Having an abundance of AQP protein in the yeast cell is the first essential requirement for robust functional evaluations and improves the detection limit in response to treatments. For example, the water permeability for AtPIP2;3 was assessed using two differentially active promoters, with greater freeze-thaw tolerance (a proxy for water permeability) achieved using the strong GPD promoter relative to the less active TPI1 promoter ( Supplemental Figure S2 ). To maximize the likelihood of high AtPIP production we (i) used high copy number plasmids with minimal load burdens on yeast growth, (ii) used a strong constitutive GPD promoter with complementing terminator, (iii) ensured codon usage compatibility between AtPIPs and yeast, and (iv) modified the Kozak sequence to enhance translational initiation (see Supplemental Materials and Methods ). A parallel collection of AtPIP-GFP transgenes that differed only in the C-terminal GFP fusion compared to the expression vectors used in the functional assays, were used for evaluating heterologous AtPIP production in vivo and subcellular localization. Of the 13 AtPIPs assessed, 12 AtPIP-GFP yeast lines repeatedly emitted strong GFP signal (440 GFP fluorescence units/cell OD1 average), indicating high AtPIP protein production ( Supplemental Figure S3 ). Relative to the other PIPs, AtPIP1;4 had the lowest abundance, emitting 113 GFP fluorescence units/cell OD1, representing 27% of the average fluorescence intensity for all AtPIPs combined (415 GFP fluorescence units/cell OD1, black dotted line, Supplemental Figure S3 ). The reduced AtPIP1;4 protein abundance relative to the other AtPIPs remains unexplained, however we could still detect growth phenotypes comparable to the other AtPIPs through our functional experiments (described below).

In addition to ample heterologous protein production, sufficient AtPIP needs to localize to the yeast plasma membrane (PM) in order to evaluate AQP-facilitated substrate transport into the cell. Sub-cellular localization of the AtPIPs was evaluated using confocal microscopy of AtPIP-GFP lines and compared against cytosolic (GFP only) and endoplasmic reticulum (ER; SEC63-RFP) markers ( Figure 2 ). Free GFP is cytosolically localized ( Figure 2A ). The SEC63-RFP marker reveals the web-like ER network, with the prominent nuclear envelope ER domain (nER) and peripheral or cortical ER domain (cER) ( Figure 2B ). The cER lies immediately adjacent to the PM but is discontinuous around the perimeter with discernible gaps, distinguishing it from PM localisation ( Figure 2B ). A sharp ring around the cell perimeter was seen for all 8 AtPIP2-GFP proteins, indicating a consistent strong targeting to the PM ( Figures 2, E, F, I, J, M, N, Q and R ). When expressed alone, the five AtPIP1-GFP proteins show a faint continuous ring around the periphery of the cell, consistent with PM localisation ( Figures 2, C, G, K, O and S ), but less efficient than observed in the AtPIP2s. In addition to localizing to the cell periphery, all 5 AtPIP1s show dual localization consisting of a patchy peripheral ring and internal webs like those observed in the SEC63-RFP ER marker ( Figures 2, C, G, K, O and S ).

PIP2 proteins can interact and guide PIP1 proteins more efficiently to the PM (Jozefkowicz et al., 2017). The Yeast-two-Hybrid mating-based Split-Ubiquitin System (Y2H mbSUS; Figure 3A ) was used to screen an AtPIP interactome library. Yeast co-expressing the bait AtPIP2;5-CubPLV and any of the AtPIP1;1-Nub to AtPIP1;5-Nub prey proteins, activated the lacZ reporter ≥ 4-fold above background levels ( Figure 3B ), demonstrating that AtPIP2;5 strongly interacted with each AtPIP1. Co-expression of AtPIP2;5 with GFP tagged versions of AtPIP1;1 to 1;5, resulted in the fluorescence signal now being predominantly associated with the PM and comparable to AtPIP2 isoforms ( Figures 2, D, H, L, P and T ).

AtPIP2;5 was chosen because, among the AtPIP2s, its expression in yeast resulted in moderate relative levels of sensitivity to the tested substrates (compared to other PIP2s), enabling further enhancements in sensitivity due to the co-expressed AtPIP1.

Yeast are sensitive to very rapid freezing events, with the formation of intracellular ice crystals causing cell damage and death. Freeze-thaw survivorship depends on how rapidly water can efflux from the cell across the PM (Cabrera et al., 2020), which in turn correlates with the water transport function of water permeable AQPs (Tanghe et al., 2002; Tanghe et al., 2004; Soveral et al., 2006). Therefore, water permeability of AQP variants can be rapidly determined in yeast by screening for improved freeze-thaw survivorship over extended time periods (To et al., 2015). We adapted a freeze-thaw assay to our micro-cultivation setup to test the permeability of AtPIPs to water. For wild type yeast carrying an empty vector, successive freeze-thaw treatments incrementally decreased ΔAUC ( Supplemental Figures S4A and B ). Freeze-thawing prolonged the lag phase (λ) ( Supplemental Figure S4C ), consistent with a reduction in viable cell count of the starting population, which delayed detection of population growth. The sensitivity of the freeze-thaw assay was improved by using the aquaporin null mutant background (aqy1 aqy2), which is compromised in tolerance to rapid freeze-thaw events (Tanghe et al., 2002; Tanghe et al., 2004). Two freeze-thaw cycles were sufficient to essentially render the entire aqy1 aqy2 starting population unviable ( Supplemental Figures S4A-C ). Heterologous expression of a water permeable AQP (AtPIP2;1) (Verdoucq et al., 2008), dramatically improved the tolerance of the aqy1 aqy2 mutant to repeated freeze-thaw treatments ( Supplemental Figures S4A-C ).

