Arabidopsis TIP1;1 (γTIP) is a member of the tonoplast family of aquaporins (AQP). Using RNA interference (RNAi) we reduced TIP1;1 to different extent in various lines. When most severely affected, miniature plants died, a phenotype partially complemented by the TIP1;1 homolog McMIP-F. Less severely affected lines produced small plants, early senescence, and showed lesion formation. The relative water content in TIP1;1 RNAi plants was not significantly affected. Global expression profiling suggested a disturbance in carbon metabolism in RNAi lines with upregulated transcripts for functions in carbon acquisition and respiration, vesicle transport, signaling and transcription, and radical oxygen stress. Metabolite profiles showed low glucose, fructose, inositol, and threonic, succinic, fumaric, and malic acids, but sucrose levels were similar to WT. Increased amounts were found for raffinose and several unknown compounds. TIP1;1 RNAi plants also contained high starch and apoplastic carbohydrate increased. A GFP-TIP1;1 fusion protein indicated tonoplast location in spongy mesophyll cells, and high signal intensity in palisade mesophyll associated with vesicles near plastids. Signals in vascular tissues were strongest not only in vesicle-like structures but also outlined large vacuoles. Compromised routing of carbohydrate and lack of sucrose provision for cell-autonomous functions seems to characterize this RNAi phenotype. We suggest a function for TIP1;1 in vesicle-based metabolite routing through or between pre-vacuolar compartments and the central vacuole. Phenotype and expression characteristics support a view of TIP1;1 functioning as a marker for vesicles that are targeted to the central vacuole.
Major intrinsic proteins (MIPs) in various membranes form a large gene family in higher plants. They are often viewed under the aspect of facilitated water movement with the direction of movement determined by osmotic differences between compartments where the proteins reside. Based on this function, the name aquaporin, AQP, has been adopted for the entire family, which includes at least 35 genes in Arabidopsis thaliana (Chaumont et al., 2001; Johanson et al., 2001; Quigley et al., 2002). An ‘aquapore’ function is well documented for many MIPs, but large differences exist between different members in how these proteins alter hydraulic conductance (Maurel et al., 1993). In addition, MIP function is complex with the facilitation of considerable water flux by some MIPs possibly predominantly correlated with either pore size or protein rigidity. The effect of overexpression of the Arabidopsis plasma membrane-located PIP1b in tobacco has been interpreted in an AQP function. The plants, grown under optimal conditions, exhibited increased growth rate, transpiration, and photosynthetic efficiency (Aharon et al., 2003). Experiments that used Xenopus oocyte expression of MIP coding regions, and their expression in Escherichia coli, yeast, erythrocytes, and protoplasts or plants (Agre et al., 1993; Kaldenhoff et al., 1998; Maurel et al., 1993, 1994) provided evidence for other functions, such as MIP-dependent passage of glycerol, urea, CO2, H2O2, and even ions through membranes (Boassa and Yool, 2002; Klebl et al., 2003; Yasui et al., 1999). For example, antisense reduction and overexpression in plants indicated a function for the Arabidopsis PIP1a as a conduit for CO2 (Uehlein et al., 2003).
Conceivably, classification along lines of metabolite preference could replace the classification that is presently based on presumptive, often unproved, cellular location that is used for two of the four sub-families in Arabidopsis. Yet even the notion of AQPs/MIPs as facilitators of small (uncharged) metabolites may no longer be tenable or sufficient. Outlined here is a hypothesis that describes a function that could replace the ‘guilty by sequence homology’ viewpoint. We argue that the passage of a number of small molecules through AQPs may be a matter of size, and that the important function of at least one MIP, TIP1;1 in Arabidopsis, may be as a guide that determines metabolite routing between the Golgi complex, pre-vacuolar compartments, and the tonoplast. TIP1;1 is concentrated or uniquely present in a particular type of plant vacuole that is targeted to the lytic compartment, while a different TIP is a marker for storage vacuoles (Jauh et al., 1998; Liu et al., 2003; Moriyasu et al., 2003; Paris et al., 1996). These results have shaped the concept of how vacuolar trafficking is regulated in plants (Bassham and Raikhel, 2000; Neuhaus and Rogers, 1998; Vitale and Galili, 2001). The absence of TIP1;1 leads to a problem in the distribution of metabolites. The apparently wrongly directed export of metabolites from sink leaves then leads to retarded growth, and cell and plant death.
The 35 AQP genes in the Arabidopsis genome have been divided into four sub-families, termed PIP, TIP, NIP, and SIP. The PIP and TIP sub-family members are labeled according to their purported location in either the plasma or tonoplast membranes. At least for some of these PIP and TIP proteins this location has been made highly likely but evidence is increasing that suggests a location in more than one membrane and compartment, and, also, that MIPs may move depending on physiological states or environmental disturbances (Barkla et al., 1999; Liu et al., 2003; Vera-Estrella et al., 2004).
