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

Learn More →

DArTseq molecular markers associated with the piping leaf margin phenotype in pineapple (Ananas comosus L.)

DArTseq molecular markers associated with the piping leaf margin phenotype in pineapple (Ananas... This study sought to understand the genetic basis of the piping leaf margin phenotype in pineapple. To achieve this aim, a genome-wide association study (GWAS) using mixed linear regression and logistic regression analysis was conducted on three pineapple diversity panels including seedling populations segregating for spiny, spiny-tip and piping leaf margins. This study identified single nucleotide polymorphism (SNP) markers associated with the piping and spiny-tip leaf margin phe- notypes. A broad quantitative trait locus (QTL) positioned on chromosome 23 between positions 240,475 and 2,369,197 bp was the most highly associated with piping leaf margin in all analyses. Major candidate genes proposed are a Zinc finger protein 2, a Zinc finger protein 3, a WUSCHEL-related homeobox 2, a WUSCHEL-related homeobox 1 and a Zinc finger protein CONSTANS-like. Some other genes of a lower association, linked or nearby genes of interest, are also considered potentially involved to varying degrees. All candidate genes are known to be involved in aspects of stem cell maintenance, cell proliferation, epidermal cell differentiation, organogenesis, leaf polarity, cell wall modification or hormone signalling. It is possible each plays a role in either differentiation or morphological aspects of the spiny-tip and piping leaf margin phenotypes. It is expected the relative role of each associated gene might vary with genetic background. Keywords Spines · Piping leaf · ZFP · Abaxial · WUSCHEL · Zinc finger · W OX · Leaf margin Abbreviations MAF Minor allele frequency ABA Abscisic acid MMLM Multi-locus mixed linear model BGLR Bayesian Genetic Linear Regression PIC Polymorphism information content BLINK Bayesian-information and Linkage- PRI Pineapple Research Institute disequilibrium Iteratively Nested Keyway QTL Quantitative trait loci FarmCPU Fixed and Random Model Circulating Prob- SNP Single nucleotide polymorphism ability Unification SAM Shoot apical meristem DArT Diversity Arrays Technology GAPIT Genomic Association and Prediction Inte- Introduction grated Tool GMMAT Generalized linear Mixed Model Association Pineapple exhibits five leaf margin types and all, except Tests complete spiny, are considered the product of domestica- GWAS Genomic-wide Association Study tion (Coppens d’Eeckenbrugge and Sanewski 2011). Geno- iPat Intelligent Prediction and Association Tool types exhibiting this phenotype are typically found in the KNNi K-nearest neighbour imputation north-western regions of Brazil (Coppens d’Eeckenbrugge LD Linkage disequilibrium et al. 2018a), the country of origin of Ananas. The piping margin typically exhibits a thin white border along the leaf margins and is devoid of any spines (Fig. 1A and B). Leaves Communicated by Robert Henry with a spiny margin are considered the natural state as all * Garth M. Sanewski wild genotypes possess this phenotype and it is reasonable garth.sanewski@daf.qld.gov.au to assume piping represents a mutation selected at some 1 stage in the estimated 6,000 to 10,000 years of domestica- Queensland Department of Agriculture and Fisheries, tion of pineapple (Coppens d’Eeckenbrugge et al. 2018b). Maroochy Research Facility, Nambour, Australia Vol.:(0123456789) 1 3 234 Tropical Plant Biology (2022) 15:233–246 cases, only present as a shallow valley. There are however no published studies describing the botanical nature of pip- ing leaf margin. Leaf margin spines do not usually occur with piping suggesting an additional genetic mechanism. According to Collins (1960), the piping gene is epistatic to the spiny and spiny-tip alleles. He goes further to say, plants homozygous for piping have a more pronounced piping margin, suggest- ing a dosage effect and plants homozygous recessive for the piping alleles will display as either spiny or spiny-tip. The piping phenotype appears dominant over spiny and spiny-tip phenotypes. The piping leaf margin appears very stable and there are few reports of reversion to spiny, unlike the spiny- tip phenotype which frequently reverts to spiny and displays considerable variability in gene penetrance (Sanewski 2020). This study examined the botanical nature of piping leaf margin and identified SNP markers associated with spiny-tip and piping leaf margin phenotypes. The use of three diver- sity panels, independently analysed, was a novel approach that enabled testing between pairs of phenotype states; viz, piping and spiny, piping and spiny-tip and spiny and spiny- tip. Putative candidate genes associated with the piping leaf margin, all located on chromosome 23, are discussed. Methods In this study, the use of the term ‘piping’ is meant to encom- Fig. 1 A and B. The piping leaf margin typically exhibits a thin white pass smooth margins with or without folding and/or an obvi- border along the leaf margin (1A) and is devoid of spines even at the ous white edge. Plants were phenotyped by visually inspect- tip. On some leaves of some genotypes, it might also be characterised by an upward fold in the margin (1B) ing the leaf margins and carefully feeling the margin to the very tip for the presence of small spines. Those without any spines anywhere on the leaf were considered as piping, those The piping leaf margin does not appear to offer any obvious with small spines at the tip as spiny-tip, and plants with evolutionary advantage and hence its widespread existence spines completely over the margins were classed as spiny. in the western Amazon is likely all attributable to mankind. Because the trait exists in several different genotypes which Histology are considered to have arisen from clonal selection and sex- ual recombination (Chen et al. 2019), the piping phenotype 10 µm thick transverse leaf margin sections were prepared most likely has been selected multiple times. from cultivars, ‘Tapiricanga’ (piping) and ‘Smooth Cayenne’ The first obvious sign of the piping phenotype occurs (spiny-tip), using a freezing microtome to examine the dif- about 8 cm behind the tip and extends to the base of the leaf ferences between a piping margin and a margin with spine where it emerges from the meristematic zone. The demar- suppression but no piping. The sections were fixed in 4% cation or boundary between adaxial and abaxial ‘domains’ formalin and mounted in Mowiol solution for brightfield appears clear but a narrow strip of abaxial surface has been microscopy. folded onto the adaxial surface to present a new ‘false’ margin (Collins 1960). The piping strip will vary in width Panel Descriptions for GWAS in different genotypes from almost non-existent to a strip approximately 2 mm wide. The width can also vary on dif- Three diversity panels were used. Diversity panel one was a ferent leaves of the same plant or on different margins of sub-set of 90 plants from a population of seedlings exhibit- the same leaf possibly revealing an environmental effect on ing piping (33 plants) or completely spiny leaf margins (57 gene penetrance, meristematic cell development or leaf cell plants). This panel was intended to reveal the genetic dif- turgor. The fold can be exaggerated and obvious or, in some ference between spiny and piping margin. The population 1 3 Tropical Plant Biology (2022) 15:233–246 235 was developed using the parents that included spiny, spiny- authors of that work to correct for an error in the original tip and piping genotypes. Sixty-two of these plants were of assembly (Chen et al. 2019). These two reference genomes the parent combination ‘PRI 59–656’ (pollen) and ‘MD- should be identical for chromosomes two to 23 and hence 2’ (seed). It is considered that the piping margin trait in the different panel results are comparable where markers all progenies used was inherited from the piping margin of interest are positioned only on chromosomes two to 23. ‘Smooth Guatemala’ component and the spiny trait was Before analysis, the SNP data sets were filtered on repro- derived from ‘MD-2’. ducibility (> 0.99), call rate (> 0.75) and MAF (> 0.025) Panel two comprised 516 genotypes with 12 full sib fami- before imputation using the LD KNNi approach in TASSEL lies and some of their parental varieties. There were 440 5.2.44 (Bradbury et al. 2007; Money et al. 2015). The accu- plants with piping leaf margins and 76 with spiny-tip. The racy of imputation was tested by masking 10% of SNPs and population had all derived from varieties from the Pineap- comparing the imputed file with the original. ple Research Institute (PRI) of Hawaiian breeding program as for panel one. Panel two differed from panel one in that Principal Component Analysis (PCoA) it represented a larger number of recombinant events and spiny leaf margin plants were not included. This panel was Principal component covariates were generated in TASSEL intended to reveal the genetic difference between spiny-tip 5.2.54 (Bradbury et al. 2007). The number of covariates for and piping margin genomes. GWAS was determined based on either the Bayesian Infor- Diversity panel three was an amalgamation of panel mation Criterion (BIC) approach (Schwarz 1978) performed two seedlings (508 seedlings) and a reciprocal bi-parental in GAPIT, or the QQ and Scree plots. The smallest num- population (367 seedlings) (‘MD2’ × ‘Cayenne’, and ‘Cay- ber of covariates that produced a QQ plot that was neither enne’ × ‘MD2’) of seedlings segregating for spiny and spiny- inflated or under-inflated and had the smallest P values and tip margin. Panel three also included 76 parental genotypes was consistent with the Scree plot (visualized in TASSEL and several unrelated genotypes including wild accessions. but not reproduced) was used in the GWAS. Panel three was the most diverse of the populations used. There were 492 plants with piping margins, 351 with spiny- Association Analyses tip and 108 with spiny margins to give a total of 951 plants. The SNP marker data was analysed by genome-wide asso- Molecular Marker Development ciation study methods (GWAS) using mixed linear models (MLM) with structure correction. The multi-locus mixed Freeze-dried leaf base samples (white tissue) were supplied linear models (MMLM) of Segura et al. (2012), FarmCPU to Diversity Arrays Pty Ltd for DArTseq marker develop- (Liu et al. 2016) and BLINK (Huang et al. 2019) and the sin- ment. Marker development followed the procedure described gle loci ECMLM model, all in the GAPIT R package version in Sanewski et al. (2017). The methylation-sensitive restric- 3 (Zhang et al. 2010; Tang et al. 2016) were used. Kinship tion enzymes PstI and MseI were used to avoid highly repet- was estimated using the Van Raden method in GAPIT (Van itive regions that bias GWAS, and to reduce the genome Raden 2008). complexity. The marker sequences were trimmed to 69 bp. The phenotype data was binary and coded as 0 and 1 Marker sequences for panel one genotypes were refer- for panels one and two and 0, 1 and 2 for panel three. The enced against the revised genome of the spiny-tip pineapple Bonferroni-adjusted P value of P ≤ 0.01 was used as the level cultivar ‘Smooth Cayenne’ clone 153 (Acomosus_321_v3; of significance. https:// genom evolu tion. org/ CoGe/ Noteb ookVi ew. pl? nid= The R package Bayesian Genomic Linear Regression 937) (Ming et al. 2015). (BGLR) was used, through the graphic user interface, Subsequent to the sequencing of panel one genotypes, a Intelligent Prediction and Association Tool (iPat v1.3) revised version of this genome sequence became available (Chen and Zhang  2017) to test SNP markers in panels and was used for panels two and three. Marker sequences for one, two and three. The Bayes A model was used with a panel two and panel three genotypes were referenced against probit link to optimise analysis of binary or ordinal data the revised genome of the spiny-tip pineapple cultivar through logistic regression. The algorithm was run with ‘Smooth Cayenne’ clone 153 (Acomosus_321_v7.1, assem- 50,000 iterations and the first 3,000 discarded. The Van bly GCA_902162155.1 at European Nucleotide Archive; Raden kinship matrix and principal components were used https:// www. ebi. ac. uk/ ena/ brows er/ home). This version of to adjust for relatedness. The marker effects posterior val - the genome is slightly different from the previous version. ues were converted to absolute values and the Z-score Chromosomes 24 and 25 have been amalgamated as a larger calculated by dividing the absolute marker effect by its chromosome 25. Chromosome one was divided and the later standard deviation. The P value was then calculated by the portion re-labelled as chromosome 24. This was done by the following equation according to Campbell et al. (2019), 1 3 236 Tropical Plant Biology (2022) 15:233–246 P = 2*(1-pnorm(Z-Score)). A Bonferroni-adjusted P ≤ 0.05 Results was used as the cut-off for significance. A mixed model logistic regression was also performed Histology using the R program Generalized Linear Mixed Model Association Test (GMMAT v1.1.2) (Chen et al. 2016) for The section for spiny-tip cv. ‘Smooth Cayenne’ (Fig. S1) panels one and two data. This model uses a Bernoulli dis- shows the co-ordinated