Genome-wide identification of the BASS gene family in four Gossypium species and functional characterization of GhBASSs against salt stress

Bile acid sodium symporter (BASS) family proteins encode a class of sodium/solute symporters. Even though the sodium transporting property of BASSs in mammals was well studied, their sodium transportability and functional roles in plant salt tolerance remained largely unknown. Here, BASS family members from 4 cotton species, as well as 30 other species were identified. Then, they were designated as members of BASS1 to BASS5 subfamilies according to their sequence similarity and phylogenetic relationships. There were 8, 11, 16 and 18 putative BASS genes in four cotton species. While whole-genome duplications (WGD) and segmental duplications rendered the expansion of the BASS gene family in cotton, BASS gene losses occurred in the tetraploid cotton during the evolution from diploids to allotetraploids. Concerning functional characterizations, the transcript profiling of GhBASSs revealed that they not only preferred tissue-specific expression but also were differently induced by various stressors and phytohormones. Gene silencing and overexpression experiments showed that GhBASS1 and GhBASS3 positively regulated, whereas GhBASS2, GhBASS4 and GhBASS5 negatively regulated plant salt tolerance. Taken together, BASS family genes have evolved before the divergence from the common ancestor of prokaryotes and eukaryotes, and GhBASSs are plastidial sodium-dependent metabolite co-transporters that can influence plant salt tolerance.


Results
Identification and phylogenetic analysis of BASS family members in different species. In this study, our main focus was four cotton species, so we first compared publicly available genome databases and finally selected Beijing Genomics Institute sequenced genome (BGI) for G. arboretum, Plant Genome Mapping laboratory (University of Georgia) sequenced genome (JGI) for G. raimondii, Nanjing Agricultural University sequenced genome (NAU-NBI) for G. hirsutum and Chinese National Human Genome Center sequenced genome (CHGC) for G. barbadense. After eliminating redundant sequences, 7 putative BASS genes and 2 pseudogenes from the G. arboretum genome, 11 putative BASS genes and 1 pseudogene from the G. raimondii genome, 15 putative BASS genes from the G. hirsutum genome and 18 putative BASS genes and 3 pseudogenes from the G. barbadense genome were respectively identified by performing the hmmsearch to their respective protein sequence databases. Since the released genome and annotation were not flawless, these BASS gene sequences were revised by comparing them with the genomic sequences, followed by correction and integration with the data from EST and SRA databases. After extensive bioinformatics analyses, a total of 8 and 11 putative BASS genes from the genomes of diploid species (G. arboretum and G. raimondii) and 16 and 18 putative BASS genes from the genomes of tetraploid species (G. hirsutum and G. barbadense) were retrieved (Tables 1, S1, S2). After performing Pfam (http:// pfam. xfam. org/) and NCBI-CDD (https:// www. ncbi. nlm. nih. gov/ Struc ture/ cdd/ wrpsb.cgi), all 53 candidate genes retained the conserved SBF and bass domain (Fig. S1), which validates the typical configuration of BASS family members. What is more, transmembrane regions and Na + -binding sites are unique characteristics of BASS genes, and all these candidates possessed these qualities also (Fig. S2). www.nature.com/scientificreports/ Additionally, we identified BASS genes from 30 other species including three bacteria, three algae, one basal land plants, one lycophytes, one gymnosperms and 21 angiosperms (Table S2). We found that BASS family genes were firmly maintained in all species from prokaryotes to eukaryotes selected in this study; however, the number of BASS members in prokaryotes and algae was less than those in embryophytes (Fig. 1). For example, there was only one BASS gene in two prokaryotic bacteria Neisseria meningitidis and Escherichia coli, but three BASS genes were identified in bacteria Yersinia frederiksenii. While there were three BASS genes in each of unicellular green algae Micromonas pusilla and Dunaliella salina, four BASS genes were observed in marine green algae Ostreococcus tauri.
To gain an insight into the evolutionary relationships of BASS genes, we aligned cotton protein sequences (53 genes) with 56 predicted BASS sequences from another 13 sequenced species including bacteria, plants and mammals for the construction of the phylogenetic tree. The BASSs could be classified into five subfamilies (Subfamily I to Subfamily V) according to the topology of the phylogenetic tree. All BASS1 genes belonged to subfamily I, BASS2 genes belonged to subfamily II, BASS3 genes belonged to subfamily III, BASS4 genes belonged to subfamily IV and BASS5 genes belonged to subfamily V, respectively (Fig. 2). Subfamily I, the BASS1 gene subfamily, stood as the largest group. The second-largest clade was the BASS4 gene subfamily, and BASS genes from bacteria and mammals were fallen into this group, which was followed by BASS3, BASS2 and BASS5 subfamily, respectively.
With respect to evolutionary relationships of BASS genes identified from different species, we found that members of the BASS4 subfamily were consistently retained in all selected species, even in prokaryotic bacteria that contained only BASS4 subfamily members (Figs. 1, 2). Intriguingly, fewer numbers of BASS4 subfamily genes were observed in almost tested angiosperm species. There were members of 2-3 BASS subfamilies in unicellular green algae, i.e., Micromonas pusilla, Dunaliella salina and Ostreococcus tauri. Members of 4 BASS subfamilies were discovered in basal land plants (Physcomitrella patens), lycophytes (Selaginella moellendorffii) and gymnosperms (Picea abies), while members of all subfamilies were kept in basal angiosperms (Amborella trichopoda). All tested dicotyledons and monocotyledons maintained all BASS subfamilies except for three C4 Table 1. Characteristics of putative BASS genes from four cotton species. BASS family genes were identified from four cotton genomes using the HMMER software version 3.0. BASS, bile acid sodium symporter. Ga Gossypium arboretum, Gr Gossypium raimondii, Gb Gossypium barbadense, Gh Gossypium hirsutum, AA Amino acid. *Pseudogenes. www.nature.com/scientificreports/ monocots (Phalaenopsis equestris, Sorghum bicolor and Zea may) in which there was a lack of the BASS3 subfamily gene. In general, the number of BASS genes was obviously larger in eudicots compared with monocots ( Fig. 1). Hence, BASS family genes are very conserved starting from the prokaryotic bacteria and unicellular lower eukaryotic green algae to the multicellular angiosperm species. Collectively, BASS4 is assumed as the oldest gene, and this gene tends to be restored to the 'singleton' state in most tested angiosperm species by recently independent genome duplication events.
