Identification and Expression Analysis of the Barley (Hordeum vulgare L.) Aquaporin Gene Family

Aquaporins (AQPs) are major intrinsic proteins (MIPs) that mediate bidirectional flux of water and other substrates across cell membranes, and play critical roles in plant-water relations, dehydration stress responses and crop productivity. However, limited data are available as yet on the contributions of these proteins to the physiology of the major crop barley (Hordeum vulgare). The present work reports the identification and expression analysis of the barley MIP family. A comprehensive search of publicly available leaf mRNA-seq data, draft barley genome data, GenBank transcripts and sixteen new annotations together revealed that the barley MIP family is comprised of at least forty AQPs. Alternative splicing events were likely in two plasma membrane intrinsic protein (PIP) AQPs. Analyses of the AQP signature sequences and specificity determining positions indicated a potential of several putative AQP isoforms to transport non-aqua substrates including physiological important substrates, and respond to abiotic stresses. Analysis of our publicly available leaf mRNA-seq data identified notable differential expression of HvPIP1;2 and HvTIP4;1 under salt stress. Analyses of other gene expression resources also confirmed isoform-specific responses in different tissues and/or in response to salinity, as well as some potentially inter-cultivar differences. The work reports systematic and comprehensive analysis of most, if not all, barley AQP genes, their sequences, expression patterns in different tissues, potential transport and stress response functions, and a strong framework for selection and/or development of stress tolerant barley varieties. In addition, the barley data would be highly valuable for genetic studies of the evolutionarily closely related wheat (Triticum aestivum L.).


Introduction
Aquaporins (AQPs) belong to the superfamily of membrane channels called the major intrinsic proteins (MIPs). The plant MIPs (often generically called 'aquaporins') are typically divided into seven subfamilies: the plasma membrane intrinsic proteins (PIPs), tonoplast intrinsic proteins (TIPs), nodulin-26-like intrinsic proteins (NIPs), small, basic intrinsic proteins (SIPs) [1] including the gene sequences, expression patterns and the potential transport substrates of the encoded proteins.

Identification of barley aquaporin sequences
Barley AQP sequences were identified from the leaf mRNA-seq dataset obtained earlier, the NCBI UniGene database, and the barley draft genome sequence, as follows. For mining the leaf mRNA-seq data (publicly available, deposited as Sequence Read Archive accession number SRA062960 ( [19]; including supplemental data) to identify the transcripts representing AQPs, keyword searches were carried out for putative protein encoding sequences annotated as 'aquaporin', 'PIP', 'TIP', 'NIP', 'SIP', 'intrinsic', 'channel', 'transmembrane' and 'nodulin-like' based on their similarity to rice proteins. For accessing the NCBI UniGenes, the accession numbers of cDNAs corresponding to each barley UniGene were retrieved from their profile in NCBI (http://www.ncbi.nlm.nih.gov/unigene/; last accessed December 2014). The cDNAs were translated in Gene Runner (http://www.generunner.net/) and inspected for conserved features such as NPA motifs. The nearest barley relatives of rice AQP genes identified earlier [8] were mined from the NCBI UniGene database. All thus-identified UniGenes were also translated and the barley cDNAs representing these identified from NCBI as above. For searching the barley draft genome, the cDNAs identified in mRNA-seq and the rice AQPs (http://rice. plantbiology.msu.edu/; last accessed December 2014) were subjected to BLASTn against the International Barley Genome Sequencing Consortium (IBGSC) [20] IPK Barley Blast Server (http://webblast.ipk-gatersleben.de/barley/; last accessed December 2014) high confidence CDS (HC_genes_CDS_Seq) and full length cDNA databases (ftp://ftpmips.helmholtzmuenchen.de/plants/barley/public_data/genes/; last accessed December 2014) (e-value threshold <0.01). All data were used to compile a comprehensive list of all identified barley AQP cDNAs. These were then subjected to BLASTn against the IBSC assembly_WGSMorex and assembly_WGSBowman databases (http://webblast.ipk-gatersleben.de/barley/; last accessed December 2014) (e-value <0.01) to obtain their genomic sequences and physical locations. Additionally, the low confidence gene set (LC_genes_CDS_Seq) was also mined to note any other potential AQPs.

