Amino acid transporter (AAT) gene family in foxtail millet (Setaria italica L.): widespread family expansion, functional differentiation, roles in quality formation and response to abiotic stresses

Background Amino acid transporters (AATs) plays an essential roles in growth and development of plants, including amino acids long-range transport, seed germination, quality formation, responsiveness to pathogenic bacteria and abiotic stress by modulating the transmembrane transfer of amino acids. In this study, we performed a genome-wide screening to analyze the AAT genes in foxtail millet (Setaria italica L.), especially those associated with quality formation and abiotic stresses response. Results A total number of 94 AAT genes were identified and divided into 12 subfamilies by their sequence characteristics and phylogenetic relationship. A large number (58/94, 62%) of AAT genes in foxtail millet were expanded via gene duplication, involving 13 tandem and 12 segmental duplication events. Tandemly duplicated genes had a significant impact on their functional differentiation via sequence variation, structural variation and expression variation. Further comparison in multiple species showed that in addition to paralogous genes, the expression variations of the orthologous AAT genes also contributed to their functional differentiation. The transcriptomic comparison of two millet cultivars verified the direct contribution of the AAT genes such as SiAAP1, SiAAP8, and SiAUX2 in the formation of grain quality. In addition, the qRT-PCR analysis suggested that several AAT genes continuously responded to diverse abiotic stresses, such as SiATLb1, SiANT1. Finally, combined with the previous studies and analysis on sequence characteristics and expression patterns of AAT genes, the possible functions of the foxtail millet AAT genes were predicted. Conclusion This study for the first time reported the evolutionary features, functional differentiation, roles in the quality formation and response to abiotic stresses of foxtail millet AAT gene family, thus providing a framework for further functional analysis of SiAAT genes, and also contributing to the applications of AAT genes in improving the quality and resistance to abiotic stresses of foxtail millet, and other cereal crops. Supplementary Information The online version contains supplementary material available at 10.1186/s12864-021-07779-9.


Background
Amino acids, as small-molecule organic compounds containing nitrogen, are the main carriers of nitrogen exchange in plants and the nutrients necessary for plant growth, development and quality [1]. The absorption and transport of amino acids are mainly mediated by amino acid transporters (AATs) [2]. Through the transmembrane transport of amino acids, amino acid transporters play essential roles in plant growth and development, including seed germination [3,4], long-distance transport of amino acids [5], quality formation [6,7], and response to pathogenic bacteria [8] and abiotic stress [9]. A total of 72, 65, 189, 85 and 297 amino acid transporters have been characterized in potato [10], Arabidopsis [1], soybean [11], rice [12] and wheat [13], respectively, suggesting that amino acid transporters are widely present in higher plants.
Based on the heterologous expression system, the functions of AAT genes in Arabidopsis have been characterized in detail [17]. There are 8 AAP genes (AtAAP1-AtAAP8), and 6 AtAAP genes transport charged and neutral amino acids with various a nities [17,18]. In endosperm, roots and cotyledon, AtAAP1 is abundantly expressed, which is essential for the synthesis of storage proteins in Arabidopsis seed and seed yield, while regulating embryonic development and amino acid input in roots [14,19]. AtAAP2 and AtAAP3 are mainly involved in the xylem-phloem transfer of amino acids and the absorption of amino acids in xylem, respectively [20,21]. AtAAP4 and AtAAP5 can transfer amino acids to developing embryos and AtAAP5 plays crucial role in the root amino acid intake system [22]. AtAAP6 affects the interaction between plants and aphids through the regulation of the phloem's amino acid composition [8]. AtAAP8 is mainly responsible for transporting amino acids into the endosperm and embryo at the early embryonic development [23]. Additionally, the functions of some AAPs in other species have also been explored. In Vicia faba, VfAAP1 is expressed abundantly in cotyledons and other pool tissues, and signi cantly affects the storage protein content (GPC) in seeds by regulating the nitrogen content of seeds [4,6], while VfAAP3 is signi cantly enriched in roots, stems, pods, pistils and seed coats at various stages of development [4]. In potato, StAAP1 is expressed during the source-sink transition process, and the free amino acid contents are decreased in transgenic plants with reduced StAAP1 expression [24], and the expression patterns of StAAP4, 5, and 6 may be associated with the long-distance transport of amino acids [10]. In barley, HvAAP2 and 6 are mainly expressed in nutritive organs, while HvAAP3 is high-expressed in seeds and contributes signi cantly to nitrogen accommodate in developing seed [25]. In rice, OsAAP1 affects plant growth and grain yield by modulating the redistribution of neutral amino acids [26], and the quantitative trait locus qPC1 (QTL) modulation of the rice GPC is linked to the expression of OsAAP6 [7]. In addition, several wheat TaAAP genes have been con rmed to co-localize with QTL of grain protein content [27].
