Plant Physiol, March 2001, Vol. 125, pp. 1206-1215

Aquaporins Constitute a Large and Highly Divergent Protein Family in Maize1

Franšois Chaumont, Franšois Barrieu,2 Eva Wojcik, Maarten J. Chrispeels,* and Rudolf Jung Physiological Biochemistry, UniversitÚ Catholique de Louvain, B-1348 Louvain-La-Neuve, Belgium (F.C.); Division of Biology, University of California, San Diego, California 92093-0116 (F.B., M.J.C.); and Pioneer Hi-Bred International, Incorporated, 7300 Northwest 62nd Avenue, Johnston, Iowa 50131-1004 (E.W., R.J.)


Aquaporins (AQPs) are an ancient family of channel proteins that transport water and neutral solutes through a pore and are found in all eukaryotes and most prokaryotes. A comparison of the amino acid sequences and phylogenetic analysis of 31 full-length cDNAs of maize (Zea mays) AQPs shows that they comprise four different groups of highly divergent proteins. We have classified them as plasma membrane intinsic proteins (PIPs), tonoplast intrinsic proteins, Nod26-like intrinsic proteins, and small and basic intrinsic proteins. Amino acid sequence identities vary from 16% to 100%, but all sequences share structural motifs and conserved amino acids necessary to stabilize the two loops that form the aqueous pore. Most divergent are the small and basic integral proteins in which the first of the two highly conserved Asn-Pro-Ala motifs of the pore is not conserved, but is represented by alanine-proline-threonine or alanine-proline-serine. We present a model of ZmPIP1-2 based on the three-dimensional structure of mammalian AQP1. Tabulation of the number of times that the AQP sequences are found in a collection of databases that comprises about 470,000 maize cDNAs indicates that a few of the maize AQPs are very highly expressed and many are not abundantly expressed. The phylogenetic analysis supports the interpretation that the divergence of PIPs through gene duplication occurred more recently than the divergence of the members of the other three subfamilies. This study opens the way to analyze the function of the proteins in Xenopus laevis oocytes, determine the tissue specific expression of the genes, recover insertion mutants, and determine the in planta function.


Aquaporins (AQPs) are an ancient family of channel proteins that transport water and certain neutral metabolites across biological membranes. Many among them regulate the hydraulic conductivity of the membranes in which they reside and potentiate a 10- to 20-fold increase in the water permeability coefficient (Pf) of those membranes. The most active are highly specific for water and may transport up to a billion water molecules per second per 28-kD protein subunit, depending on the osmotic gradient imposed. Some members of this family transport glycerol as well as water, whereas other members of the family found in bacteria and yeast (Saccharomyces cerevisiae) transport only glycerol and neutral solutes.

Since their discovery in plants (Maurel et al., 1993), the properties of these proteins, the genes that encode them, and their potential roles in plant-water relations and intra- and intercellular water transport have been intensely studied and the results have been reviewed several times in the last few years (Maurel, 1997; Kjellbom et al., 1999; Tyerman et al., 1999; Johansson et al., 2000; Maurel and Chrispeels, 2001). In plants, AQPs are present in the tonoplast, the plasma membrane, and possibly in other internal membranes (Barkla et al., 1999). An analysis of the Arabidopsis genome shows that there are 35 different AQPs grouped into four subfamilies (Weig et al., 1997, and subsequent analysis by us and independently by U. Johanson and P. Kjellbom). One subfamily corresponds to tonoplast proteins and a second one to plasma membrane proteins, but the subcellular location of the others is still uncertain.

The 250 to 300 amino acids of AQP monomers form two tandem repeats of three membrane-spanning domains each. Structural analysis of crystals of mammalian AQPs and amino acid sequence comparisons of all AQPs show that they have six membrane-spanning alpha helices with N and C termini that face the cytosol. The cytosolic loop (loop B) between the second and third transmembrane domain and the extra-cytosolic loop (loop E) between the fifth and sixth transmembrane domain also form short helices that are relatively hydrophobic and dip into the membrane from opposite sides (Fig. 1) These two loops contain conserved Asn-Pro-Ala (NPA) motifs and the two Asn residues participate in forming an aqueous channel that is 3 ┼ at its narrowest point (Mitsuoka et al., 1999; Murata et al., 2000). The structural basis for the transport of glycerol, a molecule much larger than water, has also been elucidated (Fu et al., 2000). Glycerol molecules move in a single file through an equally narrow amphipathic channel in which the NPA motifs also play a critical role.