Analysis of expression of barley aquaporins
The mRNA-seq data were analysed for relative expression of AQPs in the control and saltstressed leaf RNAs. Details of each barley EST corresponding to a UniGene (cultivar, tissue, treatment, development stage) were sourced from the NCBI EST database (http://www.ncbi. nlm.nih.gov/nucest/16322814) (data not shown). Also, the probeset IDs corresponding to AQPs were obtained from HarvEST:Barley v1.83 (http://harvest.ucr.edu/) using the 'search by GenBank number, EST name or unigene number' tool, and some also from Besse et al. [21]. The IDs were used for analysis of their expression profiles in Genevestigator (https://www. genevestigator.com/gv/). The RNA-seq data of the eight tissues related to the IBGSC high confidence gene predictions (ftp://ftpmips.helmholtz-muenchen.de/plants/barley/public_data/ expression/README) were used to obtain expression scores for genes with MLOC numbers and generate a heat map of log2 normalised expression.

Identification of barley aquaporin genes
The search for AQPs in the SUT barley leaf transcriptome developed earlier [19] [GenBank SRA062960] by keyword and rice locus identified thirty-one UniGenes. The corresponding cDNA numbers to these (e.g., http://www.ncbi.nlm.nih.gov/UniGene/clust.cgi?UGID= 2919645&TAXID=4513&SEARCH=Hv.23281) showed the cDNAs represented twenty-two annotated (eleven PIPs, seven TIPs, four NIPs) and eight unannotated genes (Table 1; S1  Table). By mining the IBGSC barley genome project CDS and genomic databases with the cDNAs identified from the MSU rice cDNA database and the barley leaf mRNA-seq data, 36 CDSs and 40 genomic sequences could be identified altogether. By comparing these to the rice and barley cDNAs, 22 genomic sequences could be annotated as above, while sixteen were unannotated, comprised of eight as above and eight additional ones. Additionally, a genomic sequence identified in Bowman and Morex genomes (contig_222714; contig_401367) was not present in the CDS database, and a BLASTn against NCBI revealed that both contigs corresponded to HvTIP5;1 cDNA (AB540227). Based on % identities to other reported barley sequences (Table 1) or to orthologs in rice, maize or sorghum (data not shown), it appears that some annotations in Genbank may need to be revised, i.e., (i) AB540229, currently annotated as HvNIP2;1, seems better annotated as HvNIP2;2; (ii) AB710142 annotated as HvNIP2;2 should be HvNIP2;1; (iii) GU584119 annotated as HvTIP1;3 should be HvTIP1;2; (iv) GU584121 annotated as HvTIP2;1 should be HvTIP2;3 (Table 1). The BLASTn of the unannotated sequences against the Rice MSU cDNA database revealed they represented seven PIP2s (tentatively called HvPIP2a-g), three TIPs (HvTIP3, HvTIP4a, HvTIP4b), four NIPs (HvNIP2, HvNIP3a, HvNIP3b, HvNIP4) and two SIPs (HvSIP1, HvSIP2). BLAST queries of PpGIP1;1 [2], PpHIP1;1 [3] and GmXIP1;1 [6] against the barley genome database led to no hits, indicating these subfamilies are absent or highly divergent in barley, consistent with reports of their absence/loss in monocots [3,5]. Additionally, searches of the low confidence (LC) gene set led to nine sequence, of which one (MLOC_43388.1) was the longest (S2 Table). However, their putative translation products indicated they were partial and lacked NPA motifs, except two sequences (MLOC_37440.1 and MLOC_43388.1) that exhibited one NPA. The LC genes need confirmation and were not included in further analyses. In summary, the methods collectively led to the barley AQP superfamily comprising of at least forty members, twenty four being annotated and sixteen unannotated (Table 1).
