Genome-wide analysis of Rf-PPR-like (RFL) genes and a new InDel marker development for Rf1 gene in cytoplasmic male sterile CMS-D2 Upland cotton

Abstract

Background

Cytoplasmic male sterility in flowering plants is a convenient way to use heterosis via hybrid breeding and may be restored by nuclear restorer-of-fertility (Rf) genes. In most cases, Rf genes encoded pentatricopeptide repeat (PPR) proteins and several Rf genes are present in clusters of similar Rf-PPR-like (RFL) genes. However, the Rf genes in cotton were not fully characterized until now.

Results

In total, 35 RFL genes were identified in G. hirsutum, 16 in G. arboreum, and 24 in G. raimondii. Additionally, four RFL-rich regions were identified; the RFL-rich region in Gh_D05 is the probable location of Rf-PPR genes in cotton and will be studied further in the future. Furthermore, an insertion sequence was identified in the promoter sequence of Gh_D05G3392 gene in the restorer line, as compared with the CMS-D2 line and maintainer lines. An InDel-R marker was then developed and could be used to distinguish the restorer line carrying Rf1 from other genotypes without the Rf1 allele.

Conclusion

In this study, genome-wide identification and analysis of RFL genes have identified the candidate Rf-PPR genes for CMS in Gossypium. The identification and analysis of RFL genes and sequence variation analysis will be useful for cloning Rf genes in the future and also for three-line hybrid breeding in cotton.

Introduction

Cotton is an important fiber crop worldwide. Improving cotton yield and quality is becoming critical to meet industrial demands. Hybrid breeding is an important strategy to increase yield and quality by efficiently exploiting heterosis and has been applied to many important crops, including rice, maize, and cotton (Huang et al. 2016). In China, more than 90% of cotton hybrids are produced by artificial emasculation and pollination (Yu et al. 2016). It is time-consuming, labor-intensive, and costly and the purity of hybrid seeds cannot be guaranteed, representing an important limiting factor for hybrid seed production. One of the major challenges is the absence of a pollination control strategy that could efficiently produce hybrid seed on a commercial level. In other crops, cytoplasmic male sterility (CMS) is an indispensable resource for commercial hybrid seed production (Schnable and Wise 1998; Hanson and Bentolila 2004; Chase 2007; Pelletier and Budar 2006).

CMS is a maternally inherited trait in flowering plants that cannot produce functional pollen (Hanson and Bentolila 2004). The CMS trait is caused by the rearrangement of the mitochondrial genome and several CMS genes have been identified in many crops (Schnable and Wise 1998; Hanson and Bentolila 2004; Chase 2007). The products of CMS genes destroy the normal function of mitochondria and cause a deficiency in the energy supply required for pollen development, resulting in aborted pollen (Schnable and Wise 1998). The CMS phenotypes could be restored by the fertility restorer (Rf) genes from the nuclear genome. Previous studies have indicated that the Rf genes identified in petunia (Bentolila et al. 2002), radish (Brown et al. 2003; Desloire et al. 2003), rice (Tan et al. 2004, 2008; Fujii et al. 2014; Igarashi et al. 2016), and sorghum (Klein et al. 2005) belong to a pentatricopeptide repeat (PPR) gene family. Exceptions are the maize Rf2, which encodes an aldehyde dehydrogenase that may be involved in the production of the plant hormone indole-3-acetyl acetate (Cui et al. 1996; Liu and Schnable 2002), and the Rf2 gene in rice for Lead-type CMS that encodes a protein containing a glycine-rich domain (Itabashi et al. 2010). Additionally, three PPR genes cosegregated with the Rf3 gene of S type CMS in maize (Xu et al. 2009), and the Rf5 gene in rice encodes a PPR protein interacting with a glycine-rich domain protein (GRP) which restores fertility in Hong-Lian CMS lines (Hu et al. 2012). These studies indicated that PPR genes have important relationships with the Rf genes in plants.

