correspond to loss in the KO. Besides, we also used SUPPA2 to carry out the differential splicing analysis where a pairwise comparison between the KO and WT was made. The differentially spliced candidates were obtained using p-val < 0.05. REFERENCES 1. Lee, Y. & Rio, D. C. Mechanisms and regulation of alternative Pre-mRNA splicing. Differential rhythmicity analysis Annu. Rev. Biochem. 84, 291–323 (2015). The detection of differentially rhythmic transcripts isoform within each 2. Chen, M. & Manley, J. L. Mechanisms of alternative splicing regulation: insights dataset was carried out using DODR R package (version 0.99.2 ). All 24 h from molecular and genomics approaches. Nat. Rev. Mol. Cell Biol. 10, 741–754 rhythmic transcript sets were used for this analysis. 24 h differentially (2009). 3. Wang, E. T. et al. Alternative isoform regulation in human tissue transcriptomes. rhythmic transcript isoform pairs per single tissue/condition type were Nature 456, 470–476 (2008). obtained using FDR corrected p-value cut-off meta. p < 0.05. To further 4. Paronetto, M. P., Passacantilli, I. & Sette, C. Alternative splicing and cell survival: examine the transcript isoform with more prominent changes in HCT116, from tissue homeostasis to disease. Cell Death Differ. 23, 1919–1929 (2016). phase difference cut-off ≥3 h was applied. 5. Ravi, S., Schilder, R. J. & Kimball, S. R. Role of precursor mRNA splicing in nutrient- induced alterations in gene expression and metabolism. J. Nutr. 145, 841–846 Acquisition of list of cancer hallmark genes (2015). 6. Du, J. X. et al. Splicing factors: Insights into their regulatory network in alter- To understand the direct influence of clock alterations on cancer hallmarks, native splicing in cancer. Cancer Lett. 501,83–104 (2021). we retrieved the complete gene list categorized according to processes 7. Scotti, M. M. & Swanson, M. S. RNA mis-splicing in disease. Nat. Rev. Genet. 17, from Cancer Hallmark Genes database . 19–32 (2016). The CHG database was established by systematically integrating 8. Venables, J. P. Aberrant and alternative splicing in cancer. Cancer Res. 64, manually curated hallmarks genes from the confirmed literature and KEGG 123 7647–7654 (2004). pathways . In parallel, based on pre-defined keywords related to 9. Wang, E. & Aifantis, I. RNA splicing and cancer. Trends Cancer 6, 631–644 (2020). hallmarks, a text mining software Lucene (https://lucene.apache.org/) 10. Ben-Dov, C., Hartmann, B., Lundgren, J. & Valcárcel, J. Genome-wide analysis of was used to identify hallmark specific pathways in the literature and in the alternative pre-mRNA splicing. J. Biol. Chem. 283, 1229–1233 (2008). KEGG database. To establish the CHG database Zhang et al. used: (1) 301 11. Oltean, S. & Bates, D. O. Hallmarks of alternative splicing in cancer. Oncogene 33, KEGG pathways for Lucene search and the extraction of pathways genes, 5311–5318 (2014). (2) Gene variant data (7075 mutation samples, 6177 methylation samples, 12. Bonnal, S. C., López-Oreja, I. & Valcárcel, J. Roles and mechanisms of alternative 9445 copy number variation samples) from TCGA (The Cancer Genome splicing in cancer—implications for care. Nat. Rev. Clin. Oncol. 