Genomic characterization of a MAGIC population containing introgressions from four tetraploid Gossypium species

Abstract

One strategy for increasing the genetic diversity of upland cotton (Gossypium hirsutum) involves the introgression of beneficial alleles from related and wild tetraploid species. Progress in this area has been limited due to undesirable outcomes associated with whole genome interspecific hybridization. To minimize genetic influences from other chromosomes, chromosome substitution lines (CSLs) were developed in upland cotton TM‐1 using Gossypium barbadense (GB), Gossypium mustelinum (GM), and Gossypium tomentosum (GT) as donor species. A multiparent advanced generation intercross (MAGIC) population comprising 797 recombinant inbred lines (RILs) was developed by crossing 32 CSLs with five upland cotton cultivars. Whole‐genome sequencing of all parental lines and MAGIC RILs was performed to assess alien genome introgressions. Of the 32 CSLs, 28 exhibited targeted chromosome substitutions, although residual alien genome introgressions were prevalent across all lines. Among the 797 MAGIC RILs, 794 displayed varying levels of introgressed segments, with each RIL containing introgressions in an average of nine chromosomes. GB‐derived introgressions predominated in 356 RILs, suggesting that GB alleles outcompeted GM or GT alleles during population development. Kinship analysis revealed a clear structure within the MAGIC population, which could be categorized into four groups. Despite five cycles of random mating, large alien genome segments persisted in many RILs, highlighting a continuing challenge posed by linkage drag in interspecific hybridization. This study offers new perspectives on the complexities of utilizing wild species in cotton improvement and provides valuable genetic materials for future breeding efforts.

Core Ideas

• A total of 32 chromosome substitution lines (CSLs) from three tetraploid donor species were crossed with five upland cotton cultivars, resulting in the development of a multiparent advanced generation intercross (MAGIC) population comprising 797 recombinant inbred lines (RILs).

• The genomes of 37 parental lines and 797 RILs were sequenced at 25–40× and 10× genome coverage, respectively.

• Alien genome introgression in each of 32 chromosome substitution lines (CSLs) and 797 MAGIC RILs was analyzed using donor species‐specific single nucleotide polymorphisms (SNPs). Twenty‐eight CSLs contained the targeted chromosome substitutions, while the remaining four CSLs failed to obtain the expected chromosome substitutions.

• Seven hundred ninety‐four MAGIC RILs exhibited different degrees of alien genome introgression. Identity by state kinship analysis showed that the population could be divided into four distinct groups.

Plain Language Summary

Increasing the genetic diversity of upland cotton remains a critical objective for cotton geneticists and breeders. One strategy involves the introgression of beneficial alleles from related and wild tetraploid species into upland cotton. Traditional interspecific hybridization has yielded limited progress, primarily due to deficiencies in progeny such as hybrid breakdown, pronounced segregation distortion, and deleterious linkage drag. A chromosome substitution line (CSL) is defined as a genetic line in which a single chromosome is replaced by that of a donor species. In this study, 32 CSLs were crossed with five upland cotton cultivars, resulting in the development of a multiparent advanced generation intercross (MAGIC) population comprising 797 recombinant inbred lines (RILs). Genome sequencing was utilized to assess alien genome introgressions within both CSLs and MAGIC RILs. The outcomes of this research provide a foundation for future genome‐wide association study of agronomic traits derived from related and wild tetraploid species using this MAGIC population.

Abbreviations

CS‐B

chromosome substitution from Gossypium barbadense

CSL

chromosome substitution line

CS‐M

chromosome substitution from Gossypium mustelinum

CS‐T

chromosome substitution from Gossypium tomentosum

GB

Gossypium barbadense

GH

Gossypium hirsutum

GM

Gossypium mustelinum

GT

Gossypium tomentosum

GWAS

genome‐wide association study

IBD

identity by descent

IBS

identity by state

InDel

insertion/deletion

MAGIC

multiparent advanced generation intercross

RIL

recombinant inbred line

SNP

single nucleotide polymorphism

UPGMA

unweighted pair‐group method with arithmetic mean

1. INTRODUCTION

In 2025, global cotton production reached about 119.8 million bales (218 kg per bale) (Hudson & Liu, 2025; Meyer & Dew, 2026). Upland cotton (Gossypium hirsutum L.) made up roughly 97% of the total, while Gossypium barbadense (GB) L. (Pima cotton or Egyptian cotton) and two diploid species (Gossypium arboreum L. and Gossypium herbaceum L.) accounted for the remaining 3%. As a result, most cotton breeding efforts focus on improving upland cotton cultivars.

G. hirsutum and GB are both allotetraploid species (AD) with 52 chromosomes each (2n = 52). In addition to these two cultivated species, the Gossypium genus also includes five other noncultivated tetraploid species (Gallagher et al., 2017). Research suggested that all seven tetraploid species originated from a single hybridization event between diploid A and D genome species about 1–2 million years ago (Gallagher et al., 2017; Wendel & Grover, 2015). Collectively, these seven species are considered as the primary gene‐pool for upland cotton improvement.

Many studies showed that upland cotton has a narrow genetic base and limited diversity due to its origin and long history of domestication (Shrestha et al., 2025; Tyagi et al., 2014). Modern cultivars have even less diversity as a result of strong human selection, repeated use of a few elite cultivars as parents in breeding, and introgression of transgenic traits via backcrossing (Esbroeck & Bowman, 1998; D. Fang et al., 2013; Y. Li et al., 2023; Y. Liu et al., 2022; Safiullina et al., 2024). Broadening the genetic base and increasing diversity of upland cotton remain key goals for cotton breeders.

The idea of using wild tetraploid species to expand the genetic base of upland cotton originated in the United States in 1930s, as discussed by Niles and Feaster (1984). These tetraploid species possess agronomic traits valuable for improving upland cotton, such as superior fiber qualities found in GB and high tolerance to abiotic (heat and drought) and biotic stresses (jassids and thrips) in Gossypium tomentosum (GT) Nutt. ex Seem. (Saha, Raska, et al., 2006; Shen et al., 2021; Waghmare et al., 2005). Additionally, alleles from Gossypium mustelinum (GM) Watt have been identified to be able to enhance fiber quality in upland cotton (Chen, Wang, et al., 2020). These tetraploid species share a similar AD genomic structure and can be readily crossed with G. hirsutum, to produce viable offsprings.

For nearly a century, researchers have aimed to transfer beneficial agronomical traits especially superior fiber quality from GB into upland cotton through hybridization and selection (Niles & Feaster, 1984). Efforts to move beneficial agronomical traits from noncultivated tetraploid species such as GM and GT into upland cotton began about three decades ago (Saha, Raska, et al., 2006; Saha et al., 2013; Waghmare et al., 2005). Although interspecific hybridization was mostly successful and F₁ progeny plants were fertile, breeders observed a variety of undesirable phenomena in the subsequent generations such as hybrid breakdown, deleterious linkage drag, genome incompatibility, severe segregation distortion and gradual loss of the introgressed alleles (Anwar et al., 2022; P. Wang et al., 2012; J. Yu et al., 2012; Y. Yu et al., 2011; J. Zhang et al., 2014). To date, there is not a single variety that is known to be bred through interspecific hybridization. This underscores a need to look beyond the whole genome interspecific hybridization and investigate alternative strategies to transfer beneficial alleles from other tetraploid species into upland cotton.