In cotton, two main CMS systems, CMS-D2–2 and CMS-D8, have been developed by transferring exotic cytoplasm from Gossypium harknessii Brandegee (D2) and G. trilobum (DC.) Skovst. (D8) into the Upland cotton (G. hirsutum, AD1) nuclear background (Meyer 1975; Yin et al. 2006; Zhang et al. 2007; Wang et al. 2010; Wu et al. 2011). So far, no studies have reported the cloning of cotton Rf genes, with most studies focusing on genetic mapping and the development of related markers. Previous studies have indicated that the Rf1 gene from G. harknessii (D2) can restore the fertility of both CMS-D2 and CMS-D8, whereas the Rf2 gene from G. trilobum only restores male fertility to CMS-D8 (Zhang and Stewart 2001a, 2001b). Additionally, the Rf1 and Rf2 genes in cotton function sporophytically and gametophytically, respectively. These two restorer genes are not allelic but tightly linked in 0.93 cM (Yin et al. 2006; Wang et al. 2009; Wu et al. 2011, 2014). Yin et al. (2006) identified that the marker NAU4047 is closely linked to Rf1 (within 0.2 cM) and delimited the Rf1 gene to a 100-kb region. Furthermore, the Rf1 gene is located on the Gh_D05 chromosome, with genetic mapping indicating that the nearest SSR markers to Rf1 are BNL3535 (within 0.049 cM) and NAU3652 on the other side (within 0.078 cM). An Rf1-specific CAPS marker was developed based on a candidate PPR gene and could ensure the purity of restorer lines (Wang et al. 2007, 2009; Wu et al. 2014). Wang et al. (2007) constructed a linkage map with nine markers flanking the Rf2 gene including a PPR-AFLP marker. A whole-genome resequence was completed for the restorer N (Rf1Rf1) and maintainer N (rf1rf1) lines that indicated that most of the InDels were distributed near the region containing the Rf1 gene in Gh_D05. Furthermore, an InDel-1891 marker was developed for fine mapping of the Rf1 gene (Wu et al. 2017).

The PPR gene family constitute a large family of RNA-binding proteins in plants and the members are involved in many cellular functions and biological processes in organelles, including gene expression, RNA stabilization, RNA cleavage, and RNA editing (Schmitzlinneweber and Small 2008; Prikryl et al. 2010). Previous studies indicated that all cloned Rf-PPR genes might have a common ancient ancestor and that Rf-CMS genes have coexisted during the evolutionary process (Geddy and Brown 2007; Fujii et al. 2011; Joanna et al. 2016; Sykes et al. 2017). For example, Rf1a and Rf1b genes in rice share 70% identity between their protein sequences (Wang et al. 2006) while in radish the Rf3 protein shows 85% similarity with the Rf0 protein (Wang et al. 2013). Additionally, several studies indicated that Rf-PPR genes are targeted to mitochondria where they prevent the accumulation of the CMS-specific gene product (Bentolila et al. 2002; Wang et al. 2006; Kazama et al. 2008). Furthermore, these Rf-PPR genes are presented in clusters of similar Rf-PPR-like (RFL) genes in almost all cases (Bentolila et al. 2002; Wang et al. 2006; Kazama et al. 2008; Uyttewaal et al. 2008; Barr and Fishman 2010). RFL genes at the same genomic region are most likely to be active restorer genes and several PPR-Rf genes present within the RFL-rich region such as the rice Rf1 and Rf4 genes presented in the RFL-rich region of rice chromosome 10 (Wang et al. 2006; Fujii et al. 2011; Luo et al. 2013). Additionally, the Rf5 gene in rice was mapped to a 200-kb region on chromosome 8 that contains three RFL genes, one of which, Os08g01870, was located within 15 kb of the marker and cosegregated with the Rf gene (Hu et al. 2012; Huang et al. 2016). In maize, the Rf8 locus was mapped to an RFL cluster on chromosome 2 (Meyer et al. 2011). The only PPR-Rf gene identified in sorghum was found to be located outside of the RFL-rich regions, however, occurs on chromosome 8. This gene most likely encodes a PPR protein belonging to the PLS (P-L-S motifs) subfamily that is involved in RNA editing events, indicating that the mechanism of fertility restoration in sorghum may be unique (Klein et al. 2005; Schmitzlinneweber and Small 2008; Dahan and Mireau 2013). This allowed us to further explore the candidate Rf genes in cotton by identifying the RFL-rich region that shows a similar pattern to other species.