17, 457–474 Atlas https://www.cancer.gov/tcga) to calculate the frequency of gene (2020). variation, (3) Human protein-protein interaction data from HPRD (Human 13. Climente-González, H., Porta-Pardo, E., Godzik, A. & Eyras, E. The functional 124 125 126 127 Protein Reference Database ), STRING , BioGRID , HTRIdb to impact of alternative splicing in cancer. Cell Rep. 20, 2215–2226 (2017). integrate a gene interaction network. CHG provides the corresponding 14. Harper, S. J. & Bates, D. O. VEGF-A splicing: the key to anti-angiogenic ther- gene sets for ten cancer hallmarks: (1) Activating invasion and metastasis apeutics? Nat. Rev. Cancer 8, 880–887 (2008). 15. Herrlich, P. et al. CD44 acts both as a growth- and invasiveness-promoting (1,149 genes), (2) Enabling replicative immortality (308 genes), (3) Evading molecule and as a tumor-suppressing cofactor. Ann. N. Y. Acad. Sci. 910, growth suppressors (709 genes), (4) Evading immune destruction (609 106–120 (2000). genes), (5) Genome instability and mutation (231 genes), (6) Inducing 16. Cohen-Eliav, M. et al. The splicing factor SRSF6 is amplified and is an onco- angiogenesis (507 genes), (7) Reprogramming energy metabolism (461 protein in lung and colon cancers. J. Pathol. 229, 630–639 (2013). genes), (8) Resisting cell death (1192 genes), (9) Sustaining proliferative 17. Park, S. et al. Differential functions of splicing factors in mammary transforma- signaling (1314 genes), and (10) Tumor promoting inflammation (648 tion and breast cancer metastasis. Cell Rep. 29, 2672–2688.e2677 (2019). genes). 18. Shakhmantsir, I. & Sehgal, A. Splicing the clock to maintain and entrain circadian rhythms. J. Biol. Rhythms 34, 584–595 (2019). 19. Patke, A. et al. Mutation of the human circadian clock gene CRY1 in familial Functional enrichment analysis delayed sleep phase disorder. Cell 169, 203–215.e213 (2017). Gene ontology analysis for selected gene sets was carried out using an R 20. Hastings, M. H., Reddy, A. B. & Maywood, E. S. A clockwork web: circadian timing based package ClusterProfiler (version 4.0 ). Enriched biological pro- in brain and periphery, in health and disease. Nat. Rev. Neurosci. 4, 649–661 cesses were obtained for each of the gene sets. Dot plots were used to (2003). visualize the top 10 enriched biological processes in the case of 21. Andreani, T. S., Itoh, T. Q., Yildirim, E., Hwangbo, D. S. & Allada, R. Genetics of differentially rhythmic isoform pairs. The enriched biological processes Circadian rhythms. Sleep. Med. Clin. 10, 413–421 (2015). for the candidate cancer hallmarks markers were manually curated after 22. Innominato, P. F., Lévi, F. A. & Bjarnason, G. A. Chronotherapy and the molecular removing the redundant biological processes. The visualization of cancer clock: clinical implications in oncology. Adv. Drug Deliv. Rev. 62, 979–1001 (2010). hallmarks connected with selected candidate markers was carried out 23. Shafi, A. A. & Knudsen, K. E. Cancer and the Circadian clock. Cancer Res. 79, using an R based package GO plot (version 1.0.2 ). 3806–3814 (2019). 24. El-Athman, R., Fuhr, L. & Relógio, A. A systems-level analysis reveals circadian regulation of splicing in colorectal cancer. EBioMedicine 33,68–81 (2018). DATA AVAILABILITY 25. El-Athman, R., Knezevic, D., Fuhr, L. & Relógio, A. A computational analysis of alternative splicing across mammalian tissues reveals circadian and ultradian The datasets analysed during the current study are available as follows: recently rhythms in splicing events. Int. J. Mol. Sci. 20, https://doi.org/10.3390/ generated RNA-sequencing data for HCT116 WT cells and KOs in this study have ijms20163977 (2019). been deposited in the ArrayExpress repository (Accession number: E-MTAB-9701). 