In addition to the in-depth studies on AAP genes, the functions of some genes in other subfamilies have also been explored [5,8,9]. AtLHT1 plays an indispensable role in the uptake of amino acids and mesophyll formation in root, while AtLHT2 mainly involves in the movements and allocation of amino acids in oral organs [28][29][30]. AtANT1 is expressed abundantly in owers and stem leaves, transporting neutral and aromatic amino acids as well as indole-3acetic acid (IAA) and arginine [31]. AtAUX1 and AtLAX3 affect the root gravitropism and lateral root formation by facilitating IAA sink-source distribution and auxin-induced expression of root cell-wall-remodeling genes, respectively [32][33][34][35]. AtProT1 and AtProT2 participate in the long-distance transportation of proline and the absorption of proline and glycine betaine in roots of Arabidopsis, respectively [36,37]. AtProT3 is mainly expressed in leaf epidermal cells, which may be involved in the regulation of proline distribution in leaves [38]. AtCAT5 mediates the reabsorption of amino acids permeated at leaf margin by transporting basic amino acids, while AtCAT3, AtCAT6 and AtCAT8 prefer to transport acidic or neutral amino acids [17]. AtBAT1 has been identi ed as a bidirectional amino acid transporter expressed on mitochondrial membrane, which mediates γ-aminobutyric acid (GABA) from cytosol to mitochondria [39].
In addition to the role of AAT genes in seed germination, long-distance amino acid transport, root gravitropism, the formation of lateral roots and reproductive organs, a number of AAT genes play an important role in reducing damage to plants under various abiotic stresses, speci cally by promoting the transportation of stress-response compounds and compatible solutes such as proline, GABA, and betaine [40]. For example, the ProT genes rapidly distribute proline to minimize damage to plants caused by water shortage [41]. AtProT2 is strongly induced by salt stress, and plays a role in nitrogen distribution under salt stress [42]. GmProT1 and GmProT2 are sensitive to stress in the roots and transgenic plants can transport more proline after spraying proline [43]. AtGAT1 is signi cantly induced by GABA enrichment in plants under mechanical damage, stress or sensitive environment [44]. OsAAP6, OsAAP11 and OsANT3 are strongly up-regulated under salt and drought stresses, to regulate the tolerance to stress of plants through the transport of stress compounds [12].
Foxtail millet (Setaria italica L.), a member of the Poaceae family, is one of the oldest cereal crops, domesticated in Northern China. It is rich in essential amino acids, fatty acids and minerals, which is of important signi cance to human health [45]. To date, the AAT gene family in foxtail millet has not been characterized. Given their essential roles in the growth and grain quality formation of foxtail millet, we carried out a detailed characterization of AAT genes in foxtail millet. The aims of this study are : (a) Accurately identify and characterize AAT genes in foxtail millet based on the latest genome sequence; (b) Explore the roles of paralogous and orthologous AAT genes in the build-up and functional differentiation of AAT family in foxtail millet through comparative analysis of their gene structure, sequence feature and expression patterns; (c) Mine the key candidate AAT genes that affect grain quality formation of foxtail millet through the comparative transcriptome analysis of two genotypes; (d) Detect the response of SiAAT genes to diverse abiotic stresses by Quantitative Real-Time PCR; (e) Predict the possible functions of the SiAAT genes based on the previous studies, sequence composition and expression pattern. This work will provide a deeper insight and understanding of the AAT genes in foxtail millet. Also, important clues for their functional analysis and applications in improving quality and resistances to abiotic stresses will be provided.

Results
The foxtail millet genome contains 94 AAT genes, which all with transmembrane regions Firstly, 104 putative AAT transcripts were identi ed in foxtail millet based on local BLASP. However, 94 high-con dence non-redundant AAT genes were con rmed after screening the conservative domains by HMMER, CDD and Interpro databases and removing the different transcripts of the same gene and sequences containing incomplete conserved domains, which was basically at the same level as other gramineous species besides hexaploid wheat (Table 1; Additional le 5: Table S1). Except for tetraploid soybean, AAT gene family in foxtail millet was expanded than that in Arabidopsis and potato. Overall, the number of AAT genes in monocots was higher than that in dicots. The AAT genes identi ed in foxtail millet were renamed according to their chromosomal localizations and phylogenetic relationships with other species. The length of AAT proteins in foxtail millet ranged from 311 to 984aa, with molecular weight (Mw) varing from 4.97 to 107.5 kD, and isoelectric point (pI) ranged from 4.94 to 9.99 (Additional le 5: Table S1). Different subfamilies of AAT genes showed abundant diversity in subcellular localization, with the subfamily members of AAP, LHT, GAT, AUX and ProT all located on plasma membrane, while those of TTP, ACT, ANT and ATLb all located on vacuole membrane. Some members of the same subfamily showed different subcellular localization, such as, different genes of the ATLa subfamily located on both membranes of plasma and vacuole, while that of the CAT and PHS subfamilies located on the three membranes of the plasma, vacuole and chloroplast. a The numbers represent the number of identi ed AAT subfamily members, and the percentages in the brackets represent the proportion to all AAT genes.
The number of predicted TM regions in SiAATs varied in different subfamilies, which ranged from 7 to 14 ( Fig. 1a; Additional le 5: Table S1). For example, all AUXs contained 10 TM regions, ATLas contained 10 or 11 TM regions, while CAT and PHS subfamilies ranged from 9 to 14 and 7 to 12, respectively.
Duplication events occurred in 62% of SiAAT genes and promoted their expansion in foxtail millet Here, 93 of the 94 SiAATs were mapped to 9 chromosomes. The remaining one was mapped to an unassembled scafford (Additional le 1: Figure S1), which were unevenly distributed, with 18 on chromosome 7 and 2 on chromosome 2 as the most and least, respectively. The apparent regional enrichment of SiAATs on some chromosomes was observed on their distributions. For instance, a large number of SiAATs were located on the ends of chromosomes 1, 5, and 7. Moreover, some SiAATs were mainly concentrated at the fronts of chromosomes 6 and 8 (Additional le 1: Figure S1).