In cotton, we have characterized the DYW (Asp-Tyr-Trp tripeptide in C terminal domain) deaminase domain-containing PPR genes belonging to PLS subfamily and have determined that these genes may not directly function in the occurrence of CMS or in fertility restoration, while P (common PPR motif) subfamily genes might have a critical role in the fertility restoration process (Zhang et al. 2017). However, no results have been reported regarding the identification and analysis of RFL genes in cotton until now. Here, to identify the candidate Rf-PPR genes for CMS in cotton, a genome-wide identification and analysis of RFL genes were completed in Gossypium. The RFL genes identified and analyzed in our study will be useful for cloning the Rf genes and for three-line cotton hybrid breeding in the future.

Materials and methods

Cotton genome and RNA-seq resources

The genome sequence and annotation information of three Gossypium species (G. raimondii, G. arboreum, and G. hirsutum) were downloaded from Cottongen (https://www.cottongen.org). The raw sequence data of a 3 mm floral bud transcriptome from three-line hybrid cotton (CMS-D2 line A, maintainer line B, and restoration line R) could be found in the National Center for Biotechnology Information (NCBI) under accession number SRX3421007.

Identification and chromosomal mapping RFL genes in Gossypium

To precisely identify the RFL genes in Gossypium, BLAST (http://www.ncbi.nlm.nih.gov/Tools/) was used to search sequences in three cotton genomes. The sequence of Rf-PPR592 from Petunia hybrida identified previously was used for searches against the whole genome database of the three cotton species. Hits with an estimated E-value under 1e^− 100 were set as threshold (Fujii et al. 2011). The number of PPR domains in the protein structure was further validated using SMART software (http://smart.embl-heidelberg.de).

The physical location data of RFL genes were retrieved from genome sequence data of three cotton species. Mapping of these RFL genes was then performed using Mapchart software (Voorrips 2002).

Subcellular location analysis

The signal peptide prediction program Target P (http://www.cbs.dtu.dk/services/TargetP/) was used to predict the subcellular location of RFL proteins.

Quantitative (q) RT-PCR validation of DEG expression

The CMS-D2 three-line hybrid cotton system was obtained from the Institute of Cotton Research, Chinese Academy of Agricultural Science (ICR, CAAS). The three lines were planted under normal production conditions. Samples were collected as described previously (Wu et al. 2011; Suzuki et al. 2013); floral buds approximately 3 mm in length (corresponding roughly to the meiosis stage) were collected with three independent biological replicates. All collected floral buds were cut above ovaries and immediately frozen in liquid nitrogen and stored at − 80 °C. Total RNAs were extracted from floral buds and reverse transcribed to cDNA using a PrimeScript RT reagent kit (Takara, Dalian) following the manufacturer’s guidelines. For qRT-PCR, reactions were performed in 20-μL volumes containing 1 μL diluted cDNA, 10 μL 2× SYBR Green Mix (Takara), 7 μL water and 1 μL each of forward primer and reverse primer. The amplifications were carried out as follows: 94 °C for 30 s, then 40 cycles of 94 °C for 5 s, 55 °C for 15 s, and 72 °C for 25 s. The cotton histone 3 (GhHIS3) was used as a reference gene for normalization. All the primers were listed in Additional file 1: Table S1.

Promoter sequence analysis and InDel marker development

Total genomic DNA from the three lines was extracted from leaves using the CTAB method (Paterson et al. 1993), respectively. Additionally, gene-specific primers were designed by using Primer Premier 5.0 software (http://www.premierbiosoft.com) to amplify the promoter sequence of Gh_D05G3392 gene in the A, B and R lines. A 20-μL mixture consisting of 1× reaction buffer, 2.0 mmol·L^− 1 MgCl₂, 0.2 mmol·L^− 1 dNTPs, 0.5 mmol·L^− 1 of each primer, 1 U Taq DNA polymerase (Takara, Japan), and 50 ng DNA template was used. The PCR procedure was as follows: 35 cycles of 94 °C for 30 s, then 58 °C for 30 s, and 72 °C for 60 s. The PCR mixture was separated and purified by TaKaRa DNA Fragment Purification Kit. Then the DNA fragment was ligated into the pEASY-T1 vector (TransGen, Beijing), following the manufacture’s guidelines. Then five clones were selected in every sample for sequencing. The MEGA7.0 was used for sequence alignment.

The cis-acting element identification in the promoter region was completed by using plant cis-acting regulatory DNA elements (https://www.dna.affrc.go.jp/htdocs/PLACE/).