Phylogenetic relationships, gene architectures and conserved motifs of cotton BASSs. BASSs from four Gossypium species were tightly clustered in every subfamily. Moreover, BASSs from Gossypium species were closely clustered with those from Theobroma cacao, indicating that they shared the high sequence similarity and had close evolutionary relationships (Fig. 2). In addition, to detect the evolutionary relationships of BASSs in G. arboreum, G. raimondii, G. barbadense and G. hirsutum, the phylogenetic tree was also built using 53 cotton BASSs in which they were divided into five subfamilies too (Fig. 3A). In theory, one BASS gene in the diploid G. arboreum corresponds to one orthologous BASS gene in G. raimondii and two orthologous genes in the tetraploid G. hirsutum and G. barbadense. However, more BASS genes were found in G. raimondii (GrBASS3-2b, GrBASS5b and GrBASS5c) and G. barbadense (GbBASS3-2b and GbBASS5b). The BASS3 gene subfamily was composed of 14 members, each of BASS1 and BASS2 subfamilies consisted of 12 members, 9 members were grouped to form the BASS5 subfamily and the other 6 were for the BASS4 clade which served as an outgroup for other subfamilies (Fig. 3A).
In order to investigate gene structures of Gossypium BASS family genes, their exon/intron organizations were analyzed by mapping the coding sequences onto their genomic sequences. Overall, exon/intron structures of Gossypium BASS genes were relatively conserved within a subfamily, but they were quite diverse among subfamilies (Fig. 3B). For further details, BASS4 subfamily members owned the longest genomic sequences (14,006 to 24,121 bp) with 11-14 exons and comparatively longer introns. Within the BASS1 subfamily, there were two subgroups of which group one members kept the resembled formation of exons and introns except for GhBASS1-1A. On the other hand, group two was composed of genes with varied structural patterns of exon/ Figure 1. Species phylogeny and the number of BASS genes in each species. The species tree was inferred from Zeng et al. 21 . The divergence time was estimated by molecular clock dating from TimeTree 22 . Stars on the branches represent WGD events. ζ, ancestral seed plant WGD; ε, ancestral angiosperms WGD; γ, triplication event; β and α, two recent WGD events in Eudicots; σ and ρ, two polyploidy events in monocots. The scientific name, common name and group are followed by the number of BASSs identified. MYA, million years ago. www.nature.com/scientificreports/ intron, in which GbBASS1-2D carried the shortest genomic sequence (1145 bp) bearing 5 exons, leading to variation in the gene length of that group. The majority of participants in the BASS2 clade were comparatively less diverse in exon/intron structures, having 12 to 14 exons. Of the 9 BASSs in the BASS5 cluster, GbBASS5Da (8 exons) and GbBASS5Db (7 exons) were slightly different in exon/intron patterns, whereas the others preserved the resembled structure. With regard to the BASS3 clade, members showed similar exon/intron structure and gene length except for GbBASS3-1D, GaBASS3-1 and GhBASS3-2A (Fig. 3B, Table S1). These results implied that numbers of exon/intron and structural organization of Gossypium BASS family genes are well correlated with the clades identified in the phylogenetic tree for each group.
To further clarify the structural feature of cotton BASS genes, the diversity of motif composition in 53 Gossypium BASS proteins was assayed by the MEME software. A total of 25 motifs designated as motif 1 to motif 25 were examined. Generally speaking, both of motif numbers and their distribution patterns in BASSs within a clade were fairly conservative (Fig. 3C). Motif numbers and patterns in the genes of the BASS1 clade were relatively varied, and there were 13 motifs of which motif 17 appeared as a unique motif of those genes belonged to this clade.11 motifs were found in the genes existed in the BASS4 cluster, and motif 15, 20 and 23 were unique for the genes of that cluster. The numbers of motifs in the members of the BASS2 and BASS5 subfamily were 16 and 14; motif 21 was exclusive to the members of the BASS2 subfamily, while motif 11 emerged only in genes of the BASS5 subfamily. The largest numbers of motifs (17 motifs) were produced by the genes involved in the BASS3 subfamily, and motif 10 and 18 were entirely restricted to these genes. Motif 1 was ubiquitous in every gene across all subfamilies, and motif 2 stood as a common motif except for the genes of the BASS4 subfamily (Fig. 3C). These results suggested that the similarity of gene structure and motifs distribution in each BASS subfamily provides evidence for the effective analysis of the phylogenetic tree. GhBASSs were anchored on the chromosome A01, A04, A06, A07, A08, A12, A13, D01, D04, D06, D07, D08, D12 and D13, and every other chromosome except A12 and D12 which had two BASSs received one gene. Apart from GbBASS3-2Db (D02) and GbBASS5Db (D10), the chromosomal localization of 16 other G. barbadense BASSs was similar to the localization of orthologous counterpart genes of G. hirsutum (Tables 1, S1). Whole-genome duplication (WGD) is an event which results in additional copies of the entire genome, which is also known as polyploidization 24 . If two or more paralogous genes were adjacent to each other on the same chromosome with no more than one intervening gene, they were defined as tandemly duplicated genes, whereas gene duplications between different chromosomes were designated as segmental duplications 25 . Following the criteria for inferring a gene duplication event 26 , we identified 3 duplicated gene pairs (GaBASS1-1/GaBASS1-2, GaBASS2-1/GaBASS2-2 and GaBASS3-1/GaBASS3-2) from G. arboretum, while 2 duplicated gene pairs (GrBASS1-1/GrBASS1-2 and GrBASS2-1/GrBASS2-2) and 2 triplicated gene pairs (GrBASS3-1/GrBASS3-2a/GrBASS3-2b and GrBASS5a/GrBASS5b/GrBASS5c) were found in G. raimondii. In tetraploid Gossypium species, BASS1, BASS2 and BASS3 genes were tetraplicate; however, there were 2 copies for BASS4 and BASS5 in G. hirsutum, as well as 2 copies for BASS4 and 3 copies for BASS5 in G. barbadense, respectively. Of the 53 genes from 4 Gossypium spp., only 2 genes were tandemly duplicated (GrBASS3-2a/GrBASS3-2b), whereas the others were derived from the segmental duplication. Besides, whole-genome duplications (WGD) resulted in www.nature.com/scientificreports/ more gene numbers of tetraploid Gossypium species, demonstrating its significant contribution to gene family expansion ( Table 1, Fig. 4).
