Transmembrane Helices 2 and 3 Determine the Localization of Plasma Membrane Intrinsic Proteins in Eukaryotic Cells

In plants, plasma membrane intrinsic protein (PIP) PIP1s and PIP2s mediate the transport of disparate substrates across plasma membranes (PMs), with a prerequisite that the proteins correctly localize to the PMs. While PIP2s can take correct localization by themselves in plant cells, PIP1s cannot unless aided by a specific PIP2. Here, we analyzed the localization of the Arabidopsis aquaporins, AtPIP1s, AtPIP2;4, and their mutants in yeast, Xenopus oocytes, and protoplasts of Arabidopsis. Most of AtPIP2;4 localized in the PM when expressed alone, whereas AtPIP1;1 failed to realize it in yeast and Xenopus oocytes. Switch of the transmembrane helix 2 (TM2) or TM3 from AtPIP1;1 to AtPIP2;4 disabled the latter’s PM targeting activity. Surprisingly, a replacement of TM2 and TM3 of AtPIP1;1 with those of AtPIP2;4 created a PM-localized AtPIP1;1 mutant, 1;1Δ(TM2+TM3)/2;4(TM2+TM3), which could act as a water and hydrogen peroxide channel just like AtPIP2;4. A localization and function analysis on mutants of AtPIP1;2, AtPIP1;3, AtPIP1;4, and AtPIP1;5, with the same replaced TM2 and TM3 from AtPIP2;4, showed that these AtPIP1 variants could also localize in the PM spontaneously, thus playing an inherent role in transporting solutes. Sequential and structural analysis suggested that a hydrophilic residue and a defective LxxxA motif are modulators of PM localization of AtPIP1s. These results indicate that TM2 and TM3 are necessary and, more importantly, sufficient in AtPIP2 for its PM localization.