[21] (Table 1). Pairwise identity scores were then obtained by alignments of the rice and barley cDNAs and putative protein sequences to identify the closest orthologs (S3 Table).
Comparison of the cDNAs (from leaf mRNA-seq data [19] and IBGSC barley genome project) and the available complete genomic sequences (from IBGSC barley genome project) using GSDS indicated that the HvPIPs had zero to three introns, as noted in rice OsPIPs [8] but unlike the typically three introns in dicot PIPs, e.g., soybean GmPIPs [6] (Table 2; AQPs which lack complete gDNA sequences are not included). The HvPIPs exhibited a diversity in intron sizes (71 to 1,245 bp), while most introns of TIPs were <200 bp ( Table 2). Similar to soybean [6], two introns were common for HvTIPs, except HvTIP1s and HvTIP4;3. The NIP introns    Characteristic features and potential transport abilities of the putative aquaporins The putative protein sequences of the barley AQPs showed conservation of both NPA motifs in all isoforms except for HvNIP3;1 which exhibited NPS and NPV, and HvSIP1;1 and HvSIP2;1 exhibited Leu or Thr at the Ala of first NPA (Table 3; S2 Fig). The ar/R substrate selectivity filter residues for PIPs were F-H-T-R, similar to those in maize [7], wheat and rice [8,10]. The product of HvPIP2;7a lacked both NPAs and had unusual ar/R residues due to the frameshift (see above). The HvTIPs exhibited six different combinations, i.e., H-I-A-V (TIP1s), and Q-V-A-R (HvTIP5;1), confirming the variability in TIPs in other plants [6,8]. HvNIP2;1 and HvNIP2;2 showed G-S-G-R, similar to other NIPIIIs such as OsNIP2;1 and OsNIP2;2 [8]. The HvNIPI residues were W-V-A-R, characteristic of NIPIs in rice [8] and soybean [6]. The HvNIP3;2 presented a new combination (V-I-A-R). The P1-P5 positions deemed important for discriminating between authentic 'aquaporins' and glyceroporins [23], were more conserved at P2-P4 in all subfamilies ( Table 3). The radish RsPIP1s have lower water permeability than RsPIP2s, and a loop E residue in RsPIP1;3 (Ile244) compared to RsPIP2;2 (Val235) is suggested to be relevant [24]. All HvPIP1s possessed Ile while the HvPIP2s had Val (S2 Fig), suggesting the HvPIP1s may have lower water permeability. Further, the difference between OsPIP1s (lower water permeability) and OsPIP2s (higher water permeability) was attributed to a residue in TM2, where OsPIP1s have Ala and OsPIP2s have Ile/Val [25]. At this position, HvPIP1;1 and HvPIP1;5 had Ala, but all HvPIP2s as well as HvPIP1;2, HvPIP1;3, HvPIP1;4 had Val (S2 Fig), suggesting the latter group may be permeable to water and supporting the observation of Besse et al. [21] that HvPIP1;2 is permeable to water. Some of the HvTIPs and HvNIPs also possessed a Val here (e.g. HvTIP2;1, HvTIP2;3, HvNIP1;1, HvNIP1;2), suggesting this residue may be functionally important for water transport of TIPs and NIPs. In line with this suggestion, a recent study observed that HvNIP1;1, HvNIP1;2, HvNIP2;1 and HvNIP2;2 all of which possessed the Val residue in TM2 (S2 Fig), were water transporters [26].