To further explore the genetic origins and evolutionary relationships of BASSs in cotton, we carried out the synteny analysis between the G. raimondii and Vitis vinifera genome, between the G. raimondii and Theobroma cacao genome, between the G. raimondii and G. arboreum genome, between the genome of G. arboreum and At-subgenomes of cultivated tetraploid cotton and between the genome of G. raimondii and Dt-subgenomes of cultivated tetraploid cotton. While G. raimondii vs. V. vinifera shared 14 inter-species syntenic regions harboring 4 orthologous gene pairs, 6 orthologous gene pairs were distributed in 16 inter-species syntenic regions of G. raimondii vs. T. cacao (Fig. 4A,B). 13 collinearity blocks with 8 orthologous gene pairs were detected between the Chromosomes of G. raimondii (Gr01-13) and T. cacao (Tc01-10) were filled with orange and purple, respectively. Purple lines connect orthologous gene pairs between T. cacao and G. raimondii. (C) Synteny analysis between G. arboreum and G. raimondii. Chromosomes of G. raimondii (Gr01-13) and G. arboreum (Ga01-13 and scaffold2621) were filled with orange and green, respectively. Blue lines bridge orthologous gene pairs between G. arboreum and G. raimondii. (D) Synteny analysis between G. hirsutum and two diploid species (G. arboreum and G. raimondii). Purple, orange and green colored bars were depicted as chromosomes of G. hirsutum (GhA01-13 and GhD01-13), G. raimondii (Gr01-13) and G. arboreum (Ga01-13 and scaffold2621), respectively. Green lines link orthologous gene pairs between G. arboreum and G. hirsutum, orange lines connect orthologous gene pairs between G. raimondii and G. hirsutum, and purple lines bridge homoeologous gene pairs between At-and Dt-subgenome of G. hirsutum. (E) Synteny analysis between G. barbadense and two diploid species (G. arboreum and G. raimondii). Blue, orange and green colored bars were depicted as chromosomes of G. barbadense (GbA01-13 and GbD01-13), G. raimondii (Gr01-13) and G. arboreum (Ga01-13 and scaffold2621), respectively. Green lines link orthologous gene pairs between G. arboreum and G. barbadense, orange lines connect orthologous gene pairs between G. raimondii and G. barbadense, and blue lines bridge homoeologous gene pairs between At-and Dt-subgenome of G. barbadense. Grey lines represent conserved synteny blocks between two different genomes and/or subgenomes of species. Putative BASS family genes were anchored to their corresponding chromosomes, and the symbol (*) means pseudogenes. www.nature.com/scientificreports/ G. raimondii and G. arboreum genome (Fig. 4C). Next, there were 8 syntenic blocks with 7 orthologous gene pairs between G. arboreum and At-subgenome of G. hirsutum, 9 blocks having 8 orthologous gene pairs between G. raimondii and Dt-subgenome of G. hirsutum and 9 blocks bearing 6 homoeologous gene pairs between At-and Dt-subgenome of G. hirsutum (Fig. 4D). Additionally, a sum of 23 collinearity blocks was identified between the diploid cotton and G. barbadense. Among them, two blocks individually possessed two orthologous gene pairs, but one harbored no gene pair. In contrast, the other collinearity blocks individually contained one gene pair (Fig. 4E). Based on these results, the analysis of syntenic blocks carrying orthologous or homoeologous genes showed important evidence to verify the evolutionary relationships of BASS genes in cotton.
To better understand the divergence and selection pressure of Gossypium BASS genes after polyploidization, values of nonsynonymous to synonymous substitution ratios (Ka/Ks) were calculated for homologous and homoeologous gene pairs among Gossypium species. Generally, Ka/Ks values of homologous and homoeologous BASS gene pairs between genomes and/or subgenomes were less than 1, in spite of the fact that the average Ka/ Ks values of homologous gene pairs between the genome of G. arboreum and Dt-subgenome of G. hirsutum was 1.004, as well as Dt-and Dt-subgenome of G. hirsutum was 1.09. The total number of homologous and homoeologous gene pairs in the present study were 115, and the average Ka/Ks value of all 115 gene pairs was 0.88, which was less than 1 (Fig. S3). These data indicated that purifying selection was the main selective force operating on Gossypium BASS genes across species evolution.
Cloning, subcellular localization and expression profiles of GhBASS genes. Based on the bioinformatics analyses, sequences of GhBASS1 (Gene ID: Gh_D01G0342), GhBASS2 (Gene ID: Gh_D06G0564), GhBASS3 (Gene ID: Gh_D07G1990), GhBASS4 (Gene ID: Gh_D12G0875) and GhBASS5 (Gene ID: Gh_ D12G2214) were retrieved for designing gene-specific primers (Table S3) in order to clone GhBASS genes from cotton cv. Zhong G5. We ultimately obtained 5 GhBASS cDNA sequences with the complete ORFs, and the cloned GhBASSs exhibited 99.5 to 99.89% identities with their corresponding query sequences derived from bioinformatics studies (Fig. S4). All of them maintained the conserved domains (SBF and bass domain), ten transmembrane regions and two Na + -binding sites, which are the characteristics of BASS family members.
According to the data from The Arabidopsis Information Resource (TAIR), Arabidopsis BASSs are predicted to localize in the chloroplast envelope. Consistently, the chloroplast localization of GhBASSs was further predicted by web server predictors such as ChloroP (http:// www. cbs. dtu. dk/ servi ces/ Chlor oP/) and LOCALIZER (http:// local izer. csiro. au/). To prove this bioinformatics information, GhBASSs-GFP-transformed protoplasts and GFP-transformed protoplasts which served as a control were observed under a confocal laser scanning microscope. While observing GhBASSs-GFP-transfected protoplasts, GhBASSs-GFP signal was targeted only to the chloroplast membrane. On the other hand, as examining GFP-transfected protoplasts, GFP fluorescence was universally distributed throughout the cell, including in the nucleus (Fig. 5). This provided evidence that GhBASSs were localized in the chloroplast envelope.
In order to examine the potential functions of GhBASS genes, their transcriptional levels in different tissues (root, stem and leaf) were monitored by qPCR with gene-specific primers (Table S3). Of these, the transcripts of GhBASS2 and GhBASS3 were expressed preferentially in leaf tissues; however, GhBASS1 and GhBASS5 showed higher expression levels in roots, and GhBASS4 exhibited ubiquitous expression with no specific pattern (Fig. 6). These results indicated that BASS family members showed diverse expression patterns in different tissues.
To determine whether BASS genes were induced by different stressors, we analyzed the expression patterns of GhBASSs during abiotic stresses, i.e., cold, heat, salt and drought. In response to NaCl treatment, GhBASS1, GhBASS2 and GhBASS3 were significantly induced; when GhBASS1 and GhBASS2 peaked at 24 h post-treatment (hpt), GhBASS3 reached the maximum point at 12 hpt. By contrast, the transcriptional levels of GhBASS4 and GhBASS5 were decreased in responding to salt except at a few time points (Fig. 7A). Generally speaking, GhBASSs exhibited up-regulation after the drought induction, of which GhBASS5 was greatly induced, while the other genes were moderately induced (Fig. 7B). When exposed to heat stress, GhBASS5 was clearly up-regulated, whereas the transcript level of GhBASS1 was increased only at 3, 12, 72 and 96 hpt. The expression of GhBASS2, GhBASS3 and GhBASS4 was increased as early as 0.2 hpt and then rapidly fell until 72 and 96 hpt during heat stress (Fig. 7C). Overall, the transcripts of GhBASS1, GhBASS2, GhBASS3 and GhBASS5 were elevated after the cold treatment; in contrast, GhBASS4 was expressed only at 0.2, 3 and 24 hpt (Figs. 7D, S6). These results suggested that the expression patterns of G. hirsutum BASS genes were widely varied in response to different abiotic stresses.
