The Gossypium stocksii genome as a novel resource for cotton improvement

Abstract

Cotton is an important textile crop whose gains in production over the last century have been challenged by various diseases. Because many modern cultivars are susceptible to several pests and pathogens, breeding efforts have included attempts to introgress wild, naturally resistant germplasm into elite lines. Gossypium stocksii is a wild cotton species native to Africa, which is part of a clade of vastly understudied species. Most of what is known about this species comes from pest resistance surveys and/or breeding efforts, which suggests that G. stocksii could be a valuable reservoir of natural pest resistance. Here, we present a high-quality de novo genome sequence for G. stocksii. We compare the G. stocksii genome with resequencing data from a closely related, understudied species (Gossypium somalense) to generate insight into the relatedness of these cotton species. Finally, we discuss the utility of the G. stocksii genome for understanding pest resistance in cotton, particularly resistance to cotton leaf curl virus.

Introduction

The cotton genus, Gossypium, is responsible for providing the majority of natural textile fiber through the cultivation of its four domesticated species. While most research and resource development is devoted to the two major polyploid crop species, i.e., Gossypium hirsutum and Gossypium barbadense, the cultivated diploid species Gossypium herbaceum and Gossypium arboreum comprise a significant share of the cotton market in certain countries (Wendel et al. 1989; Basu 1996; Guo et al. 2006; Khadi et al. 2010; Kranthi 2018). Native to Africa, these latter two species are nestled within a clade of additional African species that possess short nonspinnable fiber, but which may be valuable as sources of various disease and/or stress-resistant traits (Yik and Birchfield 1984; Rudgers et al. 2004; Nazeer et al. 2014; Rahman et al. 2017).

Gossypium stocksii is a diploid cotton species native to Eastern Africa whose subsection, Pseudopambak [E-genome cottons (Wang et al. 2018)], is thought to be earliest diverging lineage in the African clade (Wendel and Grover, 2015). E-genome cottons, including Gossypium stocksii (E₁), Gossypium somalense (E₂), Gossypium areysianum (E₃), and Gossypium incanum (E₄), may be sources of valuable traits including disease resistance. While both G. stocksii and G. somalense have resistance to reniform nematode (Yik and Birchfield 1984), only G. stocksii has reported resistance to cotton leaf curl disease (CLCuD) (Nazeer et al. 2014). Spread by white flies (Briddon and Markham 2000), the virus that causes CLCuD can have devastating effects on crop yield, as exhibited by the Pakistan epidemic in the early 1990s (Rahman et al. 2017), which resulted in massive financial losses over the course of 5 years. By some estimates, CLCuD is capable of decreasing total yield up to 90%, yet none of the major G. hirsutum cultivars exhibit resistance (Mammadov et al. 2018).

Because G. stocksii germplasm may be a useful source of resistance traits, interspecific material derived from crosses between G. stocksii and the commercially important G. hirsutum have been evaluated for a number of traits, including resistance to CLCuD and possible improvements in fiber. Research has shown that the F1 generation of a doubled G. stocksii × G. hirsutum cross not only has resistance to CLCuD but also exhibits increased fiber strength relative to the parents (Nazeer et al. 2014). More recently, comparisons among hexaploid hybrids derived from crosses between G. hirsutum and other wild diploid species suggests that four wild diploid species, including G. stocksii, are potentially valuable for fiber breeding programs (Konan et al. 2020).

Although there has been interest in G. stocksii for breeding purposes, genomic resources are virtually nonexistent for this species. Here, we describe a high-quality de novo genome sequence for G. stocksii, a valuable source of disease resistance in cotton and a potential source for improving fiber in domesticated cotton.

