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

Learn More →

Proliferative potential and resistance to immune checkpoint blockade in lung cancer patients

Proliferative potential and resistance to immune checkpoint blockade in lung cancer patients Background: Resistance to immune checkpoint inhibitors (ICIs) has been linked to local immunosuppression independent of major ICI targets (e.g., PD-1). Clinical experience with response prediction based on PD-L1 expression suggests that other factors influence sensitivity to ICIs in non-small cell lung cancer (NSCLC) patients. Methods: Tumor specimens from 120 NSCLC patients from 10 institutions were evaluated for PD-L1 expression by immunohistochemistry, and global proliferative profile by targeted RNA-seq. (Continued on next page) * Correspondence: Carl.Morrison@OmniSeq.com; Carl.Morrison@Roswellpark.org OmniSeq, Inc., 700 Ellicott Street, Buffalo, NY 14203, USA Roswell Park Comprehensive Cancer Center, Elm and Carlton Streets, Buffalo, NY 14206, USA Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 2 of 13 (Continued from previous page) Results: Cell proliferation, derived from the mean expression of 10 proliferation-associated genes (namely BUB1, CCNB2, CDK1, CDKN3, FOXM1, KIAA0101, MAD2L1, MELK, MKI67, and TOP2A), was identified as a marker of response to ICIs in NSCLC. Poorly, moderately, and highly proliferative tumors were somewhat equally represented in NSCLC, with tumors with the highest PD-L1 expression being more frequently moderately proliferative as compared to lesser levels of PD-L1 expression. Proliferation status had an impact on survival in patients with both PD-L1 positive and negative tumors. There was a significant survival advantage for moderately proliferative tumors compared to their combined highly/poorly counterparts (p = 0.021). Moderately proliferative PD-L1 positive tumors had a median survival of 14.6 months that was almost twice that of PD-L1 negative highly/poorly proliferative at 7.6 months (p =0.028). Median survival in moderately proliferative PD-L1 negative tumors at 12.6 months was comparable to that of highly/ poorly proliferative PD-L1 positive tumors at 11.5 months, but in both instances less than that of moderately proliferative PD-L1 positive tumors. Similar to survival, proliferation status has impact on disease control (DC) in patients with both PD-L1 positive and negative tumors. Patients with moderately versus those with poorly or highly proliferative tumors have a superior DC rate when combined with any classification schema used to score PD-L1 as a positive result (i.e., TPS ≥ 50% or ≥ 1%), and best displayed by a DC rate for moderately proliferative tumors of no less than 40% for any classification of PD-L1 as a negative result. While there is an over representation of moderately proliferative tumors as PD-L1 expression increases this does not account for the improved survival or higher disease control rates seen in PD-L1 negative tumors. Conclusions: Cell proliferation is potentially a new biomarker of response to ICIs in NSCLC and is applicable to PD-L1 negative tumors. Keywords: Atezolizumab, Nivolumab, Pembrolizumab, Ipilimumab, PD-1 Background diagnostic for pembrolizumab treatment, PD-L1 expres- On March 4th 2015, nivolumab (Opdivo®, from Bristol- sion levels assessed by the PD-L1 22C3 pharmDx assay Myers Squibb) became the first immune checkpoint (from Agilent) [12]. However, response prediction based inhibitor (ICI) to be approved by the US Food and Drug on PD-L1 levels is not 100% accurate. For instance, Administration for use in patients with metastatic pembrolizumab monotherapy in NSCLC patients with a nonsquamous non-small cell lung cancer (NSCLC) PD-L1 tumor proportion score (TPS) < 1% (i.e., membran- progressing on or after platinum-based chemotherapy [1], ous PD-L1 expression on < 1% malignant cells), of 1–49%, following disclosure of the results from the Phase III and ≥ 50% was associated with response rates of 10.7, 16.5, Checkmate-037 trial [2]. Since then, three other ICIs that and 45.2%, respectively [3]. Thus, a small population of inhibit the programmed cell death pathway, including NSCLC patients with low PD-L1, seemingly “negative bio- programmed cell death 1 (PDCD1 or CD279, best known marker” patients, will still respond to ICI-based therapy. as PD-1) and its ligands – CD274 (best known as PD-L1) Conversely, not all patients with high PD-L1 TPS obtain and programmed cell death 1 ligand 2 (PDCD1LG2 or clinical benefits from ICIs, which suggests the existence of CD273, best known as PD-L2) – have been licensed for alternative resistance mechanisms, such as mutations that use in NSCLC patients, namely pembrolizumab affect the ability of cancer cells to be recognized or elimi- (Keytruda®, from Merck) [3, 4], atezolizumab (Tecentriq®, nated by the immune system [9], or other mechanism of from Genentech) [5, 6], and durvalumab (Imfinzi®, from local immunosuppression in the tumor microenvironment AstraZeneca) [7]. Response rates to these ICIs employed via pathways that do not directly involve ICI targets such as single agent immunotherapeutic interventions in an as PD-L1 and PD-1 [3]. unselected population, however, is generally below 20% We employed targeted RNA sequencing of an immune [3]. Moreover, ICI-based immunotherapy has been related panel of slightly less than 400 genes to optimize estimated to cost 100,000–250,000 USD per patient (with the detection of low expressing genes as opposed to some variation depending on specific ICI, treatment whole transcriptome, that was specifically designed for regimen and duration) [8]. Thus, considerable efforts are use in formalin fixed paraffin embedded (FFPE) clinical being devoted to the elucidation of the mechanisms samples [13]. This list of genes was divided into 41 controlling the development of primary and acquired different immune function categories and analyzed for resistance to ICIs [9], as well as to the identification of response to ICIs in a cohort of NSCLC patients from biomarkers with robust predictive value [10, 11]. ten different institutions. The highest association with These observations have rapidly been translated into the response among the different immune function categories clinical management of NSCLC with the FDA companion was cell proliferation, represented by the expression of ten Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 3 of 13 unique genes. We demonstrate that either extreme of within 90 days of first dose were excluded as it could not cellular proliferation in the tumor microenvironment, i.e. be discerned whether they were rapid progressors or had highly or poorly proliferative, is associated with resistance poor performance prior to going on drug. Patients to ICIs amongst NSCLC patients, and that the expression lacking sufficient follow up time for response evaluation levels of a 10-gene set associated with cellular proliferation (less than 90 days from first dose), were also excluded can be harnessed to improve patient stratification beyond from analysis. Of the120 ICI-treated patients, for all of PD-L1 TPS. Most importantly, we show that additional which survival data was available, there were 10 patients stratification of PD-L1 negative NSCLC based upon cell not evaluable for response due to either no measurable proliferation status introduces a new approach to response disease or target lesion (n = 4), missing scans (n = 4), or to ICI therapy in NSCLC. not specified (n = 2) (Fig. 1a). For the remaining 110 pa- tients all were evaluable for response based on RECIST Methods v1.1 and were divided into a test set (n = 34) from one Patients and clinical data institution with the most patients (Duke) and a training Ten collaborating institutions obtained approval by their set (n = 76) from all other institutions. Patients whose respective institutional review boards (IRBs) to submit best response was complete response (CR), partial existing de-identified specimens and associated clinical response (PR), or stable disease (SD) with 12 months or data for use in this study. A total of 120 patients were more survival were classified as disease control (DC), included in the study (Fig. 1a), based on the following while patients whose best response was progressive criteria: (1) history of Stage IV NSCLC; (2) availability of disease (PD) or SD with less than 12 months survival adequate archival formalin-fixed paraffin-embedded were classified as no disease control (NDC). Duration of (FFPE) tissue collected prior to treatment with ICIs; (3) response was not available for all patients and not availability of sequencing data; and (4) availability of included for final analysis. demographic, diagnosis, follow-up and survival data. Table 1 summarizes the baseline clinical characteristics Immunohistochemical studies of these patients (individual patient data provided in The expression of PD-L1 on the surface of cancer cells Additional file 1: Table S1). was assessed in all cases by means of the Dako Omnis Patients who were treated with ICIs were included if Platform and the 22C3 pharmDx antibody (Agilent, they were treated by an agent approved by the FDA as Santa Clara, CA) using FDA-scoring guidelines [14]. of November 2017 and had follow up and survival from Briefly, a minimum of 100 viable tumor cells were first ICI dose (n = 120). ICI-treated patients who died assessed for membranous staining of any intensity for Fig. 1 Summary of patient disposition and exploratory analysis. a) A total of 120 patients previously treated with checkpoint inhibitors were included in the study. All patients had survival data from date of first dose of checkpoint inhibitor, while 110 were evaluable by RECIST v1.1 for response. b) Exploratory analysis using pair-wise proportion test of 41 immune-related gene functions derived from 394 genes for patients with disease control versus no disease control identifies cell proliferation as a biomarker of interest Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 4 of 13 Table 1 Patient characteristics Data analysis Immune gene expression ranks (range 0–100) from a Patients (n = 120) targeted RNA-seq immune panel of approximately 400 Age at initial diagnosis (years) genes were divided into 41 biological function categories < 30 1 (00.0) according to commercial annotations from the manufac- 30–39 1 (00.0) turer (Additional file 1: Table S2). For all 110 cases with 40–49 4 (03.3) response, distribution of each biological function was 50–59 28 (23.3) split into 3 tertiles of low (less than 33), medium 60–69 43 (35.8) (between 33 and 66) and high (greater than 66). Next, we performed a pair wise proportion test (chi-square 70–79 34 (28.3) test) to test for difference in DC rates for these three ≥ 80 9 (07.5) tertiles (i.e. low vs medium, medium vs high and low vs Mean 65 high) for each biological function (Fig. 1b). Proportion Sex test was performed with continuity correction and pair- Female 61 (50.8) wise p-values for each biological function were adjusted Male 59 (49.2) for multiple hypothesis testing using “holmes” correc- tion. We further divided the dataset into a training set Race (n = 76) consisting of samples from all data access White 96 (80.0) groups except the largest contributor. A separate test set Other 17 (14.2) (n = 34) was constituted from samples from a single Unknown 7 (05.8) largest contributing institute. Any biological function Vital status at last follow up that did not have cases representing one or more tertiles Alive 60 (50.0) was removed from further analysis due to lack of dynamic range of that biological function in the popula- Dead 60 (50.0) tion assessed in this study. The most significant gene Checkpoint inhibitor functions were utilized for further analysis. Survival atezolizumab 2 (01.7) analysis was performed using a log-rank test on 5-year ipilimumab + nivolumab 2 (01.7) Kaplan-Meier survival curves for PD-L1 levels assessed nivolumab 79 (65.8) by IHC and combined expression of 10 proliferation-re- pembrolizumab 37 (30.8) lated genes assessed