Next, we investigated the responses of GhBASS genes against the treatments of different phytohormones such as salicylic acid (SA), abscisic acid (ABA) and gibberellic acid (GA3), as well as methyl viologen (MV) which is the herbicide and experimental tool, and induces the generation of reactive oxygen species (ROS) in the chloroplast 28 . All G. hirsutum BASS genes were induced and accumulated at higher levels after MV and ABA treatments except at some time points. With the influence of these two inducers, the transcriptional abundance of GhBASS2 was the highest among the tested genes (Fig. 8A,B). Under SA conditions, transcript levels of GhBASS1 and GhBASS5 were significantly elevated, reaching their highest levels at 3 and 72 hpt, respectively, meanwhile those of GhBASS2, GhBASS3 and GhBASS4 were not clear with the fluctuated expression patterns (Fig. 8C). In response to the exogenous GA3 application, GhBASS3 and GhBASS5 were highly expressed; GhBASS3 reached its peak at 48 hpt, and GhBASS5 peaked at 72 hpt. GhBASS2 was dramatically expressed at 24 hpt, but its expression was diminished at other time points in relation to GA3. GhBASS1 and GhBASS4 showed oscillated expression levels reaching their peaks at 24 hpt when subjected to GA3 (Figs. 8D, S7). As a whole, this finding revealed that all candidate genes were induced by at least two of four inducers, implying that they might play important roles in signaling pathways.  29 . Agrobacterium-mediated VIGS has been successfully employed in a wide range of dicotyledonous and monocotyledonous species for various functional studies 30 . With the application of VIGS, a more detailed study on the functions of GhBASSs in relation to salt stress was performed. GhCLA1 is a homologous gene to DXS1/CLA1 in A. thaliana. The Arabidopsis Cloroplastos alterados 1 gene (AtCLA1) involved in chloroplast development 31 , and knocking-out AtCLA1 resulted in an albino phenotype in Arabidopsis 32 , which provides an excellent visual marker for silencing efficiency. Each of the TRV-based VIGS vectors was specifically constructed targeting to the non-conserved regions of GhBASS genes and outside the conserved domains to avoid interference with other BASS genes and other SBF proteins. Two weeks after agroinfiltration, TRV:GhCLA1 exhibited a highly uniformed photobleaching symptom in newly emerged leaves, meanwhile non-symptomatic new leaves were observed in TRV:GFP plants under the same conditions, indicating that silencing of GhCLA1 occurred specifically in TRV:GhCLA1 infected plants (Fig. 9A).  (Fig. 9B). In addition, the root performance of TRV:GhBASSs was quite similar to that of TRV:GFP under normal conditions (Fig. 9C). Although roots of TRV:GhBASS2, TRV:GhBASS4 and TRV:GhBASS5 plants were stronger and bigger than those of the controls, TRV:GhBASS1 and TRV:GhBASS3 plants possessed weakened roots after they had been exposed to salinity conditions (Fig. 9D), which matched up to the shoot performance of those plants. To confirm this result, we analyzed Na + and K + contents in all tested plants. Ions accumulation in the tested plants were enormously different between before and after the 200 mM NaCl application. Na + accumulation in GhBASS2, GhBASS4 and GhBASS5 knock-down plants was greatly lower, but in GhBASS1 and GhBASS3 knock-down plants was higher than those in the control plants after subjected to salt, in spite of the fact that their Na + contents were likely the same under non-salt exposure (Fig. 10A). By contrast, the scenario of K + accumulation was totally opposite to that of Na + accumulation during salt stress. Under normal conditions, K + content in all tested plant was not clearly different (Fig. 10B). Consequently, the K + /Na + ratio was significantly declined in GhBASS1 and GhBASS3 knock-down plants, but distinctly raised in GhBASS2, GhBASS4 and GhBASS5 knock-down plants compared with that in the controls during salt treatment, even though the difference in this ratio was not clear under the mock conditions (Fig. 10C). According to these results, it could be concluded that GhBASS1 and GhBASS3 positively regulated, whereas GhBASS2, GhBASS4 and GhBASS5 played a negative regulatory role in salt tolerance of their corresponding knock-down plants.
Overexpression of GhBASS2 and GhBASS5 impairs salt tolerance of transgenic Arabidopsis. Gene overexpression is a process which makes the abundant target protein expression substantially 33 .
By this means, phenotypes resulted from overexpressing target genes can be observed, which allows the func- www.nature.com/scientificreports/ tional characterization of those genes 34 . A. thaliana (Col-0) was used to generate transgenic plants overexpressing selected GhBASS genes in the present study, because it grows quickly, produces many seeds and is genetically well characterized 35 . Therefore, we continued to conduct the overexpression experiment for GhBASS2 and GhBASS5 which were selected based on our previous findings and information on other literatures. With the nested sequence-specific and shorter arbitrary degenerate primers, thermal asymmetric interlaced PCR (TAIL-PCR) is a highly efficient method to characterize the genomic DNA regions flanking the T-DNA insertion sites, which is applicable for the initial screening of independent transgenic lines [36][37][38] . Thus, transgenic T1 seedlings surviving on the selective medium supplemented with 50 μg ml −1 hygromycin were validated by TAIL-PCR. In the TAIL-PCR gel electrophoresis results (Fig. S10A,B), WT control showed faint non-specific bands, whereas negative control didn't reveal any band even in the tertiary TAIL-PCR products. Interestingly, only one clear and bright band was observed in each of the secondary TAIL-PCR products of OE lines, even thought there was no clear and bright band in the primary TAIL-PCR products. Consistently, clearer and brighter bands were found in the tertiary TAIL-PCR products, and the sizes of those bands were similar to the sizes of bands in the secondary TAIL-PCR products. Importantly, different molecular weight bands, which were clearly and brightly visible in the tertiary TAIL-PCR products, between different OE transgenic lines were observed, indicating different insertion sites of interest DNA sequences (GhBASS2 and GhBASS5 gene) into the genome of A. thaliana (Col-0) [36][37][38] . These gel results were corroborated by the flanking sequence tags by direct sequencing the tertiary TAIL-PCR products (Fig. S10C,D). Accordingly, these OE transgenic lines were independent with different T-DNA insertion loci in the Arabidopsis genome. After GhBASS2-OE and GhBASS5-OE plants had been validated as genuinely transgenic plants through the application of TAIL-PCR, qPCR and RT-PCR (Figs. 11D,E, S9-S10), the salt tolerance detection of transgenic plants was carried out. To determine the growth performance of transgenic seedlings, seeds of T3 homozygous lines (GhBASS2-OE1, GhBASS2-OE2, GhBASS5-OE1 and GhBASS5-OE2), loss-of-function mutants (atbass2 and atbass5) and wild-type (WT) were grown on 1/2 MS medium with or without 100 mM NaCl. During salt stress conditions, OE lines exhibited much worse phenotypes with yellowing www.nature.com/scientificreports/ leaves and weakened roots in comparison with mutants and WT, in spite of the fact that their shoot and root performances were similar under normal situations (Fig. 11A,C). For further clarification of salt sensitivity of soil-grown GhBASS2-OE and GhBASS5-OE lines, 3-week-old WT, mutants and transgenic plants were irrigated with 150 mM NaCl. After 7 days of the salt treatment, GhBASS2-OE and GhBASS5-OE plants showed more impaired phenotypes with chlorotic and wilting leaves when compared with the control plants. Nonetheless, they had much better phenotypes displaying deeper green leaves than those of the controls once exposed to the non-salt treatment (Fig. 11B).

