Directional auxin fluxes in plants by intramolecular domain–domain coevolution of PIN auxin transporters

Summary Morphogenesis and adaptive tropic growth in plants depend on gradients of the phytohormone auxin, mediated by the membrane‐based PIN‐FORMED (PIN) auxin transporters. PINs localize to a particular side of the plasma membrane (PM) or to the endoplasmic reticulum (ER) to directionally transport auxin and maintain intercellular and intracellular auxin homeostasis, respectively. However, the molecular cues that confer their diverse cellular localizations remain largely unknown. In this study, we systematically swapped the domains between ER‐ and PM‐localized PIN proteins, as well as between apical and basal PM‐localized PINs from Arabidopsis thaliana, to shed light on why PIN family members with similar topological structures reside at different membrane compartments within cells. Our results show that not only do the N‐ and C‐terminal transmembrane domains (TMDs) and central hydrophilic loop contribute to their differential subcellular localizations and cellular polarity, but that the pairwise‐matched N‐ and C‐terminal TMDs resulting from intramolecular domain–domain coevolution are also crucial for their divergent patterns of localization. These findings illustrate the complexity of the evolutionary path of PIN proteins in acquiring their plethora of developmental functions and adaptive growth in plants.

Previous analyses of PIN protein structure have identified a tripartite domain structure that includes two predicted transmembrane domains (TMDs) at both the N-and C-termini and a central hydrophilic loop (HL) (K re cek et al., 2009). Compared to the HL, the TMDs are relatively evolutionarily conserved among PIN family members (Bennett et al., 2014). From an evolutionary perspective, PIN genes have been found in almost all land plant lineages (Viaene et al., 2013(Viaene et al., , 2014Bennett, 2015), in agreement with the presence of auxin signaling components and auxin responses in land plants (Mutte et al., 2018). Recently, an ancestral form of PIN was identified in the green alga Klebsormidium, which shows plasma membrane (PM) localization in algal cells and was verified to transport auxin molecules as well (Skokan et al., 2019). In terms of phylogenetic relationships and subcellular localization, the PIN protein family members are generally grouped into three types: the canonical, PM-localized PINs that mediates cell-to-cell auxin flow for intercellular auxin homeostasis; the noncanonical, endoplasmic reticulum (ER)-localized PINs that presumably mediate auxin exchange between the cytosol and ER-lumen possibly contributing to intracellular auxin homeostasis; and dual PM-and ER-localized PINs such as Arabidopsis PIN6 with unclear function (Mravec et al., 2009;Ding et al., 2012;Cazzonelli et al., 2013;Bennett et al., 2014;Ganguly et al., 2014;Verna et al., 2015;Simon et al., 2016). Additionally, the canonical PINs are polarly distributed at the PM of specific cells to guide directional auxin flow (Wisniewska et al., 2006;Glanc et al., 2018). For instance, in the model flowering plant Arabidopsis, four of five canonical PINs, PIN1/3/4/7, are localized to the basal side of root stele cells to mediate the formation of an auxin maximum in root tips that is recognized as the pattern-or organ-organized signal (Friml et al., 2002;Billou et al., 2005), while PIN2 is apically localized in root epidermal cells to exclusively transport auxin from the root tip to the elongation zone to mediate root gravitropic growth (M€ uller et al., 1998;Baster et al., 2013;Zhang et al., 2019).
Despite great advances in our understanding of the crucial role of PIN proteins at the different subcellular destinations for plant development and adaptive growth, we still have limited knowledge of the molecular cues that determine these divergent localizations of PIN family members. Uncovering the mysteries of sequence-dependent diversification of the subcellular and polar localization of PIN proteins will not only reveal the evolutionary processes that underlie the establishment of divergent developmental roles of PIN proteins, but will also open a window into a better understanding of the evolution of higher plants with their richness of forms and multifaceted growth adaptability to different environmental conditions.