by RNA-Seq. Comparison of DC rate was performed using Chi-square test with Yate’s the 22C3 antibody. The percentage of viable tumor cells continuity correction. Multivariate analysis was per- showing partial or complete membrane staining relative formed by fitting a binomial logistic regression model to to all viable tumor cells present in the sample (positive DC labels and co-variates such as proliferation status, and negative) was then used to derive a tumor proportion PD-L1 status, histology, race, sex, and age category. score (TPS). PD-L1 levels were scored by a board-certified Analysis of variance (ANOVA) was performed on the anatomic pathologist as per published guidelines [15], fitted model to study the table of deviance to determine with a TPS ≥ 50% considered as a strongly positive result the co-variate that explains the most variance in the for different comparisons, while a result of ≥1% consid- DC rates. ered as positive result for different comparisons. PD-L1 TPS ≥ 1% to < 50% were considered weakly positive for Results additional comparative purposes. PD-L1 TPS < 1% was Immune-related gene functions considered as negative. Ki-67 positivity amongst neoplas- Among 41 different immune-related gene functions tic and immune cells was scored upon nuclear staining, (Additional file 1: Table S2) evaluated by pairwise regardless of intensity, with the M7240 (clone MIB1) anti- comparison test in the training set (n = 76), three were body from Dako (Carpentaria, CA) with the percentage of significantly differentially expressed for DC versus NDC each cell type recorded. for at least one comparison (Additional file 1: Table S3). These three functions and specific genes (see Additional RNA-seq file 1: Table S2 for full gene names) included proliferation RNA were extracted from each sample and processed [BUB1, CCNB2, CDK1, CDKN3, FOXM1, KIAA0101, for targeted RNA-seq, as previously described [13, 16]. MAD2L1, MELK, MKI67 (better known as Ki-67), and Gene expression was evaluated by amplicon sequencing of TOP2A; maximum p = 0.0092], antigen processing (CD74, 394 immune transcripts on samples that met validated HLA-A, HLA-B, HLA-C, HLA-DMA, HLA-DMB, quality control (QC) thresholds [13]. HLA-DOA, HLA-DOB, HLA-DPA1, HLA-DPB1, Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 5 of 13 HLA-DQA1, HLA-DQA2, HLA-DQB2, HLA-DRA, (adenocarcinoma n = 94, sarcomatoid carcinoma n =1, HLA-DRB1, HLA-E, HLA-F, HLA-F-AS1, HLA-G; p = squamous cell carcinoma n = 25) was used as the pri- 0.0796), and dendritic cell (HERC6, IL3RA, ITGAX, NRP1, mary indicator for the proliferative status of the tumor TLR3, ZBTB46; p = 0.0903). When both the training and microenvironment. Tumors were stratified into poorly, test set (n = 110) were used for the same comparison moderately and highly proliferative based on the tertile (Additional file 1: Table S4), proliferation was the only of rank of expression of this gene signature as compared to these three functions that was significant (Fig. 1b). Results a separate reference population of 167 patients with for the test set (n = 34) did not identify proliferation, anti- multiple tumor types (Additional file 1: Table S6) [10]. gen processing, or dendritic cell as significant (Additional Based on this analysis, poorly proliferative tumors were file 1: Table S5), presumably due to the small size of the the least frequent in all available samples tested (27/120; sample set. Proliferation was chosen for further evaluation 22.5%), followed by an equal distribution of highly (47/ based upon the identification as a significant factor in the 120; 39.2%) and moderately proliferative tumors (46/120; training set as well as the combination of the training and 38.3%), (Fig. 2a). test set. To define whether neoplastic cells, immune cells, or both constituted the source of proliferation-related Proliferative status transcripts, 7 highly proliferative and 9 poorly prolifera- NSCLC had a wide distribution of poorly, moderately, tive cases were evaluated by immunohistochemistry for and highly proliferative tumors with input by both the expression of MKI67 (best known as Ki-67), a neoplastic and immune cells that can be measured in biomarker of proliferation largely employed in the clinics more than one way. The mean expression rank values of [17]. Highly proliferative tumors (as defined by 10 proliferation-related genes in 120 NSCLC specimens RNA-seq) had > 50% of neoplastic cells staining positive Fig. 2 Results for cell proliferation as an independent biomarker. a) Proportion of 120 NSCLC patients for cell proliferation by tertiles of poorly, moderately, and highly proliferative. b) Proportion of 120 NSCLC patients positive or negative for PD-L1 IHC using a cut-off of tumor proportion score of ≥50% as a positive result. c) Proportion of 120 NSCLC patients positive or negative for PD-L1 IHC using a cut-off of tumor proportion score of ≥1% as a positive result. d) Prevalence for all combinations of strongly positive PD-L1 (TPS ≥ 50%) cases and proliferation status. e) Prevalence for all combinations of PD-L1 and proliferation status for weakly positive PD-L1 cases (TPS ≥ 1 and < 50%). f) Prevalence for all PD-L1 negative (TPS < 1%) cases and proliferation status. Number and p values are reported Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 6 of 13 for Ki-67 in 6 out of 7 cases, while their poorly prolifera- 10 proliferation-related genes were evaluated for accur- tive counterparts contained less than 40% Ki-67 neo- acy (i.e. true positive plus true negatives divided by total plastic cells in 8 of 9 cases (Additional file 1: Table S7). number of results) for each gene individually (Additional In a similar fashion, highly proliferative tumors had 5% file 1: Table S6). Accuracy ranged from a low of 52.7% or more of immune cells staining positive for Ki-67 in for FOXM1 to a high of 67.3% for TOP2A, as compared all cases, while their poorly proliferative counterparts to a value of 71.8% for the mean expression rank values showed only two cases with this degree of reactivity. of all ten proliferation-related genes (Additional file 2: Importantly, an abundant tumor CD8 T-cell infiltrate Figure S1). The accuracy of Ki-67 at 59.1% was near the did not necessarily correlate with a highly proliferative mid-value of other single gene results. tumor microenvironment. For example, in one poorly The sum of all of these results suggest that poorly, mod- proliferative adenocarcinoma (Fig. 3a) there is a lack of erately, and highly proliferative tumors are somewhat staining by Ki-67 in both malignant and immune cells equally represented in NSCLC; that both immune cells (Fig. 3b), even though there is an abundance of CD8 T and malignant cells are sources of proliferation-related cells (Fig. 3c). In comparison, for a highly proliferative transcripts, and it is possible to reach similar results for adenocarcinoma (Fig. 3d) there is frequent staining by any of the 10 genes using only single gene evaluations. Ki-67 in both malignant and immune cells (Fig. 3e), with a similar number of CD8 T cells (Fig. 3f). PD-L1 expression To evaluate the impact of single gene proliferation Tumors with the highest PD-L1 expression were more results, e.g. Ki-67, the mean expression rank values of all frequently moderately proliferative as compared to lower Fig. 3 Immunohistochemical assessment of Ki-67 positivity and CD8 T cell infiltration. Representative fields for hematoxylin/eosin (a, d), CD8 positivity (b, e) and Ki-67 positivity (c, f) are depicted. The left hand panel (a-c) of a poorly proliferative tumor shows numerous CD8+ T-cells (c), while Ki-67 (b) stains very few neoplastic or immune cells. The right hand panel (d-f) of a highly proliferative tumor like the other case shows numerous CD8 T-cells (f), while Ki-67 (e) stains a high number of neoplastic and immune cells. Scale bar = 100 μm Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 7 of 13 levels of PD-L1 expression. PD-L1 TPS, defined as the or a combination of poorly and highly proliferative (p = percentage of neoplastic cells displaying membranous 0.02317), but less so for highly proliferative (p= 0.1188), positivity of any intensity upon staining with the DAKO (Fig. 2f). Overall these results support that as PD-L1 22C3 antibody, ranged from 0 to 100 and 32/120 expression increases there is an over representation of (26.7%) of all cases were strongly positive (Fig. 2b), while moderately proliferative tumors, but as shown below 56/120 (46.7%) of all cases were positive at any level of does not account for the improved survival or higher staining (Fig. 2c). Moderately proliferative tumors were disease control rates seen in PD-L1 negative tumors. slightly enriched for strongly positive PD-L1 tumors as compared to highly proliferative tumors (p= 0.4611), Overall survival and more so as compared to poorly proliferative tumors Proliferation status had an impact on survival in patients (p = 0.01237), or a combination of the latter two (p = with both PD-L1 positive and negative tumors. There 0.07227), (Fig. 2d). For weakly positive PD-L1 tumors, was a significant survival advantage for moderately moderately proliferative were not enriched as compared proliferative tumors compared to their combined highly/ to poorly proliferative counterparts (p= 1.0), highly poorly counterparts (p = 0.021) (Fig. 4a). When highly proliferative (p= 0.2463), or a combination of the latter and poorly proliferative groups were evaluated separately two (p = 0.5417), (Fig. 2e). For PD-L1 negative tumors, there was a trend towards survival for patients with moderately proliferative were under represented as com- moderately proliferative tumors (p = 0.064) (Fig. 4b). pared to poorly proliferative counterparts (p= 0.01955), Likewise, the survival of patients with strongly positive Fig. 4 Overall survival of 120 NSCLC patients receiving an immune checkpoint inhibitor (ICI) as part of their therapy. a) Overall survival based upon stratification by cell proliferation for moderately versus combined poorly/highly proliferative. b) Overall survival based upon stratification by cell proliferation for moderately versus poorly and highly proliferative. c) Overall survival based upon stratification by PD-L1 expression levels using TPS ≥ 50% as a cut-off for a positive result. d) Overall survival based upon stratification by strongly positive PD-L1 tumors and proliferation status (PD-L1 TPS ≥ 50% moderately proliferative, PD-L1 TPS ≥ 50% highly or poorly proliferative, PD-L1 TPS ≥ 50% moderately proliferative, PD-L1 TPS ≥ 50% highly or poorly proliferative). Number at risk and p-values are reported Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 8 of 13 PD-L1 tumors was associated with a statistically signifi- DC rate when combined with any classification schema cant survival advantage (p = 0.03) (Fig. 4c). A combin- used to scorePD-L1 as apositiveresult(i.e., TPS ≥ 50% or ≥ ation of proliferation and PD-L1 resulted in a significant 1%; see Additional file 4 for full results). The value of cell survival advantage in moderately proliferative strongly proliferation as a marker of response was best displayed by positive PD-L1 tumors with a median survival of 14.6 noting that the DC rate for moderately proliferative tumors months that was almost twice that of all less than was no less than 40% for any classification of PD-L1 as a strongly positive PD-L1 highly/poorly proliferative negative result. This was critically important for the tumors at 7.6 months (p = 0.028) (Fig. 4d). Likewise, fifty-seven negative PD-L1 negative tumors for which mod- median survival in less than strongly positive PD-L1 erately proliferative tumors had a DC rate of 41.2% (7/17) moderately proliferative tumors at 12.6 months was (Fig. 5g), while the DC rate among highly and poorly prolif- comparable to that of highly/poorly proliferative strongly erative tumors combined was 17.5% (7/40, p = 0.1179). The positive PD-L1 tumors at 11.5 months (p = 0.86) (Fig. 4d), summary of all of these results support that cell prolifera- but