INTRODUCTION
Aquaporins (AQPs) are transmembrane channel proteins with a broad variety of biological functions in all kingdoms. AQPs were found to facilitate water diffusion across cell membranes when expressed in Xenopus laevis oocytes (Agre et al., 1993). In recent years, some other uncharged small solutes, such as hydrogen peroxide (H 2 O 2 ), glycerol, urea, and ammonia, are also proved to be the substrates of AQPs (Soria et al., 2010;Benga, 2012;Kirscht et al., 2016;Tian et al., 2016). Furthermore, AQPs play important roles in resistance regulation, signal transduction, nutrient uptake, and transduction in various organisms (Chaumont et al., 2017).
Plant AQPs include plasma membrane intrinsic proteins, nodulin26-like intrinsic proteins (NIPs), the small basic intrinsic proteins (SIPs), the tonoplast intrinsic proteins (TIPs), and the poorly characterized X intrinsic proteins (XIPs). (Chaumont et al., 2001;Johanson et al., 2001;Danielson and Johanson, 2008). PIPs are divided into two major groups, PIP1 and PIP2, which share about 80% amino acid identity. Despite such a high similarity, PIP1s and PIP2s show different transmembrane water fluxes when expressed in X. laevis oocytes (Kammerloher et al., 1994;Chaumont et al., 2000;Yaneff et al., 2015). In correspondence with this, PIPs show a difference in subcellular localization between PIP1 and PIP2 groups, which explains the difference of water transporting activity in X. laevis oocytes (Fetter et al., 2004;Yaneff et al., 2015;Bienert et al., 2018). PIP2s are able to massively traffic to the plasma membrane of X. laevis oocytes and serve as functional transmembrane channels, while PIP1s showed little localization in the PM and a significant lower transporting capacity if expressed alone (Kammerloher et al., 1994;Fetter et al., 2004).
Generally, the newly synthesized and properly folded PM proteins will be transported from endoplasmic reticulum (ER) to Golgi and finally to PM via the vesicle system (Demmel et al., 2011;Liu and Li, 2014;Geva and Schuldiner, 2014). However, the PIP1s are going to be detained in intracellular membranes, especially endoplasmic reticulum membranes (Zelazny et al., 2007;. In PIP1s, the missing PM trafficking signal in transmembrane helix 3 makes it hard to be successfully secreted from ER (Chevalier and Chaumont, 2015). Interestingly, when expressed together with PIP2s in oocytes, PIP1s successfully localized in the PM just like PIP2s (Fetter et al., 2004). PIP1s interact with PIP2s and form stable heterotetramers. The heterotetramerization between PIP1s and PIP2s enables PIP1s to be secreted along with PIP2s from ER to PM (Zelazny et al., 2007;Bienert et al., 2018). The massive PM localization brings PIP1s with the normal function of transporting water like PIP2s (Berny et al., 2016;Bienert et al., 2018). Therefore, PIP1 group owes an inherent capacity to transport substrates. But trafficking to the correct position is the prerequisite before PIP1 is able to carry out its full function.
Diacidic DxE (D, aspartic acid; E, glutamic acid; x, uncertain amino acid residue), buried within the N-terminal region, is characterized as a PM trafficking motif in plant PIPs (Zelazny et al., 2009;Sorieul et al., 2011). Wild-type ZmPIP2;4 and ZmPIP2;5 targeted to PM but are retained in ER if the DxE motif is mutated. However, the replacement of the N-terminal region of ZmPIP1;2 with that of ZmPIP2;5 which contains a DxE motif is not enough to enable ZmPIP1;2 in targeting to PM like ZmPIP2;5 (Zelazny et al., 2009). LxxxA (L, leucine; x, undetermined amino acid residue; A, alanine) in TM3 of ZmPIP2;5 is also essential in ER-to-Golgi trafficking, and considered to be an ER export signal . LxxxA motif mutant is not secreted from the ER. Nevertheless, replacement of both N-terminal and TM3 region of ZmPIP1;2 with those of ZmPIP2;5 did not yield successful PM localization . Some phosphorylation sites, such as S280 and S283 in the C-terminus of Arabidopsis PIP2;1, are also identified as candidates that can affect PIPs trafficking (Prak et al., 2008). In addition, some E3 ubiquitin ligases localized in the ER membrane could play a role in AtPIP2;1 trafficking by posttranslational modification (Lee et al., 2009). Moreover, the PM-localized SNARE isoform SYP121 regulates the transfer of PIPs from the vesicles to the PM (Besserer et al., 2012;Hachez et al., 2014).
However, although several motifs are characterized as affecting the localization of PIP2 group, what really determines the defect of PIP1 group is not fully understood. Here, we took AtPIP1;1 (AT3G61430) and AtPIP2;4 (AT5G60660) from Arabidopsis thaliana ecotype Col-0 on behalf of PIP1 isoforms and PIP2 isoforms respectively to investigate the contributing factors of their localization differences. In A. thaliana, there are 13 AtPIPs isoforms, 5 AtPIP1 isoforms, and 8 AtPIP2 isoforms, which are typical AQPs with six TMs and five loops. AtPIP2;4 is considered to be a water channel like other AtPIP2s and was also found to play a role in hydrogen peroxide transporting and response to abscisic acid (Kammerloher et al., 1994;Lewis et al., 1997;Dynowski et al., 2008;Kline et al., 2016). As previously mentioned, AtPIP1;1, just like other PIP1s, need to heterotetramerize with AtPIP2s to normally localize to the PM (Zelazny et al., 2007;Chevalier and Chaumont, 2015). The localization of PIPs will determine the function of the proteins and even determine the fate of the host cells. Hence, it is imperative to uncover the foremost factors that cause the difference in behavior between PIP1s and PIP2s.