adaxial and abaxial layers. The tribution and logit function with covariate adjustment by abaxial layer is characterized by a corrugated epidermal principal components as a fixed effect and kinship (Van surface which appears slightly thinner than the adaxial sur- Raden) as a random effect to adjust for population struc- face. The transition between the adaxial and abaxial layers ture. Logistic models such as GMMAT that can include dissects the vascular bundles such that the phloem bundle fixed and random effects to account for population struc- and its sclerenchyma sheath are in the abaxial layer, and ture are considered more suitable for binary traits than the xylem bundle and its sclerenchyma sheath are in the linear mixed models especially if some populations have adaxial layer. Alternating between the phloem-xylem bun- more or less ‘affected’ status rather than ‘control’ sta- dles, but only in the abaxial layer, are aeration canals. Both tus. In such situations there is an increased risk of type the abaxial and adaxial layers have a line of fibre bundles I errors with mixed linear models as they assume error relatively close to the epidermal layer. Both the adaxial terms are normally distributed (Shenstone et al. 2018). and abaxial layers converge to the leaf margin. The model 10 algorithm in GMMAT was used for to per- The piping leaf margin section shows the abaxial piping form score tests on marker associations. The data were ‘bulge’ consists of one to two repeats of the phloem vas- treated as binomial. cular bundle and fibre strands in two different directions Logistic regression was also performed in gPLINK (Fig. S2A and B). The piping section therefore appears to using a logit function (Purcell et al. 2007) for panel one be one or two repeats of the abaxial layer at the margin only. sandwiched on top of each other before the terminating epidermal cell layer. Linkage Analysis Molecular Marker Datasets Linkage disequilibrium (LD) for the linkage decay curve was estimated for the entire genome for panel one data using the The original seedling SNP dataset for panel one comprised software package TASSEL 5.2.44 using a sliding window of 3,574 markers with 18% heterozygosity and an average 100 SNPs. The r of SNP marker pair frequencies and their minor allele frequency (MAF) of 0.23. The average PIC was distance was used to generate a linkage decay curve. A Loess 0.31, one ratio 0.47 and call rate 0.85. There were 13.5% curve was fitted using the package SigmaPlot 14. Linkage null alleles before imputation. Imputation accuracy was 91% disequilibrium decay distance was calculated as the point on and there were 2,141 SNP markers with an average MAF of the x axis corresponding to where the Loess curve passed 0.23 and 0.18 heterozygosity after imputation. 77 genotypes 50% of the maximum r (Noble et al. 2018). remained for association analyses. The original SNP dataset Marker linkages were also analysed for individual chro- for panel two comprised 516 genotypes and 40,686 markers mosomes for panel one SNP markers using Haploview and contained 4.8% null alleles. After imputation, with an 4.2 (Barrett et  al. 2005). In the case of Haploview, the accuracy of 88%, 516 genotypes and 16,393 markers remained Hardy–Weinberg p value cut-off was set at zero. with heterozygosity of 0.32, average MAF of 0.29 and average PIC of 0.18. After filtering for MAF ≤ 0.05 and imputation (accuracy 0.88), the panel three dataset included 16,393 SNP Nearby Genes markers for 952 genotypes and had an average MAF of 0.29 and heterozygosity of 0.33. Genes within the LD decay distance were identified using the annotated ‘Smooth Cayenne’ v3 genome for panel one Principal Component Analysis (PCoA) genotypes and v7 genome for panels two and three in the software package ‘Geneious R10’. This differential use of Based on the PCoA analysis, zero to four principal compo- genome versions was because these different genomes were nents (PCs) were used as covariates for panel one depending used as the reference for marker positioning in those marker on the algorithm, up to nine for panel two depending on the datasets. Likely candidate genes were selected based on algorithm and three for panel three. For panel one, the first proximity and supporting published reports of gene/protein four PCs accounted for 27% of the genotypic variance. For biological function. 1 3 Tropical Plant Biology (2022) 15:233–246 237 panel two, the first 9 PCs accounted for 37% and for panel homozygous for the SNP or heterozygous in piping geno- three, the first three PCs accounted for 18%. types. SNP marker #4713909 was homozygous for the SNP allele or heterozygous in spiny genotypes and homozygous Marker Density for the reference allele in piping genotypes. Several other markers on chromosomes one and two were also signifi- Density for SNP markers for chromosome 23 using the panel cantly associated but most were intergenic and not matched three dataset after filtering and imputation is shown in Fig.  2. to genes. Several other genes of interest were however posi- Figure 2 showed that marker density was high on chromo- tioned close to these markers. The Manhattan and QQ plots some 23 at the starting end up to mid-way around position for GWAS of panel one SNPs are shown in Figs. 3A and 4,504,232 bp. Chromosome 23 appears to be telocentric. 4A. Figure 5 shows the significantly associated SNP mark - ers by ECMLM and gPlink for the main QTL on chromo- Linkage Disequilibrium some 23. SNP marker #4714214, at position 798,190 bp on chromosome 23, was the most highly associated SNP LD decay distance for the entire genome using panel one marker by GMMAT and gPlink. SNP marker #4713909 SNPs indicated an effective decay distance of 1,222 kb for was the only significantly associated SNP marker using the entire genome (Fig. S3). This LD decay distance was Bayes A in BGLR (Table 1). This intergenic marker is posi- used as a guide for selecting candidate genes in all analyses. tioned on chromosome 23 at 762,438 bp and was close to For panel one, most of the significant markers were inter - the most significantly associated SNP marker (#4714214). genic and linkages were generally not strong between asso- Both #4714214 and #4713909 are positioned close to a ciated SNPs. WOX2 (Aco015463.1) and ZFP2 (Aco015474.1) (Table S1). Linkage data for the most highly associated SNP markers in Association Analyses panel one (spiny and piping margins) is shown in table S2. The highest associated SNP marker by BGLR, #4713909, Panel one comprised piping and spiny phenotypes only. The was close to the WOX2 and ZFP2 but also highly linked highest associated SNP marker (#4714214) was the most to #100067496 which was also close to the WOX2 and significantly associated in all panel one mixed linear model ZFP2. There were no other strong linkages with associated (MLM) analyses (Table  1). Genes close to this position SNP markers. include a Zinc-finger protein 2 (ZFP2), Zinc-finger protein Panel two comprised spiny-tip and piping margin plants. 3 (ZFP3), Ras-related protein Rab-8B, and WUSCHEL- SNP marker #54316030 was the most significantly associ - related homeobox 2 (WOX2) genes (Table S1). The WOX2 ated in all panel two GWAS analyses and logistic regres- and ZFP2 are considered the main candidates associated with sion by BGLR and GMMAT (Table  2). This marker was piping and spiny leaf phenotypes in this panel of genotypes. positioned on chromosome 23 at 1,068,662 bp, 9.3 kb from SNP marker #4714214 was homozygous for the reference a WUSCHEL-related homeobox 1 (WOX1) (Aco007021.1) allele (spiny-tip ‘Smooth Cayenne’) in spiny genotypes and and 31.6 kb from a Zinc finger CONSTANS-like protein Fig. 2 SNP marker density (number of markers within 1 Mb sliding window) for all 25 chromosomes, panel 3 after filtering for MAF < 0.025 and imputation by LD-KNNi 1 3 238 Tropical Plant Biology (2022) 15:233–246 Table 1 SNP markers SNP Position Length -log10p (Bonf) Maf Algorithm Allelic significantly associated with chrom: bp (bp) effect spiny and piping leaf margin phenotypes for panel one 4714214|F|0–52:G > A 23: 798,190 59 8.3 0.33 ECMLM pos genotypes using the GAPIT 15.2 MLMM algorithms ECMLM, MLMM, 12.7 FarmCPU FarmCPU and BLINK and 17.8 BLINK logistic regressions in GMMAT, 6.7* Plink gPLINK and BGLR. All 12.4 GMMAT markers are significant at the 4713909|F|0–63:G > A 23: 762,438 69 7.7 0.45 ECMLM neg Bonferroni-adjusted P < 0.01 4.3* Plink except the BGLR and gPLINK 2.0* BGLR values* which are significant 12.4 GMMAT at P < 0.05. 17 of the SNPs 4712886|F|0–20:A > G 23: 423,562 69 7.5 0.26 ECMLM pos with the lowest association 6.4* Plink significance, all on chromosome 11.0 GMMAT 23 or not positioned, have been omitted. The direction of the 100044752|F|0–50:T > C 2: 16,332,302 69 6.9 0.07 FarmCPU pos allelic effect for the ECMLM 4712964|F|0–31:C > T 23: 296,925 69 6.7 0.27 ECMLM pos algorithm is shown 6.1* Plink 11.1 GMMAT 100058551|F|0–63:A > G NP 69 6.7 0.50 FarmCPU pos 100012184|F|0–68:A > G 1: 24,514,356 69 6.3 0.47 FarmCPU pos 100067496|F|0–23:C > A 23: 678,197 69 6.2 0.44 ECMLM neg 5.3* Plink 11.0 GMMAT 4713774|F|0–5:T > C 23: 1,970,691 69 5.6* 0.37 Plink pos 7.9 GMMAT 4716438|F|0–19:C > T 23: 1,305,909 69 5.4* 0.34 Plink neg 9.3 GMMAT 4714616|F|0–49:A > G NP 69 5.3* 0.27 Plink pos 8.6 GMMAT 100058930|F|0–31:T > G 23: 1,473,662 69 5.3* 0.41 Plink neg 9.1 GMMAT 4724522|F|0–24:C > G NP 41 5.2* 0.48 Plink pos 7.4 GMMAT 4717335|F|0–57:T > C 23: 240,475 69 5.2* 0.44 Plink pos 9.4 GMMAT 4724236|F|0–29:G > A 23: 2,058,488 69 5.1* 0.36 Plink pos 7.0 GMMAT 100044395|F|0–7:G > C 23: 1,999,412 69 5.0* 0.47 Plink pos 7.5 GMMAT (Aco007020.1) (Table S3). These are considered the best cultivars. Other markers of relatively lower significance were candidates for the piping and spiny-tip genotype data set also demonstrated by some algorithms on chromosome 5, 6, studied. This marker was homozygous or heterozygous for 12, and 13 and un-positioned. The Manhattan and QQ plots the polymorphism in piping genotypes. Spiny-tip geno- for GWAS of panel two SNP data are shown in Figs.  3B types were homozygous for the reference allele (spiny-tip and 4B. cv. ‘Smooth Cayenne’). Marker #28878068 positioned at Panel three comprised spiny, spiny-tip and piping 671,978 bp on chromosome 23 was highly associated by margin phenotypes. Marker # 54316030 at position MLM only. This marker was unmatched but positioned 1,068,662 on chromosome 23 was the most significantly 6.9 kb from a Myb domain protein 97 (Aco015454.1), 28 kb associated in three MLM algorithms (Table  3). This from an O-fucosyltransferase family protein (Aco015458.1) marker was only 9.3 kb from a WOX1 (Aco007021.1) and 66.8  kb from a WUSCHEL-related homeobox  2 and 31.6 kb from a Zinc finger CONSTANS-like protein (WOX2) (Aco015463.1). Piping genotypes were homozy- (Aco007020.1) (Table S4). These are the same genes pro- gous for the polymorphism and null or homozygous for the posed for panel two. Marker # 4723993 on chromosome reference allele (spiny-tip cv. ‘Smooth Cayenne’) in spiny-tip six at 14,330,763 bp was also highly associated by three 1 3 Tropical Plant Biology (2022) 15:233–246 239 Fig. 3 A, B and C. Manhattan plots for SNP markers for panel one corrected -log10(Bonferroni-adjusted) P < 0.01 cut-off horizont al (3A) analysed in GAPIT using the ECMLM, MLMM, BLINK and lines are shown. Markers on scaffolds and not positioned are included FarmCPU algorithms for association with spiny and piping leaf mar- as fictitious chromosomes 26 and 27 respectively. Figures reproduced gin phenotypes, panel two markers (3B) analysed for spiny-tip and in the R package ‘Memory-efficient, Vizualization-enhanced, and piping leaf margin phenotypes and panel three markers (3C) analysed Parallel-accelerated tool for genome-wide association study’ (rMVP), for spiny, spiny-tip and piping leaf margin phenotypes. The structure-https:// github. com/ xiaol eiLiu Bio/ rMVP MLM algorithms. This was close (20.5 kb) to a Zeaxanthin homozygous for the reference allele in spiny-tip and spiny epoxidase (ZEP) as identified putatively associated with genotypes and homozygous for the SNP or heterozygous spiny-tip phenotypes in a previous study (Sanewski 2020). in piping genotypes. Bear in mind, the spiny-tip ‘Smooth Manhattan and QQ plots for panel three SNP data are Cayenne’ provided the reference genome. #54316030 shown in Figs.  3C and 4C. Marker #54316030 was therefore represents a dominant allele for the piping trait. 1 3 240 Tropical Plant Biology (2022) 15:233–246 Fig. 4 A, B and C. QQ plots from the panel one (4A), panel two (4B) and panel three (4C) SNP marker GWAS analyses. Figures reproduced in the R package rMVP, https:// github. com/ xiaol eiLiu Bio/ rMVP Marker #4723993 was homozygous for the polymor- cell wall structural protein and MATE eff lux family gene phism in spiny-tip genotypes and homozygous for the ref- (Table S4). erence allele in spiny genotypes. Piping genotypes carried any of the allele states. SNP marker #4723993 which was associated with the spiny-tip phenotype in panel three, Discussion was in various allele states in piping and spiny-tip geno- types in panel two. There appeared no difference between There do not appear to be any global disturbances in leaf polar- piping and spiny-tip genotypes with respect to the allele ity. The development of a piping margin appears consistent state of this marker for the spiny-tip phenotype. SNP with a cessation of the adaxial layer while the abaxial layer is marker #28883264 was the only marker significantly repeated an additional one to two times at the margin thus creat- associated by logistic regression (BGLR) (Table 3). It is ing a thicker composite abaxial layer, an imbalance and upward positioned on chromosome six at 14,526,582 bp close to folding. Piping leaf margin appears to be a lack of co-ordination several likely candidates including a Receptor kinase 2, between development of the abaxial and adaxial tissues at the Zeaxanthin epoxidase, Cellulose synthase, Glycine-rich leaf margin rather than changes in adaxial-abaxial patterning. 