In the Arabidopsis TIP group, water transport facilitation had been characterized (Daniels et al., 1994; Maurel et al., 1993, 1995), and several TIPs have been reported as urea facilitators (Klebl et al., 2003; Liu et al., 2003). It is believed that the TIPs are predominantly located at the tonoplast, but the tight coupling between protein storage vacuole and αTIP (Jauh et al., 1998) and the GFP signals in fusion proteins (Liu et al., 2003), together with the association of αTIP with non-tonoplast vacuole membranes involved in autophagy (Moriyasu et al., 2003), suggests different TIP isoforms may have functions in addition to water transport, or functions that may be entirely different. Paris et al. (1996) reported that TIP1;1 resided in one type of vesicle only and that the TIP1;1-containing vesicles fused with the tonoplast. Recently, analysis in yeast indicated the capacity for urea transport for TIP1;1 (γTIP), TIP1;2 (γTIP2), TIP2;1, and TIP4;1. The location of at least three of these TIPs, monitored by TIP-GFP protein fusion, indicated their location in various intracellular membranes in addition to location in the tonoplast membrane (Liu et al., 2003).
Based on expressed sequence tag numbers and microarray intensity profiles, TIP1;1 is one of the most highly expressed AQPs (Chaumont et al., 2001; Quigley et al., 2002), and it is an abundant protein in membranes (Higuchi et al., 1998). To investigate its physiological function in Arabidopsis in more detail, we have generated RNAi lines directed against this transcript, because no true knockout line has been described. TIP1;1-RNAi leads to death in plants where TIP1;1 is strongly downregulated, while other AQPs are not significantly affected. High humidity did not alleviate this phenotype. The addition of a TIP1;1 homolog (McMIP-F) partially reversed the phenotype. Based on transcript and metabolite profiling, the phenotype after strong reduction in TIP1;1 points toward a role in the control of carbon distribution that alter source/sink relationships. We suggest that this function includes targeting, or marking the destination, of vesicles destined to the central vacuole. GFP-TIP1;1 fusion protein amount and cellular localization support this interpretation. The absence of AtTIP1;1 seems to reduce the export of carbohydrate from chloroplasts, while at the same time increasing apoplastic carbohydrate concentrations, and reducing carbohydrate in the vacuole of mesophyll cells. We interpret the results assuming that the absence of TIP1;1 leads to the loss of a component involved in intracellular routing of those vesicles that are targeted to the central vacuole.
TIP1;1-RNAi leads to a series of phenotypes in primary transformants
To generate an RNAi construct against TIP1;1 in Arabidopsis, the partial coding region was cloned, flanking a GUS coding region, and inserted into the binary vector pFGC1008 twice, in opposite orientations (Figure 1a). The vector was transferred into Agrobacterium tumefaciens, and Arabidopsis (Col-O) transformed by flower-dipping (Clough and Bent, 1998). Seeds from transformed plants were plated on selective medium, which resulted in a spectrum of conspicuous phenotypes among the hygromycin-resistant plants. When planted to soil, most of the selected primary transformants of TIP1;1-RNAi plants (HygR) displayed a reduced growth phenotype of varying severity between different transformants. According to severity, the RNAi lines were categorized into five groups (Figure 2, Table 1). Among 56 transformed lines, 36 (approximately 60%) showed diverse phenotype, characterized by either slow growth and late bolting and slow seed formation (group III), or early senescing leaves, reduced growth to varying degrees, and an inability to either bolt or generate siliques (groups II), or even lethality before a rosette had formed (six lines; group I). In the intermediate group of RNAi plants, group III showed senescence of leaves, which developed as chlorotic lesions from the leaf edge, while leaf thickness increased, finally resulting in expanding lesions of dead cells, followed by leaf desiccation, shrinking and abscission (Figure 2). Transformed plants in two additional groups were not readily distinguishable from wild type in size: group IV plants showed early senescence on older leaves, and group V plants grew exactly like WT (Figure 2).
Table 1. Comparison between TIP1;1-RNAi plants and complemented plants
RNAi plant number
Complementation plant number
Four-week-old primary transformants, both TIP1;1-RNAi and complementation plants growing under identical conditions, were categorized into five groups according to the severity of phenotype. TIP1;1 expression levels are indicated in the legend of Figure 3.
Lethal, death within 2–3 weeks
Very small plants, early senescing leaves
Intermediate-sized plants, early senescing leaves
Normal size, senescing leaves appear earlier than in WT
Growing like WT
Seeds that were collected from the intermediate RNAi plants (group III) were tested for the phenotype in the T2 generation. Most of the progeny of the first generation of hygromycin-resistant plants did not show the phenotype observed in the T1 generation, or showed an even less severe RNAi phenotype than the parent, suggesting instability of the RNAi effect.
