Aquaporin PIP genes are not expressed in the stigma papillae in Brassica oleracea


*For correspondence (fax +33 4 72 72 86 00; e-mail


The pollen grains of angiosperms are usually desiccated at maturity. Following pollination, pollen hydrates on the stigma surface before germination takes place. Rehydration is an essential step for the success of pollination and depends on the movement of water from the stigmatic cells. This water flow has been shown to be biologically regulated, and components of both pollen and stigma surfaces have been demonstrated to play a role in the control of pollen hydration. Regulation of water transport between animal or plant cells involves membrane proteins, designated aquaporins, which possess water-channel activity. Such molecules may be candidates for controlling pollen hydration, and consequently we investigated whether aquaporins are present in the pollen and stigma cells in Brassica oleracea. Here, we report the identification of two new aquaporin genes, Bo-PIP1b1 and Bo-PIP1b2, which are highly homologous to PIP1b from Arabidopsis thaliana. Both Bo-PIP1b1 and Bo-PIP1b2 proteins are active water channels when expressed in Xenopus oocytes. Expression of Bo-PIP1b1 and Bo-PIP1b2 was observed in reproductive organs as well as in vegetative tissues. Interestingly, the use of a Bo-PIP1b2 cDNA probe revealed that PIP1-like transcripts were not present in the pollen grains or in the stigma papillae, but were present in the stigma cell layers underlying the papillar cells. This observation suggests that water flow between the pollen and stigma papillae may be dependent on aquaporins expressed in cells that are not directly in contact with the pollen grain.


During compatible pollination, pollen grains adhere to the papillar cells of the stigma and then rehydrate to reach a water content that allows them to regain active metabolism. Once metabolically reactivated, the pollen grains germinate to emit a pollen tube which grows down through the transmitting tissue of the style to the ovary. In the ovary, the pollen tube delivers the two male gametes into the embryo sac where the double fertilization takes place.

The time required for completion of pollen rehydration may vary greatly according to the species considered. For example, pollen grain is rehydrated after 7 min in rye (Secale cereale) and 15 min in Arabidopsis, whereas it requires more than 60 min in Brassica oleracea ( Dickinson, 1995; Heslop-Harrison, 1979; Preuss et al., 1993 ). This relatively slow uptake of water by pollen landing on a stigma greatly contrasts with the almost instantaneous hydration of pollen grains when placed in liquid media. This observation indicates that the availability of water from the stigma is variable among species, and suggests that pollen or stigma components may participate in controlling water flow. In fact, a recent study of Arabidopsis mutants defective in the pollen hydration process has clearly shown the essential role played by long-chain lipids of the pollen wall in this process ( Hülskamp et al., 1995 ; Preuss et al., 1993 ; Wolters-Arts et al., 1998 ).

Since the early 1990s, water transport channels that facilitate the flow of water between cells have been reported in many animal and plant systems (reviewed by Agre et al., 1998 ; Maurel, 1997). These water channels, also called aquaporins, act mainly as water-selective channels, generally excluding the passage of ions and metabolites, although glycerol and urea have been reported to pass along with water molecules in a few cases ( Biela et al., 1999 ; Ishibashi et al., 1994 ; Ishibashi et al., 1997 ; Ishibashi et al., 1998 ; Rivers et al., 1997 ). Aquaporins facilitate the movement of water through membranes but do not act as pumps, the driving forces for water movement being hydraulic or osmotic in nature. In plants, aquaporins have been found in various tissues and are localized in the membrane of the vacuole (termed TIPs – tonoplast-intrinsic proteins; Maurel et al., 1993 ) or in the plasma membrane (termed PIPs – plasma membrane-intrinsic proteins; Kammerloher et al., 1994 ). Although they are an abundant component of the membrane and are widely distributed throughout the plant tissues or organs, involvement of TIPs or PIPs in physiological processes has not yet been clearly demonstrated ( Chrispeels and Maurel, 1994; Kaldenhoff et al., 1998 ; Maurel, 1997). With regard to their expression patterns, two functions of plant aquaporins have been postulated: a role for TIPs in intracellular osmotic equilibration, and a role for PIPs in the regulation of transcellular water transport. Recently, the analysis of self-compatible mutants in Brassica has led to the assumption that an aquaporin-like gene, MOD, is required for the self-incompatibility response of Brassica ( Ikeda et al., 1997 ). Because it may be envisaged that rehydration of pollen on stigma cells is regulated by water-channel activity, we undertook a study to determine whether aquaporin genes of the PIP type were expressed in pollen and stigma tissue. In this paper, we report the identification of two genes, Bo-PIP1b1 and Bo-PIP1b2, that belong to the PIP family and which are expressed in the anther, stigma and style of B. oleracea. The water-channel activity of the corresponding proteins was demonstrated by expression in Xenopus laevis oocytes. In situ hybridization and immunocytological analyses showed that no aquaporin transcripts or proteins are detectable in the papillar cells of the stigma, and an RNA blot analysis suggested that PIP genes may be only very weakly expressed in pollen. This observation raises the question as to whether aquaporins are actually present in the plasma membrane of these cells and involved in the rehydration process of the pollen grain.