Plant materials and growth conditions
The loss-of-function mutant pin2 was previously described (M€ uller et al., 1998). Arabidopsis thaliana seedlings were grown vertically in Petri dishes on ½ Murashige & Skoog (MS) medium (pH 5.9) containing 1% sucrose and 0.8% agar at 18°C under a long-day light regime (light intensity: 250 µmol m -2 s -1 ). The vertical growth index (VGI) of Arabidopsis roots were measured as previously described (Grabov et al., 2005). The parameters for the calculation of VGI were measured using IMAGEJ software on > 10 seedlings per genotype (NIH; http://rsb.info. nih.gov/ij).

Vector construction and plant transformation
To construct the PIN-GFP (green fluorescent protein) fusion proteins, the GFP gene was fused in-frame to the central HL domain of PIN protein open reading frames by performing overlapping PCR. To construct the chimerical PIN proteins containing combinations of TMDs and HLs from different PIN family members, the central HL and TMD from the Nor C-termini were cloned individually, after which we performed overlapping PCR to fuse the three fragments together. Following this, the PIN-GFP and chimeric PIN PCR products were individually cloned into the Gateway entry vector pDONR221 by the BP reaction. A 1.4 kb fragment containing the PIN2 promoter was cloned into the Gateway entry vector pPONRP4P1r, and the two were then fused and cloned into the Gateway destination vector pB7m24GW.3 by the LR reaction. The primers used to generate these constructs are shown in Supporting Information Table S1. Transgenic Arabidopsis plants were generated using the floral dip method and selected on solid, 0.59 MS medium containing 15 lg ml À1 of Basta herbicide (glufosinate). To confirm successful plant transformation, following herbicide selection, seedlings carrying the GFP fusion proteins were further selected by examination under the fluorescence microscope.

Confocal microscopy
To analyze the polarity and subcellular localization of proteins in the transgenic lines described above carrying the GFP-PIN fusion proteins, the plants were separately crossed with pPIN2::PIN2-mCherry and the ER marker line 35S::HDEL-RFP (Liu et al., 2011). We then observed the GFP, mCherry and RFP (red fluorescent protein) signals in 5-d-old Arabidopsis root tips using a Zeiss 800 inverted confocal laser scanning microscope.

Identification of coevolving sites
The PIN protein sequences used for the coevolution analysis were retrieved from PHYTOZOME (https://phytozome.jgi.doe.gov/pz/ portal.html#) in 15 plant species (Dataset S1), including A. thaliana, Populus trichocarpa, Capsella rubella, Brassica oleracea capitata, Carica papaya, Medicago truncatula, Crocus sativus, Gossypium raimondii, Glycine max, Vitis vinifera, Sorghum bicolor, Brachypodium distachyon, Oryza sativa, Zea mays and Ananus comosus. Identification of pairs of coevolving sites in the correlated regions of N-and C-terminal TMDs of PIN proteins was achieved by running the SPIDERMONKEY/BGM program (Poon et al., 2008), which was implemented in the HYPHY software package (Kosakovsky Pond et al., 2005). This reconstructs the substitution history of the alignment by maximum likelihoodbased phylogenetic methods, and then analyzes the joint distribution of substitution events using Bayesian graphical models to identify significant associations among sites. We set the minimum number of nonsynonymous substitutions per codon site to 1, and the maximum number of parents per node to 2. The PIN amino-acid sequences are run under default settings (a 10 4 'burnin' followed by 10 5 steps thinned to 100). A posterior probability value ≥ 0.5 was used as the definition of association between coevolving sites.

Sequence-based determinants of Arabidopsis PIN protein subcellular localizations
In Arabidopsis, there are eight PIN protein family members. In terms of sequence similarity and the length of the central HL, they are grouped into the following three evolutionary clades (Bennett et al., 2014;Adamowski & Friml, 2015): the canonical PINs (PIN1, PIN2, PIN3, PIN4 and PIN7) with long HLs (> 350 residues); two noncanonical PINs (PIN5 and PIN8) that have short HLs (< 50 residues); and another noncanonical PIN (PIN6) with an HL of intermediate length (> 250 residues) (Fig. 1a). To study the properties of their sequence-dependent subcellular localization, we used the Arabidopsis root epidermal cell as the unified cell type and expressed all these PIN proteins under control of the PIN2 promoter (Fig. 1b).