AQPs are known to function as tetramers and a highly conserved Cys (Cys80 in ZmPIP2;1) in loop A of PIP1s and PIP2s is shown to be essential for formation of disulphide bonds between PIP monomers, increasing the oligomer stability [27]. Alignments showed conservation of this Cys in all HvPIP1s and PIP2s (S2 Fig), suggesting a similar role for it in these, but its absence in TIPs, NIPs and SIPs suggests other residue(s) may be involved in this role. All PIPs were predicted to localise to the plasma membrane, as the subfamily name suggests and consistent with localisation of HvPIP2;1 [12]. However, an N-terminal diacidic motif Asp-Ile/Val-Glu [28,29] found important for exit of newly synthesised PIPs from the ER to plasma membrane was noted only in HvPIP2;1, HvPIP2;3 and HvPIP2;4. Further, the subfamily names do not always indicate location (reviewed in [8]). In congruence with this, the TIPs were predicted to localise to the plasma membrane, mitochondria and chloroplast, in addition to the expected tonoplast (Table 3). Some NIPs were predicted to localise to the plasma membrane and HvSIP1;1 was predicted in the endoplasmic reticulum. All HvPIPs except HvPIP2;7 and HvPIP2;10 possessed a His corresponding to His193 in SoPIP2;1 considered important for pH-dependent gating [30] (S2 Fig), but the TIPs, NIPs and SIPs did not show this residue. Gating of SoPIP2;1 by phosphorylation of Ser115 and Ser274 has been shown, with blocking by   [21]. *Only the highest % identity score of each sequence is listed. The identity scores were obtained based on alignments of the cDNAs and putative protein sequences by pairwise identities in BioEdit (S3 Table). The AQPs annotated in this study are shown in bold. Sub-cellular locations: Chl (chloroplast), ER (endoplasmic reticulum), M (mitochondria), PM (plasma membrane) and V (vacuole). Our earlier analysis of plant AQPs with non-aqua transport substrates had revealed distinct signature sequences and 'specificity determining positions' (SDPs) for each substrate group [32]; hence the putative barley AQPs were analysed for these. A number of isoforms were predicted to have the potential to transport micronutrients and molecules involved in plant growth and vigour (boron also being toxic in high concentrations) ( Table 3). The potential to transport urea (many PIPs, TIPs and NIPs) and boron (many HvPIPs but no TIPs) was widespread, but that for ammonia was restricted to some TIPs (HvTIP2;1 HvTIP2;2; HvTIP2;3), CO 2 to HvPIPs (all except HvPIP2;7), and silicon to NIPs (HvNIP2;1, HvNIP2;2; HvNIP2;2). The potential to transport the signalling molecule H 2 O 2 seemed restricted to the TIPs (Table 3). NIPs had the most diverse predicted substrates, as found in other studies, including the potentially toxic arsenite.

Modulation of expression of barley aquaporin genes
In mRNA-seq, gene expression level is generally measured as the number of sequence reads mapping to a gene [33]. In the present study, the normalised sequence reads (base mean expression) were used. Most PIPs and some TIPs exhibited higher expression levels than NIPs and SIPs (S3 Fig), in line with Arabidopsis [34]. HvPIP1;4 had the most abundant transcripts in control plants, which nearly doubled upon exposure to salt stress (fold change (FC) +1.93; S1 Table). FC of +1.5 or -1.5 is considered notable [35], hence in the present context, such genes may have roles in osmotic regulation under salinity stress. Several other PIPs also showed notable changes, the largest being noted for HvPIP1;3 (FC +2.93) but that in HvPIP1;2 (FC +1.97) being statistically significant. Among the TIPs, HvTIP1;2 and HvTIP4;1 were the most abundant AQP transcripts in control plants, but only HvTIP4;1 displayed a statistically significant decline (FC -15.89) under salinity, suggesting a critical role for it. HvTIP1;1 and HvTIP2;3 showed moderate expression and up-regulation (FC +1.62, +2.60, respectively). HvTIP2;1, HvTIP2;2, HvTIP3;1 and HvTIP3;2 exhibited very low expression levels, but HvTIP3;2 was significantly up-regulated (FC +2.40). HvNIP2;2 was the most abundant NIP, but expressed at much lower levels than many PIPs and TIPs, and showed little change. The five other NIPs exhibited low-moderate expression and only HvNIP2;1 showed notable down-regulation. HvSIP2;1 had higher and more differential expression than HvSIP1;1. In summary, thirteen genes showed up-regulation and five showed a decrease under salinity, of which the changes in HvTIP4;1 and HvPIP1;2 were statistically significant (p-value <0.05).
