Genomic structure and marker-trait association for plant and fruit traits in Capsicum chinense and Capsicum baccatum germplasm

Abstract

Objectives

Capsicum baccatum and C. chinense are domesticated pepper species originating from Latin America recognized for their unique flavor and taste and widely diffused as spicy food for fresh uses or for processing. Owing to their capacity for adaptation to diverse habitats in tropical regions, these species serve as a valuable resource for agronomic traits and tolerance to both biotic and abiotic challenges in breeding projects. This study aims to dissect the genetic diversity of C. baccatum and C. chinense germplasm and to detect candidate genes underlying the variation of plant morphological and fruit size and shape traits. To that goal, SNP data from genotyping by sequencing have been used to investigate the genetic diversity and population structure of 103 accessions belonging to the two species. Further, plants have been assessed with main plant descriptors and fruit imaging analysis and association between markers and traits has been performed.

Results

The population structure based on 29,820 SNPs revealed 4 subclusters separating C. chinense and C. baccatum individuals. A deeper analysis within each species highlighted three subpopulations in C. chinense and two in C. baccatum. Phenotypic characterization of 54 traits provided approximately 125 thousand datapoints highlighting main differences between species for flower and fruit traits rather than plant architecture. Marker-traits association, performed with the CMLM model, revealed a total of 6 robust SNPs responsible for change in flower traits and fruit shape. This is the first attempt for mapping morphological traits and fruit features in the two domesticated species, paving the way for further genomic assisted breeding.

Introduction

Pepper (Capsicum spp.) is one of the most important vegetable crops which has been originated in the American tropics and after European conquests has been spread throughout the world. Pepper fruits are a finest source of health-related compounds such as Vitamin C (ascorbic acid), Vitamin E (tocopherols), provitamin A (b-carotene), flavonoids [1]. Peppers are also unique among vegetables for the presence of capsaicinoids conferring the characteristic pungent aroma [2]. The genus consists of approximately forty species, out of which five, namely C. annuum L., C. baccatum L., C. chinense Jacq., C. frutescens L., and C. pubescens, are domesticated and extensively used for consumption. Among these, C. baccatum is highly consumed in the Andean regions as fresh of processed spice whereas, C. chinense, is largely grown in the Caribbean area, Africa and Northeast India, and its popularity is given by the presence of highly pungent varieties, that reached the Guinness World Record [3, 4]. Both species include a wide range of variation related to plant and fruit traits [5, 6]. Furthermore, them are recognized for their distinctive flavor and for uses in traditional cuisines [3], as well as a source of resistances to main pathogens, pests and abiotic stresses [7]. The growing demand of peppers with desirable quality and disease resistance traits has raised the interest of both C. baccatum and C. chinense in breeding programs [8]. Therefore, it becomes essential to integrate phenotypic evaluation with genetic diversity analysis with the goal of exploiting germplasm resources and improve the allele mining that plant breeders can use to create novel varieties. Dissecting plant architecture and morphological fruit traits is fundamental prior the establishment of any breeding strategy. In pepper, precise phenotyping can be achieved by plant descriptors and automated imaging tools [9]. Progress in plant genotyping have been facilitated by development of leading-edge technologies that sped up the detection and analysis of single nucleotide polymorphisms (SNPs). Genotyping-by-sequencing was the most successful one providing a cost-effective approach for exploring the diversity of crops genome-wide scale [10].

The genetic diversity and association mapping studies in Capsicum have primarily focused on C. annuum [11,12,13,14], with limited research on other domesticated species, mainly focusing on specific traits [15, 16] The objective of this work was to explore the genomic diversity and population structure of 103 accessions of C. chinense and C. baccatum and perform candidate genes scan for 54 plant and fruit traits via association mapping. Findings help to cover the gap of knowledge among domesticated peppers toward their better use for genetic improvement.