Of the 94 AAT genes in foxtail millet, 58 (62%) were involved in gene duplication events, including 36 tandem and 25 segmental duplications, of which SiATLb12, SiLHT2 and SiLHT9 were involved in both duplications (Additional le 1: Figure S1; Additional le 5: Table S1). The 36 tandem duplications were classi ed into 13 groups, of which 2 groups with 5 genes (TD10, TD11), 1 group with 4 genes (TD3), 2 groups with 3 genes (TD1, TD13), and the remaining 8 groups all with 2 genes (Additional le 5: Table S1). The 25 segmental duplications were classi ed into 12 groups, except for SD3 containing 3 genes, with other possessing 2 genes. Gene duplication events produced a large number of paralogous AAT genes, thus promoting the signi cant build-up of the AAT gene family in foxtail millet.
The phylogenetic tree and structure of AAT genes in foxtail millet The phylogenetic analysis with the AAT protein sequences from foxtail millet (94), rice (85), Arabidopsis (63) and potato (72) showed that they could be divided into twelve genetic groups with high con dence and that from Monocots and Eudicots distributed in the same group (Fig. 2). Based on their annotation information, similar exon/intron structures were observed in most of the SiAATs in the same subfamily, such as SiATLa1 and SiATLa5, SiCAT3 and SiCAT10, and SiATLb3 and SiATLb2. Variants were also found in some SiAATs of the same subfamily, such as SiAUX2 and SiAUX3, SiAAP11 and SiAAP18, and SiANT1 and SiANT2 (Fig. 3). The conserved motifs of SiAAT proteins predicted with MEME were highly consistent with the phylogenetic relationship and classi cation of SiAATs ( Fig. 3; Additional le 2: Figure S2). Similar to their gene structures, the presence of the conserved motifs in different subfamilies of SiAAT proteins varied. For instance¸Motif 1, 4, and 7 were widespread in the AAAP family, while Motif 9 existed in the APC family. Some motifs were unique to certain subfamilies, for example, AUX subfamily only possessed Motif 19, and Motif 16 and 18 were only found in the ACT subfamily.
Variations in gene structure, conserved domains and sequences of paralogous AAT genes promoted their functional differentiation The diversi cation of paralogous gene is one of the important sources of the functional differentiation of gene families [46][47][48]. The comparison on the TM regions of SiAATs found that signi cant variations in the number of TM regions occurred in the same duplicated gene groups, 60% of the duplicated gene groups (15/25) varied in the number of TM regions, which was 77% (10/13) and 42% (5/12) for the tandemly and for the segmentally duplicated gene groups, respectively ( Fig. 1b; Table 2). Table 2 The extensive variations in gene structure, protein structure and sequence of paralogous AAT genes produced by gene duplication in foxtail millet.
The data in columns 4, 5 and 6 represents the total number and its proportion (in the brackets) of mutated duplicated gene groups, while that in column 7 represents the range and average (in the brackets) of Ka/Ks values.
Multi-sequence alignment of SiAAP proteins revealed that their overall similarity was 59.65% with 11 conserved motifs, and their TM regions were highly correlated with the conserved motifs, both in length and amino acid composition (Fig. 4). Motif 1 and 14 both formed the TM 1 and TM 2 regions, TM 4 and part of TM 5 regions included Motif 4. TM regions of 3, 6, 7, 8, 9 and 10 were with Motifs of 8, 6, 2, 7, 5, 13, respectively (Fig. 4). In addition, some conserved motifs were located in the non TM region, such as Motif 11 in the extra-membrane region and Motif 3 in the intra-membrane region (Fig. 4). It was worth noting that some SiAAPs had missing transmembrane regions due to incomplete conserved motifs, such as SiAAP6 and SiAAP9. Therefore, the number variation of TM regions was mainly determined by the presence of different conserved domains. Prediction of the secondary structure of SiAAT proteins found 14 α-helixes and 4 η-helixes structures, with vast majority located in the TM regions to ensure the e cient and stable transmembrane transport of amino acids ( Fig. 4).
In terms of gene structure, 56% (14/25) of duplicated gene groups had variation in the number of introns, which was 69% (9/13) of the tandemly ones and 42% of the segmentally ones (Table 2; Additional le 3: Figure S3). In addition, the variations in the conserved motifs were also observed in 14 duplicated gene groups, 7 for each of the two types of duplication ( Table 2).
The Ka/Ks values of all paralogous SiAAT gene pairs were less than 1, ranged from 0.1 to 0.97, which suggested that these genes were subject to different levels of purifying selection (Fig. 5) 5). In general, signi cant variations in gene structure, conserved domains and sequences among the paralogous SiAAT gene groups might greatly promote the functional diversity of the AAT gene family in foxtail millet.