Figure 1. Model of the structure of an AQP showing the principal features of the protein. Alpha helices are represented as rectangles. There are six transmembrane domains (TM1-TM6) connected by five loops (A-E). Two helical domains (HB and HE) in different loops dip halfway into the membrane from opposite sides and form the aqueous pore. Loops B and E also contain the highly conserved NPA motifs that are part of the pore. In the three-dimensional structure (see Fig. 7), these two motifs are positioned one above the other.


To gain further insight into the possible physiological functions of the members of this large family, we need detailed expression analysis of the genes. This can be most easily done after we have surveyed the complexity of the family in a few plant species. Here we present an analysis of the AQP family of maize (Zea mays, abbreviated to Zm in the names of genes and proteins) based on complete sequences of the cDNAs obtained after sequencing a set of unique expressed sequence tags (ESTs) culled from a number of maize EST libraries present in the DuPont/Pioneer Hi-Bred database. Our results show that AQPs form a highly divergent gene family in maize with four subgroups and that some members are highly expressed, whereas many others are much less frequently encountered in the database. We present a model for the structure of maize PIP1-2 that is based on the recently published structure of mammalian AQP1 and discuss the evolution of this protein family.

By screening about 470,000 maize ESTs representing 215 maize cDNA libraries, we identified more than 1,300 different accessions of AQP gene family members. The longest clones representing unique sequences were obtained from the libraries and sequenced in their entirety. This resulted in 31 different complete nucleotide sequences representing 30 amino acid sequences (ZmPIP1-3 and ZmPIP1-4 present in the B73 inbred line have different nucleotide sequences encoding the same protein). These 31 sequences could be grouped in four subfamilies referred to as plasma membrane intrinsic proteins (PIPs), tonoplast intrinsic proteins (TIPs), Nod26-like intrinsic proteins (NIPs), and small and basic intrinsic proteins (SIPs; Table I). Some of these names are used for historical reasons and the names PIP and TIP are used even though all the members of this subfamily may not be located in the plasma membrane and tonoplast, respectively (Barkla et al., 1999). All sequences have six putative transmembrane heliceshere called TM1 through TM6and most have the double NPA motif in two of the loops (B and E loops) connecting the domains. The proteins vary in length from 243 to 302 amino acids.

Table I. AQPs of maize



Accession No. No. of Amino Acids
ZmPIP1-1 (ZmPIP1a) X82633 288
ZmPIP1-2 (ZmPIP1b) AF131201 289
  ZmPIP1-3 AF326487 292
  ZmPIP1-4 AF326488 292
  ZmPIP1-5 AF326489 288
  ZmPIP1-6 AF326490 296
  ZmPIP2-1 AF326491 290
  ZmPIP2-2 AF326492 292
  ZmPIP2-3 AF326493 289
  ZmPIP2-4 AF326494 288
ZmPIP2-5 (ZmPIP2a) AF130975 290
  ZmPIP2-6 AF326495 288
  ZmPIP2-7 AF326496 287
ZmTIP1-1 (ZmTIP1) AF037061 250
  ZmTIP1-2 AF326500 254
  ZmTIP2-1 AF326501 249
  ZmTIP2-2 AF326502 250
  ZmTIP2-3 AF326503 247
  ZmTIP3-1 AF326504 262
  ZmTIP4-1 AF326505 255
  ZmTIP4-2 AF326506 257
  ZmTIP4-3 AF326507 249
  ZmTIP4-4 AF326508 252
  ZmTIP5-1 AF326509 260
  ZmSIP1-1 AF326497 245
  ZmSIP1-2 AF326498 243
  ZmSIP2-1 AF326499 249
  ZmNIP1-1 AF326483 282
  ZmNIP2-1 AF326484 294
  ZmNIP2-2 AF326485 295
  ZmNIP3-1 AF326486 302

After the revision of this paper was accepted, we succeeded in cloning full-length cDNAs of two other maize AQPs. Their sequences have been submitted to GenBank and are: ZmTIP3-2 (accession no. AF342809) and ZmNIP2-3 (accession no. A342810). The nos. reflect their sequence identity to existing AQPs on a cladogram. The previous published names are in parentheses (Chaumont et al., 1998, 2000).  These are the only two maize AQPs that have been shown to have good water channel activity in oocytes. ZmPIP1-1 and ZmPIP1-2 were inactive and the others have been tested.    


Phylogenetic analysis clearly shows the presence of four subfamilies (Fig. 2). The length of the branches is an indication of the amino acid sequence relatedness. The PIP proteins are most closely related to each other and have 64% to 100% identity. However, their relatedness to the other three groups is much less: Only 16% to 35% of the amino acids are conserved with members of the other groups (Table II). We were able to identify 68 different amino acid positions that are conserved in 20 out of the 31 different maize AQPs, or nearly 25% of the total (see below).