An InDel-R marker was then developed and the primer pair (forward: 5′- GAAAGTTGGACAACAATGAGAAGTC-3′; reverse: 5′- CCAATTTCTAATAAAGAAAAGAAAGAG-3′) were designed for applications. A 20-μL mixture consisting of 1× reaction buffer, 2.0 mmol·L^− 1 MgCl₂, 0.2 mmol·L^− 1 dNTPs, 0.5 mmol·L^− 1 of each primer, 1 U Taq DNA polymerase (Takara, Japan), and 50 ng DNA template was used. PCR was performed as follows: 30 cycles of 94 °C for 30 s, then 56 °C for 30 s, and 72 °C for 10 s. The PCR products were then separated using 3.0% agarose gel electrophoresis.

Results

Genome-wide identification and chromosomal distribution of RFL genes in Gossypium

To identify potential RFL genes in the G. hirsutum, G. arboreum, and G. raimondii protein databases, the sequence of Rf-PPR592 from P. hybrida was used for BLAST searching against the three cotton genomes, as per the previous study by Fujii et al. (2011). Hits with an estimated E-value under 1e ^− 100 were collected (Fujii et al. 2011). In total, 75 RFL genes were identified, of which 35 were obtained from G. hirsutum, 16 from G. arboreum, and 24 from G. raimondii. Analysis of the 75 predicted cotton RFL proteins, which identified by homology to the known restorer genes Rf-PPR592 from P. hybrida, revealed that these proteins also belonged to the P subfamily. Further analysis indicated that the number of PPR motifs in the proteins ranged from 9 to 20 (Table 1).

Table 1 Characteristics of RFL genes and predicted properties of RFL proteins in three Gossypium species

The 35 RFL genes which identified from G. hirsutum were found to be located on 15 chromosomes, with 17 and 18 genes distributed to the A and D sub-genomes, respectively (Fig. 1), with the Gh_A04G1306 and Gh_A04G1307 genes localized to scaffold756_A04. Additionally, six and five genes were located on chromosome 5 and 10 in the D sub-genome, respectively. Chromosomes 1, 5, 6, 7, 12, and 13 in the A sub-genome and chromosomes 1, 4, 6, 7, and 12 in the D sub-genome were the exceptions and did not contain any RFL genes. Previously, the rice Rf1 (Wang et al. 2006) and Rf4 (Luo et al. 2013) genes were found to occur in the RFL-rich region of rice chromosome 10. In our study, four RFL-rich regions were identified, including three RFL genes in Gh_A04, four RFL genes in Gh_A10, six RFL genes in Gh_D05, and five RFL genes in Gh_D10. The RFL genes in these regions will be studied further.

Fig. 1

The putative chromosome location of RFL genes on G. hirsutum. The scale represents megabases (Mb). The red column represents the RFL-rich region

Expression patterns of RFL genes and qPCR validation

Additionally, because of the tissue and time-specific expression of RFL genes (Prasad et al. 2003; Tomohiko and Kinya 2014), transcriptomic data from 3 mm floral buds of three-line hybrid cotton (CMS-D2 line (A), maintainer line (B), and restoration line (R)) were used to identify candidate Rf-PPR genes within the RFL-rich region (Fig. 2) (Additional file 2: Table S2). Interestingly, three genes (Gh_D05G3356, Gh_D05G3389, and Gh_D05G3392) in Gh_D05 were up-regulated in the R line as comparing with the A and B lines. To verify the expression profiles of the RFL genes, three genes (Gh_D05G3356, Gh_D05G3389, and Gh_D05G3392) were selected for qPCR analysis using the 3 mm floral buds from the A, B, and R lines. Their gene expression patterns were similar to the RNA-seq data and indicated that all three genes were up-regulated in the R line as comparing with the A and B lines. This suggests that these genes might play critical roles in fertility restoration.