26. Genov, N., Basti, A., Abreu, M., Astaburuaga, R. & Relógio, A. A bioinformatic Other datasets used in this manuscript are available as follows: SW480 and SW620 analysis identifies circadian expression of splicing factors and time-dependent cell lines, ArrayExpress, E-MTAB-7779; HD-MY-Z and L-1236 cell lines, Gene Expression alternative splicing events in the HD-MY-Z cell line. Sci. Rep. 9, 11062 (2019). Omnibus (GEO) GSE16206; mouse Arntl-KO lung samples Pulmonary airway epithelial 27. Genov, N., Basti, A., Abreu, M. & Relógio, A. Temporal splicing switches in ele- cells, ArrayExpress, E-MTAB-6384; mouse SCN samples, GEO, GSE72095; Nr1d1/2 KO ments of the tnf-pathway identified by computational analysis of transcriptome mouse ESCs, GEO, GSE125696; murine Per1/2 KO liver samples, GEO, GSE171975; data for human cell lines. Int. J. Mol. Sci. 20, https://doi.org/10.3390/ murine Cry1/2 KO liver samples, GEO, GSE135898. ijms20051182 (2019). 28. Pluquet, O. et al. Posttranscriptional regulation of PER1 underlies the oncogenic function of IREα. Cancer Res. 73, 4732–4743 (2013). CODE AVAILABILITY 29. Koyanagi, S. et al. A molecular mechanism regulating circadian expression of vascular endothelial growth factor in tumor cells. Cancer Res. 63, 7277–7283 All programs, packages and functions used in this manuscript are freely available and (2003). are referred to in the Method section. Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17 D. Malhan et al. 30. Preußner, M. et al. Rhythmic U2af26 alternative splicing controls 61. Martín, E., Vivori, C., Rogalska, M., Herrero-Vicente, J. & Valcárcel, J. Alternative PERIOD1 stability and the circadian clock in mice. Mol. Cell 54, 651–662 (2014). splicing regulation of cell-cycle genes by SPF45/SR140/CHERP complex controls 31. Lee, J.,Park, E.,Kim, G.H., Kwon,I.&Kim, K. Asplice variant of humanBmal1 acts as a cell proliferation. RNA 27, 1557–1576 (2021). negative regulator of the molecular circadian clock. Exp. Mol. Med. 50,1–10 (2018). 62. Venables, J. P. Unbalanced alternative splicing and its significance in cancer. 32. Duffield, G. E. DNA microarray analyses of circadian timing: the genomic basis of Bioessays 28, 378–386 (2006). biological time. J. Neuroendocrinol. 15, 991–1002 (2003). 63. Zhang, J. & Manley, J. L. Misregulation of pre-mRNA alternative splicing in 33. Wu, Y. et al. Systematic studies of the circadian clock genes impact on tem- cancer. Cancer Discov. 3, 1228–1237 (2013). perature compensation and cell proliferation using CRISPR tools. Biology (Basel) 64. Gebauer, F., Schwarzl, T., Valcárcel, J. & Hentze, M. W. RNA-binding proteins in 10, https://doi.org/10.3390/biology10111204 (2021). human genetic disease. Nat. Rev. Genet. 22, 185–198 (2021). 34. Bunger, M. K. et al. Mop3 is an essential component of the master circadian 65. Aviram, R., Dandavate, V., Manella, G., Golik, M. & Asher, G. Ultradian rhythms of pacemaker in mammals. Cell 103, 1009–1017 (2000). AKT phosphorylation and gene expression emerge in the absence of the cir- 35. Relógio, A. et al. Alternative splicing microarrays reveal functional expression of cadian clock components Per1 and Per2. PLoS Biol. 19, e3001492 (2021). neuron-specific regulators in Hodgkin lymphoma cells. J. Biol. Chem. 280, 66. Weger, B. D. et al. Systematic analysis of differential rhythmic liver gene 4779–4784 (2005). expression mediated by the circadian clock and feeding rhythms. Proc. Natl 36. Seiler, M. et al. Somatic mutational landscape of splicing factor genes and their Acad. Sci. USA 118, https://doi.org/10.1073/pnas.2015803118 (2021). functional consequences across 33 cancer types. Cell Rep. 23,282–296.e284 (2018). 