Discussion
Genome-wide identification of the BASS gene family in cotton. Plants underwent a higher rate of gene duplication in comparison with other eukaryotes 39 , and gene and genome duplications are responsible for variability in gene numbers 40 ; therefore, the number of BASS genes was greatly varied in different species (Fig. 1). The peculiar distribution of the BASS gene family suggests a plausible scenario for its origin that BASS4 emerged in bacteria, later green algae obtained BASS4, BASS1 and BASS2, and then land plants (bryophytes and vascular plants) possessed all BASS members (Fig. 1). Hence, all recent BASS genes were derived from an ancestral BASS4 gene which evolved before the divergence from the common ancestor of prokaryotes and eukaryotes, indicating its essential role in living organisms. In our study, we identified 8, 11, 16 and 18 BASS genes from G. arboreum, G. raimondii, G. hirsutum and G. barbadense, respectively (Tables 1, S1, S2). Gene duplication contributing to the amplification of gene numbers is the key mechanism behind the expansion of a gene family 41 . Likewise, cotton BASSs experienced gene duplication events. Consequently, each of BASS1, BASS2 and BASS3 was duplicated at least two copies in diploids (G. arboreum and G. raimondii) and four copies in tetraploids (G. hirsutum and G. barbadense). Unlikely, BASS4 was resistant to expand its number or it tended to be restored to the 'singleton' state during polyploidization followed by diploidization events; thereby, the number of BASS4 is only one in diploids and two in tetraploids. Duplication pattern of BASS5 was varied among Gossypium spp. that there were three paralogues (GrBASS5a, GrBASS5b and GrBASS5c) in G. raimondii, two homeologs (GhBASS5A and GhBASS5D) in G. hirsutum and three homeologs (GbBASS5A, GbBASS5Da and GbBASS5Db) in G. barbadense, while only GaBASS5 was anchored in G. arboretum (Tables 1, S1). Collectively, some BASS members have expanded their numbers, whereas other www.nature.com/scientificreports/ members are expansion-resistant, hinting the selective duplication of BASS genes in cotton 42 . Additionally, since only one paralogous gene pair (GrBASS3-2a/GrBASS3-2b) derived from tandem duplication, WGD as well as segmental duplication conferred the expansion of the BASS gene family in cotton. Of the 16 GhBASSs, 8 homeologs were from the At-subgenome and another 8 homeologs were from the Dt-subgenome; hence, every gene had its corresponding ortholog in G. arboretum and G. raimondii. Likewise, GbBASSs individually possessed their orthologous counterpart genes of G. arboretum and G. raimondii. Nonetheless, neither corresponding orthologs of 3 GrBASSs (GrBASS3-2b, GrBASS5b and GrBASS5c) in the Dtsubgenome of G. hirsutum nor the corresponding ortholog of GrBASS5c in the Dt-subgenome of G. barbadense was discovered (Tables 1, S1), indicating that these genes were presumably lost in the cultivated tetraploid cotton during or after polyploidization events 42 .
For further details of gene losses in cotton, the study on evolutionary relationships of BASS family members from Gossypium and other eudicot species was demanded to explore how the BASS gene family has evolved in cotton. We, therefore, conducted the conserved synteny analysis which can help discover the origin of genes as neighborhood genes were likely persistent, even though target genes were lost during evolution 43 . V. vinifera not only has not undergone recent genome duplications but also is a palaeo-hexaploid common ancestor for many dicotyledonous plants including cotton 44 . T. cacao experienced fewer polyploidization events compared with cotton 45 , and it was very closely related to cotton (Fig. 2). For these reasons, V. vinifera, T. cacao and Gossypium spp. were used to trace the history of the BASS gene family in cotton. In a genome-wide comparison of BASS members among V. vinifera, T. cacao and G. raimondii, the same number of BASS1 should have been maintained in the genome of the last common ancestor of those species, meanwhile ancestral genes of BASS2, BASS3 and BASS5 were expanded two to three times only in G. raimondii after speciation from the last common ancestor (Fig. 4A,B). Interestingly, the ancestral gene of BASS4 was duplicated only in V. vinifera after speciation from the last common ancestor or perhaps the counterpart ortholog of this gene was lost in T. cacao and G. raimondii after speciation from the last common ancestor (There is a difficulty for this conclusion because BASS4 could not be mapped on syntenic blocks). Moreover, there was a case where BASS members were not mapped onto syntenic blocks, especially in G. raimondii vs. V. vinifera (Fig. 4A,B), suggesting that after speciation from the www.nature.com/scientificreports/ last common ancestor, each of those genomes evolved through the extensive genome fractionation accompanied by multiple rounds of chromosome breakages, rearrangements and fusions, and it was followed by the selective loss of genes, which makes a severe ambiguity in the identification of chromosomal syntenies 46 .