Figure 2. Phylogenetic analysis of 31 maize AQP proteins. The distance scale represents the evolutionary distance, expressed in the number of substitutions per amino acid. National Center for Biotechnology Information accession numbers are shown in Table I.


Table II.   Percentage of sequence identity between the different subfamily members



TIPs 36-94 - - -
PIPs 25-35 64-100 - -
NIPs 23-34 21-28 39-91 -
SIPs 17-28 16-24 16-21 30-78


AQP genes have been identified in many land plant species (Johansson et al., 2000). We conducted a phylogenetic analysis that included a number of AQPs from other plants to determine how the maize sequences relate to sequences of other proteins whose properties have already been studied.

Only four of these proteins have been examined for water channel activity in oocytes. ZmTIP1-1 (Chaumont et al., 1998) and Zm PIP2-5 (Chaumont et al., 2000) have high activity, whereas ZmPIP1-1 and ZmPIP1-2 are inactive in the oocyte assay. The activities of all others remain to be studied.

Phylogenetic Analysis of ZmPIPs

The PIP subfamily differs from the TIP subfamily by the presence of an additional 20 to 38 amino acids at the N terminus of the PIP proteins. In addition, there are a large number of amino acid positions (142, or about 50%) that are conserved in all PIPs. The PIPs can be divided into two major groups, referred to as PIP1 and PIP2 (Fig. 3), in accordance with the work of Kammerloher et al. (1994) and others subsequently. All PIP2 proteins examined for water channel activity in Xenopus laevis oocytes show good activity, but PIP1 proteins are often inactive in oocytes (Chaumont et al., 2000). The reason for this difference is not known. Some other AQPs such as soybean (Glycine max) Nod26 and mammalian AQP0 have very low activity in the X. laevis oocyte assay. PIP2 proteins are characterized by a shorter N-terminal extension than PIP1 proteins and a longer C-terminal end that contains putative phosphorylation sites (Schńffner, 1998; Chaumont et al., 2000; Johansson et al., 2000). In addition to several conservative amino acid substitutions between PIP1s and PIP2s, some positions show single nonconservative exchange associated with each subgroup (i.e. Gly 56 Val 44 in TM1, Gln 90 Leu 86 in TM2, and Met 140 Ala136 in TM3 in ZmPIP1-1 and ZmPIP2-1, respectively).

Figure 3. Phylogenetic analysis of the maize ZmPIPs and other plant PIPs. Accession numbers of ZmPIPs are shown in Table I. Other plant PIP accession numbers are indicated in the tree or (in parentheses): NtAQP1 (AJ001416), AtPIP1c (AAF81320), AtPIP1b (AAC28529), AtPIP1a (CAB71073), AtRD28 (AAD18141), AtPIP2a (CAB67649), and AtPIP3 (CAA17774). The distance scale represents the evolutionary distance, expressed in the number of substitutions per amino acid.

The phylogenetic tree of PIPs indicates that the multiplicity of most maize and Arabidopsis PIPs must have emerged relatively late during evolution, after the monocot-dicot divide. Five of the six maize PIP1 sequences cluster in one branch, separate from all Arabidopsis PIP1 sequences. ZmPIP1-6 forms a third PIP1 branch. In a similar manner, six of the seven maize PIP2 sequences cluster in one group. ZmPIP2-7 forms a branch with two Arabidopsis sequences (AtPIP3 and AAC64216.1), and all other 11 AtPIPs cluster in two groups, apart from the maize sequences (Fig. 3; for simplicity of representation, only selected Arabidopsis sequences are shown in the tree).

Phylogenetic Analysis of ZmTIPs
The ZmTIP cladogram (Fig. 4) shows that TIPs can be divided into five groups. TIP1 corresponds to the highly expressed and active -TIPs found in many plants (Maurel et al., 1993; Chaumont et al., 1998). TIP2 corresponds to -TIP of Arabidopsis (Daniels et al., 1996). Vacuoles containing -TIP proteins may act as storage compartments for pigments and vegetative storage proteins (Jauh et al., 1998). TIP3 corresponds to -TIP, first found in the common bean (Phaseolus vulgaris) and highly expressed in cotyledons where it is a component of the membrane that delimits the protein storage vacuole (Johnson et al., 1990). TIP4 represents a family that also contains NtTIPa, a protein that transports water and glycerol in X. laevis oocytes (Gerbeau et al., 1999). A closely related sequence is found in the Arabidopsis database (Fig. 4). The TIP5 group includes a not-yet-characterized Arabidopsis protein and two proteins from maize and barley (Hordeum vulgare), respectively. ZmTIP5-1 is characterized by an eight-amino acid residue insertion in the third loop. In contrast to PIPs, the TIP cladogram (Fig. 4) reveals monocot and dicot sequences clustered together in each of the five major TIP branches.

