A new LxxxA motif in the transmembrane Helix3 of maize aquaporins belonging to the plasma membrane intrinsic protein PIP2 group is required for their trafficking to the plasma membrane.

Aquaporins play important roles in maintaining plant water status under challenging environments. The regulation of aquaporin density in cell membranes is essential to control transcellular water flows. This work focuses on the maize (Zea mays) plasma membrane intrinsic protein (ZmPIP) aquaporin subfamily, which is divided into two sequence-related groups (ZmPIP1s and ZmPIP2s). When expressed alone in mesophyll protoplasts, ZmPIP2s are efficiently targeted to the plasma membrane, whereas ZmPIP1s are retained in the endoplasmic reticulum (ER). A protein domain-swapping approach was utilized to demonstrate that the transmembrane domain3 (TM3), together with the previously identified N-terminal ER export diacidic motif, account for the differential localization of these proteins. In addition to protoplasts, leaf epidermal cells transiently transformed by biolistic particle delivery were used to confirm and refine these results. By generating artificial proteins consisting of a single transmembrane domain, we demonstrated that the TM3 of ZmPIP1;2 or ZmPIP2;5 discriminates between ER and plasma membrane localization, respectively. More specifically, a new LxxxA motif in the TM3 of ZmPIP2;5, which is highly conserved in plant PIP2s, was shown to regulate its anterograde routing along the secretory pathway, particularly its export from the ER.

Aquaporins are of major importance to plant physiology, being essential for the regulation of transcellular water movement during growth and development Gomes et al., 2009;Heinen et al., 2009;Prado and Maurel, 2013;Chaumont and Tyerman, 2014). Aquaporins are small membrane proteins consisting of six transmembrane (TM) domains connected by five loops (A-E), and N and C termini facing the cytosol (Fig. 1A). They assemble as homotetramers and/or heterotetramers in the membrane, with each monomer acting as an independent water channel (Murata et al., 2000;Fetter et al., 2004;Gomes et al., 2009). Aquaporins form a highly divergent protein family in plants (Chaumont et al., 2001;Johanson et al., 2001), and this work focuses on the maize (Zea mays) plasma membrane intrinsic protein (ZmPIP) family (Chaumont et al., 2001). The regulation of the subcellular localization of these proteins is a key process controlling their density in the plasma membrane (PM) and, hence, their physiological roles (Hachez et al., 2013).
PIPs were originally thought to be exclusively localized in the PM and were named accordingly (Kammerloher et al., 1994). However, recent experiments have shown that not all PIPs are located to the PM under all conditions, and that regulation of PIP subcellular localization is a highly dynamic process involving protein interactions (Boursiac et al., 2005(Boursiac et al., , 2008Zelazny et al., 2007Zelazny et al., , 2009Uehlein et al., 2008;Besserer et al., 2012;Luu et al., 2012). Figure 1. Swapping TM3 of ZmPIP2;5 with that of ZmPIP1;2 retains the protein in intracellular structures. A, Cartoons representing the chimeric proteins composed of ZmPIP2;5, in which each TM has been replaced by the corresponding TM from ZmPIP1;2. All proteins are drawn with the cytosolic domains facing down. ZmPIP2;5 and ZmPIP1;2 portions are shown in black and white, respectively. All chimeras were fused to the C terminus of mYFP, which is not displayed for clarity purposes. B, Confocal microscopy images of maize mesophyll protoplasts transiently coexpressing mYFP-tagged ZmPIP2;5-PIP1;2 TM chimeric proteins (green) and the ER marker mCFP:HDEL (cyan). FM4-64 was added as a PM marker (red). Arrowheads in image 13 indicate accumulation of the protein in punctate structures that are not labeled by mCFP:HDEL. The localization When expressed singly in maize leaf mesophyll protoplasts, fluorescently tagged ZmPIP1s and ZmPIP2s differ in their subcellular localization. ZmPIP1s are retained in the endoplasmic reticulum (ER), whereas ZmPIP2s are targeted to the PM (Zelazny et al., 2007). However, upon coexpression, ZmPIP1s are relocalized from the ER to the PM, where they perfectly colocalize with ZmPIP2s. This relocalization results from their physical interaction as demonstrated by Förster resonance energy transfer/ fluorescence lifetime imaging microscopy and immunoprecipitation experiments (Zelazny et al., 2007). These results indicate that ZmPIP2s, but not ZmPIP1s, possess signals that allow them to be delivered to the PM, and that hetero-oligomerization is required for ZmPIP1 trafficking to the PM. Interestingly, a diacidic motif (DxE, Asp-any amino acid-Glu) located in the N terminus of ZmPIP2;4, ZmPIP2;5, and Arabidopsis (Arabidopsis thaliana) AtPIP2;1 was shown to be required to exit the ER (Zelazny et al., 2009;Sorieul et al., 2011). Diacidic motifs interact with Secretory protein24, which is thought to be the main cargo-selection protein of the Coat proteinII complex that mediates vesicle formation at ER export sites (Miller et al., 2003). However, not all PM-localized PIP2s contain a diacidic ER export signal (Zelazny et al., 2009). In addition, swapping the N-terminal region of ER-retained ZmPIP1;2 with that of PM-localized ZmPIP2;5, which contains the functional diacidic motif, is not sufficient to trigger ER export of the protein (Zelazny et al., 2009). This result suggests that other export signals might be present in PIP2s and/or ER retention signals might be present in PIP1s elsewhere than in the N terminus.
To identify new signals regulating ZmPIP1 and ZmPIP2 protein trafficking along the secretory pathway, we used a protein domain swapping-based approach and identified the TM3 as an important region that discriminates between ER-retained ZmPIP1;2 and PM-localized ZmPIP2;5. Specific mutations in the TM3 region of ZmPIP2;5 allowed the identification of a new ZmPIP2-conserved LxxxA motif, which regulates its export from the ER.

