, Volume 238, Issue 4, pp 669–681

Genome-wide expression analysis of rice aquaporin genes and development of a functional gene network mediated by aquaporin expression in roots


  • Minh Xuan Nguyen
    • Department of Plant Molecular Systems Biotechnology and Crop Biotech InstituteKyung Hee University
  • Sunok Moon
    • Department of Plant Molecular Systems Biotechnology and Crop Biotech InstituteKyung Hee University
    • Department of Plant Molecular Systems Biotechnology and Crop Biotech InstituteKyung Hee University
Original Article

DOI: 10.1007/s00425-013-1918-9

Cite this article as:
Nguyen, M.X., Moon, S. & Jung, K. Planta (2013) 238: 669. doi:10.1007/s00425-013-1918-9


The world population continually faces challenges of water scarcity for agriculture. A common strategy called water-balance control has evolved to adapt plant growth to these challenges. Aquaporins are a family of integral membrane proteins that play a central role in water-balance control. In this study, we identified 34 members of the rice aquaporin gene family, adding a novel member to the previous list. A combination of phylogenetic tree and anatomical meta-expression profiling data consisting of 983 Affymetrix arrays and 209 Agilent 44 K arrays was used to identify tissue-preferred aquaporin genes and evaluate functional redundancy among aquaporin family members. Eight aquaporins showed root-preferred expression in the vegetative growth stage, while 4 showed leaf/shoot-preferred expression. Integrating stress-induced expression patterns into phylogenetic tree and semi-quantitative reverse transcriptase polymerase chain reaction (RT-PCR) analyses revealed that 3 rice aquaporin genes were markedly downregulated and 4 were upregulated by water deficiency in the root, suggesting that these candidate genes are key regulators of water uptake from the soil. Finally, we constructed a functional network of genes mediated by water stress and refined the network by confirming the differential expression using RT-PCR and real-time PCR. Our data will be useful to elucidate the molecular mechanism of water-balance control in rice root.


Drought stress in rootFunctional gene networkGene redundancyMeta-profiling analysis











Low silicon rice


Molecular evolutionary genetics analysis




Nodulin 26-like intrinsic proteins


Asparagine–proline–alanine region


Plasma membrane intrinsic proteins


Semi-quantitative reverse transcription polymerase chain reaction


Rice genome annotation project database


Rice water channel


Shoot apical meristem




Small basic intrinsic proteins


Tonoplast intrinsic proteins


Uncategorized X intrinsic proteins


Climate changes, such as drought, salinity, and fluctuating temperature, that cause abiotic stress, have been highlighted as one of the most harmful variables in affecting the crop growth and productivity, resulting in agricultural loss worldwide (Josine et al. 2011). A recent study reported that the average annual loss caused by drought in eastern India alone is $162 million (Pandey et al. 2007). In addition, increased salinity can reduce yields to 10–90 % of normal production for wheat and 30–50 % for rice (Eynard et al. 2005). In challenging climates, plants must adjust their water-balance control for normal development and maintenance of proper yield (Luu and Maurel 2005).

Aquaporins, a family of integral membrane proteins, rapidly transport water and small molecules, such as glycerol or volatile substances, across cell membranes in all organisms (Kaldenhoff and Fischer 2006; Zardoya and Villalba 2001). Since AQUAPORIN 1 (AQP1), an integral membrane protein in erythrocytes, was first discovered in mammals, many studies have been conducted to identify the diverse roles of aquaporins (Denker et al. 1988; Preston and Agre 1991). Molecular and functional characterization of aquaporins has revealed the importance of their regulation in response to stress stimuli from various environmental conditions (Luu and Maurel 2005). Specifically, aquaporins play a central role in plant water transport by regulating the water transport system in the root to protect against a variety of environmental stimuli and facilitate water transport through inner leaf tissues during transpiration (Maurel et al. 2008). Aquaporins also play several important roles in the physiology and pathophysiology of many organisms (Ishibashi et al. 2011).

Over 30 members of the aquaporin family are found in higher plants, e.g., 33 in rice (Sakurai et al. 2005), 35 in Arabidopsis (Johanson et al. 2001; Quigley et al. 2002), and 36 in maize (Chaumont et al. 2001). Aquaporin subfamilies contain distinct substrate specificities and subcellular localizations (Maurel et al. 2008). On the basis of these features, the aquaporin family has been divided into 5 subfamilies: plasma membrane intrinsic proteins (PIPs), tonoplast intrinsic proteins (TIPs), nodulin 26-like intrinsic proteins (NIPs), small basic intrinsic proteins (SIPs), and uncategorized X intrinsic proteins (XIPs) (Sakurai et al. 2008; Ishibashi et al. 2011). Although several studies have analyzed this gene family in plants, a systematic understanding of this family is lacking, especially in rice (Oryza sativa), a model crop plant.