The ESTs corresponding to the forty genes above were analysed and grouped as follows; apex, callus, epidermis, flower (anther, carpel, inflorescence and pistil), leaf, maternal, root, seed (caryopsis, coleoptile, embryo, endosperm, pericarp, testa), shoot, and spike (rachis) (S4 Table). Based on the total number of ESTs, the PIPs (2,043) and TIPs (1,260) appeared to have much higher expression than the NIPs (51) and SIPs (40), confirming other reports [34]. The PIPs and TIPs also typically showed higher expression in the leaf (PIPs: 285; TIPs: 112), root (546; 305), seed (215; 417) and shoot (798; 281) compared to other tissues (S4 Fig). Some ESTs were noted in response to stresses such as cold, drought, salinity or waterlogging (S5 Table). The heat-map generated for the IBGSC RNA-seq data of eight tissues related to the high-confidence gene predictions supported high expression levels of PIPs and TIPs in shoots (including tillers), inflorescences and developing grain, but not roots and germinating seed (S5 Fig). HvTIP1;1, HvTIP2;1, HvTIP2;3 and HvPIP1;1 were highly expressed in a number of tissues. In many tissues the expression levels of NIPs were lower than or the same as the TIPs and PIPs, except for high levels of HvNIP3;1 (developing grain) and HvNIP4;1 (inflorescence). A number of isoform-specific patterns were also noted.

Discussion
In recent years, the function of plant AQPs has been extended from transport of water alone to the transport of other substrates and responding to diverse physiological processes and environmental stresses (reviewed in [1]). However, while significant information exists on the AQP families in rice and soybean, a complete picture of this family is missing in barley and wheat, two of the most important cereal crops. In this work, we address this gap by providing comprehensive analysis of what we believe is the full gene set of the barley AQPs superfamily, including new annotations of sixteen transcripts, removal of some redundancies and possible inaccuracies of four annotations (HvNIP2;1, HvNIP2;2, HvTIP1;2 and HvTIP2;3) ( Table 1). The annotation of HvNIP2;1 may be regarded as somewhat equivocal, as the sequences AB447482 (registration 14-JUL-2008; Ma, direct submission) and GQ496520 (18-AUG-2009; Sutton et al., direct submission) were annotated as HvNIP2;1, followed by registration of AB540229 (13-JAN-2010; Ligaba et al., direct submission;) also as HvNIP2;1. However, AB540229 is 100% identical to AB447484 and not to the former two. Our annotations of HvNIP2;1 and HvNIP2;2 are based on the dates of registration as well as % identities to the other annotated barley sequences (listed in Table 1) and to the orthologs in rice [37], maize [7] and/or sorghum (SbNIP2;1 -EF373651; SbNIP2;2 -EF408053). Hence we suggest that AB540229 represents HvNIP2;2, and likewise, AB710142 (annotated as HvNIP2;2; 02-APR-2012, Shibasaka et al, direct submission; also see [26]) represents HvNIP2;1. These suggestions are open to further reports and/or Genbank updates. In all, forty genes (nineteen PIPs, eleven TIPs, eight NIPs, two SIPs) were identified, making a significant advancement on the twentythree (ten PIPs, eight TIPs, three NIPs, two SIPs) estimated by Katsuhara and Hanba [11] and twenty-five (eleven PIPs, eight TIPs, four NIPs, two SIPs) estimated by Besse et al. [21]. This family size is consistent with maize (36; [7]), rice (38; [8]) and soybean (66; [6]). Subject to confirmation, nine more sequences identified in the low confidence dataset may also belong. The AQP diversity in plants is attributed to the higher degree of compartmentalization of plant cells and the ability to fine-tune water control in situ, under different environmental conditions [34]. The gene identifications and analyses are also relevant to wheat, a genetically close relative; e.g., a previous study could not amplify TIP5 from wheat, with primers based on rice TIP5 [10]; the data on HvTIP5 will address this limitation. A wheat NIP is reported to be involved in salinity tolerance [38]; the present work will enable analysis of its orthologue in barley.