Identification of the aquaporin genes Bo-PIP1b1 and Bo-PIP1b2

To isolate aquaporin genes putatively expressed in the pollen and stigma in B. oleracea, we screened anther and stigma cDNA libraries with a probe (BOPC 39) derived from a B. oleracea aquaporin-like gene. Hybridizing clones were further analysed by a PCR strategy using primers WC1 and WC3 (see Figure 1) designed from nucleotide sequences conserved within animal and plant aquaporins. Following this second screening, two cDNAs of about 1200 bp, one from the anther-derived library and the other from the stigma-derived library, which both contained a putative initiator and terminator codon, were completely sequenced. Comparison of the two cDNA sequences with sequences in the GenBank database revealed that both genes belong to the superfamily of MIPs (membrane-intrinsic protein genes). Within this superfamily, the highest homologies were found with members of the PIP1 aquaporin family, and we called these two clones Bo-PIP1b1 and Bo-PIP1b2, respectively, although the water-channel activity of the corresponding proteins remained to be demonstrated.

Figure 1.

Position on the Bo-PIP1b2 cDNA sequence of the different primers (arrowheads) and probes (bold lines) used.

The numbering of nucleotides is relative to the first base of the ATG initiation codon. MO7 and MO8 primers were localized on the Bo-PIP1b1 cDNA at the same position as MO3 and MO4.

At the nucleotide sequence level, Bo-PIP1b1 and Bo-PIP1b2 are 98% homologous in the coding region and 60% homologous in the 3′ untranslated region (UTR). At the protein level, the deduced amino acid sequences of Bo-PIP1b1 and Bo-PIP1b2 only differ by one amino acid residue at position 78 (Ser78 → Ala) ( Figure 2). Within the PIP family, the highest homologies were found with the aquaporin At-PIP1b from A. thaliana ( Kaldenhoff et al., 1993 ) and with three aquaporin-like genes that were recently isolated from Brassica species: MIPA and MIPB from B. oleracea ( Ruiter et al., 1997 ) and MOD from B. campestris ( Ikeda et al., 1997 ).

Figure 2.

Comparison of the deduced amino acid sequences of Bo-PIP1b1 and Bo-PIP1b2 with those of PIP1b ( Kammerloher et al., 1994 ), MIPA, MIPB ( Ruiter et al., 1997 ) and MOD ( Ikeda et al., 1997 ).

Amino acid residues identical to the Bo-PIP1b1 sequence are indicated by dots. Underlined regions are predicted to form transmembrane α-helices.

All aquaporins that have been characterized so far contain two conserved NPA motifs and have a predicted topology comprising six membrane-spanning domains connected by five loops and short N-terminal and C-terminal domains in the cytoplasm. The predicted amino acid sequences of Bo-PIP1b1 and Bo-PIP1b2 presented these general features. For this reason, Bo-PIP1b1 and Bo-PIP1b2 proteins were considered to constitute putative water channels.

Bo-PIP1b1 and Bo-PIP1b2 are water channels

The first plant membrane protein reported to have water-transport activity was the vacuolar membrane γTIP protein from A. thaliana ( Maurel et al., 1993 ). In that study, the water-channel activity was demonstrated by using the Xenopus oocyte system, allowing expression of heterologous proteins from RNAs. To determine whether Bo-PIP1b1 and Bo-PIP1b2 also exhibited water-channel activity, we cloned both corresponding cDNAs in a vector for in vitro synthesis of capped complementary RNA (cRNA), and injected the Bo-PIP1b1 cRNA or Bo-PIP1b2 cRNA into Xenopus laevis oocytes. Two, three and four days after injection, the oocytes were used for swelling assays by exposing them to a hypotonic solution. Figure 3 shows that oocytes micro-injected with cRNAs of Bo-PIP1b1, Bo-PIP1b2 or γTIP (this latter being used as a positive control) exhibited an increase in the osmotic water permeability coefficient (Pf) of their plasma membrane; this increase was sixfold for Bo-PIP1b2 and Bo-PIP1b1 and eightfold for γTIP in comparison with the values for negative control oocytes (injected with water). Taking into account the similarity in the responses observed with γTIP, Bo-PIP1b1 and Bo-PIP1b2 cRNAs, the oocyte swelling assays clearly show that Bo-PIP1b1 and Bo-PIP1b2 encode functional aquaporins.

Figure 3.

Osmotic water permeability (Pf) of oocytes injected with Bo-PIP1b1 and Bo-PIP1b2 cRNAs.

Representative data from the same batch of oocytes. Pf values are expressed in 10−4 cm sec−1. Bars show mean ± SE.