PIN1 and PIN2 are localized to the PM of root epidermal cells ( Fig. 1b-

Research
New Phytologist cell-type-specific determinants, these diverse localizations are based on sequence-specific signals.
Role of PIN hydrophilic loop in determination of canonical PIN polarity at the PM As shown above, the canonical PINs, PIN1 and PIN2, localize to the PM in root epidermal cells, but they show differences with respect to the polarity of their localization. Specifically, PIN2 is apically localized in root epidermal cells, as revealed by anti-PIN2 immunolocalization (M€ uller et al., 1998) and by tagging with GFP or mCherry (Figs 1c, S2a). PIN1, however, shows a basal or nonpolar localization pattern demonstrated by the failure of PIN1-GFP to be intimately colocalized with apically localized PIN2-mCherry in epidermal cells (Figs 1d,S2b). We then wanted to know which topological structural parts of PIN proteins confer the molecular signals that determine this differential polarization in the PM.
Intragenic domain swapping experiments revealed that when either the N-or C-terminal TMD of PIN2 was replaced with the N-or C-terminal TMD from PIN1, the two chimeric PIN proteins (named V1 and V2, respectively) still showed the predominantly apical cellular localization in epidermal cells, as verified by their colocalization with PIN2-mCherry (Figs 1e,f, S2c,d). Additionally, when both the N-and the Cterminal TMDs of PIN2 were simultaneously replaced with the N-and C-TMDs from PIN1, this chimeric PIN protein (V3) still showed apical localization in the PM, again confirmed by its colocalization with PIN2-mCherry (Figs 1g, S2e). Together, these results suggest that for the canonical PINs, the HL rather than the TMDs probably underwent evolutionary selection and thus determines their polar localization in the PM. To verify this hypothesis further, we replaced the central HL of PIN2 with that of PIN1, and this chimeric PIN (V4) indeed lost the ability to localize to the apical side of the epidermal cells, as indicated by its uncoupled colocalization with PIN2-mCherry (Figs 1h,j, S2f). Moreover, besides in the root epidermal cells, we also analyzed the cellular localizations of these PIN proteins and chimeras in cells of the lateral root cap (LRC) and cortex under control of the PIN2 promoter ( Fig. S3). Consistent with their polarity analysis in root epidermal cells, the PIN1 and chimeric protein V4 failed to be apically localized in LRC cells, while the PIN2 and chimeras V1-V3 with the HL of PIN2 were localized to the apical side in this cell type (Fig. S3). These findings revealed a crucial role of the HL in the polarity of canonical PIN localization. Notably, both PIN1 and PIN2 showed a basal localization in cortical cells as previously reported (Wisniewska et al., 2006), and all chimeric PIN proteins V1-V4 were basally localized as well (Fig. S3). This indicates that cell type also contributes to the determination of PIN polarity.
The localization of PIN proteins on the basal vs apical sides of the cells is tightly linked to their physiological functions in plant adaptive growth (Glanc et al., 2018). To analyze the functions of these chimeric PIN proteins, we introduced them into the pin2 mutant under control of the PIN2 promoter. The PIN2-GFP fusion protein was able to restore the pin2 defect in root gravitropism (Figs 1c, S4), whereas PIN1-GFP failed to rescue this defect (Fig. 1d). The apically localized chimeric PIN proteins V1, V2 and V3 nicely complemented the pin2 mutant phenotype ( Fig. 1e-g), whereas the basally PM-localized chimeric PIN (V4) failed to complement the pin2 mutant gravitropic growth defect (Fig. 1h,i).
Moreover, in contrast to PIN2-GFP, expression of the PIN1-GFP fusion protein or the chimeric PIN (V4) protein under control of the PIN2 promoter in the wild-type background resulted in agravitropic root growth (Fig. 1k), suggesting that the basally PM-localized PIN1-GFP and V4-GFP fusion proteins may interfere with apical (shootward) auxin flow directed by apically localized endogenous PIN2 protein and therefore interfere with root gravitropism.