RESULTS
AtPIP1;1 Is Unable to Localize in the PM Like AtPIP2;4 When Expressed Alone To investigate the localization of AtPIP1;1 and AtPIP2;4 in different situations, AtPIP1;1 and AtPIP2;4 were expressed in different eukaryotic cells. AtPIP1;1 and AtPIP2;4 were fused with an enhanced green fluorescent protein (eGFP) and transformed into Saccharomyces cerevisiae NMY51 cells to examine the localization of each one when expressed alone. Confocal images showed that most of AtPIP2;4-eGFP reached the PM of S. cerevisiae cells, while AtPIP1;1-eGFP gathered in the cytoplasm and the fluorescent intensity was significantly lower, which meant most of the expressed proteins might be degraded due to the unstable form and failed PM localization ( Figure 1A upper). Xenopus oocyte expression of cRNA coding AtPIP1;1-eGFP or AtPIP2;4-eGFP gave similar results as that for expression in yeast cells ( Figure 1A middle). The only difference in oocytes is that the fluorescence of AtPIP2;4 was all at the PM. The immunoblot of the PM proteins and the intracytoplasmic proteins (including cytosolic proteins and intracellular membrane proteins) from AtPIP1;1-6xHis-or AtPIP2;4-6xHis-transformed S. cerevisiae cells showed that most of AtPIP2;4-6xHis was in the PM, while AtPIP1;1-6xHis gave a weak signal in the CP fraction and an almost imperceptible signal in the PM fraction ( Figure 1B). In fact, there was no significant difference in mRNA levels among AtPIP1;1-6xHis, AtPIP2;4-6xHis, and other AtPIP1;1 or AtPIP2;4 variants (Supplemental Figure 1). In the oocyte expression experiments, all the proteins could be detected by immunoblotting (Supplemental Figure 2). However, in Xenopus oocyte expression, AtPIP1;1 or AtPIPs mutants seemed to be soluble in cytosolic space which might be confusing, since PIPs should integrate into membranes ( Figures 1B, 2C, 3C, and 5A). In fact, the ER cluster massively distributes in cytoplasm of Xenopus oocytes (Terasaki et al., 2001). The cytosolic fluorescent signals might be due to the widely distributed ER membrane system in Xenopus oocytes.
The examinations in S. cerevisiae and X. laevis oocytes indicated that AtPIP2;4 can localize in PM when expressed alone, while nearly all the AtPIP1;1 failed to realize it. However, the AtPIP1;1 and AtPIP2;4 fused with a yellow fluorescent protein (YFP) showed transformed protoplasts of (B) Immunoblot of plasma membrane proteins (PM fraction) and intracytoplasmic proteins (including cytosolic proteins and intracellular membrane proteins, CP fraction), extracted from AtPIP1;1-6xHis-or AtPIP2;4-6xHis-transformed yeast cells, with the antibody of anti-6xHis (HT501). (C) Interaction between AtPIP1;1 and other AtPIPs using split ubiquitin yeast two hybrid (SUB Y2H) system showing AtPIP1;1 interacted with all the AtPIPs. The transformed yeast cells were series-diluted, and 5 ml of the yeast suspension was spotted on the SD-WLAH plates.
A. thaliana Col-0 with almost all the fluorescent signal at the PM ( Figure 1A nether). It was reported that PIP1 could heteroligomerize with PIP2 to realize its PM localization (Zelazny et al., 2007). Therefore, the eight expressed AtPIP2s in wild-type Arabidopsis might carry the transiently expressing AtPIP1;1 to the PM. Other than this, the PIPs were promoted by CaMV 35S promoter, which could cause a massive overexpression that might lead to an artifact result. To validate the former assumption, a split ubiquitin yeast two hybrid system (SUB Y2H system) designed for testing interaction of membrane proteins (Obrdlik et al., 2004) was introduced to our study. For testing the strength of their oligomerization, 10-fold series dilution of the transformed yeast was conducted. The results demonstrated that AtPIP1;1 had the capacity of heteroligomerization with all the AtPIP2s ( Figure 1C). It was most likely that heteroligomerization between AtPIP1;1 and AtPIP2s could provide an explanation for the PM localization of AtPIP1;1 in the AtPIP1;1-YFP transformed protoplasts of A. thaliana Col-0. The heteroligomerization and homoligomerization between AtPIP1;1 and AtPIP1s were also found in this test ( Figure 1C). This kind of heteroligomerization or homoligomerization was also demonstrated for maize ZmPIP1;2 in plant cells by FRET experiments, and in this system, ZmPIP1;2 physically interacts with itself and ZmPIP1;1 (Zelazny et al., 2007). Besides, the AtPIP1;1-CubPLV and NubG-AtPIP2s co-transformed yeast grown much better than AtPIP1;1-CubPLV and NubG-AtPIP1s co-transformed ones on SD-WLAH [synthetic dextrose minimal media without W (tryptophan), L (leucine), H (histidine), and A (alanine)] plates ( Figure 1C). This result further demonstrated that the affinity between AtPIP1;1 and AtPIP1s was significantly lower than that between AtPIP1;1 and AtPIP2s.
Yeast growth and viability assays were carried out to determine the H 2 O 2 transport activity. AtPIP2;4-6xHis-and 1;1D(TM2+TM3)/2;4(TM2+TM3)-6xHis-transformed S. cerevisiae cells showed significant growth deficiency due to the toxicity of excess H 2 O 2 uptake compared with the control when spotted on the YPDA+ 3 mM H 2 O 2 plates ( Figure 4B). A cytoplastic H 2 O 2 fluorescent dye named H2DCFDA was applied to evaluate H 2 O 2 accumulation in S. cerevisiae cells. H 2 O 2 was added to the suspension of yeast cells after H2DCFDA installed in cytoplasm. The fluorescence intensity of AtPIP2;4and 1;1D(TM2+TM3)/2;4(TM2+TM3)-transformed cells are all significantly stronger than control ( Figure 4C), which implied 1;1D(TM2+TM3)/2;4(TM2+TM3) has similar ability to regulate H 2 O 2 across the membrane. And like the results of water transport assay, H 2 O 2 transport activity of AtPIP1;1-transformed yeast cells was slightly stronger than the control ( Figure 4C). This slight difference could not even be detected by the yeast growth and viability assay. Previous studies show PIP1s own an original capacity of transporting solutes (Fetter et al., 2004;Bienert et al., 2018). The switch of TM2 and TM3 from AtPIP2;4 to AtPIP1;1 did not alter the original potential of AtPIP1;1 as a water and H 2 O 2 transporter, but truly enabled the localization in PM.
For a further comprehension of the importance of the correct localization of AtPIP1s, H 2 O and H 2 O 2 transport assays were carried out. The capacity of transporting H 2 O was tested, since water channel was the inherent function of AQPs. Pf values were measured based on the water swelling of the Xenopus oocytes with AtPIP1 variants expressed in. The Pf values of AtPIP1;2, AtPIP1;3, AtPIP1;4, and AtPIP1;5 showed a slight difference from the control while the Pf values of (TM2+TM3) AtPIP2;4 replaced AtPIP1 variants were significantly higher than the control and the wild-type AtPIP1s, just like AtPIP2,4 ( Figure  5B). To test the capacity of transporting H 2 O 2 , yeast growth and viability assays were conducted and showed AtPIP1 mutanttransformed yeast cells grew more slowly compared with wildtype AtPIP1 isoforms and the control ( Figure 5C). DCF fluorescence assay also showed that AtPIP1 mutanttransformed yeast cells had stronger fluorescence signal compared with wild-type AtPIP1 isoforms and the control, which meant a faster uptake of H 2 O 2 ( Figure 5D). Altogether, the AtPIP1 mutants, with TM2 and TM3 replaced by those of AtPIP2;4, had an inherent capacity to transport H 2 O and H 2 O 2 when they spontaneously reached the PM, like the performance of 1;1D(TM2+TM3)/2;4(TM2+TM3).