Regarding the synteny analysis between the G. raimondii and G. arboretum genome, ancestral genes of BASS1 and BASS2 were similarly extended to make two copies of those genes in both species after speciation from the last common ancestor, while ancestral genes of BASS3 and BASS5 were expanded two times in G. arboretum, but three times in G. raimondii after speciation from the last common ancestor (Fig. 4C). Contrary to other members, BASS4 was not found in the syntenic region, which is probably due to the imperfect genome assembly and annotation of G. arboretum because GaBASS4 anchored in scaffold2621 (Fig. 4C). According to the synteny analysis between diploids and allotetraploids, some BASS family members showed the conservative evolution, e.g., BASS1 and BASS2 orthologs could maintain the conservative evolution, which was corroborated by the preserved collinearity of those orthologous gene pairs between diploids and allotetraploids (Fig. 4D,E). On the other hand, some members of BASS3 and BASS5 were not found in allotetraploids, even though they shared syntenic blocks (Fig. 4D,E). The A and D diploid ancestral species diverged from a common ancestor 5-10 million years ago 47 . And approximately 1.5 million years ago, these A-and D-genome progenitor were reunited by hybridization, as well as polyploidization events occurred 47 , which led to the evolution of the allotetraploid cotton. Thus, G. hirsutum and G. barbadense had undergone whole-genome duplications followed by the gene degeneration and loss, which causes a loss of some BASS family members in allotetraploids. For example, corresponding orthologs of GrBASS3-2b, GrBASS5b and GrBASS5c in the Dt-subgenome of G. hirsutum, as well as the orthologous counterpart gene of GrBASS5c in the Dt-subgenome of G. barbadense was lost. Unlikely, the www.nature.com/scientificreports/ conserved synteny of BASS4 was observed between the D-genome (G. raimondii) and Dt-subgenomes of allotetraploids, in spite of the fact that GaBASS4 was not localized in the chromosome (Fig. 4D,E), which indicates that the G. arboretum genome database (BGI v1.0 assembly genome) has a flaw in relation to the BASS gene family, and if GaBASS4 showed the shared synteny, it would really be helpful to trace the evolutionary history of BASS family members in cotton. Studying the patterns of exon/intron organization can provide an additional insight into the structural architecture of genes 48 . In our results, not any big difference in patterns and structures of exon/intron was observed within a subfamily, but the distinct difference in their structures was found among subfamilies (Fig. 3B). Such a resembled gene structure in a subfamily suggests that Gossypium spp. underwent gene duplications and produced many copies of each subfamily member during their long evolutionary processes. Moreover, exon gains or exon losses occurred during their evolution. For instance, GaBASS1-2 lost two exons, but GaBASS5 gained one additional exon after speciation from the last common ancestor of G. arboreum and G. raimondii. With regard to the evolution from diploids to allotetraploids, BASSs of allotetraploids experienced both exon gains and losses (Fig. 3B). A motif is the most basic unit of a protein, and gene structure analyses based on motif patterns are increasingly demanded 25 . In our study, cotton BASSs with similar motif distribution patterns were closely clustered and made up a subfamily (Fig. 3A,C), hinting the high degree of sequence similarity within each subfamily, which reflects that BASS orthologs in each subfamily have similar functions and modes of action. Interestingly, each subfamily possessed unique motifs that probably have specific functions, that is, perhaps, the reason for a broad range of substrate specificity. Besides, Motif 1 and motif 2 on which two Na + -binding sites were situated (Fig. 3C) were distributed in almost all of the BASS members, indicating that cotton BASS genes are persistent to carry the two conserved Na + -binding sites.
Functional characterization of GhBASSs in response to salt stress. Since the gene expression pattern is important information for the prediction of gene functions 49 , we characterized expression patterns of GhBASSs not only in different tissues but also under abiotic stresses and hormone elicitors. Except for GhBASS4 which was ubiquitously expressed, other GhBASS genes exhibited the tissue-specific expression (Fig. 6). For www.nature.com/scientificreports/ example, GhBASS1 and GhBASS5 were highly expressed in roots (Fig. 6), hinting that these genes preferentially function in roots, i.e., functioning in the ion and mineral absorption and transportation from roots. By contrast, the transcripts of GhBASS2 and GhBASS3 were obviously high in leaf tissues (Fig. 6), which reflects that these genes participate in the functions of leaves. Abiotic stresses such as salt, drought and extreme temperatures seriously affect cotton growth and yield 15 , so our studies focused on molecular mechanisms concerning these stresses in cotton. In the present study, all GhBASS genes showed up-or down-regulation responding to different abiotic stresses (Fig. 7). This result can be speculated that different GhBASS genes play different functional roles in response to various abiotic stresses. ROS are important signal transduction molecules, as well as toxic oxidants which are by-products of stress metabolisms 50 . The ABA signaling involves in salinity and drought resistance of Arabidopsis 51 . In addition to having important roles in plant growth and development processes, GAs have regulatory functions in the plant cellular network of stress signaling 52 . SA is a phytohormone and improves the plant resistance to biotic and abiotic stresses 53 . Here, the elevated expression of GhBASS genes was imposed by various phytohormones, especially by ABA and MV (Fig. 8). This finding can be explained that GhBASS genes may involve in the signaling of many hormones, especially in ABA and ROS signal transduction pathways. Taken together, GhBASS genes might play a significant role in hormone signaling pathways which contribute to an interactive network that coordinates responses to different abiotic stresses. However, further studies are still needed to confirm the conclusions mentioned here. Sodium is the main culprit of soil salinity, which is a major deleterious environmental factor limiting the global cotton production, and approximately 20% of the world's agricultural land is presumably contaminated by Na + ions 54 . The high level of soil salinity imposes osmotic stress and ion toxicity, which results in an imbalance in cellular homeostasis and functional disturbance of Ca 2+ and K +3 . Virus-induced gene silencing (VIGS) is a particularly useful tool for rapid and large-scale gene analyses in functional genomics and widely used to investigate the functions of genes 29 . BASS family members of different species retained the conserved Na + binding sites 8,55 , which was consistent with our bioinformatics results (Figs. 3, S2). Besides, the transcriptional expression of GhBASSs (Fig. 7A) indicated that they tend to involve in salt stress. Hence, GhBASS genes were silenced and subjected to salt stress to uncover their functional roles concerning plant salt tolerance. In our results, GhBASS1 and GhBASS3 positively regulated, whereas GhBASS2, GhBASS4 and GhBASS5 negatively regulated the response of their corresponding knock-down plants to salt stress (Fig. 9). Na + content is direct evidence for measuring the degree of salt stress in plants. Here, the silencing of GhBASS1 and GhBASS3 resulted in an increased Na + accumulation, while the silencing of GhBASS2, GhBASS4 and GhBASS5 reduced Na + content in comparison with the control (Fig. 10A), which matched up to the phenotypes of their corresponding knock-down plants (Fig. 9). In plants, Na + is taken up at the root level, loaded into the xylem and then accumulated in the shoots, and Na + transporters are key players in all of these steps 3 . BASS family genes are sodium-dependent metabolite co-transporters 7,8 . For example, AtBASS2 was shown to mediate the co-transport of sodium and pyruvic acid into the chloroplast 9 , as well as AtBASS5 was reported as a co-transporter of Na + and 2-keto acids into plastids 11 . Next, AtBASS1 localized in the plastid inner envelope and exported pantoate from the plastid 56 . However, the substrates of BASS3 and BASS4 are still unknown. Taken together, it is conceivable that GhBASSs are capable of transporting Na + , no matter what type of substrates they co-transport.