These results are congruent with the cellular polar localization of these chimeric PIN proteins and confirm the crucial role of the apical PIN2 localization in its function for root gravitropism. Fig. 1 The divergent hydrophilic loop (HL) domains confer the polarity of canonical PIN. (a) Phylogenetic analysis of the eight Arabidopsis PIN gene proteins based on their sequence similarities and subcellular localization. The unrooted tree is divided into three clades as indicated by the three black dots. (b) Subcellular localization analysis of four PIN proteins from three clades in Arabidopsis root epidermal cells: PIN genes tagged with GFP were expressed under control of the PIN2 gene promoter. PIN1 and PIN2 represent the canonical PIN clade, while PIN6 and PIN5 represent the other two noncanonical clades, respectively. (c) Coexpression of PIN2-GFP with PIN2-mCherry showed their intimate colocalization to the apical side of the root epidermal cells (middle panel), and this apically localized PIN2-GFP successfully rescued the defective root gravitropism phenotype of the pin2 mutant (bottom panel). (d) the PIN1-GFP fusion protein shows nonpolar localization in root epidermal cells as indicated by its uncoupled colocalization with PIN2-mCherry (middle panel), and it also failed to rescue the pin2 mutant phenotype (bottom panel). (e-h) Top panels: schemes for the intragenic domain swapping experiments between PIN1 and PIN2. These chimeric PIN proteins are composed of the transmembrane domains (TMDs) at the N-and C-termini and a central HL from PIN1 or PIN2. Middle panels: coexpression analyses of these chimeric PIN proteins with apically PM-localized PIN2-mCherry; V1 has the N-TMD of PIN1 and the C-TMD/HL of PIN2 (e), V2 has the C-TMD of PIN1 and the N-TMD/HL of PIN2 (f), and V3 has the C-and N-TMDs of PIN1 and the HL of PIN2 (g). The nonpolar localization of V4, which consists of the C-and N-TMDs of PIN2 and the HL of PIN1, is shown in (h). All experiments were conducted in Arabidopsis root epidermal cells. Bottom panels: genetic complementation experiments show that only the apically localized PIN proteins (PIN2, V1, V2 and V3) complement the defective gravitropism phenotype in the pin2 mutant (c, e-g). (i) Quantification of the vertical growth index (VGI, as explained in Supporting Information Fig. S2b) of the transgenic Arabidopsis lines. Center lines show the medians; box limits indicate the 25 th and 75 th percentiles as determined by R software; whiskers extend 1.5 times the interquartile range from the 25 th and 75 th percentiles, and outliers are represented by dots. (j) Polarity index of the PIN protein localization in the root epidermal cells (n = 100 cells from 10 roots). (k) In contrast to the apically localized PIN2, expression of the PIN1 and V4 proteins in the wild-type background under control of the PIN2 promoter may antagonize the endogenous PIN2 function and lead to the agravitropic phenotype. The white arrowheads (b-h) indicate the direction that the PIN proteins localize to. Scale bars, 10 µm.

Role of hydrophilic loop in determining PM vs ER localization of PIN proteins
Compared with the PM-localized canonical PIN, the noncanonical PIN6, which possesses an HL of intermediate length, displays a dual localization in both the PM and the ER membrane (Fig. 1b). To reveal which topological structural part between PIN2 and PIN6 leads to their divergent subcellular localization, we systematically constructed various hybrid PIN proteins by domain swapping. First, we replaced the HL of PIN6 with the PIN2-HL and called this chimeric PIN protein X1. When compared with PIN6, the X1 chimera displayed the predominant PM localization similar to PIN2 (Figs 2a-c, S5), suggesting that the HL harbors molecular cues for the ER vs PM subcellular localization. Additionally, similar to PIN2, the hybrid PIN X1 showed the predominantly apical subcellular localization, as manifested by colocalization of X1-GFP and PIN2-mCherry in root epidermal cells (Fig. S5a-c). Consequently, X1 was also able to partially rescue the defective root gravitropism phenotype of the pin2 mutant (Fig. 2c).