TM3 Plays a Role in the Delivery of ZmPIP2s to the PM
To identify new trafficking signals, the loops and termini of PM-localized ZmPIP2;1 were first replaced, individually or in combinations, by the corresponding portions of ER-retained ZmPIP1;2 (Supplemental Fig. S1A). The loop and TM regions were defined on the basis of a multiple alignment of all of the ZmPIP sequences, and on the topological model of ZmPIP1;2 (Chaumont et al., 2001). Maize mesophyll protoplasts were transiently transformed with genetic constructs encoding the chimeric ZmPIPs fused downstream of the monomeric yellow fluorescent protein (mYFP; see the "Materials and Methods"). Previous studies demonstrated that tagging ZmPIPs with GFP variants at their N terminus does not affect their activity (Fetter et al., 2004;Besserer et al., 2012;Bienert et al., 2012), nor does it prevent N-terminal trafficking motifs to be functional (Zelazny et al., 2009). In plant cells, PM, ER, and cytosol are sometimes difficult to distinguish because of the presence of the central vacuole that occupies most of the cell volume and pushes all other cell compartments against the PM. For this reason, mCFP:HDEL and FM4-64 have been used throughout this study as widely accepted markers of the ER and the PM, respectively. The use of those markers confirmed previous results obtained in protoplasts demonstrating that ZmPIP1s are retained in the ER, whereas ZmPIP2s reach the PM (Fig. 1B, images 1-4; Supplemental Fig. S1B, images 1-8; Zelazny et al., 2007Zelazny et al., , 2009. Even though some chimeric proteins strongly labeled various intracellular structures, the PM remained clearly stained by all of them (Supplemental Fig. S1B). This led us to conclude that the N and C termini and the intracellular and extracellular loops of ZmPIP2;1 and ZmPIP1;2 did not contain dominant trafficking signals. However, one chimeric protein (mYFP:ZmPIP2;1-PIP1;2Mix, which contains the TM3 and TM4 of ZmPIP1;2; Supplemental Fig. S1A), was completely retained in intracellular structures (Supplemental Fig. S1B, images 33-36), colocalizing with the ER marker mCFP:HDEL. This observation points to a possible role for the TM domains in the routing of ZmPIPs to the PM.
A similar domain-swapping approach was used to assess the possible involvement of TMs in the regulation of ZmPIP trafficking to the PM. In this case, ZmPIP2;5 was used instead of ZmPIP2;1, because the former exhibited a more defined PM localization compared with ZmPIP2;1. Indeed, when expressed in protoplasts, mYFP:ZmPIP2;5 strongly labeled the PM, and the signal was weaker in the intracellular structures than that of mYFP:ZmPIP2;1 (compare Fig. 1B, images 1-4, with Supplemental Fig. S1B, images 1-4; Zelazny et al., 2007).
The six TMs of ZmPIP2;5 were individually replaced by those of ZmPIP1;2 (Fig. 1A). Because the length of the TM is known to influence the localization of type I membrane proteins along the secretory pathway (Brandizzi et al., 2002), the swapped regions were slightly longer than the predicted TMs. Figure 1B shows the subcellular localization of the mYFP-tagged chimeric proteins transiently expressed in maize mesophyll protoplasts. Strikingly, ZmPIP2;5, whose TM3 was replaced by that of ZmPIP1;2 (mYFP:ZmPIP2;5-TM3 PIP1;2 ; Fig. 1A), was unable to reach the PM (Fig. 1B, images 13-16), which is in contrast with the wild-type mYFP:ZmPIP2;5 (Fig. 1B, images 1-4). All of the other chimerae (Fig. 1B, images 5-12 and 17-28) sharply labeled the cell periphery and colocalized with the styrl dye FM4-64, showing that they were still, at least to some extent, able to reach the PM. The control experiment showed that the position of the fluorescent tag did not modify the localization of ZmPIP2;5, ZmPIP1;2, and ZmPIP2;5-TM3 PIP1;2 (Supplemental Fig. S2).
As shown by the partial colocalization with the ER marker mCFP:HDEL, mYFP:ZmPIP2;5-TM3 PIP1;2 was retained both in the ER (Fig. 1B, images 13-16) and in bright punctate intracellular structures (Fig. 1B, arrowheads in image 13). To investigate the nature of these structures, mYFP:ZmPIP2;5-TM3 PIP1;2 was transiently coexpressed with the Golgi marker ST:mCFP (Boevink et al., 1998;Batoko et al., 2000). Figure 1C shows the colocalization of mYFP:ZmPIP2;5-TM3 PIP1;2 and ST:mCFP in Golgi stacks (arrowheads), demonstrating that mYFP:ZmPIP2;5-TM3 PIP1;2 is retained in the Golgi apparatus in addition to the ER. To identify the amino acid residues responsible for the retention of mYFP:ZmPIP2;5-TM3 PIP1;2 in intracellular structures, each TM3 residue differing between ZmPIP2;5 and ZmPIP1;2 was individually mutated. For each position, the residue in ZmPIP2;5 was mutated into its ZmPIP1;2 counterpart ( Fig. 2A). The final three amino acid residues (S148, A149, and F150) of the exchanged region were mutated together (ZmPIP2;5SAF/ QGL). These mutated proteins were tagged with mYFP and expressed in protoplasts (Fig. 2B;Supplemental Fig. S3). For most of the mutants, the fluorescent signal in the intracellular structures was stronger than for the wild-type protein. However, all showed a sharp signal at the periphery of the cell, which colocalized with FM4-64, and were therefore able to reach the PM, at least partially. These results indicated that two or more residues were involved in the full intracellular retention observed for mYFP:ZmPIP2;5-TM3 PIP1;2 (Fig. 1).
To analyze more precisely which amino acid residues of the TM3 could together form a trafficking motif, a previously published homology model of the ZmPIP2;5 homotetramer was utilized ( Fig. 2C; Bienert et al., 2012). This model showed that the TM3 was located on the outer surface of the ZmPIP2;5 tetramer. Furthermore, among the 10 amino acids of ZmPIP2;5 that differed from ZmPIP1;2, V123, L127, V130, and A131 might possibly be involved in protein-protein or protein-lipid interactions in the membrane, because they were facing the lipid bilayer and were in close proximity to each other. Based on these observations, a mYFP:ZmPIP2;5 variant combining the four point mutations was generated (mYFP:ZmPIP2;5V123T/L127F/V130I/A131M). As hypothesized, this protein was totally absent from the PM (Fig. 2D, images 1-3). Interestingly, reintroducing either L127 (mYFP:ZmPIP2;5V123T/V130I/A131M) or A131 (mYFP:ZmPIP2;5V123T/L127F/V130I) was sufficient to restore a sharp signal in the PM (Fig. 2D, images 4-9), indicating that these amino acid residue exchanges were responsible for the full intracellular retention of mYFP:ZmPIP2;5-TM3 PIP1;2 . A construct combining both the L127F and A131M mutations was then generated (mYFP:ZmPIP2;5L127F/A131M). As shown in Figure 2D (images 10-12), this double mutation was sufficient to prevent the protein from reaching the PM. Quantification of the relative fluorescent signal in the PM confirmed these results ( Fig. 2E; Supplemental Fig. S4).