Bo-PIP1b1 and Bo-PIP1b2 are single-copy genes

Twenty-three MIP genes have been reported to be expressed in A. thaliana, 11 of which are PIP family members ( Weig et al., 1997 ). To investigate whether such a polymorphism of PIP genes existed in B. oleracea, we undertook a DNA blot analysis of genomic DNA of B. oleracea digested with three restriction enzymes. DNA fragments containing PIP sequences were identified by hybridization with a probe corresponding to the entire Bo-PIP1b2 cDNA. Figure 4 illustrates the different hybridization patterns obtained with the three enzymes tested. In all digest patterns, three major DNA fragments of almost similar intensity were detected by the probe, suggesting the presence of at least three homologous aquaporin genes in the Brassica genome. To determine which of the digested fragments contained Bo-PIP1b1 and Bo-PIP1b2 genes, we amplified part of the 3′ UTR sequences of the corresponding cDNAs, which are usually the most divergent regions within genes, and used them as putative specific probes. Both 3′ UTR probes detected only a single specific fragment for each digest pattern, excepted for the EcoRI digest hybridized with the Bo-PIP1b1 3′ UTR probe, which shows three very weak hybridizing signals. These results suggest that both Bo-PIP1b1 and Bo-PIP1b2 are single-copy genes in the Brassica genome and that their 3′ UTR sequences constitute highly specific probes.

Figure 4.

Genomic DNA blot analysis of Brassica oleracea hybridized with PIP probes.

Genomic DNA was digested with HindIII, EcoR1 or BamH1 as shown at the top. The DNA blot was probed with the Bo-PIP1b2 coding sequence, 3′ UTR Bo-PIP1b2 or 3′ UTR Bo-PIP1b1. Molecular length markers are indicated on the left in base pairs.

Expression of Bo-PIP1b1, Bo-PIP1b2 and other PIP transcripts

As Bo-PIP1b1 and Bo-PIP1b2 genes were initially isolated, respectively, from cDNA libraries derived from anther and stigma mRNAs, we wished to establish whether the expression of both genes was restricted to the reproductive tissues or could also extend to other parts of the plant. Thus, we looked for Bo-PIP1b1 and Bo-PIP1b2 transcripts in a number of different tissues by carrying out an RNA gel blot analysis using the specific Bo-PIP1b1 and Bo-PIP1b2 3′ UTR probes. The same blot was also hybridized with a probe spanning the coding sequence of Bo-PIP1b2 to reveal the expression of other PIP genes. Figure 5(a) shows that transcripts of about 1200 nucleotides, which all hybridized with the three probes, were found in sepals, petals, ovaries, stigmas and in the three developmental stages of the flower bud. During the development of the anther, PIP genes were more strongly expressed at early (unicellular) and late (tricellular) stages of development.

Figure 5.

Expression of Bo-PIP1b1 and Bo-PIP1b2 in various plant tissues.

(a) RNA blot analysis of total RNA (25 µg) from sepals, petals, ovaries, stigmas, buds at different stages of development, and anthers at different stages of development. The stage of bud development is expressed with regard to that of pollen, with ‘uni’, ‘bi’ and ‘tri’ corresponding to unicellular, bicellular and tricellular pollen grains, respectively. Blots were hybridized with the Bo-PIP1b2 coding sequence or specific Bo-PIP1b2 or Bo-PIP1b1 3′ UTR probes, as indicated on the right. The length of the transcripts is indicated on the left in nucleotides. A photograph of the ethidium-bromide stained gel is shown below each blot (rRNA). (b) RNA blot analysis of total RNA (25 µg) from leaves, stigmas, pollen grains and anthers. The blot was hybridized with the Bo-PIP1b2 coding sequence, specific Bo-PIP1b2 3′ UTR, rRNA and Brassica actin cDNA probes, as indicated on the right. The length of the transcripts is indicated on the left in nucleotides. The absence of signal in total pollen RNA was observed in three individual experiments.

To determine whether PIP genes were expressed in the pollen and contributed to the hybridizing signal observed in whole anthers, we isolated total RNA from mature pollen grains and carried out an RNA gel blot analysis by comparing the pollen RNA with RNA from whole anthers at the tricellular stage, i.e. just prior to anthesis. RNAs from leaves and stigmas were also tested on the same blot to estimate the relative abundance of Bo-PIP transcripts in reproductive and vegetative tissues. Figure 5(b) shows that, whatever the probe used (coding Bo-PIP1b2 or specific 3′ UTR), and even after long exposure times, we did not detect any hybridizing signal in pollen RNA. This absence of signal was not due to the degradation of mRNA during the procedure of pollen RNA extraction, as an actin probe gave a clear detectable signal with pollen RNA. However, the presence of PIP transcripts in pollen was detected by an RT–PCR analysis using specific PIP1 primers (data not shown). Together, these results suggest that PIP genes are only very weakly expressed in mature pollen in comparison with other tissues, and that, in the whole anther, expression of PIP genes derives mainly from the sporophytic tissues of the anther.