Since the subcellular localization of a protein is directly correlated to its function, the subcellular localization of GhBASS proteins will have important effects on their roles in relation to Na + transportation 57 . To the best of our knowledge, all existing studies of plant BASS genes reported that they consistently located in chloroplast envelopes 9,10,12,56 . For instance, Arabidopsis BASS1 protein was solely localized in the plastid inner envelope 56 , TaBASS2 located in the chloroplast membrane 13 and AtBASS5 specifically existed in the plastid envelope 10 . Consistently, GhBASS proteins were predominately localized in the chloroplast envelope (Fig. 5). Although the chloroplast contains its own genome, many of its genes are encoded in the nuclear genome. Most nuclear-encoded chloroplast proteins carry transit peptides (cTPs) at their N-terminus that guide them to the chloroplast 58 . Web server predictors such as ChloroP (A neural network-based method of predicting transit peptides) 58 and LOCALIZER (A method for predicting plant and effector protein targeting to chloroplasts, mitochondria or nuclei) 59 were developed for the prediction of a protein localization. By means of ChloroP (http:// www. cbs. dtu. dk/ servi ces/ Chlor oP/) and LOCALIZER (http:// local izer. csiro. au/), the chloroplast localization of GhBASSs were predicted because they held transit peptides (cTPs) (Fig. S5), which makes the experimentally subcellular localization result sound that GhBASSs are localized in chloroplast envelopes. Taken the sodium transportability and plastid-localized character of GhBASSs into consideration, it is speculated that GhBASS2, GhBASS4 and GhBASS5 import Na + into chloroplasts, which is agreed by previous reports in which BASS2 and BASS5 genes co-transported Na + and their respective substrates into chloroplasts [9][10][11]13 , while GhBASS1 and GhBASS3 export Na + from chloroplasts, which is consistently found by Huang et al. 56 that AtBASS1 exported pantoate from the plastid by the Na + dependent manner. This notion is further approved by the measurement of cytosolic and chloroplastic Na + levels using the fluorescent dye SBFI-AM, as well as by the detection of net Na + fluxes using Non-invasive Micro-test Technology (NMT) (Data not showed). Following this concept, when the expression of GhBASS1 and GhBASS3 is suppressed, their functions are weakened, probably resulting in more Na + ions inside the chloroplast. By contrast, if GhBASS2, GhBASS4 and GhBASS5 are silenced, levels of Na + ions inside the chloroplast will be feasibly decreased. Chloroplast serves as not only a factory for energy assimilation but also a generator for the synthesis of phytohormones, e.g., ABA and SA 13,60 and important metabolites such as soluble sugar and proline 61 . In addition, chloroplast is one of the major sites for ROS production, and latter species of ROS can be produced in the plastid of root tissues also. The enhanced ROS generation in the chloroplast by salt stress changes the redox state of cells, which is instrumental in regulating the chloroplast metabolic activities 62,63 . Importantly, chloroplasts can be damaged by an excess of Na + accumulated in chloroplasts 64 . Hence, chloroplast is a very important organelle in plant cells, and it is responsible for responding to salt stress 61 . Thereby, silencing of these genes affected the levels of Na + accumulation inside chloroplasts, which in turn interfered with the www.nature.com/scientificreports/ functions of chloroplasts in their respective knock-down plants, and consequently, these knock-down plants suffered from salt stress with the varying degree of salt severity (Figs. 9, 10). From our previous studies observed by Guo et al. 65 and Myo et al. 66 , as well as literatures reported by Furumoto et al. 9 , Gigolashvili et al. 10 , Sawada et al. 11 and Zhao et al. 13 , GhBASS5 and GhBASS2 are significantly promising candidate genes. Based on this information, we continued our study on these two genes employing the overexpression experiment. In our results, both of GhBASS5 and GhBASS2 impaired salt tolerance of transgenic Arabidopsis (Fig. 11), which was consistent with the increased Na + accumulation in root tissues of OE plants (Fig. 12A). These results confirmed again that GhBASS5 and GhBASS2 are Na + transporters, as well as negatively regulate the plant salt tolerance, which is agreed with the gene silencing results (See the above paragraphs). With regard to the sodium-coupled substrate transported by BASS5, Sawada et al. 11 presented that AtBASS5 cotransported Na + :2-keto acids into the chloroplast for the synthesis of Glucosinolates (GSLs). Moreover, Hossain et al. 67 stated that GLS degradation products stimulated the generation of ROS which can cause oxidative stress if they are at the excessively accumulated level. Taken GhBASS5's co-transporter role into consideration, if the GhBASS5 gene is overexpressed, high levels of Na + and 2-keto acids will be expectedly accumulated inside the chloroplast, which will harm the functions of the chloroplast. By this means, disturbing chloroplast functions resulted in an impaired tolerance of the plant to salt stress (Figs. 11, 12). If GhBASS5 serves as a co-transporter of Na + and 2-keto acids, a question of what biosynthesis pathway occurs in cotton will be interested in this context because GSL compounds are restricted to the Brassicaceae 11 . Pyruvic acid is the precursor of abscisic acid (ABA), and ABA is generated inside the chloroplast 13,61 . ABA acts as an endogenous messenger functioning in the regulating and signaling network of salt response 68 . Regarding the BASS2's symporter model of Na + :pyruvic acids, Furumoto et al. 9 reported that AtBASS2 imported pyruvate into the plastid via the Na + dependent manner. From the view of the sodium-coupled pyruvate transporter role of GhBASS2, it co-transports a couple of pyruvic acids and Na + ions into the chloroplast. Accordingly, pyruvic acids and Na + ions should be highly accumulated inside the chloroplasts of GhBASS2-OE plants because the numbers of GhBASS2's transcripts are multiplied in those overexpressing experiments. As a result, such an excess of Na + accumulated inside the chloroplast damages the functions of the chloroplast, rendering a weakened tolerance of the plant against salt stress (Figs. 11,12). However, there is a cross-talk that TaBASS2 served as a pyruvate transporter into the chloroplast, which resulted in salt tolerance of wheat 13 . In such a case, more ABA will be possibly synthesized by the chloroplast, and there will be similarly more Na + ions inside the chloroplast, but they will be under the detrimental level. Logically, the accumulated levels of pyruvate and Na + inside the chloroplast should have the threshold level. In accordance with this notion, if they are above this level, the excessive accumulation of Na + will disturb the functions of chloroplasts, even though ABA can enhance the plant tolerance to salt stress. Accordingly, it is comprehensible that GhBASS2 might play a dual role, and whether having the role of positive or negative regulation in plant salt tolerance may depend on the expression levels of this gene. Although this report cannot explore exactly where Na + was accumulated, i.e., in the cytoplasm or plastid, how a plastidial sodium-dependent transporter regulated other genes participated in the pathway network responding to salt stress and what detailed metabolic mechanisms occurred concerning salt stress for each gene, this paper is the second report in relation to a plastidlocalized sodium-coupled metabolite transporter that can influence Na + transportation and plant salt tolerance, which is fundamental information for further detailed analyses of salt-tolerant mechanisms.
In conclusion, BASS family genes are very conserved across all tested organisms, and GhBASSs are sodiumdependent metabolite co-transporters. To the best of our knowledge, this paper is the second report to validate that a plastidial Na + -coupled metabolite transporter carries out Na + transporting function in plants and has a powerful impact on plant salt tolerance. We believe that our findings will provide programs for the production of salt-tolerant cotton varieties with fundamental information, even though precise metabolic mechanisms of GhBASS genes concerning salt stress require further investigation.