Materials and methods

Plant material and sequencing methods

Mature leaves from G. stocksii (E₁) grown under greenhouse conditions at Brigham Young University (BYU) were collected for PacBio sequencing. A CTAB-based method was used to extract high-quality DNA (Kidwell and Osborn 1992), which was quantified on a Qubit Fluorometer (ThermoFisher, Inc.; Waltham, MA, USA). A BluePippen instrument (Sage Science, LLC; Beverly, MA, USA) was used to size-select for only fragments >18 kb, as verified using a Fragment Analyzer (Advanced Analytical Technologies, Inc; Ankeny, IA, USA). Size-selected DNA was sent to the BYU DNA Sequencing Center (DNASC; Provo, UT, USA) for PacBio (Pacific Biosciences; Menlo Park, CA) library construction and sequencing on a total of 20 PacBio cells. Canu V1.6 was used to assemble the raw sequencing reads using default parameters (Koren et al. 2017).

Young leaf tissue was also used for DNA extraction and HiC library construction (Belton et al. 2012) by PhaseGenomics LLC (Seattle, WA, USA). These HiC libraries were sequenced on an Illumina HiSeq 2500 (2 × 125 bp) at the BYU DNASC. Resulting HiC reads were used to join contigs, and the association frequency between paired-ends was used to correct the assembly using JuiceBox (Durand et al. 2016). The final genome sequence of G. stocksii was generated via a custom python script available through PhaseGenomics LLC, yielding 13 assembled chromosomes.

Repeat and gene annotation

Transposable elements (TEs) were annotated using a combination of RepeatMasker (Smit et al. 2015) and “One code to find them all” (Bailly-Bechet et al. 2014). A custom library of Repbase 23.04 (Bao et al. 2015) was combined with cotton-specific repeats (Grover et al. 2020) to mark repeats in the genome using RepeatMasker. Adjacent matches were merged using “One code to find them all,” and the output was aggregated and summarized in R/4.0.3 (R Core Team 2017) using dplyr /0.8.1 (Wickham et al. 2015). All codes are available at https://github.com/Wendellab/stocksii (last accessed 4/23/2021).

The G. stocksii genome was annotated using existing RNA-seq data from various tissues of closely related species (Supplementary Table S1). Specifically, the following tissues were used: G. arboreum developing seeds and seedling (SRR617075, SRR617073, SRR617068, SRR617067, and SRR959508), Gossypium davidsonii roots and leaves (SRR2132267), G. herbaceum seed and developing fiber (SRR959585, SRR10675236, SRR10675235, SRR10675234, and SRR10675237), Gossypium longicalyx leaf, stem, and flower (SRR1174182, SRR1174179, SRR6327757, SRR6327758, and SRR6327759), Gossypium raimondii leaf, seed, stem, petal, meristem, and floral tissues (SRR617009, SRR617011, SRR617013, SRR8267554, SRR8267566, SRR8878565, SRR8878526, SRR8878661, SRR8878800, SRR8878534, and SRR8878745), Gossypium thurberi leaf, root, and stem (SRR8267623, SRR8267616, and SRR8267619), and Gossypium trilobum leaf, root, and stem (SRR8267606, SRR8267582, and SRR8267601). Each library was downloaded from the Short Read Archive (SRA), and all RNA-seq data were mapped to the hard-masked G. stocksii genome using hisat2 [v2.1.0] (Kim et al. 2015). BRAKER2 [v2.1.2] (Hoff et al. 2019) was trained with GeneMark [v4.38] (Borodovsky and Lomsadze 2011) generated annotations, which were also used to train Augustus [v3.3.2] (Stanke et al. 2006). StringTie [v2.1.1] (Pertea et al. 2015) and Cufflinks [v2.2.1] (Ghosh et al. 2016) generated de novo RNA-seq assemblies were combined with a Trinity [v2.8.6] (Grabherr et al. 2011) reference-guided assembly and splice junction information from Portcullis [v1.2.2] (Mapleson et al. 2018) in Mikado [v1.2.4] (Venturini et al. 2018). MAKER2 [v2.31.10] (Holt and Yandell 2011; Campbell et al. 2014) was used to integrate gene predictions from (1) BRAKER2 trained Augustus, (2) GeneMark, and (3) Mikado, also using evidence from all Gossypium ESTs available from NCBI (nucleotide database filtered on “txid3633” and “is_est”) and a database composed of all curated proteins in Uniprot SwissProt [v2019_07] (UniProt Consortium 2008) combined with the annotated proteins from the G. hirsutum (https://www.cottongen.org/species/Gossypium_hirsutum/jgi-AD1_genome_v1.1, last accessed 4/23/21) and G. raimondii (Paterson et al. 2012) genomes. SNAP [v2013-02-16] and Augustus were trained with the predicted annotations from Maker. Maker was run a second time with the newly trained Augustus and SNAP models, along with the other inputs from the first iterations. Annotation edit distance (AED) ( Eilbeck et al. 2009; Holt and Yandell 2011; Yandell and Daniel 2012) was used to score each gene model relative to EST and protein evidence, and gene models with an AED <0.35 were retained. Gene models were functionally annotated using InterProScan [v5.47-82.0] (Jones et al. 2014) and BlastP [v2.9.0+] (Camacho et al. 2009) searches against the Uniprot SwissProt database. Orthologous relationships between G. stocksii and other diploid cottons were determined via OrthoFinder (Emms and Kelly 2015, 2019). Proteins from G. longicalyx (Grover et al. 2020), G. arboreum (Li et al. 2014; Du et al. 2018; Huang et al. 2020), G. herbaceum (Huang et al. 2020), G. raimondii (Paterson et al. 2012; Udall et al. 2019a), G. turneri (Udall et al. 2019a), and G. australe (Cai et al. 2020) were downloaded from CottonGen (https://www.cottongen.org; Yu et al. 2014, last accessed, 4/23/21) and run using default parameters. Code is available from https://github.com/Wendellab/stocksii.