The Steady-State Subcellular Localization of ZmPIPs in Protoplasts Is Reached after 16 h
Some of the proteins of interest were localized in intermediate compartments of the secretory pathway (ER and/or Golgi) 16 h after protoplast transformation. To determine whether this localization was their steadystate destination or whether it represented intermediate trafficking steps, a 3-d time course experiment was performed (Supplemental Fig. S5). Protoplasts were transfected with genetic constructs encoding mYFP:ZmPIP2;5, mYFP:ZmPIP1;2, or mYFP:ZmPIP2;5L127F/A131M, and the subcellular localization of the proteins was observed 16, 40, and 64 h after transfection. Cycloheximide (CHX) was added after the first observation in order to follow the evolution of the pool of proteins observed after 16 h. FM4-64 was added to the protoplast suspension just before observation. Pearson's correlation coefficients between the yellow fluorescent protein (YFP) and FM4-64 signals were calculated to measure to which extent the proteins reached the PM. The localization of all three proteins of interest appeared not to change over time (Supplemental Fig. S5). Even 60 h after transfection, mYFP:ZmPIP1;2 and mYFP:ZmPIP2;5L127F/A131M were unable to reach the PM, whereas ZmPIP2;5 stayed in the PM at all time points.

Transient Transformation of Maize Leaf Epidermal Cells by Biolistic Particle Delivery
The above-described experiments were repeated using biolistic transient transformation of maize leaf epidermal cells (Fig. 3). The major advantage of using this technique was to express the chimeric proteins in intact living cells of known origin, still surrounded by their wall and neighboring tissue. As expected, when expressed alone, mYFP:ZmPIP2;5 labeled the cell periphery ( Fig. 3A), whereas mYFP:ZmPIP1;2 showed a network-like ER localization (Fig. 3B). However, upon coexpression, both proteins colocalized in the PM (Fig. 3D), confirming previous observations performed in maize mesophyll Figure 2. The double mutant mYFP:ZmPIP2;5L127F/A131M is retained in intracellular structures. A, Alignment of the TM3 region exchanged between ZmPIP2;5 and ZmPIP1;2. The residues that differ between both proteins are highlighted in gray. Mutations introduced in ZmPIP2;5 are indicated below the alignment. The dashed-line box highlights the predicted TM3. The alignment was generated using ClustalW2 (http://www.ebi.ac.uk/Tools/msa/clustalw2/). B, Confocal microscopy images of maize mesophyll protoplasts transiently expressing wild-type or mutated mYFP:ZmPIP2;5 proteins (green). FM4-64 was added as a PM marker (magenta). C, Homology model of the ZmPIP2;5 homotetramer. One subunit is highlighted in green, with its TM3 shown in blue. The residues of ZmPIP2;5 that differ from ZmPIP1;2 in this region are represented in yellow and labeled. D, Confocal microscopy images of maize mesophyll protoplasts transiently expressing YFP-tagged mutated ZmPIP2;5 proteins (green) chosen on the basis of the model in C. FM4-64 was added as a PM marker (magenta). E, Quantification of the relative mYFP fluorescence intensity in the PM of protoplasts expressing wild-type or mutated mYFP:ZmPIP2;5 proteins as shown in D. The y axis shows the fluorescence ratio between the PM and the whole cell. Error bars are confidence intervals (a = 0.05). The letter above each bar represents statistical classes determined by a Bonferroni test (P , 0.001). The localization patterns of the proteins of interest are representative of a total of at least 29 cells from a minimum of two independent experiments (B and D). Fluorescence calculations in E were performed using the same data set. Bar = 5 mm.

DISCUSSION
This work aimed to identify new protein motifs regulating the subcellular localization of ZmPIPs. By analyzing the localization of fluorescently tagged ZmPIP1-ZmPIP2 chimeric or mutated proteins in two different homologous systems, we showed that the TM3 plays an important role in regulating ZmPIP delivery to the PM. Furthermore, we identified two residues within TM3 that are critical for the routing of ZmPIPs between the ER and the PM.