Main text

Materials and methods

Plant material

Plant material consisted of 41 C. baccatum and 62 C. chinense accessions retrieved from main international genebanks (Centre for Genetic Resources, CGN, Wageningen, The Netherlands, and the Leibniz-Institut for Plant Research, IPK, Gatersleben, Germany) and seed companies (Esasem) and highly representative of the variability existing in the origin centres and other World regions. Details are in Additional File 1: Table S1. Prior to experiments, accessions were subjected to cycles of controlled self-fertilization to increase homozygosity and remove any off types. For minimizing the environmental effect, all genotypes were grown in controlled conditions under glasshouse at the Research Centre for Vegetable and Ornamental Crops (Pontecagnano, Salerno, Italy; Latitude: 40°39′N, Longitude 14°53′E). The cultivation cycle was carried out during spring-summer season with a climate control consisting of an evaporative fan cooling system and a shading screen able to ensure an internal temperature range was from 14 °C (night) to 28 °C (day). Plants were grown in pots filled with peat and pumice (3:1 v/v) with water and nutrients administered via fertigation through a drip system. A complete randomized design with three replicates was adopted.

Phenotypic characterization and data analysis

Fifty-four traits were considered including qualitative and pseudo-qualitative descriptors conform to the guidelines of the International Plant Genetic Resources Institute [17], colour CIELAB (L*a*b*) coordinates, fruit size and shape parameters from imaging analysis. Details of traits scored are in Additional File 1: Table S2. All traits were subjected to subjected to ANOVA (analysis of variance) to detect significant differences among the two pepper species. Analyses were performed in R version 4.3.3. [18].

Genomic diversity and association analysis

Genotyping data were retrieved from the dataset described in Taranto et al. [11]. The ApeKI restriction enzyme was used for library preparation. The resultant fragments from all samples were directly ligated to a pair of enzyme-specific adapters and then combined into pools. PCR amplification was carried out to generate the GBS library, which was submitted to a single Illumina HiSeq 2500 run (Illumina Inc., USA). Illumina reads were trimmed with cutadapt [19] then aligned to the C. annuum CM334 reference genome v1.6 [20] with the Burrows–Wheeler Aligner BWA-MEM v0.7.17 [21] with default parameters. The GBS analysis pipeline implemented in TASSEL (version 3.0.166) was used to call SNPs [22]. SNP calling implemented within the TASSEL-GBS pipeline produced a raw HapMap genotypic data file. A two-step filtering procedure was used to filter high quality SNPs. SNPs were filtered out with VCFtools version 0.1.17 [23] considering a call rate of 95% able to retain SNPs with less than 5% missing data and minor allele frequency of 5%. Three datasets were obtained: (i) a matrix of 29,820 SNPs for the entire collection including 103 individuals of both species, (ii) a matrix of 22,739 SNPs for C. chinense, (iii) a matrix of 24,290 SNPs for C. baccatum. Homozygosity and heterozygosity of individuals of the two collections was inferred by Tassel v5.2.15 [24]. Genetic differentiation parameters including F statistic and private and shared alleles were computed in VCFtools version 0.1.17.

Population structure and ancestry were investigated in the whole collection and keeping separated the two species using ADMIXTURE version 1.3.0 [25] with K ranging from 1 to 10, 1,000 bootstrap replicates to estimate parameter standard errors, 10-fold cross-validation (CV) with five iterations. The optimal number of ancestral populations was detected with the lowest value of the cross-validation error. Individuals were considered to belong to a specific K population if its membership coefficient (qi) was ≥ 0.5, whereas the genotypes with qi lower than 0.5 at each assigned K were considered as admixed. Principal component analysis (PCA) was performed in Tassel v5.2.15 [19] and the biplot was drawn using the ggplot2 R package [26]. The Neighbor-Joining (NJ) phylogenetic tree was drawn using the Maximum Composite Likelihood methodnmethod with 1000 bootstraps. Analyses were conducted with the MEGA 11 software [27]. Association between traits scored and markers was performed using Tassel v5.2.15 software. Phenotypic data from three replications were used. The compressed mixed linear model (CMLM) considering principal components covariates and the familial relatedness based on the SNP data model was used. The significance of association between traits and markers was estimated by using an adjusted P value (Bonferroni correction) according to the formula B = 0,05/n° of SNP. Manhattan and quantile–quantile (Q–Q) plots for association mapping results were produced using the R package CMplot. Functional annotation of significant association signals was performed on C. annum CM334 genome http://peppergenome.snu.ac.kr/ [20] For associated SNPs mapping within intron regions, the nearest genes located upstream or downstream of the significant marker.