67. Zhang, Y., Qian, J., Gu, C. & Yang, Y. Alternative splicing and cancer: a systematic 37. Cvitkovic, I. & Jurica, M. S. Spliceosome database: a tool for tracking components review. Signal Transduct. Target Ther. 6, 78 (2021). of the spliceosome. Nucleic acids Res. 41, D132–141 (2013). 68. Bechara, E. G., Sebestyen, E., Bernardis, I., Eyras, E. & Valcarcel, J. RBM5, 6, and 10 38. Piva, F., Giulietti, M., Burini, A. B. & Principato, G. SpliceAid 2: a database of differentially regulate NUMB alternative splicing to control cancer cell pro- human splicing factors expression data and RNA target motifs. Hum. Mutat. 33, liferation. Mol. Cell 52, 720–733 (2013). 81–85 (2012). 69. Su, Z. et al. Lentiviral vector-mediated RBM5 overexpression downregulates 39. Kinsella, R. J. et al. Ensembl BioMarts: a hub for data retrieval across taxonomic EGFR expression in human non-small cell lung cancer cells. World J. Surg. Oncol. space. Database (Oxf.) 2011, bar030 (2011). 12, 367 (2014). 40. Wojtuszkiewicz, A. et al. Pre-mRNA splicing in cancer: the relevance in onco- 70. Jiang, Y. et al. RBM5 inhibits tumorigenesis of gliomas through inhibition of Wnt/ genesis, treatment and drug resistance. Expert Opin. Drug Metab. Toxicol. 11, β-catenin signaling and induction of apoptosis. World J. Surg. Oncol. 15, 9 (2017). 673–689 (2015). 71. Prabhu, V. V. & Devaraj, N. Regulating RNA binding motif 5 gene expression—a 41. Zhang, D. et al. CHG: a systematically integrated database of cancer hallmark novel therapeutic target for lung cancer. J. Environ. Pathol. Toxicol. Oncol. 36, genes. Front. Genet. 11, https://doi.org/10.3389/fgene.2020.00029 (2020). 99–105 (2017). 42. Gallego-Paez, L. M. et al. Alternative splicing: the pledge, the turn, and the 72. Florea, L., Song, L. & Salzberg, S. L. Thousands of exon skipping events differentiate prestige: the key role of alternative splicing in human biological systems. Hum. among splicing patterns in sixteen human tissues. F1000Res 2, 188 (2013). Genet. 136, 1015–1042 (2017). 73. Pan, B. et al. SNX3 suppresses the migration and invasion of colorectal cancer 43. Baralle, F. E. & Giudice, J. Alternative splicing as a regulator of development and cells by reversing epithelial-to-mesenchymal transition via the β-catenin path- tissue identity. Nat. Rev. Mol. Cell Biol. 18, 437–451 (2017). way. Oncol. Lett. 18, 5332–5340 (2019). 44. McCahill, A., Warwicker, J., Bolger, G. B., Houslay, M. D. & Yarwood, S. J. The 74. Zhang, H. et al. High expression of leucine‑rich repeat‑containing 8A is indica- RACK1 scaffold protein: a dynamic cog in cell response mechanisms. Mol. tive of a worse outcome of colon cancer patients by enhancing cancer cell Pharm. 62, 1261–1273 (2002). growth and metastasis. Oncol. Rep. 40, 1275–1286 (2018). 45. Bottini, S. et al. Post-transcriptional gene silencing mediated by microRNAs is 75. Lenzken, S. C., Loffreda, A. & Barabino, S. M. L. RNA splicing: a new player in the controlled by nucleoplasmic Sfpq. Nat. Commun. 8, 1189 (2017). DNA damage response. Int. J. Cell Biol. 2013, 153634 (2013). 