Whole-genome sequencing enabled us to identify and classify various members of the aquaporin family in rice. A recent study identified 33 rice aquaporin genes: 11 PIPs, 10 TIPs, 10 NIPs, and 2 SIPs (Sakurai et al. 2005). Using semi-quantitative reverse transcription-PCR (qRT-PCR), Sakurai et al. (2005) monitored the expression of 33 rice aquaporin genes in rice leaf blades, roots, and anthers. The tissue- and cell-specific localization of rice aquaporin was also identified by immunoblotting and immunocytochemistry (Sakurai et al. 2008). In addition, the expression of 10 rice TIPs was assayed in roots, shoots, germinating seeds, and ABA-treated seedlings by real-time PCR (Li et al. 2008a).

Although systematic expression analyses have been performed for the aquaporin family in rice, actual functional identification was rare. Until now, 3 aquaporin genes have been functionally identified. The Low silicon rice 1(Lsi1) gene controls silicon accumulation in rice (Ma et al. 2006). The role of Rice water channel 3(RWC3) during drought avoidance in rice was identified by RWC3 overexpression in transgenic lowland rice (O. sativa L. spp japonica cv. Zhonghua 11), which is sensitive to water stress (Lian et al. 2004). Aquaporin OsPIP1;1 stimulates salt resistance and seed germination in rice (Liu et al. 2013). The expression of 11 OsPIP genes was observed in upland and lowland rice species under normal and water-deficient conditions, as well as in chilling-tolerant and chilling-sensitive rice under chilling stress and recovery (Lian et al. 2006; Yu et al. 2006). In addition, overexpression of either OsPIP1 or OsPIP2 in wild-type Arabidopsis, a heterologous expression system, triggered enhanced tolerance to both salt and drought stress (Guo et al. 2006). Transgenic rice plants overexpressing OsPIP1;3 enhanced the chilling tolerance, and the protein accumulated at low temperatures (Matsumoto et al. 2009). Furthermore, OsPIP2;7 induced an enhanced transpiration rate and tolerance to low temperature when overexpressed in rice (Li et al. 2008b). The functions of the remaining rice aquaporin genes remain unknown, possibly due to functional redundancy and limited information guiding functional studies.

Recently, we reported that selecting a single gene or predominant gene family member in response to light induction is effective in identifying defective phenotypes in loss-of-function mutants (Jung et al. 2008). In this respect, phylogenomic analysis integrating diverse meta-expression profiling data into a phylogenetic tree is effective to evaluate functional redundancy within this gene family.

In this study, we expanded the number of known rice aquaporin genes from the 33 identified by Sakurai et al. (2005) to 34 based on updated annotation information from the Rice Genome Annotation Project ( and Greenphyl ( Then, we generated and analyzed phylogenomic data integrating anatomical meta-expression profiling data based both on 983 Affymetrix arrays and 209 Agilent 44 K arrays, as well as differential expression patterns in response to abiotic stresses, into the context of the phylogenetic tree. These systemic overviews of aquaporin gene expression in rice were used to evaluate functional redundancy or dominance of each family member and to identify the aquaporin genes activated in response to water stress. A functional network of genes mediated by water stress-responsive aquaporin genes is proposed and discussed.

Materials and methods

Database search for rice aquaporin genes

Rice aquaporin members were identified by searching “aquaporin” as a key word in the Greenphyl ( (Rouard et al. 2011) and the Rice Genome Annotation Project (RGAP) databases ( (Ouyang et al. 2007). As a result, 34 aquaporin genes were obtained. Suppl. Table S1 provides more detailed information on the 34 aquaporin genes, including RGAP locus identifiers (IDs), RGAP database locus IDs, common names, gene accession numbers, protein accession numbers, and cDNA clone accession numbers.

Chromosomal positions of rice aquaporin genes

The RAGP locus IDs of 34 rice aquaporin genes were mapped on the chromosomes using the map tool function in the Oryzabase database ( (Fig. S1).

Sequences alignment, phylogenetic tree, and conserved motifs of rice aquaporin genes

The predicted protein sequences of the 34 rice aquaporin genes were downloaded from RGAP by querying the 34 locus IDs against the bulk download tool in RGAP. These sequences were then aligned with ClustalX version 2.0.11 (Larkin et al. 2007). A phylogenetic tree of aligned aquaporin sequences was constructed by Molecular Evolutionary Genetics Analysis (MEGA) version 5 (Tamura et al. 2011) using the following parameters: neighbor-joining tree method, complete deletion, and bootstrap with 500 replicates. Next, the conserved motifs among the full-length protein sequences were identified using the surveyed conserved motif alignment diagram and associating dendrogram database ( (Mihara et al. 2010).