Application of two freeze-thaw treatments to aqy1 aqy2 yeast expressing one of the 13 AtPIP genes or an empty vector, differentially affected the growth curves ( Figure 4A ). All of the AtPIP2 proteins had sufficient capacity to transport water across the PM to confer freeze-thaw tolerance, but their response varied with AtPIP2;7, the most, and AtPIP2;2, the least tolerant to freeze-thaw ( Figure 4B ). At ф, growth was not detected for any AtPIP1 expressing lines. However, inspection of the raw growth curves indicated that over the course of the 42h growth period there was a low level of growth recovery post freezing treatments. By calculating AUC at ф + 1000 mins, freeze-thaw tolerance associated with AtPIP1s was revealed, but resolution between AtPIP2 isoforms was lost ( Figure 4C ). The survivorship of AtPIP1 expressing yeast after freeze-thaw treatment were substantially lower than the AtPIP2s.

Water transport of AtPIP1s was further assessed by increasing their abundance in the PM through co-expression with AtPIP2;5. Yeast co-transformed with AtPIP2;5 + Empty vector served as a base-level control, with less freeze-thaw tolerance than yeast carrying the AtPIP2;5 vector alone or co-expressing two copies of AtPIP2;5 ( Figure 4D ). This is consistent with AtPIP2;5 + Empty vector yeast having reduced expression of AtPIP2;5 as only half the plasmid load carries AtPIP2;5. Co-expression of AtPIP1;1, 1;2, 1;3, 1;4 or 1;5 with AtPIP2;5 substantially improved freeze-thaw survivorship over the AtPIP2;5 + Empty vector control, with AtPIP1+AtPIP2;5 co-expression increasing relative survivorship at ~40-75% compared to AtPIP2;5 alone (i.e. AtPIP2;5 + AtPIP2;5; Figure 4D ). Co-expression revealed that all five AtPIP1 isoforms are capable of significant water transport, but they appear less effective than AtPIP2s.

Water permeability was also assessed using the traditional, but more laborious, yeast spheroplast bursting method ( Figure 4E ). During a 5-sec exposure of spheroplasts to a hypotonic solution, there is an initial phase marked by a rapid decrease in measured OD₆₅₀ as spheroplasts swell and burst, followed by a slower reduction phase reflective of spheroplast settling. A greater value for the initial phase kinetic parameter A₁ is indicative of more efficient water influx into yeast spheroplasts. Spheroplast bursting rates were ranked AtPIP2;7 > AtPIP2;1 > AtPIP1;5 > empty, matching the order and approximate relative differences in growth phenotypes obtained from the freeze-thaw assay. The consistency in ranking between the two methods, validated assessment of water permeability across the AtPIP family using the freeze-thaw assay which provides a platform to rapidly and comparatively evaluate relative water transport function of AQPs.

The reactive oxygen species (ROS) hydrogen peroxide (H₂O₂), can impair yeast growth (decreasing μ and κ) and trigger cell death (prolonging λ) when internalized levels exceed the protective mechanisms of the cell (Jamieson, 1998; Madeo et al., 1999). In our setup, H₂O₂ treatments impaired growth of the empty vector aqy1 aqy2 yeast ( Figure 5A ), impacting all three growth traits (λ, μ, and κ; Supplemental Figure S5 ). The effects were more prominent when using the skn7 yeast, which is compromised in its antioxidant buffering capacity ( Figure 5A ; Supplemental Figure S5 ). 0.5mM and 1mM H₂O₂ were chosen as treatment concentrations as they occur at the commencement of pronounced growth inhibition (i.e. linear range of the dose response curves) ( Figure 5B ), and thus provide a greater range of detection and resolution between AQP isoforms. Testing at two concentrations (i.e. modulating diffusion potential) extends the detectable range and ability to identify weaker permeabilities and variation in the toxicity phenotypes between isoforms.

The impact on λ, μ, and κ for AtPIP-expressing yeast in response to 0.5mM and 1mM H₂O₂ treatment were consistent with the empty vector control exposed to increasing concentrations of H₂O₂ ( Figures 5A, C and D ). As expected, the differences between AtPIP associated H₂O₂ sensitivities were greater in skn7 compared to aqy1 aqy2 yeast ( Figures 5C, D , and Supplemental Figures S6A, B ).