Variations in expression levels of paralogous and orthologous SiAAT genes together drove the functional differentiation of AAT family in foxtail millet The pattern of expression of all paralogous AAT genes were investigated using the public transcriptome data. Compared with the ancestral genes, according to their expression patterns, the newly duplicated genes could be classi ed into three main types of new-functionalization, sub-functionalization and nonfunctionalization, which were all observed in the tandemly duplicated gene groups with SiAAP8/9 (TD2) possessing new functions. SiAAP9 and SiAAP8 were highly expressed during grain development, and in leaves at the lling stage, respectively. SiAAP14/15/16/17 (TD3) and SiCAT1/2 (TD12) showed subfunctionalization, as the expressions of SiAAP16/17 and SiCAT2 were down-regulated signi cantly. In addition, SiATLb12 and SiLHT3 losing their functions, as they were not expressed in any tissues (Fig. 6a). Though no new function was observed, similar functional differentiation were also observed in the segmentally duplicated gene groups, such as sub-functionalization of SiATLb1/6/10 (SD3), SiBAT1/SiBAT8 (SD8) and SiCAT3/10 (SD9), and nonfunctionalization of SiATLa1/5 (SD1), SiLHT1/2 (SD6) and SiCAT4/11 (SD10) (Fig. 6b).
There were 62, 51, 52 and 35 orthologous AAT genes found in sorghum, rice, wheat and Arabidopsis, respectively, and their corresponding relationships with foxtail millet were investigated ( Fig. 7; Additional le 6: Table S2). The wheat genome was much larger than other species, so it was not displayed in Fig. 7.
The collinear relationship between different species was clearly distinguishable and showed good collinearity between chromosome 9 of foxtail millet and chromosome 1 of sorghum, chromosomes 3, 4, 5 of foxtail millet and chromosomes 8, 9, 10 of sorghum, respectively, etc. Moreover, the collinearity of the orthologous AAT genes among different species was consistent with that of the whole genomes. The number of orthologous subfamilies LHTs and LATs in millet and sorghum were signi cantly more than those in wheat and rice, which suggested that their expansion in millet and sorghum completed later than that in sorghum and wheat ( Fig. 8; Additional le 6: Table S2). Compared with Arabidopsis, the AAP subfamily in grass species was signi cantly expanded. For instance, the orthologous of SiAAP2, SiAAP12 and SiAAP14 only could be found in grass species, which suggested that they might be produced after the differentiation of monocotyledon and dicotyledon.
The expression characteristics of those orthologous AAT genes in root, stem, leaf, in orescence/spike and grain of foxtail millet, sorghum, wheat, rice and Arabidopsis were compared using their intergrated transcriptome datasets. Owing to the heterohexaploid nature of wheat with almost three homologous copies for each gene, the average expression values of these copies were used. The expression patterns of SiAATs in foxtail millet and their orthologous in other species were generally conservative, especially in the grass species (Fig. 8). The correlation coe cients on the expression patterns of AAT genes in gramineous species were 0.483, 0.470, and 0.481 between foxtail millet with sorghum, wheat and rice, respectively, while it was only 0.21 between foxtail millet and Arabidopsis. Several high-expressed AAT genes in foxtail millet were also highly conserved in other species, such as SiCAT3 and SiATLa5. In addition to the relatively conservative expression patterns of some important orthologous AAT genes among different species, the expression patterns of some genes were differentiated, such as SiATLa6 and SiATLb6 (Fig. 8). Compared with the orthologous genes in Arabidopsis, SiAATs showed greater variation in expression patterns, and several genes changed their tissue speci cities, such as SiATLb1 and SiAUX3.
SiAATs showed abundant spatiotemporal expression patterns, and some of them were speci cally expressed in developing grains The expression patterns of SiAATs were analyzed using the RNA-seq data of different tissues at mutiple growth stages collected from the public online database. The heat map was displayed with the normalized log 2 (FPKM + 1) values (Fig. 9a). According to the expression patterns, the SiAATs were clustered into three groups. The 8 SiAAT genes in the rst group included 7 AAAP family genes and one APC family gene with relatively high expression level, and were stably expressed in almost all tissues at different developmental stages. The 37 SiAAT genes in the second group were expressed in low abundance in most tissues, but expressed explicitly in some tissues, such as SiAAP9 and SiAAP7 were highly expressed in germinating seeds, while SiATLb3, SiAUX3 and SiAUX4 were relatively high expressed in panicle. The 49 SiAAT genes in the third group showed diverse spatiotemporal expression characteristics, such as SiATLa6 and SiATLa3 were highly expressed in the leaves at the seedling stage, but lower in the leaves at lling stage; SiAAP20 was speci cally high-expressed in the stem, while SiAAP3, SiBAT7, SiCAT11, SiBAT2, SiLAT7 and SiLAT8 were expressed abundantly in leaf tissues including leaf sheath and mesophyll. In addition, some genes were highly expressed in two or more tissues, such as SiAAP1, SiAAP13 and SiAUX1 in stem and root; SiANT1, SiCAT3 and SiATLb1 were expressed abundantly in organs involved in the entire source-sink circulation including root, stem, leaf and grain. The expressions of the 20 selected SiAATs by qRT-PCR analysis were not completely consistent with that from the transcriptome data, but their expression characteristics and trends were similar, which veri ed the reliability of the transcriptome data ( Fig. 9a; Additional le 4: Figure S4).