A phylogenetic tree of rice aquaporins identified in this study was developed and included Arabidopsis and maize aquaporins identified in previous global analyses (Chaumont et al. 2001; Quigley et al. 2002). These data were used to assign rice aquaporins to subfamilies (Fig. 1).
Fig. 1

Phylogenetic analysis of aquaporin proteins in Arabidopsis, maize, and rice. 97 aquaporin proteins were classified as 35 plasma membrane intrinsic proteins (PIPs) in the brown tree, 31 tonoplast intrinsic proteins (TIPs) in the pink tree, 20 nodulin 26-like intrinsic proteins (NIPs) in the green tree, and 8 small basic intrinsic proteins (SIPs) in the light blue tree. At denotes Arabidopsis; Os rice, and Zm maize. Dark red bold letters denote rice aquaporin proteins. Numbers indicate estimated evolutionary distances

Integration of expression patterns to the phylogenic tree of aquaporin family

To investigate the anatomical expression of rice aquaporin genes, anatomical meta-profiling databases consisting of 983 Affymetrix arrays (Affymetrix, Santa Clara, CA, USA) and 209 Agilent 44 K arrays (Agilent Technologies, Santa Clara, CA, USA) were used. The log2-normalized intensity data in a tab-delimited text file format were then uploaded to Multi Experiment Viewer ( (Saeed et al. 2006) to generate a heat map (Fig. 2). In addition, a heat map was generated based on log2-fold change data in response to abiotic stresses, such as drought, salt, cold, heat, and submergence (Figs. 3, 4). The heat map was then integrated into the phylogenetic tree. To refine significant expression, genes were chosen that had average expression values greater or less than twofold from a coefficient of variation of less than 1.
Fig. 2

Anatomical expression of rice aquaporin genes by meta-anatomical expression profiles and RT-PCR. Root preference genes are denoted by asterisk. The Affymetrix anatomical expression database was based on 983 arrays of 11 anatomical tissues/organs. The Agilent 44 K anatomical expression database was based on 209 arrays of 10 anatomical tissues/organs. Heat maps were generated with average log2 intensity values in each tissue/organ. Blue in these heat maps denotes low expression and yellow denotes high expression. mRNAs from root, shoot, and leaf in the seedling stage were used for RT-PCR analysis. Rice Ubiquitin 1 (OsUbi1) and Rice Ubiquitin 5 (OsUbi5) were used as internal controls for RT-PCR. Cycles denote the number of PCR cycles. "−" denotes samples that were not amplified by RT-PCR
Fig. 3

Expression patterns of rice aquaporin genes in response to abiotic stresses using Affymetrix arrays. A heat map of changes in expression was generated using 5 replicates of roots under drought, 6 replicates of leaves under drought, 3 replicates of seedlings in salinity, 3 replicates of seedlings in cold, 3 replicates of seedlings in heat, and 4 replicates of seedlings submerged. Genes showing at least two fold upregulation or downregulation with less than 1 coefficient of variation (COV) are denoted as red or green, respectively. Gray denotes a less significant change data or data not analyzed by microarray analysis
Fig. 4

Identification of rice aquaporin genes responsive to drought stress in roots through microarray and real-time PCR. Microarray data for roots under drought were selected from Fig. 3. For real-time PCR analyses, we used rice roots 7 days after germination (DAG) with a 2-h drought treatment as without treatment. OsUbi5 and rice elongation factor 1α (OsEF1α) were used as internal controls. Refer to Figs. 2 and 3 for additional abbreviations. In the bar chart, blue bars denote no treatment and red bars denote drought treatment; "+" denotes upregulation and "−" denotes downregulation

In silico network analysis

A hypothetical functional network of genes mediated by rice aquaporin genes was developed from the RiceNet webtool ( (Lee et al. 2011). RGAP locus IDs of 9 genes (OsNIP2;2, OsNIP3;1, OsPIP2;5, OsPIP2;7, OsSIP1;1, OsTIP1;2, OsTIP3;1, OsTIP3;2, and OsTIP4;1) with significant drought responses were queried in the webtool ( Sixteen interactions among 15 genes including 5 aquaporin genes were predicted (Fig. 5). The network was further edited by Cytoscape version 2.8.1 (Shannon et al. 2003) by integrating drought stress-responsible expression into the nodes and refining the remaining genes to those with more than a twofold change. In addition, the network was confirmed by RT-PCR and real-time PCR analyses.
Fig. 5

Development and refinement of a functional gene network associated with rice aquaporin genes responsive to drought in roots. RiceNet was used to develop a functional network of genes mediated by OsSIP1;1, OsNIP3;1, or OsNIP2;2, which were downregulated in root in response to drought, and OsTIP3;1 and OsTIP3;2, which were upregulated in roots in response to drought. Four interactors related to OsSIP1;1, OsNIP3;1, and OsNIP2;2 and 6 interactors related to OsTIP3;1 and OsTIP3;2 are indicated in nodes. Green nodes denote at least twofold downregulation in response to drought in the root and red denotes upregulation. Significant expression of interactors was confirmed through RT and real-time PCR analyses to refine the functional network. The internal controls were the same as those used in Fig. 4. In the bar chart, blue bars denote no treatment and red bars denote drought treatment

Plant materials and drought treatment

Rice (O. sativa L. cv. Dongjin) seeds (obtained from the Rural Development Administration, Korea) were germinated on Murashige–Skoog medium (Murashige and Skoog 1962) for 7 days at 28 °C after sterilizing with 50 % (w/v) sodium hypochlorite (NaOCl) solution for 30 min with gentle shaking. Subsequently, the seedlings were washed with sterile water to completely remove the agar and were air-dried for 2 h at 28 °C under continuous 150 mol μm2 s−1 light to induce drought stress (Oh et al. 2009; Jeong et al. 2010).