Results and discussion

Admixture analysis of the entire collection based on 29,820 SNPs, revealed the presence of 4 subclusters mostly separating C. chinense and C. baccatum individuals, as from the CV error results (Additional File 1: Table S3). The first cluster (K1) included 39 C. chinense genotypes highly diversified in terms of geographical origin (Fig. 1a, Additional File 1: Table S4). The second one, included 11 C. chinense and 1 C. baccatum accessions highlighting specific group of individuals from Perú. The third subcluster (K3) revealed admixtures between the two species, including 6 accessions of C. baccatum and 10 of C. chinense, while the fourth one (K = 4) exclusively consisted of 33 individuals of C. baccatum. The principal component analysis (PCA) projections revealed clear differentiation of most of the individuals of the two species, showing furthermore a greater diversity within the group individuals belonging to K = 3. This observation agrees with previous studies showing defined genetic differentiation between the two species [28]. The first two principal components explained a combined 76.31% of the genetic variance (PC1, 58.63% and PC2, 17.68%) with the first component separating the two species, while the second highlighting substructures within each species. Given the results on the whole collection, population structure has been thoroughly investigated within each species. Based on 24,290 SNPs, two subpopulations were detected in C. baccatum (Additional File 1: Table S3, Fig. 1b). The first subpopulation (K = 1) included 8 genotypes, while the remaining 33 were clustered in the second one (K = 2). The PCA revealed a clear separation of the two subclusters on the first axis (PC1, 66.09%); the second component exhibiting 5.37% of the variation showed a further subdivision of K1 in two subgroups mostly distinguishing accessions from the internal region of Perù. For C. chinense, a filtered dataset of 22,739 SNPs allowed to distinguish three subpopulations including 39, 9 and 13 accessions, respectively with one accession classified as admixed (Additional File 1: Table S3, Fig. 1b). The PCA revealed on the first axis (PC1, 45.70%) a high differentiation of accessions belonging to cluster 2 from those of clusters 1 and 3, while the second component (PC2, 7.21%) better separated the latter two subpopulations. The fixation index (Fst) considering the entire collection and individual species, revealed a relatively lower genetic differentiation between C. chinense clusters compared to C. baccatum ones (Additional File 1: Table S5). Indeed, at entire population level, K1 and K2 mostly including C. chinense genotypes exhibited a moderate Fst = 0.349 whereas the remaining clusters showed high Fst values above 0.7 with peaks between K4 (C. baccatum) and the rest of the clusters. The trend was confirmed when ADMIXTURE analysis was performed at single-species level with the two clusters of C. baccatum showing Fst = 0.770 while for C. chinense, the greatest differentiation was observed only between the largest cluster (K1) and the other two (0.732 < Fst < 0.762) whereas between K2 and K3 an Fst = 0.326 was detected.

Fig. 1

Genomic diversity of the 103 pepper accessions using GBS-SNPs derived markers. In each subfigure, from the left to the right are shown the population structure considering the most informative value according to the cross-validation error implemented in ADMIXTURE, the Loading plot in the first two components with annotated the number of K subpopulation and the species. (a) Population structure in the whole collection based on 29,820 SNPs, the population was divided into four (K = 4) groups according to the most informative K value; (b) Population structure in C. baccatum based on 24,290 SNPs, the population was divided into four (K = 2) groups according to the most informative K value; (c) Population structure in C. chinense based on 22,739 SNPs, the population was divided into four (K = 3) groups according to the most informative K value

The observed (Ohom) and expected (Ehom) homozygosity in C. baccatum were 88.713% and 69.951%, respectively. In C. chinense Ohom was 90.099%, whereas Ehom was 68.979%.