46. Mora Gallardo, C. et al. Dido3-dependent SFPQ recruitment maintains efficiency 76. Xu, L., Yu, W., Xiao, H. & Lin, K. BIRC5 is a prognostic biomarker associated with in mammalian alternative splicing. Nucleic Acids Res. 47, 5381–5394 (2019). tumor immune cell infiltration. Sci. Rep. 11, 390 (2021). 47. Jaafar, L., Li, Z., Li, S. & Dynan, W. S. SFPQ•NONO and XLF function separately and 77. Siffroi-Fernandez, S. et al. Functional genomics identify Birc5/survivin as a together to promote DNA double-strand break repair via canonical non- candidate gene involved in the chronotoxicity of cyclin-dependent kinase homologous end joining. Nucleic Acids Res. 45, 1848–1859 (2017). inhibitors. Cell Cycle 13, 984–991 (2014). 48. Klotz-Noack, K. et al. SFPQ depletion is synthetically lethal with BRAF(V600E) in 78. Jo, S. et al. Inhibition of PCGF2 enhances granulocytic differentiation of acute colorectal cancer cells. Cell Rep. 32, 108184 (2020). promyelocytic leukemia cell line HL-60 via induction of HOXA7. Biochem. Bio- 49. Pellarin, I. et al. Splicing factor proline- and glutamine-rich (SFPQ) protein reg- phys. Res. Commun. 416,86–91 (2011). ulates platinum response in ovarian cancer-modulating SRSF2 activity. Onco- 79. Davis, W. & Tew, K. D. ATP-binding cassette transporter-2 (ABCA2) as a ther- gene 39, 4390–4403 (2020). apeutic target. Biochem. Pharmacol. 151, 188–200 (2018). 50. Robles, M. S., Boyault, C., Knutti, D., Padmanabhan, K. & Weitz, C. J. Identification 80. Sawada, G. et al. CHD8 is an independent prognostic indicator that regulates of RACK1 and protein kinase Calpha as integral components of the mammalian Wnt/β-catenin signaling and the cell cycle in gastric cancer. Oncol. Rep. 30, circadian clock. Science 327, 463–466 (2010). 1137–1142 (2013). 51. Becquet, D. et al. Evidence for an internal and functional circadian clock in rat 81. Masoud, G. N. & Li, W. HIF-1α pathway: role, regulation and intervention for pituitary cells. Mol. Cell Endocrinol. 382, 888–898 (2014). cancer therapy. Acta Pharm. Sin. B 5, 378–389 (2015). 52. Liao, Y., Tong, L., Tang, L. & Wu, S. The role of cold-inducible RNA binding 82. Warzecha, C. C., Sato, T. K., Nabet, B., Hogenesch, J. B. & Carstens, R. P. ESRP1 and protein in cell stress response. Int J. Cancer 141, 2164–2173 (2017). ESRP2 are epithelial cell-type-specific regulators of FGFR2 splicing. Mol. Cell 33, 53. White, E. J., Brewer, G. & Wilson, G. M. Post-transcriptional control of gene 591–601 (2009). expression by AUF1: mechanisms, physiological targets, and regulation. Biochim. 83. Kaprio, T., Hagström, J., Andersson, L. C. & Haglund, C. Tetraspanin CD63 inde- Biophys. Acta 1829, 680–688 (2013). pendently predicts poor prognosis in colorectal cancer. Histol. Histopathol. 35, 54. Hoekstra, M. M., Emmenegger, Y., Hubbard, J. & Franken, P. Cold-inducible RNA- 887–892 (2020). binding protein (CIRBP) adjusts clock-gene expression and REM-sleep recovery 84. Lai, X. et al. Decreased expression of CD63 tetraspanin protein predicts elevated following sleep deprivation. Elife 8, https://doi.org/10.7554/eLife.43400 (2019). malignant potential in human esophageal cancer. Oncol. Lett. 13,4245–4251 (2017). 55. Kim, H. R. et al. SRSF5: a novel marker for small-cell lung cancer and pleural 85. Tummers, B. & Green, D. R. Caspase-8: regulating life and death. Immunol. Rev. metastatic cancer. Lung Cancer 99,57–65 (2016). 277,76–89 (2017). 