Tissue localization of PIP transcripts and proteins in the pistil

To establish which cells of the pistil surface were actually expressing PIP transcripts at maturity, we performed an in situ hybridization analysis on pistil sections by using an antisense probe corresponding to the coding sequence of Bo-PIP1b2. Tissue sections of pistils collected 1 day prior to anthesis, a stage where the pistil is functionally receptive to pollen but still protected from pollination by the perianth parts, were hybridized with either the sense (negative control) or antisense labelled probe. Figure 6(a,c) shows that strong expression of PIP genes (detected as black staining) was specifically observed in the stigmatic cells underlying the papillar cell layer and in tissues of the style, but not in the papillar cells of the stigma. PIP transcripts were also localized, although to a lower extent, in the ovules and other ovary tissues (data not shown). To check whether the lack of signal in the papillar cells was artifactual, we investigated the presence of transcripts known to be highly expressed in the papillae. This is the case for the products of the S locus glycoprotein (SLG) gene, a gene which is associated with the self-incompatibility response in Brassica (see Nasrallah and Nasrallah, 1993). We used an antisense RNA probe specific for the S3 allele of SLG (SLG3) in B. oleracea, which was characterized in a previous study ( Delorme et al., 1995 ). Figure 6(b,d) shows that SLG3 transcripts were specifically and abundantly expressed in the papillar cells but not in cells underlying the stigmatic papillae. This result supports the assumption that the lack of hybridizing signal observed with the antisense Bo-PIP1b2 probe reflects the fact that PIP genes are not expressed in the papillar cells at maturity.

Figure 6.

Localization of PIP and SLG3 transcripts in pistil sections.

(a,c) Tissue sections were hybridized with Bo-PIP1b2 sense and antisense RNA probes. (b,d) Tissue sections were hybridized with SLG3 sense and antisense RNA probes. The scale bar corresponds to 10 µm.

It still remained possible that PIP genes may be expressed in stigmatic papillae at early stages of development of the pistil, or, by contrast, late, following pollination of the stigma. To address this question, we localized PIP transcripts on sections of pistils collected from young flower buds (unicellular stage), from flower buds at about 4 days before anthesis (bicellular stage) or from mature flowers which had been pollinated for 1 h. Figure 7 illustrates the pattern of expression of PIP genes during these different stages. PIP transcripts were not detected at any stage in the papillar cells of the stigma, whereas PIP transcripts were always found in tissues of the style. PIP gene expression was particularly high in the region of the stigma underneath the papillar cell layer, in pistils collected from bicellular flower buds and in pollinated flowers. A more diffuse labelling was visible in most tissues of the style, except in the epidermal cells where no transcript was detected. The highest level of PIP expression was observed in pistils from flower buds at 4 days before anthesis (bicellular stage), although the relatively high background of the control (sense probe) section indicates that PIP transcripts may not be so differently expressed as compared to the other developing stages.

Figure 7.

Localization of PIP transcripts during stigma development. Pistil sections from young flower buds (a,d), from buds at 4 days before anthesis (b,e) and from mature flowers after compatible pollination (c,f) were hybridized with Bo-PIP1b2 sense (a–c) and antisense (d–f) RNA probes. The scale bar corresponds to 10 µm.

Finally, to determine whether the localization of PIP proteins followed that of transcripts, we carried out an immunocytological study on sections of pistils collected from flower buds 1 day prior to anthesis. We used immunoglobulins purified from a rabbit immune serum raised against the N-terminus of PIP1-like aquaporins to detect PIPs in pistil tissues. Figure 8 shows that, following incubation of sections with the anti-PIP1 antibody, a detectable labelling (dark staining) was visible in stigma cells underneath the papillar cell layer and in stylar cells, although no strong labelling was detected in the central part of the transmitting tissue. The labelling observed in the papillar cells was the same as the background signal detected in control (pre-immune serum) sections, suggesting that no PIP1 proteins are present in papillae.

Figure 8.

Immunolocalization of PIPs in the stigma.

Pistil sections from flower buds at 1 day before anthesis were incubated with the pre-immune serum (a) or purified anti-PIP1 antibody (b). The scale bar corresponds to 10 µm.

Identification of PIPs in different tissues

Because PIP transcripts were only detected in mature pollen grains by RT–PCR, we investigated whether the corresponding proteins were present in pollen. Protein extracts from leaves, stigmas, pollen grains and whole anthers of B. oleracea were analysed by protein immunoblotting. Figure 9 shows that PIPs were detected in all the tissues tested, although the weakest intensities of immunostaining were obtained with proteins extracted from pollen.

Figure 9.

Expression of PIPs in different tissues.

Proteins (10 µg) extracted from leaves, stigmas, pollen grains and whole anthers of Brassica oleracea were separated on an SDS–PAGE gel, electroblotted and immunostained with purified anti-PIP1 immunoglobulins. The estimated molecular mass of the PIPs, expressed in kDa, is indicated on the right.