TM2 and TM3 Play Different Roles in Localization
TM2 and TM3 are important segments for the localization of PIPs. For further understanding about the role of TM2 and TM3 in determining subcellular localization of PIPs, structural models were set up based on the previous AQP structure models. AtPIP1;1 and AtPIP2;4, being the focus of this study and the representatives of AtPIP1 group and AtPIP2 group, were logically chosen for the structural analysis. We set up a tetramer model of AtPIP2;4, which compromises four predicted monomers ( Figure 6A). Afterward, four monomers of AtPIP1;1 were aligned with this tetramer of AtPIP2;4, and from the model, we found that TM2 is in inside the tetramer, while TM3 faces the outside. Heterotetramer of AtPIP1;1 and AtPIP2;4 is far more likely to share the same framework as that of the AtPIP2;4 homotetramer to keep the tetramer in a stable form. The physical position of TM2 indicates its roles in oligomerization. The TM2 of AtPIP1s and AtPIP2s is highly conserved respectively, especially the residues extending outward (Supplemental Figure 3), which could be an explanation that the interaction of AtPIP1;1 with its AtPIP2 partners was similar ( Figure 1C and Supplemental Figure 4). Besides, TM2s of AtPIP1;1 and AtPIP2;4 were extracted from the tetramer model with Q88 and L81 in the stick form ( Figure 6B). TM3s