Figure 4. Phylogenetic analysis of the maize ZmTIPs and other plant TIPs. Accession numbers of ZmTIPs are shown in Table I. Other plant TIP accession numbers are indicated in the tree or (in parentheses): At-TIP (AAF18716), AtTIP (AAF97261), AtTIP (BAB1264), AtTIP (AAD31569), and NtTIPa (CAB40742).

Phylogenetic Analysis of ZmNIPs

The Nod26-like major intrinsic protein (MIP) subfamily (NIP) is represented by four members in maize (Fig. 5). One of these (ZmNIP1-1) is most closely related to four other proteins (GmNod26, LjLIMP2, AtNLM1, and AtNLM2) that have been found to transport glycerol as well as water (Rivers et al., 1997; Dean et al., 1997; Guenther and Roberts, 2000; Weig and Jakob, 2000). Two of the other sequences (ZmNIP2-1 and ZmNIP2-2) are closely related and may be the result of a more recent duplication. The NIPs and PIPs are the longest proteins, but NIPs differ mainly from PIPs by the presence of longer C-terminal tails (eight-30 amino acid residues) that are highly charged in the case of ZmNIP2-1 (16 out of 41 amino acids) and ZmNIP2-2 (12 out of 35 amino acids).

Figure 5. Phylogenetic analysis of the maize ZmNIPs and other plant NIPs. Accession numbers of ZmNIPs are shown in Table I. Other plant NIP accession numbers are indicated in the tree or (in parentheses): LjLIMP2 (AAF82791), GmNOD26 (AAA02946), AtNLM1 (CAA16760), and AtNLM2 (CAB78893). The distance scale represents the evolutionary distance, expressed in the number of substitutions per amino acid.

On cladograms that include water-specific AQPs as well as glycerol transporters from bacteria, yeast, and mammals, all the sequences cluster into two groups: an AQP cluster (true AQPs) and a glycerol facilitator-like protein (GLP) cluster (Park and Saier, 1996; Heymann and Engel, 1999). A similar cladogram that includes the plant AQPs that transport glycerol, whether in the PIP, TIP, or NIP cluster, shows that these glycerol transporters are grouped with the AQP cluster. Froger et al. (1998) identified five conserved amino acid positions that differ consistently between the two groups. However, in the NIP glycerol transporters of plants only two out of these five positions follow this rule (Guenther and Roberts, 2000; Weig and Jakob, 2000). Also, in ZmNIP1-1 and ZmNIP3-1, aromatic (Phe) and aliphatic (Leu/Val) residues are found in positions P1 and P5, respectively, in accordance with the glycerol transporter rule identified by Froger et al. (1998). However, ZmNIP2-1 and ZmNIP2-2 have residues that are typical of orthodox AQP. As seen in the cladogram (Fig. 5), maize NIP1 and NIP3 sequences have counterparts in dicot species. No dicot orthologs for the NIP2 have been detected in the databases, including the entire Arabidopsis genomic sequence. However, it is interesting that a close NIP homolog that clusters in a cladogram to the NIP2 branch sequences has been reported from the fern Adiantum capillus-veneris (accession no. BAB12437).

Phylogenetic Analysis of ZmSIPs

The SIPs constitute a new small subfamily that has been recently identified in Arabidopsis (U. Johanson and P. Kjellbom, personal communication at the MIP 2000 meeting in G÷teborg, Sweden, July 2000) and in maize by one of us (R. Jung). The amino acid sequence of this group is the most highly diverged, showing only 16% to 28% identity with the three other groups (Table II). Not only is there a general divergence over the entire sequence, but there is a striking lack of conservation in the short helix of loop B, which contains the first NPA motif (Fig. 6). This motif is represented by Asn-Pro-Thr or Asn-Pro-Leu in the ZmSIP sequences. In Arabidopsis, the third position can be occupied by Thr, Cys, or Leu. The second NPA motif is conserved in all maize and Arabidopsis AQPs. Overall in helix B, there is complete conservation of 6 other amino acids in all the ZmTIPs and ZmPIPs: SGGHXNPAVT. Of these six positions, only one is conserved in the ZmSIPs. Therefore, it is likely that ZmSIPs evolved separately from an ancestral gene. The ZmNIPs also show some divergence in this highly conserved loop (Fig. 6). The SIP cladogram (Fig. 7) shows that SIPs can be divided into two groups, each of them including the same number of maize and Arabidopsis proteins.