Comparison to G. somalense

Three DNA libraries of G. somalense (E₂; SRA: SRR3560160-SRR3560162), a close relative of G. stocksii (Chen et al. 2016), were used to provide a preliminary comparison of the two species. Raw reads were mapped to the newly generated G. stocksii genome using the Spack (Gamblin et al. 2015) implementation of bwa v0.7.17-rgxh5dw (Li and Durbin 2009). Single-nucleotide polymorphisms (SNPs) in G. somalense were called relative to G. stocksii using the Sentieon pipeline (Kendig et al. 2019) (Spack version sentieon-genomics/201808.01-opfuvzr), which is an optimization of existing methods, such as Genome Analysis ToolKit (GATK) (McKenna et al. 2010). This pipeline included read deduplication, indel realignment, and genotyping. The three libraries represent technical replicates of the G. somalense sequencing and were therefore merged after read deduplication. Parameters for mapping and SNP calling follow standard practices, and are available in detail at https://github.com/Wendellab/stocksii. The resulting variant file was filtered for read depth using vcftools (Spack version 0.1.14-v5mvhea) (Danecek et al. 2011), only retaining sites with a minimum of 10 reads and a maximum of 100 reads. GenomeTools (Gremme et al. 2013) was used to convert the annotation file to gtf format, which was used in conjunction with SnpEff (Cingolani et al. 2012) to annotate and predict the effects of the SNP differences between G. stocksii and G. somalense.

Divergence between the two species was estimated using –window-pi from vcftools in 100 kb, nonoverlapping windows, which estimated the average number of differences per window. Diversity was parsed by region by first intersecting the filtered VCF with the relevant feature (e.g., exon, intron, etc.) from the G. stocksii annotation using intersectBed from bedtools2 (Spack version 2.27.1-s2mtpsu) (Quinlan 2014) to get a list of SNP sites associated with that region. The original, filtered VCF was then used in conjunction with vcftools –window-pi and the flag –positions, which limits the analysis to only the specified sites (e.g., exon, intron, intergenic). Diversity/divergence results were parsed in R/4.0.3 using dplyr (Wickham et al. 2015) and plotted using ggplot2 (Wickham 2016). Relevant code and detailed pipeline analysis can be found at https://github.com/Wendellab/stocksii.