The isoform-specific response of barley AQPs noted here support the observations in maize [41]. Tissue-specific responses were also observed, e.g., HvPIP1;2 was down-regulated in the root [11,13], but up-regulated in leaf mRNA-seq (S3 Fig; S1 Table). Differences were also noticed in results from different techniques; e.g., HvPIP2;1 was found largely unchanged in the microarray (Table 4), but down-regulated in roots [12,13] and up-regulated in shoots by real-time PCR [12] and leaf mRNA-seq [19] and in IBGSC shoot + tillers RNA-Seq data (S5 Fig). In addition, the response of aquaporins to stimuli is often species-dependent. For example, PIP1;1 from rice was shown to be down-regulated in response to 150 mM NaCl [42], while its ortholog in barley (HvPIP1;1) was up-regulated (our current study). The expression of AQPs during water stress is reported to be dependent on cultivars [43] as well as time course and intensity of stress [41]. Thus investigations of wider germ-plasm for allelic differences are essential for assessing the genetic versus environmental factors (including experimental techniques) to such studies.
An N-terminal diacidic motif Asp-Ile-Glu in ZmPIP2;4 and ZmPIP2;5 [28] or Asp-Val-Glu in AtPIP2;1 [29], found important for exit of PIPs from the ER to plasma membrane, was noted in HvPIP2;1, HvPIP2;3 and HvPIP2;4, suggesting they may use this mechanism. It is unclear how the other HvPIPs lacking this motif but also predicted to be plasma membrane-located would be exported. For PIP1s, hetero-tetramer formation with PIP2s has been suggested as means of ER export [28]. The plant cell vacuole is associated with turgor regulation, osmotic adjustment, storage, pH regulation, cell signalling and protein degradation. HvTIP4;1 was predicted to be vacuolar (Table 3), similar to AtTIP4;1 [44]. OsTIP4;1 has been confirmed to be water-permeable and its transcripts fluctuate, possibly in response to turgor status [45], and the orthology suggests that HvTIP4;1 may play a similar role. In addition, other TIPs predicted to be localised to the vacuole could also play this role.
AQP activity can be regulated transcriptionally, and/or post-translationally by pH, phosphorylation and vesicle trafficking [1]. Protonation of a His in loop D of PIPs is shown to be important for pH-dependent gating during flooding (e.g., His 193 in SoPIP2;1; [30]). All HvPIPs except HvPIP2;7 and HvPIP2;10 possessed a His corresponding to His193 (S2 Fig), but the TIPs, NIPs and SIPs did not, suggesting HvPIPs may be regulated by pH, but not other isoforms. Tornroth-Horsefield et al. [30] also demonstrated the gating of SoPIP2;1 by phosphorylation of Ser115 and Ser274, with Leu197 as the key blocking residue. The equivalents of Ser115, Ser274, and Leu197 in most HvPIPs (S2 Fig) suggest a similar regulation. In contrast, most TIPs lacked these Sers, and Leu197 occurred only in HvTIP4;3 and HvTIP5;1, suggesting the gating may not apply. The Ser at Ser262 in GmNOD26 [46] was noted in all NIPs (except HvNIP3;2) suggesting their phosphorylation-mediated regulation also. The regulation of AQPs is critical in maintaining the plant water status during normal and stress conditions, and this ability could be employed in the development of transgenic plants. AtPIP2;1 is mono-and di-phosphorylated at two C-terminal sites, S280 (i.e. S274 of SoPIP2;1) and S283 under salt stress [31]. Alignments with AtPIP2;1 showed a conserved Ser at S280 and S283 in most HvPIP2s (S2 Fig), suggesting that these, although not exhibiting transcriptional modulations under salt stress, may be regulated post-translationally.