RNAi expression strength correlates with the phenotype
The severity of the phenotype correlated with the expression level of TIP1;1 RNAi vector. Real-time RT-PCR analyzed transcript amounts in RNAi lines of plants assessed in groups II–V (Table 1). The GUS fragment within the RNAi construct (Figure 1a) was amplified to monitor RNAi expression strength in the different size groups. Group II plants showed highest GUS expression. Compared with group II plants, the mean GUS level was reduced to 0.50 (group III), 0.24 (group IV), and 0.20 (group V), respectively, that is a correlation existed between the GUS level and phenotype (Figure 3). Compared with wild type, TIP1;1 mRNA levels varied between 7.4 and 9.8% in all four groups of plants (Figure 3). Analysis of plants in groups III–V by antibody detection of TIP1;1 showed TIP1;1 protein signals reduced to less than 10% of wild type in the three groups for which sufficient protein could be obtained (Figure 3). It appears that the phenotype is solely dependent on the strength of RNAi expression and not on the steady-state transcript levels (see Discussion). Affected by the TIP1;1 RNAi construct, TIP1;2 transcript amounts in the RNAi plants were reduced to between 30 and 50% in the phenotypically different groups of TIP1;1 RNAi plants.
The TIP1;1-RNAi construct might have interfered with TIP1;2 expression, based on an overall identity of 81%, and a 29-nucleotide stretch of absolute identity. To investigate whether the severe phenotype of group II/III plants (for which sufficient material could be obtained) was due to this reduction, Arabidopsis RNAi lines targeting TIP1;2 were generated, by a procedure similar to that used for TIP1;1 RNAi, except that the TIP1;1 partial coding region was replaced by a part of the TIP1;2-coding region (Figure 1a). The TIP1;2-RNAi lines did not show any growth or developmental phenotype when growing under the condition where TIP1;1-RNAi lines exhibited the defect outlines before (S. Ma, unpublished data). Real-time RT-PCR data indicated that the TIP1;2 transcript was reduced in the lines to less than 5% of wild type in two of five lines analyzed. In the TIP1;2-RNAi lines transcript levels of TIP1;1 were slightly affected, but in the most severe cases at least 30% of the TIP1;1 mRNA was still present (Table 2). Similarly, a reduction in TIP1;3, which shows a stretch of 19 bp identical to the RNAi segment used, cannot completely be excluded although its reduction, as seen in microarrays, was less pronounced than that observed for TIP1;2. We conclude that the defect is based on the post-transcriptional silencing of TIP1;1 alone.
Table 2. Relative mRNA amounts in TIP1;2-RNAi lines
The relative amounts of TIP1;1 (γTIP) and TIP1;2 (γTIP2) transcripts in five TIP1;2 RNAi plants. Data were obtained by real-time RT-PCR and normalized against actin-3. Data are expressed as percentage transcript remaining relative to WT. A construct similar to that shown in Figure 1 was established for TIP1;2 (S. Ma, PhD thesis, unpublished data). The phenotype of the TIP1;2 line shows no growth reduction of the plants, indicating that TIP1;2 could have a different function (S. Ma, unpublished data).
Complementation of the RNAi phenotype
To confirm that the loss of TIP1;1 mRNA caused phenotypic changes in the RNAi lines, we generated a complementation population. This population contained a T-DNA cassette to express a TIP1;1 homolog, McMIP-F (accession number U43291) with a nucleotide sequence identity of 73% (76% protein), driven by the Arabidopsis TIP1;1 promoter, in addition to the cassette that contained the TIP1;1 RNA construct (Figure 1b). Individual transformants of the TIP1;1-RNAi line and the complementation line were grown alongside each other (Figure 4) in low light (60 μmol m−2 sec−1). From among 56 plants of the complementation line, 30 grew like wt (group V), while the same number for TIP1;1 RNAi plants was 12 of 56 lines. The percentage of wt-like plants in the complementation population (53.6%) was significantly higher than in the TIP1;1-RNAi population (21.4%), based on a one-sided Z-test of proportions (α = 0.05).
Real-time RT-PCR data showed that McMIP-F was expressed at varied levels in complementation plants. In the plants with minor phenotype, McMIP-F was 0.9% (±0.1%; n = 3) compared with the TIP1;1 level in WT, while in the wt-like complemented plants McMIP-F amounted to 16% (±9.4%; n = 3). While the amount of TIP1;1 transcript in the RNAi lines was in the order of 5–10% of wild type (Figure 3), in the wt-like complemented plants this level was reduced to 1.6% (±0.8%; n = 3), although their GUS level (0.161) was similar to group V RNAi plants. The expression of McMIP-F, replacing TIP1;1, apparently counteracted an activation of TIP1;1 expression in the RNAi lines. In complementation lines growing like wt (group V) a low level of McMIP-F and intermediate TIP1;1 amounts coincided (data not shown). However, at higher irradiation (200 μmol m−2 sec−1), the complemented plants developed phenotypes as severe as the RNAi lines (data not shown). The results indicate that the expression of McMIP-F, and hence the MIPF protein, could only partially compensate for the severe loss of TIP1;1.