TM3 Is Involved in ZmPIP Trafficking
The replacement of the third TM of ZmPIP2;5 with the corresponding region of ZmPIP1;2 strikingly prevented PM delivery of the chimeric protein (Fig. 1). Point-mutation analysis of the diverging residues in the TM3 of ZmPIP2;5 and ZmPIP1;2 showed that the combination of only two single point mutations (mYFP: ZmPIP2;5L127F/A131M) induced the complete intracellular retention phenotype observed for ZmPIP2;5-TM3 PIP1;2 (Fig. 2). Trafficking of those mutants to the PM is fully blocked and not simply delayed, as mYFP: ZmPIP2;5L127F/A131M is still exclusively found in intracellular structures 3 d after protoplast transfection (Supplemental Fig. S5).
Trafficking signals are usually located in the N-or Cterminal soluble regions of membrane proteins (Kappeler et al., 1997;Nishimura and Balch, 1997;Sevier et al., 2000;Ma et al., 2001;Otte and Barlowe, 2002) or in cytosolic loops (Takano et al., 2010). The accessibility of the sorting motifs from the cytoplasm was previously believed to be crucial for the interaction of membrane cargo proteins  . Swapping the TM3 of ZmPIP1;2 with that of ZmPIP2;5 does not allow the protein to reach the PM. A, Cartoons representing ZmPIP1;2-N PIP2;5 , ZmPIP1;2-TM3 PIP2;5 , and ZmPIP1;2-N+TM3 PIP2;5 . ZmPIP1;2 and ZmPIP2;5 are shown in white and black, respectively. The proteins are drawn with the cytosolic domains facing down. They were fused to the C-terminal end of the mYFP, which is not displayed for clarity purposes. B, Confocal microscopy images of maize mesophyll protoplasts transiently coexpressing mYFP:ZmPIP1;2, mYFP:ZmPIP1;2-N PIP2;5 , mYFP:ZmPIP1;2-TM3 PIP2;5, or mYFP:ZmPIP1;2-N+TM3 PIP2;5 (green) and the ER marker mCFP:HDEL (magenta). Arrowheads in images 4 and 10 show punctate structures that are seen only in the YFP channel. The localization patterns of the proteins of interest are representative of a total of at least 27 cells from three independent experiments. Bar = 5 mm.
with trafficking regulators (Barlowe, 2003;Miller et al., 2003;Mossessova et al., 2003;Contreras et al., 2004;Robinson et al., 2007;Sato and Nakano, 2007;Bassham et al., 2008;Sieben et al., 2008). However, more and more examples of the regulation of membrane protein sorting by alternative mechanisms (Springer et al., 2014), and specifically by TMs (Cosson et al., 2013), have been reported. For instance, the TM6 of the tonoplast aquaporin a-tonoplast intrinsic protein from bean (Phaseolus vulgaris) contains sufficient information for its transport to the vacuolar membrane (Höfte and Chrispeels, 1992). The rice secretory carrier membrane protein1 relies on its TM2 and TM3 for Golgi export, and on its TM1 for trans-Golgi network (TGN) to PM trafficking (Cai et al., 2011). Together with our data, these results provide evidence for the involvement of TMs in intracellular sorting processes. The exact mechanisms underlying TM-based sorting are not yet fully understood, but likely involve recognition by sorting receptors and targeting to specific membrane subdomains (Cosson et al., 2013).
The length of the TM has been shown to regulate the progression of single TM integral membrane proteins through the secretory pathway in plants, vertebrates, and fungi (Brandizzi et al., 2002;Ronchi et al., 2008;Sharpe et al., 2010). Whether such a TM length-dependent sorting process might, to some extent, explain the results obtained here with single TM reporters (Fig. 5) was investigated. A possible hypothesis could be that the TM3 of ZmPIP1;2 is shorter than that of ZmPIP2;5, resulting in a different progression through the secretory pathway. Interestingly, TM3 (and TM4) of ZmPIP1s are generally predicted to be shorter than their ZmPIP2 counterparts by the algorithm TMHMM 2.0 (http://www.cbs.dtu.dk/ services/TMHMM/; Table I). This first prediction seemed to confirm our hypothesis. However, the calculated length of the TM3 of both ZmPIP1;2 and ZmPIP2;5 in the context of the