in both instances less than that of moderately prolifer- tion is a relevant biomarker in all groups of NSCLC, but is ative strongly positive PD-L1 tumors. The results for all unique and clinically useful for patients with PD-L1 PD-L1 positive tumors by a TPS ≥ 1% criteria were very negative tumors. Further support of this conclusion was a similar (Additional file 3: Figure S2). The summary of multivariate analysis on all co-variates using binomial these results support that moderately proliferative tumors logistic regression model showed that moderately pro- have a survival advantage beyond PD-L1 positive tumors liferative tumors to have a significant association with for NSCLC patients treated with checkpoint inhibitors. probability of disease control (Table 3; p = 0.0071). Furthermore, analysis of deviance of each co-variate Disease control rate (Table 3) suggests that adding proliferation to a null Similar to survival, proliferation status had an impact on model improved it significantly (p = 0.0009) followed disease control in patients with both PD-L1 positive and by a second most informative co-variate of PD-L1 negative tumors. The overall objective of evaluating status (p = 0.0337). Collectively these results suggest disease control was to show this intersection of response that, the proliferative status of the tumor microenvir- to checkpoint inhibition for cell proliferation versus the onment can be harnessed to improve patient stratifi- current standard of PD-L1 IHC. The results (Table 2, cation based on PD-L1 expression levels. Importantly, Fig. 5) show that patients with moderately versus those cell proliferation seems to have value as a biomarker with poorly or highly proliferative tumors have a superior of response in PD-L1 negative tumors. Table 2 Disease control for cell proliferation and PD-L1 IHC Cell Proliferation PD-L1 IHC DC NDC Total pts DC rate χ2 test Moderately 22 22 44 50.0% Highly 9 33 42 21.4% p = 0.0146 Poorly 4 20 24 16.7% p = 0.0113 Poorly/highly 13 53 66 19.7% p = 0.0017 Strongly positive (TPS ≥ 50%) 16 16 32 50.0% Not strongly positive (TPS < 50%) 19 59 78 24.4% p = 0.0009 Positive (TPS ≥ 1%) 21 32 53 39.6% Negative (TPS < 1%) 14 43 57 24.6% p = 0.1363 Moderate Strongly positive (TPS ≥ 50%) 10 7 17 58.8% Poorly/highly 6 9 15 40.0% p = 0.4786 Moderately Not strongly positive (TPS < 50%) 12 15 27 44.4% Highly 4 25 29 13.8% p = 0.0250 Poorly 3 19 22 13.6% p = 0.0438 Poorly/highly 7 44 51 13.7% p = 0.0063 Moderately cold tumors (CD8 rank < 15%) 7 10 17 41.2% Poorly/highly cold tumors (CD8 rank < 15%) 7 33 40 17.5% p = 0.1179 Moderately cold tumors (CD8 rank < 33%) 5 5 10 50.0% Poorly/highly cold tumors (CD8 rank < 33%) 0 11 11 0.0% p = 0.3298 Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 9 of 13 Fig. 5 Impact of PD-L1 levels and proliferative status on disease control rate in 110 NSCLC patients receiving an immune checkpoint inhibitor (ICI) as part of their therapy. a) Prevalence and DC rate for moderately versus highly and poorly proliferative tumors, as well as combined of the latter two. b) Prevalence and DC rate for strongly positive PD-L1 (TPS ≥ 50%). c) Prevalence and DC rate for PD-L1 negative (TPS < 1%). d) Prevalence and DC rate for strongly positive PD-L1 combined with moderately versus highly/poorly proliferative tumors. e) Prevalence and DC rate for PD-L1 positive (TPS ≥ 1%) combined with moderately versus highly/poorly proliferative tumors. f) Prevalence and DC rate for PD-L1 less than strongly positive (TPS < 50%) combined with moderately versus highly/poorly proliferative tumors. g) Prevalence and DC rate for PD-L1 negative (TPS < 1%) combined with moderately versus highly/poorly proliferative tumors. h) Prevalence and DC rate for weakly positive PD-L1 (TPS ≥ 1% and < 50%) combined with moderately versus highly/poorly proliferative tumors. i) Prevalence and DC rate for minimal tumor infiltration by CD8 T cells (so-called “cold” tumors) combined with moderately versus highly/poorly proliferative tumors Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 10 of 13 Table 3 Multivariate analysis Variable Estimate Std. Error z value p value (Intercept) 17.3526 2712.1561 0.006 0.9949 Proliferation Moderately 1.3503 0.5013 2.694 0.00707 PD.L1. status Positive 0.5169 0.547 0.945 0.34468 Histology SCC.or.Other −0.5898 0.6417 −0.919 0.35801 race Black or African American −34.8319 3301.0002 −0.011 0.99158 race Black or African American −16.7776 2712.156 −0.006 0.99506 race Other −35.6219 4796.5772 −0.007 0.99407 race Other −34.6736 4796.5771 −0.007 0.99423 race Unknown −18.5693 2712.1561 −0.007 0.99454 race White −17.9126 2712.1559 −0.007 0.99473 race White −18.3781 2712.1559 −0.007 0.99459 sex M 0.1522 0.5119 0.297 0.76616 age_cat 1–29 −35.0709 4796.5772 −0.007 0.99417 age_cat 40–49 −1.201 1.6759 −0.717 0.47359 age_cat 50–59 −0.6471 0.9129 −0.709 0.47843 age_cat 60–69 −0.9142 0.8863 −1.031 0.30233 age_cat 70–79 −1.1416 0.9098 −1.255 0.20955 Analysis of deviance of each co-variate Co-variate Df Deviance Resid. Df Resid. Dev P value(>Chi) NULL 109 137.61 Proliferation 1 11.1163 108 126.49 0.0008557 PD.L1.status 1 4.5112 107 121.98 0.0336733 Histology 1 0.0593 106 121.92 0.8076295 race 7 7.4867 99 114.44 0.3800195 sex 1 0.1064 98 114.33 0.7442778 age_cat 5 4.2582 93 110.07 0.5128654 Proliferative status and cold tumors for the more stringent cut-off value (Fig. 5i). Most Proliferation status had an impact on disease control in importantly, the DC rate was greater than 50% for any patients with factors other than PD-L1 positive or nega- grouping of moderately proliferative cold tumors, while tive status, impacting response to checkpoint inhibitors. the rate was less than 20% for poorly/highly proliferative In this regard, cell proliferation was further evaluated for counterparts. PD-L1 status did not associate with value beyond PD-L1 status in the emerging recognition response in cold tumors (Additional file 1: Table S8), of inflammatory status [16], and more specifically the again supporting that cell proliferation is a unique degree of CD8 infiltration. Response was evaluated for biomarker of response in NSCLC. tumors with reduced levels of CD8-coding transcripts as compared to a reference population of 167 patients with Discussion multiple tumor types, which we previously demonstrated Our findings suggest that a highly or poorly proliferative to indicate minimal tumor infiltration by CD8 T cells tumor microenvironment is associated with limited (so-called “cold” tumors) [16]. As there is no current ab- sensitivity to ICIs amongst NSCLC patients, and that solute criteria to define cold tumors we first arbitrarily targeted RNA-seq can be employed to assess the prolif- defined this group by a CD8 rank less than 15, and then erative status of the tumor microenvironment at diagno- compared to those results to an non-arbitrary cut-off of sis, with the ultimate goal of improving clinical decision the lower tertile of CD8 rank, or a value less than 33. making based on PD-L1 only. Most importantly, these Irrespective of the cut-off, DC was accurately predicted findings suggest that some highly or poorly proliferative by the proliferative status of the tumor microenviron- tumors may be resistant to ICIs independent of PD-L1 ment (Table 2), although the numbers are quite small or inflamed status and that both PD-L1 positive and Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 11 of 13 PD-L1 negative tumors at any TPS value can be strati- data from patients receiving PD-1- or PD-L1-targeting fied more accurately by cell proliferation. Moving for- agents (nivolumab, pembrolizumab atezolizumab), ward the need for standardization of cell proliferation CTLA4-targeting agents (ipilimumab), or both (nivolu- will be vitally important in comparing response among mab + ipilimumab) as it complicated subgroup analysis. various studies. In that regard the proliferative potential As a retrospective study across multiple institutions, there of malignant cells (assessed by Ki-67 positivity or were also limitations for data collection. Smoking status enumeration of mitotic figures) has been extensively was not available from all sites and as such was not a vari- employed over the past 3 decades for prognostic able in the multi-variate analysis. The exclusion of purposes in a number of tumors [18–21]. In our study, ICI-treated patients who died in less than 90 days post Ki-67 as measured by RNA-seq analysis was not the first dose checkpoint inhibitor did not allow for an most accurate predictor of disease control as a single analysis of this important group due to the lack of gene result, but rather was TOP2A. At such a formative collection ECOG performance score and our subsequent stage of development we did not evaluate proliferation inability to distinguish rapid progressors from poor health as a continuous variable for any single gene or the mean performance. rank of 10 genes, but this is factor that will need to eval- uated further in future studies. We also did not evaluate Conclusion K-67 or TOP2A IHC as a predictor of disease control In summary, we demonstrated that a poorly or highly and is another potential future study. proliferative potential in the tumor microenvironment is In a recent study, RNA-seq was employed to investi- associated with resistance to ICI-based immunotherapy gate the effect of proliferation on the survival of 6581 amongst NSCLC patients, and that assessing the expres- patients with 19 different cancers, as catalogued by The sion levels of ten proliferation-related genes by RNA-seq Cancer Genome Atlas (TCGA) [22]. In this setting, a in diagnostic biopsies stands out as a promising strategy low proliferation index was associated with improved for improving clinical decision making based on PD-L1 patient survival in 7 of 19 malignancies (including lung expression only. Additional studies are ongoing to test adenocarcinoma) which were subsequently defined as these observations in other tumor types commonly “proliferation-informative cancers” [22]. Most recently, treated with ICIs. another TCGA study evaluating the immune landscape of cancer in more than 10,000 tumors identified six im- Additional files mune subtypes hypothesized to define immune response patterns impacting prognosis [23]. Two of these six sub- Additional file 1: Table S1. Clinical characteristics. Table S2. RNA-seq gene function list. Table S3. Gene function analysis training set (Proportion types, C1 and C2, were noted for a high proliferation rate, test). Table S4. Gene function analysis training and test set combined with both having a substantial immune component but (Proportion test). Table S5. Gene function analysis test set (Proportion test). the least favorable outcomes. In this study tumor types Table S6. Accuracy for 10 proliferation immune-related genes. Table S7. Immunohistochemical assessment of Ki-67 positivity. Table S8. Disease over represented by C1 and C2 subtypes included bladder control rate in cold tumors by proliferation status. (XLSX 82 kb) cancer, breast cancer, cervical cancer, colon cancer, head Additional file 2: Figure S1. Gene specific proliferation values. (TIFF 274 kb) and neck squamous cell carcinoma, lung squamous cell Additional file 3: Figure S2. Disease control rates for PD-L1 positive carcinoma, mesothelioma, ovarian cancer, gastric adeno- (TPS > 1%) and negative tumors combined with cell proliferation status. carcinoma, and endometrial cancer. Moreover, in NSCLC, (TIFF 345 kb) a dormant tumor-infiltrating lymphocytes (TIL) signature Additional file 4: Supplementary tables with clinical annotations and data analysis results. (DOCX 32 kb) characterized by low activation (Granzyme B) and prolif- eration markers (Ki-67) in CD3 + TILs was also recently Abbreviations demonstrated to be associated with survival benefit in pa- CR: Complete Response; DC: Disease Control; FFPE: Formalin-Fixed Paraffin- tients treated with ICI [24]. These studies support that cell Embedded; ICI: Immune Checkpoint Inhibitor; IHC: Immunohistochemistry; proliferation should be evaluated further as an integral IRB: Institutional Review Board; NDC: No Disease Control; ORR: Objective Response Rate; OS: Overall Survival; PD: Progressive Disease; PR: Partial component of the immune response to ICIs and that re- Response; QC: Quality Control; RECIST: Response Evaluation Criteria In Solid sults may be tumor type dependent. Tumors; SD: Stable Disease; TCGA: The Cancer Genome Atlas; TPS: Tumor While our work was not based upon a single, Proportion Score well-structured clinical trial, samples were obtained from Acknowledgements 10 different institutions across the US and Europe, and re- The following individuals participated with the collection of biospecimens sults stood the test of such a heterogeneous, real-world and associated clinical data: Larson Hsu (Roswell Park Cancer Institute), Ryan clinical scenario. One of the major limitations of the Winters (Biosample Repository Facility at Fox Chase Cancer Center), Mary Shields and Ashley Gibbs (Northwest Oncology), Rosemary Makar and Amy present study is that response data (based on RECIST Fricke (Oregon Health & Science University Knight BioLibrary), Pearl v1.1) was available for a relatively small number of cases Abernathy (Mission Health System), and Stephanie Kaufman (acting as (110 patients), which obliged us to operate on pooled honest broker from OmniSeq, Inc). Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 12 of 13 Funding Author details 1 2 This research was funded by OmniSeq, Inc. (Buffalo, NY). OmniSeq, Inc., 700 Ellicott Street, Buffalo, NY 14203, USA. Roswell Park Comprehensive Cancer Center, Elm and Carlton Streets, Buffalo, NY 14206, 3 4 USA. Dartmouth-Hitchcock Medical Center, Lebanon, NH 03756, USA. Duke Availability of data and materials University, Durham, NC 27708, USA. Fox Chase Cancer Center, Philadelphia, The datasets generated and/or analyzed during the current study are not PA 19111, USA. Hospital Universitario Virgen Macarena, 41009 Sevilla, Spain. publicly available due to a non-provisional patent filing covering the 7 8 Medical College of Wisconsin, Milwaukee, WI 53226, USA. Meharry Medical methods used to analyze such datasets but are available from the College, Nashville, TN 37208, USA. Mission Health System, Asheville, NC corresponding author upon reasonable request. 10 11 28801, USA. Community Hospital, Munster, IN 46321, USA. Center for Personalized Cancer Therapy, Moores Cancer Center, La Jolla, CA 92093, USA. Authors’ contributions Department of Radiation Oncology, Weill Cornell Medical College, New GD, EY, AE, HC, WB, KM, KS, KD, LJT, DM, JZ, JC, ML, SM, TZ, MZ, PG, IAF, AS, York, NY 10065, USA. Sandra and Edward Meyer Cancer Center, New York, BG, ACM, JT, RS, RJ, LD, MS, OB, DP, and NS collected and submitted de- NY 10065, USA. Université Paris Descartes/Paris V, 75006 Paris, France. identified patient data with corresponding clinical cases from their respective institutions with IRB approval. JC, MN, DD, STG, CM, SP, and LG contributed Received: 14 November 2018 Accepted: 13 January 2019 to the experimental design of this analysis. CM, SP, JMC, MKN, STG, APS, BB, JA, VG, MQ, YW, FLL, LG, KR, ME, and MG prepared and analyzed patient datasets and corresponding clinical cases and were major contributors to writing and revising the manuscript. All authors read and approved the final References manuscript. 1. Vanpouille-Box C, Lhuillier C, Bezu L, Aranda F, Yamazaki T, Kepp O, et al. Trial watch: immune checkpoint blockers for cancer therapy. Oncoimmunology. 2017;6:e1373237. https://doi.org/10.1080/2162402X.2017.1373237. Ethics approval and consent to participate 2. Borghaei H, Paz-Ares L, Horn L, Spigel DR, Steins M, Ready NE, et al. Nivolumab Existing de-identified specimens and associated clinical data was submitted versus docetaxel in advanced nonsquamous non–small-cell lung Cancer. N to OmniSeq from Roswell Park Comprehensive Cancer Center (Buffalo, NY), Engl J Med. 2015;373:1627–39. https://doi.org/10.1056/NEJMoa1507643. Fox Chase Cancer Center (Philadelphia, PA), Dartmouth-Hitchcock Medical 3. Garon EB, Rizvi NA, Hui R, Leighl N, Balmanoukian AS, Eder JP, et al. Center (Lebanon, NH), Northwest Oncology (Munster, IN), Hospital Universitario Pembrolizumab for the treatment of non–small-cell lung Cancer. N Engl J Virgen Macarena (Sevilla, Spain), Duke Cancer Institute (Durham, North Carolina), Med. 2015;372:2018–28. https://doi.org/10.1056/NEJMoa1501824. Mission Health (Ashville, North Carolina), Meharry Medical College (Nashville, 4. Merck. Keytruda (pembrolizumab) [package insert]. Whitehouse Station, NJ: Tennessee), and Medical College of Wisconsin (Milwaukee, WI) upon IRB U.S. Food and Drug Administration; 2018. https://www.accessdata.fda.gov/ approval from each individual institution. OmniSeq studies based on scripts/cder/daf/index.cfm?event=overview.process&ApplNo=125514. de-identified data were considered non-human subjects research under 5. Genentech. Tecentriq (atezolizumab) [package insert]. South San IRB-approved protocol (BDR #080316) at Roswell Park Comprehensive Francisco, CA: U.S. Food and Drug Administration; 2018. https://www. Cancer Center (Buffalo, NY). accessdata.fda.gov/scripts/cder/daf/index.cfm?event=overview. process&ApplNo=761034. Consent for publication 6. Fehrenbacher L, Spira A, Ballinger M, Kowanetz M, Vansteenkiste J, Mazieres Not applicable. J, et al. Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 Competing interests randomised controlled trial. Lancet. 2016;387:1837–46. https://doi.org/10. MZ, PG, KD, KS, KGM, LJT, NS, DK, IAF, DM, JZ, ML, ACMK, JT, RS, LD, AS, SM, 1016/S0140-6736(16)00587-0. RJ, OB, and MS have no competing interests to disclose. SP, JMC, MKN, STG, 7. AstraZeneca. Imfinzi (durvalumab) [package insert]. Cambridge. England: U.S. APS, BB. JH, VG, JA, FLL, YW, MG, and CM are employees of OmniSeq, Inc. Food and Drug Administration; 2018. https://www.accessdata.fda.gov/ (Buffalo, NY) and hold restricted stock in OmniSeq, Inc. GD, EY, AE, HC, WB, scripts/cder/daf/index.cfm?event=overview.process&ApplNo=761069 CM, STG, MSE, KO, and JC are employees of Roswell Park Comprehensive 8. Andrews A. Treating with Checkpoint Inhibitors-Figure $1 Million per Patient. Cancer Center (Buffalo, NY), which is the majority shareholder of OmniSeq, Am Heal Drug Benefits. 2015;8 Spec Issue:9. http://www.ncbi.nlm.nih.gov/ Inc.. LG provides remunerated consulting to OmniSeq, Inc., and is supported pubmed/26380599%5Cn http://www.pubmedcentral.nih.gov/articlerender. by a startup grant from the Department of Radiation Oncology at Weill fcgi?artid=PMC4570079. Cornell Medicine (New York, US) and by donations from Sotio a.s. (Prague, 9. Galluzzi L, Chan TA, Kroemer G, Wolchok JD, Lopez-Soto A. The hallmarks of Czech Republic), Phosplatin (New York, NY, US), and the Luke Heller TECPR2 successful anticancer immunotherapy. Sci Transl Med. 2018; in press. Foundation (Boston, MA, US). JC is supported by grants from Bristol-Myeres 10. Morrison C, Pabla S, Conroy JM, Nesline MK, Glenn ST, Dressman D, et al. Squibb and Medpacto, is a site PI for Bristol-Myers Squibb, Genentech, Predicting response to checkpoint inhibitors in melanoma beyond PD-L1 Spectrum, Adaptimmune, Medpacto, Bayer, AbbVie, and Moderna, and has and mutational burden. J Immunother Cancer. 2018;6:32. https://doi.org/10. spoken on behalf of Merck and Guardant. TZ has received consulting fees 1186/s40425-018-0344-8. from Genentech Roche, Sanofi-Aventis, Astra Zeneca, Bayer, Janssen, Pfizer, 11. Nishino M, Ramaiya NH, Hatabu H, Hodi FS. Monitoring immune-checkpoint Foundation Medicine, Pharmacyclics, Bristol Myers Squibb, and MAA blockade: response evaluation and biomarker development. Nat Rev Clin Laboratories, promotional service fees from Genentech Roche and Exelixis, Oncol. 2017;14:655–68. https://doi.org/10.1038/nrclinonc.2017.88. contracted research with Janssen, Pfizer, OmniSeq, PGDx, Novartis, 12. Sul J, Blumenthal GM, Jiang X, He K, Keegan P, Pazdur R. FDA approval Merrimack, Abbvie/StemCentrx, Acerta, Merck, and Regeneron, and summary: Pembrolizumab for the treatment of patients with metastatic non- ownership interest in Capio Biosciences. RK has research funding from Incyte, small cell lung Cancer whose tumors express programmed death-ligand 1. Genentech, Merck Serono, Pfizer, Sequenom, Foundation Medicine, Guardant Oncologist. 2016;21:643–50. https://doi.org/10.1634/theoncologist.2015-0498. Health,Grifols, and Konica Minolta, as well as consultant fees from LOXO, 13. Conroy JM, Pabla S, Glenn ST, Burgher B, Nesline M, Papanicolau-Sengos A, X-Biotech, Actuate Therapeutics, Roche and NeoMed, and receives speaker et al. Analytical validation of a next-generation sequencing assay to monitor fees from Roche, and has equity in IDbyDNA, and CureMatch, Inc.. BG is a immune responses in solid tumors. J Mol Diagnostics. 2018;20:95–109. consultant for Celgene, Cook Medical, Merrimack, Foundation Medicine, https://doi.org/10.1016/j.jmoldx.2017.10.001. Ipsen, Bristol Myers Squibb, Exelixis, Terumo Interventional Systems, and 14. Dako. PD-L1 IHC 22C3 pharmDx: Non-Small Cell Lung Cancer [interpretation Taiho Oncology. manual]. Santa Clara, CA: Dako; 2015. https://www.accessdata.fda.gov/cdrh_ docs/pdf15/P150013c.pdf. 15. Büttner R, Gosney JR, Skov BG, Adam J, Motoi N, Bloom KJ, et al. Publisher’sNote Programmed death-ligand 1 immunohistochemistry testing: a review of Springer Nature remains neutral with regard to jurisdictional claims in analytical assays and clinical implementation in non–small-cell lung Cancer. published maps and institutional affiliations. J Clin Oncol. 2017;35:3867–76. https://doi.org/10.1200/JCO.2017.74.7642. Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 13 of 13 16. Paluch BE, Glenn ST, Conroy JM, Papanicolau-Sengos A, Bshara W, Omilian AR, et al. Robust detection of immune transcripts in FFPE samples using targeted RNA sequencing. Oncotarget. 2017;8:3197–205. https://doi.org/10. 18632/oncotarget.13691. 17. Kriegsmann M, Warth A. What is better/reliable, mitosis counting or Ki67/ MIB1 staining? Transl Lung Cancer Res. 2016;5:543–6. https://doi.org/10. 21037/tlcr.2016.10.11. 18. Shi W, Hu J, Zhu S, Shen X, Zhang X, Yang C, et al. Expression of MTA2 and Ki- 67 in hepatocellular carcinoma and their correlation with prognosis. Int J Clin Exp Pathol. 2015;8:13083–9 http://www.ncbi.nlm.nih.gov/pubmed/26722504. 19. Pan H, Gray R, Braybrooke J, Davies C, Taylor C, McGale P, et al. 20-year risks of breast-Cancer recurrence after stopping endocrine therapy at 5 years. N Engl J Med. 2017;377:1836–46. https://doi.org/10.1056/NEJMoa1701830. 20. Briest F, Wang Y, Arsenic R, Elezkurtaj S, Berg E, Greshake S, et al. Immunohistochemical study of mitosis-regulatory proteins in Gastroenteropancreatic neuroendocrine neoplasms. Anticancer Res. 2018;38:3863–70. https://doi.org/10.21873/anticanres.12670. 21. Jakobsen JN, Sørensen JB. Clinical impact of ki-67 labeling index in non- small cell lung cancer. Lung Cancer. 2013;79:1–7. https://doi.org/10.1016/j. lungcan.2012.10.008. 22. Ramaker RC, Lasseigne BN, Hardigan AA, Palacio L, Gunther DS, Myers RM, et al. RNA sequencing-based cell proliferation analysis across 19 cancers identifies a subset of proliferation-informative cancers with a common survival signature. Oncotarget. 2017;8:38668–81. https://doi.org/10.18632/ oncotarget.16961. 23. Thorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Ou Yang T-H, et al. The immune landscape of Cancer. Immunity. 2018;48:812–830.e14. https:// doi.org/10.1016/j.immuni.2018.03.023. 24. Gettinger SN, Choi J, Mani N, Sanmamed MF, Datar I, Sowell R, et al. A dormant TIL phenotype defines non-small cell lung carcinomas sensitive to immune checkpoint blockers. Nat Commun. 2018;9:3196. https://doi.org/10. 1038/s41467-018-05032-8. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal for ImmunoTherapy of Cancer Springer Journals

Loading next page...
 
/lp/springer-journals/proliferative-potential-and-resistance-to-immune-checkpoint-blockade-9HRwb8JQoN
Publisher
Springer Journals
Copyright
Copyright © 2019 by The Author(s).