Water relations and leaf gas-exchange analysis
Water content of excised shoots from mature RNAi category III, IV, and V seedlings was determined gravimetrically and compared with wild-type plants of the same age. To eliminate variability resulting from plant size or dry weight, water content was standardized as a percentage relative to the initial shoot water content. When all three RNAi lines were considered collectively, their water loss was significantly greater than that measured in wild-type plants, with RNAi lines exhibiting 5–10% lower water content at any given time point (Figure 5a). Further, water loss of RNAi MIPF-complemented lines was identical to wild type. To evaluate the influence of the level of RNAi expression on plant water balance, whole-plant water loss was measured on intact seedlings during 30-h diurnal time courses and expressed on the basis of shoot dry weight. The experiment was repeated three times, and representative results are shown in Figure 5(b). In each repetition, relative ranking of water loss was consistent among the RNAi lines and controls evaluated, yet the rankings did not correlate with the level of RNAi expression. Plants of category V lost water most rapidly, followed by those of category IV, wild type, and the MIPF-complemented RNAi line. Because category III plants were extremely small, their water loss rate could not be reliably determined in this assay. Although relative ranking among lines was consistent over repetitions, absolute differences were small at any given time point. Net CO2 assimilation rate (A) of fully expanded leaves of category V plants was significantly higher than in wild type; it was suppressed, but not entirely complemented, by MIPF (Figure 5c). Similarly, stomatal conductance (gs) of RNAi category V plants was significantly higher (P < 0.05) than in wild type, and MIPF expression restored gs of RNAi to wild-type levels (Figure 5d).
Metabolites were analyzed in extracts from individual plants of the RNAi category III (growth inhibition and early leaf senescence) and from wild-type plants grown in the same tray (n = 6 each). Each metabolite profile was established for a single plant. Analysis of the GC-MS data indicated several changes in metabolite distribution and amount between the RNAi lines and wild type (Figure 6). RNAi lines had significant lower amounts for at least 10 metabolites, many of which represent major peaks in wild-type plant extracts. A reduced amount was observed in the RNAi lines for glucose, fructose, inositol, threonic acid, succinic acid, fumaric acid, and malic acid (Figure 6; Table 3). All of these represented abundant metabolites in wild-type plants. Higher abundance was found for raffinose and more than 10 other metabolites, whose identities remain to be determined or verified. The distribution and difference for the metabolites listed was identical in all RNAi and wild-type plants, although in several (individual) plant extracts not all compounds were detected. Several compounds at high retention times elute at positions where a number of complex carbohydrates can be expected but their identity has not been verified.
Table 3. Metabolite profiles
Metabolite profiles of TIP1;1 RNAi and WT plants. Individual RNAi plants belonging to groups III (expt 1) and IV (expt 2), respectively, were visually selected. Compounds levels are expressed as mg g−1 FW. Values are mean ± SD (n = 3 for each experiment). ND, not detected.
0.91 ± 0.22
0.37 ± 0.09
4.11 ± 0.29
2.40 ± 1.70
6.36 ± 1.93
0.22 ± 0.13
4.32 ± 0.62
0.34 ± 0.40
0.39 ± 0.09
0.06 ± 0.002
0.72 ± 0.14
0.15 ± 0.09
0.27 ± 0.25
1.06 ± 0.43
0.23 ± 0.16
0.03 ± 0.01
0.55 ± 0.27
0.54 ± 0.27
0.06 ± 0.02
0.17 ± 0.08
0.05 ± 0.02
2.89 ± 0.28
0.37 ± 0.26
2.93 ± 0.87
0.74 ± 0.31
2.21 ± 0.51
0.48 ± 0.19
4.07 ± 0.90
2.10 ± 0.90
0.06 ± 0.04
0.25 ± 0.07
0.23 ± 0.13
4.29 ± 1.60
2.25 ± 0.31
3.11 ± 0.81
4.71 ± 0.76
Plants grown at high humidity
Growth at high humidity did not reverse the phenotype of any TIP1;1 RNAi line. Transformants growing in agar with 0.5 × MS (with or without 2% sucrose) in vented Magenta boxes did not show any water deficit compared with wild type. Within approximately 3 weeks, severe RNAi lines growing without sucrose remained small, the leaves showed slight chlorosis, and the root system barely increased in size, while wild-type plants grew normally to a much larger size (Figure 7a). RNAi lines with a severe phenotype growing with external sucrose as a carbon source developed chlorotic leaves that senesced rapidly, while the root biomass increased and the plants developed to a larger size than the RNAi lines without sucrose (Figure 7a).
Starch accumulation in TIP1;1 plants
Phenotype and metabolite analysis indicated that in severely affected lines the distribution or transport of carbohydrates appeared to be affected. The RNAi lines of intermediate size growing in soil developed thick, slightly chlorotic leaves. KJ-staining of the leaves revealed an abundance of starch grains in the chloroplasts of RNAi lines compared with wild type. In addition, in stark contrast to wild type, iodine staining of leaves of the RNAi lines showed signals that increased in the vicinity of the vascular system and around phloem cells in particular (Figure 7b).