Figure 6. Amino acid residues in the NPA motifs of ZmAQPs. The amino acid residues in the structural loops B and E of each maize subfamily are indicated. Residues in bold are found in 20 or more of 31 ZmAQPs.


Figure 7. Phylogenetic analysis of the maize and Arabidopsis SIPs. Accession numbers of ZmSIPs are shown in Table I. Arabidopsis accession numbers are indicated in the tree. The distance scale represents the evolutionary distance, expressed in the number of substitutions per amino acid.

Conserved Motifs and Amino Acids

The AQP monomers of maize contain 243 to 302 amino acid residues that form two tandem repeats of three membrane-spanning -helices (TM1-TM6) and with amino and carboxy termini located on the cytoplasmic side of the membrane. Analysis of the crystal structures of mammalian AQP1 (Murata et al., 2000) and bacterial GlpF (Fu et al., 2000) shows that portions of two loops (loop B and loop E) form -helices that dip halfway into the membrane from opposite sides and form the aqueous pore. Figure 8 shows a model of ZmPIP1-2 that is based on the structure of AQP1 (Mitsuoka et al., 1999; Murata et al., 2000) and on an analysis of putative transmembrane domains based on a hydrophobicity plot. The amino acids that are conserved in 20 or more of the 31 maize AQPs are indicated in color. The highest degree of conservation is in the transmembrane domains and more particularly in the two loops that form the aqueous pore and contain the two NPA motifs.

Figure 8. A topological model of the maize ZmPIP1-2. The representation is based on the human AQP1 and bacterial GlpF structures (Fu et al., 2000; Murata et al., 2000), and shows the six transmembrane helices (TM1-TM6) and the two short helices in the structural loops B and E (HB and HE). Residues in yellow with thick red circles are highly conserved among ZmAQPs (found in 20 or more of the 31 ZmAQPs). Residues in pink indicate the position of a highly conserved residue present in 20 or more of the 31 ZmAQPs but absent from ZmPIP1-2 (97Ser and 140Met of ZmPIP1-2 are replaced by an Ala).

A number of amino acid positions are conserved in the two tandem repeats of the protein. These include Glu 60 and 184 in TM1 and TM4, Thr 64 and 188 in TM1 and TM4, Gly 145 and 264 in TM3 and TM6, and Gly 113 and 234 in loops B and E, respectively. In addition to these residues, a large number of amino acid positions are conserved either in the first half or the second half of the protein. Many of the positions conserved in maize AQPs are also conserved in other AQPs (Heymann and Engel, 2000; Murata et al., 2000). Maize has some conserved positions not conserved in other AQPs; For example, Arg 55 and Ala 56 in TM1 are highly conserved, as are Ala 94 and Phe 102 in TM2; Leu 144 in TM3; Phe 189 in TM4; and Ala 214, Leu 216, and Leu 227 in TM5.

In mammalian AQP1, the positions of the two functional loops that form the aqueous pore are stabilized through ion pairs and hydrogen bonds of highly conserved amino acids. These are also conserved in the maize AQPs and on the basis of the structure of AQP1 (Mitsuoka et al., 1999, 2000) we can predict the following interactions. His 115 in loop B forms an ion pair with Glu 60 in TM1 and Arg 241 in HE is connected by a salt bridge to Glu 184 in TM4. Ser 112 in loop B forms a hydrogen bond with Tyr 137 (TM3), further stabilizing loop B.

Expression of Maize AQP Genes

Because we used so many different libraries prepared at different times from plants grown under slightly different conditions, the number of times a specific cDNA appears gives only a rough estimate of its abundance in the mRNA population. Most abundantly expressed are ZmTIP1-1 (a -TIP-like sequence), ZmTIP2-1, most of the members of the ZmPIP1 family, and ZmPIP2-1 (Table III). Ten of the sequences were found only a few times (between one-10 times) and another eight were found less than 20 times. These include all the NIPs and SIPs, which were also the last groups of AQPs to be identified in plants. Most of the more abundantly expressed sequences were found in the various plant organs that were examined. The tissue distribution of the 124 ESTs of ZmPIP1-3 and ZmPIP1-4 is shown in combined numbers (Table III). Because of their very close relationship (approximately 98% identity of the nucleotide sequence), only one single cluster was formed by the ESTs of these two PIPs. Using computer algorithms, it was not possible to faithfully deconvolute this cluster. However, by visual inspection of a representative sample set of high-quality ESTs and by considering signature sequences, we can conclude that about 25% of the ESTs appear to represent ZmPIP1-3 and 75% represent ZmPIP1-4. The distribution of both cDNAs did not appear to differ significantly.