RNA extraction, semi-quantitative RT-PCR, and real-time PCR

Roots of rice seedlings were frozen in liquid nitrogen and ground with a Tissue Lyser II (Qiagen, Hilden, Germany). Subsequently, RNA was extracted with RNAiso Plus Kit (Takara Bio Inc., Shiga, Japan). Root, shoot, and leaf mRNA at the seedling stage were used for RT-PCR analysis. Anatomical expression and drought stress analyses by RT-PCR were compared using 2 control genes: rice ubiquitin 1 (OsUbi1, LOC_Os03g13170) and rice ubiquitin 5 (OsUbi5, LOC_Os01g22490). Real-time PCR analyses were compared using OsUbi5 and OsEF1α (LOC_Os03g08010) as internal controls. The other primers are listed in Table S2.


Identification of rice aquaporin genes

Using the 2,771 rice gene families from Greenphyl (Rouard et al. 2011) and gene annotation by RGAP (Ouyang et al. 2007), we identified aquaporin genes in 37 loci, of which, 33 genes were identified by a previous phylogenetic analysis of the rice aquaporin family (Sakurai et al. 2005). These genes included 10 NIPs, 11 PIPs, 10 TIPs, and 2 SIPs. In contrast to the 3 other newly identified candidate genes, the full coding sequence of a novel candidate gene was validated using RT and real-time PCR analyses (data not shown). Detailed characteristics of the 34 rice aquaporin family genes are displayed in Table S1. We mapped the 34 rice aquaporin genes onto the 10 chromosomes to identify the physical location of this gene family on the rice chromosomes (Fig. S1).

Phylogenetic analysis of rice aquaporin family

We generated a phylogenetic tree for rice, Arabidopsis (Quigley et al. 2002), and maize (Chaumont et al. 2001) aquaporins using the MEGA5 program (Fig. 1). The aquaporin proteins were clearly divided into 4 subfamilies: NIP, TIP, PIP, and SIP. This observation is consistent with previous phylogenetic analyses of plant aquaporins (Chaumont et al. 2001; Luu and Maurel 2005; Sakurai et al. 2005). In addition, a phylogenetic tree of the 34 rice aquaporins was constructed (Fig. S2). In this phylogenetic tree, a novel gene—LOC_Os07g26640—belongs to the PIP2 sub-type. Several rice aquaporins had potential partial coding sequences, as mentioned previously (Sakurai et al. 2005).

Integration of anatomical expression patterns into the phylogenic tree of aquaporin family

By integrating the phylogenomic data and anatomical expression patterns consisting from 983 Affymetrix and 209 Agilent 44 K array platforms, we provided an overview of the expression patterns for this gene family (Fig. 2, Fig. S4). Using independent, meta-anatomical expression data from 2 sources, we evaluated the consistency of the anatomical expression patterns of this family.

There are 12 members in the PIP subfamily (Fig. 2), of which OsPIP1;1, OsPIP1;2, and OsPIP2;1 were highly expressed in most organs/tissues, suggesting a role as housekeeping genes. OsPIP2;3 and OsPIP2;5 were more highly expressed in the roots, while OsPIP2;7 was more highly expressed in the leaves and shoots. OsPIP1;3 and OsPIP2;2 were most highly expressed in the during vegetative growth. OsPIP2;6 was more highly expressed in the root, leaf sheath, inflorescence, anther, ovary, pistil, and embryo than in other organs/tissues. Expression in flag leaves, however, was inconsistent between Affymetrix and Agilent 44 K array data. OsPIP2;4 showed peak expression in seedling roots and high of expression in the flag leaf, anther, ovary, and endosperm. The LOC_Os07g26640 expression pattern determined by Agilent 44 K was similar to OsPIP2;4, although Affymetrix array data was unavailable. OsPIP2;8 showed low expression levels in most anatomical samples, except for its peak expression in root samples assessed by Affymetrix arrays.

The NIP subfamily has 10 members (Fig. 2), of which OsNIP1;1 was most highly expressed in the reproductive growth stage by both arrays, while OsNIP1;3 and OsNIP2;2 were most highly expressed in the vegetative growth stage. OsNIP2;1 and OsNIP3;1 were most highly expressed in the root. Affymetrix arrays showed that OsNIP3;2 was most highly expressed in the root. Both arrays showed that OsNIP4;1 was most highly expressed in the developing anther. Agilent 44 K arrays showed that OsNIP3;3 reached peak expression in embryos, while Affymetrix arrays did not show any tissue or organ-preference. Affymetrix arrays showed only low levels of OsNIP1;4 expression.