There were signi cant differences in multiple grain quality traits between the two foxtail millet genotypes "JG21" and "YG1". Except for crude fat content, the basic amino acid contents of glutenin, cysteine, alanine, methionine, leucine, tryptophan and serine in "JG21" grain was signi cantly lower than "YG1" (Fig. 11b). The developing grains (including glumes) of "JG21" and "YG1" at lling stage in 2019 were separated for RNA extraction and RNA-sEq. Finally, 348 differentially expressed genes (DEGs) in the developing grains of these two genotypes were identi ed, of which 164 and 184 DEGs were down-regulated and up-regulated in "JG21", respectively. GO analysis showed that DEGs were mainly enriched in multiple terms such as responses to fungus or viruses, ribonuclease activity, stress response, and transmembrane transport of various substances (Fig. 11a). Interestingly, 9 AAT genes were enriched into four terms related to amino acid transport, including 5 SiAAPs, 1 SiANT, 1 SiATLb, 1 SiAUX and 1 SiBAT (Fig. 11c). The differential expression of these SiAATs suggested that these genes might directly affect the formation of grain quality.

Some SiAATs actively responded to various abiotic stresses
The qRT-PCR results of JG21 and the transcriptome data of the foxtail millet cultivars of Yugu1 and Yugu2 showed a relatively consistent trend of a similar response of SiAAT genes to abiotic stress (Fig. 12a). The qRT-PCR analysis of 12 SiAATs in foxtail millet seedlings (15 days after sowing) to simulated drought (20% PEG 6000, 1 h and 5 h) and salt (200 mM NaCl, 1 h and 5 h) revealed that various subfamilies of SiAATs responded differently to abiotic stresses (Fig. 12b). Seven SiAAT genes were up-regulated more than 5 times after drought or salt stress, of which SiANT1 was up-regulated by about 60 times at 5 h after drought stress. SiANT1, SiCAT10 and SiATLa1 were mainly induced by drought stress, and their expression all reached the highest levels at 5 h after drought treatment. Except for SiAUX2 and SiATLb5, the other 7 SiAAT genes reached their highest expression levels at the late stage of salt stress (5 h). SiLHT12, SiATLb1 and SiLAT3 were speci cally and strongly induced by long-time salt stress, while SiANT1, SiATLa1 and SiATLb5 had similar expression patterns under both drought and salt stresses.
The response of different SiAATs to abiotic stress also showed temporal speci city. SiAAP3 was up-regulated under 1 h drought stress, while it was upregulated at 5 h under salt stress. SiLHT12 was down-regulated under 1 h drought and salt stresses, but then was strongly up-regulated at 5 h (Fig. 12b). SiANT1 was continually up-regulated with the increase of drought stress, and was down-regulated under 5 h salt stress. SiATLb5 and SiATLa1 responded at the early and late stages of drought and salt stress, respectively. Further, some SiAATs showed a sustainable response to abiotic stresses; for example, both SiATLb2 and SiATLa1 were continually up-regulated under both salt and drought stresses. These genes might enhance the adaptability of foxtail millet to abiotic stresses through their active responsiveness.

Discussion
A large number of tandem and segmental duplication events were the direct cause of AAT gene family expansion in foxtail millet Amino acid transporters (AATs) play indispensable roles in plant growth by transporting and distributing different types of amino acids, and therefore are considered as important targets for crop improvement [15]. Here, a total of 94 AAT genes were characterized in foxtail millet, which were divided into 12 subfamilies and was consistent with the previous reports in other plants (Table 1). Different subfamilies of AATs in monocotyledons and dicotyledons were corresponding, thus con rming that they were formed before the differentiation in monocotyledons and dicotyledons (Fig. 2). The number of SiAATs in foxtail millet was less than those in hexaploid wheat (294) and maize (107) [13], but more than that in sorghum (77), rice (85) [12], and Brachypodium [13], and much higher than that in Arabidopsis (63) [10], respectively. As observed in this study, 58 of the 94 (62%) SiAATs were duplicated in foxtail millet, including 36 and 25 tandemly and segmentally duplicated genes, respectively. All these further veri ed the in uence of gene duplication of the AAT family expansion in monocots. Furthermore, obvious imbalances in the expansion of different AAT subfamilies in monocots and eudicots were found, as AAP, ATLa/b, ACT and CAT subfamilies expanded greatly in monocots, while more ProT genes in eudicots (Table 2; Fig. 13). These increased AATs may play important roles in speci c functions, and promoted further differentiation of monocots and eudicots.
The multi-levels variations of paralogous AAT genes generated from gene duplication promoted the functional diversity of AAT family in foxtail millet Gene duplication and subsequent functional differentiation of new genes promoted the expansion and new-functionalization of important gene families, thereby enabling plants to better adapt to the environment [46]. Variations in the sequence, structure and expressions of genes were the three main sources of functional differentiation [13]. Base mutation was the most direct evidence of sequence variation. In general, the Ka/Ks values of all duplicate groups except for TD12 were far less than 1, indicating that most of the paralogous AAT genes undergone puri cations in sequence variation to ensure the stability of key biological functions (Fig. 5). The Ka/Ks values of different AAT subfamilies showed different median and dispersion, such as AAP, PHS and CAT subfamilies had higher dispersion, which indicated that these subfamilies may have the opportunity to produce new features. However, the Ka/Ks values of ACT subfamily, both tandemly and segmentally duplicated gene pairs, showed smaller dispersion, suggesting that their functions might be more conservative. In addition, tandemly duplicated gene pairs showed higher Ka/Ks values than segmentally duplicated gene pairs, suggesting they were under stronger selection pressure, which con rmed that tandem duplication played more crucial roles in functional diversity of foxtail millet AAT genes. Similar results were also found in the study of wheat AAT family genes [13]. These results con rmed that the sequence variations of paralogous genes in different AAT subfamilies of foxtail millet increased the diversity of AAT genes.