Identification of the aquaporins Bo-PIP1b1 and Bo-PIP1b2

The first objective of this work was to investigate whether water channels are expressed in pollen and stigmatic cells, which are structures that face water flows during the early steps of the fertilization process, i.e. pollen hydration on the stigma papillae. The approach used, based on the screening of two cDNA libraries, allowed us to identify two aquaporin-like genes, namely Bo-PIP1b1 and Bo-PIP1b2, which were isolated from anther and stigma cDNA libraries, respectively. From the comparison of Bo-PIP1b1 and Bo-PIP1b2 polypeptide sequences with sequences in the GenBank database, we show that Bo-PIP1b1 and Bo-PIP1b2 are highly homologous to PIP1b from A. thaliana ( Kaldenhoff et al., 1993 ), and to MIPA, MIPB ( Ruiter et al., 1997 ) and MOD ( Ikeda et al., 1997 ) from Brassica species. However, the water-channel activity of the proteins encoded by these three latter genes has not yet been demonstrated. Here, we report that Bo-PIP1b1 and Bo-PIP1b2 proteins are actually functional aquaporins as can be deduced from oocyte swelling assays. Whether only water molecules can be transported through Bo-PIP1b1 and Bo-PIP1b2 remains to be proven. It is known that some aquaporins can also transport small molecules besides water. For example, the plant aquaporin NtAQP1, identified in Nicotiana tabacum and which shares sequence homology to the At-PIP1 protein family, has been shown to mediate glycerol transport in addition to water flow ( Biela et al., 1999 ). Thus, we cannot rule out the possibility that Bo-PIP1b1 or Bo-PIP1b2 may also display such a dual transport activity.

Expression of Bo-PIP1b1, Bo-PIP1b2 and other PIP aquaporins

A common feature of plant and mammal aquaporins is the number of isoforms found in various tissues, with some being expressed in a ubiquitous manner while others seem to be highly specific to given tissues. The DNA blot analyses performed in the present study show that B. oleracea Bo-PIP1b1 and Bo-PIP1b2 are members of a small multi-gene family and are present as single-copy genes in the Brassica genome. The studies by Ruiter et al. (1997) on two other Brassica PIP-like genes, MIPA and MIPB, are consistent with our observation. Taking into account the fact that the highest sequence homologies between the five known Brassica PIP genes (Bo-PIP1b1, Bo-PIP1b2, MIPA, MIPB and MOD) and sequences from the GenBank database were found with PIP1b from A. thaliana, we may consider that these five genes are members of the PIP1 sub-family in B. oleracea. Interestingly, a similar number of PIP1-related genes has been reported in A. thaliana ( Weig et al., 1997 ). In that study, of the 23 MIP genes reported and analysed, 11 were identified as PIP genes, 11 as TIP genes, and one, At-NLM1, was shown to encode an active aquaporin related to the Gm-NOD26 protein found in the bacteroid membranes of soybean (Glycin max) root nodules. In A. thaliana, the expression of PIP1b (previously named AthH2) is induced by blue light and abscisic acid ( Kaldenhoff et al., 1993 ) and by osmotic shock ( Shagan and Bar-Zvi, 1993). It would be most interesting to investigate whether Bo-PIP1b1 and Bo-PIP1b2 are similarly expressed.

The use of 3′ UTR probes specific for Bo-PIP1b1 and Bo-PIP1b2 genes allowed us to demonstrate that these two aquaporin genes are not specifically expressed in the anther or stigma. The genes present a similar pattern of expression in the different tissues we tested, with the highest level of expression found in petals. It is worth noting that the expression of PIP genes in the developing anther is observed before as well as during the dehydration process that occurs in the anther. The maximum level of PIP expression is detected in anthers at the tricellular stage, i.e just prior to the dispersion of dehydrated mature pollen grains. A striking common feature concerning the expression of Bo-PIP1b1, Bo-PIP1b2, MIPA and MIPB genes (see Ruiter et al., 1997 ) is that no transcripts of these genes were detected in pollen grains by total RNA blot analysis. The absence of hybridizing signal observed with the probe corresponding to the coding sequence of Bo-PIP1b2 strongly suggests that PIP transcripts in general may only be present in very low amounts in the pollen at maturity. Moreover, PIP transcripts are never detected in isolated bicellular ( Ruiter et al., 1997 ) or tricellular pollen grains ( Ruiter et al., 1997 ; this study, see Figure 5) by total RNA blot analysis. The relative abundance of PIP transcripts in the anther in comparison with their absence or very low abundance in developing pollen grains suggests that PIPs from the anther may be the only such molecules implicated in controlling water movements leading to a final low water content of the mature pollen grain. The molecular mechanism by which water leaves the vegetative cell of the pollen cannot be determined at present, but it is likely to depend on the existence of a low water potential in the medium surrounding the pollen grains. This low water potential could be generated by the activity of PIPs present in the tissues of the anther. Interestingly, we have shown by immunodetection that PIPs are present in pollen extracts, despite the very low expression of PIP genes in pollen. The immunodetected PIPs probably derived from the tapetum of the anther. Indeed, the pollen coat that fills the cavities of the pollen wall originates from the tapetal cells that degenerate at the end of pollen development ( Dickinson and Lewis, 1973). So, most probably, PIPs are synthesized in the tapetal cells and then migrate to the pollen wall cavities. This assumption is supported by the fact that MIPA and MIPB were isolated as sequences encoding pollen coat proteins ( Ruiter et al., 1997 ). At the surface of the Brassica pollen wall, two distinct membrane-like layers have been described, the exinic outer layer (EOL; Gaude and Dumas, 1984) and the coating superficial layer (CSL; Dickinson and Elleman, 1985). We envisage that PIPs of the pollen coat may be located in these membranous structures. Whether these pollen coat PIPs are actually involved in the dehydration process of Brassica pollen remains to be demonstrated.