yeast cells (left) and
Xenopus oocytes (right), which were expressing AtPIP1;2-eGFP, AtPIP1;3-eGFP, AtPIP1;4-eGFP, AtPIP1;5-eGFP, and their eGFP-fused mutants . White scale lines represent the length of 5 mm (left) and 100 mm (right). (B) Bar graph of Pf values of Xenopus oocytes injected with 2 ng cRNA coding eGFP, AtPIP1;2-eGFP, AtPIP1;3-eGFP, AtPIP1;4-eGFP, AtPIP1;5-eGFP, and their eGFP-fused mutants. On bar graphs, different letters indicate significant differences in multiple comparisons of each Pf values using an one-way ANOVA analysis, P < 0.01; for each sample, 10 technical replicates were measured, and the experiments were independently repeated three times. (C) Growth status of yeast cells transformed with AtPIP1;2, AtPIP1;3, AtPIP1;4, AtPIP1;5, or their mutants spotted on YPDA plates plus 0 or 3 mM H 2 O 2 . (D) DCF curves of AtPIP1;2, AtPIP1;3, AtPIP1;4, AtPIP1;5, or their mutant-transformed yeast cells with DCF stored and 3 mM H 2 O 2 added, showing fluorescent signal detected by microplate reader. Data are the mean ± SDs. For each sample, triplicates (technical replicates) were measured, and the experiments were independently repeated three times. of AtPIP1;1 and AtPIP2;4 were also highlighted in the superimposed monomers, and the outfacing residues are in stick forms ( Figure 6C). Due to the difference in physical position of the TM2 and TM3 and chemical property of the residues, TM2 and TM3 may play different roles in localization.
PM-targeting PIPs undergo ER-Golgi trafficking and finally reach the PM. Wide studies on localization of PIPs have been done. But the localization of PIPs is still not fully uncovered. When PIP1s were expressed in plant cells, almost all the PIP1s localized in the PM (Figure 1A nether). This phenomenon could be explained by the heterotetramerization between PIPs and PIP2s. PIP1s lack a PM ER secreting signal. But the heterotetramerization with PIP2s makes up for the defect of PIP1 group in ER-Golgi trafficking and even drives PIP1 to the PM. Therefore, whether the oligomerization happens between PIP1s and PIPI2s tremendously impacts the localization of PIP1s. The interfaces between the monomers in the tetrameric ZmPIP1;2 contains TM1-TM2 and TM4-TM5 (Vajpai et al., 2018). Residual amino acids in TM2 could play dominant roles in the tetramerization.
Several motifs and residual amino acids are demonstrated to importantly impact the localization. Among them, LxxxA motif in TM3 plays a vital role in localization process of PIPs. LxxxA motif may act as an ER secreting signal . PIP1s, lacking LxxxA motif, is thus expected to be detained in ER. However, the detaining does not happen to all the PIP1s molecules, since certain amount of PIP1s do localize in the PM in Xenopus oocytes . This is also FIGURE 6 | Structure analysis and localization working model of AtPIP1;1 and AtPIP2;4. (A) Tetrameric structure of AtPIP2;4 with AtPIP1;1 aligned in. TM2s and TM3s are highlighted with cyan (AtPIP1;1) and green (AtPIP2;4). (B) Structure of four TM2s from AtPIP2;4 tetramer (green) with those from AtPIP1;1 (cyan) aligned in. The Q residue and the corresponding L residue are in the form of sticks. Red in the sticks represents oxygen atom, while blue represents nitrogen atom. (C) Sequences alignment of TM3 region of AtPIP1;1 with cyan background and AtPIP2;4 with green background (upper) and structure of AtPIP1;1 and AtPIP2;4 with TM3 highlighted in cyan (AtPIP1;1) or green (AtPIP2;4) (nether). Red letters are the varying residues, and LxxxA region is in the square frame (upper). All outfacing residues are in the form of sticks (nether). (D) Working model of the subcellular localization of plasma membrane intrinsic proteins. Other relative residues are some of the different residues except the L81 and LxxxA motif in TM2 and TM3 between PIP1s and PIP2s that might also take part in localization of PIPs. demonstrated by our results. The mechanism in this process is not well interpreted.