Previous studies have shown that variations in gene structure are also important markers of gene family evolution [48,49]. In this study, 56% (14/25) of the paralogous gene groups had signi cant changes in the number of introns, resulted in 9 and 5 groups of tandem and segmental duplications, respectively, accounting for 69% and 42% of the total (Table 2; Additional le 3: Figure S3). The variations in the number of TM regions were found in 77% (10/13) and 42% (5/12) of the tandemly and segmentally duplicated gene groups, respectively, which were caused by the deletion of conserved domains, such as SiAAP5/6/7 (TD1), SiAAP8/9 (TD2) ( Fig. 1b; Fig. 4). There were signi cant variations in gene structure and the number of TM regions which was observed in 56% (14/25) and 60% (15/25) of the two duplicated gene groups, con rming that the structural variation of paralogous genes was one of the important sources of the functional categorization of the AAT family genes in foxtail millet.
Changes in expressions of paralogous genes are another key sources of functional categorization of important gene family [49]. Sub-functionalization and new-functionalization of new duplicated genes directly lead to signi cant differences in expression level and spatiotemporal expression patterns, compared to their ancestor genes [48]. As observed in rice [12], soybean [11] and potato [10], the changes of the three expression types of paralogous genes were also observed in this study. For instance, SiAAP8/9 (TD2) and SiAAP19/20 (TD4) were speci cally expressed in different tissues, (Fig. 6), con rming that the expression changes of paralogous genes in AAT family were common in higher plants. Interestingly, the paralogous genes with new functions were only found in tandemly duplicated gene groups, suggesting that tandem duplication has a greater contribution to the evolution of the AAT gene family in foxtail millet.

The changes of orthologous AAT genes in foxtail millet play important roles in differentiation of new functions
Along with the separation of species and the in uence of natural selection, though the orthologous genes of different species are largely conserved, the variations in their regulatory elements, epigenetic markers, etc will cause changes in the pattern of expression of orthologous genes in various species, resulting in their functional differentiation [50][51][52][53][54]. For some C 4 genes in C 4 species, such as PEPC, MDH and PPDK, their C 4 functions were directly obtained by the differences in the subcellular localization and expression level of orthologous genes [55]. The completion of genome sequencing allowed us to quickly identify the orthologous genes of different plants at genome level (Fig. 7). There were 62, 51, 52, 35 orthologous of the SiAATs identi ed in sorghum, rice, wheat and Arabidopsis, which were consistent with the genetic distance and differentiation time of these species [56]. The AAP subfamily of grass species expanded signi cantly after the differentiation of monocots and eudicots, which was consistent with previous studies [12,13]. Within the grass family, the LHT and LAT subfamilies expanded signi cantly in foxtail millet and sorghum, which may have improved their adaptability to the environment. Most SiAATs and their orthologous in other species had similar expression patterns, especially for those of grass family with high correlation coe cients ranged from 0.470 to 0.483 (Fig. 8). Compared with their orthologous in Arabidopsis, the tissue expression characteristics of some SiAATs had changed, which may be related to the long genetic distance and complete differentiation between foxtail millet and Arabidopsis.
Many important AAT genes were expressed conservatively in multiple species to retain essential biological functions Spatiotemporal expression analysis of important gene family members will help to predict their putative functions [57]. Some SiAATs, such as SiProT1 and SiATLa5, were abundantly expressed during the whole period of growth, which were essential for the growth and development of foxtail millet (Fig. 9). Some SiAATs were abundantly expressed in speci c organs at certain developmental stages, such as SiAAP9, SiAAP7 and SiAAP2 in germinating seeds, SiAAP1 and SiAUX1 in stems and roots, and SiLAT7 and SiLAT8 in leaf tissues, while, SiAAT6 was highly expressed only in leaves at seedling and heading stages, but low at lling stage. The different spatiotemporal expression patterns of these genes suggested that they might have speci c functions.