Subject
Medicine & Public Health; Oncology; Immunology
eISSN
2051-1426
DOI
10.1186/s40425-019-0506-3
Publisher site
See Article on Publisher Site

Abstract

Background: Resistance to immune checkpoint inhibitors (ICIs) has been linked to local immunosuppression independent of major ICI targets (e.g., PD-1). Clinical experience with response prediction based on PD-L1 expression suggests that other factors influence sensitivity to ICIs in non-small cell lung cancer (NSCLC) patients. Methods: Tumor specimens from 120 NSCLC patients from 10 institutions were evaluated for PD-L1 expression by immunohistochemistry, and global proliferative profile by targeted RNA-seq. (Continued on next page) * Correspondence: Carl.Morrison@OmniSeq.com; Carl.Morrison@Roswellpark.org OmniSeq, Inc., 700 Ellicott Street, Buffalo, NY 14203, USA Roswell Park Comprehensive Cancer Center, Elm and Carlton Streets, Buffalo, NY 14206, USA Full list of author information is available at the end of the article © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 2 of 13 (Continued from previous page) Results: Cell proliferation, derived from the mean expression of 10 proliferation-associated genes (namely BUB1, CCNB2, CDK1, CDKN3, FOXM1, KIAA0101, MAD2L1, MELK, MKI67, and TOP2A), was identified as a marker of response to ICIs in NSCLC. Poorly, moderately, and highly proliferative tumors were somewhat equally represented in NSCLC, with tumors with the highest PD-L1 expression being more frequently moderately proliferative as compared to lesser levels of PD-L1 expression. Proliferation status had an impact on survival in patients with both PD-L1 positive and negative tumors. There was a significant survival advantage for moderately proliferative tumors compared to their combined highly/poorly counterparts (p = 0.021). Moderately proliferative PD-L1 positive tumors had a median survival of 14.6 months that was almost twice that of PD-L1 negative highly/poorly proliferative at 7.6 months (p =0.028). Median survival in moderately proliferative PD-L1 negative tumors at 12.6 months was comparable to that of highly/ poorly proliferative PD-L1 positive tumors at 11.5 months, but in both instances less than that of moderately proliferative PD-L1 positive tumors. Similar to survival, proliferation status has impact on disease control (DC) in patients with both PD-L1 positive and negative tumors. Patients with moderately versus those with poorly or highly proliferative tumors have a superior DC rate when combined with any classification schema used to score PD-L1 as a positive result (i.e., TPS ≥ 50% or ≥ 1%), and best displayed by a DC rate for moderately proliferative tumors of no less than 40% for any classification of PD-L1 as a negative result. While there is an over representation of moderately proliferative tumors as PD-L1 expression increases this does not account for the improved survival or higher disease control rates seen in PD-L1 negative tumors. Conclusions: Cell proliferation is potentially a new biomarker of response to ICIs in NSCLC and is applicable to PD-L1 negative tumors. Keywords: Atezolizumab, Nivolumab, Pembrolizumab, Ipilimumab, PD-1 Background diagnostic for pembrolizumab treatment, PD-L1 expres- On March 4th 2015, nivolumab (Opdivo®, from Bristol- sion levels assessed by the PD-L1 22C3 pharmDx assay Myers Squibb) became the first immune checkpoint (from Agilent) [12]. However, response prediction based inhibitor (ICI) to be approved by the US Food and Drug on PD-L1 levels is not 100% accurate. For instance, Administration for use in patients with metastatic pembrolizumab monotherapy in NSCLC patients with a nonsquamous non-small cell lung cancer (NSCLC) PD-L1 tumor proportion score (TPS) < 1% (i.e., membran- progressing on or after platinum-based chemotherapy [1], ous PD-L1 expression on < 1% malignant cells), of 1–49%, following disclosure of the results from the Phase III and ≥ 50% was associated with response rates of 10.7, 16.5, Checkmate-037 trial [2]. Since then, three other ICIs that and 45.2%, respectively [3]. Thus, a small population of inhibit the programmed cell death pathway, including NSCLC patients with low PD-L1, seemingly “negative bio- programmed cell death 1 (PDCD1 or CD279, best known marker” patients, will still respond to ICI-based therapy. as PD-1) and its ligands – CD274 (best known as PD-L1) Conversely, not all patients with high PD-L1 TPS obtain and programmed cell death 1 ligand 2 (PDCD1LG2 or clinical benefits from ICIs, which suggests the existence of CD273, best known as PD-L2) – have been licensed for alternative resistance mechanisms, such as mutations that use in NSCLC patients, namely pembrolizumab affect the ability of cancer cells to be recognized or elimi- (Keytruda®, from Merck) [3, 4], atezolizumab (Tecentriq®, nated by the immune system [9], or other mechanism of from Genentech) [5, 6], and durvalumab (Imfinzi®, from local immunosuppression in the tumor microenvironment AstraZeneca) [7]. Response rates to these ICIs employed via pathways that do not directly involve ICI targets such as single agent immunotherapeutic interventions in an as PD-L1 and PD-1 [3]. unselected population, however, is generally below 20% We employed targeted RNA sequencing of an immune [3]. Moreover, ICI-based immunotherapy has been related panel of slightly less than 400 genes to optimize estimated to cost 100,000–250,000 USD per patient (with the detection of low expressing genes as opposed to some variation depending on specific ICI, treatment whole transcriptome, that was specifically designed for regimen and duration) [8]. Thus, considerable efforts are use in formalin fixed paraffin embedded (FFPE) clinical being devoted to the elucidation of the mechanisms samples [13]. This list of genes was divided into 41 controlling the development of primary and acquired different immune function categories and analyzed for resistance to ICIs [9], as well as to the identification of response to ICIs in a cohort of NSCLC patients from biomarkers with robust predictive value [10, 11]. ten different institutions. The highest association with These observations have rapidly been translated into the response among the different immune function categories clinical management of NSCLC with the FDA companion was cell proliferation, represented by the expression of ten Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 3 of 13 unique genes. We demonstrate that either extreme of within 90 days of first dose were excluded as it could not cellular proliferation in the tumor microenvironment, i.e. be discerned whether they were rapid progressors or had highly or poorly proliferative, is associated with resistance poor performance prior to going on drug. Patients to ICIs amongst NSCLC patients, and that the expression lacking sufficient follow up time for response evaluation levels of a 10-gene set associated with cellular proliferation (less than 90 days from first dose), were also excluded can be harnessed to improve patient stratification beyond from analysis. Of the120 ICI-treated patients, for all of PD-L1 TPS. Most importantly, we show that additional which survival data was available, there were 10 patients stratification of PD-L1 negative NSCLC based upon cell not evaluable for response due to either no measurable proliferation status introduces a new approach to response disease or target lesion (n = 4), missing scans (n = 4), or to ICI therapy in NSCLC. not specified (n = 2) (Fig. 1a). For the remaining 110 pa- tients all were evaluable for response based on RECIST Methods v1.1 and were divided into a test set (n = 34) from one Patients and clinical data institution with the most patients (Duke) and a training Ten collaborating institutions obtained approval by their set (n = 76) from all other institutions. Patients whose respective institutional review boards (IRBs) to submit best response was complete response (CR), partial existing de-identified specimens and associated clinical response (PR), or stable disease (SD) with 12 months or data for use in this study. A total of 120 patients were more survival were classified as disease control (DC), included in the study (Fig. 1a), based on the following while patients whose best response was progressive criteria: (1) history of Stage IV NSCLC; (2) availability of disease (PD) or SD with less than 12 months survival adequate archival formalin-fixed paraffin-embedded were classified as no disease control (NDC). Duration of (FFPE) tissue collected prior to treatment with ICIs; (3) response was not available for all patients and not availability of sequencing data; and (4) availability of included for final analysis. demographic, diagnosis, follow-up and survival data. Table 1 summarizes the baseline clinical characteristics Immunohistochemical studies of these patients (individual patient data provided in The expression of PD-L1 on the surface of cancer cells Additional file 1: Table S1). was assessed in all cases by means of the Dako Omnis Patients who were treated with ICIs were included if Platform and the 22C3 pharmDx antibody (Agilent, they were treated by an agent approved by the FDA as Santa Clara, CA) using FDA-scoring guidelines [14]. of November 2017 and had follow up and survival from Briefly, a minimum of 100 viable tumor cells were first ICI dose (n = 120). ICI-treated patients who died assessed for membranous staining of any intensity for Fig. 1 Summary of patient disposition and exploratory analysis. a) A total of 120 patients previously treated with checkpoint inhibitors were included in the study. All patients had survival data from date of first dose of checkpoint inhibitor, while 110 were evaluable by RECIST v1.1 for response. b) Exploratory analysis using pair-wise proportion test of 41 immune-related gene functions derived from 394 genes for patients with disease control versus no disease control identifies cell proliferation as a biomarker of interest Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 4 of 13 Table 1 Patient characteristics Data analysis Immune gene expression ranks (range 0–100) from a Patients (n = 120) targeted RNA-seq immune panel of approximately 400 Age at initial diagnosis (years) genes were divided into 41 biological function categories < 30 1 (00.0) according to commercial annotations from the manufac- 30–39 1 (00.0) turer (Additional file 1: Table S2). For all 110 cases with 40–49 4 (03.3) response, distribution of each biological function was 50–59 28 (23.3) split into 3 tertiles of low (less than 33), medium 60–69 43 (35.8) (between 33 and 66) and high (greater than 66). Next, we performed a pair wise proportion test (chi-square 70–79 34 (28.3) test) to test for difference in DC rates for these three ≥ 80 9 (07.5) tertiles (i.e. low vs medium, medium vs high and low vs Mean 65 high) for each biological function (Fig. 1b). Proportion Sex test was performed with continuity correction and pair- Female 61 (50.8) wise p-values for each biological function were adjusted Male 59 (49.2) for multiple hypothesis testing using “holmes” correc- tion. We further divided the dataset into a training set Race (n = 76) consisting of samples from all data access White 96 (80.0) groups except the largest contributor. A separate test set Other 17 (14.2) (n = 34) was constituted from samples from a single Unknown 7 (05.8) largest contributing institute. Any biological function Vital status at last follow up that did not have cases representing one or more tertiles Alive 60 (50.0) was removed from further analysis due to lack of dynamic range of that biological function in the popula- Dead 60 (50.0) tion assessed in this study. The most significant gene Checkpoint inhibitor functions were utilized for further analysis. Survival atezolizumab 2 (01.7) analysis was performed using a log-rank test on 5-year ipilimumab + nivolumab 2 (01.7) Kaplan-Meier survival curves for PD-L1 levels assessed nivolumab 79 (65.8) by IHC and combined expression of 10 proliferation-re- pembrolizumab 