Growth relative to the empty vector control was inhibited by 0.5mM H₂O₂ for all AtPIP2 expressing aqy1 aqy2 yeast lines except AtPIP2;6 ( Figure 5E ). At the higher concentration (and thus diffusion potential) of 1mM H₂O₂, all AtPIP2 expressing aqy1 aqy2 yeast grew worse than empty vector control ( Supplemental Figure S6A ). Greater growth inhibition and differentiation between isoforms was observed in the more sensitive skn7 background, with differences between select AtPIP2 isoforms (AtPIP2;2, 2;5, and 2;6) especially prominent at 1mM compared to their evaluation in the aqy1 aqy2 background ( Supplemental Figure S6B ). The results indicated that all AtPIP2 proteins can facilitate enhanced diffusion of H₂O₂ across the PM to some extent, AtPIP2;6-expressing yeast was least sensitive to H₂O₂ treatment, while AtPIP2;7 was the most sensitive ( Figure 5E ).

AtPIP1 expressing aqy1 aqy2 yeast showed no indication of enhanced H₂O₂ uptake across the PM beyond the passive background diffusion rate, represented by the empty vector control, except for a small effect with AtPIP1;1 at 1mM H₂O₂ ( Supplemental Figure S6A ). When expressed in skn7, AtPIP1;3, 1;4 and 1;5 conferred greater sensitivity to H₂O₂ (at 1mM) than empty vector control, with growth reductions sitting between the most and least sensitive AtPIP2;2 and AtPIP2;6 respectively ( Supplemental Figure S6A ), indicating that these isoforms also facilitate H₂O₂ transport across the PM ( Supplemental Figure S6B ). Intriguingly, the skn7 AtPIP1;2-expressing yeast grew consistently better than empty vector control (several independent transformation events, and a marginally discernable effect in the aqy1 aqy2 background), suggesting that expression of AtPIP1;2 alone in skn7 somehow protects against H₂O₂ treatment ( Figure 5F and Supplemental Figure S6B ).

When AtPIP1 PM targeting was improved through co-expression with AtPIP2;5, all AtPIP1s dramatically increased the sensitivity of skn7 yeast to H₂O₂ over the AtPIP2;5 + Empty vector control. The effect was clearly evident at 0.5mM ( Figure 5G ) and even as low as 0.25mM H₂O₂ ( Supplemental Figure S6C ), whereas 1mM H₂O₂ was required to observe a significant increase in skn7 sensitivity beyond the empty vector control when AtPIP1s were expressed in skn7 yeast alone ( Supplemental Figure S6B ). Of note, the enhanced growth phenotype observed in AtPIP1;2-expressing skn7 yeast at 1mM H₂O₂ was not observed when co-expressed with AtPIP2;5 ( Figure 5G ), suggesting that localization to the ER or a homotetrameric state may be required for this off-target effect. AtPIP2;5 + AtPIP1;3 and AtPIP2;5 + AtPIP1;4 skn7 lines were the most sensitive to 0.5mM H₂O₂, with AUC values relative to AtPIP2;5 + Empty vector control of 0.1 and 0.06, respectively ( Figure 5G ). In order to confirm that co-expression of AtPIP1 was not associated with some form of hyperactivation of the AtPIP2;5 through hetero-oligomerization, we designed a mutant version of AtPIP1;4 (AtPIP1;4H207K) with hindered monomeric channel activity but retained hetero-oligomerization capacity. The histidine at position 207 in AtPIP1;4 represents a strongly conserved residue located in a cytosolic loop of PIP proteins that is involved in gating of the monomeric pore (Törnroth-Horsefield et al., 2006). Mutation to a positively charged Lysine(K) mimics histidine protonation, which favors pore closure and reduces transport of substrates, including H₂O₂ (Tournaire-Roux et al., 2003; Verdoucq et al., 2008; Bienert et al., 2014). In an independent collection of H₂O₂ toxicity assays, increasing PM abundance of AtPIP1;4 through AtPIP2;5 + AtPIP1;4 co-expression, once again dramatically sensitized skn7 yeast to H₂O₂ ( Figure 5H ). However, when AtPIP2;5 was co-expressed with the AtPIP1;4H207K closed/gated mutant, the ΔAUC values resembled growth levels more similar to AtPIP2;5 + Empty control ( Figure 5H ). This supports the interpretation that AtPIP1;4 was responsible for the enhanced H₂O₂ sensitivity of the AtPIP2;5 + AtPIP1;4 yeast. Collectively, the co-expression results suggest that all AtPIP1 proteins transport H₂O₂ and provide greater sensitivity phenotypes than AtPIP2 isoforms.