1 3 Tropical Plant Biology (2022) 15:233–246 241 Fig. 5 Significantly associated SNP markers on chromosome 23 for the panel one dataset using the ECMLM and gPLINK analyses. The -log10(P) line of significance (Bonferroni- adjusted) P < 0.01 is shown. The vertical arrows delineate the proposed main QTL Table 2 SNP markers SNP Posn -log10P (Bonf) Length Maf Algorithm Allelic significantly associated with chrom: bp bp effect spiny-tip and piping leaf margin phenotypes for panel 54316030|F|0–11:A > C 23: 1,068,662 34.6 29 0.46 MLMM neg two genotypes using the 28.6 Blink GAPIT algorithms ECMLM, 27.2 ECMLM MLMM, FarmCPU and 26.3 FarmCPU BLINK and logistic regression 5.5* BGLR in GMMAT and BGLR. All 27.1 GMMAT markers are significant at 28878068|F|0–65:G > A 23: 671,978 28.4 69 0.25 MLMM neg the Bonferroni-adjusted 21.9 Blink P < 0.01 except the BGLR 20.9 ECMLM value* which is significant at FarmCPU 18.0 P < 0.05. The direction of the 28878068|F|0–66:C > A 18.1 69 0.32 Blink pos allelic effect for the ECMLM 16.6 MLMM algorithm is shown. Seventy- 12.2 FarmCPU one markers were found 14.1 GMMAT significant but only the most significant markers for each 4713909|F|0–63:G > A 23: 762,438 20.3 69 0.36 ECMLM neg chromosome are shown 17.7 GMMAT 4710125|F|0–59:A > G 23: 701,343 18.6 69 0.45 ECMLM neg 21.5 GMMAT 28877781|F|0–34:C > G 23: 506,814 16.0 69 0.23 ECMLM neg 11.3 FarmCPU 6.7 Blink 10.6 GMMAT 54317320|F|0–18:T > A 5: 6,727,065 10.9 69 0.20 Blink neg 4727322|F|0–43:A > T 12: 5,152,594 7.1 69 0.36 FarmCPU neg 6.9 Blink 54316897|F|0–37:G > A 6: 9,696,245 9.7 69 0.14 ECMLM neg 28877802|F|0–5:T > G 13: 2,431,298 6.0 69 0.08 Blink pos 1 3 242 Tropical Plant Biology (2022) 15:233–246 Table 3 SNP markers SNP Position -log10p Length (bp) Maf Algorithm Allelic significantly associated with chromo: bp effect spiny, spiny-tip and piping leaf margin phenotypes for panel 54316030|F|0–11:A > C 23: 1,068,662 60.3 29 0.33 BLINK pos three genotypes using the 60.0 FarmCPU GAPIT algorithms MLMM, 24.2 MLMM FarmCPU and BLINK and 4723993|F|0–39:G > C 6: 14,330,763 56.0 69 0.47 FarmCPU neg logistic regression in BGLR. All 24.3 MLMM markers are significant 8.4 BLINK at the Bonferroni-adjusted 28883264|F|0–10:C > A 6: 14,526,582 50.2 69 0.20 BLINK neg P < 0.01 except the BGLR 40.0 FarmCPU value* which is significant 30.0 MLMM at P < 0.05. The direction of 7.4* BGLR the effect, either positive or negative for the ECMLM 4726743|F|0–6:A > T 6: 14,355,558 20.9 69 0.27 BLINK pos algorithm is shown, even 15.2 FarmCPU though this algorithm is not 12.3 MLMM otherwise represented in the 54314089|F|0–40:G > A 6: 14,334,819 23.7 69 0.47 BLINK neg table 6.7 FarmCPU 54314200|F|0–17:G > A 6: 13,977,148 18.5 69 0.33 BLINK neg 28881980|F|0–27:G > A 6: 13,967,023 14.0 69 0.35 FarmCPU neg 54314646|F|0–9:G > T 13: 10,965,585 11.7 69 0.46 BLINK neg 4716772|F|0–47:A > C 6: 14,556,931 11.2 69 0.15 MLMM pos 28881708|F|0–8:G > A NP 11.0 23 0.30 FarmCPU neg 6.4 BLINK 4714193|F|0–41:C > T 12: 5,141,097 11.0 69 0.29 BLINK neg 54309331|F|0–37:T > G 6: 14,525,401 10.8 69 0.36 BLINK neg 28876898|F|0–59:G > A 6: 13,991,970 9.9 69 0.47 MLMM pos 28874267|F|0–22:T > C 18: 10,110,153 9.7 69 0.06 BLINK pos 54312110|F|0–55:T > C 7: 12,303,848 9.7 69 0.18 BLINK neg 28882074|F|0–12:G > A 5: 10,126,663 9.5 69 0.08 BLINK neg 4716438|F|0–19:C > T 23: 1,305,909 8.9 69 0.23 BLINK neg 7.3 MLMM 54310912|F|0–23:A > G 12: 418,631 8.9 69 0.23 BLINK pos 4726657|F|0–31:T > G 12: 4,768,435 8.7 69 0.28 FarmCPU neg 54313753|F|0–50:G > A 7: 10,461,432 9.3 69 0.23 BLINK neg 6.6 FarmCPU 54316608|F|0–39:G > A 6: 13,991,970 9.3 69 0.31 BLINK pos 4710241|F|0–17:A > T 6: 14,732,936 8.8 69 0.35 BLINK neg 4709656|F|0–11:G > T 6: 14,242,751 8.6 69 0.18 FarmCPU pos 28882522|F|0–38:A > G 4: 1,202,604 8.1 58 0.09 FarmCPU neg 54315064|F|0–6:C > T NP 7.7 24 0.06 BLINK neg 4715867|F|0–68:A > T 25: 8,053,599 7.2 69 0.46 BLINK neg 4718200|F|0–6:G > A 6: 13,796,356 7.0 69 0.32 FarmCPU pos 28877469|F|0–16:G > C 2: 13,859,177 7.0 69 0.31 FarmCPU neg 4713341|F|0–56:G > A 5: 11,677,507 6.9 69 0.07 MLMM neg 28879929|F|0–25:C > T 25: 6,103,800 6.9 65 0.11 BLINK pos 4719816|F|0–30:C > G NP 6.8 69 0.46 BLINK neg 4717292|F|0–51:G > T 6: 14,468,303 6.7 62 0.16 BLINK pos 28883299|F|0–59:G > A 25: 8,864,898 6.7 69 0.27 FarmCPU pos 4715794|F|0–33:C > G NP 6.6 5.1 0.35 BLINK pos 28877973|F|0–6:T > C 4: 1,026,724 6.4 69 0.06 BLINK neg 1 3 Tropical Plant Biology (2022) 15:233–246 243 JAGGED, have also been shown to be involved in boundary Diversity Panel One maintenance and/ or cell proliferation. In both Arabidopsis and tomato, the KNUCKLES transcription factor indirectly Panel one comprised a population segregating for spiny and piping margin and as such is considered the definitive data represses WUSCHEL to allow a temporary inhibition of meristematic activity (Bollier et al. 2018). Ubiquitin E3 set for examining polymorphisms associated with the selec- tion of piping-leafed variants from the presumed original ligases are also ZFP family genes. Plant U-box E3 ligases have been implicated in plant development and morphol- spiny genotypes. The assumption at the commencement of this study is that the piping phenotype arose from the spiny ogy. As an example, a U-box E3 ligase mutation in barley has been associated with semi-dwarfing, wavy leaf margins phenotype. Discussion of candidate genes within this context is therefore based primarily on the findings from panel one. and erect growth (Braumann et al. 2018). Cullin4-based E3 ligases are somehow involved in aspects of plant mor- WUSCHEL‑Related Homeobox 2 (WOX2) phology given that mutation of the gene in Arabidopsis causes stunted growth and aberrant leaf shapes (Chen and The WOX2 (Aco015463.1) gene is 59 kb from the highest Hellmann 2013). associated SNP by all four MLM models and two logistic regressions and hence considered a highly likely principal Ras‑related Protein Rab‑8B causal variant, along with the ZFP2. There are ten WOX genes in pineapple, spread across six chromosomes with The Ras-related Rab-8B gene (Aco015473.1) is positioned only 4.0 kb from the highest associated SNP marker in panel three on chromosome 23 (Rahman et al. 2017). Research on WOX genes in other species supports a potential role one. Again, data in other species suggests a potential role in leaf margin phenotype in pineapple. The Rab GTPase fam- in leaf margin morphology. Rice plants carrying a muta- tion in a WOX3 called LSY1, exhibited curled leaf margins ily of proteins plays an important role in cell wall synthesis and modification through trafficking of cell wall polymers to suggesting greater cell growth or proliferation in the abax- ial leaf tissues compared with the adaxial side. In another specific target membranes. Cell wall synthesis and modifica- tion such as might occur during cell expansion, are integral example, a WUS ortholog in rice is expressed mostly in leaf margins (Somssich et al. 2016). Also, overexpression to plant development and possibly architecture (Speth et al. 2009; Lycell 2008). A study in Arabidopsis looked at the of a WOX3 in rice results in adaxial curling of leaf margins. Similarly, a WOX1 double mutant in rice exhibits reduced membrane where RabA localizes, and in conjunction with down-regulation manipulation, demonstrated a correspond- leaf blades and thickened leaf margins (Honda et al. 2018). WOX1 promotes cell proliferation (Tadege 2016) suggest- ing change in cell shape (Kirchhelle et al. 2016). Follow- ing on from this it was postulated by Rahni and Birnbaum ing a loss or reduced-function polymorphism would reduce cell proliferation. (2016) that Rab proteins might mediate local wall stiff- ness thus affecting cell shape and consequently tissue and Zinger Finger Proteins (ZFPs) organ morphology. This potential role in specific cell edge mechanics, rather than that of the entire cell wall, might The highest associated SNP marker was 1.8 kb from a ZFP2 have implications in pineapple piping leaf margin where a weakness in one edge might allow bending or folding, given (Aco015474.1). ZFPs form a large family of transcription factors with roles in plant development and stress response growth and turgor pressure in surrounding tissues, especially if cell proliferation is increased in some surrounding tissues. and appear conserved across plant evolution (Li et al. 2013). As in the case of WOX genes, ZFPs have also been impli- Although not studied here, it might also explain the position and direction of spines. cated in leaf morphology in other species. Superman (SUP), one of the better known ZFPs affecting morphology, sup- Diversity Panel Two presses cell division at a specific stage and is associated with flower organogenesis and plant morphogenesis (Jiang Panel two results represent a difference between spiny-tip et al. 2008). SUP overexpression in Nicotiana tabacum pro- duces a percentage of plants with curled-up leaf margins and and piping leaf margin phenotypes. It is known that the pip- ing margin is dominant over spiny-tip (Collins 1960). The thickened roots (Nibau et al. 2011). In plants with moderate expression of SUP and curly leaves, the organisation of the results here (table two, table S3) suggest the piping mar- gin plants might carry a polymorphism in either a WOX1, adaxial/ abaxial mesophyll layers was disrupted with more cells and cells of variable shape throughout. Overexpression WOX2 and/ or ZFP that the spiny-tip plants do not. There is no difference between the genotypes represented in panel of a ZFP in rice, OsZHD1, induced abaxial curling due to an increase in the number and abnormal arrangement of bul- two with respect to the spiny-tip polymorphisms puta- tively associated in a previous study (Sanewski 2020). This liform cells (Xu et al. 2014). Other ZFPs, KNUCKLES and 1 3 244 Tropical Plant Biology (2022) 15:233–246 suggests that both piping and spiny-tip genotypes in panel The most likely causal gene was a ZEP, a precursor to ABA, two carry the spiny-tip polymorphism. Because piping mar- although other genes including a MATE efflux were also gin is dominant over spiny-tip, piping margin genotypes as implicated. A much larger diversity panel used in the current derived in this diversity panel must have arisen from spiny- study comprised of many segregating populations and unre- tip genotypes, not spiny genotypes. lated genotypes, and using additional analytical methods, but identified the same broad QTL for the spiny-tip phenotype Diversity Panel Three thus validating the earlier study to some extent. No major SNP markers associated with a piping mar- Because panel three contains all three phenotypes, piping, gin were positioned on chromosome six near the spiny-tip spiny-tip and spiny, it is expected that the SNPs associated loci in the current study. The piping margin phenotype is with both piping margins and spiny-tip will be identified as a however known to suppress other leaf margin phenotypes. difference between them and/or spiny margins. Accordingly, This suggests the piping leaf genes function upstream of the two main QTLs were identified (Table  3), both identical to causal gene on chromosome six, as far as determination of those identified previously in panels one and two. Of some leaf margin is concerned, or cause a morphological change interest is the fact that three die ff rent WOX genes were found that prevents spine development as well as induce marginal close to markers associated with piping margins. These were rolling in the case of piping specifically. What is evident is all positioned on chromosome 23 and some distance apart. that piping does not usually occur with spines or spiny-tip. WOX1 is positioned at 1,057,435–1,059,380 bp, WOX2 is at The putative WOX and ZFP polymorphisms are therefore 738,739–739,215 bp and WOX5 is at 159,566–160,802 bp. upstream or dominant over the putative ABA pathway poly- There were no linkages observed between markers near morphisms regarding leaf morphology. these genes and, as such, it is assumed they are complimen- There were no significantly associated markers identi- tary or independently functional. If we assume these differ - fied in panel two genotypes (spiny-tip and piping) on chro- ent WOX genes are each independently associated with the mosome six where the spiny-tip QTL has previously been piping margin phenotype, then they are likely independent shown to be positioned. This suggests there is no difference selection events. The piping margin phenotype has therefore in the spiny-tip allele between spiny-tip and piping margin likely been selected multiple times rather than once. It is plants. This then suggests most of the piping margin