The TIP subfamily contains 10 members (Fig. 2). OsTIP1;1 and OsTIP2;2 were constitutively expressed in all organs/tissues, suggesting a role as housekeeping genes during development. OsTIP1;1 expression was significantly reduced in pollen, while both arrays showed that OsTIP2;1 was more highly expressed in roots. In contrast, Affymetrix arrays indicated that OsTIP1;2 and OsTIP4;1 were more highly expressed in the leaf and flag leaf. Both OsTIP4;2 and OsTIP4;3 showed increased expression in the developing panicle. Furthermore, OsTIP4;3 was highly expressed throughout panicle development, while OsTIP4;2 was more highly expressed in mature panicles than in earlier stages. OsTIP4;2 and OsTIP4;3 were differentially expressed in leaf blade and sheath or ovary and pistil. Affymetrix arrays showed that OsTIP3;1 and OsTIP3;2 were more highly expressed in the embryo and endosperm, while OsTIP5;1 was more highly expressed in pollen (Agilent 44 K arrays did not contain pollen). OsTIP3;1 and OsTIP5;1 were pollen-preferred, and similarly, there are two pollen specific TIP isoforms in Arabidopsis (Soto et al. 2008). In addition, OsTIP4s were more highly expressed in the leaf or leaf sheath at the seedling stage, while OsTIP3s and OsTIP5 were more highly expressed in reproductive organs, such as the anther, embryo, and endosperm.

The SIP subfamily includes 2 members (Fig. 3), of which OsSIP1;1 was constitutive expression in all organs/tissues, except maturing pollen, and Affymetrix array showed that OsSIP2;1 is expressed in mature pollen, like OsTIP5;1 (Fig. S4). These analyses indicate that OsSIPs functions are differentiated during pollen development (Fig. S4).

In summary, anatomical meta-profiling data integrated into the phylogenetic tree revealed that 8 of the 34 aquaporin genes were more highly expressed in root than in shoot and leaf samples at the seedling stage. We confirmed the expression patterns of these aquaporin genes in seedlings by RT-PCR analysis (Fig. 2). The anatomical meta-profiling data and RT-PCR analyses suggest that OsNIP2;1, OsTIP2;1, and OsPIP2;3 have more of a root preference than the other aquaporins.

Integration of diverse stress-responsive expression data into the aquaporin phylogenic tree

We also measured the expression of rice aquaporin genes in response to abiotic stresses. To accomplish this work, we developed abiotic stress fold-change databases using publicly available Affymetrix array data. In total, the database included fold-change data from 24 comparisons related to abiotic stresses (Table S3). From this database, all rice aquaporin genes except 5 were indicated as displaying upregulation or downregulation in response to drought, cold, heat, and submergence (Fig. 3). Five of these genes were downregulated by drought in seedling root and 4 were upregulated. In the rice leaf under drought stress, 13 genes were downregulated and 1 was upregulated. In response to submergence, the differential expression patterns were similar to those in rice leaf under drought stress. Heat stress in seedlings triggered the opposite expression pattern (Fig. 3). Three genes were downregulated and 5 were upregulated in seedlings under salinity stress. In cold stress, only 1 gene (OsTIP1;2) was markedly downregulated in seedlings. Interestingly, OsTIP1;2 showed opposite expression patterns in the root and leaf in response to drought. In addition, OsTIP3;1 was markedly upregulated in all abiotic stresses in this study, except for cold, suggesting that this gene may regulate multiple abiotic stress responses. The drought response of aquaporin genes in root strongly overlapped with the salt response in rice seedlings, indicating that salinity at the root may make a larger contribution to differential expression patterns than salinity at the leaf. In response to drought in the root and salinity in seedlings OsNIP2;2 was downregulated, but OsTIP3;3, OsTIP3;1, OsTIP4;2 and OsTIP1;2 were commonly upregulated.

In response to abiotic stress response, OsNIP expression was negatively related to drought in root, OsTIP expression was positively associated with drought in root and salinity in seedlings, OsPIP expression was negatively associated with drought and submergence in seedlings but positively associated with heat in seedlings, and OsSIP expression was negatively associated with drought and submergence in seedlings.

Since aquaporins in roots are important in taking up water from the soil, we investigated their response under drought conditions, especially in the root (Boursiac et al. 2008; Postaire et al. 2010). We identified 9 candidate genes that showed markedly different expression in response to drought stress in root (Fig. 3). Among these genes, 5 (OsNIP2;2, OsNIP3;1, OsPIP2;5, OsPIP2;7, OsSIP1;1) were downregulated and 4 (OsTIP1;2, OsTIP3;1, OsTIP3;2, OsTIP4;1) were upregulated. This result clearly indicates that OsTIP expression is positively associated with drought in root, while expression of other subfamilies is negatively associated with this stress response.