Table III. Distribution of cDNAs in maize libraries


Name TCa RTb Root Vc Sum

ZmPIP1-1 17 65 38 78 198
ZmPIP1-2 3 45 9 26 83
ZmPIP1-3 8 66 12 38 124
ZmPIP1-5 1 0 18 6 25
ZmPIP1-6 0 0 0 1 1
ZmPIP2-1 29 99 17 56 201
ZmPIP2-2 2 24 9 8 43
ZmPIP2-3 1 11 8 9 29
ZmPIP2-4 2 0 27 9 38
ZmPIP2-5 1 2 8 6 17
ZmPIP2-6 12 6 9 12 39
ZmPIP2-7 0 0 0 3 3
ZmTIP1-1 29 114 27 66 236
ZmTIP1-2 0 6 0 3 9
ZmTIP2-1 5 29 56 27 117
ZmTIP2-2 0 0 27 1 28
ZmTIP2-3 2 0 11 0 13
ZmTIP3-1 9 6 0 0 15
ZmTIP4-1 0 0 3 8 11
ZmTIP4-2 0 3 0 8 11
ZmTIP4-3 0 6 0 0 6
ZmTIP4-4 0 5 0 0 5
ZmTIP5-1 0 1 0 0 1
ZmNIP1-1 0 14 0 3 17
ZmNIP2-1 0 0 0 3 3
ZmNIP2-2 0 0 0 1 1
ZmNIP3-1 2 12 3 0 17
ZmSIP1-1 1 6 0 0 7
ZmSIP1-2 3 12 0 2 17
ZmSIP2-1 0 4 0 0 4
  Total 127 536 282 374 1,319

    a Tissue culture. b Reproductive tissue. c Aereal vegetative tissue.

It is clear that a more detailed analysis is needed to determine the tissue and cell type specific expression of the individual genes. The information given in Table III provides a sound basis from which to proceed with such an analysis.


By screening a very large database of maize ESTs we identified a number of AQP genes and for 31 of these we were able to obtain complete nucleotide sequences. This large number is not surprising because Arabidopsis contains 35 AQP genes (Weig et al., 1997; Kjellbom et al., 1999, and our own analysis of the genome). AQP cDNAs are often very difficult to clone in Escherichia coli and we expect that some additional maize AQPs will be forthcoming. We obtained four partial sequences (one PIP, two TIPs, and one NIP) for which we were unable, after repeated attempts, to obtain full-length sequences. This is probably caused by the hydrophobicity of the proteins, which could disrupt bacterial membranes and impair bacterial growth, even if they are expressed at a very low level. If we obtain these full sequences they will be submitted to GenBank as we obtain them.

Maize AQPs Form Four Subfamilies

The maize AQP sequences can be grouped into four subfamilies, which we named PIPs, TIPS, NIPs, and SIPs. This nomenclature preserves the earlier names of TIPS and PIPs, which denote sequence similarity rather than subcellular location. We do not know if all the PIPs and TIPs are located in plasma membranes and tonoplasts, respectively. The name NIP refers to the Nod26-like MIPs previously called Nod-like MIP (NLM; Weig et al., 1997; Weig and Jakob, 2000). The name SIP refers to small basic integral proteins, identified in maize by one of us (R. Jung) and in Arabidopsis (by U. Johanson in an oral communication at the MIP2000 meeting in Gothenburg, Sweden in July 2000). These proteins are not smaller than TIPs but appear to be more basic and as a group have a slightly higher pI than the TIPs.

The distribution of sequences between the four major subfamilies is very similar in maize and Arabidopsis, where 13 PIPs, 10 TIPs, nine NIPs, and three SIPs can be found in the recently completed sequence of the entire genome. In maize, if we count full-length and partial sequences, we have 14 PIPs, 13 TIPs, five NIPs, and three SIPs. Unlike the Arabidopsis genome, the maize genome has not yet been sequenced. The broad representation of AQPs in the four categories gives us confidence that we have obtained most of the maize AQPs. Moreover, AQPs from maize and Arabidopsis are found in each group of the different subfamilies, except for the NIP2 group where no Arabidopsis counterpart has been detected in its genome. Overall, this observation suggests that the separation into the different groups had occurred before the monocot-dicot divergence and that the ancestral gene of these groups encoded a protein with a specific biological role. The persistence of these groups in monocots and dicots is also an indication of the crucial role of AQPs in water and solute relations in plants.