Boron is essential for yeast growth, but at high concentrations is toxic. At moderate concentrations (< 80mM) it acts as a fungistatic agent, slowing down proliferation by disrupting cell wall synthesis, but not killing the cell (Bennett et al., 1999; Schmidt et al., 2010). A range of moderate boric acid (BA; H₃BO₃) concentrations were tested on aqy1 aqy2 empty vector yeast to determine treatment doses. Consistent with reports, moderate BA treatments mainly reduced the rate of growth (μ) with little impact on lag-phase (λ) ( Figure 6A ; Supplemental Figure S7 ). ΔAUC at ф relative to untreated cultures followed a single dose response curve and 20mM and 30mM BA were selected as optimal treatment concentrations ( Figure 6B ).

Changes in the growth curve characteristics (λ, μ, and κ) of AtPIP-expressing yeast in response to 20mM and 30mM BA treatment were consistent with the empty vector control exposed to increasing BA concentrations ( Figures 6A and C ). Five of the 13 AtPIP yeast lines were more sensitive to BA than the empty vector control ( Figures 6D and E ). AtPIP1;1 expressing yeast were by far the most sensitive to BA exposure, with dramatic growth reductions even at 20mM BA ( Supplemental Figure S8 ). Yeast expressing AtPIP2;2, 2;7 and 2;8 had BA-induced sensitivities similar to the HvPIP1;4 positive control (Fitzpatrick and Reid, 2009). AtPIP1;5 expressing yeast showed a small increase in BA sensitivity compared to Empty vector, which was significant in three of the four experiments ( Figure 6E ; Supplemental Figure S8A and B ). Co-expression of AtPIP1s with AtPIP2;5 did not alter BA sensitivity compared to the yeast expressing AtPIP1s alone at 20mM BA ( Supplemental Figure S8B ), whereas at 30mM BA, AtPIP1;2 + AtPIP2;5 co-expression resulted in an increased BA sensitivity compared to AtPIP1;2 alone ( Figure 6E ).

Truncation of the cytosolic N-terminal domain of PIP1, PIP2, and NIP isoforms from different plant species, has enabled boron, or similar metalloid, uptake in yeast (Bienert et al., 2008; Fitzpatrick and Reid, 2009; Kumar et al., 2014; Mosa et al., 2016). We generated and tested several PIP1 isoforms with truncations of the cytosolic N-terminal domain (AtPIP1;2_Δ2-47 , AtPIP1;4_Δ2-47 and AtPIP1;5_Δ2-48 ). The truncated versions had similar sensitivity to BA as their full-length counterparts ( Supplemental Figure S8C ). Overall, the results indicate that five members across both the AtPIP1 and AtPIP2 sub-families are candidates for BA transport.

Growth of the empty vector ynvw1 (dur3) urea uptake deficient mutant was enhanced by increasing concentrations of urea; specifically through increased maximum growth rate (μ) and carrying capacity (κ) ( Figures 7A, B ; Supplemental Figure S9 ). When urea was supplied at high concentration (i.e. high diffusion potential), all yeast lines grew similarly to the empty vector control ( Figure 7C ), indicating that the AtPIP and AtTIP2;3 (positive control) yeast cultures were healthy and capable of growing better when exposed to a urea/nitrogen concentration that imposes a higher permeability across the plasma membrane. However, in lower concentrations of urea (4 mM), none of the AtPIP expressing yeast showed improved growth compared to Empty vector, whereas the positive urea transporting control, AtTIP2;3 (Dynowski et al., 2008a), clearly complemented the dur3 growth phenotype ( Figure 7C ). Therefore, none of the AtPIPs appear to promote notable urea uptake.

To assess AtPIP potential for Na⁺ transport, we quantified Na⁺ accumulation in AtPIP-expressing yeast, following short-term exposure to 70mM NaCl treatments (Qiu et al., 2020). Exposing empty vector control aqy1 aqy2 yeast to 70mM NaCl resulted in a ~40-fold increase in the Na⁺ content relative to yeast from media without additional NaCl ( Figure 8 ). The five AtPIP1 isoforms and AtPIP2;5 accumulated Na⁺ similar to the empty vector control. Yeast expressing AtPIP2;1, 2;2, 2;6 and 2;7 accumulated more Na⁺, while yeast expressing AtPIP2;3, 2;4, and 2;8 accumulated less Na⁺ than empty vector control. AtPIP2;1 served as a positive control (Byrt et al., 2017).