plants possible however, given the panels used here, that some of might carry the same polymorphism for spiny-tip and there- this selection has occurred in the post-Columbian period. fore are descendants or closely related to a spiny-tip culti- In the case of the largest and most diverse set of geno- var such as ‘Smooth Cayenne’. As previously stated, it is types used here, panel three, SNP marker #54316030 believed the piping margin genetics incorporated into the was the most highly associated. This marker was close to genotypes represented in this study originated from ‘Monte a WOX1 and ZFP CONSTANS. When referenced against Lirio’. This suggests ‘Monte Lirio’ is related to a spiny-tip the spiny-tip ‘Smooth Cayenne’, piping genotypes were cultivar like ‘Smooth Cayenne’. homozygous or heterozygous for a polymorphism. Spiny- tip and spiny genotypes were homozygous for the ‘Smooth Cayenne’ variant. Conclusion Leitão (2018), using a small F1 population, estimated the piping locus to be between 1,177,327 and 1,976,804 bp on All MLM and logistic regression models were in good chromosome 23. That study used an unrelated population agreement regarding the broad QTL and in most cases and different mapping strategy but was in very good agree- specific markers most significantly associated with piping ment thus supporting the current study. The QTL identi- leaf margin. Positions within 0.6–1.1 mb on chromosome fied in that study was close to the second QTL found in the 23 contained the most significantly associated loci using current study which was positioned close to a WUSCHEL three logistic regression models, four mixed linear models related homeobox 1 (Aco007021.1) and Zinc finger CON- and three diversity panels. In all three diversity panels, a STANS-like protein (Aco007020.1). Nashima et al. (2022) WOX and ZFP were positioned close to the most signifi- used a fine mapping strategy to identify approximately the cantly associated markers. While other significantly associ- same QTL on chromosome 23 between positions 2,100,182 ated markers on other chromosomes were demonstrated, it is and 2,262,401 and nominated one of the same candidate likely they are of less importance in most genotypes studied. WOX genes (Aco007021.1) with supporting RNAi data but It appears there are two discreet QTLs associated depending described the gene as a WOX3 not a WOX1 as annotated in on the diversity panel. One QTL encompasses a WOX2 and the reference genome (Ming et al. 2015). ZFP2, the other a WOX1 and ZFP CONSTANS. It is highly In a previous study, Sanewski (2020) identie fi d a QTL on probable a WOX and/ or a ZFP are the main gene(s) control- chromosome six associated with the spiny-tip leaf margin. ling the piping leaf margin in modern pineapple varieties 1 3 Tropical Plant Biology (2022) 15:233–246 245 Bollier N, Sicard A, Leblond J et al (2018) AT-MINI ZINC FINGER2 that originate from the Hawaiian PRI pineapple breeding and S1-INHIBITOR OF MERISTEM ACTIVITY, a conserved program. The only piping leaf margin variety used in that missing link in the regulation of floral meristem termination in program was ‘Smooth Guatemala’ (syn. ‘Monte Lirio’) (PRI Arabidopsis and tomato. Plant Cell. https:// doi. org/ 10. 1105/ tpc. breeding records) and it is evident that this is the source of 17. 00653 Bradbury PJ, Zhang Z, Kroon DE et al (2007) TASSEL: Software for the piping phenotype in the current study. association mapping of complex traits in diverse samples. Bioin- The relatively high level of heterozygosity in pineap- formatics 23:2633–2635 ple, coupled with strong selection pressure for this trait Braumann I, Urban W, Preus A et al (2018) Semi-dwarf barley (Hordeum by plant breeders and early domesticators has likely led to vulgare L.) brh2 and ari-I mutants are deficient in a U-box ubiquitin ligase. Plant Growth Regul 86:223–234 the accumulation of genes which play a role. This raises Campbell M, Momen M, Walia H, Morota G (2019) Leveraging values the possibility that other piping or smooth leaf genotypes obtained from random regression models for genetic inference of not studied in detail here might possess a slightly different longitudinal traits. The Plant Genome 12(2):8p. https:// doi. org/ genetic determinism as geographical separation and sexual 10. 3835/ plant genom e2018. 10. 0075 Chen L, Hellmann H (2013) Plant E3 ligases: Flexible enzymes in a recombination has likely played a role in their development. sessile world. Mol Plant 6(5):1388–1404 Most polymorphisms exhibiting a highly significant associa- Chen H, Wang C, Conomos M et al (2016) Control for population tion for the piping phenotype in this study are all positioned structure and relatedness for binary traits in genetic association on chromosome 23. studies via logistic models. AJHG 98(4):653–666. https:// doi. org/ 10. 1016/j- ajhg. 2016. 02. 012 Homologues of several of the proposed candidate genes Chen J, Zhang Z (2017) User manual for intelligent prediction and are known to have functions in important plant processes association tool, version 1.3.  https: // doi. or g/ z zlab. ne t/ iP at. other than plant morphogenesis in other species. Further Accessed on 07 Jan 2020 study is needed to better understand the role of ‘piping leaf Chen L-Y, Van Buren R, Paris M et  al (2019a) The bracteatus pineapple genome and domestication of clonally propagated genes’ in not only that phenotype but also other plant func- crops. Nat Genet 51:1549–1558. https:// doi. or g/ 10. 1038/ tions important to horticulture. s41588- 019- 0506-8 Collins JL (1960) The pineapple. Botany, cultivation and utilization. Supplementary Information The online version contains supplemen- Interscience Publishers Inc, New York. tary material available at https://doi. or g/10. 1007/ s12042- 022- 09317-7 . Coppens d’Eeckenbrugge G, Sanewski G (2011) Leaf margin in pine- apple. Pineapple News 18: 32–37. https://w ww.i shs-h ortic ultur e. Acknowledgements The project was funded by Horticulture Innova- org/ worki nggro ups/ pinea pple tion Australia Pty Ltd (Hort Innovation) using voluntary contributions Coppens d’Eeckenbrugge G, Duval M-F, Leal F (2018a) The pineap- from the Department of Agriculture and Fisheries and matched funds ple success story: From domestication to pantropical diffusion. from the Australian Government. In: Genetics and genomics of pineapple. Eds: Ming R. Springer Nature. Switzerland. 281 Coppens d’Eeckenbrugge G, Duval MF, Leal F (2018b) Origin and Funding Open Access funding enabled and organized by CAUL and nd evolution. In: The pineapple, 2 edition; Botany, Production and its Member Institutions. Uses. Eds: Sanewski GM, Bartholomew DP and Paull RE. CABI Publishing. Oxfordshire 341 Declarations Honda E, Yew C-L, Yoshikawa T et al (2018) LEAF LATERAL SYM- METRY1, a member of the WUSCHEL-RELATED HOME- Conflicting Interests No potential personal funding, competing or OBOX3 gene family, regulates lateral organ development differ - conflicting interests are reported by the author. entially from other paralogs, NARROW LEAF2 and NARROW LEAF3 in rice. Plant Cell Physiol 59(2):376–391 Huang M, Liu X, Zhou Y, Summers RM, Zhang Z (2019) BLINK: A Open Access This article is licensed under a Creative Commons Attri- package for the next level of genome-wide association studies bution 4.0 International License, which permits use, sharing, adapta- with both individuals and markers in the millions. Gigascience tion, distribution and reproduction in any medium or format, as long 8(2):giy154 as you give appropriate credit to the original author(s) and the source, Jiang C-J, Sugano S, Takatsuji H (2008) An Arabidopsis SUPERMAN- provide a link to the Creative Commons licence, and indicate if changes like gene, AtZFP12, expressed at shoot organ boundaries sup- were made. The images or other third party material in this article are presses cell growth. J Plant Biol 51(6):413–417 included in the article's Creative Commons licence, unless indicated Kirchhelle C, Chow C-M, Foucart C, Neto H et al (2016) The specifi- otherwise in a credit line to the material. If material is not included in cation of geometric edges by a plant Rab GTPase is an essential the article's Creative Commons licence and your intended use is not cell-patterning principle during organogenesis in Arabidopsis. permitted by statutory regulation or exceeds the permitted use, you will Dev Cell 36:386–400 need to obtain permission directly from the copyright holder. To view a Leitão JM (2018) Genetic mapping in pineapple. In: Ming R (ed) copy of this licence, visit http://cr eativ ecommons. or g/licen ses/ b y/4.0/ . Plant Genetics and Genomics: crops and models. Genetics and Genomics of Pineapple. Springer, Cham. https://doi. or g/10. 1007/ 978-3- 030- 00614-3 Li W-T, He M, Wang J, Wang V-P (2013) Zinc finger protein (ZFP) in References plants – A review. Plant Omics 6(6):474–480 Liu X, Huang M, Fan B, Buckler ES, Zhang Z (2016) Iterative usage Barrett JC, Fry B, Maller, Daly MJ (2005) Haploview: analysis and of fixed and random effect models for powerful and efficient visualization of LD and haplotype maps. Bioinformatics 1 3 246 Tropical Plant Biology (2022) 15:233–246 genome-wide association studies. PLoS Genet 12(2):e1005767. Sanewski GM (2020) DArTseq molecular markers associated with the https:// doi. org/ 10. 1371/ journ al. pgen. 10057 67 spiny-tip leaf margin in pineapple (Ananas comosus L.) Trop Plant Lycell G (2008) The role of Rab GTPases in cell wall metabolism. J Biol. https:// doi. org/ 10. 1007/ s12042- 019- 09245-z Exp Bot 59(15):4061–4074. https:// doi. org/ 10. 1093/ jxb/ ern255 Schwarz G (1978) Estimating the dimension of a model. Ann Statist Ming R, VanBuren R, Wai C-M et al (2015) The pineapple genome and 6(2):461–464 the evolution of CAM photosynthesis. Nat Genet 47:1435–1442. Segura V et al (2012) An efficient multi-locus mixed-model approach https:// doi. org/ 10. 1038/ ng. 3435 for genome-wide association studies in structured populations. Money D, Gardner K, Migiovsky Z et al (2015) LinkImpute: Fast Nat Genet 44(7):825–832. https:// doi. org/ 10. 1038/ ng. 2314 and accurate genotype imputation for non-model organisms. G3: Shenstone E, Cooper J, Rice B, Bohn M, Jamann TM, Lipka AE (2018) Genes. Genomes, Genetics 5:2383–2390 An assessment of the performance of the logistic mixed model for Nashima K, Shirawas K, Isobe S et al (2022) Gene prediction for leaf analyzing binary traits in maize and sorghum diversity panels PLoS margin phenotype and fruit flesh color in pineapple (Ananas One. https:// doi. org/ 10. 1371/ journ al. pone. 02077 52 comosus) using haplotype-resolved genome sequencing. Plant J Somssich M, Je BI, Simon R, Jackson D (2016) CLAVATA-WUSCEL 110:720–734. https:// doi. org/ 10. 1111/ tpj. 15699 signalling in the shoot meristem. Development 143(18):3238–3248 Nibau C, Stilio VS, Wu H-M et al (2011) Arabidopsis and tobacco Speth EB, Imboden L, Hauck P, He SY (2009) Subcellular localization SUPERMAN regulate hormone signalling and mediate cell pro- and functional analysis of the Arabidopsis GTPase RabE. Plant liferation and differentiation. J of Exp Botany 62(3):949–961. Physiol 149(4):1824–37. https://doi. or g/10. 1104/ pp. 108. 132092 . https:// doi. org/ 10. 1093/ jxb/ erq325 Tadege M (2016) WOX3 in the scene: intimacy with hormones. J Exp Noble TJ, Tao Y, Mace ES et al (2018) Characterization of linkage Bot 67(6):1605–1607. https:// doi. org/ 10. 1093/ jxb/ erw073 disequilibrium and population structure in a mung bean diversity Tang Y, Liu X, Wang J et al (2016) GAPIT version 2: An enhanced panel. Front Plant Sci. https:// doi. org/ 10. 3389/ fpls. 2017. 02102. integrated tool for genomic association and prediction. The Plant Purcell S, Neale B, Todd-Brown K et al (2007) PLINK: A tool set for Genome 9(2):1–9 whole-genome association and population-based linkage analysis. Van Raden PM (2008) Efficient methods to compare genomic predictions. Am J Hum Genet 81(3):559–575 J Dairy Sci 91:4414–4423 Rahman Z, Azam SM, Liu Y, Cheng Y (2017) Expression profiles Xu Y, Wu Y, Long Q et al (2014) Over-expression of OsZHD1, a zinc of Wuschel-related Homeobox gene family in pineapple (Ananas finger homeobox class homeobox transcription factor, induces comosus L.) Trop Plant Biol 10(4). https:// doi. or g/ 10. 1007/ abaxially curled and drooping leaf in rice. Planta 239(4):803–816. s12042- 017- 9192-9https:// doi. org/ 10. 1007/ s00425- 013- 2009-7 Rahni R, Birnbaum KD (2016) Plant cell shape: trafficking gets edgy. Zhang Z, Ersov E, Lai CQ et al (2010) Mixed linear model approach Dev Cell 36(4):353–354. https:// doi. org/ 10. 1016/j. devcel. 2016. adapted for genome-wide association studies. Nat Genet 02. 005 42(4):355–360 Sanewski G, Ko L, Innes D, Kilian A, Carling J, Song J (2017) DArT- seq molecular markers for resistance to Phytophthora cinnamomi Publisher's Note Springer Nature remains neutral with regard to in pineapple (Ananas comosus L.) Australasian Plant Pathol 46(5): jurisdictional claims in published maps and institutional affiliations. 499–509. https:// doi. org/ 10. 1007/ s13313- 017- 0512-1 1 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Tropical Plant Biology Springer Journals

DArTseq molecular markers associated with the piping leaf margin phenotype in pineapple (Ananas comosus L.)

Tropical Plant Biology , Volume 15 (3) – Sep 1, 2022

Loading next page...