To provide a comparative framework for qualitative interpretation of the amount of divergence between G. stocksii and G. somalense, two other species pairs (i.e., G. herbaceum–G. arboreum and G. raimondii–G. gossypioides) were also subjected to SNP calling/filtering and calculation of π in 100-kb windows, as outlined above for G. stocksii–G. somalense. Here, the genome of G. herbaceum (Huang et al. 2020) was used as a reference for G. arboreum reads (SRR8979980; Page et al. 2013), and G. raimondii (Udall et al. 2019a) was used as a reference for reads from G. gossypioides (SRR3560148 and SRR3560149). Genomes and annotations were both downloaded from CottonGen (Yu et al. 2014).

Data availability

The G. stocksii genome sequence is available at NCBI under PRJNA701967 and through CottonGen (https://www.cottongen.org/). Raw data are available from the SRA under PRJNA701967. Supplementary files are available from figshare: https://doi.org/10.25387/g3.14080361.

Results and discussion

Genome assembly and annotation

We report a high-quality de novo genome sequence for G. stocksii covering 93% of the 1531-Mb genome (Hendrix and Stewart 2005). PacBio reads (58Χ coverage) were initially assembled into 316 contigs with an N50 of 17.8 Mb. These contigs were then ordered and oriented using both HiC and Bionano evidence to produce a chromosome level assembly (n = 13) with an average length of 110 Mb (1424 Mb total) and containing only 5.7 kb of gap sequence across all chromosomes. BUSCO (Waterhouse et al. 2018) analysis of the genome (Table 1) indicates a general completeness with only 2.4% of BUSCOs either fragmented (0.9%) or missing (1.5%). Over 97% complete BUSCOs were recovered, most of which were single copy (88.9%, versus 8.7% duplicated). The LTR Assembly Index (LAI) (Ou et al. 2018) was also within guidelines for “reference-quality” genomes (LAI = 10–20; G. stocksii LAI = 15.4), and dotplots (Figure 1) with existing high-quality cotton genome assemblies (Paterson et al. 2012; Du et al. 2018; Udall et al. 2019a, 2019b; Grover et al. 2020; Huang et al. 2020) further indicates the high-quality nature of this genome.

Table 1.

BUSCO results for the genome and annotation

	Genome	Annotation
Complete BUSCOs (C)	2,271 (97.6%)	2,227 (95.8%)
Complete and single-copy BUSCOs (S)	2,068 (88.9%)	1,888 (81.2%)
Complete and duplicated BUSCOs (D)	203 (8.7%)	339 (14.6%)
Fragmented BUSCOs (F)	20 (0.9%)	26 (1.1%)
Missing BUSCOs (M)	35 (1.5%)	73 (3.1%)
Total BUSCO groups searched	2,326

Figure 1.

Pairwise comparisons of G. stocksii with G. herbaceum (A1; Huang et al. 2020), G. raimondii (D5; Udall et al. 2019a), G. longicalyx (F1; Grover et al. 2020), G. arboreum (A2; Huang et al. 2020), G. turneri (D10; Udall et al. 2019a), and G. australe (G2; Cai et al. 2020).

Annotation of the G. stocksii genome revealed 37,889 transcripts representing 34,928 unique genes, similar to other cotton diploid genomes ( range 37,505 to 43,952; Paterson et al. 2012; Du et al. 2018; Udall et al. 2019a; Grover et al. 2020; Huang et al. 2020). BUSCO analysis of the annotation (Table 1) exhibits recovery, similar to the whole-genome BUSCO. Ortholog analysis between G. stocksii and these previously published cotton diploids produces 23,399 orthogroups (Supplementary File S1) containing at least one G. stocksii gene (range 18,785 in G. australe to 27,913 in G. arboreum; Huang et al. 2020), comprising 68.5% of the total orthogroups. Notably, five species-specific orthogroups were recovered containing a total of 68 genes (Table 2), 62 of which are argonaute-like proteins (Supplementary Table S2). On average, over half of the transcripts (22,403) are placed in a simple 1:1 relationship in pairwise comparisons between G. stocksii and another cotton diploid genome (Supplementary Table S3).

