1. Interspecific and intraspecific variations in the stem trichome phenotypes of two cultivated tetraploid cottons
There were two kinds of trichomes grown from trichopores, single and tufted/complex dendritic, as seen under a stereo microscope (Fig. 1), and they are similar to those from leaves . The single trichomes were normally long and thin with sharp ends, while the tufted trichomes were normally shorter and grew from the same trichopore. The trichome numbers per trichopore varied from two to more than 10. Different cotton plants, even from the same species, differed in trichome number per trichopore, trichome density, and trichome length, forming clearly different trichome phenotypes. Stem trichome densities were grouped into six classes (0–5) from none to pilose (Fig. S2). As a whole, the cotton stem trichomes were categorized into four types (I–IV) (Table 1, Fig. 2). In Type I, most were singular long trichomes per trichopore, plus a few clusters, and the trichome density differed among different varieties. The stems with low trichome density levels (grades 1 and 2) were further classified as Type Ia and those with high trichome density (grade 4) as Type Ib. In Type II, both singular and clustered trichomes occurred. Trichomes were shorter and coarser than those in Type I, with higher density resulting in a pilose phenotype over the whole plant. In Type III, multiple trichomes formed clusters from one trichopore. They grew sparsely on stems and were shorter than those of Type II. In Type IV, the tufted/complex dendritic trichomes were extremely short and dense, looking as if they were covered by a layer of powder.
The trichome phenotypic data of 11,442 Gh and 4,323 Gb varieties/accessions were downloaded from the USDA National Cotton Germplasm Collection, one of the largest collections of cotton germplasm resources  (Table S6), and used to determine the percentages of cultivars having different stem trichome densities. The two species differed significantly in the proportions of glabrous varieties (Fig. 3a). In total, 55.4% of the Gb varieties (accessions) did not grow or had very few trichomes on their stems, while 97% of the Gh varieties had trichomes in grades 1–5, and only 3% had glabrous stems. In addition, the two cultivated tetraploid cottons were clearly different in all aspects of the trichome phenotype. Type I was the main type on Gh stems, accounting for 95% (Table 1, Fig. 2), while Type III was the main type on Gb stems, accounting for 31.4%. In addition, 5.6% of Gb varieties had Type IV trichomes (grades 4 and 5), while less than 1% of Gh had Type II trichomes (grades 4 and 5). Similar results were obtained from 295 Gh and 246 Gb varieties (Table 2, Fig. 3b, Table S1 and S4). Therefore, the two cultivated species had distinctive stem trichome phenotypes, especially their trichome density levels and trichome types.
2. Wild hirsutum accessions differed significantly in stem trichome morphology from modern cultivated varieties.
We observed the stem trichome phenotypes of wild Gh accessions planted in Hainan, China (Table S2) and found significant differences between wild forms and modern varieties. Compared with modern upland cotton, the wild accessions showed the following trichome characteristics (Table 2, Table S2 and Fig. S4): (1) No or fewer trichomes on stems. Among 180 wild accessions of seven wild races, 30 (16.8%) were glabrous, while only about 3% of the modern varieties in the data from the USDA National Cotton Germplasm Collection (Fig. 3) and none of the varieties we collected were glabrous (Table 2, Table S1). When stem pubescence was compared among the wild accessions based on flowering time, those that flowered later had a greater proportion of glabrous accessions than those that flowered at regular time as modern varieties. Among 81 late-flowering accessions, 37 (45.6%) were glabrous, or had grade 1 trichomes, whereas among 99 regular flowering accessions, only 19 (19.2%) were glabrous, or had grade 1 trichomes (Table 2). When different wild races were compared, race marie-galante had more glabrous accessions than the other wild races. (2) A large number of wild accessions had shorter and denser trichomes than modern varieties. Among the 149 accessions having stem trichomes, 66 (44.3%) were shorter than the Type I modern varieties, including 28 similar to Type II and 6 that were even shorter, having the pilose phenotype (Table S2, Fig. S4), especially among the morrillii race. As a whole, wild G.hirsutum accessions either have more ratio of glabrous stem, or more ratio of shorter trichome compared to modern varieties.