GFP localization of TIP1;1
To locate the TIP1;1 protein at the tissue and sub-cellular levels, Arabidopsis lines expressing a GFP-TIP1;1 fusion protein driven by the native TIP1;1 promoter were generated and analyzed by confocal microscopy. All cells showed the GFP signal, albeit at different intensities with the highest intensity in the phloem tissue of the vasculature. Within cells, a generally strong signal was obtained with a membrane that is most likely the tonoplast, but a strong or even stronger signal of the GFP-TIP1;1 fusion protein was also associated with regions adjacent to chloroplasts in mesophyll cells (Figure 8). In these cells, the fusion protein also seemed to form a net-like image and was additionally highlighting small vesicles. Differences in amount and possibly also location are suggested by differences in the signal between palisade and spongy mesophyll cells. In sieve elements and companion cells of the phloem, but not (or extremely low) in xylem parenchyma, the fusion protein generated the strongest signals (Figure 8). GFP expression in roots generated signals that identify cells that are different from those observed in the leaves. GFP in roots is mainly located in the pericycle and endodermis but missing in cells of the phloem and xylem.
Global changes in transcript abundance
Wild-type plants and TIP1;1-RNAi plants were grown alongside each other in two independent experiments. RNAi plants from categories III and IV were visually selected. From each set of plants RNAs were isolated, converted to cDNAs, labeled with cy3 and cy5 fluorescent dyes and used in hybridizations to oligonucleotide microarray slides containing approximately 26 000 DNA elements for genes located in the Arabidopsis thaliana (Col) genome (http://www.ag.arizona.edu/microarray/). The analysis of four hybridizations (two sets of independently grown plants for wild type and RNAi plants each) revealed 65 significantly (>2.5-fold) up- or down-regulated transcripts (Table 4, see also Table S1). The most obvious changes in upregulation were found in categories that identified carbon metabolism, general defense reactions, redox regulation, signaling, transcription, and vesicle transport functions (Table 4). Downregulated were the transcripts TIP1;1 (the intended RNAi target), TIP1;2 (partially targeted by the RNAi construct), and TIP1;3 (not targeted by the RNAi construct) of the TIP sub-family of AQPs. Other downregulated transcripts were found in different categories, defense, metabolism, and redox control. Strongest reduction in transcript amount was found for the glycolysis-related enzyme fructose-1,6-bisphosphate aldolase, several pathogen defense components, a gibberellin-responsive protein, Fe-SOD, a putative receptor kinase, an RNA export factor binding function, and a glutamine-dependent asparagine synthetase (Table 4). In the vesicle transport category, four transcripts were among the most highly upregulated functions. A complete rendition of the microarray data is presented in Table S1.
Table 4. Regulated transcripts revealed by global expression profiling
Log10 (Mu/WT) mean
Data obtained from four microarray slides, which included one biological repeat. Shown are the average log10 values [RNAi line class 3 versus wild type (Mu/Wt)] and SD. Genes are grouped according to functional annotations (TIGR/MIPS databases, December 2003). A negative sign identifies downregulated transcripts.
aThe downregulated TIP1;3 transcript does not show significant nucleotide sequence identity with the partial sequence of TIP1;1 chosen for the RNAi construction.
RNAi suppression or elimination of specific transcripts has opened a powerful way to analyze gene function by generating the reduction in a specific transcript or group of sequence-related transcripts. The approach has been applied to study, for example, animal (Caenorhabditis elegans) development and cell function (Caplen et al., 2001; Zamore, 2002). In plants, several approaches have been reported (Chuang and Meyerowitz, 2000), and, recently, an inducible RNAi expression system has been presented (Guo et al., 2003).
The surprising complexity of the TIP1;1-RNAi phenotype required a number of different experimental approaches to arrive at a hypothesis to explain this phenotype. This suggests that TIP1;1 may have a function in vesicle routing, possibly apart from its transport function. In primary transformants that expressed the RNAi construct strongly, based on growth retardation and real-time PCR analyses, a dominant phenotype could be recognized, and experiments could be carried out with such phenotyped T0 plants. The plants were categorized into different groups based on the severity of the phenotypic changes. Plant size, and survival under optimal water and nutrient conditions, correlated with expression strength of the RNAi construct, monitored by the GUS insert, but not with TIP1;1 mRNA amount per se. The residual TIP1;1 mRNA seems to be an indication of transcription rate, but not the rate of translation. Possibly, the small interference RNA (siRNA) generated from TIP1;1 RNAi inhibits TIP1;1 translation, apart from cleaving the mRNA. When complemented with target mRNA, siRNAs block translation (Doench et al., 2003; Zeng et al., 2003). Completely matched siRNA may block protein synthesis, but this effect will generally be masked by the loss of the targeted mRNA. In plants with a severe phenotype, the decline of plant growth may be strong enough to affect normal RNAi interference, while the siRNA will still inhibit translation. This view is supported by the following observation. While in the RNAi plants for which enough material could be obtained the resident amount of TIP1;1 transcript was between 5 and 10% of the wild-type level, plants complemented by McMIP-F had an even more reduced amount of TIP1;1 mRNA at similarly high RNAi expression. Also possible is that the TIP1;1 promoter in the RNAi lines is upregulated leading to increased initial expression, while this is not the case in most of the lines that expressed McMIP-F in addition to the RNAi construct. The phenotype correlates linearly with the expression of the GUS transcript that is flanked by the TIP1;1 RNAi inverted repeat. This correlation, observed for virtually all transformants, and supported by the complementation experiment, indicates that the growth phenotype was a consequence of the presence of the RNAi construct.