Next, we swapped the N-TMD of PIN2 with that of PIN6 to make a chimeric protein that we called X2. Notably, as compared to the X1 chimera with both N-and C-TMDs from PIN6, the chimeric X2 protein has much greater domain similarity to PIN2; nonetheless, X2 showed increased localization at the lateral sides of the root epidermal cell PM, as shown by coexpressing it with PIN2-mCherry, and accordingly failed to rescue the defective root gravitropism of pin2 (Figs 2d, S5d). Unexpectedly, in root epidermal cells, the chimeric PIN X3, in which we substituted the C-TMD of PIN2 with that of PIN6, was predominantly localized to the ER, as confirmed by its coexpression analysis with PIN2-mCherry and the ER marker HDEL-RFP, which was different from the subcellular localization of PIN2, X1 and PIN6 (Figs 2e,g-i, S5e). Consequently, the ER-localized X3 protein also failed to complement the pin2 mutant phenotype (Fig. 2e,f). Together, these findings imply that besides the HL, the pairwise matching of N-and C-terminal TMDs contributes to the subcellular localization of PIN proteins. Consistent with this idea, compared with the PM localization of X1 and X2, the chimeric PIN protein X3 was also localized to the ER in LRC/ cortex cells similar to its subcellular localization in root epidermis (Fig. S6). Notably, while the expression of ER-localized X3 under control of the PIN2 promoter in the wild-type background showed normal gravitropic root growth (Fig. 2j), expression of nonpolar PIN6 might interfere with the apically PM-localized PIN2 function and therefore affect normal root gravitropism.
Overall, these results suggest that while the PIN hydrophilic loop contains the main sequence-based determinants of ER vs PM localization, the N-and C-terminal trans-membrane domains also need to be mutually matched to allow correct PIN subcellular localization.
Crucial role of pairwise matching between the N-and C-TMDs in PIN subcellular localization Next, we performed further domain swapping experiments between PM-localized PIN2 and ER-localized PIN5 (Fig. 3a,b).
When the short central HL of PIN5 was replaced by the long HL from PIN2, the chimeric PIN protein, called C1, displayed PM localization, further demonstrating the crucial role of the HL domain in subcellular localization of the PIN proteins (Fig. 3c). The colocalization of C1-GFP with PIN2-mCherry confirmed its apical localization at the PM, although C1 also shows more lateral localization in epidermal cells as compared to PIN2. Moreover, C1-GFP partially restores the pin2 mutant defects in root gravitropism (Figs 3c, S7a-c).
To further test the possible role of the pairwise matched Nand C-TMDs in the subcellular localization of PIN proteins, we replaced either the N-or the C-TMD of PIN2 individually with the corresponding part of PIN5, to generate the chimeric proteins called C2 and C3, respectively (Fig. 3d,e). When compared to C1, the C2 and C3 chimeras shared many more common domains with PIN2; nonetheless, unexpectedly, unlike PIN2 and C1, both C2 and C3 were found to be predominantly localized to the ER but not to the PM in epidermal cells, as indicated by colocalization analysis with PIN2-mCherry (Figs 3a-e, S7a-e) and HDEL-RFP (Fig. 3h-j). Likewise, in the LRC/cortex cell types, both the chimeric PIN proteins C2 and C3 were localized to the ER while the C1 was predominantly localized to the PM (Fig. S8). Consistent with their ER localization pattern, both C2 and C3 failed to rescue the gravitropic defect of the pin2 mutant (Fig. 3a,d-e). These results support the suggestion that the mutual matching of TMDs between the N-and C-termini is crucial for the correct subcellular localization of PIN proteins, and also strongly implies that the PIN proteins underwent an intramolecular domain-domain coevolution to ensure compatibility between the domains and thus jointly determine the subcellular membrane localization of PIN proteins.