 
/lp/springer-journals/dartseq-molecular-markers-associated-with-the-piping-leaf-margin-p2811M1Q6F
Publisher
Springer Journals
Copyright
Copyright © Crown 2022
ISSN
1935-9756
eISSN
1935-9764
DOI
10.1007/s12042-022-09317-7
Publisher site
See Article on Publisher Site

Abstract

This study sought to understand the genetic basis of the piping leaf margin phenotype in pineapple. To achieve this aim, a genome-wide association study (GWAS) using mixed linear regression and logistic regression analysis was conducted on three pineapple diversity panels including seedling populations segregating for spiny, spiny-tip and piping leaf margins. This study identified single nucleotide polymorphism (SNP) markers associated with the piping and spiny-tip leaf margin phe- notypes. A broad quantitative trait locus (QTL) positioned on chromosome 23 between positions 240,475 and 2,369,197 bp was the most highly associated with piping leaf margin in all analyses. Major candidate genes proposed are a Zinc finger protein 2, a Zinc finger protein 3, a WUSCHEL-related homeobox 2, a WUSCHEL-related homeobox 1 and a Zinc finger protein CONSTANS-like. Some other genes of a lower association, linked or nearby genes of interest, are also considered potentially involved to varying degrees. All candidate genes are known to be involved in aspects of stem cell maintenance, cell proliferation, epidermal cell differentiation, organogenesis, leaf polarity, cell wall modification or hormone signalling. It is possible each plays a role in either differentiation or morphological aspects of the spiny-tip and piping leaf margin phenotypes. It is expected the relative role of each associated gene might vary with genetic background. Keywords Spines · Piping leaf · ZFP · Abaxial · WUSCHEL · Zinc finger · W OX · Leaf margin Abbreviations MAF Minor allele frequency ABA Abscisic acid MMLM Multi-locus mixed linear model BGLR Bayesian Genetic Linear Regression PIC Polymorphism information content BLINK Bayesian-information and Linkage- PRI Pineapple Research Institute disequilibrium Iteratively Nested Keyway QTL Quantitative trait loci FarmCPU Fixed and Random Model Circulating Prob- SNP Single nucleotide polymorphism ability Unification SAM Shoot apical meristem DArT Diversity Arrays Technology GAPIT Genomic Association and Prediction Inte- Introduction grated Tool GMMAT Generalized linear Mixed Model Association Pineapple exhibits five leaf margin types and all, except Tests complete spiny, are considered the product of domestica- GWAS Genomic-wide Association Study tion (Coppens d’Eeckenbrugge and Sanewski 2011). Geno- iPat Intelligent Prediction and Association Tool types exhibiting this phenotype are typically found in the KNNi K-nearest neighbour imputation north-western regions of Brazil (Coppens d’Eeckenbrugge LD Linkage disequilibrium et al. 2018a), the country of origin of Ananas. The piping margin typically exhibits a thin white border along the leaf margins and is devoid of any spines (Fig. 1A and B). Leaves Communicated by Robert Henry with a spiny margin are considered the natural state as all * Garth M. Sanewski wild genotypes possess this phenotype and it is reasonable garth.sanewski@daf.qld.gov.au to assume piping represents a mutation selected at some 1 stage in the estimated 6,000 to 10,000 years of domestica- Queensland Department of Agriculture and Fisheries, tion of pineapple (Coppens d’Eeckenbrugge et al. 2018b). Maroochy Research Facility, Nambour, Australia Vol.:(0123456789) 1 3 234 Tropical Plant Biology (2022) 15:233–246 cases, only present as a shallow valley. There are however no published studies describing the botanical nature of pip- ing leaf margin. Leaf margin spines do not usually occur with piping suggesting an additional genetic mechanism. According to Collins (1960), the piping gene is epistatic to the spiny and spiny-tip alleles. He goes further to say, plants homozygous for piping have a more pronounced piping margin, suggest- ing a dosage effect and plants homozygous recessive for the piping alleles will display as either spiny or spiny-tip. The piping phenotype appears dominant over spiny and spiny-tip phenotypes. The piping leaf margin appears very stable and there are few reports of reversion to spiny, unlike the spiny- tip phenotype which frequently reverts to spiny and displays considerable variability in gene penetrance (Sanewski 2020). This study examined the botanical nature of piping leaf margin and identified SNP markers associated with spiny-tip and piping leaf margin phenotypes. The use of three diver- sity panels, independently analysed, was a novel approach that enabled testing between pairs of phenotype states; viz, piping and spiny, piping and spiny-tip and spiny and spiny- tip. Putative candidate genes associated with the piping leaf margin, all located on chromosome 23, are discussed. Methods In this study, the use of the term ‘piping’ is meant to encom- Fig. 1 A and B. The piping leaf margin typically exhibits a thin white pass smooth margins with or without folding and/or an obvi- border along the leaf margin (1A) and is devoid of spines even at the ous white edge. Plants were phenotyped by visually inspect- tip. On some leaves of some genotypes, it might also be characterised by an upward fold in the margin (1B) ing the leaf margins and carefully feeling the margin to the very tip for the presence of small spines. Those without any spines anywhere on the leaf were considered as piping, those The piping leaf margin does not appear to offer any obvious with small spines at the tip as spiny-tip, and plants with evolutionary advantage and hence its widespread existence spines completely over the margins were classed as spiny. in the western Amazon is likely all attributable to mankind. Because the trait exists in several different genotypes which Histology are considered to have arisen from clonal selection and sex- ual recombination (Chen et al. 2019), the piping phenotype 10 µm thick transverse leaf margin sections were prepared most likely has been selected multiple times. from cultivars, ‘Tapiricanga’ (piping) and ‘Smooth Cayenne’ The first obvious sign of the piping phenotype occurs (spiny-tip), using a freezing microtome to examine the dif- about 8 cm behind the tip and extends to the base of the leaf ferences between a piping margin and a margin with spine where it emerges from the meristematic zone. The demar- suppression but no piping. The sections were fixed in 4% cation or boundary between adaxial and abaxial ‘domains’ formalin and mounted in Mowiol solution for brightfield appears clear but a narrow strip of abaxial surface has been microscopy. folded onto the adaxial surface to present a new ‘false’ margin (Collins 1960). The piping strip will vary in width Panel Descriptions for GWAS in different genotypes from almost non-existent to a strip approximately 2 mm wide. The width can also vary on dif- Three diversity panels were used. Diversity panel one was a ferent leaves of the same plant or on different margins of sub-set of 90 plants from a population of seedlings exhibit- the same leaf possibly revealing an environmental effect on ing piping (33 plants) or completely spiny leaf margins (57 gene penetrance, meristematic cell development or leaf cell plants). This panel was intended to reveal the genetic dif- turgor. The fold can be exaggerated and obvious or, in some ference between spiny and piping margin. The population 1 3 Tropical Plant Biology (2022) 15:233–246 235 was developed using the parents that included spiny, spiny- authors of that work to correct for an error in the original tip and piping genotypes. Sixty-two of these plants were of assembly (Chen et al. 2019). These two reference genomes the parent combination ‘PRI 59–656’ (pollen) and ‘MD- should be identical for chromosomes two to 23 and hence 2’ (seed). It is considered that the piping margin trait in the different panel results are comparable where markers all progenies used was inherited from the piping margin of interest are positioned only on chromosomes two to 23. ‘Smooth Guatemala’ component and the spiny trait was Before analysis, the SNP data sets were filtered on repro- derived from ‘MD-2’. ducibility (> 0.99), call rate (> 0.75) and MAF (> 0.025) Panel two comprised 516 genotypes with 12 full sib fami- before imputation using the LD KNNi approach in TASSEL lies and some of their parental varieties. There were 440 5.2.44 (Bradbury et al. 2007; Money et al. 2015). The accu- plants with piping leaf margins and 76 with spiny-tip. The racy of imputation was tested by masking 10% of SNPs and population had all derived from varieties from the Pineap- comparing the imputed file with the original. ple Research Institute (PRI) of Hawaiian breeding program as for panel one. Panel two differed from panel one in that Principal Component Analysis (PCoA) it represented a larger number of recombinant events and spiny leaf margin plants were not included. This panel was Principal component covariates were generated in TASSEL intended to reveal the genetic difference between spiny-tip 5.2.54 (Bradbury et al. 2007). The number of covariates for and piping margin genomes. GWAS was determined based on either the Bayesian Infor- Diversity panel three was an amalgamation of panel mation Criterion (BIC) approach (Schwarz 1978) performed two seedlings (508 seedlings) and a reciprocal bi-parental in GAPIT, or the QQ and Scree plots. The smallest num- population (367 seedlings) (‘MD2’ × ‘Cayenne’, and ‘Cay- ber of covariates that produced a QQ plot that was neither enne’ × ‘MD2’) of seedlings segregating for spiny and spiny- inflated or under-inflated and had the smallest P values and tip margin. Panel three also included 76 parental genotypes was consistent with the Scree plot (visualized in TASSEL and several unrelated genotypes including wild accessions. but not reproduced) was used in the GWAS. Panel three was the most diverse of the populations used. There were 492 plants with piping margins, 351 with spiny- Association Analyses tip and 108 with spiny margins to give a total of 951 plants. The SNP marker data was analysed by genome-wide asso- Molecular Marker Development ciation study methods (GWAS) using mixed linear models (MLM) with structure correction. The multi-locus mixed Freeze-dried leaf base samples (white tissue) were supplied linear models (MMLM) of Segura et al. (2012), FarmCPU to Diversity Arrays Pty Ltd for DArTseq marker develop- (Liu et al. 2016) and BLINK (Huang et al. 2019) and the sin- ment. Marker development followed the procedure described gle loci ECMLM model, all in the GAPIT R package version in Sanewski et al. (2017). The methylation-sensitive restric- 3 (Zhang et al. 2010; Tang et al. 2016) were used. Kinship tion enzymes PstI and MseI were used to avoid highly repet- was estimated using the Van Raden method in GAPIT (Van itive regions that bias GWAS, and to reduce the genome Raden 2008). complexity. The marker sequences were trimmed to 69 bp. The phenotype data was binary and coded as 0 and 1 Marker sequences for panel one genotypes were refer- for panels one and two and 0, 1 and 2 for panel three. The enced against the revised genome of the spiny-tip pineapple Bonferroni-adjusted P value of P ≤ 0.01 was used as the level cultivar ‘Smooth Cayenne’ clone 153 (Acomosus_321_v3; of significance. https:// genom evolu tion. org/ CoGe/ Noteb ookVi ew. pl? nid= The R package Bayesian Genomic Linear Regression 937) (Ming et al. 2015). (BGLR) was used, through the graphic user interface, Subsequent to the sequencing of panel one genotypes, a Intelligent Prediction and Association Tool (iPat v1.3) revised version of this genome sequence became available (Chen and Zhang  2017) to test SNP markers in panels and was used for panels two and three. Marker sequences for one, two and three. The Bayes A model was used with a panel two and panel three genotypes were referenced against probit link to optimise analysis of binary or ordinal data the revised genome of the spiny-tip pineapple cultivar through logistic regression. The algorithm was run with ‘Smooth Cayenne’ clone 153 (Acomosus_321_v7.1, assem- 50,000 iterations and the first 3,000 discarded. The Van bly GCA_902162155.1 at European Nucleotide Archive; Raden kinship matrix and principal components were used https:// www. ebi. ac. uk/ ena/ brows er/ home). This version of to adjust for relatedness. The marker effects posterior val - the genome is slightly different from the previous version. ues were converted to absolute values and the Z-score Chromosomes 24 and 25 have been amalgamated as a larger calculated by dividing the absolute marker effect by its chromosome 25. Chromosome one was divided and the later standard deviation. The P value was then calculated by the portion re-labelled as chromosome 24. This was done by the following equation according to Campbell et al. (2019), 1 3 236 Tropical Plant Biology (2022) 15:233–246 P = 2*(1-pnorm(Z-Score)). A Bonferroni-adjusted P ≤ 0.05 Results was used as the cut-off for significance. A mixed model logistic regression was also performed Histology using the R program Generalized Linear Mixed Model Association Test (GMMAT v1.1.2) (Chen et al. 2016) for The section for spiny-tip cv. ‘Smooth Cayenne’ (Fig. S1) panels one and two data. This model uses a Bernoulli