3. Stem trichome phenotypes of diploid ancestors of tetraploid cotton
To compare the stem trichome phenotypes of diploid ancestors with those of cultivated tetraploid cottons, 77 individual plants from 12 accessions of G. arboreum (Table S7) and 5 individual plants from G. raimondii were observed. All the G. raimondii plants had very short and dense clustered trichomes on their stems, similar to Gb Type IV (Fig. 4a). Compared with G. raimondii, G. arboreum had more variation in stem trichome phenotypes (Fig. 4b). As in Gh, G. arboreum accessions also had singular and clustered trichomes. However, the clustered trichomes of G. arboreum were shorter than those of Gh and even shorter than Gb Type IV. As a result, the clusters were difficult to visualize. On the basis of their growth situations, the stem trichome phenotype of G. arboreum was classified into four types: single plus cluster, only single, only cluster, and glabrous. The stem trichome phenotypes of 12 G. arboreum accessions were investigated. Six had singular and clustered trichomes, four only had clustered trichomes, one only had singular trichomes, and one had the glabrous phenotype. Two diploid ancestors differ significantly. G.raimondii has the similar stem trichome as Gb, while G.arboreum has the stem trihomce like Gh.
4. Genetic mapping of stem trichomes
To reveal the genetic basis regulating different types of stem trichomes, five crosses involving varieties having different types of stem trichomes were made (Table S5). Cotton stem trichome initiation is believed to be mainly controlled by genes on Chrs. 06 and 24 [15, 25, 16, 22, 26]. Therefore, our mapping efforts focused on these two genes and 77 polymorphic DNA markers, 23 on Chr. 06 and 49 on Chr. 24, among them 54 were newly developed based on reference genome sequences of G. hirsutum var. TM-1 [27,20] (Table S8). Definitely, it is an efficient way to find the gene(s) related to trichome phenotype, but also will miss some other loci.
Type I trichomes in Gh: QF-10/1 is a Gh variety with Type Ia stem trichomes, grade 2, similar to genetic standard line TM-1. To map the genes conferring this type of stem trichome, QF-10/1 was crossed with Tu 75-37, a glabrous Gb variety. No trichomes grew on the stems of the F1 plants. An F2 population consisting of 118 plants showed continuous variation, and most of them (60%) had grade 0 or I trichome, while some had grade IV trichomes, surpassing the Gh parent (Fig. 5). Using this population, genetic maps were constructed for Chrs. 06 and 24, consisting of 17 and 24 DNA markers, respectively (Fig. 6a, b, Some cosegregation markers were not displayed in the figures. All marker information was listed in the Table S5). Combining the trichome phenotyping and genotyping data from this population, a QTL that had a LOD value of 4.4 and explained 18,4% of the phenotypic variation was detected in the region between NAU5433 and JESPR194 on Chr. 06, and was tightly associated with AF2, a SSCP marker developed by the GhHD1 sequence (Table 3, Fig. 6a, Table S8). Another QTL was detected in the region between D08-163 and L4-239 on Chr. 24 that had a LOD value of 8.6 and explained 30.6% of the phenotypic variation (Table 3, Fig. 6b).