The phylogenetic analysis of the four subfamilies showed that, in many cases, the proteins in a given species are found on a single branch (without members from another species), indicating a number of recent DNA duplication events arising after the monocot-dicot separation. As already outlined above, this is particularly striking in the PIP tree, where five ZmPIP1 and six ZmPIP2 proteins are found on the same branch in each respective group. The same phenomenon is much less pronounced in the TIP cladogram, where most subgroups contain monocot as well as dicot sequences. This independent, and apparently evolutionary, late emergence of novel PIP genes could indicate similar adaptive advantages that were gained in the two large angiosperm clades, probably in response to similar selective pressures.

This leads to the crucial question: Why are there so many different AQPs in a single plant species? The separation of AQPs in different subfamilies and groups may reflect a specialization of the function and the localization. The water channel activity of only four maize AQPs has been tested in X. laevis oocytes, but these data and those obtained in other species indicate a differential regulation of the activity. ZmTIP1-1 is the most highly expressed tonoplast AQP, corresponding to the Arabidopsis -TIP (Barrieu et al., 1998; Chaumont et al., 1998). Members of the TIP4 group may transport solutes in addition to water as demonstrated for NtTIPa, which transports urea and glycerol in addition to water (Gerbeau et al., 1999). As is the case for the other plasma membrane PIP2 proteins tested so far, ZmPIP2-5 is a good water channel, but ZmPIP1-1, ZmPIP1-2, and other PIP1 homologs show a poor water-transport activity in oocytes (Chaumont et al., 2000). NIP proteins have been found to transport glycerol as well as water (Dean et al., 1997; Rivers et al., 1997; Guenther and Roberts, 2000; Weig and Jakob, 2000).

More recent duplication events giving rise to close isoforms in a single species could be a way to control specific expression according to developmental and environmental conditions. For instance, ZmPIP2-2 is mostly expressed in reproductive tissue and ZmPIP2-4 in roots. This could also be true for proteins present in a specific group, such as the TIP3 members, which are found in seed and cotyledons (At-TIP).

ZmPIP1-3 and ZmPIP1-4 obviously are also the result of a very recent gene duplication. Both genes encode identical proteins and their transcripts show a similar tissue distribution. The result of having two genes may be an increased expression level due to a gene dosage effect of the duplicated genes. A recent study (Lynch and Conery, 2000) based on genome-wide analyses of different organisms discusses frequent gene duplications and their importance for the evolution of a species and its evolutionary fate. Because of its large size, the AQP gene family in plants is well suited to test such a phylogenetic hypothesis.

It is interesting that we obtained an aberrant SIP cDNA that also contained exonic sequences of an unrelated gene. This cDNA is not a cloning artifact, but rather it is derived from a pseudogene transcript. Several clones of this cDNA were isolated from independently constructed libraries, all originating from a cultured cell line of the maize cv Black Mexican Sweet. The derived amino acid sequence of this pseudogene contains several stop codons and was not included in the phylogenetic analysis.

Structural Features of Water and Glycerol Transporters
Mammalian AQP1 has recently been crystallized and its three-dimensional structure determined at 3.8-┼ resolution (Mitsuoka et al., 1999; Murata et al., 2000). We modeled ZmPIP1-2 based on the structure of AQP1 (Fig. 8). Sequence analysis revealed the conservation of many amino acid positions (about 60), especially in the two loops that create the aqueous pore. In addition, amino acid residues in different domains that stabilize the structure of the pore through a salt bridge, an ion pair, and a hydrogen bond are completely conserved in all maize AQPs. Determination of the structure of a plant AQP (common bean -TIP) at much lower resolution (7.7 ┼) has indicated that this plant AQP has the same general structure as AQP1 (Daniels et al., 1999).

Some AQPs are water specific; others, such as GlpF from E. coli, are glycerol specific, and yet others have a mixed function. Several AQPs that transport glycerol have been identified in plants (Rivers et al., 1997; Biela et al., 1999; Dean et al., 1999; Gerbeau et al., 2000; Guenther and Roberts, 2000; Weig and Jakob, 2000), although it is not known that they transport glycerol in planta. On a comprehensive cladogram that includes all AQPs, these glycerol transporters are grouped with the water-transporting AQPs and not with the glycerol transporters of prokaryotes and the mixed-function AQPs of mammals (Heymann and Engel, 1999).