37 (30.8) lated genes assessed by RNA-Seq. Comparison of DC rate was performed using Chi-square test with Yate’s the 22C3 antibody. The percentage of viable tumor cells continuity correction. Multivariate analysis was per- showing partial or complete membrane staining relative formed by fitting a binomial logistic regression model to to all viable tumor cells present in the sample (positive DC labels and co-variates such as proliferation status, and negative) was then used to derive a tumor proportion PD-L1 status, histology, race, sex, and age category. score (TPS). PD-L1 levels were scored by a board-certified Analysis of variance (ANOVA) was performed on the anatomic pathologist as per published guidelines [15], fitted model to study the table of deviance to determine with a TPS ≥ 50% considered as a strongly positive result the co-variate that explains the most variance in the for different comparisons, while a result of ≥1% consid- DC rates. ered as positive result for different comparisons. PD-L1 TPS ≥ 1% to < 50% were considered weakly positive for Results additional comparative purposes. PD-L1 TPS < 1% was Immune-related gene functions considered as negative. Ki-67 positivity amongst neoplas- Among 41 different immune-related gene functions tic and immune cells was scored upon nuclear staining, (Additional file 1: Table S2) evaluated by pairwise regardless of intensity, with the M7240 (clone MIB1) anti- comparison test in the training set (n = 76), three were body from Dako (Carpentaria, CA) with the percentage of significantly differentially expressed for DC versus NDC each cell type recorded. for at least one comparison (Additional file 1: Table S3). These three functions and specific genes (see Additional RNA-seq file 1: Table S2 for full gene names) included proliferation RNA were extracted from each sample and processed [BUB1, CCNB2, CDK1, CDKN3, FOXM1, KIAA0101, for targeted RNA-seq, as previously described [13, 16]. MAD2L1, MELK, MKI67 (better known as Ki-67), and Gene expression was evaluated by amplicon sequencing of TOP2A; maximum p = 0.0092], antigen processing (CD74, 394 immune transcripts on samples that met validated HLA-A, HLA-B, HLA-C, HLA-DMA, HLA-DMB, quality control (QC) thresholds [13]. HLA-DOA, HLA-DOB, HLA-DPA1, HLA-DPB1, Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 5 of 13 HLA-DQA1, HLA-DQA2, HLA-DQB2, HLA-DRA, (adenocarcinoma n = 94, sarcomatoid carcinoma n =1, HLA-DRB1, HLA-E, HLA-F, HLA-F-AS1, HLA-G; p = squamous cell carcinoma n = 25) was used as the pri- 0.0796), and dendritic cell (HERC6, IL3RA, ITGAX, NRP1, mary indicator for the proliferative status of the tumor TLR3, ZBTB46; p = 0.0903). When both the training and microenvironment. Tumors were stratified into poorly, test set (n = 110) were used for the same comparison moderately and highly proliferative based on the tertile (Additional file 1: Table S4), proliferation was the only of rank of expression of this gene signature as compared to these three functions that was significant (Fig. 1b). Results a separate reference population of 167 patients with for the test set (n = 34) did not identify proliferation, anti- multiple tumor types (Additional file 1: Table S6) [10]. gen processing, or dendritic cell as significant (Additional Based on this analysis, poorly proliferative tumors were file 1: Table S5), presumably due to the small size of the the least frequent in all available samples tested (27/120; sample set. Proliferation was chosen for further evaluation 22.5%), followed by an equal distribution of highly (47/ based upon the identification as a significant factor in the 120; 39.2%) and moderately proliferative tumors (46/120; training set as well as the combination of the training and 38.3%), (Fig. 2a). test set. To define whether neoplastic cells, immune cells, or both constituted the source of proliferation-related Proliferative status transcripts, 7 highly proliferative and 9 poorly prolifera- NSCLC had a wide distribution of poorly, moderately, tive cases were evaluated by immunohistochemistry for and highly proliferative tumors with input by both the expression of MKI67 (best known as Ki-67), a neoplastic and immune cells that can be measured in biomarker of proliferation largely employed in the clinics more than one way. The mean expression rank values of [17]. Highly proliferative tumors (as defined by 10 proliferation-related genes in 120 NSCLC specimens RNA-seq) had > 50% of neoplastic cells staining positive Fig. 2 Results for cell proliferation as an independent biomarker. a) Proportion of 120 NSCLC patients for cell proliferation by tertiles of poorly, moderately, and highly proliferative. b) Proportion of 120 NSCLC patients positive or negative for PD-L1 IHC using a cut-off of tumor proportion score of ≥50% as a positive result. c) Proportion of 120 NSCLC patients positive or negative for PD-L1 IHC using a cut-off of tumor proportion score of ≥1% as a positive result. d) Prevalence for all combinations of strongly positive PD-L1 (TPS ≥ 50%) cases and proliferation status. e) Prevalence for all combinations of PD-L1 and proliferation status for weakly positive PD-L1 cases (TPS ≥ 1 and < 50%). f) Prevalence for all PD-L1 negative (TPS < 1%) cases and proliferation status. Number and p values are reported Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 6 of 13 for Ki-67 in 6 out of 7 cases, while their poorly prolifera- 10 proliferation-related genes were evaluated for accur- tive counterparts contained less than 40% Ki-67 neo- acy (i.e. true positive plus true negatives divided by total plastic cells in 8 of 9 cases (Additional file 1: Table S7). number of results) for each gene individually (Additional In a similar fashion, highly proliferative tumors had 5% file 1: Table S6). Accuracy ranged from a low of 52.7% or more of immune cells staining positive for Ki-67 in for FOXM1 to a high of 67.3% for TOP2A, as compared all cases, while their poorly proliferative counterparts to a value of 71.8% for the mean expression rank values showed only two cases with this degree of reactivity. of all ten proliferation-related genes (Additional file 2: Importantly, an abundant tumor CD8 T-cell infiltrate Figure S1). The accuracy of Ki-67 at 59.1% was near the did not necessarily correlate with a highly proliferative mid-value of other single gene results. tumor microenvironment. For example, in one poorly The sum of all of these results suggest that poorly, mod- proliferative adenocarcinoma (Fig. 3a) there is a lack of erately, and highly proliferative tumors are somewhat staining by Ki-67 in both malignant and immune cells equally represented in NSCLC; that both immune cells (Fig. 3b), even though there is an abundance of CD8 T and malignant cells are sources of proliferation-related cells (Fig. 3c). In comparison, for a highly proliferative transcripts, and it is possible to reach similar results for adenocarcinoma (Fig. 3d) there is frequent staining by any of the 10 genes using only single gene evaluations. Ki-67 in both malignant and immune cells (Fig. 3e), with a similar number of CD8 T cells (Fig. 3f). PD-L1 expression To evaluate the impact of single gene proliferation Tumors with the highest PD-L1 expression were more results, e.g. Ki-67, the mean expression rank values of all frequently moderately proliferative as compared to lower Fig. 3 Immunohistochemical assessment of Ki-67 positivity and CD8 T cell infiltration. Representative fields for hematoxylin/eosin (a, d), CD8 positivity (b, e) and Ki-67 positivity (c, f) are depicted. The left hand panel (a-c) of a poorly proliferative tumor shows numerous CD8+ T-cells (c), while Ki-67 (b) stains very few neoplastic or immune cells. The right hand panel (d-f) of a highly proliferative tumor like the other case shows numerous CD8 T-cells (f), while Ki-67 (e) stains a high number of neoplastic and immune cells. Scale bar = 100 μm Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 7 of 13 levels of PD-L1 expression. PD-L1 TPS, defined as the or a combination of poorly and highly proliferative (p = percentage of neoplastic cells displaying membranous 0.02317), but less so for highly proliferative (p= 0.1188), positivity of any intensity upon staining with the DAKO (Fig. 2f). Overall these results support that as PD-L1 22C3 antibody, ranged from 0 to 100 and 32/120 expression increases there is an over representation of (26.7%) of all cases were strongly positive (Fig. 2b), while moderately proliferative tumors, but as shown below 56/120 (46.7%) of all cases were positive at any level of does not account for the improved survival or higher staining (Fig. 2c). Moderately proliferative tumors were disease control rates seen in PD-L1 negative tumors. slightly enriched for strongly positive PD-L1 tumors as compared to highly proliferative tumors (p= 0.4611), Overall survival and more so as compared to poorly proliferative tumors Proliferation status had an impact on survival in patients (p = 0.01237), or a combination of the latter two (p = with both PD-L1 positive and negative tumors. There 0.07227), (Fig. 2d). For weakly positive PD-L1 tumors, was a significant survival advantage for moderately moderately proliferative were not enriched as compared proliferative tumors compared to their combined highly/ to poorly proliferative counterparts (p= 1.0), highly poorly counterparts (p = 0.021) (Fig. 4a). When highly proliferative (p= 0.2463), or a combination of the latter and poorly proliferative groups were evaluated separately two (p = 0.5417), (Fig. 2e). For PD-L1 negative tumors, there was a trend towards survival for patients with moderately proliferative were under represented as com- moderately proliferative tumors (p = 0.064) (Fig. 4b). pared to poorly proliferative counterparts (p= 0.01955), Likewise, the survival of patients with strongly positive Fig. 4 Overall survival of 120 NSCLC patients receiving an immune checkpoint inhibitor (ICI) as part of their therapy. a) Overall survival based upon stratification by cell proliferation for moderately versus combined poorly/highly proliferative. b) Overall survival based upon stratification by cell proliferation for moderately versus poorly and highly proliferative. c) Overall survival based upon stratification by PD-L1 expression levels using TPS ≥ 50% as a cut-off for a positive result. d) Overall survival based upon stratification by strongly positive PD-L1 tumors and proliferation status (PD-L1 TPS ≥ 50% moderately proliferative, PD-L1 TPS ≥ 50% highly or poorly proliferative, PD-L1 TPS ≥ 50% moderately proliferative, PD-L1 TPS ≥ 50% highly or poorly proliferative). Number at risk and p-values are reported Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 8 of 13 PD-L1 tumors was associated with a statistically signifi- DC rate when combined with any classification schema cant survival advantage (p = 0.03) (Fig. 4c). A combin- used to scorePD-L1 as apositiveresult(i.e., TPS ≥ 50% or ≥ ation of proliferation and PD-L1 resulted in a significant 1%; see Additional file 4 for full results). The value of cell survival advantage in moderately proliferative strongly proliferation as a marker of response was best displayed by positive PD-L1 tumors with a median survival of 14.6 noting that the DC rate for moderately proliferative tumors months that was almost twice that of all less than was no less than 40% for any classification of PD-L1 as a strongly positive PD-L1 highly/poorly proliferative negative result. This was critically important for the tumors at 7.6 months (p = 0.028) (Fig. 4d). Likewise, fifty-seven negative PD-L1 negative tumors for which mod- median survival in less than strongly positive PD-L1 erately proliferative tumors had a DC rate of 41.2% (7/17) moderately proliferative tumors at 12.6 months was (Fig. 5g), while the DC rate among highly and poorly prolif- comparable to that of highly/poorly proliferative strongly erative tumors combined was 17.5% (7/40, p = 0.1179). The positive PD-L1 tumors at 11.5 months (p = 