Expression of PIPs in the stigma and pollen hydration

Most strikingly, the cytological analyses we carried out to locate PIP expression in the stigmatic tissue show that PIP transcripts and proteins are not detected in the papillar cells, i.e. the cells that directly contact the pollen grains following pollination. Expression of PIP aquaporins does not seem to be developmentally regulated in the papillae, as a similar pattern of expression is found at the different stages of development analysed, from young stigmas to mature unpollinated or pollinated stigmas. Moreover, no PIP transcripts were detected in the epidermis of the stigma or style. This latter observation was not unexpected as the stigma papillae and the epidermal cells originate from the same layer (L1) during the development and differentiation of reproductive organs. The level of expression of aquaporins in the epidermis has been shown to be variable according to the tissue and the type of aquaporins considered. For example, MIPA (a PIP1b-like aquaporin) from the ice plant Mesembryanthemum crystallinum is not expressed in the epidermis and epidermal bladder cells in leaves ( Yamada et al., 1995 ), while PIP1b from A. thaliana is poorly expressed in the epidermal cells of leaves but expressed at much higher levels in the guard cells ( Kaldenhoff et al., 1995 ). In roots, MIPA proteins from the ice plant are present in the epidermis ( Yamada et al., 1997 ), and the tonoplast aquaporin ZmTIP1 from maize is highly expressed in the epidermal cells of the root tip ( Barrieu et al., 1998 ). From the different studies that have reported expression patterns of aquaporins, it seems that there is general accordance between the detection of transcripts and the distribution of the corresponding proteins. Our immunolocalization analysis confirms this assumption , and it is likely that the plasma membrane of the stigma papillae does not contain PIPs, although this remains to be clearly demonstrated by electron microscopy analysis.

By contrast, high levels of PIP transcripts and of the corresponding PIPs are detected in the first cell layers underneath the stigma papillae. Given that the water flow between the stigma surface and the dehydrated pollen grain is biologically regulated (see Introduction), the presence of PIPs in the stigmatic cells that underlie the papillar cell layer suggests that they may play a crucial role in controlling water movements necessary for pollen rehydration to occur. We propose that the apoplastic water potential in the stigma is under the control of these PIPs. Intriguingly, the aquaporin-like MOD gene has been shown to be required for establishment of the self-incompatibility response in Brassica campestris, as mod mutants that do not express MOD transcripts are self-compatible ( Ikeda et al., 1997 ). These authors have proposed that the MOD aquaporin is activated by the S locus receptor protein kinase (SRK), which is a plasma membrane receptor-like kinase involved in the self-incompatibility response (see Nasrallah and Nasrallah, 1993). Once activated, the MOD protein would turn the water away from the incompatible pollen, thus preventing an adequate rate of pollen hydration. The hypothesized function of MOD does not exclude the possibility that other functional PIPs (or TIPs) from the stigma may participate in the regulation of water flow during compatible pollination. Taken together, our data and those from the analysis of mod mutants emphasize the great complexity of molecular mechanisms that regulate the transfer of water from the stigma to pollen. It seems important now, in order to elucidate how aquaporins may actually be involved in the pollen hydration process, to determine the precise cellular location of the different PIPs identified in the stigma (e.g. Bo-PIP1b1, Bo-PIP1b2, MOD) and determine how they may interact with other components (e.g. SRK) of the pollen–stigma interaction.

Experimental procedures

Plant material

Plants of B. oleracea var acephala homozygous for the S3 haplotype were grown either in a greenhouse at 25°C or under field conditions.