Two primary approaches were utilized. The first was to create chromosome substitution lines (CSL) or chromosome segment substitution lines (Jenkins et al., 2006; Saha, Jenkins, et al., 2006; Saha et al., 2004). The second was to create chromosome segment introgression lines (L. Fang, Tian, et al., 2014; Sanamyan et al., 2022; Song et al., 2017). These two approaches differed greatly in development methodologies. The former used a chromosome‐specific hypoaneuploid as a donor parent (more detail below), the latter first made a cross between an upland cotton variety and a GB (or other species) line, then backcrossed four or five cycles using the upland cotton variety as the recurrent parent. DNA markers were used to track the intregressions in each backcross cycle. The materials used in this research were developed using the first approach.

Hypoaneuploid refers to the loss of a chromosome, or part of a chromosome. After years of study, cotton researchers succeeded in identifying and creating hypoaneuploids for most of the 26 chromosomes (Endrizzi & Ramsay, 1980; Endrizzi et al., 1984; Myles & Endrizzi, 1989). In 1990s, Stelly and coworkers began to use these hypoaneuploids to develop CSLs in the upland cotton TM‐1 background. In 2005, Stelly et al. (2005) released 17 CSLs in TM‐1 containing a chromosome or segment from GB Pima cotton 3–79. These lines were named as chromosome substitution from G. barbadense (CS‐B). Later, Saha et al. developed CSLs from GT (CS‐T) in 2006 and from GM (CS‐M) in 2013, both in TM‐1 background (Saha, Raska, et al., 2006; Saha et al. 2013). The CS‐B lines underwent field evaluations and some demonstrated significant improvement in yield and fiber quality, for instance, CS‐B02 had better lint percentage and higher fiber strength than TM‐1 (Jenkins et al., 2007; Saha et al., 2004). To date, field evaluations of CS‐M and CS‐T lines have not yet been reported.

Following the development of the CSLs mentioned earlier, Jenkins and colleagues crossed 12 CS‐B, 8 CS‐M, and 12 CS‐T lines with five upland cotton cultivars. They then random‐mated for five generations, and released this as a new population under the name Random Mated Barbadense, Hirsutum, Mustelinum, Tomentosum, Upland Population (RMBHMTUP‐C4; Reg. no. GP‐1061, PI 690471) (Jenkins et al., 2019). We further self‐pollinated five generations via single seed decent and obtained a new multiparent advanced generation intercross (MAGIC) population consisting of 797 recombinant inbred lines (RILs). We sequenced the genomes of the 37 parent lines and the 797 MAGIC RILs. Our goal will be to use this MAGIC population to identify quantitative trait loci for important agronomic traits derived from GB, GM, or GT through a genome‐wide association study (GWAS). Although we have begun collecting agronomic trait data for this population, currently very limited information is available. As the first research report from this project, this paper focuses on using the whole genome sequences to analyze introgression levels in the chromosome substitution parental lines and the entire 797 MAGIC RILs, as well as to evaluate the structure of the MAGIC population.

Core Ideas

• A total of 32 chromosome substitution lines (CSLs) from three tetraploid donor species were crossed with five upland cotton cultivars, resulting in the development of a multiparent advanced generation intercross (MAGIC) population comprising 797 recombinant inbred lines (RILs).

• The genomes of 37 parental lines and 797 RILs were sequenced at 25–40× and 10× genome coverage, respectively.

• Alien genome introgression in each of 32 chromosome substitution lines (CSLs) and 797 MAGIC RILs was analyzed using donor species‐specific single nucleotide polymorphisms (SNPs). Twenty‐eight CSLs contained the targeted chromosome substitutions, while the remaining four CSLs failed to obtain the expected chromosome substitutions.

• Seven hundred ninety‐four MAGIC RILs exhibited different degrees of alien genome introgression. Identity by state kinship analysis showed that the population could be divided into four distinct groups.

2. MATERIALS AND METHODS

2.1. Plant materials

Plant materials used in this research included 37 parental lines and 797 RILs derived from crosses between these parental lines. The 37 parental lines were 12 CS‐B, 8 CS‐M, 12 CS‐T lines, and five upland cotton cultivars (Table 1). All CSLs were in the upland cotton TM‐1 background. In theory, each CSL was near isogenic to TM‐1, and to each other, except the substituted chromosome or chromosome segment. There are three types of chromosome substitutions: whole‐chromosome (primary monosomic, e.g., CS‐B02), arm‐specific (acrocentric, or monotelodisomic, e.g., CS‐B08sh), and double segmental (tertiary monosomic, e.g., CS‐B04‐15). The tertiary monosomic lines were obtained from rare nonhomologous chromosomal monosomic translocation events resulting from atomic radiation. The five upland cotton cultivars were developed by different seed companies in the United States or Australia. They were Deltapine Acala 90 (DP90) (PVP 8100143, by Delta & Pine Land company), FiberMax 966 (FM966) (PI 619097; PVP 200100209, by BASF company), Phytogen PSC 355 (PSC355) (PI 612974; PVP 200000167, by Corteva company), Sure Grow 747 (SG747) (PI 656375; PVP 9800118, by SureGrow company), and Stoneville 474 (ST474) (PI 578877; PVP 9400152, by Stoneville Pedigree company). Cotton chromosomes were previously named as chromosome (Chr.) 1 through Chr. 26. When tetraploid upland cotton genome sequences were first published in 2015 (F. Li et al., 2015; T. Zhang et al., 2015), the chromosomes were renamed to A01–A13 for A subgenome chromosomes, and D01–D13 for D subgenome chromosomes. The correspondence between the old and new chromosome nomenclatures is shown in Table 1.

TABLE 1.

The 32 chromosome substitution lines and five upland cotton cultivars used to create the BHMT multiparent advanced generation intercross (MAGIC) population.