56. Kandoth, C. et al. Mutational landscape and significance across 12 major cancer 86. Fianco, G. et al. Caspase-8 contributes to angiogenesis and chemotherapy types. Nature 502, 333–339 (2013). resistance in glioblastoma. Elife 6, https://doi.org/10.7554/eLife.22593 (2017). 57. Lee, Y. H., Jhuang, Y. L., Chen, Y. L., Jeng, Y. M. & Yuan, R. H. Paradoxical 87. De Blasio, A. et al. Unusual roles of caspase-8 in triple-negative breast cancer cell overexpression of MBNL2 in hepatocellular carcinoma inhibits tumor growth line MDA-MB-231. Int J. Oncol. 48, 2339–2348 (2016). and invasion. Oncotarget 7, 65589–65601 (2016). 88. Comoglio, P. M., Trusolino, L. & Boccaccio, C. Known and novel roles of the MET 58. Anczuków, O. & Krainer, A. R. Splicing-factor alterations in cancers. RNA 22, oncogene in cancer: a coherent approach to targeted therapy. Nat. Rev. Cancer 1285–1301 (2016). 18, 341–358 (2018). 59. Blijlevens, M., Li, J. & van Beusechem, V. W. Biology of the mRNA splicing 89. Park, M. et al. A novel isoform of met receptor tyrosine kinase blocks hepatocyte machinery and its dysregulation in cancer providing therapeutic opportunities. growth factor/Met signaling and stimulates skeletal muscle cell differentiation. J. Int. J. Mol. Sci. 22, https://doi.org/10.3390/ijms22105110 (2021). Biol. Chem. 290, 1804–1817 (2015). 60. da Silva, M. R. et al. Splicing regulators and their roles in cancer biology and 90. Guo, C., Liu, S., Wang, J., Sun, M. Z. & Greenaway, F. T. ACTB in cancer. Clin. Chim. therapy. Biomed. Res. Int. 2015, 150514 (2015). Acta 417,39–44 (2013). npj Systems Biology and Applications (2022) 17 Published in partnership with the Systems Biology Institute D. Malhan et al. 91. Matsuda, Y., Hagio, M., Seya, T. & Ishiwata, T. Fibroblast growth factor receptor 2 123. Kanehisa, M., Sato, Y. & Kawashima, M. KEGG mapping tools for uncovering IIIc as a therapeutic target for colorectal cancer cells. Mol. Cancer Therapeutics hidden features in biological data. Protein Sci. 31,47–53 (2022). 11, 2010–2020 (2012). 124. Keshava Prasad, T. S. et al. Human Protein Reference Database-2009 update. 92. Isacke, C. M. & Yarwood, H. The hyaluronan receptor, CD44. Int J. Biochem. Cell Nucleic Acids Res. 37, D767–772 (2009). Biol. 34, 718–721 (2002). 125. Snel, B., Lehmann, G., Bork, P. & Huynen, M. A. STRING: a web-server to retrieve 93. Coppola, D. et al. CD44V6 expression in human colorectal carcinoma. Hum. and display the repeatedly occurring neighbourhood of a gene. Nucleic Acids Pathol. 29, 627–635 (1998). Res. 28, 3442–3444 (2000). 94. Mikami, T. et al. Decreased expression of CD44, alpha-catenin, and deleted colon 126. Stark, C. et al. BioGRID: a general repository for interaction datasets. Nucleic Acids carcinoma and altered expression of beta-catenin in ulcerative colitis-associated Res. 34, D535–539 (2006). dysplasia and carcinoma, as compared with sporadic colon neoplasms. Cancer 127. Bovolenta, L. A., Acencio, M. L. & Lemke, N. HTRIdb: an open-access database for 89, 733–740 (2000). experimentally verified human transcriptional regulation interactions. BMC 95. Gotley, D. C. et al. Alternatively spliced variants of the cell adhesion molecule CD44 Genomics 13, 405 (2012). and tumour progression in colorectal cancer. Br. J. Cancer 74,342–351 (1996). 128. Yu, G., Wang, L. G., Han, Y. & He, Q. Y. clusterProfiler: an R package for comparing 96. Todaro, M. et al. CD44v6 is a marker of constitutive and reprogrammed cancer biological themes among gene clusters. Omics 16, 284–287 (2012). stem cells driving colon cancer metastasis. Cell Stem Cell 14, 342–356 (2014). 