In recent years, plant AQPs, predominantly NIPs, have been shown to be permeable to twelve different substrates, many of which are of physiological significance, e.g., ammonia and CO 2 [1], and the signalling molecule hydrogen peroxide [47], but some are toxic, e.g., arsenite. Hence, the involvement of the barley AQPs in transporting any non-aqua substrates was predicted using the specificity determining positions identified earlier [32]. Based on these predictions, a number of PIP isoforms have the potential to transport boron, CO 2 and urea which are important for plant development (boron also being toxic in high concentrations), the TIPs share the urea and ammonia transport potential but not CO 2 , and transport of the signalling molecule H 2 O 2 seems restricted to TIPs ( Table 3). Some of these are experimentally proven substrates, e.g., boron for HvPIP1;3 and HvPIP1;4 [48] and HvNIP2;1 [49]; CO 2 for HvPIP2;1-2;3 and HvPIP2;5 [50,40]; arsenite for HvNIP1;2 [26], and silicon for HvNIP2;1 [51]. Many NIPs appear to have the potential to transport diverse substrates including the micronutrient silicon, but also the toxic arsenite. Further, in our previous work [32] the transport of non-aqua substrates was predicted to generally compromise the water permeability of such AQPs; however, this did not seem to apply to AtTIP1;1, AtTIP2;1, AtNIP1;1, HvPIP2;1, OsPIP2;4, OsPIP2;6, OsPIP2;7, OsTIP1;2, PtNIP1;1 and TaTIP2;2. Thus, their orthologs in barley may also be water transporters.
HvPIP1;3 was found up-regulated in response to boron, while HvPIP1;4 was unchanged [48]. In the leaf mRNA-seq data, both of these were up-regulated under salinity stress (S3 Fig; S1 Table). Down-regulation of HvNIP2;1 is suggested to induce boron tolerance [48]; this isoform was found down-regulated by salinity in our study [19]. Orthologs often retain their function during evolution; thus HvNIP2;1 may transport the substrates shown for OsNIP2;1, i.e., boron, silicon and urea [52] and arsenite and antimony [53], and may not necessarily show significant differential expression under salinity stress (alone) (see below). Boron toxicity and salinity both significantly reduce crop yield [54], and may occur together, e.g., in parts of Western Australia. Thus isoforms such as HvPIP1;3 and HvNIP2;1 would be invaluable for identification/development of lines tolerant to both stresses. HvPIP1;1, HvPIP1;2 and HvPIP2;1 are also of interest due to being salt-responsive and possible boron transporters, and HvPIP1;5, HvPIP2;3 HvPIP2;5, HvNIP2;2 and the newly identified HvNIP3;1 are also likely boron transporters. Urea is abundant in nature and application of urea fertilisers is a common agricultural practice; hence there is a need for plants to 'load and unload' urea [44], and some of the barley isoforms may be urea transporters. Silicon is found to alleviate abiotic and biotic stresses and improve light interception and canopy photosynthesis, and HvNIP2;1 and HvNIP2;2 being silicon transporters ( [55] and within) was supported by our predictions. Isoforms that transport antimony or arsenite could be valuable in their detoxification. The rice orthologs could assist in testing the functions of barley genes, and the sequences and physical locations of barley genes are important for marker development and studying wheat orthologs.

Conclusions
Forty AQPs is the largest number reported for barley so far. The data acquired on their coding sequences, innate tissue and/or cultivar specific variations in expression, modulations under salinity, and the potential roles of various genes in transport of other substrates and in salinity and other stress responses, have collectively yielded a wealth of molecular information. This will be vital for assessing and/or transgenically developing germplasm of barley and possibly of wheat with improved abiotic stress tolerance and yield.