Materials and methods
All the experiments have been done in accordance with relevant institutional, national, and international guidelines and legislation.

RNA extraction, cDNA synthesis and qPCR analyses. Total RNA was extracted using the Plant
Total RNA Isolation Kit Plus (Foregene, Chengdu, China) with the manufacturer's guidelines. The TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix kit (TransGen Biotech, Beijing, China) was used for the first-strand cDNA synthesis following the supplier's protocols. The relative quantification of GhBASSs by qPCR was conducted with SYBR Green on a LightCycler 480 (Roche Molecular Systems, Inc., Indianapolis, Indiana, USA). Following the instructions of the TransStart Tip Green qPCR SuperMix kit (TransGen Biotech, Beijing, China), the reaction mixture and PCR cycles were performed. The cotton UBQ 7 (GenBank accession no. DQ116441) and Arabidopsis UBQ 10 (GenBank accession no. AT4G05320) were applied as internal references for normalization. The transcript levels of GhBASSs were calculated by the comparative 2 −ΔΔCT method 81 , and three independent biological replicates were done for each assay. Primers used for qPCR analyses were listed in Table S3.
Cloning and subcellular localization of GhBASSs-GFP fusion proteins. All GhBASS genes were amplified from cotton with specific primers (Table S3) designed from the identified deduced sequences resulting from bioinformatics analyses. The PCR amplified products were independently cloned into the pMD19-T Vector (Takara Biomedical Technology Co., Ltd., Beijing, China) following the manufacturer's instructions. www.nature.com/scientificreports/ full-length cDNA sequences of GhBASSs were examined by performing the alignment analysis with their corresponding query sequences derived from bioinformatics studies. Next, GhBASS genes were separately inserted into the pCaMV35S:GFP vector with the use of specific primers (Table S3) to generate pCaMV35S:GhBASSs:GFP recombinant plasmids. Then, not only those fusion proteins (pCaMV35S:GhBASSs:GFP) but also the empty vector control (pCaMV35S:GFP) were transiently expressed in A. thaliana mesophyll protoplasts using the PEG-mediated transformation method 27 . After incubating the PEG transfected mesophyll protoplasts in the dark for 16 h, GFP fluorescence was observed under a confocal laser scanning microscope (Carl Zeiss Co., Ltd., Jena, Germany).
Agrobacterium-mediated virus-induced gene silencing and salt tolerance analysis. The construction of tobacco rattle virus (TRV)-based VIGS vectors and inoculation of Agrobacterium tumefaciens strain GV3101 harboring recombinant TRV vectors were performed as described previously 30 . Briefly, cDNA fragments of GhCLA1 (420 bp) and GhBASSs (250-450 bp) were amplified by PCR with specific primers (Table S3). The resulting PCR products were separately cloned into pTRV2 to produce recombinant vectors, pTRV:GhCLA1 and pTRV:GhBASSs. These recombinant vectors were independently introduced into A. tumefaciens strain GV3101, and then cultured at 28 °C for 24 h. Afterword, Agrobacterium carrying pTRV1 was independently mixed with Agrobacterium having recombinant vector pTRV:GhCLA1 or pTRV:GhBASSs in a 1:1 ratio just before inoculated into cotyledons of cotton seedlings. TRV:GFP was included as a negative control, whereas TRV:GhCLA1 served as a positive control. The agroinfiltrated plants were maintained in a greenhouse (23 °C, 120 μmol photons m -2 s -1 light, 12 h light/12 h dark cycle). After the photobleaching phenotype of TRV:GhCLA1 had been observed, gene silencing efficiency was verified by examining the expression level of endogenous BASS genes using RT-PCR and qPCR. The silenced plants were then subjected to 200 mM NaCl treatment for the salttolerant test. Regarding the plant materials in VIGS experiments, 16 plants for one gene for one replication were applied to a completely randomized design (CRD) with two replications. Accordingly, the results were validated by three successive times.
Transgenic plant generation and salt tolerance analysis. To construct pCAMBIA1302:GhBASS2 and pCAMBIA1302:GhBASS5 recombinant vectors, each of GhBASS2 and GhBASS5 was separately introduced into the pCAMBIA1302 vector under the control of a cauliflower mosaic virus 35S promoter. Agrobacterium tumefaciens strain GV3101 harboring pCAMBIA1302:GhBASS2 or pCAMBIA1302:GhBASS5 was separately transformed into A. thaliana (Col-0) by the floral dip method with minor modifications 82 . Transgenic T1 seedlings were screened by growing on the 1/2 MS medium supplemented with 50 μg ml −1 hygromycin, and surviving seedlings on the selective medium were validated by PCR. To confirm whether these selected T1 lines were independent with different T-DNA insertion loci in the Arabidopsis genome, they were then screened by thermal asymmetric interlaced PCR (TAIL-PCR) following previously described methods [36][37][38] with some modifications. Of these independent transgenic lines, two T3 homozygous lines obtained from T2 lines showing the correct segregation ratio (3:1) were confirmed by qPCR, as well as RT-PCR and then the correct ones were selected for further functional studies. Seedlings and adults of WT, loss-of-function mutants and T3 homozygous transgenic lines were exposed to salt stress for the observation of salt-stressed symptoms and salt-tolerant phenomenon. Primers used in this study were included in Table S3. Regarding the plant materials in these experiments, 30 seedlings/plants for one overexpressing line for one replication were applied to a completely randomized design (CRD) with two replications. Accordingly, the results were validated by three successive times.
Measurement of Na + and K + contents. Salt-treated and none salt-treated plants were sampled for the measurement of their ions concentration as described by Rus et al. 83 with some modifications. In brief, harvested samples were dried inside a drying oven at 70 °C for 4 days. The oven-dried samples were digested with 0.1 M HNO 3 and boiled at 90 °C for an hour, then filtered through the Whatman no. 1 filter paper. Na + and K + content in the filtrated solutions were determined by an atomic absorption spectrophotometer (Triad Scientific, Inc., Manasquan, New Jersey, USA).

Statistical analysis.
To perform statistical analysis, IBM SPSS version 16.0 (SPSS Inc., Armonk, New York, USA) and GraphPad Prism version 8.0 (GraphPad Software Inc., San Diego, CA, USA) were used. Data come from three independent biological replicates were analyzed by applying independent sample t-test or one-way ANOVA based on Duncan's multiple range test. The significant difference level was signified with a single star (*, P = 0.05).

Data availability statement
The data in relation to this study are obtainable from corresponding authors on reasonable request.

Permission statement for plant materials
All required approvals for collection of plant or seed specimens were obtained for the study, which complied with relevant institutional, national, and international guidelines and legislation.