cDNA cloning and sequencing

Two cDNA libraries, one constructed from mRNA extracted from stigmas and another constructed from mRNA extracted from immature anthers containing bicellular pollen grains, were used in this study and have been described elsewhere ( Cock et al., 1997 ; Giranton et al., 1995 ). The two libraries were screened at high stringency using standard procedures ( Sambrook et al., 1989 ) with a probe corresponding to the BOPC 39 fragment, which was kindly supplied by Dr René Ruiter ( University of Nijmegen, Nijmegen, The Netherlands). This fragment is a 422 bp sequence including the 3′ untranslated region (UTR) and part of the coding region of an aquaporin-like gene (MIPA) isolated from B. oleracea ( Ruiter et al., 1997 ). Positive cDNA clones were further analysed by a PCR strategy using two oligonucleotides designated WC1 and WC3. The degenerative oligonucleotide WC1 (sense strand, 5′-GGiGGiCA(CT)iTiAA(CT)CCiGCiGT(ACGT)AC-3′) corresponds to a conserved region within the MIP sequences and includes the first NPA motif ( Figure 1), whereas WC3 (antisense strand, 5′-GCATTAGGCAAAAGAAGAGGAA-3′) is located in the 3′ UTR of the BOPC 39 fragment, close to the polyA+ tail. PCR amplification was carried out for 25 cycles of denaturation at 94°C for 30 sec, annealing at 50°C for 30 sec and extension at 72°C for 1 min, with a final extension for 7 min. Of the positive clones, the longest cDNAs (about 1100 bp) were partially sequenced by the chain termination method ( Sanger et al., 1977 ). Bo-PIP1b1 and Bo-PIP1b2 cDNAs were completely sequenced on both strands using custom-synthesized oligonucleotides. DNA and protein sequence analyses were performed using Lasergene programs (DNASTAR, London, UK).

Xenopus oocyte micro-injection and swelling assay

The coding sequences of Bo-PIP1b2 and Bo-PIP1b1 cDNA in pBluescript were amplified by PCR using two oligonucleotides ( Figure 1), WC4 (sense strand, 5′-GGCGGGATCCATGGAAGGCAARGAAGAAG-3′) and WC5 (antisense strand, 5′-GGCGGGATCCTCAGCTTCTGGACTTGAAT-3′). BamHI restriction sites were incorporated into the 5′ ends of WC4 and WC5 in order to insert the coding region of Bo-PIP1b2 or Bo-PIP1b1 into the BglII site of pXΒG-ev1. pXΒG-ev1 is a Bluescript-derived vector containing the 5′ and 3′ untranslated regions of the β-globin cDNA of Xenopus ( Preston et al., 1992 ). pXBG-ev1 containing the γTIP cDNA was kindly donated by Dr Christophe Maurel (Institut des Sciences Végétales, Gif-sur-Yvette, France). Capped complementary RNAs (cRNAs) were then synthesized in vitro using T3 RNA polymerase (Promega, Charbonnières, France) as previously described ( Preston et al., 1992 ), yielding cRNAs encoding γTIP, Bo-PIP1b2 or Bo-PIP1b1 proteins. Oocytes were injected with either 50 nl of nuclease-free water or 50 nl of γTIP, Bo-PIP1b2 or Bo-PIP1b1 cRNA (1 mg ml−1) and were incubated at 18°C in Barth's solution (10 m m HEPES–NaOH, pH 7.4, 88 m m NaCl, 1 m m KCl, 2.4 m m NaHCO3, 0.33 m m Ca(NO3)2, 0.41 m m CaCl2, 0.82 m m MgSO4) containing 50 mg ml−1 gentamycin.

The osmotic water permeability (Pf) of the oocyte plasma membrane was determined by transferring the injected oocytes from Barth's solution (200 mosmol  kg−1) to the same solution diluted to 40 mosmol kg−1 with distilled water, 2, 3 and 4 days after injection as described by Maurel et al. (1993) .

DNA and RNA blots

Genomic DNA was isolated from young leaves according to the procedure described by Vallejos et al. (1992) and digested with the restriction enzymes BamH1, EcoR1 or HindIII. DNA fragments were then separated on 1% agarose gels and transferred to nylon Hybond N+ in 0.4 m NaOH. The protocols used for these different steps followed standard procedures ( Sambrook et al., 1989 ). DNA probes corresponding to part of the 3′ UTR of Bo-PIP1b2 and Bo-PIP1b1 cDNAs (namely 3′ UTR Bo-PIP1b2 and 3′ UTR Bo-PIP1b1 probes, respectively) were obtained by PCR amplification using two pairs of primers, designated MO3 (sense strand, 5′-GCTTTATGCTATTATTA-3′) and MO4 (antisense strand, 5′-ATAGAAGAGAGATTACA-3′) for the 3′ UTR Bo-PIP1b2 probe, and MO7 (sense strand, 5′-ACTTGGCTTTTTGTTCTA-3′) and MO8 (antisense strand, 5′-ATTATCTCCGACTCTGTT-3′) for the 3′ UTR Bo-PIP1b1 probe. PCR products were purified on 1% agarose gel using a Qiaquick gel extraction kit (Qiagen, Courtaboeuf, France) and labelled with 32P-dCTP by specific priming. These probes correspond to nucleotides 883–1030 and 880–1021 of the Bo-PIP1b2 and Bo-PIP1b1 gene sequences, respectively (numbering relative to the first base of the ATG initiation codon). A DNA probe spanning the coding region of Bo-PIP1b2 (designated coding Bo-PIP1b2 probe) was produced by PCR amplification of Bo-PIP1b2 cDNA using WC4 and WC5 primers and was labelled by random priming. Another DNA probe (designated the entire Bo-PIP1b2 probe) corresponding to the entire Bo-PIP1b2 cDNA (nucleotides −36 to 1082) was generated by random primed labelling of single-stranded cDNA. Filters were pre-hybridized and hybridized at 42°C in a solution containing 50% formamide, 6 × SSC (1 × SSC is 0.15 m NaCl, 0.015 m sodium citrate), 0.5% SDS, 0.1% Ficoll, 0.1% polyvinylpyrrolidone, 0.1% BSA, 100 µg ml−1 denatured herring sperm DNA, and washed in 2 × SSC and 0.1% SDS, successively at 55°C for 30 min and at 64°C for 1 h.