129. Walter, W., Sánchez-Cabo, F. & Ricote, M. GOplot: an R package for visually 97. Saito, S. et al. CD44v6 expression is related to mesenchymal phenotype and poor combining expression data with functional analysis. Bioinformatics 31, prognosis in patients with colorectal cancer. Oncol. Rep. 29, 1570–1578 (2013). 2912–2914 (2015). 98. Li, M. et al. Drug delivery systems based on CD44-targeted glycosaminoglycans for cancer therapy. Carbohydr. Polym. 251, 117103 (2021). 99. Burnichon, N. et al. SDHA is a tumor suppressor gene causing paraganglioma. ACKNOWLEDGEMENTS Hum. Mol. Genet. 19, 3011–3020 (2010). We are gratefull to all members of Relógio group for their critical remarks and 100. Yu, W., Tang, L., Lin, F., Yao, Y. & Shen, Z. DGKZ acts as a potential oncogene in feedback. Vector icons used in this manuscript were obtained from Flaticon (https:// osteosarcoma proliferation through its possible interaction With ERK1/2 and www.flaticon.com/) and modified accordingly. MYC pathway. Front Oncol. 8, 655 (2018). 101. Heldin, C. H. Targeting the PDGF signaling pathway in tumor treatment. Cell Commun. Signal 11, 97 (2013). AUTHOR CONTRIBUTIONS 102. Dallmann, R., Okyar, A. & Levi, F. Dosing-time makes the poison: Circadian A.R. conceptualized the study, attained the funding, carried out the investigation, regulation and pharmacotherapy. Trends Mol. Med 22, 430–445 (2016). provided supervision, prepared the initial draft, reviewed and revised the manuscript. 103. Lévi, F., Okyar, A., Dulong, S., Innominato, P. F. & Clairambault, J. Circadian timing D.M. carried out the transcriptome data analysis, visualization, prepared the initial in cancer treatments. Annu. Rev. Pharm. Toxicol. 50, 377–421 (2010). draft, reviewed and revised the manuscript. A.B. carried out in vitro experiments, 104. Giacchetti, S. et al. Phase III trial comparing 4-day chronomodulated therapy reviewed and revised the manuscript. All authors have read and agreed to the final versus 2-day conventional delivery of fluorouracil, leucovorin, and oxaliplatin as version of the manuscript. first-line chemotherapy of metastatic colorectal cancer: the European Organi- sation for Research and Treatment of Cancer Chronotherapy Group. J. Clin. Oncol. 24, 3562–3569 (2006). FUNDING 105. Alamoudi, A. A. Why do cancer cells break from host circadian rhythm? Insights from unicellular organisms. Bioessays 43, e2000205 (2021). The work in the group of A.R. was funded by the Dr. Rolf M. Schwiete Stiftung. A.B. 106. Pliatsika, V. & Rigoutsos, I. “Off-Spotter”: very fast and exhaustive enumeration of was additionally funded by the Berlin School of Integrative Oncology (BSIO) graduate genomic lookalikes for designing CRISPR/Cas guide RNAs. Biol. Direct 10, 4 (2015). program funded within the German Excellence Initiative, and the German Cancer 107. Hodgkins, A. et al. WGE: a CRISPR database for genome engineering. Bioinfor- Consortium (GCC). Open Access funding enabled and organized by Projekt DEAL. matics 31, 3078–3080 (2015). 108. Zhang, Z. et al. Genome-wide effect of pulmonary airway epithelial cell-specific Bmal1 deletion. FASEB J. 33, 6226–6238 (2019). COMPETING INTERESTS 109. Pembroke, W. G., Babbs, A., Davies, K. E., Ponting, C. P. & Oliver, P. L. Temporal The authors declare no competing interests. transcriptomics suggest that twin-peaking genes reset the clock. Elife 4, https:// doi.org/10.7554/eLife.10518 (2015). 