Table 2.

Orthogroup relationships between G. stocksii and other cotton diploid genomes

	G. stocksii	G. arboreum			G. herbaceum	G. raimondii			G. turneri	G. longicalyx	G. australe	G. anomalum
		Li (2014)	Du (2018)	Huang (2020)	Huang (2020)	Paterson (2012)	Wang et al. (2012)	Udall (2019a)	Udall (2019a)	Grover (2020)	Cai (2020)	Unpublished
Number of genes	37,889	40,134	40,960	43,278	43,952	37,505	40,976	41,030	38,871	38,378	38,281	38,480
Genes in orthogroups	36,005 (95%)	37,824 (94%)	40,502 (99%)	42,521 (98%)	42,665 (97%)	36,702 (98%)	38,124 (93%)	37,116 (91%)	35,297 (91%)	36,324 (95%)	34,492 (90%)	36,720 (95%)
Unassigned genes	1,884 (5%)	2,310 (6%)	458 (1%)	757 (2%)	1,287 (3%)	803 (2%)	2,852 (7%)	3,914 (10%)	3,574 (10%)	2,054 (5%)	3,789 (10%)	1,760 (5%)
Orthogroups including species representatives	23,399 (69%)	26,541 (78%)	27,330 (80%)	27,146 (80%)	27,913 (82%)	26,323 (77%)	25,498 (75%)	24,944 (73%)	25,286 (74%)	24,800 (73%)	18,785 (55%)	24,186 (71%)
Species-specific orthogroups	5	8	0	1	6	1	13	3	7	5	13	3
Genes in species- specific orthogroups	68	36	0	2	13	7	59	11	16	23	60	9

TE content was assessed by de novo TE prediction via RepeatMasker (Bailly-Bechet et al. 2014; Smit et al. 2015), indicating that repeats occupy ∼43% of the 1531-Mbp genome (Table 3). Consistent with other plant genomes, Ty3/gypsy predominate the G. stocksii genome, comprising over 90% of the detected repetitive elements. Ty1/copia elements and DNA elements (as a whole) were substantially less represented, accounting for only 43 and 13 Mb, respectively, in the present analysis.

Table 3.

Repeat types and predicted copy numbers in the G. stocksii genome

Element type	Fragments	Copies	SoloLTR	Total_Mb
DNA	15,047	9,190	0	13.37
DNA/EnSpmCACTA	1,138	682	0	2.02
DNA/Harbinger	1	1	0	0.00
DNA/hAT	1,858	1195	0	0.84
DNA/hAT-Tip100	18	11	0	0.02
DNA/L1	923	461	0	1.08
DNA/MarinerTc1	71	39	0	0.05
DNA/MuDR	11,030	6,797	0	9.36
DNA/MULE-MuDR	6	3	0	0.00
DNA/PIF-Harbinger	2	1	0	0.00
LTR	906,998	487,232	269,000	650.36
LTR	34	33	0	0.00
LTR/Copia	45,806	27,117	9,567	42.96
LTR/Gypsy	861,158	460,082	259,433	607.39
Total	922,045	496,422	0	663.73