#	CSL a	Donor species	Affected chromosome b	Affected chromosome c	Upland cotton parent d
1	CS‐B02	G. barbadense	Chr. 2	A02	SG747
2	CS‐B04‐15	G. barbadense	Chr. 4 and 15	A04 and D01	PSC355
3	CS‐B04	G. barbadense	Chr. 4	A04	SG747
4	CS‐B06	G. barbadense	Chr. 6	A06	ST474
5	CS‐B08sh	G. barbadense	Chr. 8	A08	ST474
6	CS‐B10‐19	G. barbadense	Chr. 10 and 19	A10 and D05	PSC355
7	CS‐B10	G. barbadense	Chr. 10	A10	FM966
8	CS‐B11sh	G. barbadense	Chr. 11sh	A11	DP90
9	CS‐B15sh	G. barbadense	Chr. 15sh	D01	FM966
10	CS‐B16‐15	G. barbadense	Chr. 16 and 15	D07 and D01	PSC355
11	CS‐B17	G. barbadense	Chr. 17	D03	PSC355
12	CS‐B22sh	G. barbadense	Chr. 22sh	D04	SG747
13	CS‐M02	G. mustelinum	Chr. 2	A02	DP90
14	CS‐M04‐15	G. mustelinum	Chr. 4 and 15	A04 and D01	SG747
15	CS‐M04	G. mustelinum	Chr. 4	A04	ST474
16	CS‐M06	G. mustelinum	Chr. 6	A06	FM966
17	CS‐M08sh	G. mustelinum	Chr. 8sh	A08	SG747
18	CS‐M15sh	G. mustelinum	Chr. 15sh	D01	ST474
19	CS‐M16‐15	G. mustelinum	Chr. 16 and 15	D07 and D01	SG747
20	CS‐M22sh	G. mustelinum	Chr. 22sh	D04	PSC355
21	CS‐T02	G. tomentosum	Chr. 2	A02	FM966
22	CS‐T04‐15	G. tomentosum	Chr. 4 and 15	A04 and D01	FM966
23	CS‐T04	G. tomentosum	Chr. 4	A04	DP90
24	CS‐T06	G. tomentosum	Chr. 6	A06	SG747
25	CS‐T08sh	G. tomentosum	Chr. 8sh	A08	FM966
26	CS‐T10‐19	G. tomentosum	Chr. 10 and 19	A10 and D05	FM966
27	CS‐T10	G. tomentosum	Chr. 10	A10	SG747
28	CS‐T11sh	G. tomentosum	Chr. 11sh	A11	PSC355
29	CS‐T15sh	G. tomentosum	Chr. 15sh	D01	SG747
30	CS‐T16‐15	G. tomentosum	Chr. 16 and 15	D07 and D01	FM966
31	CS‐T17	G. tomentosum	Chr. 17	D03	DP90
32	CS‐T22sh	G. tomentosum	Chr. 22sh	D04	FM966
	Following are upland cotton cultivars
33	DP90
34	FM966
35	PSC355
36	SG747
37	ST474

Abbreviation: CSL, chromosome substitution line.

^a CSL parent designations indicate the source donor species, that is, G. barbadense (B), G. mustelinum (M), or G. tomentosum (T). As an example, CS‐B02 is upland cotton inbred TM‐1 with disomic substitution for chromosome 2 of G. barbadense line 3–79. A single two‐digit number alone (e.g., 02) indicates the substituted whole chromosome, an “sh” suffix indicates an arm‐specific segment, and a hyphenated pair of two‐digit numbers indicates a double segmental substitutions.

^b Chromosome names based on old nomenclature.

^c Chromosome names based on new nomenclature.

^d The upland cotton cultivar used to make the cross. For example, CS‐B02 was crossed to SG747 to obtain F1 progeny.

2.2. MAGIC population development

The development of RMBHMTUP‐C4, the precursor of the MAGIC population, was described by Jenkins et al. (2019). Each CSL was crossed with one of the five upland cotton cultivars (Table 1) resulting in a total of 32 F₁ families. These F₁ families were planted in two‐row plots with 100 seeds per row and random‐mated for five cycles using a bulked pollen method described in a previously published protocol (Jenkins et al., 2008). Briefly, we collected four flowers from one row of each of the 32 F₁ family plots, mixing pollens and applying to 10 emasculated flowers in each of the 32 plots on alternate rows. The following day, male and female rows were switched and the process was repeated. This procedure was applied for about 21 days of pollination. We harvested approximately 70 crossed bolls per generation from each of 32 families. The crossed bolls were harvested based on their original F₁ family. Two hundred randomly selected seeds from each family were used to plant 32 pairs of rows the second year. Random mating of the F₁ plants was called cycle zero (C₀). We continued the same random‐mating procedure for four additional cycles. We did not make any selections during the random mating cycles. After five cycles of random mating, self‐pollination was followed for five generations using a single seed descent method. Twenty‐five lines were selected from each of the 32 families representing 32 CSLs, and a new C₅S₅ population comprising 797 RILs (three lines were lost during the development) was created. This new population is designated BHMT MAGIC population.

2.3. Genomic DNA isolation and whole genome sequencing

Seeds from the 37 parents and 797 RILs were sown in small pots in a greenhouse in New Orleans, LA, in 2023. For each line, young leaves were collected from 10 seedling plants and bulked. Leaves were stored at −80°C. Total DNA was extracted from the frozen leaves following a protocol described before (D. Fang et al., 2010) with an additional RNAase A digestion step before binding DNA to a purification column. DNA quantity and quality was measured using a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific) as well as on a 1.5% agarose gel. DNAs were sent to Dante Labs or Novogene Corporation for library preparation, and whole genome sequencing using Illumina HiSeq 2500 with paired‐end 150 bp reads. Parental lines were sequenced at 25–40x genome coverage, corresponding to approximately 62.5–100 Gb of data per line, while each RIL was sequenced at 10x genome coverage, generating roughly 25 Gb per sample.

2.4. Processing of sequence data and variant calling

For the 37 parental lines, we aligned the sequencing reads to the newest JGI version 3 G. hirsutum cv. TM‐1 genome (Sreedasyam et al., 2024) and to the GB, GM, and GT genomes (Chen, Sreedasyam, et al., 2020) with Burrows–Wheeler‐Aligner software (the BWA‐MEM command) using the default parameters (H. Li & Durbin, 2009). For the 797 MAGIC RILs, we only aligned the sequencing reads to the JGI version 3 G. hirsutum cv. TM‐1 genome. Sequencing reads resulting from PCR duplicates for each sample were filtered with the Picard package, and uniquely mapping reads were retained in BAM format. We used SAMtools to convert the mapping result to the BAM format (H. Li et al., 2009). DNA sequence variants (single nucleotide polymorphisms [SNPs] and insertion/deletions [InDel]) calling was performed using SAMtools mpileup | BCFtools call (H. Li, 2011) to generate variant call format files. We first filtered sequence variants based on their quality scores. Only variants with quality scores higher than 300 were kept for further analysis. Second, we removed heterozygous SNPs. In the MAGIC population, we filtered SNPs further if a SNP had more than 20% (>159 RILs) missing rate or minor allele frequency (MAF) was lower than 5% (∼40 RILs).

Diagnostic SNPs are SNPs that are species‐specific and can be used to identify introgression segments in CSLs and MAGIC RILs. To identify a diagnostic SNP, in the parental VCF files obtained by aligning to the G. hirsutum TM‐1 genome, at a particular SNP locus, all five upland cotton parents must be homozygous and same as TM‐1, that is “0/0” (two copies of the reference base), and a CSL parent must be homozygous on the alternate allele, that is, “1/1” (two copies of the alternate base). An SNP was discarded if one or more upland parental lines had missing data or were heterozygous. The informative diagnostic SNPs were retained and used for analysis of the sequence data from the 797 MAGIC RILs.