Type II trichomes in Gh: Previously, we mapped the Type II stem trichomes of T586 to Chr. 06 and confirmed that its pilose phenotype was probably controlled by the HD1 gene, using an F2 population derived from the cross between T586, a typical Gh genetic standard line with grade 5 stem trichomes and Pima S6 with glabrous stems . To further clarify the genetic relationship between Type I and II stem trichomes, we crossed the Gh variety Liao1779, having a pilose morphology and grade 5 stem trichomes like T586, with Xinyan96-48, a variety having Type Ia stem trichomes (grades 1 and 2). The F1 plants had stems similar to those of Liao1779, covered with heavy trichomes. In total, 66 F2 plants showed clearly different trichome densities, most (52) having a high density of trichomes like Liao1779 (grades 4 and 5), and 11 had stem trichomes like Xinyan96-48, grades 1 and 2. Only three plants had intermediate trichome density levels (Table S9). These results indicated that the pilose stem (Type II) was probably controlled by a dominant gene. Using AF2, we found that both parents showed very good polymorphism, and the segregation of stem trichome density in this F2 population was closely correlated with the genotype of the SSCP marker (Table S9), suggesting that the pilose stem trichomes of Liao1779 were also controlled by GhAt-HD1, as in T586 .
Type III trichomes in Gb: This type had sparse clustered trichomes on the stems, grading 0 or 1. To map this type of trichome, N73DeltapineNGO, a Gh variety with glabrous stems, was crossed with VIR-59TV, a Gb variety having the typical Type III stem trichomes. The F1 did not have, or had only a few trichomes on stems, like the Gh parent. In the F2 population, most plants (86 of 130) did not have, or had only a few trichomes, like the F1 plants. Approximately 1/3 of F2 plants had stem trichomes of grade 1, and a few were graded at >2 or 3 (Fig. 5). Thus, the glabrous stem of N73DeltapineNGO is probably controlled by a dominant gene. To map this gene, a genetic linkage map of Chr. 24, consisting of 28 markers, was constructed, and a QTL in the region between L4-211 and NAU3904 that had a LOD value of 17.6 and explained 51.5% of phenotypic variance was identified (Table 3, Fig. 6c).
Type IV trichomes in Gb varieties: Bazhou 03 and L-7009 had typical Type IV stem trichomes. To map the genes correlated with this phenotype, Bazhou 03 was crossed with N73DeltapineNGO. The F1 plants showed a trichome phenotype similar to that of N73DeltapineNGO. The F2 plants showed continuous segregation in trichome density, but displayed two peaks at grades 0 and 4 (Fig. 5). Among 383 F2 plants, 243 had few trichomes, graded 0 and 1, similar to N73DeltapineNGO; and 77 had many trichomes, graded 4, similar to Bazhou 03. Only 44 plants had trichomes of grade 2, and very few plants had trichomes of grade 3. When a trichome density of less or equal to grade 2 was used to form a group and a trichome density greater than grade 2 formed a second group, the ratio of the two groups of plants was close to 3:1, again suggesting that the glabrous stem trichomes of N73DeltapineNGO were probably controlled by a dominant gene. To map this gene, 21 polymorphic SSR and SSCP markers were used to genotype the F2 plants, and a linkage map covering 3.74 cM was constructed (Fig. 6d). A significant QTL having a LOD value of 149.1 and explaining 83.7% of the phenotype variation was detected in the region between L4-209 and T10 (Table 3, Table S8). All these mapping results strongly suggested that the gene on Chr. 24 is essential in regulating stem trichome initiation.
To reveal the genetic relationship between the Types III and IV stem trichomes of Gb varieties, Aken 4154 (Type III) was crossed with Bazhou 03 (Type IV). The F1 plants showed a trichome phenotype similar to that of Aken 4154, with a few clustered trichomes, and some had almost no trichomes. Like the (Bazhou 03 x N73DeltapineNGO)F2 population, this F2 population showed clear segregation in stem trichome density (Table S10). Among 64 F2 plants, 44 had stem trichomes of grades 0 and 1, whereas 18 had stem trichomes of grades 3 and 4, and only two had trichomes of grade 2. This result again suggests that the pilose phenotype of Bazhou 03 was controlled by a recessive gene. A polymorphic SSR marker (D08-98, Table S8) which had a very close correlation with stem trichome phenotype (Fig. 6c) was used to genotype the F2 plants, also revealing a close correlation with stem trichome density, with only three exceptions (Table S10), suggesting that Types III and IV stem trichomes in Gb were controlled by the same gene but different alleles, and that Type III was dominant to Type IV. Above mapping efforts revealed that the different types of Gh and Gb stem trichome were mainly controlled by different combinations of genes or alleles on Chr.06 and 24. The one on Chr.24 probably plays even more important function compared to one on Chr.06, especially in Gb.