Comparison of G. stocksii with G. somalense

Gossypium stocksii is part of a clade of approximately seven species (subsection Pseudopambak), but relatively little is known about the members of this subsection, including questions regarding species circumscription and the possibility of unrecognized taxa (Fryxell 1979, 1992; Vollesen 1987). A comparison between Gossypium stocksii and the closely related G. somalense (Supplementary Figure S1) reveals considerable divergence between these two species, with 39.7 M interspecific SNPs evenly distributed among the 13 chromosomes (Table 4). As expected, most of the variation (94% or 37.1 M SNPs) is found in the intergenic space, only 30% of which is found near genes (±5 kb up- or down-stream). An assessment of nucleotide distance between G. stocksii and G. somalense (here measured as π in VCFtools) reveals a modest distance between these two species (mean π = 0.0116; 100-kb windows) that is intermediate between the very closely related sister species G. arboreum and G. herbaceum (Renny-Byfield et al. 2016; Huang et al. 2020) and the more distantly related species G. gossypioides and G. raimondii (subgenus Houzingenia; Grover et al. 2019). On a per-chromosome basis, the pairwise G. stocksii–G. somalense π estimates range from an average of 0.0098 on chromosome E05 to 0.0126 on chromosome E03 (Figure 2).

Table 4.

Comparison of G. somalense resequencing with the G. stocksii genome

Chromosome	Length	Number of variants	Variant rate (%)	Average pi
E01	116,888,287	3,307,654	2.83	0.0124
E02	88,432,363	2,545,417	2.88	0.0120
E03	125,861,959	3,604,952	2.86	0.0126
E04	106,357,252	3,070,741	2.89	0.0124
E05	106,784,523	2,598,022	2.43	0.0098
E06	118,523,473	3,312,902	2.80	0.0120
E07	94,613,334	2,855,058	3.02	0.0118
E08	122,663,403	3,390,477	2.76	0.0117
E09	82,831,125	2,285,194	2.76	0.0106
E10	114,014,049	3,343,978	2.93	0.0120
E11	117,232,604	3,111,632	2.65	0.0104
E12	114,623,165	3,068,541	2.68	0.0112
E13	115,591,314	3,227,954	2.79	0.0121
Total	1,424,416,851	39,722,522	2.79	0.0116
Total (genic)	128,641,547	2,578,738	2.00	0.0007
	SNP location	Number of variants	Proportion of variants (%)
	Intergenic	37,188,781	93.62
	Upstream	5,461,093	13.75
	Downstream	5,121,580	12.89
			0.00
	Exon	831,986	2.09
	Missense	471,909	1.19
	Silent	352,203	0.89
	Nonsense	14,780	0.04
	Intron	1,746,752	4.40
	UTR, 5'	63,862	0.16
	UTR, 3'	71,297	0.18

Figure 2.

Pairwise comparisons of π for G. somalense and G. stocksii, with G. arboreum vs G. herbaceum and G. gossypioides vs G. raimondii for comparison. Here, π is calculated individually for each species pair for the entire dataset (all) and for the specified subset of SNPs (i.e., exonic, genic, intragenic, and intronic). Colors reflect the individual comparisons, with lines to represent the mean and points to represent outliers. Because π is calculated between two samples each, the values here reflect the pairwise divergence between samples.

Although genic regions have far fewer SNPs, SNPs in these regions still account for 2.6 M of the 39.7 M total (Table 3). Intron-based SNPs outweigh exon-based SNPs in a 2:1 ratio, accounting for 4.4% and 2.1% of the overall SNPs, respectively. Most exon-based SNPs are minimally disruptive, either conferring silent (352,203) or missense (471,909) changes (Table 3); very few (14,780) produced predicted nonsense changes. Similar to other species pairs in Gossypium, the average nucleotide distance in genes was far lower than the overall distance (0.007 vs 0.0116, respectively), indicating a close relationship between these two species in their gene space. Given that G. somalense does not exhibit the sample level of resistance to CLCuD (Nazeer et al. 2014; Anjum et al. 2015), but does show other forms of pest resistance (Yik and Birchfield 1984; Shim et al. 2018), future comparisons including multiple accessions of both species may shed insight into the evolution of natural pest resistance in cotton species.