2.5. Mosaic map tracing alien introgression in the MAGIC RILs via identity by descent

A hidden Markov model (HMM) was employed to reconstruct the identity by descent (IBD) ancestry mosaic of each MAGIC RIL across all 26 chromosomes. The model consisted of four hidden states corresponding to the three donor species (GB, GM, and GT) and the GH host. At each diagnostic SNP locus, genotype observations were classified as homozygous reference, homozygous alternate, heterozygous, or missing. Transition probabilities between adjacent loci were modeled as a function of physical distance, reflecting expected recombination frequencies in MAGIC RILs following five generations of random mating and subsequent inbreeding. The most probable sequence of ancestral states along each chromosome was determined using Viterbi algorithm (Rabiner, 1990). An IBD segment was reported when it was supported by at least five consecutive diagnostic SNPs. This approach follows the HMM‐based IBD tracing framework described by Mott et al. (2000) and applied to MAGIC populations by H. Liu et al. (2020).

Before inference, positions present in more than one donor genome are removed (not genome‐specific). Positions from multiple CSLs of the same donor are deduplicated, leaving only unambiguously donor‐specific sites.

2.6. Identity by state kinship tree and population structure

To evaluate population structure and genetic relatedness among the 797 MAGIC RILS, a pairwise kinship matrix was constructed based on the identity by state (IBS). IBS measures the similarity of two individuals at each SNP locus regardless of its origin. All biallelic SNPs passing quality filters (MAF ≥ 0.05 and missing rate ≤ 0.20) were used. The IBS coefficient between each pair of RILs was estimated as mean proportion of alleles shared across all non‐missing loci, calculated as IBS(i, j) = 1 − |gi − gj|/2, where g denotes the alternated allele dosage (0, 1, or 2) at each SNP (VanRaden, 2008). The unweighted pair‐group method with arithmetic mean (UPGMA) was applied to the IBS distance matrix (1 − IBS) to construct the Kinship tree (Sneath & Sokal, 1973). Principal Coordinate analysis (PCoA) was performed via eigen decomposition of kinship matrix to visualize population structure.

3. RESULTS

3.1. Genomic characterization of the 32 chromosome substitution parental lines

3.1.1. Alien genome introgressions in the 32 CSLs when aligned to the G. hirsutum TM‐1 genome

The 32 CSLs along with five upland cotton cultivars were sequenced at 25–40x genome coverage, and each had about 62.5–100 Gb sequence data. When aligned to the G. hirsutum TM‐1 genome (JGI v3), we identified an average of 2,756,359 SNPs and 305,679 InDels in a CSL line (Table 2). The proportion of InDels relative to SNPs was consistently around 11% across all CSLs. Generally, CSLs with double chromosome segment substitutions, such as CS‐B04‐15, had higher number of SNPs and InDels compared to those containing only one chromosome or segment substitution. For comparison, the number of homozygous SNPs found in the five upland cotton parental lines ranged from 240,059 in FM966 to 682,431 in DP90, with an average of 542,895.

TABLE 2.

Number of single nucleotide polymorphisms (SNPs), insertions/deletions (InDels), and diagnostic SNPs in the chromosome substitution lines (CSLs) when comparing to the Gossypium hirsutum TM‐1 genome.

#	CSLs	# SNPs	# InDels	# Diagnostic SNPs	% diagnostic SNPs	% InDels of SNPs
1	CS‐B02	2,276,682	289,776	244,931	10.76	12.70
2	CS‐B04‐15	3,797,624	363,979	725,289	19.10	9.60
3	CS‐B04	2,571,720	303,946	447,783	17.41	11.80
4	CS‐B06	2,102,886	285,266	621	0.03	13.60
5	CS‐B08sh	2,890,371	311,703	213,696	7.39	10.80
6	CS‐B10‐19	3,432,482	338,703	684,669	19.95	9.90
7	CS‐B10	3,015,458	321,143	583,778	19.36	10.60
8	CS‐B11sh	2,304,586	284,906	133,148	5.78	12.40
9	CS‐B15sh	2,142,837	279,725	2844	0.13	13.10
10	CS‐B16‐15	2,798,962	310,951	312,115	11.15	11.10
11	CS‐B17	2,307,468	286,985	115,562	5.01	12.40
12	CS‐B22sh	2,428,134	293,567	126,348	5.20	12.10
13	CS‐M02	2,627,547	294,572	227,355	8.65	11.20
14	CS‐M04‐15	2,705,747	295,526	2722	0.10	10.90
15	CS‐M04	2,697,811	296,673	257,114	9.53	11.00
16	CS‐M06	3,138,002	313,115	234,807	7.48	10.00
17	CS‐M08sh	3,285,672	317,783	540,820	16.46	9.70
18	CS‐M15sh	2,718,395	303,718	294,883	10.85	11.20
19	CS‐M16‐15	2,763,774	311,922	343,344	12.42	11.30
20	CS‐M22sh	2,394,225	288,645	85,749	3.58	12.10
21	CS‐T02	2,580,470	292,742	242,800	9.41	11.30
22	CS‐T04‐15	2,973,871	316,993	344,552	11.59	10.70
23	CS‐T04	2,442,287	289,573	11,911	0.49	11.90
24	CS‐T06	2,848,630	298,896	112,360	3.94	10.50
25	CS‐T08sh	2,919,488	310,609	394,407	13.51	10.60
26	CS‐T10‐19	3,346,596	334,055	676,641	20.22	10.00
27	CS‐T10	3,324,122	334,247	487,156	14.66	10.10
28	CS‐T11sh	2,592,296	291,249	206,468	7.96	11.20
29	CS‐T15sh	2,501,923	298,207	99,468	3.98	11.90
30	CS‐T16‐15	2,834,536	317,100	271,201	9.57	11.20
31	CS‐T17	3,012,141	317,687	192,094	6.38	10.50
32	CS‐T22sh	2,426,752	287,775	22,325	0.92	11.90
	Average	2,756,359	305,679			11.23

We conducted a detailed analysis of introgression for each CSL using the diagnostic SNPs (Table 2; Table S1). In a CSL, a high percentage of diagnostic SNPs relative to the total number of SNPs shown in Table 2 suggested a substantial rate of alien genome introgression and likely an indication of a successful targeted chromosome substitution. Among the 12 CS‐B lines, two lines, that is, CS‐B06 and CS‐B15sh, contained only 621 (0.03% of total SNPs) and 2844 (0.13% of total SNPs) diagnostic SNPs, respectively. Specifically, CS‐B06 had just 35 diagnostic SNPs on the target Chr. A06 while CS‐B15sh had only 18 on Chr. D01 (Table S1), clearly demonstrating that these two lines lacked the intended chromosome substitutions. The remaining 10 CS‐B lines had successful targeted chromosome substitutions (Table 2; Table S1; Figure S1.). In the case of the eight CS‐M lines, seven displayed successful substitution of the target chromosomes, whereas CS‐M04‐15 exhibited only 30 and 22 diagnostic SNPs on the target Chr. A04 and D01, respectively, indicating unsuccessful targeted chromosome substitutions (Table S1; Figure 1). Of the 12 CS‐T lines, CS‐T04 was the sole line that did not obtain successful replacement of the targeted chromosome A04 (Table S1; Figure S2).

FIGURE 1.