single TM YFP reporters was strictly identical (Table I). Therefore, the different destination of our single TM proteins is not likely to be the consequence of a different TM length. It has to be noted that the output of the TMHMM 2.0 algorithm differs according to the sequence context of the TM helix. Whether the length of TM regions actually changes depending on their sequence context or whether this is a calculation artifact is not known. The question is even more complicated regarding full-length proteins. TM segments in aquaporins are slightly tilted in the membrane (Murata et al., 2000;Törnroth-Horsefield et al., 2006). In a membrane of a given thickness, aquaporin TM helices are therefore likely to be longer than those of single-pass membrane proteins. Although examples of TM length-based sorting have been reported for single TM proteins, whether and how such a mechanism could occur in multiple TM proteins has, to our knowledge, not been addressed. Because TM length can regulate partitioning of single TM proteins in different lipid subdomains of the ER and, therefore, their export to the Golgi (Ronchi et al., 2008), whether similar mechanisms occur in multispanning membrane proteins should be studied.
Although the shorter predicted TM3 (18 amino acids) in ZmPIP1s might seem responsible for their retention in the ER, the TM3 of ER-retained ZmPIP1;6 ( Fig. 4) had a predicted length similar to PIP2 TM3 (Table I). A length effect could therefore not by itself explain the intracellular retention of mYFP:ZmPIP2;5-TM3 PIP1;6 . Furthermore, in the context of ZmPIP2;5-TM3 PIP1;2 , the length of the TM3 was predicted to be 23 amino acid residues (Table I); however, the protein was not able to reach the PM (Figs. 1 and 3). Together, these predictions indicate that regulation of the subcellular localization of Figure 7. Water transport activity of the ZmPIP2;5-TM3 PIP1;2 chimera. Water permeability coefficients of X. laevis oocytes injected with water (negative control), or cRNA encoding ZmPIP2;5 (positive control), ZmPIP2;5 and ZmPIP1;2 (positive control for the synergistic effect), ZmPIP2;5-TM3 PIP1;2 , or ZmPIP2;5-TM3 PIP1;2 and ZmPIP2;5. Error bars are confidence intervals (a = 0.05). The letter above each bar represents statistical classes determined by a Bonferroni test (P , 0.001). At least 12 oocytes injected with each cRNA were assayed, except for the negative control (water), for which the swelling of eight cells was recorded. Three independent experiments were performed. The mutated ZmPIP2;5 proteins might be retained in intracellular structures as a consequence of an improper folding. Water transport activity assays were performed to rule out this possibility (Fig. 7). When expressed alone, ZmPIP2;5-TM3 PIP1;2 was unable to induce an increase in oocyte P f . However, when ZmPIP2;5-TM3 PIP1;2 was coexpressed with the wild-type ZmPIP2;5, the P f was significantly higher than the P f of ZmPIP2;5-expressing cells, similarly to what was previously reported with ZmPIP1;2 (Fetter et al., 2004). The mutated protein was therefore still able to interact with wild-type ZmPIP2;5 proteins to induce this synergistic P f increase. A minor conformational change, turning ZmPIP2;5 into a ZmPIP1-like isoform, might happen as a consequence of TM3 exchange, but the overall structure and topology must still be consistent with those of a ZmPIP to allow water to be transported. These results, in addition to the fact that the trafficking defects were specifically observed for TM3 and not for other TMs of ZmPIP2;5, suggest that the altered localization of mYFP:ZmPIP2;5-TM3 PIP1;2 is due to specific trafficking properties of the TM3, rather than to a nonspecific response retaining the mutated protein inside the cell for degradation. The fact that mYFP:ZmPIP2;5L127F/A131M was still detected in protoplasts 3 d after transformation and in the presence of CHX (Supplemental Fig. S5) further supports this conclusion.