On average the level of heterozygosity of the two species was similar with slightly higher values ​​observed for C. baccatum (He = 11.286%) compared to C. chinense (He = 9.900%) (Additional File 2; Figure S1). Only few accessions showed heterozygosity higher than 20% with the maximum values observed in two C. baccatum genotypes. High inbreeding coefficients (Fisand Fit) were observed for the two species with values in C. baccatum and C. chinense of Fis = 0.710, Fit = 0.837 and Fis = 0.698, Fit = 0.831 respectively.

To better understand the genetic diversity of the collection, a Neighbor-Joining tree was constructed (Fig. 2). The phylogenetic diversity analysis was generally in agreement with the population structure one. Two main subpopulations which separated the two species were detected, with a third group that included a both C. chinense and C. baccatum accessions (Fig. 2a) and few C. baccatum accessions grouping with the main C. chinense cluster. The separation of the subpopulations obtained with the structure analysis was equally evident by redrawing the tree in accordance with the ADMIXTURE clusters (Fig. 2b).

Fig. 2

Neighbor-joining phylogenetic tree (unrooted) using 29,820 SNPs. The evolutionary distances were computed using the Maximum Composite Likelihood method model with 1,000 bootstraps. (a) Tree with annotated species and horticultural type. (b) Tree with annotated grouping revealed by the population structure analysis

These results can be explained by the role of C. chinense as a bridge species in interspecific crosses between cultivated pepper (C. annuum) and C. baccatum, thus making possible a gene-flow between these two species [29] as highlighted by the number of shared alleles between the two species with a percentage of 39.357% in C. baccatum and 43.201% in C. chinense.

Phenotypic characterization has involved the analysis of 18 morphological and colour traits as well as 38 fruit size and shape descriptors from the scan of over 3000 fruit sections. Overall, over 125.000 data points were obtained. The distribution of phenotypic data between the two species are shown in Fig. 3, while detailed results of ANOVA, average, maximum and minimum values for each trait, are reported in the Additional File 1: Table S6. Six out of twelve morphological traits showed significant differences (P < 0.05) between species, with colours traits turned out to be the most discriminating ones. Among these, the greenish/yellowish spots are distinctive characters of C. baccatum from other domesticated species [30]. Contrariwise, C. chinense does not present flowers with either spot or purple coloured corolla flowers [30]. Most significant differences were found for fruit size and shape traits. C. baccatum accessions showed in average a higher size (perimeter and area) and height of the fruits. The width was instead more similar between the two species. As a result, C. baccatum displayed a more elongated fruit shape, resulting in greater dimensions compared to C. chinense, whose fruits had square dimensions. This was also confirmed by ellipsoid (E) and circular (C) higher in C. baccatum and rectangular (R) higher in C. chinense. The parameters defining the internal shape of the fruit (e.g. PMA, PIA, DMA) were higher in C. chinense, thus evidencing the higher irregularity of berries.

Fig. 3

Box plots showing mean values and quartiles (first and third) for 54 traits assayed in C. baccatum (Cb) and C. chinense (Cc) accessions. Plant descriptors are indicated with green and violet boxplots, colour coordinates with light-blue and yellow boxplots, fruit size and shape traits with blue and red boxplots. Details of traits in Table S2. The red asterisk indicates traits significantly different (P < 0.05) between species

The Spearman rank correlation coefficients after Bonferroni correction calculated for qualitative and quantitative traits revealed how some traits were rather independent, whereas a group of traits clustered together because of a reciprocal tight correlation for the same category of measures (e.g., fruit size and shape traits, colour traits) (Fig. 4). The strongest positive correlations in both species were found between fruit shape indexes and height/width related traits as well as between these traits and eccentricity area index and lobedness degree. Negative correlations were found between height, width, fruit shape index traits and proximal and distal angles of fruits. As for morphological traits, in C. chinense moderate correlations were found between stem pubescence, leaf shape, leaf pubescence and corolla shape with fruit shape and size traits, while the same trend was not observed in C. baccatum.