Gossypium mustelinum specific single nucleotide polymorphisms (SNPs) of the eight CS‐M lines when aligned to Gossypium hirsutum reference genome. X‐axis: chromosomes, Y‐axis: number of SNPs per Mb. CS‐M04‐15 does not contain the targeted substituted chromosomes. CS‐M, chromosome substitution from Gossypium mustelinum.

The overall percentage of diagnostic SNPs in CS‐B22sh, CS‐M22sh, and CS‐T22sh was notably low, with values of 5.20%, 3.58%, and 0.92%, respectively (Table 2). This is likely attributable to the fact that Chr. 22 (Chr. D04) is the third shortest chromosome comprising a total length of 56,598,092 bp, or 2.6% of the entire genome. Nevertheless, each of these three CSLs contained the substituted target chromosome or segment that reached more than 76% of the Chr. D04 length (Figure 1; Figures S1 and S2; Table S2).

The length of the substituted target chromosome in each CSL is shown in Table S2. The substituted target chromosome length in 19 CSLs reached more than 90% of the total chromosome length, an indication of very successful target chromosome substitution. In CS‐B04‐15, substitutions in both Chr. A04 and D01 reached more than 90% of their respective chromosome length.

3.1.2. Alien genome introgressions in the 32 CSLs when aligned to their respective donor species genome

We aligned the genome sequences of 37 parental lines against the GB, GM, and GT reference genomes (Chen, Sreedasyam, et al., 2020) to further validate alien chromosome substitutions in the CSLs. A successful targeted chromosome substitution should result in a minimal number of SNPs on the target chromosome when a CSL genome was compared to its own donor species reference genome. For example, in CS‐B02, the number of SNPs on Chr. A02 was notably low upon alignment with the GB reference genome (Table S3). Tables S3–S5 show the SNP numbers aligned to GB, GM, and GT reference genome, respectively. In Table S3, it clearly shows that CS‐B06 and CS‐B15sh lacked their intended chromosome substitutions. Conversely, the remaining 10 CS‐B lines had successful chromosome substitutions as shown by the lowest number of SNPs in their respective target chromosomes. For example, CS‐B02 had only 1913 SNPs on Chr. A02, but more than 300,000 SNPs in each of the remaining 25 chromosomes (Figure 2; Table S3). In CS‐B10‐19, the lowest SNP numbers were 15,871 on Chr. A10 and 121,593 on Chr. D05 (Table S3). Alignment against the GM reference genome confirmed minimal or absent GM introgression in the target chromosomes A04 and D01 for CS‐M04‐15 (Table S4). Similarly, alignment to the GT reference genome revealed no introgression on Chr. A04 in CS‐T04 (Table S5).

FIGURE 2.

Homozygous single nucleotide polymorphism (SNP) density per Mb in each chromosome of CS‐B02 when aligned to the Gossypium barbadense reference genome. X‐axis: chromosomes, Y‐axis: number of SNPs per Mb.

In summary, SNP analysis using GH, GB, GM, or GT reference genomes consistently showed that CS‐B06, CS‐B15sh, CS‐M04‐15, and CS‐T04 did not acquire the intended chromosome substitutions from their respective donor species. The remaining 28 CSLs were developed successfully as anticipated.

3.1.3. Residual alien genome introgressions in the 28 true CSLs

In all 28 true CSLs, we found residual alien genome introgressions scattered throughout the entire genome, not just on their target chromosomes (Table S1). Table 3 highlights introgression blocks that cover at least half of their residing chromosome length. These blocks were formed by merging diagnostic SNP positions within 5 Mb of each other, and SNP density must be higher than 10 per Mb. We found 24 large, nontarget introgression blocks present in nine CSLs including all seven double‐segmental CSLs and two single chromosomal CSLs (CS‐M15sh and CS‐T02). Typically, double‐segmental CSLs had a greater number of large introgression blocks, for example, CS‐T10‐19 contained seven blocks, the highest count. Noticeably, diagnostic SNP density per Mb in these nontarget introgression blocks was much lower than on the actual substituted target chromosome. For example, CS‐T10‐19's A09 block had the highest diagnostic SNP density (564.3 per Mb) among nontarget blocks, while the diagnostic SNP densities on its target chromosomes, A10 and D05, were 4507 and 1490 per Mb, respectively. CS‐B16‐15, CS‐M16‐15, and CS‐T16‐15 had only one large block on Chr. A07, which is homologous to Chr. D07 (Chr. 16), a target chromosome in these three CSLs. Most large introgressed genome blocks in other CSLs appeared on chromosomes that are not homologous to the target chromosomes.

TABLE 3.

Presence of large nontarget alien introgression block (>50% of a chromosome) in the 28 true chromosome substitution lines (CSLs).

#	Line	Target chromosome	Residual chromosome	Block length (bp)	Chromosome length (bp)	Introgression (%)	# SNPs in block	SNP density per Mb
1	CS‐B04‐15	A04/D01	A02	61,546,679	90,410,113	68.07	3987	64.8
2	CS‐B04‐15	A04/D01	A11	66,417,198	114,316,535	58.1	4102	61.8
3	CS‐B04‐15	A04/D01	A13	71,143,041	104,534,465	68.06	22,517	316.5
4	CS‐B04‐15	A04/D01	D09	45,800,076	53,439,012	85.71	4628	101
5	CS‐B04‐15	A04/D01	D10	44,569,983	67,973,905	65.57	1940	43.5
6	CS‐B04‐15	A04/D01	D12	49,765,103	63,843,421	77.95	3232	64.9
7	CS‐B10‐19	A10/D05	A01	66,304,827	104,050,246	63.72	1611	24.3
8	CS‐B10‐19	A10/D05	A09	79,562,465	81,831,061	97.23	32,321	406.2
9	CS‐B10‐19	A10/D05	A12	64,708,379	103,038,859	62.8	1084	16.8
10	CS‐B10‐19	A10/D05	D08	38,702,386	69,935,236	55.34	1372	35.5
11	CS‐B10‐19	A10/D05	D10	66,114,245	67,973,905	97.26	6060	91.7
12	CS‐B16‐15	D07/D01	A07	71,485,165	93,829,987	76.19	2681	37.5
13	CS‐M15sh	D01	A09	60,716,450	81,831,061	74.2	23,011	379
14	CS‐M16‐15	D07/D01	A07	81,831,943	93,829,987	87.21	2723	33.3
15	CS‐T02	A02	A12	72,625,247	103,038,859	70.48	40,039	551.3
16	CS‐T04‐15	A04/D01	A05	60,193,566	100,714,657	59.77	14,195	235.8
17	CS‐T10‐19	A10/D05	A01	79,484,764	104,050,246	76.39	1838	23.1
18	CS‐T10‐19	A10/D05	A05	78,594,259	100,714,657	78.04	6467	82.3
19	CS‐T10‐19	A10/D05	A09	61,692,760	81,831,061	75.39	34,812	564.3
20	CS‐T10‐19	A10/D05	A12	83,321,905	103,038,859	80.86	3101	37.2
21	CS‐T10‐19	A10/D05	A13	53,988,450	104,534,465	51.65	22,702	420.5
22	CS‐T10‐19	A10/D05	D08	38,807,629	69,935,236	55.49	1273	32.8
23	CS‐T10‐19	A10/D05	D10	64,665,688	67,973,905	95.13	4991	77.2
24	CS‐T16‐15	D07/D01	A07	75,148,134	93,829,987	80.09	2696	35.9

3.2. DNA sequence variations on chromosome A08 between the five upland cotton parental lines and TM‐1 reference genome

In this research, we used the latest version (JGI v3) of G. hirsutum TM‐1 genome as our reference. We detected a significant number of SNPs on Chr. A08 across all five upland cotton parental lines. The average SNP density per Mb in these five cultivars was 1588 on Chr. A08, as compared to 174 in the other 25 chromosomes (Table S6), a striking ninefold difference. Figure 3 illustrates the chromosomal distribution of SNPs in PSC355 as a representative example. The findings show a marked genetic difference between the obsolete TM‐1 and more recently developed cultivars with respect to Chr. A08. Section 4 explores potential explanations in depth.