dis- shows the co-ordinated adaxial and abaxial layers. The tribution and logit function with covariate adjustment by abaxial layer is characterized by a corrugated epidermal principal components as a fixed effect and kinship (Van surface which appears slightly thinner than the adaxial sur- Raden) as a random effect to adjust for population struc- face. The transition between the adaxial and abaxial layers ture. Logistic models such as GMMAT that can include dissects the vascular bundles such that the phloem bundle fixed and random effects to account for population struc- and its sclerenchyma sheath are in the abaxial layer, and ture are considered more suitable for binary traits than the xylem bundle and its sclerenchyma sheath are in the linear mixed models especially if some populations have adaxial layer. Alternating between the phloem-xylem bun- more or less ‘affected’ status rather than ‘control’ sta- dles, but only in the abaxial layer, are aeration canals. Both tus. In such situations there is an increased risk of type the abaxial and adaxial layers have a line of fibre bundles I errors with mixed linear models as they assume error relatively close to the epidermal layer. Both the adaxial terms are normally distributed (Shenstone et al. 2018). and abaxial layers converge to the leaf margin. The model 10 algorithm in GMMAT was used for to per- The piping leaf margin section shows the abaxial piping form score tests on marker associations. The data were ‘bulge’ consists of one to two repeats of the phloem vas- treated as binomial. cular bundle and fibre strands in two different directions Logistic regression was also performed in gPLINK (Fig. S2A and B). The piping section therefore appears to using a logit function (Purcell et al. 2007) for panel one be one or two repeats of the abaxial layer at the margin only. sandwiched on top of each other before the terminating epidermal cell layer. Linkage Analysis Molecular Marker Datasets Linkage disequilibrium (LD) for the linkage decay curve was estimated for the entire genome for panel one data using the The original seedling SNP dataset for panel one comprised software package TASSEL 5.2.44 using a sliding window of 3,574 markers with 18% heterozygosity and an average 100 SNPs. The r of SNP marker pair frequencies and their minor allele frequency (MAF) of 0.23. The average PIC was distance was used to generate a linkage decay curve. A Loess 0.31, one ratio 0.47 and call rate 0.85. There were 13.5% curve was fitted using the package SigmaPlot 14. Linkage null alleles before imputation. Imputation accuracy was 91% disequilibrium decay distance was calculated as the point on and there were 2,141 SNP markers with an average MAF of the x axis corresponding to where the Loess curve passed 0.23 and 0.18 heterozygosity after imputation. 77 genotypes 50% of the maximum r (Noble et al. 2018). remained for association analyses. The original SNP dataset Marker linkages were also analysed for individual chro- for panel two comprised 516 genotypes and 40,686 markers mosomes for panel one SNP markers using Haploview and contained 4.8% null alleles. After imputation, with an 4.2 (Barrett et  al. 2005). In the case of Haploview, the accuracy of 88%, 516 genotypes and 16,393 markers remained Hardy–Weinberg p value cut-off was set at zero. with heterozygosity of 0.32, average MAF of 0.29 and average PIC of 0.18. After filtering for MAF ≤ 0.05 and imputation (accuracy 0.88), the panel three dataset included 16,393 SNP Nearby Genes markers for 952 genotypes and had an average MAF of 0.29 and heterozygosity of 0.33. Genes within the LD decay distance were identified using the annotated ‘Smooth Cayenne’ v3 genome for panel one Principal Component Analysis (PCoA) genotypes and v7 genome for panels two and three in the software package ‘Geneious R10’. This differential use of Based on the PCoA analysis, zero to four principal compo- genome versions was because these different genomes were nents (PCs) were used as covariates for panel one depending used as the reference for marker positioning in those marker on the algorithm, up to nine for panel two depending on the datasets. Likely candidate genes were selected based on algorithm and three for panel three. For panel one, the first proximity and supporting published reports of gene/protein four PCs accounted for 27% of the genotypic variance. For biological function. 1 3 Tropical Plant Biology (2022) 15:233–246 237 panel two, the first 9 PCs accounted for 37% and for panel homozygous for the SNP or heterozygous in piping geno- three, the first three PCs accounted for 18%. types. SNP marker #4713909 was homozygous for the SNP allele or heterozygous in spiny genotypes and homozygous Marker Density for the reference allele in piping genotypes. Several other markers on chromosomes one and two were also signifi- Density for SNP markers for chromosome 23 using the panel cantly associated but most were intergenic and not matched three dataset after filtering and imputation is shown in Fig.  2. to genes. Several other genes of interest were however posi- Figure 2 showed that marker density was high on chromo- tioned close to these markers. The Manhattan and QQ plots some 23 at the starting end up to mid-way around position for GWAS of panel one SNPs are shown in Figs. 3A and 4,504,232 bp. Chromosome 23 appears to be telocentric. 4A. Figure 5 shows the significantly associated SNP mark - ers by ECMLM and gPlink for the main QTL on chromo- Linkage Disequilibrium some 23. SNP marker #4714214, at position 798,190 bp on chromosome 23, was the most highly associated SNP LD decay distance for the entire genome using panel one marker by GMMAT and gPlink. SNP marker #4713909 SNPs indicated an effective decay distance of 1,222 kb for was the only significantly associated SNP marker using the entire genome (Fig. S3). This LD decay distance was Bayes A in BGLR (Table 1). This intergenic marker is posi- used as a guide for selecting candidate genes in all analyses. tioned on chromosome 23 at 762,438 bp and was close to For panel one, most of the significant markers were inter - the most significantly associated SNP marker (#4714214). genic and linkages were generally not strong between asso- Both #4714214 and #4713909 are positioned close to a ciated SNPs. WOX2 (Aco015463.1) and ZFP2 (Aco015474.1) (Table S1). Linkage data for the most highly associated SNP markers in Association Analyses panel one (spiny and piping margins) is shown in table S2. The highest associated SNP marker by BGLR, #4713909, Panel one comprised piping and spiny phenotypes only. The was close to the WOX2 and ZFP2 but also highly linked highest associated SNP marker (#4714214) was the most to #100067496 which was also close to the WOX2 and significantly associated in all panel one mixed linear model ZFP2. There were no other strong linkages with associated (MLM) analyses (Table  1). Genes close to this position SNP markers. include a Zinc-finger protein 2 (ZFP2), Zinc-finger protein Panel two comprised spiny-tip and piping margin plants. 3 (ZFP3), Ras-related protein Rab-8B, and WUSCHEL- SNP marker #54316030 was the most significantly associ - related homeobox 2 (WOX2) genes (Table S1). The WOX2 ated in all panel two GWAS analyses and logistic regres- and ZFP2 are considered the main candidates associated with sion by BGLR and GMMAT (Table  2). This marker was piping and spiny leaf phenotypes in this panel of genotypes. positioned on chromosome 23 at 1,068,662 bp, 9.3 kb from SNP marker #4714214 was homozygous for the reference a WUSCHEL-related homeobox 1 (WOX1) (Aco007021.1) allele (spiny-tip ‘Smooth Cayenne’) in spiny genotypes and and 31.6 kb from a Zinc finger CONSTANS-like protein Fig. 2 SNP marker density (number of markers within 1 Mb sliding window) for all 25 chromosomes, panel 3 after filtering for MAF < 0.025 and imputation by LD-KNNi 1 3 238 Tropical Plant Biology (2022) 15:233–246 Table 1 SNP markers SNP Position Length -log10p (Bonf) Maf Algorithm Allelic significantly associated with chrom: bp (bp) effect spiny and piping leaf margin phenotypes for panel one 4714214|F|0–52:G > A 23: 798,190 59 8.3 0.33 ECMLM pos genotypes using the GAPIT 15.2 MLMM algorithms ECMLM, MLMM, 12.7 FarmCPU FarmCPU and BLINK and 17.8 BLINK logistic regressions in GMMAT, 6.7* Plink gPLINK and BGLR. All 12.4 GMMAT markers are significant at the 4713909|F|0–63:G > A 23: 762,438 69 7.7 0.45 ECMLM neg Bonferroni-adjusted P < 0.01 4.3* Plink except the BGLR and gPLINK 2.0* BGLR values* which are significant 12.4 GMMAT at P < 0.05. 17 of the SNPs 4712886|F|0–20:A > G 23: 423,562 69 7.5 0.26 ECMLM pos with the lowest association 6.4* Plink significance, all on chromosome 11.0 GMMAT 23 or not positioned, have been omitted. The direction of the 100044752|F|0–50:T > C 2: 16,332,302 69 6.9 0.07 FarmCPU pos allelic effect for the ECMLM 4712964|F|0–31:C > T 23: 296,925 69 6.7 0.27 ECMLM pos algorithm is shown 6.1* Plink 11.1 GMMAT 100058551|F|0–63:A > G NP 69 6.7 0.50 FarmCPU pos 100012184|F|0–68:A > G 1: 24,514,356 69 6.3 0.47 FarmCPU pos 100067496|F|0–23:C > A 23: 678,197 69 6.2 0.44 ECMLM neg 5.3* Plink 11.0 GMMAT 4713774|F|0–5:T > C 23: 1,970,691 69 5.6* 0.37 Plink pos 7.9 GMMAT 4716438|F|0–19:C > T 23: 1,305,909 69 5.4* 0.34 Plink neg 9.3 GMMAT 4714616|F|0–49:A > G NP 69 5.3* 0.27 Plink pos 8.6 GMMAT 100058930|F|0–31:T > G 23: 1,473,662 69 5.3* 0.41 Plink neg 9.1 GMMAT 4724522|F|0–24:C > G NP 41 5.2* 0.48 Plink pos 7.4 GMMAT 4717335|F|0–57:T > C 23: 240,475 69 5.2* 0.44 Plink pos 9.4 GMMAT 4724236|F|0–29:G > A 23: 2,058,488 69 5.1* 0.36 Plink pos 7.0 GMMAT 100044395|F|0–7:G > C 23: 1,999,412 69 5.0* 0.47 Plink pos 7.5 GMMAT (Aco007020.1) (Table S3). These are considered the best cultivars. Other markers of relatively lower significance were candidates for the piping and spiny-tip genotype data set also demonstrated by some algorithms on chromosome 5, 6, studied. This marker was homozygous or heterozygous for 12, and 13 and un-positioned. The Manhattan and QQ plots the polymorphism in piping genotypes. Spiny-tip geno- for GWAS of panel two SNP data are shown in Figs.  3B types were homozygous for the reference allele (spiny-tip and 4B. cv. ‘Smooth Cayenne’). Marker #28878068 positioned at Panel three comprised spiny, spiny-tip and piping 671,978 bp on chromosome 23 was highly associated by margin phenotypes. Marker # 54316030 at position MLM only. This marker was unmatched but positioned 1,068,662 on chromosome 23 was the most significantly 6.9 kb from a Myb domain protein 97 (Aco015454.1), 28 kb associated in three MLM algorithms (Table  3). This from an O-fucosyltransferase family protein (Aco015458.1) marker was only 9.3 kb from a WOX1 (Aco007021.1) and 66.8  kb from a WUSCHEL-related homeobox  2 and 31.6 kb from a Zinc finger CONSTANS-like protein (WOX2) (Aco015463.1). Piping genotypes were homozy- (Aco007020.1) (Table S4). These are the same genes pro- gous for the polymorphism and null or homozygous for the posed for panel two. Marker # 4723993 on chromosome reference allele (spiny-tip cv. ‘Smooth Cayenne’) in spiny-tip six at 14,330,763 bp was also highly associated by three 1 3 Tropical Plant Biology (2022) 15:233–246 239 Fig. 3 A, B and C. Manhattan plots for SNP markers for panel one corrected -log10(Bonferroni-adjusted) P < 0.01 cut-off horizont al (3A) analysed in GAPIT using the ECMLM, MLMM, BLINK and lines are shown. Markers on scaffolds and not positioned are included FarmCPU algorithms for association with spiny and piping leaf mar- as fictitious chromosomes 26 and 27 respectively. Figures reproduced gin phenotypes, panel two markers (3B) analysed for spiny-tip and in the R package ‘Memory-efficient, Vizualization-enhanced, and piping leaf margin phenotypes and panel three markers (3C) analysed Parallel-accelerated tool for genome-wide association study’ (rMVP), for spiny, spiny-tip and piping leaf margin phenotypes. The structure-https:// github. com/ xiaol eiLiu Bio/ rMVP MLM algorithms. This was close (20.5 kb) to a Zeaxanthin homozygous for the reference allele in spiny-tip and spiny epoxidase (ZEP) as identified putatively associated with genotypes and homozygous for the SNP or heterozygous spiny-tip phenotypes in a previous study (Sanewski 2020). in piping genotypes. Bear in mind, the spiny-tip ‘Smooth Manhattan and QQ plots for panel three SNP data are Cayenne’ provided the reference genome. #54316030 shown in Figs.  3C and 4C. Marker #54316030 was therefore represents a dominant allele for the piping trait. 1 3 240 Tropical Plant Biology (2022) 15:233–246 Fig. 4 A, B and C. QQ plots from the panel one (4A), panel two (4B) and panel three (4C) SNP marker GWAS analyses. Figures reproduced in the R package rMVP, https:// github. com/ xiaol eiLiu Bio/ rMVP Marker #4723993 was homozygous for the polymor- cell wall structural protein and MATE eff lux family gene phism in spiny-tip genotypes and homozygous for the ref- (Table S4). erence allele in spiny genotypes. Piping genotypes carried any of the allele states. SNP marker #4723993 which was associated with the spiny-tip phenotype in panel three, Discussion was in various allele states in piping and spiny-tip geno- types in panel two. There appeared no difference between There do not appear to be any global disturbances in leaf polar- piping and spiny-tip genotypes with respect to the allele ity. The development of a piping margin appears consistent state of this marker for the spiny-tip phenotype. SNP with a cessation of the adaxial layer while the abaxial layer is marker #28883264 was the only marker significantly repeated an additional one to two times at the margin thus creat- associated by logistic regression (BGLR) (Table 3). It is ing a thicker composite abaxial layer, an imbalance and upward positioned on chromosome six at 14,526,582 bp close to folding. Piping leaf margin appears to be a lack of co-ordination several likely candidates including a Receptor kinase 2, between development of the abaxial and adaxial tissues at the Zeaxanthin epoxidase, Cellulose synthase, Glycine-rich leaf margin rather than changes in adaxial-abaxial patterning. 