Gossypium stocksii as a resource for disease resistance

Whereas domesticated varieties of G. hirsutum are highly susceptible to CLCuD (Rehman et al. 2017), G. stocksii exhibits natural resistance (Nazeer et al. 2014). The molecular basis of CLCuD resistance in cotton is not well understood (Rahman et al. 2017), although genetic analyses indicate that CLCuD resistance is likely controlled by one or few dominant genes with possible epistatic modifiers (Knight 1948; Ali 1997; Haidar et al. 2003; Rahman et al. 2005; Ahuja et al. 2006), thereby making it a prime target for breeding programs and/or genetic modification. While the success of CRISPR/Cas9 in controlling similar viral diseases and the continued lack of success in controlling CLCuD using conventional methods (Iqbal et al. 2016) has piqued interest in genome modification enhancing resistance, little research has focused on the genomic basis of CLCuD resistance.

Preliminary research in a CLCuD-resistant accession of G. arboreum identified 1062 differentially expressed genes (DEG) between challenged and unchallenged plants (Naqvi et al. 2017), 17 of which were considered prime candidates for conferring disease resistance. Of those 17 genes, 16 were placed in orthogroups that also contained one or more G. stocksii homologs (Table 5), with the sole exception of the gene putatively encoding “phytosulfokines 3” (i.e., Cotton_A_25246_BGI-A2_v1.0), which plays a role in pathogen response in lotus (Wang et al. 2015). Most orthogroups were comparable in size between the G. arboreum genome used to detect DEG and our G. stocksii annotation, aside from OG0000284 (the cysteine protease ervatamin-B like genes), which was composed of five tandemly arrayed genes in G. arboreum, but only two in G. stocksii; the relevance of these genes to CLCuD defense is unclear. The largest orthogroup that contained one of the top DEG candidates was orthogroup OG0000074, which is composed of resistance gene (i.e., R-gene) analogs (Naqvi et al. 2017); notably, G. stocksii appears to have one additional copy of this gene. Similarity at the protein level between the G. arboreum DEG and its closest G. stocksii homolog is generally high (i.e., 95%, on average), although it drops as low as 73.4% in the poorly conserved ervatamin-B like orthogroup (Table 5). These results indicate that similar genes may operate in CLCuD resistance in G. stocksii; however, comparative expression data from infected and uninfected plants are required to understand whether the two species use similar pathways to avoid infection by the CLC virus.

Table 5.

Ortholog identification for previously identified CLCuD candidates and copy numbers in other published genomes