FIGURE 3.

Homozygous single nucleotide polymorphism (SNP) density per Mb in each chromosome of PSC355 when aligned to the JGI version 3 Gossypium hirsutum TM‐1 reference genome. X‐axis: chromosomes, Y‐axis: number of SNPs per Mb.

The five upland cotton cultivars had an average of 542,896 homozygous SNPs (i.e., 1/1) when aligned against TM‐1 genome (Table S6). When using GB, GM, or GT reference genome, this number was 13,362,138; 15,459,651; and 11,833,275, respectively. These results further indicate that GT is phylogenetically more closely related to GH than GB or GM does (Tables S3–S5). This finding is in agreement with a previous study (Gallagher et al., 2017).

3.3. Genomic characterization of 797 MAGIC RILs

3.3.1. Alien genome introgressions and IBD‐based mosaic map of 797 MAGIC RILs

Each MAGIC RIL was sequenced at 10x genome coverage (∼25 Gb), yielding approximately 20 Tb sequence data. The resulting sequences were aligned to the JGI v3 G. hirsutum TM‐1 reference genome (Sreedasyam et al., 2024). Following the application of filtering criteria described in Section 2, 7,333,111 SNPs were identified within the MAGIC population consisting of 797 RILs (Table 4). The A sub‐genome chromosomes (A01–A13) accounted for 4,787,566 SNPs and D sub‐genome chromosomes (D01–D13) contained 2,551,545 SNPs, aligning with the respective sizes of the two sub‐genomes. Chr. D09 had the lowest number of SNPs (132,250), whereas Chr. A08 had the highest count (685,580).

TABLE 4.

Number of single nucleotide polymorphisms (SNPs) in the 797 multiparent advanced generation intercross (MAGIC) recombinant inbred lines (RILs).

Chromosome	SNPs	Chromosome	SNPs
A01	387,778	D01	258,209
A02	299,603	D02	239,739
A03	353,776	D03	139,060
A04	407,830	D04	148,493
A05	294,859	D05	325,062
A06	377,495	D06	185,628
A07	286,896	D07	198,693
A08	685,580	D08	250,782
A09	275,841	D09	132,250
A10	344,096	D10	165,345
A11	330,627	D11	160,645
A12	336,287	D12	178,857
A13	400,898	D13	168,782
Subtotal	4,781,566	Subtotal	2,551,545
Grand total			7,333,111

The range of donor species‐specific SNPs observed in each RIL varied from 8 in RIL4389 to 1,016,117 in RIL4610, with an average of 77,872 and a median value of 31,465 (Table S7). Notably, three RILs (RIL4471, 4483, and 4664) did not contain donor species‐specific SNPs and were therefore excluded from subsequent IBD analysis.

To identify IBD segments, we employed a HMM to infer the parental origin of each alien introgression segment in 794 RILs (M. Wang et al., 2022). In total, 25,809 IBD segments were detected across 794 RILs (Table S8). Among these, 608 RILs exhibited more than 1% genome introgression, with an average of 2.93%. RIL4677 showed the highest introgression rate at 24.05%. The length of the IBD segments ranged from 478 bp (RIL4139 on Chr. A10) to 45.4 Mb (RIL4260 on Chr. A04), with a mean length of 1.96 Mb. The number of IBD segments per line varied from 1 in four RILs to 278 in RIL4544, averaging 32.5 segments per RIL.

Twenty‐five RILs contained one or more introgressed segments exceeding 50% of the corresponding chromosome. They are highlighted in yellow in Table S8. This phenomenon and the three RILs without receiving any introgression suggest that random mating was insufficient to completely disrupt chromosomal linkage in certain RILs, which may be attributable to genome incompatibility, recombination suppression, or other unidentified factors. More is discussed in Section 4.

On average, nine chromosomes in each inbred line had introgressions. Seven RILs had introgressions on only one chromosome. RIL4610 exhibited introgressions on 25 out of 26 chromosomes from all three donor species, covering 14.47% of the genome (Figure 4). The biological consequences of such extensive introgression in one single line remain to be investigated. Figure 5 displays the IBD‐based introgression haplotypes of 794 RILs. As shown in Figure 5 and Table S8, introgressions were detected in all 26 chromosomes across various MAGIC RILs, even though the CSLs had target substitutions in just 11 chromosomes. When combining introgression percentage across all 794 RILs, Chr. A09 had the highest introgression, equivalent to 49.46 times its chromosome length. Chr. D02 had the lowest, only 10.9 chromosome length equivalence (Table 5). Interestingly, neither chromosome had an associated CSL. Calculating averages per RIL, introgression coverage ranged from 1.37% on Chr. D02 to 6.23% on Chr. A09.

FIGURE 4.

Identity‐by‐descent‐based ancestry mosaic of RIL4610.

FIGURE 5.

Identity‐by‐descent blocks from three Gossypium species in each of 794 BHMT multiparent advanced generation intercross (MAGIC) recombinant inbred lines (RILs).

TABLE 5.

Genome coverage of alien introgressions in multiparent advanced generation intercross (MAGIC) recombinant inbred lines (RILs).

Chr.	Chromosome equivalence of the combined introgression	Average coverage per RIL (%)
A01	21.13	2.66
A02	23.16	2.92
A03	17.54	2.21
A04	43.51	5.48
A05	20.25	2.55
A06	15.70	1.98
A07	21.64	2.73
A08	27.26	3.43
A09	49.46	6.23
A10	15.74	1.98
A11	33.95	4.28
A12	18.23	2.30
A13	41.14	5.18
D01	41.15	5.18
D02	10.91	1.37
D03	23.97	3.02
D04	24.92	3.14
D05	11.30	1.42
D06	22.41	2.82
D07	17.05	2.15
D08	17.14	2.16
D09	33.06	4.16
D10	15.69	1.98
D11	13.41	1.69
D12	37.41	4.71
D13	19.93	2.51
Average	24.50

Among the 794 RILs containing alien introgressions, 783, 698, and 720 had GB, GM, and GT introgressions, respectively. GB, GM, and GT introgressions were dominant in 356, 172, and 266 RILs, respectively (Table S9). This finding suggests that GB introgressions had the best, and GM introgressions the least competitive advantage during the population development when upland cotton acted as a host. Figure 6 shows the distribution of RILs with different levels of alien introgressions.