Protein sequence alignments reveal the high homology between AtPIPs ( Supplemental Figure S10 ). Motifs associated with substrate selectivity (i.e. NPA, ar/R and Froger’s positions) are essentially identical among the AtPIPs ( Supplemental Table S2 ). Gross differences are seen in the longer N-terminal and shorter C-terminal domains of AtPIP1s compared to AtPIP2s, and variation in the length of loop A ( Supplemental Table S2 ). Phylogenetic analysis shows that AtPIPs divide into discrete sub-clades with distinct relationships with their substrate profiles and organ level gene expression ( Figure 9 ). For example, the AtPIP1;1 and 1;2 paralogs appear to have undergone substantial functional diversification based on their gene expression patterns. AtPIP1;2 is the most abundantly and constitutively expressed of all AtPIPs, even detected at high levels in dry seed. AtPIP1;1, is mainly expressed in roots, being ~6-fold less prevalent in aerial tissues. This diversification in expression patterns could relate to BA transport being present in AtPIP1;1, but absent in AtPIP1;2 ( Figure 9 ). The AtPIP1;3 and 1;4 paralog pair, may have evolved to transport H₂O₂ in preference to water given the relative phenotype rankings of higher H₂O₂ and lower water transport compared to other AtPIPs. Both genes are broadly expressed with largely overlapping expression domains, which together with their similar transport profiles, points towards possible functional redundancy. AtPIP1;3 differs from AtPIP1;4 by being more highly expressed in general, especially in the root and stem. AtPIP1;3 expression is also up-regulated during seed imbibition and seedling germination, whereas AtPIP1;4 is only weakly expressed at this stage of development ( Figure 9 ). Intriguingly, AtPIP1;5 sits as a phylogenetic outgroup within the AtPIP1 clade, and is a candidate for water, H₂O₂ and BA transport. AtPIP1;5 had the highest relative phenotype ranking amongst the AtPIP1 for water transport ( Figure 9 ) and AtPIP1;5 transcripts are particularly abundant in elongating siliques and the developing seed within.

Among the AtPIP2 isoforms, AtPIP2;7 has the most diverse substrate transport profile and expression patterns, and putatively transports water, H₂O₂, BA, and Na⁺ ions at comparatively high efficiency based on yeast growth rankings. AtPIP2;7 is expressed at high levels in most tissues, with the exception of mature leaves and dry seed, but is upregulated during seed imbibition and germination ( Figure 9 ). Its closest relative, AtPIP2;8, is a candidate for water, H₂O₂, and BA transport, but AtPIP2;8 has relatively low expression under non-stressed growth conditions ( Figure 9 ). This reveals that AtPIP2;8 is either highly cell specific, conditionally expressed, or that AtPIP2;7 is the dominant isoform of this closely related pair. The AtPIP2;5 and AtPIP2;6 phylogenetic pair are noteworthy as being apparently the least effective H₂O₂ transporters of all AtPIPs based on the yeast growth phenotypic rankings ( Figure 9 ). AtPIP2;5 and AtPIP2;6 are distinctly not expressed in roots, and AtPIP2;5 is expressed in meristematic tissue and developing seed, and AtPIP2;6 expression is localized to aerial vegetative and reproductive tissues ( Figure 9 ).

Using yeast-based systems for heterologous expression and functional assessment of AQPs offers numerous advantages over other systems such as oocytes, liposomes, and artificial membranes. Key advantages include: a large range of well-characterized mutant S. cerevisiae strains which can be used for testing different compounds; simple monitoring of growth; scalable to high-throughput processing; and enabling power of sampling a yeast population versus single cell/event sampling in other systems.

Since liquid cultures provide improved exposure of yeast cells to substrates and enable accurate detection of smaller phenotypic growth changes relative to yeast grown on solid plates (Toussaint et al., 2006; Marešová and Sychrová, 2007; Hung et al., 2018), we developed a liquid micro-cultivation system enabling high-throughput, quantitative monitoring of yeast growth in response to treatments. The 96-well plate format offers capacity to screen multiple samples in one experiment, simplifying statistical evaluation. Optical density measurement removed the element of human subjectivity used to assess yeast spot assay phenotypes.

Although an indirect measurement, monitoring of growth inhibition/promotion in response to treatment is considered a reliable proxy for chemical uptake in yeast (Denny and Steel, 2015; Denny, 2018). This includes a growing list of yeast-based AQP studies showing that altered growth of AQP expressing yeast in response to chemical treatment reflects an enhanced intracellular accumulation of the tested substrate (Bienert et al., 2007; Bienert et al., 2008; Dynowski et al., 2008b; Fitzpatrick and Reid, 2009; Bienert et al., 2011; Kumar et al., 2014; Mao and Sun, 2015; To et al., 2015; Mosa et al., 2016; Rhee et al., 2017; Wang et al., 2019). We did not detect any indirect effects of AQP expression on yeast susceptibility to chemical treatments ( Supplemental Note 2 ), and chose substrate concentrations that were at the commencement of pronounced growth effects (i.e. linear range of the dose response curves) to more ideally correlate changes in growth with improved permeability.

Curve transformation to compensate for nonlinearity in optical density measurements, provided a more accurate representation of growth parameters and culture health. The implementation of a dynamic measuring point ф, enabled standardized evaluation between different AQP-expressing yeast lines. Differential growth responses due to increased substrate diffusion into the yeast were dependably captured by the single parameter, AUC.