The TM3 of ZmPIPs Regulates ER Export
The chimeric proteins that were unable to reach the PM as a consequence of TM3 exchange were located both in the ER and in punctate structures, some of them being Golgi bodies (Figs. 1, 3 and 4), raising the question of whether the TM3 regulates ER or Golgi export of ZmPIPs. If PIP2-TM3 PIP1 chimeras are impaired in their ER export function, Golgi particles might be partially stained as a consequence of ER leakage, due to the presence of the ER export diacidic motif in the chimeric proteins. In support of this hypothesis, fewer punctate structures were detected for mYFP:ZmPIP2;1-TM3 PIP1;2 than for mYFP:ZmPIP2;5-TM3 PIP1;2 (Supplemental Fig. S6; compare Figs. 3, E and F, and 4F), in accordance with the occurrence of a functional ER export motif in the N terminus of ZmPIP2;5 but not ZmPIP2;1. The localization of mYFP:TM3 ZmPIP1;2 in the ER, and not in punctate structures (Fig. 5), also sustains this hypothesis. In a similar manner, mutation of the diacidic ER export motif of the Golgi nucleotide sugar transporter1 caused the retention of the protein in the ER, but the protein was still partially able to escape to the Golgi and post-Golgi punctate structures (Hanton et al., 2005). These results suggest that the TM3 regulates ER export rather than Golgi export of ZmPIPs. However, because the intense punctate structures were stained by mYFP:ZmPIP2;5-TM3 PIP1;2 and mYFP:ZmPIP2;5L127F/A131M (Figs. 1-3), it is possible that TM3 might, to some extent, take part in Golgi export regulation.
The observation that some of the punctate structures labeled by mYFP:ZmPIP2;5-TM3 PIP1;2 and mYFP: ZmPIP2;5L127F/A131M did not colocalize with the Golgi marker ST:mCFP suggests that those proteins were located in other subcellular compartments, in addition to the ER and Golgi apparatus. This hypothesis is further supported by the difference in the size of the punctate structures observed (Fig. 1B, image 13). Because the proteins were retained in the early steps of the secretory pathway, it is particularly tempting to speculate that the proteins accumulate at endoplasmic reticulum export sites (ERESs), leading to an intense labeling of these structures, even if we cannot exclude a TGN location. In that case, the protein would be distributed almost all along the secretory pathway (ER, Golgi, and TGN). Export Signal of ZmPIP2;5 or Retention Signal of ZmPIP1;2?
Our domain-exchange experiments allowed us to identify the critical role of the TM3 in the trafficking of ZmPIPs, but whether this domain contains an anterograde signal of ZmPIP2;5 or a retrograde signal of ZmPIP1;2 remains an open question. The replacement of the TM3 of ZmPIP1;2 with the corresponding region of ZmPIP2;5 or the double mutation F137L/M141A did not cause the export of the protein out of the ER ( Fig. 6; Supplemental Fig. S7), as would be expected if an ER-retention motif was mutated, suggesting that the new trafficking motif is an LxxxA anterograde signal of ZmPIP2;5 rather than an FxxxM ER retention signal for ZmPIP1;2. In support of this hypothesis, mYFP:ZmPIP2;5-TM3 PIP1;6 was absent from the PM even though ZmPIP1;6 does not contain the ZmPIP1;2 FxxxM motif (Fig. 4), demonstrating that the ER retention was due to the absence of the LxxxA motif of ZmPIP2;5. In addition, ZmPIP2;1, which contains the TM3 LxxxA motif but does not contain the N-terminal diacidic motif, was able to reach the PM, pointing to a role for the LxxxA motif in ER export. Altogether, these results confirm that the trafficking motif identified in this study is an LxxxA ER export motif located in the TM3 of ZmPIP2;5. Hanton et al. (2005) suggested that diacidic ER export motifs are dominant over TM-based ER retention signals. According to this theory, one would expect that proteins containing the diacidic motif of ZmPIP2;5 and the TM3 of ZmPIP1;2 (mYFP:ZmPIP2;5-TM3 PIP1;2 ; mYFP:ZmPIP1;2-N PIP2;5 ; Fig. 6; Supplemental Fig. S7; Zelazny et al., 2009) would be able to reach the PM. However, the results showed that these proteins were retained inside the cell, suggesting that the TM3 of ZmPIP2;5 contains an LxxxA ER export signal that is required, together with the N-terminal signal, to reach the PM. The reason why ZmPIP1;2 did not reach the PM even when fused to both ER export motifs is unknown. Another, yet-unidentified motif might be necessary, in addition to the diacidic and LxxxA motifs, for ER exit. Alternatively, the trafficking signals present in the N terminus and the TM3 of ZmPIP2;5 might not be fully functional when present in ZmPIP1;2 due to a nonoptimal structural context.