FIGURE 6.

Identity‐by‐descent‐based ancestry proportions in each BHMT multiparent advanced generation intercross (MAGIC) recombinant inbred lines (RILs). Each BHMT MAGIC RIL has over 95% of its genome from Gossypium hirsutum, indicated in gray at the base.

3.3.2. IBS‐based kinship and population structure of the 797 MAGIC RILs

All 7,333,111 SNPs from 797 RILs were used to calculated identity by IBS‐based kinships among the RILs. The resulting kinship matrix is shown in Table S10 while the UPGMA heatmap can be found in Figure S3. Each pair of RILs shared at least 73.26% of their alleles, with an average allele‐sharing rate of 80.9%. This outcome was expected given that all RILs were derived from the same five G. hirsutum parents and 32 CSLs. In each CSL, only one or two chromosomes were not from G. hirsutum. The observed 19.1% variation within the population reflects the genetic diversity contributed by recombination, inherent genetic differences among upland cotton parents, and various donor introgression events.

The UPGMA kinship heat map delineated four distinct groups: Group 1 comprised nine RILs, Group 2 included six RILs, Group 3 consisted of 434 RILs, and Group 4 contained 348 RILs. Detailed group membership is provided in Table S9. The 15 RILs in Groups 1 and 2 were identified as outliers. They represented RILs that received atypical combinations of donor introgression. The average introgression percentage in Groups 1 and 2 was 7.81 and 8.98, respectively. In contrast, RILs in Groups 3 and 4 had more typical patterns with average introgression percentage of 2.63 and 3.33, respectively, as detailed in Table S9. Examining the agronomic performance of the RILs across distinct groups will be a valuable direction for subsequent research phases.

PCoA of the same information grouped the 797 RILs into four clusters, with Clusters 3 and 4 being the largest (Figure 7). The cumulative proportion of variance explained (PVE) by the principal coordinates was 18.6%. The top three coordinates, PC1, PC2, and PC3 had PVE 7.3%, 2.6%, and 1.6%, respectively.

FIGURE 7.

Identity‐by‐state‐based principal coordinate analysis (PCoA) plots of BHMT multiparent advanced generation intercross (MAGIC) recombinant inbred lines (RILs).

The kinship heat map generally agreed with the PCoA plot, but Clusters 1 and 2 in PCoA included more members than Groups 1 and 2 in the kinship heat map (Table S9).

4. DISCUSSION

The development of CSLs occurred during the 1990s and early 2000s. At that time, DNA markers with known chromosomal locations were limited in availability and challenging to utilize, as most were restriction fragment length polymorphism markers. Developers primarily relied on a small set of plant morphological characteristics and metaphase I analysis of microsporocytes to determine whether a CSL contained the target chromosome substitution (Stelly et al., 2005). Despite significant efforts, it was unavoidable that some CSLs did not achieve successful target chromosome substitution. In the present research, four out of 32 CSLs were identified as lacking the targeted chromosome substitution. In our previous research (D. Fang et al., 2023), we characterized 18 CS‐B lines using genome sequences. Five of them were found not to contain targeted chromosome substitutions. Of the 12 CS‐B lines used in this research, eight lines overlapped with those used previously, and the findings from both studies were consistent. The Stelly laboratory in Texas A&M University developed more than 60 CSLs including CS‐B, CS‐M, and CS‐T (Hitzelberger, 2022). Many of them have not been fully characterized using molecular or genomic tools. It is advisable to verify the chromosome substitution status of a CSL before using it in breeding or genetic investigations.

Cotton interspecific CSLs were developed by crossing a given donor species accession to various quasi‐isogenic hypoaneuploids of upland cotton line (most likely TM‐1) followed by five generations of backcrossing to the upland cotton parent. Finally, at least one round of self‐pollination was conducted to recover euploid CSLs that were homozygous and thus true breeding for the chromosome substitution (Stelly et al., 2005). Theoretically, each nontarget chromosome should contain ∼0.8% of the donor chromosome. If excluding the target chromosome, we observed that the percentage of residual SNPs to the total SNPs in each CSL was 1.02%, which is close to 0.8% (Table S1).

Reviewing Table 3 revealed several noteworthy findings. First, large alien introgression blocks spanning over 50% of the residing chromosome were detected in many CSLs, indicating that five cycles of backcrossing did not fully break the linkage, and confirming the challenges associated with interspecific hybridization in cotton breeding. Second, double‐segmental CSLs contained high number of large nontarget introgressed blocks than single chromosomal CSLs. The double segmental CSLs were tertiary monosomies involving unbalanced translocations of two nonhomologous chromosomes. Development of this type of CSLs was more complicated than developing primary monosomic CSLs. Third, all three CSLs involving Chr. 16 and 15 (i.e., CS‐B16‐15, CS‐M16‐15, and CS‐T16‐15) showed only one major introgression block on Chr. A07 that is homologous to Chr. D07 (Chr. 16). Further investigation is warranted to determine whether the original tertiary monosomic line exhibited abnormalities on Chr. A07.

Yang et al. (2019) conducted a comparative analysis of sequence assemblies between TM‐1 and ZM24, a modern Chinese cultivar, and revealed three very long structural inversions on Chr. A08. He et al. (2021) further analyzed 3248 cotton genomes and observed extensive chromosomal inversions on Chr. A08 among the upland cotton cultivars. Sreedasyam et al. (2024) updated the TM‐1 reference genome assembly and released it as JGI version 3. They also reported considerable sequence and structural variations on Chr. A08 between TM‐1 and three recently developed upland cotton cultivars (UGA230, UA48, and CSX8308) from the United States and Australia. In this research, we used JGI v3 TM‐1 genome as a reference, and revealed very large number of SNPs on Chr. A08 in all five upland parental lines (Figure 3). In our previous research (D. Fang et al., 2023), we used HAU TM‐1 genome (M. Wang et al., 2019) as a reference. We did not observe exceptionally large number of SNPs on Chr. A08 in FM966, SG747, or ST474. It is important to note that numerous cotton lines are referred to as “TM‐1”; the sample sequenced by Sreedasyam et al. (2024) might not be the same as the one sequenced by M. Wang et al. (2019). In He et al. (2021) research, they reported five samples under the “TM‐1” designation, and found two types of genomic haplotypes on Chr. A08. TM‐1, an inbred line derived from the obsolete Deltapine14 that was bred by Delta and Pine Land Company in early 1940s (May et al., 1995), differs considerably from modern cultivars developed in the United States, China, and Australia with respect to Chr. A08. The structural inversion on Chr. A08 likely represents a selection signature associated with breeding contemporary cotton cultivars. We filtered out all homozygous SNPs (i.e., 1/1) present in the five upland cotton parents when we determined a diagnostic SNP in a CSL. Thus, the extremely high number of SNPs on Chr. A08 present in the five upland parents do not affect our analysis and results reported here. However, we do need to be aware of A08 structural variations when we do trait GWAS analysis in the future.