Fig. 4

Spearman’s rank correlation coefficients between pairs of phenotypes. In each cell, coloured ellipses are those with P < 0.05 after Bonferroni correction. The strength of correlations is inversely proportional to the ellipses dimension. The colour intensity is directly proportional to the coefficients. On the right side of the correlogram, the colour legend shows the correlation coefficients and the corresponding colours. Correlogram for traits scored in C. baccatum is placed upside the diagonal the correlogram for traits scored in C. chinense is placed below the diagonal. Trait acronyms are listed in Table S2

Considering the outcomes obtained with population structure and genetic diversity analysis and the observed variability at phenotypic level, the collection can be considered suitable for association analysis. Marker trait association was performed in individual species as well as in the entire collection. This strategy allows to detect those robust SNPs across species considering the high level of heritability of the traits studied. In terms of computational and statistical capabilities, the compressed mixed linear model utilized has been demonstrate effective in reducing false positives caused by population structure [31]. The Bonferroni correction for multiple testing have been also used in order to manage the family-wise error [32]. In total, 6 robust associations on 5 chromosomes, were consistently identified in both species (Fig. 5). Four peaks fell within intergenic regions at a distance of 0.13 to 42.41 kb from known genes (Table 1).

Fig. 5

Association results for flowers and fruit traits. On the left the Manhattan plots showing SNP–trait associations using the CMLM model; the analysis consider SNPs identified in individual C. baccatum and C. chinense accessions as well as in the entire population. The black horizontal line indicates a significant threshold (− log10 p-value) according to Bonferroni. The X-axis indicates the chromosome position. On the right, Quantile–quantile plots for each trait. FAC = Flower anther colour; FCS = Flower corolla shape; CS = Flower corolla spot; OB = Obovoid.

In pepper, the accumulation of pigments in the vegetative parts of plants (e.g., anthocyanins) exert important biological functions by aiding in the resistance to diseases and pest attack [33]. Both C. baccatum and C. chinense are reported a valuable source of resistances to many pathogens and insect vectors of viruses [34]. Corolla shape may play role in the selection of pollinators [35], although no evidence has been reported in pepper. Among intergenic associations, the monothiol glutaredoxin-S14, chloroplastic-like GRXS14 play important roles in the control of vegetative growth and possessing specific functions in the maintenance of chlorophyll content upon depending on environmental and light conditions [36], although this does not appear to affect flower morphology. The metal transporter CNNM4 associated to corolla spot colour has been found to mineral homeostasis in barley seeds [37], while the H0515C11.6 protein have been found in Oryza sativa play a role in ion transport photosynthesis metabolism [38]. However, these two genes do not seem to have any function on the presence of corolla spots.

Only for flower anther colour and obovoid associations within unknown proteins were found. In both instances, the minor allele determined an increase of the trait. While these associations were not previously reported, the significant association detected for obovoid (trait related to fruit shape determined by the ratio of the area of the fruit above and below middle height) confirm chromosome 10 as a main one holding candidate genes responsible for fruit size and shape in pepper [13, 39, 40]. This outcome demonstrates the reliability of associations found and of the pursued methodological approach.

Table 1 Significant associations detected by a compressed mixed linear model. For each trait, chromosomal position (in bp), main statistic and effects of each association, corresponding annotated gene or close annotated gene are shown

Limitations

• Passport data not available in all accessions.

• A higher number of accessions could provide better understanding on the geographical stratification of the two species under study, increasing furthermore, the power of detecting novel associations.

• The two species represent valuable source of quality traits and resistances/tolerance to pathogens and environmental stress: phenotyping additional traits would provide better insight of the value of the collection for breeding purposes.