Furthermore, to evaluate the roles of sequence variation of TMDs between the canonical and noncanonical PINs in subcellular localization, we replaced both N-and C-TMDs of PIN5 with those of PIN2 to create a chimeric protein called C4. Unexpectedly, in contrast to the predominantly ER-localized PIN5 (Fig. 3b), C4, with the short HL of PIN5, showed a dual localization to both the PM and the ER, as indicated by its colocalization with both PIN2-mCherry and HDEL-RFP (Fig. 3f,k). This suggests that not only the HL, as previously shown (Figs 2c, 3c), but also the sequence variation in the TMDs from different clades of PIN proteins contributes to their differential subcellular localizations. Consistent with the nonpolar distribution of the C4 protein, which localizes to the PM (Fig. S7f), expressing C4 in the wild type background under control of the PIN2 promoter probably interferes with the apically PM-localized endogenous PIN2 function and therefore leads to the observed defect in root gravitropism (Fig. 3l). By contrast, the transgenic lines expressing the ER-localized PIN5 and chimeric PIN proteins (C2 and C3) in wild-type background driven by the PIN2 promoter showed normal gravitropic root growth (Fig. 3l).
Thus, again PM-localized but not ER-localized PIN proteins can interfere with the function of PM-based endogenous PIN proteins. It also confirms that the sequence variation of TMDs and the compatibility between N-and C-terminal TMDs Unmatched N-and C-TMDs between PIN5 and PIN6 As shown above, the N-and C-TMDs of PIN6 and PIN5 do not match the counterpart TMDs of PIN2, leading to the retention of X3, C2 and C3 in the ER. We therefore postulated that after the evolutionary divergence of different PIN clades (Fig. 1a), intramolecular domain-domain coevolution occurred independently within every clade, resulting in unmatched N-and C-TMDs, including between PIN5 and PIN6 clades as well.
To test this hypothesis, we further swapped the TMDs between PM-localized X1 and C1, whose N-/C-terminal TMDs came from PIN6 and PIN5, respectively (Fig. 4a,b). Two hybrid PIN proteins were then constructed, named XC1 and XC2, which had one terminal TMD derived from PIN5 and the other from PIN6 (Fig. 3c,d). Both XC1 and XC2 completely failed to rescue the pin2 phenotype, and showed a predominantly ER subcellular localization in epidermal cells, as indicated by coexpressing them with PIN2-mCherry or HDEL-RFP ( Fig. 4c-g). Moreover, analyzing the subcellular localization of the XC1 and XC2 in cell files of LRC and cortex further confirmed the prominent ER localization of the two chimeric PIN proteins (Fig. S8g, Fig. 3 The sequence-divergent transmembrane domains (TMDs) contribute to the subcellular localization of PIN proteins to the PM or ER. (a-f) Top panels: structural diagrams of PIN2, PIN5 and the chimeric PIN proteins (C1, C2, C3 and C4) with the domains swapped between a canonical PIN (PIN2) and a noncanonical PIN (PIN5). Middle panels: subcellular localization analyses of the PINs in root epidermal cells by their coexpression with the apically PMlocalized PIN2-mCherry. The chimeric PIN, C1, which consists of the C-and N-TMDs from PIN5 and the hydrophilic loop (HL) domain from PIN2, is localized to the PM as shown by its coexpression with PIN2-mCherry (c), while the chimeric C2 protein, with the N-TMD from PIN5 and the C-TMD and HL from PIN2, and C3, with the C-TMD from PIN5 and the N-TMD and HL from PIN2, are mainly localized to the ER (b, d, e). Unlike both PIN2 and PIN5, the chimeric C4 protein, which has the C-and N-TMDs from PIN2 and the HL domain from PIN5, is localized to both the PM and the ER (f). Bottom panels: the results of genetic complementation experiments indicating that C2, C3 and C4 failed to complement the pin2 mutant phenotype ( . In addition to the sequence divergence of the TMD and hydrophilic loop (HL) domains among these PIN clades, we highlight the crucial role of the pairwise matching between the N-and C-TMDs in subcellular localization, which strongly suggests independent intramolecular coevolution of PIN proteins after evolutionary divergence of the PIN clades. Scale bars, 10 µm.