Confirmation of expression of significant aquaporin genes in root by RT-PCR and real-time PCR after drought stress

We confirmed the expression patterns of the 9 aquaporin genes by RT-PCR and real-time PCR (Fig. 4). Seven of the 9 genes were consistent with the microarray data in Fig. 3. OsTIP1;2, OsTIP3;1, OsTIP3;2, and OsTIP4;1 were upregulated, while OsNIP2;2, OsNIP3;1, and OsSIP1;1 were downregulated in response to drought in root (Fig. 4).

Development and validation of a functional network of genes mediated by drought-responsible aquaporin genes

We developed a functional network of the 5 rice aquaporin genes differentially regulated by drought stress in root using RiceNet (Lee et al. 2011). The network contains 16 interactions mediated by 5 aquaporin proteins (Fig. 5). The differential expression in response to drought stress in root was then integrated into the network (Fig. 5). Significant changes in expression, at least twofold upregulation or downregulation, in root under drought stress were selected and displayed in green (downregulation) or red (upregulation) nodes: 3 genes (OsNIP2;2, OsNIP3;1, and OsSIP1;1) were downregulated and 2 (OsTIP3;1 and OsTIP3;2) were upregulated. Five of 10 genes in the predicted functional gene network showed greater than twofold upregulation by drought stress in root. These genes include oleosin (LOC_Os03g49190), heat shock protein 90s (HSP90s, LOC_Os09g30412 and LOC_Os09g30418), peroxiredoxin (LOC_Os07g44430), and DUF1264 domain protein (LOC_Os05g49440). The expression of the 5 aquaporin genes in the network was confirmed by RT-PCR and real-time PCR (Fig. 5). We also re-analyzed the expression of 5 genes that were significantly upregulated in response to drought in root (Fig. 5). Finally, we refined our network by developing functional networks of genes mediated by 2 aquaporin genes (OsTIP3;1 and OsTIP3;2) associated with oleosin, DUF1264, and peroxiredoxin and 3 aquaporin genes (OsSIP1;1, OsNIP2;2, and OsNIP3;1) associated with 2 HSP90 genes (Fig. 5).


Estimation of functional redundancy among aquaporin genes

From the physical map, we found 3 clusters of tandem duplicated aquaporin genes. The first cluster, on chromosome 1, contained 2 genes (OsTIP4;2 and OsTIP4;3); the second cluster was on chromosome 7 and contained 4 genes (OsPIP2;1, OsPIP2;4, OsPIP2;5, and LOC_Os07g26640); and the third cluster was on chromosome 8 and contained 2 genes (OsNIP3;2, and OsNIP3;3). Functional redundancy is more likely following gene duplication, with tandem duplication as the simplest gene duplication (Mendonca et al. 2011). As functional redundancy is caused by duplicated genes with duplicated functionality, we previously evaluated the functionality of duplicated genes by integrating gene expression data in response to light and performing a functional analysis of the light response among duplicated genes using related gene-indexed mutants. Duplicated genes with similar expression patterns were prone to functional redundancy (Jung et al. 2008). In addition, AtPIP2 genes are highly homologous, suggesting functional redundancy. However, AtPIP2 genes show differential expression, as well as overlapping expression, in promoter::GUS transgenic cell lines, suggesting gene-specific and non-redundant functions (Javot et al. 2003). Although the AtPIP1 genes are highly homologous, a single knock-out phenotype for the PIP1;2 gene was identified (Postaire et al. 2010; Heckwolf et al. 2011). We expect that PIP1;2 has a dominant role among related subfamily members because it is one of the most abundantly expressed PIP isoforms (Postaire et al. 2010).

Phylogenomic data integrating anatomical expression patterns into the phylogenetic tree allow us to more precisely evaluate functional redundancy among aquaporin genes. On the basis of our anatomical meta-profiles and phylogenomic data, we expect that OsTIP3;1 and OsTIP3;2 may be functionally redundant during embryo and endosperm development; OsTIP1;1, OsTIP2;2, OsPIP1;1, and OsPIP1;2 may have conserved functions as housekeeping genes; and OsPIP2;4 and LOC_Os07g26640 may be functionally redundant in various tissues/organs, including the seedling root, which has the highest expression levels (Fig. S4).

We further evaluated the functional redundancy of the tandem duplicated genes indicated in Fig. S1. Although OsTIP4;2 and OsTIP4;3 showed similar expression in developing panicle, their expression differed in leaf blade and sheath or pistil and ovary, suggesting they have unique roles in these tissues. All genes in the OsPIP2;1, OsPIP2;4, OsPIP2;5, and LOC_Os07g26640 cluster, except for OsPIP2;1, showed similar expression in the root, while OsPIP2;1 was constitutively expressed in all tissues/organs except for pollen. This observation suggests a functional convergence during root development and related stress responses. These results also indicate that the former are functionally redundant in the root, and the latter is a housekeeping gene in all tissues except for pollen.