Since the detection of abundant proteins with shared structural elements in membranes (MIPs), the search for their function has generated much discussion. Especially, their identification as water channels, AQPs, was not easily accepted, and in fact such a water channel function was deemed by many as either superfluous, or as restricted to specific tissues, as for example mammalian kidney cells (King and Agre, 1996; Nielsen et al., 2002). Many AQPs have now been analyzed and many act as facilitators of water, glycerol, and a number of other small (typically), uncharged, ubiquitous molecules. The passage of these, and possibly (many) other small metabolites, may be considered an opportunistic exploitation of pore size (Hill et al., 2004). This view then requires additional assumptions about a function or functions that are not necessarily related to small-metabolite facilitation and this then necessitates an explanation of several quandaries. One problem is how the flow of small compounds is controlled, for surely unregulated flows would be detrimental. In addition, the large number of AQP genes in plants requires explanation. This number may be a matter of demand for cell, tissue, development, or environment-specificity driving gene amplification, but as yet, experiments that measure expression specificity are incomplete, restricted to a few members of the family. Considering that the structure of the 35 AQP-deduced proteins is sufficiently diverse to generate the documented water, glycerol, peroxide, or CO2-specific holes, among others, in membranes, it could be expected that their real functions may be equally diverse. Whether a hypothesized specific function is unique to each AQP, or shared by more than one of the proteins, is not known. The presence of two TIPs in vacuoles distinguished by their destination may be an indication for TIP-specific functions (Jauh et al., 1998; Moriyasu et al., 2003; Neuhaus and Rogers, 1998; Paris et al., 1996). A comparison of the RNAi phenotype of TIP1;1 and TIP1;2 is indeed surprising because TIP1;2 RNAi plants do not show the same phenotype as TIP1;1 RNAi plants. TIP1;2 is 81% identical to TIP1;1 and is partially affected by the RNAi construct. In both RNAi lines, the reciprocal inhibition of the other TIP is approximately 50%, which obviously did not generate the reciprocal phenotype. TIP1;1 and TIP1;2 are phylogenetically the most closely related TIPs and the ice plant homolog, McMIP-F, is their nearest neighbor (Kirch et al., 2000). TIP1;1 and McMIP-F seem to have the same function but the ice plant form seems to be less efficiently recognized, less stable, or less strongly expressed in Arabidopsis.
A final task, then, was to find what the real function of any single AQP, or group of similar AQPs, might be. Our analysis focuses on an abundant tonoplast AQP, TIP1;1, that has been studied in detail before (Ludevid et al., 1992; Maurel et al., 1993). It has been analyzed in its transcript and protein expression, cellular localization, and response to water deficit. All results have been, with good justification, related to the presumed water facilitation function. Indeed, TIP1;1 allows water to pass into Xenopus oocytes, and compensates the deficiency of a urea transporter in yeast (Klebl et al., 2003; Maurel et al., 1993). The missing experiment has up to now been silencing or eliminating the TIP1;1 gene, transcript or reducing protein amount experimentally.
What may be the basis of the phenotype is unclear in detail but the results outline several possibilities. Carbon metabolism is affected significantly, as is the routing of carbohydrates. Moreover, transcripts are upregulated for functions in vesicle transport, in mitochondrial functioning, and in oxygen radical and redox homeostasis. Physiological and growth experiments indicate that water deficit may not be a major factor but given the accumulation of carbohydrates in the apoplast might in itself generate an osmotic problem for cells. We reason that the obvious change in carbohydrate distribution might indicate an effect of the absence of TIP1;1 on vacuolar loading of sucrose, which can explain the absence of fructose and glucose in metabolite profiles.
There are several possibilities. TIP1;1 could itself be involved in sucrose loading, example by acting as a subunit in a complex with sucrose (or hexose) transporters, but it is highly unlikely that sucrose (or hexoses) is transported by TIP1;1. Another possibility is that the absence of TIP1;1 could disturb water relations to a degree that may lead to the absence or shrinkage of the central vacuole, but we have not seen an indication for this possibility. Finally, TIP1;1 could be involved in the routing of vesicles toward the vacuole or in the docking and fusion process between pre-vacuoles and the central vacuole. The location and relative signal strength of the GFP-TIP1;1 fusion protein (under control of the TIP1;1 promoter) in mesophyll cells and the phloem may suggest such a role. Confocal microscopy showed strong signals adjacent to chloroplasts, a network-like structure, small vesicles, and at the central vacuole. This dynamic distribution of the TIP1;1 protein is similar to that reported for McMIP F (Vera-Estrella et al., 2004). The known location of TIP1;1, TIP2;1, TIP3;1 in lytic, delta and protein storage vacuoles, respectively, indicates that they may have different functions (Jauh et al., 1998). The RNAi-based loss of TIP1;1 adds a new indicator of its possible function. The accumulation of carbohydrate material in both chloroplasts and the extracellular space could indicate a problem in routing of carbohydrate toward the Golgi and pre-vacuolar complexes. Accumulation in the apoplastic space may then be the result of a default routing and unloading process for vesicles that no longer are recognized as targeted to the vacuole. It appears that, instead of being a ‘passenger’, TIP1;1 could act as a ‘driver’ that is involved directly or by the interaction with other proteins in deciding where some vesicles should go. In phloem-associated cells, the abundance of TIP1;1 in internal membranes, possibly tonoplast-like membranes, seems to suggest a similar role, which – in the absence of sufficient TIP1;1 protein – can explain apoplastic carbohydrate accumulation. In this first report we outlined the utility of RNAi-based gene silencing applied to the functional characterization of an abundant MIP. The unexpected results, to be documented in more detail by additional experiments, may add a new function to the increasing complexity of the AQP protein family.