The functions of some AAT genes have been extensively studied in the model species, such as AtAAP1 for seed germination and amino acid intake in root [9,14,58]. It was found that SiAAP7 and SiAAP2, OsAAP1 [26] ppossessed similar functions; AtAAP3 in the uptake of amino acids in xylem, and SiAAP4 highly expressed in stems and veins [21]; StAAP1 was expressed in multiple organs of source-sink transport in potato [24,10]; AtLHT1 located in root epidermis and mesophyll cells in Arabidopsis, which was primarily responsible for the amino acid uptake in soil [28,29]; AtAUX1, AtLAX3 and OsAUX1 related to the gravitropism of root and lateral root formation [12,32,33]. As similar expression patterns were also observed for orthologous genes, which maintain the important process of plant growth and development. Therefore, we preliminary predicted their functions in foxtail millet based on these previous studies, the phylogenetic relationships and the spatiotemporal expression characteristics of SiAATs in foxtail millet (Table 3). The intake of amino acid in endosperm and embryo development; Amino acid absorption in root and transport in stems; Play a role in early embryo development and formation of oral organs and grain quality; Participate in the transport and distribution of amino acids in leaf and grain formation; Participate in the transfer and distribution of amino acids from stems to leaf AtAAP8,1 [58,61] ; AtAAP7,1 [19] ; AtAAP8,1,VfAAP1,OsAAP6 [62,4,7] ; HvAAP3,AtAAP7 [25] ;AtAAP2,3,6,OsAAP8,15;HvAAP2,6 [20,21,8,12] GAT SiGAT3; Amino acid absorption in root and long-distance transportation of GABA AtGAT1 [44] ProT SiProT1 ST,L,GS Participate in the long-distance transport of proline and regulate the distribution of proline in leaf AtProT1,AtProT3 [5,38] LHT SiLHT11 SiLHT2,5,10 Transport and distribution of amino acids in leaf; Long-distance transportation of amino acids from root to leaf AtLHT1,9 [28,63] AUX SiAUX1; SiAUX2,3,4; ST,R; SP; G Participate in the geotropism of root and the transport of auxin in the stem; Participate in spikelet development and grain formation AtAUX1 [32][33][34] ; OsAUX1,StAUX [12,10]  , with a HMMER tool [64]. Third, the existence and integrity of the conserved domains in candidate proteins were further analyzed through conserved domain database (CDD) (https://www.ncbi.nlm.nih.gov/cdd) and InterPro Multiple AAT genes may contribute directly to grain quality traits in foxtail millet The AAP genes are closed related to grain development and quality formation in many species. AtAAP1 in Arabidopsis and OsAAP6 in rice are directly related to grain protein content [7,14], multiple AAT genes were located in or near the QTL regions associated with quality traits in wheat by genome-wide association analysis (GWAS) [27]. The RNA-seq analysis of the foxtail millet grains at development stages revealed that 70% AAPs (14/20), 67% CATs (8/12), 63% BATs (5/8), 83% ATLa (5/6) and 75% AUXs (3/4) were highly expressed during grain development, including SiAAP9 and SiAAP8. The orthologous genes of OsAAP6, suggested that those SiAATs played a major role in the development and quality formation of foxtail millet grain (Fig. 10). The DEGs between the two cultivars were enriched into four GO terms related to amino acid transport (Fig. 11a). Nine SiAATs, including SiAAP8, belonging to AAP, ANT, ATLa, AUX and PHS subfamilies, were differentially expressed in the developing grains of the two genotypes, which further con rmed that they might directly affect the grain quality traits (Fig. 11c). All these suggested that it might be feasible to improve the quality traits by manipulating the expressions of those important SiAAT genes in foxtail millet, which also provide a base for further analysis of their function in quality formation in foxtail millet.
SiAATs enhanced the adaptability of foxtail millet to abiotic stresses Drought and salt seriously affect crop yield and grain quality, and are the most important abiotic stresses in agricultural production. In Arabidopsis and other species, it has been con rmed that AAT genes in higher plants regulate the balance of osmotic potential to resist the impact of stress on plant growth, mainly through the transport of a variety of stress-response compounds and compatible solutes, such as betaine, GABA, and proline [40][41][42][43][44]. AtAAP4 and AtAAP6 were down-regulated under salt stress [42], and OsAAP4, OsAAP8 and OsBAT4 were down-regulated both under salt and drought stresses [12]. AtProT2 [42], McAAT1 [59] and HvProT [60] were greatly triggered by salt stress. In this study, active responsiveness to simulated drought and salt stresses were observed in 12 SiAAT genes investigated, and various response patterns to different stresses were found in different subfamily members (Fig. 12). It should be noted that SiAATs located on vacuole membranes such as SiANT1 and SiALTb2 responded stronger to these abiotic stress, which may relate to their roles in rapid adjustment of cell osmotic potential to maintain cell water balance. These genes could be used as key candidate genes to improve resistance and tolerance under abiotic stresses.

Conclusions
In this study, 94 AAT genes in foxtail millet were systematically identi ed and characterize and were divided into 12 subfamilies. The paralogous genes generated by tandem duplication and segmental duplication promoted AAT gene family expansion. The variations in sequence, structure and the pattern of expression of tandemly duplicated paralogous AAT genes had signi cant impacts on their functional differentiation. The functional differentiation of SiAATs was the result of the unequal expansion and differentiation of paralogous genes in different subfamilies. The transcriptome analysis revealed the contributions of some AAT genes in the formation of millet quality. Multiple SiAAT genes actively responded to drought and salt stresses as revealed by qRT-PCR. The possible functions of the SiAAT family members were predicted. These ndings might provide valuable information for further functional analysis of AAT genes and their applications in the improvement of foxtail millet. database (http://www.ebi.ac.uk/interpro/scan.html). Finally, the AAT gene family members in foxtail millet were identi ed after removing those containing incomplete conservative domains and redundant sequences.
Chromosomal mapping, duplication and selective pressure analysis The locations of SiAAT genes on the chromosomes were obtained from the ensemble plants database. The gene duplication was determined by MCScanX, and manual screening was performed according to Wang et al. [67,68]. The distributions of AAT genes and their duplicated paralogous on chromosomes were ploted with Btools software [69]. The ratios (Ka/Ks) of non-synonymous (Ka) and synonymous (Ks) of these paralogous AAT genes were analyzed using KaKs_Calculator 2.0 and displayed by an R package ggplot2 [70].