1 3 Tropical Plant Biology (2022) 15:233–246 241 Fig. 5 Significantly associated SNP markers on chromosome 23 for the panel one dataset using the ECMLM and gPLINK analyses. The -log10(P) line of significance (Bonferroni- adjusted) P < 0.01 is shown. The vertical arrows delineate the proposed main QTL Table 2 SNP markers SNP Posn -log10P (Bonf) Length Maf Algorithm Allelic significantly associated with chrom: bp bp effect spiny-tip and piping leaf margin phenotypes for panel 54316030|F|0–11:A > C 23: 1,068,662 34.6 29 0.46 MLMM neg two genotypes using the 28.6 Blink GAPIT algorithms ECMLM, 27.2 ECMLM MLMM, FarmCPU and 26.3 FarmCPU BLINK and logistic regression 5.5* BGLR in GMMAT and BGLR. All 27.1 GMMAT markers are significant at 28878068|F|0–65:G > A 23: 671,978 28.4 69 0.25 MLMM neg the Bonferroni-adjusted 21.9 Blink P < 0.01 except the BGLR 20.9 ECMLM value* which is significant at FarmCPU 18.0 P < 0.05. The direction of the 28878068|F|0–66:C > A 18.1 69 0.32 Blink pos allelic effect for the ECMLM 16.6 MLMM algorithm is shown. Seventy- 12.2 FarmCPU one markers were found 14.1 GMMAT significant but only the most significant markers for each 4713909|F|0–63:G > A 23: 762,438 20.3 69 0.36 ECMLM neg chromosome are shown 17.7 GMMAT 4710125|F|0–59:A > G 23: 701,343 18.6 69 0.45 ECMLM neg 21.5 GMMAT 28877781|F|0–34:C > G 23: 506,814 16.0 69 0.23 ECMLM neg 11.3 FarmCPU 6.7 Blink 10.6 GMMAT 54317320|F|0–18:T > A 5: 6,727,065 10.9 69 0.20 Blink neg 4727322|F|0–43:A > T 12: 5,152,594 7.1 69 0.36 FarmCPU neg 6.9 Blink 54316897|F|0–37:G > A 6: 9,696,245 9.7 69 0.14 ECMLM neg 28877802|F|0–5:T > G 13: 2,431,298 6.0 69 0.08 Blink pos 1 3 242 Tropical Plant Biology (2022) 15:233–246 Table 3 SNP markers SNP Position -log10p Length (bp) Maf Algorithm Allelic significantly associated with chromo: bp effect spiny, spiny-tip and piping leaf margin phenotypes for panel 54316030|F|0–11:A > C 23: 1,068,662 60.3 29 0.33 BLINK pos three genotypes using the 60.0 FarmCPU GAPIT algorithms MLMM, 24.2 MLMM FarmCPU and BLINK and 4723993|F|0–39:G > C 6: 14,330,763 56.0 69 0.47 FarmCPU neg logistic regression in BGLR. All 24.3 MLMM markers are significant 8.4 BLINK at the Bonferroni-adjusted 28883264|F|0–10:C > A 6: 14,526,582 50.2 69 0.20 BLINK neg P < 0.01 except the BGLR 40.0 FarmCPU value* which is significant 30.0 MLMM at P < 0.05. The direction of 7.4* BGLR the effect, either positive or negative for the ECMLM 4726743|F|0–6:A > T 6: 14,355,558 20.9 69 0.27 BLINK pos algorithm is shown, even 15.2 FarmCPU though this algorithm is not 12.3 MLMM otherwise represented in the 54314089|F|0–40:G > A 6: 14,334,819 23.7 69 0.47 BLINK neg table 6.7 FarmCPU 54314200|F|0–17:G > A 6: 13,977,148 18.5 69 0.33 BLINK neg 28881980|F|0–27:G > A 6: 13,967,023 14.0 69 0.35 FarmCPU neg 54314646|F|0–9:G > T 13: 10,965,585 11.7 69 0.46 BLINK neg 4716772|F|0–47:A > C 6: 14,556,931 11.2 69 0.15 MLMM pos 28881708|F|0–8:G > A NP 11.0 23 0.30 FarmCPU neg 6.4 BLINK 4714193|F|0–41:C > T 12: 5,141,097 11.0 69 0.29 BLINK neg 54309331|F|0–37:T > G 6: 14,525,401 10.8 69 0.36 BLINK neg 28876898|F|0–59:G > A 6: 13,991,970 9.9 69 0.47 MLMM pos 28874267|F|0–22:T > C 18: 10,110,153 9.7 69 0.06 BLINK pos 54312110|F|0–55:T > C 7: 12,303,848 9.7 69 0.18 BLINK neg 28882074|F|0–12:G > A 5: 10,126,663 9.5 69 0.08 BLINK neg 4716438|F|0–19:C > T 23: 1,305,909 8.9 69 0.23 BLINK neg 7.3 MLMM 54310912|F|0–23:A > G 12: 418,631 8.9 69 0.23 BLINK pos 4726657|F|0–31:T > G 12: 4,768,435 8.7 69 0.28 FarmCPU neg 54313753|F|0–50:G > A 7: 10,461,432 9.3 69 0.23 BLINK neg 6.6 FarmCPU 54316608|F|0–39:G > A 6: 13,991,970 9.3 69 0.31 BLINK pos 4710241|F|0–17:A > T 6: 14,732,936 8.8 69 0.35 BLINK neg 4709656|F|0–11:G > T 6: 14,242,751 8.6 69 0.18 FarmCPU pos 28882522|F|0–38:A > G 4: 1,202,604 8.1 58 0.09 FarmCPU neg 54315064|F|0–6:C > T NP 7.7 24 0.06 BLINK neg 4715867|F|0–68:A > T 25: 8,053,599 7.2 69 0.46 BLINK neg 4718200|F|0–6:G > A 6: 13,796,356 7.0 69 0.32 FarmCPU pos 28877469|F|0–16:G > C 2: 13,859,177 7.0 69 0.31 FarmCPU neg 4713341|F|0–56:G > A 5: 11,677,507 6.9 69 0.07 MLMM neg 28879929|F|0–25:C > T 25: 6,103,800 6.9 65 0.11 BLINK pos 4719816|F|0–30:C > G NP 6.8 69 0.46 BLINK neg 4717292|F|0–51:G > T 6: 14,468,303 6.7 62 0.16 BLINK pos 28883299|F|0–59:G > A 25: 8,864,898 6.7 69 0.27 FarmCPU pos 4715794|F|0–33:C > G NP 6.6 5.1 0.35 BLINK pos 28877973|F|0–6:T > C 4: 1,026,724 6.4 69 0.06 BLINK neg 1 3 Tropical Plant Biology (2022) 15:233–246 243 JAGGED, have also been shown to be involved in boundary Diversity Panel One maintenance and/ or cell proliferation. In both Arabidopsis and tomato, the KNUCKLES transcription factor indirectly Panel one comprised a population segregating for spiny and piping margin and as such is considered the definitive data represses WUSCHEL to allow a temporary inhibition of meristematic activity (Bollier et al. 2018). Ubiquitin E3 set for examining polymorphisms associated with the selec- tion of piping-leafed variants from the presumed original ligases are also ZFP family genes. Plant U-box E3 ligases have been implicated in plant development and morphol- spiny genotypes. The assumption at the commencement of this study is that the piping phenotype arose from the spiny ogy. As an example, a U-box E3 ligase mutation in barley has been associated with semi-dwarfing, wavy leaf margins phenotype. Discussion of candidate genes within this context is therefore based primarily on the findings from panel one. and erect growth (Braumann et al. 2018). Cullin4-based E3 ligases are somehow involved in aspects of plant mor- WUSCHEL‑Related Homeobox 2 (WOX2) phology given that mutation of the gene in Arabidopsis causes stunted growth and aberrant leaf shapes (Chen and The WOX2 (Aco015463.1) gene is 59 kb from the highest Hellmann 2013). associated SNP by all four MLM models and two logistic regressions and hence considered a highly likely principal Ras‑related Protein Rab‑8B causal variant, along with the ZFP2. There are ten WOX genes in pineapple, spread across six chromosomes with The Ras-related Rab-8B gene (Aco015473.1) is positioned only 4.0 kb from the highest associated SNP marker in panel three on chromosome 23 (Rahman et al. 2017). Research on WOX genes in other species supports a potential role one. Again, data in other species suggests a potential role in leaf margin phenotype in pineapple. The Rab GTPase fam- in leaf margin morphology. Rice plants carrying a muta- tion in a WOX3 called LSY1, exhibited curled leaf margins ily of proteins plays an important role in cell wall synthesis and modification through trafficking of cell wall polymers to suggesting greater cell growth or proliferation in the abax- ial leaf tissues compared with the adaxial side. In another specific target membranes. Cell wall synthesis and modifica- tion such as might occur during cell expansion, are integral example, a WUS ortholog in rice is expressed mostly in leaf margins (Somssich et al. 2016). Also, overexpression to plant development and possibly architecture (Speth et al. 2009; Lycell 2008). A study in Arabidopsis looked at the of a WOX3 in rice results in adaxial curling of leaf margins. Similarly, a WOX1 double mutant in rice exhibits reduced membrane where RabA localizes, and in conjunction with down-regulation manipulation, demonstrated a correspond- leaf blades and thickened leaf margins (Honda et al. 2018). WOX1 promotes cell proliferation (Tadege 2016) suggest- ing change in cell shape (Kirchhelle et al. 2016). Follow- ing on from this it was postulated by Rahni and Birnbaum ing a loss or reduced-function polymorphism would reduce cell proliferation. (2016) that Rab proteins might mediate local wall stiff- ness thus affecting cell shape and consequently tissue and Zinger Finger Proteins (ZFPs) organ morphology. This potential role in specific cell edge mechanics, rather than that of the entire cell wall, might The highest associated SNP marker was 1.8 kb from a ZFP2 have implications in pineapple piping leaf margin where a weakness in one edge might allow bending or folding, given (Aco015474.1). ZFPs form a large family of transcription factors with roles in plant development and stress response growth and turgor pressure in surrounding tissues, especially if cell proliferation is increased in some surrounding tissues. and appear conserved across plant evolution (Li et al. 2013). As in the case of WOX genes, ZFPs have also been impli- Although not studied here, it might also explain the position and direction of spines. cated in leaf morphology in other species. Superman (SUP), one of the better known ZFPs affecting morphology, sup- Diversity Panel Two presses cell division at a specific stage and is associated with flower organogenesis and plant morphogenesis (Jiang Panel two results represent a difference between spiny-tip et al. 2008). SUP overexpression in Nicotiana tabacum pro- duces a percentage of plants with curled-up leaf margins and and piping leaf margin phenotypes. It is known that the pip- ing margin is dominant over spiny-tip (Collins 1960). The thickened roots (Nibau et al. 2011). In plants with moderate expression of SUP and curly leaves, the organisation of the results here (table two, table S3) suggest the piping mar- gin plants might carry a polymorphism in either a WOX1, adaxial/ abaxial mesophyll layers was disrupted with more cells and cells of variable shape throughout. Overexpression WOX2 and/ or ZFP that the spiny-tip plants do not. There is no difference between the genotypes represented in panel of a ZFP in rice, OsZHD1, induced abaxial curling due to an increase in the number and abnormal arrangement of bul- two with respect to the spiny-tip polymorphisms puta- tively associated in a previous study (Sanewski 2020). This liform cells (Xu et al. 2014). Other ZFPs, KNUCKLES and 1 3 244 Tropical Plant Biology (2022) 15:233–246 suggests that both piping and spiny-tip genotypes in panel The most likely causal gene was a ZEP, a precursor to ABA, two carry the spiny-tip polymorphism. Because piping mar- although other genes including a MATE efflux were also gin is dominant over spiny-tip, piping margin genotypes as implicated. A much larger diversity panel used in the current derived in this diversity panel must have arisen from spiny- study comprised of many segregating populations and unre- tip genotypes, not spiny genotypes. lated genotypes, and using additional analytical methods, but identified the same broad QTL for the spiny-tip phenotype Diversity Panel Three thus validating the earlier study to some extent. No major SNP markers associated with a piping mar- Because panel three contains all three phenotypes, piping, gin were positioned on chromosome six near the spiny-tip spiny-tip and spiny, it is expected that the SNPs associated loci in the current study. The piping margin phenotype is with both piping margins and spiny-tip will be identified as a however known to suppress other leaf margin phenotypes. difference between them and/or spiny margins. Accordingly, This suggests the piping leaf genes function upstream of the two main QTLs were identified (Table  3), both identical to causal gene on chromosome six, as far as determination of those identified previously in panels one and two. Of some leaf margin is concerned, or cause a morphological change interest is the fact that three die ff rent WOX genes were found that prevents spine development as well as induce marginal close to markers associated with piping margins. These were rolling in the case of piping specifically. What is evident is all positioned on chromosome 23 and some distance apart. that piping does not usually occur with spines or spiny-tip. WOX1 is positioned at 1,057,435–1,059,380 bp, WOX2 is at The putative WOX and ZFP polymorphisms are therefore 738,739–739,215 bp and WOX5 is at 159,566–160,802 bp. upstream or dominant over the putative ABA pathway poly- There were no linkages observed between markers near morphisms regarding leaf morphology. these genes and, as such, it is assumed they are complimen- There were no significantly associated markers identi- tary or independently functional. If we assume these differ - fied in panel two genotypes (spiny-tip and piping) on chro- ent WOX genes are each independently associated with the mosome six where the spiny-tip QTL has previously been piping margin phenotype, then they are likely independent shown to be positioned. This suggests there is no difference selection events. The piping margin phenotype has therefore in the spiny-tip allele between spiny-tip and piping margin likely been selected multiple