Vectors and plant transformation
The partial coding region for Arabidopsis TIP1;1 (γTIP) gene was cloned into pFGC1008 as described (Gong et al., 2002), resulting in the TIP1;1-RNAi vector construct. The primers used were (forward) 5′-TTACTAGTGGCGCGCCGGTTCAGGCTCTGGCATG-3′, and (backward) 5′-TTGGATCCATTTAAATTCCACCGCCGACGAGAGG –3′. For complementation, a fragment that included the Mesembryanthemum crystallinum AQP McMIPF (U43291) under control of the Arabidopsis TIP1;1 promoter (1.5 kb upstream of the TIP1;1 ATG codon) was inserted into the TIP1;1 RNAi vector via the SphI restriction enzyme sites. The fragment contained a T-DNA right border identical to that of pFGC1008. Both TIP1;1-RNAi and complementation vectors were transferred into A. tumefaciens, GV1101, by electroporation and used to transformed wild-type Arabidopsis (Col-0) by flower dipping (Clough and Bent, 1998) to generate the RNAi and a complementation population, respectively.
Seeds of primary transformants (RNAi and complementation) were surface-sterilized with Agribrom for 3 days at 4°C and then germinated on 0.5 MS agar (0.7%) containing 3% sucrose, 20 μg ml−1 hygromycin B and 200 μg ml−1 Cefatoxime. One week later, seedlings with longer roots and larger leaves were selected and transferred either to vented Magenta boxes with 0.5 MS agar or to soil for further growth. Insertions were verified by PCR amplification of genomic DNA with two fragments inside the RNAi vectors.
Plants, both RNAi and complementation transformants, together with wild-type plants, were grown in 8 × 4 trays in soil (fine vermiculite: potting soil = 1:1). The trays, covered by hoods to maintain high humidity, were kept in a growth chamber at 12 h daylight and controlled temperature (day/night) 23/19°C. The light intensity was approximately 60 μmol m−2 sec−1. Plants were watered with Hoagland's nutrient solution every week and water as needed.
Four-week-old plants were collected to isolate total RNA (RNeasy; Qiagen, Valencia, CA, USA). For the small RNAi plants, two to three transformants were put together to get enough RNA. One μg of total RNA from each sample was used to synthesize first-strand cDNA in a 20-μl total volume (SuperScript II; Invitrogen, Carlsbad, CA, USA), using oligo(dT) and GUSB (specific to GUS fragment in pFGC1008) as primers. The cDNAs were diluted 10 times, and 2 μl of the diluted solution was used as template in each real-time PCR reaction. Real-time PCR were conducted in a Cepheid Smart-Cycle (Cepheid, Sunnyvale, CA, USA) using the QuantiTectTM SYBR kit (Qiagen) with primers specific to the interested gene. The Arabidopsis actin-3 (NM179953) was used as an internal control. The primers used for amplification are as follows: TIP1;1, forward – GTGGAATCGCTGGACTCATC; back – TGATTCGAAATTACACAAACGG; TIP1;2, forward – AAGCTGGACGTGGACCAAC; back – GCCAGAAACCCATTACGATG; actin-3, forward – AATACTCTGTCTGGATTGAGGGTC, back – GCGGTGCTTCTTCTCTGAAAAATG; GUS, forward – ATCGTGGTGATTGATGAAAC; back – CGTCGCAGAACATTACATTG; McMIPF, forward – CGTTTGGGGCGTTTATTGG, back – AGAGCGAAGGCGGATGTTTC. Standard curves were established for all genes investigated using a series of amplicon dilutions. For each sample, mean TIP1;1 value was gotten from four runs of reaction, and mean value for other genes were obtained from two runs.
Total protein was isolated from group III, IV, V RNAi plants and wild-type plants according to the method of Schaller and DeWitt (1995). Western blotting was made with an ECL Western blotting system (Amersham Biosciences, Piscataway, NJ, USA). The TIP1;1 antibody, raised against a synthetic peptide corresponding to the C-terminal 9 amino acids of TIP1;1 was kindly provided by Drs N. Raikhel and E. Avila-Teeguarden, UC Riverside.