TM2 May Affect the Tetramerization at L81
As previously referred, alignment of TM2 fragments from AtPIP1;1 and AtPIP2;4 showed only two different residue sites ( Figure 3D). Glutamine is an uncharged polar amino acid, while leucine is a non-polar hydrophobic amino acid. Theoretically, embedding the hydrophobic side groups exposed to the outside of a protein is the motive power of the oligomerization. Here, if the L81 at the central position of the tetramer was replaced to Q88, it would be an impediment to the tetramerization due to the hydrophilic side group ( Figure 6B). In interaction test, leucine in TM2 was also demonstrated to affect the oligomerization of PIPs. In maize, when glutamine in TM2 of ZmPIP1;2 was mutated into leucine, the dimeric form was more abundant than monomeric form in western blot experiment, also suggesting leucine could contribute to oligomerization (Berny et al., 2016). Furthermore, the glutamine and leucine at the corresponding site are totally conservative in AtPIP1 group and AtPIP2 group (Supplemental Figure 3). As one can imagine, this site in TM2 is far more likely to be an important difference between AtPIP1s and AtPIP2s.
Residues in TM3 Could Affect the ER-Golgi-PM Trafficking, Especially the LxxxA Motif TM3, reported previously as an important domain of PIPs to affect the localization, contains an LxxxA motif in ZmPIP2;5 which is considered to be a secreting signal . Our work enforces the role of TM3 in localization. Alignment of TM3 fragments of AtPIP1;1 and AtPIP2;4 showed eight different residue sites, including the LxxxA motif region ( Figure 5C upper). In TM3 of AtPIP1;1, the LxxxA region becomes YxxxM (Y, tyrosine; M, methionine), giving AtPIP1;1 a significant difference. In AtPIP2 group, AtPIP2;1, AtPIP2;2, AtPIP2;3, and AtPIP2;4 have LxxxA motif, others with MxxxA, SxxxA (S, serine), GxxxA (G, glycine), or AxxxA, while AtPIP1 isoforms only have YxxxM or FxxxM (F, phenylalanine) motifs (Supplemental Figure 3). Y and F, the first residue of this region in AtPIP1 isoforms, are typical aromatic amino acids, but the corresponding L, M, S, G, A residues in AtPIP2 isoforms are not. The last residue in this region is M in AtPIP1 group, but an A in AtPIP2 isoforms. Although they are all hydrophobic residues, a longer side group and an additional sulfur should give M another image. In summary, the distinction between AtPIP1 group and AtPIP2 group in the LxxxA region could make a huge difference in their ER-Golgi-PM trafficking.

Other Residues in TM2 and TM3 Are Also Important for Localization
It is not concluded that the residues except L81 and LxxxA in TM2 and TM3 do not impact the localization of PIPs. In fact, 1;1D(Q88L+LxxxA), which has the L and LxxxA motif, cannot localize into the PM like AtPIP2;4 (Supplemental Figure 7). In TM2, the other couple of different residues, corresponding to valine and alanine, are similar with each other in side groups. And the relative position of them is completely different from that of L81 and Q88. Therefore, it might affect the localization by a method other than tetramerization. Besides, the other different residues in TM3 between AtPIP1 and AtPIP2 may also play some role in the localization. The orientation, length, and hydrophilicity of the side groups of these residues show a large difference that might determine the affinity to the phospholipid membrane, based on TM3 being outfacing in the tetramer structure ( Figure 6C nether).

Working Model for the Localization of PIPs
Overall, TM2 and TM3 determine which AtPIPs is able to be released from ER to Golgi and finally localize in the PM. AtPIP1 isoforms cannot completely implement the ER-Golgi-PM trafficking when expressed alone due to the defective homotetramerization and trafficking signal. A heterotetramerization with AtPIP2 isoforms will provide AtPIP1s with a stable tetrameric form and a strong ER secreting signal, which carry the AtPIP1 molecule to the PM. Besides, some other relative residues in TM2 and TM3 will also influence the localization of PIPs in some other way.

MATERIALS AND METHODS
Cloning and Generating cDNAs of AtPIP1;1, AtPIP1;2, AtPIP1;3, AtPIP1;4, AtPIP1;5, and AtPIP2;4 were amplified from RNA templates of 3-week-old A. thaliana ecotype Col-0. Mutants of AtPIP1;1, AtPIP1;2, AtPIP1;3, AtPIP1;4, AtPIP1;5, and AtPIP2;4 were generated through fusion polymerase chain reaction (fusion-PCR, Szewczyk et al., 2006) or single-residue mutation method. For fusion-PCR, two or more fragments of AtPIP1;1, AtPIP1;2, AtPIP1;3, AtPIP1;4, AtPIP1;5, or AtPIP2;4 were amplified with homologous recombination arms which were paired with complementary targeting fragments. The complementary pairing fragments were incubated together with DNA polymerase for 12 PCR cycles. Then the full-length primers of the mutated AQPs were added in the products to conduct a PCR procedure. Products were purified with the Cycle Pure Kit (D6492-02). For single-residue mutation, pairs of partially complementary primers were designed at the sites of the targeting residues. The whole vectors with wild-type AtPIPs genes inserted were amplified with the primers. After digesting the template vectors by Dpn I, the purified products were transformed into DH5a. Positive transformants would be the single-residue mutants.