Froger et al. (1998) identified five amino acid positions that appear to be highly conserved in the glycerol transporters, but only two of these are conserved in the glycerol transporters of plants (see also Weig and Jakob, 2000). Also in maize, only two of these positions are conserved in ZmNIP1-1 and ZmNIP3-1, which are the putative glycerol transporters, but are not conserved in ZmNIP2-1 and 2-2. The structural analysis of GlpF (Fu et al., 2000) does not indicate why these conserved residues might be important. Fu et al. (2000) identified the residues that interact with glycerol in the channel as well as the hydrophobic residues that line the amphipathic channel, and there is better conservation of these between GlpF and maize NIPs. Of seven channel residues that interact with glycerol via hydrogen bonds, five are conserved. There is a substitution of Phe200 (GlpF numbering) with Gly or Ala in ZmNIPs and of Ala201 with Ser. However, it is the carbonyl groups of these residues that are important for hydrogen bonding to glycerol rather than their functional groups. The carbonyls are oriented by hydrogen bonding of backbone NH to a highly conserved Glu in TM4. Future functional analysis will reveal whether the maize NIPs are glycerol channels.


cDNA Libraries and EST Databases

In toto, 215 maize (Zea mays) cDNA libraries were constructed covering all major maize tissue types, including different developmental stages of these tissues, tissues from plants under biotic and abiotic stress conditions, tissues and cell cultures responding to chemical treatments, and tissues isolated from mutant maize lines. RNA was isolated from maize tissues using TriZol Reagent (Gibco-BRL, Gaithersburg, MD). cDNA synthesis was performed using the SuperScript II kit and cloned into NotI/SalI sites of the pSPORT1 vector (Gibco-BRL) or by using the cDNA Synthesis Kit (Stratagene, La Jolla, CA) and cloning into XhoI/EcoRI sites of the pBluescript SK+ vector (Stratagene). Inserts of randomly picked clones were sequenced by Human Genome Systems (Rockville, MD) or at the DuPont genomics facility (Newark, DE) from the 5' end to obtain ESTs. To a limited amount clones were picked after library normalization or from subtracted libraries. The sequence information of 472,890 maize EST accessions is maintained in the central DuPont/Pioneer Hi-Bred genomics database and is accessible via database interface software.

Clustering, Identification, and Analysis of Maize AQP EST and cDNA Sequences

EST sequences were first evaluated using PHRED-assigned quality scores (Ewing et al., 1998) and, after removal of short, low-complexity, and low-quality sequences, the similarity relationship (clustering) between ESTs was established using the BLAST algorithm (Altschul et al., 1990). EST clusters were subsequently subjected to a sequence assembly process using the PHRAP algorithm ( and about 38,000 contigs and 92,000 singletons were made. The resulting database of contigs and singletons was systematically searched (BLAST) for AQP-related sequences using publicly available AQPs from Escherichia coli, yeast (Saccharomyces cerevisiae), plants, and animals. As a result, about 1,300 EST accessions formed an initial group of about 200 AQP-related sequences, singletons, and contigs. These sequences were thoroughly inspected for over- and under-clustering, visually and by pair-wise sequence alignments (BLAST, ClustalW, Gap), which resulted in 84 putatively unique sequences. The longest clone representing each of these sequences was obtained from the libraries and both strands of each insert were sequenced either by primer walking or by sequencing of nested sets of deletion sub-clones after transposition. The identification and clustering of unique maize AQPs was further refined by repeated pair-wise BLAST searches and by searches of the entire EST database using both the obtained full-length insert nucleotide sequences and their deduced encoded amino acid sequences. After several rounds of reiterative analysis and sequencing, 36 complete nucleotide sequences of unique maize AQPs, 32 encoding full-length cDNA, and four encoding partial cDNA were identified in the current DuPont/Pioneer Hi-Bred maize genome database. An analysis of maize ESTs in the public databases did not turn up additional unique AQP sequences.

The transcript expression profile of each unique maize AQP was estimated by tallying the tissue distribution of clustering ESTs.

The sequence alignments were performed by CLUSTAL X (1.8; Thompson et al., 1997). The pair-wise alignments were calculated by the dynamic programming method. The multiple alignments used the series of GONNET matrices. Trees were calculated using the Neighbor-Joining method and displayed using TreeView (Page, 1996). We used the gene product names when available and the systematic open reading frame names otherwise. Figures were prepared in Illustrator 9.0 (Adobe Systems Incorporated, San Jose, CA).