In addition, we identified differences in OsPIP2;4 and OsPIP2;5 expression in developing panicles and floral organs, suggesting functional divergence between the aquaporins in these tissues/organs (Fig. S4). The 2 genes on chromosome 8 (OsNIP3;2, and OsNIP3;3) had conserved expression in pollen and embryo, while functional redundancy in other tissues/organs remains unclear (Fig. S4). This understanding of functional redundancy among aquaporin genes will be useful in establishing an experimental plan for further functional analysis. To test aquaporin genes with predicted functional redundancy, RNA silencing for a gene family (Miki et al. 2005) and artificial miRNA targeting redundant aquaporin genes (Liang et al. 2012) may be necessary.

Functional implication of rice aquaporin genes using orthologs in Arabidopsis

Information about orthologs of rice aquaporins was incorporated into the phylogenetic tree (Fig. S3). Known functions of the orthologs are good references for a functional analysis of aquaporins belonging to the same subfamily. For example, AtPIP1;4 and AtPIP2;5 control water-loss under dehydration stress (Jang et al. 2007); AtPIP2;2 is involved in water uptake by roots (Javot et al. 2003); and AtNIP1;1 determines arsenite sensitivity in plants (Kamiya et al. 2009). A functional analysis of rice orthologs might clarify gene functions.

In addition, we assigned a gene ontology (GO) term in the cellular component category to all aquaporins (Fig. S3). All aquaporins have GO terms related to membranes, supporting their general role as water transporters (Maurel 1997). Each subfamily has a somewhat different subcellular localization: PIPs are targeted to the plasma membrane (Johanson et al. 2001), TIPs to vacuolar membranes (Maurel et al. 2002), NIPs to the symbiosome membrane of nitrogen-fixing soybean nodules and plasma membrane in several plant species (Weaver et al. 1994; Ma et al. 2006; Yamaji et al. 2008; Yamaji and Ma 2009; Pang et al. 2010), and SIPs to the endoplasmic reticulum (ER) (Ishikawa et al. 2005). In cellular component GO assignment from RGAP database (, some OsTIPs and OsPIPs are predicted to be localized to vacuoles, plastids, and 2 SIPs localized to the ER (Fig. S3). Although subcellular localizations of some aquaporins remain to be clarified, these results indicate that most subcellular organelles in rice may have dedicated aquaporin proteins for intracellular transport.

Evaluation and usefulness of integrating aquaporin expression patterns into the phylogenic tree

By assessing expression data from 2 independent, meta-anatomical assays, we evaluated the consistency of anatomical expression patterns for the aquaporins and compared the results with previous research. In the PIP subfamily, OsPIP2 genes were primarily expressed in the root at the seedling stage, which agrees with the data from Lian et al. (2006). Sakurai et al. (2005) also reported higher OsPIP2;3 and OSPIP2;5 expression in roots than leaf blades. In Arabidopsis, AtPIP1;1, AtPIP1;3, AtPIP2;2, AtPIP2;3, AtPIP2;4, AtPIP2;7, and AtPIP2;8 were more abundantly expressed in the roots than aerial segments (Jang et al. 2004). These results suggest that OsPIP2 genes may play significant roles in root development and the response to environmental challenges, including water deficiency.

OsNIP1;1 was also highly expressed in the shoot in the Affymetrix array and in the leaf sheath in the Agilent 44 K array. Expression in the flag leaf, however, was inconsistent between the Affymetrix and Agilent 44 K arrays. Interestingly, this differential expression may be due to differences in the genetic backgrounds of the flag leaves used in the Affymetrix and Agilent 44 K arrays. The Affymetrix samples were prepared from both japonica and indica cultivars, while the Agilent 44 K samples were prepared from the Nipponbare cultivar, a japonica variety. Figure S4 shows the expression of all samples in the Affymetrix array platform, suggesting that expression was far greater in leaves and flag leaves of the indica cultivar than the japonica. Thus, the aquaporin gene may be a biomarker to determine evolutionary differences between the japonica and indica cultivars.

Anatomical expression of OsNIP2;1 and OsNIP3;1 was also determined by analyzing the expression of 33 aquaporin genes in rice (Sakurai et al. 2005). OsNIP3;1 expression was stimulated and OsNIP2;1 expression was suppressed, in the shoot apical meristem (SAM). OsNIP3;1 expression was inconsistent with Sakurai et al. (2005). Since OsNIP2;1 and OsNIP3;1 retain high expression in root (Fig. S4), the expression patterns are more reproducible. OsNIP3;1 expression, on the other hand, was low, and we only identified high expression in several root samples (Fig. S4). Reproducing expression patterns of genes expressed at a low level or those that are conditionally regulated is typically challenging. OsNIP1;4 expression suggested a possible disruption of gene function during evolution. Comparing DNA methylation, histone acetylation, and miRNA analyses may explain the reasons for the silenced expression.