Localization Assay in Yeast
cDNAs fused with eGFP were constructed in S. cerevisiae expression vector, pYES2, with restriction enzyme combinations of Hind III and EcoR I or Kpn I and EcoR I. The resulting plasmids were transformed into S. cerevisiae NMY51 competent cells in transformation solution (0.1 M LiCl, 30% w/v PEG4000). The positive transformants were cultured in SD-Ura (SD without uracil) media with 2% w/v galactose overnight at 30°C and harvested in phosphate buffer solution (PBS; 0.2 mM, pH 7.4). Cells were observed between 495 and 520 nm using 488 nm argon-ion laser excitation with a Zeiss LSM700 laser scanning confocal microscope.

Protein Localization Assay in Xenopus Oocytes
The use of Xenopus oocytes was evaluated and approved by the ethics committee of Nanjing Agricultural University and carried out in accordance with the guidelines provided by this committee.
A Kozak sequence, GCCACC, was placed in front of initiator codon ATG to enhance the translational efficiency (Raif et al., 2010). Kozak sequence containing cDNA-fused eGFP were constructed into pGH19 with restriction enzyme combination of EcoR I and Xba I. The generated plasmids were linearized with Not I restriction enzyme and purified in RNase free water. One microgram for each linearized DNA was applied for a in vitro transcription using RiboMAX™ Large Scale RNA Production Systems-T7 (P1300). Two nanograms resulting cRNA (without DNA) was injected into the Dumont stage V Xenopus oocytes. The generated oocytes were cultured in sterile ND96 solution (93.5 mM NaCl, 2 mM KCl, 1.8 mM CaCl2, 2 mM MgCl2, 5 mM Hepes, pH 7.50) with penicillin and streptomycin added at 18°C for 36 h. Oocytes were fixed with paraformaldehyde and then sectioned into halves. Oocytes were finally observed between 495 and 520 nm using 488 nm argon-ion laser excitation with a Zeiss LSM700 laser scanning confocal microscope.

Pf Examination
The osmotic water permeability coefficient (Pf) value was determined based on the expression of AQPs in Xenopus oocytes. Oocytes were transferred from ND96 solution to a fivefold diluted ND96 solution after 2 days' culture at 18°C. The swelling of the oocytes was monitored by a stereomicroscope linked to a black-and-white digital camera. Images were captured at 5 s intervals for 30 cycles, and the diameter of the oocytes for every cycle was measured for the calculation of the Pf values. The Pf was calculated using the equation Pf = V 0 [d(V/V 0 )/dt]/[S × V w (Osm in − Osm out )], in which the initial volume (V 0 ) equals 9 × 10 −4 cm 3 , the initial oocyte surface area is 0.045 cm 2 , and the molar volume of water (V w ) is 18 cm 3 /mol (Preston et al., 1992;Fetter et al., 2004).

Purification of Proteins
The transformed yeast cells were pre-cultured in a shaker at 180 rpm and 30°C overnight. The culture of yeast was started with an OD600 of 0.1 adjusted with the pre-cultured yeast suspension. Cells were harvested after 24 h and then broken by vortex with 0.5 mm glass beads in 0.3 M sucrose containing 0.1 mM phenylmethanesulfonyl fluoride and 0.1 mM leupeptin. Cell debris was removed by a 15,000 × g spin at 4°C for 2 × 20 min. The PM fractions were prepared by differential centrifugation as previously described (Marinelli et al., 1997;García et al., 2001). Briefly, the PM fraction was obtained by centrifugation at 200,000 × g for 1 h on a discontinuous 1.3 M sucrose gradient. The PM band was isolated and subsequently mixed with 200 mM NaCl, 50 mM Tris pH 8, 2% n-dodecyl-N, N-dimethylamine-N-oxide (PM fraction). The left fraction was cytoplastic proteins and intracellular membranes (CP fraction). The PM fraction and CP fraction were then applied to western blot assay.
To extract the protein of Xenopus oocytes, cRNA injected oocytes were broken byvortex at 4°C for 5 min and homogenized in PBS with 2% SDS.
Localization Assay in Protoplast cDNAs were constructed in a re-modified plant expression vector pCAMBIA 1300 35S::YFP::poly(A), with restriction enzyme combinations of Kpn I and Xba I or Sac I and Xba I. As a result, cDNAs ligased with YFP sequence were inserted between CaMV 35S promoter and CaMV poly(A) signal sequence, thus forming a highly efficient expression system. Protoplast isolation and transient expression were conducted based on the previous study (Hoffman et al., 1994). Leaves of Arabidopsis were cut into 0.5 × 5 mm strips and thrown in enzyme solution (1% w/v cellulase R10, 0.2% w/v Macerozyme R10, 0.4M mannitol, 20 mM KCl, 20 mM MES, pH 5.7, 10 mM CaCl 2 , 0.1% BSA). After a 30 min vacuum treatment and a 2 h incubation at 40 rpm and 23°C, protoplast was released in the solution. The protoplast containing solution was added with W5 solution (154 mM NaCl, 125 mM CaCl 2 , 5 mM KCl, 2 mM MES, pH 5.7) to stop the enzyme reaction, along with a 0.1 mm aperture nylon mesh filtration. Protoplast was spun down at 100 rpm and 4°C for 5 min, washed with W5 solution, and finally resuspended in MMG solution (0.4 M mannitol, 15 mM MgCl 2 , 4 mM MES, pH 5.7). The obtained protoplast was incubated with corresponding plasmids in PEG solution (40% w/v PEG, 0.2 M mannitol, 100 mM CaCl 2 ) at 23°C for 15 min. The reaction was stopped by adding equal volume of W5 solution. Transformed protoplast was obtained by 100 rpm spin at 4°C for 5 min, resuspended in W5 solution, and cultured in darkness at 23°C for 16 h. Resulting protoplast was observed between 520 and 535 nm using 514 nm argon-ion laser excitation with a Zeiss LSM700 laser scanning confocal microscope.