The critical residues identified in the TM3 of ZmPIP2;5 are highly conserved among plant PIPs. The L127 and A131 residues of ZmPIP2;5 are conserved in all maize PIP2s, and all ZmPIP1s contain Phe and Met residues at these positions, except for ZmPIP1;6 in which the Phe is replaced by a Tyr residue. The LxxxA motif is present in the TM3 of 18 of 23 PIP2 isoforms from rice (Oryza sativa), maize, and Arabidopsis (Bansal and Sankararamakrishnan, 2007). This motif is present in some tonoplast intrinsic proteins and a few Nodulin26-like intrinsic proteins, but is replaced by IxxxA in all uncharacterized X intrinsic proteins. On the other hand, an LxxxA sequence is also found in the TM3 of some small basic intrinsic proteins, which are reported to be localized in the ER (Ishikawa et al., 2005), suggesting that the functionality of LxxxA sequences in intracellular sorting relies on the molecular context.
Working Model for a TM-Based LxxxA Sorting Signal in ZmPIP2;5 Figure 8 presents four hypothetical models describing how TM-based trafficking motifs, as identified in this study, could integrate into well-known secretory pathway mechanisms. In the first two scenarios (Fig. 8,  A and B), the TM trafficking motif acts as a proteinprotein interaction motif with another membrane protein. In a first case (Fig. 8A), this interacting protein acts as an intermediate between the protein of interest and a , which in turn interacts with a transport protein (orange circle) that exports the complex from its compartment. The membrane of the compartment is shown in light gray. B, Proteinprotein interaction: conformational change. The protein of interest interacts with another membrane protein via the TM-based motif. This interaction induces a conformational change, which exposes a classical, cytosol-exposed, export motif (yellow circle). A transport protein is then recruited by classical mechanisms. C, Protein-lipid interaction: membrane domain bulk-flow. The TM-based motif segregates the protein of interest in a specific, export-competent, domain of the membrane (dark gray). All proteins present in this domain are exported due to the interaction of a transport protein with classical export signals present on some of the proteins present in this membrane domain. D, Proteinlipid interaction: membrane domain segregation and conformational change. The TM-based motif segregates the protein of interest in a specific, export-competent, domain of the membrane. This induces a conformational change, releasing a classical, cytosol-exposed, sorting signal. The transport machinery is then recruited according to well-known mechanisms.
cargo-selecting protein (e.g. Secretory24 in the case of ER export). In the second scenario, the interaction induces a conformational change in the protein of interest, making a classical soluble trafficking motif (e.g. the N-terminal diacidic motif of ZmPIP2;5) accessible to the export machinery (Fig. 8B). The third scenario relies on protein-lipid interaction. The presence of the TM trafficking signal causes the protein of interest to segregate in a specific, export-competent domain of the membrane (e.g. the ERESs, in the case of ER export; Fig. 8C). Finally, the fourth model combines scenarios 2 and 3. The protein reaches a specific membrane subdomain due to a protein-lipid interaction. Upon interaction with lipids, the protein undergoes a conformational change that releases a classical export motif (Fig. 8D).
In the case of the ER export of ZmPIP2;5, the second hypothesis seems particularly attractive. Given previous results in regard to ZmPIP trafficking and interaction (Zelazny et al., 2007(Zelazny et al., , 2009, it is tempting to assume that ZmPIP homotetramerization or heterotetramerization could be involved in this process. However, as depicted in Figure 2C, the LxxxA motif of ZmPIP2;5 faces the membrane side of the PIP tetramer and therefore cannot be directly involved in tetramerization events. On the other hand, the fourth hypothesis (Fig. 8D) could partially explain our results. The LxxxA motif could direct ZmPIP2;5 to an export-competent domain of the ER membrane, most likely the ERES. There, the N-terminal diacidic motif could be released, recruiting Coat proteinII particles for export toward the Golgi (Fig. 8D). However, no evidence for a ZmPIP2;5-lipid interaction is available. Finally, none of our four scenarios can explain how the single TM reporter mYFP:TM3 ZmPIP2;5 is able to reach the PM in the absence of a diacidic motif (Fig. 5). Our current research aims at discriminating between these hypotheses and refining this working model. ZmPIP ER export regulation might depend on other more complex factors, such as the number of export-competent units among a heterotetramer.