Fig. 2

qRT-PCR analysis of RFL gene expression compared with the RNA-seq data from three-line hybrid cotton lines (CMS-D2 line A, maintainer line B, and restoration line R). a The RNA-seq data of RFL genes in A, B and R lines. b The qRT-PCR analysis of three differentially expressed RFL genes. The red columns represented the relative expression levels of the genes; the black lines represented the FPKM number. A: CMS-D2 line, B: maintainer line, R: restorer line

Sequence variation of DEGs on Chr_05

Furthermore, the above transcriptomic data were further used to identify single nucleotide polymorphism (SNPs) in the three differentially expressed RFL genes (Gh_D05G3356, Gh_D05G3389, and Gh_D05G3392) on Chr_05. In total, 37 SNP loci were identified between the sequences from the R line and that from the non-restoring genome A and B lines (Additional file 3: Table S3). The results implied that these SNPs might be linked to the fertility restoring gene on Chr_05. In addition, promoter sequence analysis of Gh_D05G3392 gene among the A, B, and R lines was also conducted. Consistent with the coding region between the R line and the A and B lines, a high level of polymorphisms was observed in the promoter region (Fig. 3). Multiple alignments indicated that several SNP loci and seven InDels specifically exist between the restoration R line and the non-restoring genome A and B lines. Furthermore, there was a 12 nt insertion “TAGAAGACTGGA” in the restorer line as comparing with the A and B lines.

Fig. 3

Multiple alignments for promoter sequence of Gh_D05G3392 gene in CMS-D2 line A, maintainer line B, and restoration line R. The red color represent the translation start site. The yellow color represent the insertion in the restorer R line. The blue color represent specific cis-regulatory elements in the promoter region

A search for cis-acting elements in the promoter region of Gh_D05G3392 gene was completed by using plant cis-acting regulatory DNA elements (https://www.dna.affrc.go.jp/htdocs/PLACE/). Except for the core promoter element “TATA” box, we also found other motifs associated with light responsiveness (GA-motif (AAGGAAGA) and I-box (GATATGG)) and a TCA-element (CCATCTTT) involved in salicylic acid responsiveness. Furthermore, five copies of the pollen specific motifs POLLEN1LELAT52 (AGAAA) (Filichkin and Nonogaki 2004) were also identified, which indicated that transcriptional activation of Gh_D05G3392 gene might be controlled by the pollen specific cis-regulatory elements.

An InDel-R marker was then developed for this insertion sequence that was verified as a co-dominant marker in the three lines. A total of 24 randomly selected individual BC₅F₂ plants were checked using this InDel-R marker. As shown in Fig. 4, the InDel-R marker could be used to distinguish the restorer line carrying Rf1 from other genotypes without the Rf1 allele. The result showed three different PCR band models in which a single PCR band of nearly 149 base pairs (bp) represented plants homozygous for the Rf gene allele N(Rf1Rf1) and a single PCR band of nearly 137 bp represented plants lacking the Rf gene allele (rf1rf1). Plants containing both PCR bands were considered heterozygous at the Rf gene locus N(Rf1rf1). These results indicated that this InDel-R marker could be used in the marker-assisted breeding of fertility restoration lines carrying the Rf1 gene.

Fig. 4

BC₅F₂ plants were screened with InDel-R. M: DNA marker III, a plants lacking the restorer gene Rf1, b Rf1 homozygous plants Rf1, c Rf1 heterozygous plants

Discussion

Previous studies have indicated that most Rf genes came from the same small clade of PPR genes, with many similarities and are usually presented as clusters of similar Rf-PPR-like (RFL) genes in many plants (Bentolila et al. 2002; Kazama et al. 2008; Uyttewaal et al. 2008; Barr and Fishman 2010; Fujii et al. 2011). The importance of the Rf gene in the CMS/Rf system of cotton resulted in many studies aiming to identify molecular markers linked to the Rf gene; there have been no reports regarding cloning of the Rf gene until now. In this study, we performed genome-wide identification and analysis of RFL genes in G. hirsutum, G. arboreum, and G. raimondii to identify candidate Rf genes for CMS in cotton.

The RFL genes in Gossypium

In the draft genome sequence of cotton, a total of 35 RFL genes were identified from G. hirsutum; this is in contrast to previous studies that have suggested the presence of around 10–30 RFL genes per plant genome (Andrés et al. 2007; Fujii et al. 2011; Joanna et al. 2016; Sykes et al. 2017). This difference may be associated with the polyploidization of Upland cotton that has resulted in whole genome duplication (WGD). Additionally, 16 and 24 RFL genes were identified from G. arboreum and G. raimondii, respectively. Gene structure analysis revealed that RFL genes only contain the PPR domain and that these genes belong to the P subfamily.