Proteins Interaction Assay Using SUB Yeast Two Hybrid System
This experiment was carried out based on the previous study (Obrdlik et al., 2004). Briefly, the cDNAs of AtPIP1;1 and other AtPIPs were constructed in pMetYCgate and pXNgate respectively. The combination of pMetYCgate-AtPIP1;1 and pXNgate-AtPIP were transformed into S. cerevisiae NMY51. Transformants were spread on SD-WL (SD without W and L), SD-WLH (SD without W, L, and H), and SD-WLAH plates. Positive transformants were 10-fold serially diluted and spotted on SD-WLAH plates and incubated at 30°C for 2 days. And the positive transformants with original OD 600 = 0.01 were cultured in liquid SD-WLAH medium at 200 rpm, 30°C for 2 days. The final OD 600 of resulting yeast culture was then measured by UV-VIS spectrophotometer (UV-2700).
Yeast Growth and Viability Assay cDNAs fused with 6xHis tag were ligated to S. cerevisiae expression vector pYES2, with restriction enzyme combinations of Hind III and EcoR I or Kpn I and EcoR I. The resulting plasmids were transformed into S. cerevisiae strain NMY51, competent cells in transformation solution (0.1 M LiCl, 30% w/v PEG4000). And the positive transformants were cultured in SD-Ura media with 2% w/v galactose overnight at 30°C and harvested in PBS (0.2 mM, pH 7.4). The cell pellets were washed with PBS and diluted to a normalized cell density of OD 600 = 1. Three repeating 10-fold serial dilutions were prepared from the normalized suspension for each transformant. Six dilutions, 10 ml of each, from every transformant were spotted on the YPD agar plates with or without 3mM H 2 O 2 . The yeast was incubated at 30°C for 3-4 days.

DCF Fluorescence Assay
The transformants in yeast growth and viability assays were applied to DCF fluorescent assay. The yeast cells were resuspended in PBS along with an addition of H2DCFDA (D399) at a final concentration of 10 mM (Saito et al., 2003;Tian et al., 2016). After an incubation for 30 min at 30°C, yeast cells were washed twice by PBS and resuspended in PBS with H 2 O 2 at a final concentration of 3 mM. Fluorescence densities in transformed yeast cells were evaluated with a SpectraMax M5 96 microplate reader to estimate the H 2 O 2 -transporting capacity of AtPIP1;1, AtPIP2;4, and variant isoforms of them.

Structure Analysis
The structures of AtPIP1;1 and AtPIP2;4 were predicted and set up by the PHYRE2 Protein Fold Recognition Server. The tetramer model of AtPIP2;4 was set up based on the model of SoPIP2;1 tetramer (PDB ID 4IA4) using PyMOL software. Four AtPIP1;1 monomers were aligned in the model of AtPIP2;4 tetramer that formed an AtPIP1;1 tetramer.

DATA AVAILABILITY STATEMENT
All datasets generated for this study are included in the article/ Supplementary Material.

ETHICS STATEMENT
The use of Xenopus oocytes was evaluated and approved by the ethics committee of Nanjing Agricultural University and carried out in accordance with the guidelines provided by this committee.