In Arabidopsis, most NIP genes had transcript levels close to or below the detection limit of the microarray experiments. Expression of several NIP genes was identified: AtNIP1;1 showed root-preferred expression like OsNIP2;1 (Weig et al. 1997); and AtNIP4;1, AtNIP4;2, and AtNIP7;1 showed flower-preferred expression (Alexandersson et al. 2005). AtNIPs transport a diverse range of substrates, such as silicic acid, boric acid, lactic acid, urea, and formamide, depending on the protein sequence (Mitani-Ueno et al. 2011). Although OsNIP2;1 and OsNIP2;2 have been identified as silicon transporters (Ma et al. 2006), the substrate specificity of the other OsNIPs should be investigated.

OsTIP4s expression was confirmed by expression analysis of 6 OsTIPs in the roots and shoots (Li et al. 2008a). These results suggest that OsTIP4s in leaf or leaf sheath development and OsTIP3s and OsTIP5 in reproductive organs evolved to respond to environmental challenges, including water deficiency. Because 2 OsSIPs are expressed in all tissues/organs, we expect that their cell-specific functions are independent from other aquaporins with different cellular sub-localizations (Ishikawa et al. 2005).

Development and validation of a functional network of genes mediated by drought-responsible aquaporins

The refined network consists of 5 interactions [oleosin (LOC_Os03g49190), heat shock protein 90 s (HSP90s, LOC_Os09g30412 and LOC_Os09g30418), peroxiredoxin (LOC_Os07g44430), and DUF1264 domain of unknown function (LOC_Os05g49440)] mediated by 5 aquaporin proteins (OsNIP2;2, OsNIP3;1, OsSIP1;1, OsTIP3;1, and OsTIP3;2). HSP70 interaction with AQUAPORIN-2 (AQP2) was identified by direct binding assays using purified human HSP70 and aquaporin 2 C-terminal fusion protein (Lu et al. 2007). Hsp70–Hsp90 organizing protein (HOP) modulates HSP70/HSP90 protein folding interactions (Johnson et al. 1998). These results support a possible association between aquaporin and HSP90 proteins. Aquaporin-3 can facilitate the uptake of H2O2 metabolized by enzymes such as catalase, glutathione peroxidase, and peroxiredoxin in mammalian cells to mediate downstream intracellular signaling (Miller et al. 2010; Finkel 2011), suggesting a possible association between aquaporins and peroxiredoxin. Oil bodies from Arabidopsis thaliana retain a limited number of proteins, including 4 different oleosins and a probable aquaporin (Jolivet et al. 2004), suggesting a possible linkage between oleosin and aquaporin in oil bodies. The possible relationships between aquaporins and oleosin, peroxiredoxin, and HSP90 in response to drought in root have already been investigated (Rizhsky et al. 2004; Torres et al. 2007; Hayano-Kanashiro et al. 2009; van der Schoot et al. 2011), but the biological significance of their associations had not been determined. Therefore, the network developed in our study will help shed light on the molecular mechanisms involved through further functional analysis.


This work was supported by the Next-Generation BioGreen 21 Program of South Korea (PJ008079 and PJ008173) to KHJ and a Young Scientist Program through the National Research Foundation of Korea (Grant no. 20120003801) to KHJ.

Supplementary material

425_2013_1918_MOESM1_ESM.jpg (1.3 mb)
Fig. S1 Chromosomal distribution of 34 rice aquaporin genes. Red boxes denote a new member (JPEG 1311 kb)
425_2013_1918_MOESM2_ESM.jpg (1.2 mb)
Fig. S2 The phylogenetic tree of rice aquaporin proteins (JPEG 1231 kb)
425_2013_1918_MOESM3_ESM.eps (3.7 mb)
Fig. S3 The phylogenomic data of the rice aquaporins with orthologs from Arabidopsis and maize (EPS 3828 kb)
425_2013_1918_MOESM4_ESM.jpg (1.6 mb)
Fig. S4 Anatomical expression of rice aquaporin genes using 1150 Affymetrix and 209 Agilent 44 K arrays. A black bar denotes an indica cultivar and a grey bar a japonica cultivar (JPEG 1598 kb)
425_2013_1918_MOESM5_ESM.xlsx (13 kb)
Table S1 Detailed information for 34 rice aquaporin genes (XLSX 12 kb)
425_2013_1918_MOESM6_ESM.docx (19 kb)
Table S2 Primer sequences for RT-PCR and real-time PCR of aquaporin genes and genes functionally associated with aquaporins, used in Figs. 2, 4, and 5 (DOCX 19 kb)
425_2013_1918_MOESM7_ESM.xlsx (34 kb)
Table S3 Detailed comparison of microarrays used for Fig. 3 (XLSX 34 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2013