Total RNA was extracted from a range of tissues using the method described by Jackson and Larkins (1976). RNAs (25 µg per lane) were separated under denaturing conditions by electrophoresis on agarose gels containing formaldehyde ( Sambrook et al., 1989 ) and stained with ethidium bromide to ensure that equal amounts of RNA had been loaded. The RNA was then transferred to nylon Hybond N membrane by capillary blotting in 10 × SSC and membranes were baked at 80°C for 2 h. Equal transfer was controlled by visualizing RNA with ethidium bromide. Pre-hybridization and hybridization of the RNA blots were performed as above for DNA blots, using the same 32P-labelled DNA probes. Membranes were washed twice for 20 min each in 0.1 × SSC, 0.1% SDS at 50°C.

In situ hybridization and immunolocalization

In situ hybridization was carried out on sections of pistils collected at different stages of flower development. Within a single flower, development of the pollen and pistil are coordinated, and we determined the stages of flower development according to those of the pollen. Buds of various sizes were collected and the respective stages of pollen development (uninucleate microspore, binucleate and trinucleate pollen grains) were assessed by UV fluorescence microscopy using DAPI staining as described previously ( Vergne et al., 1987 ), excepted that 50 m m Tris–HCl pH 7.0 containing 0.5% Triton X-100 was used in place of citrate/phosphate buffer. Methods for tissue preparation, digoxigenin labelling of RNA probes and in situ hybridization were as described by Ingram et al. (1997) . A sense and an antisense RNA probe corresponding to the entire Bo-PIP1b2 RNA (nucleotides −36 to 1082) were prepared as described by Simon et al. (1994) .

Immunolocalization was performed as described by Delichère et al. (1999) , on fixed and embedded sections of pistils collected from flower buds at 1 day prior to anthesis. Sections were incubated with 500 µl of the purified anti-PIP1 antibody (see below) for 2 h at room temperature. Antigen–antibody interaction was detected with a secondary antibody conjugated to alkaline phosphatase by using nitroblue tetrazolium and 5-bromo-4-chloro-3-indolyl phosphate as colour development substrates.

Antibody production and protein analysis

Production of anti-PIP1 antibodies was performed by immunizing two rabbits with the peptide MEGKEEDVR cross-linked to KLH by glutaraldehyde (CovalAb, Oullins, France). This peptide sequence corresponds to the first nine amino acid residues conserved in the N-terminus of Bo-PIP1b1, Bo-PIP1b2 and other PIP1-like aquaporins. Immunoglobulins binding to PIP1-like proteins of the stigma were affinity-purified from immune serum by using the procedure of desorption on immunoblots described by Smith and Fisher (1984), excepted that a solution of 0.1 m glycine–HCl pH 2.2 containing 0.5% BSA was used as the elution buffer. Proteins were extracted, separated by gel electrophoresis and immunoblotted as previously described ( Gaude et al., 1991 ; Gaude et al., 1993 ). Before immunostaining, equal protein transfer was controlled by staining the blot with the reversible Ponceau S protein stain ( Harlow and Lane, 1988).


We are very grateful to Dr Christophe Maurel for his help in micro-injecting the Xenopus oocytes, Dr René Ruiter for supplying the BOPC 39 fragment, Dr Michel Charbonneau (ENS , Lyon, France) for use of Xenopus colonies, Dr Mark Cock for supplying the cDNA libraries and Brassica actin probe, and Dr Philippe Heizmann for supplying the rRNA probe. We thank Dr Françoise Monéger for stimulating discussions, Dr Charlie Scutt for reading the manuscript, and Dr Martine Pastuglia for her help with the Northern blot analysis. We would also like to thank Anne-Marie Thierry and Richard Blanc for their technical assistance. Thierry Gaude and Isabelle Fobis-Loisy are members of the Centre National de la Recherche Scientifique. Marianne Marin-Olivier and Tony Chevalier were supported as graduate students by the Ministère de l’Enseignement Supérieur et de la Recherche and by the Région Rhône-Alpes, respectively .

EMBL accession numbers AF299050(Bo-PIP1b1) and AF299051(Bo-PIP1b2).