Identification of an RFL-rich region

Previous studies have indicated that Rf-PPR genes are targeted to mitochondria where they prevent the accumulation of the CMS-specific gene products (Bentolila et al. 2002; Kazama et al. 2008; Uyttewaal et al. 2008; Barr and Fishman 2010; Fujii et al. 2011). RFL genes in the same genomic region are most likely active restorer genes, with several PPR-Rf genes presenting within the RFL-rich region, such as the rice Rf1 and Rf4 genes in the RFL-rich region of rice chromosome 10 (Wang et al. 2006; Fujii et al. 2011; Luo et al. 2013; Huang et al. 2016; Sykes et al. 2017). Additionally, the Rf6 gene in rice was mapped to a 200-kb region on chromosome 8 that contains three RFL genes. Of these, Os08g01870 was located within 15 kb of the marker and cosegregated with the Rf gene (Hu et al. 2012; Huang et al. 2016). The only identified PPR-Rf gene in sorghum is, however, located outside the RFL-rich regions on chromosome 8. This gene most likely encodes a PPR protein belonging to the PLS subfamily that is involved in RNA editing events, indicating that the mechanism of fertility restoration in sorghum may be unique (Klein et al. 2005; Schmitzlinneweber and Small 2008; Dahan and Mireau 2013). This allowed us to further refine the candidate Rf genes in cotton by identifying the RFL-rich region common to other species. Previous studies indicated that Rf1 and Rf2 in cotton functioned sporophytically and gametophytically, respectively, and that the two Rf genes are not allelic but are tightly linked in 0.93 cM (Wang et al. 2007; Wang et al. 2009; Wu et al. 2011). Furthermore, the Rf1 gene is located on chromosome Gh_D05 and genetic mapping has indicated that the nearest SSR marker to Rf1 was BNL3535 (within 0.049 cM) and NAU3652 on the other side (within 0.078 cM) (Wang et al. 2007; Wu et al. 2014). In this study, four RFL-rich regions were identified in four chromosomes with six RFL genes found to cluster in the Gh_D05 chromosome near the Rf region. Contrary to our expectations, six RFL genes were not targeted to the mitochondria based on the TargetP software prediction. This may be because some RFL genes were overlooked because of assembly errors and gaps in the draft genome or because of repetitive features in the RFL-rich genomic regions. For example, most of the InDels were distributed near the region of the Rf1 gene on chromosome Gh_D05 in cotton (Wu et al. 2017). In barley, an RFL gene was identified on an unordered contig from the chromosome 6HS containing a recently mapped Rf locus that could not be associated with an RFL cluster (Tsai et al. 2010; Ui et al. 2014).

Furthermore, a Rf1-specific CAPS marker was developed based on a SNP occurring within a PPR gene and an InDel-1891 marker was developed for fine mapping of the Rf1 gene (Wu et al. 2014; Wu et al. 2017). The application of these markers could ensure the purity of restorer lines in cotton. In this study, three genes (Gh_D05G3356, Gh_D05G3389, and Gh_D05G3392) were up regulated in the R line as compared with the A and B lines. In total, 37 SNP loci in these three genes were identified between the R line and the A and B lines. Furthermore, a 12 nt insertion “TAGAAGACTGGA” was identified in the promoter region of Gh_D05G3392 in the restorer R line as comparing with the A and B lines. An InDel-R marker was then developed for this insertion sequence that could be used to distinguish the restorer line carrying Rf1 from other genotypes without the Rf1 allele. The results implied that these SNPs and InDels might be used for fine mapping of the Rf1 gene in cotton.

Conclusion

In our study, we tried to identify candidate Rf-PPR genes for CMS in cotton via genome-wide identification and analysis of RFL genes in G. hirsutum, G. arboreum, and G. raimondii. Furthermore, four RFL-rich regions were identified. Within one of these regions on Gh_D05, expression of three RFL genes was up-regulated in the R line as comparing with the A and B lines. Sequence variation analyses indicated that several SNPs and InDels exist in the R line as comparing with the non-restoring genome A and B lines, providing excellent sites for marker development and further mapping approaches. An InDel-R marker was then developed that could be used to distinguish the restorer line carrying Rf1 from other genotypes without the Rf1 allele. These results will not only be useful for guiding future identification and cloning of Rf genes responsible for CMS but will also be useful in heterosis in cotton.