High AQP-abundance in heterologous systems is critical for accurate assessment of functional capacity and to avoid false-negative transport assignment (Bienert et al., 2014). Protein abundance was quantified in living yeast cells through capturing GFP fluorescence emission of C-terminal GFP translational fusions (AQP-GFP), which is an efficient high-throughput screen that correlates well with traditional Westerns (Albano et al., 1998; Ghaemmaghami et al., 2003; Drew et al., 2006; Drew et al., 2008; Scharff-Poulsen and Pedersen, 2013). High AtPIP production was achieved by careful design of our AtPIP yeast expression constructs. Protein abundance was not correlated with phenotypic responses, suggesting that the observed phenotypes were independent of protein production. The AtPIPs must also integrate into the yeast plasma membrane in order to affect substrate transport into the yeast cell. We found that AtPIP2s had strong PM integration, while AtPIP1s co-localize to the PM and ER. Poor PM localization of PIP1s expressed alone in heterologous systems is a common phenomenon (Yaneff et al., 2015), likely due to sequence differences in the diacidic motif, LxxxA and C-terminal phosphorylation protein motifs known to control PIP2 PM trafficking ( Supplemental Table 2 ) (Chevalier and Chaumont, 2015). The exact composition of diacidic and LxxxA motifs vary, particularly between the phylogenetically distinct [2;1, 2;2, 2;3, 2;4] and [2;5, 2;6, 2;7, 2;8] groups ( Supplemental Table 2 ), yet all AtPIP2s localized efficiently to the yeast PM. In plants, the phylogenetically distinct AtPIP2;1 and AtPIP2;7 also localize efficiently to the PM (Sorieul et al., 2011; Hachez et al., 2014). This reveals flexibility in these motif sequences that must work together with other domains (e.g. TMH2; Wang et al., 2019) to control ER to PM trafficking. PIP2 proteins can physically interact with PIP1s and facilitate PM integration in both host and heterologous systems (Jozefkowicz et al., 2017). We enhanced AtPIP1 PM localization by co-expression with AtPIP2;5, thereby enabling a more robust assessment of substrate permeability between the two PIP sub-types. Protein levels and PM targeting efficiency (e.g. using co-expression) were comparable between the AtPIPs, meaning any differences in permeability profiles were independent of these factors and may be attributed to the intrinsic differences in the sequences of the AtPIPs.

Water permeability is the most extensively studied AQP function across species. Most AtPIPs have been confirmed to transport water (AtPIP1;1, 1;2, 1;3, 2;1, 2;2, 2;3, 2;4, 2;6, and 2;7) (Kammerloher et al., 1994; Tournaire-Roux et al., 2003; Heckwolf et al., 2011; Byrt et al., 2017; Kourghi et al., 2017; Wang et al., 2020a). These assessments are from different studies and systems making it difficult to directly compare relative transport ranking. Here, water permeability was assessed for the complete set of AtPIPs using a freeze-thaw assay that we refined for rapidly evaluating water transport capacity of AQPs. We found that all AtPIP isoforms were capable of water transport, with AtPIP2s having higher apparent water transport capacity than AtPIP1s, even when PIP1s are efficiently targeted to the PM through co-expression with AtPIP2;5. Past studies concluding that PIP1s have low/no permeability to water, likely reflect the inefficient PM targeting of PIP1s when expressed alone in heterologous systems, hiding their true water transport ability [reviewed in (Yaneff et al., 2015)].

PIPs provide a transcellular route for water flow in the plant, from water uptake by roots to transpiration loss from aerial tissues (Groszmann et al., 2017). Both AtPIP1 and AtPIP2 isoforms play major roles in water flow in Arabidopsis (Javot et al., 2003; Prado et al., 2013; Sade et al., 2014). Overlapping expression patterns suggest substantial functional redundancy, which limits the ability of reverse genetic studies to resolve the contribution of each AtPIP to water flow. For example, single loss-of-function mutants of high leaf-expressing isoforms Atpip1;2, Atpip2;1 and Atpip2;6 each show a ~20% reduction in rosette hydraulic conductivity, which worsens to ~39% in the triple mutant (Prado et al., 2013). Our observations that AtPIP2;7 has high apparent water permeability and its transcripts are abundantly expressed in developing leaves ( Figure 9 ), suggests it may also contribute to rosette hydraulic conductivity. Similarly, redundancy for root hydraulic conductance is likely given that the 10-20% reductions seen in single Atpip mutants falls short of the ~64% decrease achieved using AQP chemical blockers (Maurel et al., 2015). Four of the seven AtPIPs abundantly expressed in roots ( Figure 9 ), had high apparent water permeability (AtPIP2;1, 2;2, 2;4, 2;7) and are strong candidates for multiple knock-out mutant studies.

More intricate developmental processes relying on cell-to-cell water movement through AtPIPs are emerging. For example, guard cell closure (Grondin et al., 2015), lateral root emergence (Péret et al., 2012), and pollen germination on stigmatic papillae (Windari et al., 2021). A number of AtPIPs are expressed in the flower, developing silique and seeds. In these tissues, AtPIP water transport could have roles in petal expansion, anther and pollen development, and assist in the supply of nutrients to the developing seed as seen in other species (Wang et al., 2020b; Hoai et al., 2020).