CLCuD candidatea	Total genes	Average per genome	Closest match	Protein similarity (%)	Orthogroup	G. arboreum Li et al. (2014)	G. stocksii	G. herbaceum Huang et al. (2020)	G. arboreum Du et al. (2018)	G. arboreum Huang et al. (2020)	G. raimondii Wang et al. (2012)	G. raimondii Paterson et al. (2012)	G. raimondii Udall et al. (2019a)	G. turneri Udall et al. (2019a)	G. longicalyx Grover et al. (2020)	G. anomalum unpublished	G. australe Cai et al. (2020)
Cotton_A_03097 _BGI-A2_v1.0	140	11.7	Gosto.013G 220600-1	88.6	OG0000074	8 (6 + 1  + 1)	9 (2 + 1 + 1 + 4)	12 (2 + 1 + 1 + 8)	15 (1 + 1 + 12)	12 (1 + 1 + 10)	12 (2 + 1 + 1 + 8)	6 (6)	9 (1 + 8)	31 (31)	15 (1 + 1 + 13)	10 (1 + 9)	1
Cotton_A_22786 _BGI-A2_v1.0	63	5.3	Gosto.005 G342100-1	73.4	OG0000284	5 (5)	2 (2)	5 (5)	5 (5)	5 (5)	6 (6)	7 (7)	6 (6)	7 (7)	6 (6)	6 (6)	3 (3)
Cotton_A_34570 _BGI-A2_v1.0	43	3.6	Gosto.006 G064800-1	87.6	OG0000679	3	4 (1 + 3)	5 (1 + 4)	5 (1 + 3 + 1)	5 (1 + 4)	4 (1 + 1 + 2)	5 (3 + 2)	2	4 (1 + 3)	2	3	1
Cotton_A_07057 _BGI-A2_v1.0	43	3.6	Gosto.005 G102100-1	100.0	OG0000687	4	2	3	4	4	4	4	4	4	3	4	3
Cotton_A_11914 _BGI-A2_v1.0	38	3.2	Gosto.005 G051300-1, Gosto.013 G208400-1	100.0	OG0000861	3	2	4	4	4	4	4	2	4	2	4	1
Cotton_A_23939 _BGI-A2_v1.0	30	2.5	Gosto.001 G196400-1	95.6	OG0001667	3	2	1	3	3	3	3	2	3	3	4	1
Cotton_A_28295 _BGI-A2_v1.0	23	1.9	Gosto.006 G027000-1	97.9	OG0002853	2	2	2	2	2	2	2	2	2	2	4	0
Cotton_A_40086 _BGI-A2_v1.0	22	1.8	Gosto.011 G284200-1	95.9	OG0004295	2	1	2	2	2	2	2	2	2	2	2	1
Cotton_A_19720 _BGI-A2_v1.0	21	1.8	Gosto.013 G033400-1	100.0	OG0004345	2	2	2	2	2	2	2	1	1	2	2	1
Cotton_A_00757 _BGI-A2_v1.0	14	1.2	Gosto.003 G217500-1	97.7	OG0006575	2 (2)	1	1	1	1	1	1	1	2 (2)	1	1	1
Cotton_A_00151 _BGI-A2_v1.0	12	1.0	Gosto.001 G010500-1	98.2	OG0008663	1	1	1	1	1	1	1	1	1	1	1	1
Cotton_A_00108 _BGI-A2_v1.0	12	1.0	Gosto.001 G014500-1	97.7	OG0008676	1	1	1	1	1	1	1	1	1	1	1	1
Cotton_A_19100 _BGI-A2_v1.0	12	1.0	Gosto.007 G179500-1	96.0	OG0012201	1	1	1	1	1	1	1	1	2 (2)	1	1	0
Cotton_A_29104 _BGI-A2_v1.0	12	1.0	Gosto.009 G260900-1	97.2	OG0013132	1	1	1	1	1	1	1	1	1	1	1	1
Cotton_A_01472 _BGI-A2_v1.0	12	1.0	Gosto.012 G280100-1	95.2	OG0015696	1	1	1	1	1	1	1	1	1	1	1	1
Cotton_A_25246 _BGI-A2_v1.0	11	0.9	none	—-	OG0018038	1	0	1	1	1	1	1	1	1	1	1	1
Cotton_A_17591 _BGI-A2_v1.0	10	0.8	Gosto.004 G048300-1	98.0	OG0021151	1	1	1	1	1	0	1	0	1	1	1	1

Numbers in parentheses indicate the number(s) of genes from that orthogroup that are genomically clustered. Orthogroups without parenthetical notation indicate all members from that species are genomically dispersed.

^aCandidates derived from Naqvi et al. (2017); reference genome is Li et al. (2014).

Conclusion

Cotton leaf curl virus is an important cotton pathogen that results in thickening and yellowing of small leaf veins, ultimately leading to the characteristic leaf “curling” phenotype, as well as stunted growth, delayed onset of flowering and/or fruiting, and reductions in yield quantity and quality (Rahman et al. 2001; Farooq et al. 2015; Rehman et al. 2017). Here, we report a genome sequence for Gossypium stocksii, one of the poorly understood “E-genome” species, which is also a source of CLCuD resistance. This resource provides a new foundation for understanding CLCuD resistance in cotton and represents a new resource for future evolutionary and taxonomic work in this group of cotton species.