5. Interspecific and intraspecific variations in stem trichome phenotypesreflected in F1 hybrids of crosses between Gh and Gb varieties having different types of stem trichomes.
To further clarify the classification of stem trichomes, more interspecific crosses between Gh and Gb varieties with different types of stem trichomes were performed (Table 4). Interestingly, when Gh and Gb varieties having the same type of stem trichomes were crossed, the F1 hybrids always showed a similar stem trichome morphology (Table 4, Fig. S5). However, F1 plants from crosses between Gh and Gb varieties with different types of stem trichomes always displayed different stem trichomes (Table 4, Fig. S5). For example, when Gh varieties with Type Ia stem trichomes were crossed with Gb variety H7124, a representative of the glabrous varieties, the F1 hybrids did not grow trichomes on their stems. However, when these Gh varieties were crossed with Aken 4154 or L-7009, representatives of Gb varieties with Types III and IV stem trichomes, respectively, the resultant F1 always had stem trichome phenotypes close to Types III and II, respectively (Table 4, Fig. S5). These results suggested that the stem trichome alleles in Gb are dominant to those of Gh Type Ia trichomes, except for those responsible for the lengths of the F1 hybrids of Type I × Type IV, which were intermediate between the two parents. When three Gh varieties with Type Ib stem trichomes were crossed independently with the above three Gb varieties, most of the F1 plants showed different stem phenotypes from the above crosses with the same Gb varieties. When Gh varieties with Type II trichomes were crossed with H7124 and L-7009, the F1 plants had a stem trichome phenotype similar to Type II of Gh, indicating that Gh trichome genes are dominant to those of Gb. These results further confirmed that classification of stem trichome established in this study was correct and can be further used in gene cloning and cotton breeding.
6. Stem trichome density is closely related to fiber initiation and development in both hirsutum and G.barbadense
We analyzed the correlation between the densities of fuzz fibers and stem trichomes. First, we examined the fuzz fibers of Gh wild species, which showed significant differences in stem trichome phenotypes compared with modern varieties. A much greater proportion of them had no or fewer fuzz fibers than the modern varieties (~27% vs 1.4%, Table 2, Table S2), which suggested a correlation with trichome density. In addition, more wild species had shorter and denser stem trichomes than modern varieties, which was also in accordance with fiber length, because most wild Gh accessions have very short lint fibers. Thus, fuzzy/lint fibers and stem trichomes may have at least partly overlapping genetic bases for regulating their initiation and development.
To further investigate the relationship between stem trichomes and fiber densities, we observed the stem trichomes of 37 Gh fuzzless mutants and found that there was a much higher percentage of the mutants having no or few trichomes (grades 0 and 1) than in the common varieties (56.8% vs 3.7%) (Table 2, Table S3). Only 11% of the fuzzless mutants had denser trichomes (grades 3 and 4), compared with the common varieties (78%).
As indicated above, most Gb varieties did not have, or had very few, stem trichomes. The presence of naked seeds in most Gb varieties was another characteristic that differed from the Gh varieties. Out of 121 Gb varieties, 61 (50%) had less fuzz fiber on seeds (fuzz fiber density less than 2) than the modern Gh varieties. An association analysis revealed very close correlation between the densities of stem trichomes and fuzz fibers (r = 0.501**, P < 0.05) (Table S4), suggesting that the genetic foundation of Gb stem trichome development was at least partially the same as that of fibers found in Gh.