When expressed in aqy1 aqy2 and/or skn7 yeast, all AtPIP2s and AtPIP1s (co-expressed with AtPIP2;5) led to enhanced toxicity phenotypes indicative of H₂O₂ transport in yeast ( Figure 5 ), which is consistent with the similar physicochemical properties of H₂O₂ and water (Almasalmeh et al., 2014). AtPIP1 expression alone did not lead to increased yeast cell sensitivity to H₂O₂ exposure but an unexpected ‘protective’ growth phenotype in AtPIP1;2 yeast lines was observed. Previous growth-based assessments with yeast did not assign H₂O₂ permeability to AtPIP1 isoforms and showed mixed results for AtPIP2 isoforms (Hooijmaijers et al., 2012; Wang et al., 2019; Wang et al., 2020a). Variation in previously observed growth-based assessments may have been due to inadequate protein production, insufficient PM targeting (as shown in our experiments), choice of yeast strain, sub-optimal H₂O₂ concentrations, or use of solid medium spot growth assays.

The potential for H₂O₂ transport through AtPIP1s was recently hinted at using AtPIP1/2 chimeric proteins that more effectively localize to the PM (Wang et al., 2019). However, in addition to harboring PM targeting motifs, the substituted PIP2 domains also contribute to the pore lining, making it uncertain how representative these chimeric proteins are of native AtPIP1 function. In our system, we found that native AtPIP1 proteins were indeed capable of transporting H₂O₂, and when efficiently targeted to the PM through co-expression with AtPIP2;5, appeared more effective at transporting H₂O₂ than AtPIP2 isoforms.

H₂O₂ is an indispensable signaling molecule involved in many aspects of plant growth, biotic defense and abiotic stress responses, reliant on AQPs to facilitate its movement between sub-cellular compartments and cells (Černý et al., 2018; Fichman et al., 2021). The diversity of AtPIP expression patterns in planta and AtPIP ability to facilitate H₂O₂ transport, may enable fine tuning of H₂O₂ signaling. Direct physiological evidence in Arabidopsis is emerging, with H₂O₂ transport through AtPIP2;1 involved in triggering stomatal closure (Rodrigues et al., 2017) and mediating systemic acquired acclimation to abiotic stress (Fichman et al., 2021), and AtPIP1;4 mediating H₂O₂ triggered immunity against pathogen attack (Tian et al., 2016). Our results showing the extreme H₂O₂ sensitivity conferred by expression of AtPIP1;3/1;4 paralogs in skn7 yeast suggests these paralogous AtPIPs have evolved high apparent H₂O₂ transport capacity with largely overlapping tissue-specific expression patterns in planta. This functional and tissue expression redundancy suggests that AtPIP1;3 could also mediate H₂O₂ signaling for plant immunity. Supporting this idea, H₂O₂ translocation into the cell is decreased but not eliminated in the atpip1;4 single mutant (Tian et al., 2016); and only AtPIP1;4 and AtPIP1;3 are rapidly up-regulated in response to H₂O₂ treatment of leaves (Hooijmaijers et al., 2012). The latter would be a consistent response to the apoplastic H₂O₂ produced upon pathogen recognition and facilitating its entry into the cell to trigger immune responses (Tian et al., 2016). AtPIP1;3 transcripts are not present in dry seed, but are substantially induced during seed imbibition and germination. Hydrating seed releases H₂O₂ as a signal to promote germination, and may involve AtPIP1;3, which would be consistent with the involvement of AQPs in the germination process (Hoai et al., 2020). Further investigation into a role for AtPIP1;3 in plant immunity and seed germination appears warranted.

Five AtPIPs increased yeast sensitivity to BA exposure, indicating transport of this substrate. The relative phenotypic rankings of most to least sensitive was AtPIP1;1 > AtPIP2;2 = AtPIP2;7 = AtPIP2;8 > AtPIP1;5. AQP-mediated BA transport is generally associated with NIP-type AQPs (Pommerrenig et al., 2015). However, a growing number of PIP isoforms from different species are being found capable of transporting BA in heterologous systems; ZmPIP1;1 (Dordas et al., 2000), OsPIP1;3 and OsPIP2;6 (Mosa et al., 2016), OsPIP2;4 and OsPIP2;7 (Kumar et al., 2014), and HvPIP1;3 and HvPIP1;4 (Fitzpatrick and Reid, 2009).

The AtPIPs identified in this study as candidates for BA transport, are expressed in all tissue types in planta and may help coordinate uptake and distribution of this essential micronutrient, and provide tolerance via efflux under toxic concentrations. Interestingly, AtPIP1;1 increased yeast sensitivity to BA, but its paralog AtPIP1;2 did not. In Arabidopsis, AtPIP1;1 expression is unaltered in roots and minimally in shoots under toxic boron conditions, whereas AtPIP1;2 is substantially repressed (Macho-Rivero et al., 2018). This suggests AtPIP1;1 may have undergone substantial functional diversification since duplication with AtPIP1;2. AtPIP1;2 is widely and highly expressed throughout all tissues and facilitates CO₂ diffusion into chloroplasts for photosynthesis (Heckwolf et al., 2011), whereas we suggest AtPIP1;1 may be specialized for micronutrient uptake from the soil.