Random‐mating procedures have been used to break undesirable linkage associations and to form new combinations in many plant crops (Scott et al., 2020). Previously, we used the random‐mating method to create an upland cotton MAGIC population that showed no or little population structure (D. Fang, Jenkins, et al., 2014; Islam et al., 2016). In this research, we observed that the 797 MAGIC RILs contained a wide range of alien introgressions (Figure 5), and one RIL had introgressions in 25 chromosomes and three RILs did not have any introgression. Furthermore, 25 RILs still contained a large segment that could reach 50% of its residing chromosome. IBS‐based kinship analysis showed a clear structure present among the 797 MAGIC RILs (Figure 7; Figure S3). It is clear that five cycles of random mating procedure were not able to break the alien chromosomal linkage when an interspecific cross was involved. We observed a similar phenomenon in the CS‐B MAGIC population (D. Fang et al., 2023). This may be due to localized genome incompatibility, recombination suppression, segregation distortion, or other unknown reasons. In the next phase of this research, we will use this MAGIC population to investigate how this uneven alien introgression affects agronomic performance in cotton.

A total of true 28 true CSLs (10 CS‐B, 7 CS‐M, and 11 CS‐T) served as parental lines for the MAGIC population. Among the 794 RILs with introgressions, 356 were GB dominant, 172 were GM dominant, and 266 were GT dominant. This showed that competition bias tended to favor GB over GM or GT within the population. GM was phylogenetically furthest from GH, followed by GB and then GT based on the SNP numbers and a previous study (Gallagher et al., 2017). Therefore, while genetic relatedness explains some of the competition bias toward GB, it does not account for all of it. The GB donor for CS‐B lines was “3‐79” that was a cultivated cotton variety, whereas the GM and GT donors were truly wild species. As a result, the GB 3–79 may be better suited to natural selection than GM or GT in an environment where they are not native. It would be of considerable interest to investigate the genetic and molecular basis underlying the preferential transmission of GB 3–79 chromosomal segments relative to the segments from GM and GT in upland cotton.

The use of tetraploid species to improve upland cotton cultivars has presented considerable challenges. Cotton researchers have encountered several adverse effects associated with whole genome interspecific hybridization, including hybrid breakdown, pronounced segregation distortion, among others. In this study, we implemented a new strategy to introgress alleles from wild species into upland cotton by using CSLs as parents. Additionally, five cycles of random mating were employed to disrupt chromosomal linkage. We did achieve a goal of introgression as shown by 794 RILs exhibiting varying degrees of alien genetic materials. To our knowledge, this is the first study in which introgressions from four tetraploid Gossypium species were targeted via random mating in upland cotton. We observed competitive interactions among species during introgression, with a discernible bias favoring GB within the population. Contrary to our initial expectations that fewer large alien chromosome blocks would be present in the MAGIC population, substantial chromosome segments spanning multiple chromosomes were detected in numerous RILs. In our future research, we will conduct in‐depth trait analysis of the MAGIC RILs to understand how the large alien chromosome segments affect agronomic traits such as fiber quality and yield.

5. CONCLUSION

Whole genome sequence analysis revealed that 28 out of 32 CSLs contained targeted chromosome substitutions, while four did not. Residual alien genome introgressions were widespread across the genome in all CSLs. Large introgressed segments were present in most CSLs. Among the 797 MAGIC RILs, most lines showed varying degrees of introgression, with GB introgressions predominating. IBS kinship analysis showed a clear structure within the MAGIC population. Despite multiple rounds of random mating, large alien chromosome blocks persisted in many RILs. The research advances our understanding of wild tetraploid Gossypium introgressions in upland cotton and provides valuable resources for future breeding, with ongoing field experiments to identify beneficial traits through GWAS.

AUTHOR CONTRIBUTIONS

Ganesh Pinnika: Formal analysis; methodology; software. Ping Li: Methodology. Jack C. McCarty: Conceptualization; investigation; writing—review and editing. Johnie N. Jenkins: Conceptualization; funding acquisition; investigation; resources; supervision. David D. Fang: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; resources; supervision; validation; visualization; writing—original draft; writing—review and editing.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

Supporting information

Figure S1. Gossypium barbadense specific SNPs of the 12 CS‐B lines when aligned to the G. hirsutum reference genome. X‐axis: chromosomes, Y‐axis: number of SNPs per Mb. CS‐B06 and CS‐B15sh do not contain the targeted substituted chromosomes.

Figure S2. Gossypium tomentosum specific SNPs of the 12 CS‐T lines when aligned to the G. hirsutum reference genome. X‐axis: chromosomes, Y‐axis: number of SNPs per Mb. CS‐T04 does not contain the targeted substituted chromosome.

Figure S3. A heat map showing the relatedness between RILs. Relationship matrix was estimated for the relationships among the lines using all SNP data, which the output serves as the matrix for representing familial relatedness. The heat map displays the relationships among the 797 RILs. The red diagonal represents perfect relationship of each line with itself, and the symmetric off‐diagonal elements represent relationship measures [in this case identity by state (IBS)] for pairs of lines. There is an obviously population structure showing 4 groups.

Table S1. Number of diagnostic SNPs on each chromosome in the CSLs when aligned to the G. hirsutum TM‐1 genome.

Table S2. Lenth of target introgressed chromosome or chromosome segments in CSLs.

Table S3. Number of SNPs on each chromosome in the CSLs when aligned to the G. barbadense reference genome.

Table S4. Number of SNPs on each chromosome in the CSLs when aligned to the G. mustelinum reference genome.

Table S5. Number of SNPs on each chromosome in the CSLs when aligned to the G. tomentosum reference genome.

Table S6. SNP numbers and density per Mb on each chromosome in the five upland cotton parents when aligned to the G. hirsutum TM‐1 reference genome.

Table S7. Donor species‐specific SNPs in 797 BHMT MAGIC RILs (G. barbadense = GB, G. mustelinum = GM, G. tomentosum = GT).

Table S8. Identity‐by‐descent (IBD) genomic segments in 794 BHMT MAGIC RILs (G. barbadense = GB, G. mustelinum = GM, G. tomentosum = GT).

Table S9: Per‐RIL complete introgression profile for 797 BHMT MAGIC RILs. Kinship_Cluster from IBS kinship UPGMA (Stage 4). Donor bp and Introgression_pct from IBD segments (Stage 2).

Table S10: Pairwise identity‐by‐state (IBS) kinship matrix for 797 BHMT MAGIC RILs. Values represent proportion of shared alleles (0–1).

Pinnika, G. , Li, P. , McCarty, J. C. , Jenkins, J. N. , & Fang, D. D. (2026). Genomic characterization of a MAGIC population containing introgressions from four tetraploid Gossypium species. The Plant Genome, 19, e70308. 10.1002/tpg2.70308

DATA AVAILABILITY STATEMENT

All relevant data reported in this paper are within the paper and its online Supporting Information. The original DNA sequences of the 37 parental lines were deposited to NCBI database with ID# PRJNA1524532.