110. Ikeda, R. et al. REV-ERBα and REV-ERBβ function as key factors regulating ADDITIONAL INFORMATION mammalian Circadian output. Sci. Rep. 9, 10171 (2019). Supplementary information The online version contains supplementary material 111. Fuhr, L. A. The Circadian Clock Modulates Tumour Progression and Drug Response available at https://doi.org/10.1038/s41540-022-00225-w. in Colorectal Cancer Cells through Metabolic Phenotype Rewiring (Humboldt- Universität zu Berlin, 2019). Correspondence and requests for materials should be addressed to Angela Relógio. 112. Simon, A. FastQC: a quality control tool for high throughput sequence data. Avail- able online: https://www.bioinformatics.babraham.ac.uk/projects/fastqc/ (2010). Reprints and permission information is available at http://www.nature.com/ 113. Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: a flexible trimmer for Illumina reprints sequence data. Bioinformatics 30, 2114–2120 (2014). 114. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29,15–21 Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims (2013). in published maps and institutional affiliations. 115. Patro, R., Duggal, G., Love, M. I., Irizarry, R. A. & Kingsford, C. Salmon provides fast and bias-aware quantification of transcript expression. Nat. Methods 14, 417–419 (2017). 116. Soneson, C., Love, M. I. & Robinson, M. D. Differential analyses for RNA-seq: Open Access This article is licensed under a Creative Commons transcript-level estimates improve gene-level inferences. F1000Res 4, 1521 (2015). Attribution 4.0 International License, which permits use, sharing, 117. Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: a Bioconductor package adaptation, distribution and reproduction in any medium or format, as long as you give for differential expression analysis of digital gene expression data. Bioinformatics appropriate credit to the original author(s) and the source, provide a link to the Creative 26, 139–140 (2009). Commons license, and indicate if changes were made. The images or other third party 118. Thaben, P. F. & Westermark, P. O. Detecting rhythms in time series with RAIN. J. material in this article are included in the article’s Creative Commons license, unless Biol. Rhythms 29, 391–400 (2014). indicated otherwise in a credit line to the material. If material is not included in the 119. Cornelissen, G. Cosinor-based rhythmometry. Theor. Biol. Med. Model. 11,16(2014). article’s Creative Commons license and your intended use is not permitted by statutory 120. Carlucci, M. et al. DiscoRhythm: an easy-to-use web application and R package regulation or exceeds the permitted use, you will need to obtain permission directly for discovering rhythmicity. Bioinformatics 36, 1952–1954 (2019). from the copyright holder. To view a copy of this license, visit http://creativecommons. 121. Trincado, J. L. et al. SUPPA2: fast, accurate, and uncertainty-aware differential org/licenses/by/4.0/. splicing analysis across multiple conditions. Genome Biol. 19, 40 (2018). 122. Thaben, P. F. & Westermark, P. O. Differential rhythmicity: detecting altered © The Author(s) 2022 rhythmicity in biological data. Bioinformatics 32, 2800–2808 (2016). Published in partnership with the Systems Biology Institute npj Systems Biology and Applications (2022) 17

Journal

npj Systems Biology and ApplicationsSpringer Journals

Published: May 12, 2022

There are no references for this article.