times rather than once. It is plants. This then suggests most of the piping margin plants possible however, given the panels used here, that some of might carry the same polymorphism for spiny-tip and there- this selection has occurred in the post-Columbian period. fore are descendants or closely related to a spiny-tip culti- In the case of the largest and most diverse set of geno- var such as ‘Smooth Cayenne’. As previously stated, it is types used here, panel three, SNP marker #54316030 believed the piping margin genetics incorporated into the was the most highly associated. This marker was close to genotypes represented in this study originated from ‘Monte a WOX1 and ZFP CONSTANS. When referenced against Lirio’. This suggests ‘Monte Lirio’ is related to a spiny-tip the spiny-tip ‘Smooth Cayenne’, piping genotypes were cultivar like ‘Smooth Cayenne’. homozygous or heterozygous for a polymorphism. Spiny- tip and spiny genotypes were homozygous for the ‘Smooth Cayenne’ variant. Conclusion Leitão (2018), using a small F1 population, estimated the piping locus to be between 1,177,327 and 1,976,804 bp on All MLM and logistic regression models were in good chromosome 23. That study used an unrelated population agreement regarding the broad QTL and in most cases and different mapping strategy but was in very good agree- specific markers most significantly associated with piping ment thus supporting the current study. The QTL identi- leaf margin. Positions within 0.6–1.1 mb on chromosome fied in that study was close to the second QTL found in the 23 contained the most significantly associated loci using current study which was positioned close to a WUSCHEL three logistic regression models, four mixed linear models related homeobox 1 (Aco007021.1) and Zinc finger CON- and three diversity panels. In all three diversity panels, a STANS-like protein (Aco007020.1). Nashima et al. (2022) WOX and ZFP were positioned close to the most signifi- used a fine mapping strategy to identify approximately the cantly associated markers. While other significantly associ- same QTL on chromosome 23 between positions 2,100,182 ated markers on other chromosomes were demonstrated, it is and 2,262,401 and nominated one of the same candidate likely they are of less importance in most genotypes studied. WOX genes (Aco007021.1) with supporting RNAi data but It appears there are two discreet QTLs associated depending described the gene as a WOX3 not a WOX1 as annotated in on the diversity panel. One QTL encompasses a WOX2 and the reference genome (Ming et al. 2015). ZFP2, the other a WOX1 and ZFP CONSTANS. It is highly In a previous study, Sanewski (2020) identie fi d a QTL on probable a WOX and/ or a ZFP are the main gene(s) control- chromosome six associated with the spiny-tip leaf margin. ling the piping leaf margin in modern pineapple varieties 1 3 Tropical Plant Biology (2022) 15:233–246 245 Bollier N, Sicard A, Leblond J et al (2018) AT-MINI ZINC FINGER2 that originate from the Hawaiian PRI pineapple breeding and S1-INHIBITOR OF MERISTEM ACTIVITY, a conserved program. The only piping leaf margin variety used in that missing link in the regulation of floral meristem termination in program was ‘Smooth Guatemala’ (syn. ‘Monte Lirio’) (PRI Arabidopsis and tomato. Plant Cell. https:// doi. org/ 10. 1105/ tpc. breeding records) and it is evident that this is the source of 17. 00653 Bradbury PJ, Zhang Z, Kroon DE et al (2007) TASSEL: Software for the piping phenotype in the current study. association mapping of complex traits in diverse samples. Bioin- The relatively high level of heterozygosity in pineap- formatics 23:2633–2635 ple, coupled with strong selection pressure for this trait Braumann I, Urban W, Preus A et al (2018) Semi-dwarf barley (Hordeum by plant breeders and early domesticators has likely led to vulgare L.) brh2 and ari-I mutants are deficient in a U-box ubiquitin ligase. Plant Growth Regul 86:223–234 the accumulation of genes which play a role. This raises Campbell M, Momen M, Walia H, Morota G (2019) Leveraging values the possibility that other piping or smooth leaf genotypes obtained from random regression models for genetic inference of not studied in detail here might possess a slightly different longitudinal traits. The Plant Genome 12(2):8p. https:// doi. org/ genetic determinism as geographical separation and sexual 10. 3835/ plant genom e2018. 10. 0075 Chen L, Hellmann H (2013) Plant E3 ligases: Flexible enzymes in a recombination has likely played a role in their development. sessile world. Mol Plant 6(5):1388–1404 Most polymorphisms exhibiting a highly significant associa- Chen H, Wang C, Conomos M et al (2016) Control for population tion for the piping phenotype in this study are all positioned structure and relatedness for binary traits in genetic association on chromosome 23. studies via logistic models. AJHG 98(4):653–666. https:// doi. org/ 10. 1016/j- ajhg. 2016. 02. 012 Homologues of several of the proposed candidate genes Chen J, Zhang Z (2017) User manual for intelligent prediction and are known to have functions in important plant processes association tool, version 1.3.  https: // doi. or g/ z zlab. ne t/ iP at. other than plant morphogenesis in other species. Further Accessed on 07 Jan 2020 study is needed to better understand the role of ‘piping leaf Chen L-Y, Van Buren R, Paris M et  al (2019a) The bracteatus pineapple genome and domestication of clonally propagated genes’ in not only that phenotype but also other plant func- crops. Nat Genet 51:1549–1558. https:// doi. or g/ 10. 1038/ tions important to horticulture. s41588- 019- 0506-8 Collins JL (1960) The pineapple. Botany, cultivation and utilization. Supplementary Information The online version contains supplemen- Interscience Publishers Inc, New York. tary material available at https://doi. or g/10. 1007/ s12042- 022- 09317-7 . Coppens d’Eeckenbrugge G, Sanewski G (2011) Leaf margin in pine- apple. Pineapple News 18: 32–37. https://w ww.i shs-h ortic ultur e. Acknowledgements The project was funded by Horticulture Innova- org/ worki nggro ups/ pinea pple tion Australia Pty Ltd (Hort Innovation) using voluntary contributions Coppens d’Eeckenbrugge G, Duval M-F, Leal F (2018a) The pineap- from the Department of Agriculture and Fisheries and matched funds ple success story: From domestication to pantropical diffusion. from the Australian Government. In: Genetics and genomics of pineapple. Eds: Ming R. Springer Nature. Switzerland. 281 Coppens d’Eeckenbrugge G, Duval MF, Leal F (2018b) Origin and Funding Open Access funding enabled and organized by CAUL and nd evolution. In: The pineapple, 2 edition; Botany, Production and its Member Institutions. Uses. Eds: Sanewski GM, Bartholomew DP and Paull RE. CABI Publishing. Oxfordshire 341 Declarations Honda E, Yew C-L, Yoshikawa T et al (2018) LEAF LATERAL SYM- METRY1, a member of the WUSCHEL-RELATED HOME- Conflicting Interests No potential personal funding, competing or OBOX3 gene family, regulates lateral organ development differ - conflicting interests are reported by the author. entially from other paralogs, NARROW LEAF2 and NARROW LEAF3 in rice. Plant Cell Physiol 59(2):376–391 Huang M, Liu X, Zhou Y, Summers RM, Zhang Z (2019) BLINK: A Open Access This article is licensed under a Creative Commons Attri- package for the next level of genome-wide association studies bution 4.0 International License, which permits use, sharing, adapta- with both individuals and markers in the millions. Gigascience tion, distribution and reproduction in any medium or format, as long 8(2):giy154 as you give appropriate credit to the original author(s) and the source, Jiang C-J, Sugano S, Takatsuji H (2008) An Arabidopsis SUPERMAN- provide a link to the Creative Commons licence, and indicate if changes like gene, AtZFP12, expressed at shoot organ boundaries sup- were made. The images or other third party material in this article are presses cell growth. J Plant Biol 51(6):413–417 included in the article's Creative Commons licence, unless indicated Kirchhelle C, Chow C-M, Foucart C, Neto H et al (2016) The specifi- otherwise in a credit line to the material. If material is not included in cation of geometric edges by a plant Rab GTPase is an essential the article's Creative Commons licence and your intended use is not cell-patterning principle during organogenesis in Arabidopsis. permitted by statutory regulation or exceeds the permitted use, you will Dev Cell 36:386–400 need to obtain permission directly from the copyright holder. To view a Leitão JM (2018) Genetic mapping in pineapple. In: Ming R (ed) copy of this licence, visit http://cr eativ ecommons. or g/licen ses/ b y/4.0/ . Plant Genetics and Genomics: crops and models. Genetics and Genomics of Pineapple. Springer, Cham. https://doi. or g/10. 1007/ 978-3- 030- 00614-3 Li W-T, He M, Wang J, Wang V-P (2013) Zinc finger protein (ZFP) in References plants – A review. Plant Omics 6(6):474–480 Liu X, Huang M, Fan B, Buckler ES, Zhang Z (2016) Iterative usage Barrett JC, Fry B, Maller, Daly MJ (2005) Haploview: analysis and of fixed and random effect models for powerful and efficient visualization of LD and haplotype maps. Bioinformatics 1 3 246 Tropical Plant Biology (2022) 15:233–246 genome-wide association studies. PLoS Genet 12(2):e1005767. Sanewski GM (2020) DArTseq molecular markers associated with the https:// doi. org/ 10. 1371/ journ al. pgen. 10057 67 spiny-tip leaf margin in pineapple (Ananas comosus L.) Trop Plant Lycell G (2008) The role of Rab GTPases in cell wall metabolism. J Biol. https:// doi. org/ 10. 1007/ s12042- 019- 09245-z Exp Bot 59(15):4061–4074. https:// doi. org/ 10. 1093/ jxb/ ern255 Schwarz G (1978) Estimating the dimension of a model. Ann Statist Ming R, VanBuren R, Wai C-M et al (2015) The pineapple genome and 6(2):461–464 the evolution of CAM photosynthesis. Nat Genet 47:1435–1442. Segura V et al (2012) An efficient multi-locus mixed-model approach https:// doi. org/ 10. 1038/ ng. 3435 for genome-wide association studies in structured populations. Money D, Gardner K, Migiovsky Z et al (2015) LinkImpute: Fast Nat Genet 44(7):825–832. https:// doi. org/ 10. 1038/ ng. 2314 and accurate genotype imputation for non-model organisms. G3: Shenstone E, Cooper J, Rice B, Bohn M, Jamann TM, Lipka AE (2018) Genes. Genomes, Genetics 5:2383–2390 An assessment of the performance of the logistic mixed model for Nashima K, Shirawas K, Isobe S et al (2022) Gene prediction for leaf analyzing binary traits in maize and sorghum diversity panels PLoS margin phenotype and fruit flesh color in pineapple (Ananas One. https:// doi. org/ 10. 1371/ journ al. pone. 02077 52 comosus) using haplotype-resolved genome sequencing. Plant J Somssich M, Je BI, Simon R, Jackson D (2016) CLAVATA-WUSCEL 110:720–734. https:// doi. org/ 10. 1111/ tpj. 15699 signalling in the shoot meristem. Development 143(18):3238–3248 Nibau C, Stilio VS, Wu H-M et al (2011) Arabidopsis and tobacco Speth EB, Imboden L, Hauck P, He SY (2009) Subcellular localization SUPERMAN regulate hormone signalling and mediate cell pro- and functional analysis of the Arabidopsis GTPase RabE. Plant liferation and differentiation. J of Exp Botany 62(3):949–961. Physiol 149(4):1824–37. https://doi. or g/10. 1104/ pp. 108. 132092 . https:// doi. org/ 10. 1093/ jxb/ erq325 Tadege M (2016) WOX3 in the scene: intimacy with hormones. J Exp Noble TJ, Tao Y, Mace ES et al (2018) Characterization of linkage Bot 67(6):1605–1607. https:// doi. org/ 10. 1093/ jxb/ erw073 disequilibrium and population structure in a mung bean diversity Tang Y, Liu X, Wang J et al (2016) GAPIT version 2: An enhanced panel. Front Plant Sci. https:// doi. org/ 10. 3389/ fpls. 2017. 02102. integrated tool for genomic association and prediction. The Plant Purcell S, Neale B, Todd-Brown K et al (2007) PLINK: A tool set for Genome 9(2):1–9 whole-genome association and population-based linkage analysis. Van Raden PM (2008) Efficient methods to compare genomic predictions. Am J Hum Genet 81(3):559–575 J Dairy Sci 91:4414–4423 Rahman Z, Azam SM, Liu Y, Cheng Y (2017) Expression profiles Xu Y, Wu Y, Long Q et al (2014) Over-expression of OsZHD1, a zinc of Wuschel-related Homeobox gene family in pineapple (Ananas finger homeobox class homeobox transcription factor, induces comosus L.) Trop Plant Biol 10(4). https:// doi. or g/ 10. 1007/ abaxially curled and drooping leaf in rice. Planta 239(4):803–816. s12042- 017- 9192-9https:// doi. org/ 10. 1007/ s00425- 013- 2009-7 Rahni R, Birnbaum KD (2016) Plant cell shape: trafficking gets edgy. Zhang Z, Ersov E, Lai CQ et al (2010) Mixed linear model approach Dev Cell 36(4):353–354. https:// doi. org/ 10. 1016/j. devcel. 2016. adapted for genome-wide association studies. Nat Genet 02. 005 42(4):355–360 Sanewski G, Ko L, Innes D, Kilian A, Carling J, Song J (2017) DArT- seq molecular markers for resistance to Phytophthora cinnamomi Publisher's Note Springer Nature remains neutral with regard to in pineapple (Ananas comosus L.) Australasian Plant Pathol 46(5): jurisdictional claims in published maps and institutional affiliations. 499–509. https:// doi. org/ 10. 1007/ s13313- 017- 0512-1 1 3

Journal

Tropical Plant BiologySpringer Journals

Published: Sep 1, 2022

Keywords: Spines; Piping leaf; ZFP; Abaxial; WUSCHEL; Zinc finger; WOX; Leaf margin

References