Phylogenetic and conserved motifs analysis
Multi-sequence alignment of the AAT protein sequences identi ed in Arabidopsis, potato, rice and foxtail millet was conducted with ClustalW, and phylogenetic tree was created using the maximum likelihood (ML) method, JTT model and a bootstrap of 1000 using MEGA 6.06 [71]. The phylogenetic tree was displayed with iTOL v3 (http://itol.embl.de/#).
The conserved motifs were determined using Motif-based sequence analysis (MEME) program (http://meme-suite.org) with the parameters of motif width of 6-200, the maximum motif number of 20 and any number of repeating motif sites [72]. TM regions and the possible protein secondary structures obtained from Protein Data Bank (PDB) database (https://www.rcsb.org/) were annotated and displayed by ESPript 3.0 [73].
Identi cation of the orthologous AAT genes in sorghum, wheat, rice and Arabidopsis, and their expression analysis The orthologous AAT genes in sorghum, wheat, rice and Arabidopsis were identi ed as those in foxtail millet, and by OrthFinder2 [74]. Their expression data in the main organs of root, stem, leaf, in orescence/spike, and grain was obtained from the public transcriptome databases, respectively (wheat, http://www.wheat-expression.com; rice, http://expression.ic4r.org; sorghum, http://structuralbiology.cau.edu.cn/sorghum/; Arabidopsis, https://www.arabidopsis.org). Then, the conservation of the expression patterns of AAT orthologous genes among those species was investigated through the tissue-to-tissue and gene-to-gene correlation analysis, using the software of SPSS 19.0 (SPSS, Inc., Chicago, IL, USA).
To study the spatiotemporal expression patterns of SiAATs in various tissues and grain development, a merged transcriptome dataset of foxtail millet was constructed with the RNA-seq reads downloaded from Multi-omics Database for Setaria italica (MDSi, http://foxtail-millet.biocloud.net/home), including 13 tissues at the seedling, heading and lling stages, and 2 spikelets and 5 grains. The FPKM (Fragments Per Kilobase per Million) values were calculated using an R package Edge R [75]. The heat maps of log 2 (FPKM + 1) values were drawn by TBtools [69].
Quality traits measurement and transcriptome analysis of two foxtail millet cultivars Two foxtail millet cultivars "JG21" and "YG1" used in this experiment were widely grown in northern China and provided by Mr Junjie Wang of Center for agricultural genetic resources research, shanxi agricultural university. They were grown in the experimental farm of the Institute of Water Saving Agriculture in Arid Areas of China, Northwest A&F University, Yangling, Shaanxi, China (34°7'N, 108°4'E) for two cropping seasons (May to September of 2017-2018 and 2018-2019), followed the local management practices. After harvest, the mature seeds were taken for measuring quality traits, including glutenin content, crude ber content, crude fat content, and the contents of 18 essential amino acids of methionine, cysteine, phenylalanine, alanine, glycine, glutamic acid, arginine, lysine, tyrosine, leucine, proline, tryptophan, Serine, threonine, aspartic acid, valine, isoleucine, histidine, with a near-infrared analyzer DA7250 (Perten, Sweden).
The developing grains (include glumes) of "JG21" and "YG1" were separated at lling stage in 2019, and the total RNA was isolated using a Quick RNA Isolation kit (Takara Corporation, Dalian, China) [76], RNA purity and integrity was con rmed by an RNA Nano 6000 Assay Kit and the Bioanalyzer 2100 system (Agilent Technologies, CA, USA). The RNA-seq library construction and sequencing were performed with the 150 paired-end sequencing on Illumina HiSeq platform following the standard methods by Novogene Life Sciences Pvt. Ltd., Beijing, China. After ltered the raw data, Clean reads were then mapped to the reference genome using Hisat2 [77]. Differentially expressed genes (DEGs) between "JG21" and "YG1" were identi ed using the R package DESeq [78], and the FPKM values were also calculated [79]. The KEGG (Kyoto Encyclopedia of Genes and Genomes and GO (Gene Ontology, http://geneontology.org/), http://www.genome.jp/kegg) annotation of DEGs were further performed with GOseq and KOBAS software, respectively [80].
The expressions of some AAT genes in response to abiotic stresses To verify the reliability of the transcriptome data from RNA-seq, the ag leaves, peduncles, roots, developing grains (include glumes) of "JG21" and developing grains (include glumes) of "YG1" were collected at the lling stage in 2019 for RNA extraction. For abiotic stresses, the raw read data of two datasets (SRA062640 and PRJNA545871) of foxtail millet under drought stress and salt stress were sourced from the NCBI SRA database to analyze the response of SiAATs to abiotic stresses [81,82]. The detailed analysis process was described above.
The seedlings, 15 days after sowing (two leaves with a terminal bud) were exposed to 20% polyethylene glycol (PEG 6000) solution and NaCl solution (200 mM) for 1 h and 5 h to mimic salt and drought stress, respectively, and all seedling plants were then collected for RNA isolation. PrimeScriptTM II 1st Strand cDNA Synthesis was used to synthesize the rst strand of cDNA and Quantitative Real-time PCR (qRT-PCR) was conducted by SuperReal PreMix Color kit (SYBR Green) with the primers listed in Table 3 and Additional le 7: Table S3, with the SiAct-7 gene as an internal reference. Each sample contained three biological replicates, and the 2 −ΔΔCT method was used for the statistical analysis.