0.86) (Fig. 4d), summary of all of these results support that cell prolifera- but in both instances less than that of moderately prolifer- tion is a relevant biomarker in all groups of NSCLC, but is ative strongly positive PD-L1 tumors. The results for all unique and clinically useful for patients with PD-L1 PD-L1 positive tumors by a TPS ≥ 1% criteria were very negative tumors. Further support of this conclusion was a similar (Additional file 3: Figure S2). The summary of multivariate analysis on all co-variates using binomial these results support that moderately proliferative tumors logistic regression model showed that moderately pro- have a survival advantage beyond PD-L1 positive tumors liferative tumors to have a significant association with for NSCLC patients treated with checkpoint inhibitors. probability of disease control (Table 3; p = 0.0071). Furthermore, analysis of deviance of each co-variate Disease control rate (Table 3) suggests that adding proliferation to a null Similar to survival, proliferation status had an impact on model improved it significantly (p = 0.0009) followed disease control in patients with both PD-L1 positive and by a second most informative co-variate of PD-L1 negative tumors. The overall objective of evaluating status (p = 0.0337). Collectively these results suggest disease control was to show this intersection of response that, the proliferative status of the tumor microenvir- to checkpoint inhibition for cell proliferation versus the onment can be harnessed to improve patient stratifi- current standard of PD-L1 IHC. The results (Table 2, cation based on PD-L1 expression levels. Importantly, Fig. 5) show that patients with moderately versus those cell proliferation seems to have value as a biomarker with poorly or highly proliferative tumors have a superior of response in PD-L1 negative tumors. Table 2 Disease control for cell proliferation and PD-L1 IHC Cell Proliferation PD-L1 IHC DC NDC Total pts DC rate χ2 test Moderately 22 22 44 50.0% Highly 9 33 42 21.4% p = 0.0146 Poorly 4 20 24 16.7% p = 0.0113 Poorly/highly 13 53 66 19.7% p = 0.0017 Strongly positive (TPS ≥ 50%) 16 16 32 50.0% Not strongly positive (TPS < 50%) 19 59 78 24.4% p = 0.0009 Positive (TPS ≥ 1%) 21 32 53 39.6% Negative (TPS < 1%) 14 43 57 24.6% p = 0.1363 Moderate Strongly positive (TPS ≥ 50%) 10 7 17 58.8% Poorly/highly 6 9 15 40.0% p = 0.4786 Moderately Not strongly positive (TPS < 50%) 12 15 27 44.4% Highly 4 25 29 13.8% p = 0.0250 Poorly 3 19 22 13.6% p = 0.0438 Poorly/highly 7 44 51 13.7% p = 0.0063 Moderately cold tumors (CD8 rank < 15%) 7 10 17 41.2% Poorly/highly cold tumors (CD8 rank < 15%) 7 33 40 17.5% p = 0.1179 Moderately cold tumors (CD8 rank < 33%) 5 5 10 50.0% Poorly/highly cold tumors (CD8 rank < 33%) 0 11 11 0.0% p = 0.3298 Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 9 of 13 Fig. 5 Impact of PD-L1 levels and proliferative status on disease control rate in 110 NSCLC patients receiving an immune checkpoint inhibitor (ICI) as part of their therapy. a) Prevalence and DC rate for moderately versus highly and poorly proliferative tumors, as well as combined of the latter two. b) Prevalence and DC rate for strongly positive PD-L1 (TPS ≥ 50%). c) Prevalence and DC rate for PD-L1 negative (TPS < 1%). d) Prevalence and DC rate for strongly positive PD-L1 combined with moderately versus highly/poorly proliferative tumors. e) Prevalence and DC rate for PD-L1 positive (TPS ≥ 1%) combined with moderately versus highly/poorly proliferative tumors. f) Prevalence and DC rate for PD-L1 less than strongly positive (TPS < 50%) combined with moderately versus highly/poorly proliferative tumors. g) Prevalence and DC rate for PD-L1 negative (TPS < 1%) combined with moderately versus highly/poorly proliferative tumors. h) Prevalence and DC rate for weakly positive PD-L1 (TPS ≥ 1% and < 50%) combined with moderately versus highly/poorly proliferative tumors. i) Prevalence and DC rate for minimal tumor infiltration by CD8 T cells (so-called “cold” tumors) combined with moderately versus highly/poorly proliferative tumors Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 10 of 13 Table 3 Multivariate analysis Variable Estimate Std. Error z value p value (Intercept) 17.3526 2712.1561 0.006 0.9949 Proliferation Moderately 1.3503 0.5013 2.694 0.00707 PD.L1. status Positive 0.5169 0.547 0.945 0.34468 Histology SCC.or.Other −0.5898 0.6417 −0.919 0.35801 race Black or African American −34.8319 3301.0002 −0.011 0.99158 race Black or African American −16.7776 2712.156 −0.006 0.99506 race Other −35.6219 4796.5772 −0.007 0.99407 race Other −34.6736 4796.5771 −0.007 0.99423 race Unknown −18.5693 2712.1561 −0.007 0.99454 race White −17.9126 2712.1559 −0.007 0.99473 race White −18.3781 2712.1559 −0.007 0.99459 sex M 0.1522 0.5119 0.297 0.76616 age_cat 1–29 −35.0709 4796.5772 −0.007 0.99417 age_cat 40–49 −1.201 1.6759 −0.717 0.47359 age_cat 50–59 −0.6471 0.9129 −0.709 0.47843 age_cat 60–69 −0.9142 0.8863 −1.031 0.30233 age_cat 70–79 −1.1416 0.9098 −1.255 0.20955 Analysis of deviance of each co-variate Co-variate Df Deviance Resid. Df Resid. Dev P value(>Chi) NULL 109 137.61 Proliferation 1 11.1163 108 126.49 0.0008557 PD.L1.status 1 4.5112 107 121.98 0.0336733 Histology 1 0.0593 106 121.92 0.8076295 race 7 7.4867 99 114.44 0.3800195 sex 1 0.1064 98 114.33 0.7442778 age_cat 5 4.2582 93 110.07 0.5128654 Proliferative status and cold tumors for the more stringent cut-off value (Fig. 5i). Most Proliferation status had an impact on disease control in importantly, the DC rate was greater than 50% for any patients with factors other than PD-L1 positive or nega- grouping of moderately proliferative cold tumors, while tive status, impacting response to checkpoint inhibitors. the rate was less than 20% for poorly/highly proliferative In this regard, cell proliferation was further evaluated for counterparts. PD-L1 status did not associate with value beyond PD-L1 status in the emerging recognition response in cold tumors (Additional file 1: Table S8), of inflammatory status [16], and more specifically the again supporting that cell proliferation is a unique degree of CD8 infiltration. Response was evaluated for biomarker of response in NSCLC. tumors with reduced levels of CD8-coding transcripts as compared to a reference population of 167 patients with Discussion multiple tumor types, which we previously demonstrated Our findings suggest that a highly or poorly proliferative to indicate minimal tumor infiltration by CD8 T cells tumor microenvironment is associated with limited (so-called “cold” tumors) [16]. As there is no current ab- sensitivity to ICIs amongst NSCLC patients, and that solute criteria to define cold tumors we first arbitrarily targeted RNA-seq can be employed to assess the prolif- defined this group by a CD8 rank less than 15, and then erative status of the tumor microenvironment at diagno- compared to those results to an non-arbitrary cut-off of sis, with the ultimate goal of improving clinical decision the lower tertile of CD8 rank, or a value less than 33. making based on PD-L1 only. Most importantly, these Irrespective of the cut-off, DC was accurately predicted findings suggest that some highly or poorly proliferative by the proliferative status of the tumor microenviron- tumors may be resistant to ICIs independent of PD-L1 ment (Table 2), although the numbers are quite small or inflamed status and that both PD-L1 positive and Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 11 of 13 PD-L1 negative tumors at any TPS value can be strati- data from patients receiving PD-1- or PD-L1-targeting fied more accurately by cell proliferation. Moving for- agents (nivolumab, pembrolizumab atezolizumab), ward the need for standardization of cell proliferation CTLA4-targeting agents (ipilimumab), or both (nivolu- will be vitally important in comparing response among mab + ipilimumab) as it complicated subgroup analysis. various studies. In that regard the proliferative potential As a retrospective study across multiple institutions, there of malignant cells (assessed by Ki-67 positivity or were also limitations for data collection. Smoking status enumeration of mitotic figures) has been extensively was not available from all sites and as such was not a vari- employed over the past 3 decades for prognostic able in the multi-variate analysis. The exclusion of purposes in a number of tumors [18–21]. In our study, ICI-treated patients who died in less than 90 days post Ki-67 as measured by RNA-seq analysis was not the first dose checkpoint inhibitor did not allow for an most accurate predictor of disease control as a single analysis of this important group due to the lack of gene result, but rather was TOP2A. At such a formative collection ECOG performance score and our subsequent stage of development we did not evaluate proliferation inability to distinguish rapid progressors from poor health as a continuous variable for any single gene or the mean performance. rank of 10 genes, but this is factor that will need to eval- uated further in future studies. We also did not evaluate Conclusion K-67 or TOP2A IHC as a predictor of disease control In summary, we demonstrated that a poorly or highly and is another potential future study. proliferative potential in the tumor microenvironment is In a recent study, RNA-seq was employed to investi- associated with resistance to ICI-based immunotherapy gate the effect of proliferation on the survival of 6581 amongst NSCLC patients, and that assessing the expres- patients with 19 different cancers, as catalogued by The sion levels of ten proliferation-related genes by RNA-seq Cancer Genome Atlas (TCGA) [22]. In this setting, a in diagnostic biopsies stands out as a promising strategy low proliferation index was associated with improved for improving clinical decision making based on PD-L1 patient survival in 7 of 19 malignancies (including lung expression only. Additional studies are ongoing to test adenocarcinoma) which were subsequently defined as these observations in other tumor types commonly “proliferation-informative cancers” [22]. Most recently, treated with ICIs. another TCGA study evaluating the immune landscape of cancer in more than 10,000 tumors identified six im- Additional files mune subtypes hypothesized to define immune response patterns impacting prognosis [23]. Two of these six sub- Additional file 1: Table S1. Clinical characteristics. Table S2. RNA-seq gene function list. Table S3. Gene function analysis training set (Proportion types, C1 and C2, were noted for a high proliferation rate, test). Table S4. Gene function analysis training and test set combined with both having a substantial immune component but (Proportion test). Table S5. Gene function analysis test set (Proportion test). the least favorable outcomes. In this study tumor types Table S6. Accuracy for 10 proliferation immune-related genes. Table S7. Immunohistochemical assessment of Ki-67 positivity. Table S8. Disease over represented by C1 and C2 subtypes included bladder control rate in cold tumors by proliferation status. (XLSX 82 kb) cancer, breast cancer, cervical cancer, colon cancer, head Additional file 2: Figure S1. Gene specific proliferation values. (TIFF 274 kb) and neck squamous cell carcinoma, lung squamous cell Additional file 3: Figure S2. Disease control rates for PD-L1 positive carcinoma, mesothelioma, ovarian cancer, gastric adeno- (TPS > 1%) and negative tumors combined with cell proliferation status. carcinoma, and endometrial cancer. Moreover, in NSCLC, (TIFF 345 kb) a dormant tumor-infiltrating lymphocytes (TIL) signature Additional file 4: Supplementary tables with clinical annotations and data analysis results. (DOCX 32 kb) characterized by low activation (Granzyme B) and prolif- eration markers (Ki-67) in CD3 + TILs was also recently Abbreviations demonstrated to be associated with survival benefit in pa- CR: Complete Response; DC: Disease Control; FFPE: Formalin-Fixed Paraffin- tients treated with ICI [24]. These studies support that cell Embedded; ICI: Immune Checkpoint Inhibitor; IHC: Immunohistochemistry; proliferation should be evaluated further as an integral IRB: Institutional Review Board; NDC: No Disease Control; ORR: Objective Response Rate; OS: Overall Survival; PD: Progressive Disease; PR: Partial component of the immune response to ICIs and that re- Response; QC: Quality Control; RECIST: Response Evaluation Criteria In Solid sults may be tumor type dependent. Tumors; SD: Stable Disease; TCGA: The Cancer Genome Atlas; TPS: Tumor While our work was not based upon a single, Proportion Score well-structured clinical trial, samples were obtained from Acknowledgements 10 different institutions across the US and Europe, and re- The following individuals participated with the collection of biospecimens sults stood the test of such a heterogeneous, real-world and associated clinical data: Larson Hsu (Roswell Park Cancer Institute), Ryan clinical scenario. One of the major limitations of the Winters (Biosample Repository Facility at Fox Chase Cancer Center), Mary Shields and Ashley Gibbs (Northwest Oncology), Rosemary Makar and Amy present study is that response data (based on RECIST Fricke (Oregon Health & Science University Knight BioLibrary), Pearl v1.1) was available for a relatively small number of cases Abernathy (Mission Health System), and Stephanie Kaufman (acting as (110 patients), which obliged us to operate on pooled honest broker from OmniSeq, Inc). Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 12 of 13 Funding Author details 1 2 This research was funded by OmniSeq, Inc. (Buffalo, NY). OmniSeq, Inc., 700 Ellicott Street, Buffalo, NY 14203, USA. Roswell Park Comprehensive Cancer Center, Elm and Carlton Streets, Buffalo, NY 14206, 3 4 USA. Dartmouth-Hitchcock Medical Center, Lebanon, NH 03756, USA. Duke Availability of data and materials University, Durham, NC 27708, USA. Fox Chase Cancer Center, Philadelphia, The datasets generated and/or analyzed during the current study are not PA 19111, USA. Hospital Universitario Virgen Macarena, 41009 Sevilla, Spain. publicly available due to a non-provisional patent filing covering the 7 8 Medical College of Wisconsin, Milwaukee, WI 53226, USA. Meharry Medical methods used to analyze such datasets but are available from the College, Nashville, TN 37208, USA. Mission Health System, Asheville, NC corresponding author upon reasonable request. 10 11 28801, USA. Community Hospital, Munster, IN 46321, USA. Center for Personalized Cancer Therapy, Moores Cancer Center, La Jolla, CA 92093, USA. Authors’ contributions Department of Radiation Oncology, Weill Cornell Medical College, New GD, EY, AE, HC, WB, KM, KS, KD, LJT, DM, JZ, JC, ML, SM, TZ, MZ, PG, IAF, AS, York, NY 10065, USA. Sandra and Edward Meyer Cancer Center, New York, BG, ACM, JT, RS, RJ, LD, MS, OB, DP, and NS collected and submitted de- NY 10065, USA. Université Paris Descartes/Paris V, 75006 Paris, France. identified patient data with corresponding clinical cases from their respective institutions with IRB approval. JC, MN, DD, STG, CM, SP, and LG contributed Received: 14 November 2018 Accepted: 13 January 2019 to the experimental design of this analysis. CM, SP, JMC, MKN, STG, APS, BB, JA, VG, MQ, YW, FLL, LG, KR, ME, and MG prepared and analyzed patient datasets and corresponding clinical cases and were major contributors to writing and revising the manuscript. All authors read and approved the final References manuscript. 1. Vanpouille-Box C, Lhuillier C, Bezu L, Aranda F, Yamazaki T, Kepp O, et al. Trial watch: immune checkpoint blockers for cancer therapy. Oncoimmunology. 2017;6:e1373237. https://doi.org/10.1080/2162402X.2017.1373237. Ethics approval and consent to participate 2. Borghaei H, Paz-Ares L, Horn L, Spigel DR, Steins M, Ready NE, et al. Nivolumab Existing de-identified specimens and associated clinical data was submitted versus docetaxel in advanced nonsquamous non–small-cell lung Cancer. N to OmniSeq from Roswell Park Comprehensive Cancer Center (Buffalo, NY), Engl J Med. 2015;373:1627–39. https://doi.org/10.1056/NEJMoa1507643. Fox Chase Cancer Center (Philadelphia, PA), Dartmouth-Hitchcock Medical 3. Garon EB, Rizvi NA, Hui R, Leighl N, Balmanoukian AS, Eder JP, et al. Center (Lebanon, NH), Northwest Oncology (Munster, IN), Hospital Universitario Pembrolizumab for the treatment of non–small-cell lung Cancer. N Engl J Virgen Macarena (Sevilla, Spain), Duke Cancer Institute (Durham, North Carolina), Med. 2015;372:2018–28. https://doi.org/10.1056/NEJMoa1501824. Mission Health (Ashville, North Carolina), Meharry Medical College (Nashville, 4. Merck. Keytruda (pembrolizumab) [package insert]. Whitehouse Station, NJ: Tennessee), and Medical College of Wisconsin (Milwaukee, WI) upon IRB U.S. Food and Drug Administration; 2018. https://www.accessdata.fda.gov/ approval from each individual institution. OmniSeq studies based on scripts/cder/daf/index.cfm?event=overview.process&ApplNo=125514. de-identified data were considered non-human subjects research under 5. Genentech. Tecentriq (atezolizumab) [package insert]. South San IRB-approved protocol (BDR #080316) at Roswell Park Comprehensive Francisco, CA: U.S. Food and Drug Administration; 2018. https://www. Cancer Center (Buffalo, NY). accessdata.fda.gov/scripts/cder/daf/index.cfm?event=overview. process&ApplNo=761034. Consent for publication 6. Fehrenbacher L, Spira A, Ballinger M, Kowanetz M, Vansteenkiste J, Mazieres Not applicable. J, et al. Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 Competing interests randomised controlled trial. Lancet. 2016;387:1837–46. https://doi.org/10. MZ, PG, KD, KS, KGM, LJT, NS, DK, IAF, DM, JZ, ML, ACMK, JT, RS, LD, AS, SM, 1016/S0140-6736(16)00587-0. RJ, OB, and MS have no competing interests to disclose. SP, JMC, MKN, STG, 7. AstraZeneca. Imfinzi (durvalumab) [package insert]. Cambridge. England: U.S. APS, BB. JH, VG, JA, FLL, YW, MG, and CM are employees of OmniSeq, Inc. Food and Drug Administration; 2018. https://www.accessdata.fda.gov/ (Buffalo, NY) and hold restricted stock in OmniSeq, Inc. GD, EY, AE, HC, WB, scripts/cder/daf/index.cfm?event=overview.process&ApplNo=761069 CM, STG, MSE, KO, and JC are employees of Roswell Park Comprehensive 8. Andrews A. Treating with Checkpoint Inhibitors-Figure $1 Million per Patient. Cancer Center (Buffalo, NY), which is the majority shareholder of OmniSeq, Am Heal Drug Benefits. 2015;8 Spec Issue:9. http://www.ncbi.nlm.nih.gov/ Inc.. LG provides remunerated consulting to OmniSeq, Inc., and is supported pubmed/26380599%5Cn http://www.pubmedcentral.nih.gov/articlerender. by a startup grant from the Department of Radiation Oncology at Weill fcgi?artid=PMC4570079. Cornell Medicine (New York, US) and by donations from Sotio a.s. (Prague, 9. Galluzzi L, Chan TA, Kroemer G, Wolchok JD, Lopez-Soto A. The hallmarks of Czech Republic), Phosplatin (New York, NY, US), and the Luke Heller TECPR2 successful anticancer immunotherapy. Sci Transl Med. 2018; in press. Foundation (Boston, MA, US). JC is supported by grants from Bristol-Myeres 10. Morrison C, Pabla S, Conroy JM, Nesline MK, Glenn ST, Dressman D, et al. Squibb and Medpacto, is a site PI for Bristol-Myers Squibb, Genentech, Predicting response to checkpoint inhibitors in melanoma beyond PD-L1 Spectrum, Adaptimmune, Medpacto, Bayer, AbbVie, and Moderna, and has and mutational burden. J Immunother Cancer. 2018;6:32. https://doi.org/10. spoken on behalf of Merck and Guardant. TZ has received consulting fees 1186/s40425-018-0344-8. from Genentech Roche, Sanofi-Aventis, Astra Zeneca, Bayer, Janssen, Pfizer, 11. Nishino M, Ramaiya NH, Hatabu H, Hodi FS. Monitoring immune-checkpoint Foundation Medicine, Pharmacyclics, Bristol Myers Squibb, and MAA blockade: response evaluation and biomarker development. Nat Rev Clin Laboratories, promotional service fees from Genentech Roche and Exelixis, Oncol. 2017;14:655–68. https://doi.org/10.1038/nrclinonc.2017.88. contracted research with Janssen, Pfizer, OmniSeq, PGDx, Novartis, 12. Sul J, Blumenthal GM, Jiang X, He K, Keegan P, Pazdur R. FDA approval Merrimack, Abbvie/StemCentrx, Acerta, Merck, and Regeneron, and summary: Pembrolizumab for the treatment of patients with metastatic non- ownership interest in Capio Biosciences. RK has research funding from Incyte, small cell lung Cancer whose tumors express programmed death-ligand 1. Genentech, Merck Serono, Pfizer, Sequenom, Foundation Medicine, Guardant Oncologist. 2016;21:643–50. https://doi.org/10.1634/theoncologist.2015-0498. Health,Grifols, and Konica Minolta, as well as consultant fees from LOXO, 13. Conroy JM, Pabla S, Glenn ST, Burgher B, Nesline M, Papanicolau-Sengos A, X-Biotech, Actuate Therapeutics, Roche and NeoMed, and receives speaker et al. Analytical validation of a next-generation sequencing assay to monitor fees from Roche, and has equity in IDbyDNA, and CureMatch, Inc.. BG is a immune responses in solid tumors. J Mol Diagnostics. 2018;20:95–109. consultant for Celgene, Cook Medical, Merrimack, Foundation Medicine, https://doi.org/10.1016/j.jmoldx.2017.10.001. Ipsen, Bristol Myers Squibb, Exelixis, Terumo Interventional Systems, and 14. Dako. PD-L1 IHC 22C3 pharmDx: Non-Small Cell Lung Cancer [interpretation Taiho Oncology. manual]. Santa Clara, CA: Dako; 2015. https://www.accessdata.fda.gov/cdrh_ docs/pdf15/P150013c.pdf. 15. Büttner R, Gosney JR, Skov BG, Adam J, Motoi N, Bloom KJ, et al. Publisher’sNote Programmed death-ligand 1 immunohistochemistry testing: a review of Springer Nature remains neutral with regard to jurisdictional claims in analytical assays and clinical implementation in non–small-cell lung Cancer. published maps and institutional affiliations. J Clin Oncol. 2017;35:3867–76. https://doi.org/10.1200/JCO.2017.74.7642. Pabla et al. Journal for ImmunoTherapy of Cancer (2019) 7:27 Page 13 of 13 16. Paluch BE, Glenn ST, Conroy JM, Papanicolau-Sengos A, Bshara W, Omilian AR, et al. Robust detection of immune transcripts in FFPE samples using targeted RNA sequencing. Oncotarget. 2017;8:3197–205. https://doi.org/10. 18632/oncotarget.13691. 17. Kriegsmann M, Warth A. What is better/reliable, mitosis counting or Ki67/ MIB1 staining? Transl Lung Cancer Res. 2016;5:543–6. https://doi.org/10. 21037/tlcr.2016.10.11. 18. Shi W, Hu J, Zhu S, Shen X, Zhang X, Yang C, et al. Expression of MTA2 and Ki- 67 in hepatocellular carcinoma and their correlation with prognosis. Int J Clin Exp Pathol. 2015;8:13083–9 http://www.ncbi.nlm.nih.gov/pubmed/26722504. 19. Pan H, Gray R, Braybrooke J, Davies C, Taylor C, McGale P, et al. 20-year risks of breast-Cancer recurrence after stopping endocrine therapy at 5 years. N Engl J Med. 2017;377:1836–46. https://doi.org/10.1056/NEJMoa1701830. 20. Briest F, Wang Y, Arsenic R, Elezkurtaj S, Berg E, Greshake S, et al. Immunohistochemical study of mitosis-regulatory proteins in Gastroenteropancreatic neuroendocrine neoplasms. Anticancer Res. 2018;38:3863–70. https://doi.org/10.21873/anticanres.12670. 21. Jakobsen JN, Sørensen JB. Clinical impact of ki-67 labeling index in non- small cell lung cancer. Lung Cancer. 2013;79:1–7. https://doi.org/10.1016/j. lungcan.2012.10.008. 22. Ramaker RC, Lasseigne BN, Hardigan AA, Palacio L, Gunther DS, Myers RM, et al. RNA sequencing-based cell proliferation analysis across 19 cancers identifies a subset of proliferation-informative cancers with a common survival signature. Oncotarget. 2017;8:38668–81. https://doi.org/10.18632/ oncotarget.16961. 23. Thorsson V, Gibbs DL, Brown SD, Wolf D, Bortone DS, Ou Yang T-H, et al. The immune landscape of Cancer. Immunity. 2018;48:812–830.e14. https:// doi.org/10.1016/j.immuni.2018.03.023. 24. Gettinger SN, Choi J, Mani N, Sanmamed MF, Datar I, Sowell R, et al. A dormant TIL phenotype defines non-small cell lung carcinomas sensitive to immune checkpoint blockers. Nat Commun. 2018;9:3196. https://doi.org/10. 1038/s41467-018-05032-8.

Journal

Journal for ImmunoTherapy of CancerSpringer Journals

Published: Feb 1, 2019

References