Although a native physiological role for PIP boron transport is not yet confirmed in any species, improved tolerance to boron toxicity in Arabidopsis over-expressing boron permeable rice PIPs, points towards a possible role (Kumar et al., 2014; Mosa et al., 2016).

Urea differs massively from water with respect to size, polarity, and other physicochemical properties. When expressed in the dur3 yeast strain, none of the AtPIPs improved yeast growth like that seen for the urea permeable positive control AtTIP2;3, suggesting no AtPIP was capable of permeating urea. This is consistent with urea being too large to pass through the narrow aperture of the AtPIP a/R filter ( Supplemental Table S2 ) (Dynowski et al., 2008a; Dynowski et al., 2008b).

Yeast tolerance of salt toxicity is associated with osmo-resistance (Stratford et al., 2019), meaning that AtPIP water transport could confound growth data for AtPIP expressing yeast grown at high salt concentrations. Therefore, assessment of AtPIP Na⁺ permeability from yeast growth requires a tailored mutant (Sychrova, 2004). Instead, to screen for AtPIP Na⁺ transport, we quantified intracellular yeast Na⁺ content directly. We confirmed previous reports of Na⁺ permeability for AtPIP2;1 and AtPIP2;2 (Byrt et al., 2017; Qiu et al., 2020), and observed that AtPIP2;6 and AtPIP2;7 also appear permeable to Na⁺. The latter is at odds with previous electrophysiological experiments on AtPIP2;7 expressing oocytes that report AtPIP2;7 is not permeable to Na⁺ (Kourghi et al., 2017). The contrasting findings could reflect different heterologous expression systems and detection techniques, but investigation of post-translational regulation of AtPIP2;7 function is warranted since phosphorylation of the AtPIP2;1 and HvPIP2;8 C-terminal domains have been shown to influence ion permeability (Qiu et al., 2020; Tran et al., 2020).

We observed no enhanced Na⁺ accumulation in yeast expressing AtPIP1s alone. Since the central pore, formed in the middle of tetrameric AQP complex, is the proposed pathway for monovalent ions (Yool and Weinstein, 2002), we did not screen yeast co-expressing AtPIP1s with AtPIP2;5. This would change the structure of the central pore and make interpretation of results ambiguous, as seen for CO₂ and Na⁺ transport through the central pore of PIP hetero-tetramers (Otto et al., 2010; Byrt et al., 2017).

The dual permeability to water and solutes of certain AtPIPs may help build high turgor during cell expansion. For example, AtPIP2;1 is involved in lateral root emergence where the primordia pushes through the overlying tissues (Péret et al., 2012). Our observations that AtPIP2;7 has dual water and solute transport capacity and is upregulated during seed imbibition and germination, implies a role aiding the massive influx of water needed for the radicle to puncture through the seed coat. Moreover, expression of AtPIP2;7 in seeds responds to two antagonistically acting phytohormones (GA and ABA) that regulate seed dormancy versus germination (Hoai et al., 2020).

We observed differences among AtPIPs in their relative phenotypic rankings when tested for water, H₂O_2, BA or Na⁺ transport. This is puzzling given the near identical residue signatures of motifs classically considered to govern substrate selectivity (i.e. NPA, ar/R, and Froger’s positions) ( Supplemental Table S2 ; Figure S10 ), and indicates the involvement of other domains yet to be defined. Variation in transport efficiency for water and H₂O₂ is likely to be associated with subtle differences in residues forming the monomeric pore that alter the number of hydrogen bonds with the substrate, or that shift, even slightly, the spatial configuration of the pore diameter (Horner et al., 2015; Mom et al., 2021). Differences in the sensitivity of gating regulation and the degree of ‘openness’ or ‘open probability’ is another possible factor not only for regulating the capacity for transport but also switching between substrate preferences possibly through shifting between monomeric versus central pores (Kourghi et al., 2017; Vitali et al., 2019; Qiu et al., 2020; Lu et al., 2022).

Increased Na⁺ accumulation was only detected for some AtPIPs, pointing to differences in central pore features (Yool and Weinstein, 2002). The route for BA through PIPs is unknown, but mutant analysis suggests the monomeric pore is most likely (Dynowski et al., 2008b). However, we cannot exclude the central pore given its hydrophobic profile and hypothesized ability to open wider through helix rotation (Tyerman et al., 2021). Structural changes to the central pore of hetero-tetramers would also account for the inability to improve AtPIP1;1 and AtPIP1;5 boric acid permeability when co-expressed with AtPIP2;5.

The limited sequence differences between the AtPIPs ( Supplemental Figure S10 ), should make identification of substrate specificity residues easier and feasible to explore through mutation approaches.