Water relations and leaf gas-exchange analysis
T0 plants selected for hygromycin resistance were transplanted to Scotts Metromix 360 potting media in 7.5-cm pots. Plants were grown under 16-h, long-day conditions with cool-white fluorescent lighting providing 100 μmol m−2 sec−1. Shoots from 27 well-watered 4-week-old plants were excised and placed on three balances for wild type (Col-0) and RNAi lines. Weight was logged by computer in 30-min intervals from each of six balances using Software Wedge v1.2 (TAL Technologies, Inc., Philadelphia, PA, USA). To eliminate variability resulting from plant size or dry weight, water content was standardized as a percentage relative to the initial shoot water content. It was calculated for each interval as: [(FWi−DW)/(FW0−DW)]×100, where FWi and FW0 are fresh weights for any given interval and original fresh weight, respectively, and DW is dry weight. Mean separation and standard errors were calculated using anova. To determine diurnal whole-plant water loss, pots containing 4-week-old seedlings were sealed with clear plastic wrap to prevent water loss from the soil. Three wrapped pots for each line were then placed on a balance, and weights were logged in 1-h intervals for 48 h. The rate of water loss was calculated on the basis of shoot dry weight and recorded as g H2O h−1 g DW−1. Leaf gas-exchange was measured in wild type (Col-0), RNAi category V and MIPF-complemented RNAi plants. Net CO2 assimilation rate (A) and stomatal conductance (gs) were measured on fully expanded leaves of six plants for each line using an ADC Bioscientific Leaf Chamber Analyzer, Model LCA-4 (ADC Bioscientific Ltd, Hoddesdon, UK). Comparisons of differences among mean values was by anova (α = 0.05).
Above-ground plant tissues (100–200 mg) were collected at the early bolting stage, and used for the extraction of metabolites, separated into polar and lipid phases, as described (Fiehn et al., 2000a). Derivatizations were according to an established protocol (Fiehn et al., 2000a,b). Injections were performed manually without the use of an autosamples. Samples were stored at −20°C before injection to avoid degradation. Metabolites were separated and analyzed by GC-MS analyses in the Department of Food Sciences, University of Illinois, using an HP6890 gas chromatograph and an HP5973 mass selective detector. Metabolite identification used the protocol and a metabolite library available at MPI Golm (http://www.mpimp-golm.mpg.de/mms-library/index-e.html) (kindly provided by Dr O. Fiehn). Quantification was obtained by establishing a standard curve, and then comparing the total signal of a compound with the C31 signal added to the extract and the standard curve signal.
The TIP1;1 coding region was cloned into the pRZ238 vector via the BamHI and SacI sites, and the CaMV 35S promoter in pRZ238 was replaced by the TIP1;1 promoter (1.5 kbp upstream of the ATG codon) via the NcoI and XhoI sites. The promoter-GFP-TIP1;1 fragment was cloned into the binary vector pRZ265 via the XhoI and SacI sites. The binary vector was transferred into A. tumefaciens, strain GV1101, together with the helper plasmid pSoup, and then transformed wild-type Arabidopsis (Col). Leaves from the KanR transformants were observed using an Olympus (Olympus America, Chicago, IL, USA) confocal microscope (College Veterinary Science, UIUC). Stacks of optical section imagines were recorded and assembled to generate a 3-D reconstruction imagine, using the FluoView 3.0 software. The vector construct used here is composed of various elements combined into pRZ238, a vector based on pGEM4Z backbone, assembled with a CaMV35S promoter, GFP and NOS-t (Jefferson et al., 1987; Restrepo et al., 1990; Sheen et al., 1995). Vector pRZ265 is based on pGreen II (Hellens et al., 2000).
Four-week-old RNAi plants (n = 4) from groups III and IV, together with WT plants growing alongside, were harvested and used to isolate total RNAs (Trizol; Invitrogen). Total RNAs (100 μg each) were purified (RNeasy; Qiagen), and converted into cDNA. The procedures for cDNA synthesis, labeling and hybridization to the Arabidopsis long-oligonucleatide array has been described (Inan et al., 2004). Two slides, including one dye swap, were used for each set of RNA sample. The experiment was repeated with a set of independently grown plants. Data representing the mean of four hybridizations, expressed as the log10 ratio of RNAi/wild type, are presented.
We thank Drs Vera Lozovaya and Jack Widholm (Department of Crop Sciences, UIUC) for discussions and encouragement, Dr Ray Zielinski (Department of Plant Biology, UIUC) for providing the pRZ265 and pRZ238 vectors, and Dr Oliver Fiehn, Max-Planck-Institut für Molekulare Physiologie, Golm, Germany, for providing the metabolite analysis library. We are especially grateful to Drs Natasha Raikhel and Emily Avila-Teeguarden (UC Riverside) for providing antiserum raised against a carboxy-terminal peptide of TIP1;1. The work was supported by institutional grants from UIUC and Purdue University. A progress report was presented at the 2003 ASPB meeting.