Genetic Constructs
Complementary DNAs encoding ZmPIP1;2, ZmPIP1;6, ZmPIP2;1, and ZmPIP2;5 were amplified by PCR and cloned into the uracil-excision vectors pCAMBIA2300 35S N-terminal mYFP and/or pCAMBIA2300 35S N-terminal monomeric cyan fluorescent protein (mCFP; Nour-Eldin et al., 2006;Bienert et al., 2011). Genetic constructs encoding the chimerical proteins were created using the USER fusion method (Geu-Flores et al., 2007). The same methods were used to clone the sequences of interest into the pNB1u vector (Nour-Eldin et al., 2006) for water transport assays in Xenopus laevis oocytes, and to transfer constructs encoding ST: GFP and GFP:HDEL into the mYFP and mCFP USER vectors (Bienert et al., 2011). Point mutations were created by overlapping PCRs. The primers used are listed in Supplemental Table S1. All constructs were verified by sequencing. The plasmids were amplified in Escherichia coli and purified using the Nucleobond PC 500 kit (Macherey-Nagel) following the manufacturer's instructions.

Plant Materials
Maize (Zea mays) B73 seeds were germinated at 28°C for 2 d. The seedlings were transferred to soil and grown under an 8-h-dark/16-h-light regime at 25°C for 24 to 48 h, and then transferred to dark conditions for 5 to 7 d. The median portion of the third leaf was used for protoplast preparation or leaf epidermal cell transformation.

Transient Transfection
Maize mesophyll protoplasts were prepared and transiently transfected as previously described (Zelazny et al., 2007). For maize leaf epidermal cells transfected by biolistic particle delivery, plasmid DNA (1-2 mg) was precipitated on 480 mg of 0.6-mm diameter gold beads (Bio-Rad) and the leaves were bombarded under a 28 inches Hg vacuum (95 kPa), using a 1100 PSI (7600 kPa) rupture disk, at a shooting distance of 3 cm with a Biolistic PDS1000/He device (Bio-Rad). The leaf pieces were transferred on solid Hoagland medium and incubated for 16 to 40 h at 25°C under dark conditions.

Microscopy and Image Processing
Confocal laser scanning microscopy experiments were performed using a LSM 710 microscope (Zeiss). Zen 2010 (Zeiss) was used for image acquisition. mCFP and mYFP were excited at 445 and 514 nm, and detected from 450 to 510 nm and from 520 to 620 nm, respectively. FM4-64 (Life Technologies) was added at a concentration of 16 mM for a duration of 5 min for PM labeling of protoplasts. FM4-64 was excited at 514 nm and detected from 600 nm to 760 nm. A 363 (numerical aperture = 1.4) oil immersion objective and a 340 (numerical aperture = 1.2) water immersion objective were used for protoplast and leaf cell experiments, respectively. ImageJ (http://rsbweb.nih.gov/ij/) and Fiji (http://fiji.sc) were used for image processing. Maximum displayed values were set to 100 to 200 to ensure an optimal display both on a computer screen and as a printout. When necessary, the images were smoothed by the application of a median filter. The relative protoplast PM fluorescence signal was calculated as the ratio of PM to total fluorescence, as described in . Briefly, the outer limit of the fluorescent signal was converted into a first region of interest (ROI), representing the total fluorescence of the cell. A second ROI, reflecting intracellular fluorescence, was created by narrowing the first ROI 0.35 mm in diameter. The fluorescent signal in the PM was calculated as the difference between the overall signal from the whole cell (first ROI), and the signal originating from intracellular structures (second ROI).
To measure the evolution of the localization of the proteins of interest over time, Pearson's R correlation coefficients (above threshold) between YFP and FM4-64 signals were calculated with the Coloc 2 plugin included in the Fiji package (http://fiji.sc). Nontreated samples were observed 16, 40, and 64 h after protoplast transformation. CHX was added at a concentration of 50 mM (from a 200 mM stock in ethanol; Luu et al., 2012) after the first observation (20 h after transfection) to block protein synthesis. CHX-treated samples were observed 40 and 64 h after protoplast transfection.

Complementary RNA Synthesis and Water Transport Assays in Oocytes
Complementary RNA (cRNA) synthesis, injection of oocytes, oocyte swelling assays, and calculation of the membrane water permeability coefficient (P f ; m/s) were performed as described in Fetter et al. (2004).

Statistical Analyses
Relative PM fluorescence, Pearson's coefficient, and oocyte P f data were submitted to a one-way ANOVA, followed by a Bonferroni post test comparing all data sets, using PRISM 3.0 (http://www.graphpad.com/scientific-software/prism/). Confidence intervals with a = 0.05 were calculated and displayed on bar charts using Microsoft Excel. Three independent experiments were performed.

Supplemental Data
The following materials are available in the online version of this article.
Supplemental Figure S1. TMs are involved in the trafficking of ZmPIPs to the PM.
Supplemental Figure S4. Quantification of the signal in the PM of protoplasts expressing mutated mYFP:ZmPIP2;5 proteins.
Supplemental Table S1. PCR primers used to create the genetic constructs encoding chimerical and mutated proteins.