Ó2020 The Authors
New Phytologist Ó2020 New Phytologist Trust New Phytologist (2020) 227: 1406-1416 www.newphytologist.com h). Additionally, wild-type plants expressing XC1 and XC2 under control of the PIN2 promoter showed normal root gravitropic growth, consistent with the ER-subcellular localization of these chimeric proteins (Fig. S9). These results show that not only are the N-and C-TMDs between PIN2 and PIN5 or PIN6 not pairwise matched, but also the TMDs between PIN5 and PIN6 do not match, suggesting that after the PIN5 and PIN6 clades diverged (Fig. 1a), the TMDs of these two clades of PIN proteins also underwent independent evolution, leading to the uncoupling of pairwise matched TMDs between PIN5 and PIN6.
Finally, to further confirm the intramolecular coevolution of PIN protein, we performed intraprotein coevolutionary analysis using several PIN members from various plant species. Bioinformatics analysis identified a multitude of pairs of coevolving sites (11 pairs of amino acids in total) at the correlated regions of Nand C-TMDs (Table S2), suggesting strongly that the intramolecular coevolution of N-and C-transmembrane domains of PIN auxin transporter occurred during PIN evolution, which is consistent with our experimental results. Additionally, it also showed clearly that for most pairs of identified coevolving sites, the two correlated amino acids, separately located at the N-and C-TMDs, were simultaneously substituted/mutated among three different PIN clades (i.e. PIN2, PIN6 and PIN5) ( Table S2). The correlated mutations/substitutions of the coevolving sites imply the important role of the pairwise matching of N-and C-terminal TMDs for localization and function of PIN proteins.

Discussion
PIN protein family members, which comprise the crucial polar auxin transporters, are ubiquitous across the plant kingdom and evolved to be localized to diverse membrane compartments including the ER and the different polar domains at the PM. There is a long-standing question as to why the PIN protein family members with similar topological structures evolved to localize to different cellular compartments and thus flexibly transport auxin within or between cells.
Our work provided initial insights into how the variation in different domains of PIN proteins contribute to their differential membrane localization (Fig. 4h): the central hydrophilic loops of PIN proteins are the main determinants of PIN's subcellular localization, as evidenced by the fact that the ER-localized PIN6 and PIN5 proteins are targeted to the PM after swapping their HLs with that of PM-localized PIN2; the transmembrane domain modules also contribute to the subcellular localization, as replacing the N-and C-TMDs of the predominantly ER-localized PIN5 with those of PIN2 results in the dual localization of the chimeric PIN protein C4 to both the PM and the ER (Fig. 3f); and the pairwise matching between the N-and C-TMDs also plays a crucial role in determining the subcellular localization, which is evidenced by the predominantly ER localization of a series of chimeric PINs combining the 'long' HL domain of the PMlocalized PIN2 with the hybrid N-and C-TMDs from two different PIN clades. This points to the intramolecular domain-domain coevolution after the divergence of the PIN protein clades. In the future, revealing the PIN structure would be very helpful in understanding the underlying mechanism of PIN subcellular localization determined by pairwise matched N-and C-TMDs. Additionally, TMDs exist ubiquitously in membrane proteins, so an outstanding question remains whether the coevolution and matching of the TMDs are more commonly implicated in determining the subcellular localization and function of other membrane proteins.
Our systematic domain swapping experiments among the PIN family members not only elucidated the divergent PM/ER localization of the PIN proteins, but also provided insights into how the polar PM localization (apical vs basal) is determined by the domain variation in PIN proteins (Fig. 4h). The HL plays an important role in the polar localization of the canonical PIN, as shown by the intergenic swapping of the HL domain between PIN2 and PIN1 (Fig. 1). Interestingly, replacing the HL domain of the ancestral form of the canonical PIN, Marchantia polymorpha PINZ (MpPINZ), with the HL of Arabidopsis PIN2, caused its polarity to switch from the basal side to the apical side in epidermal cell PM, and therefore successfully restored the defective root gravitropism of the pin2 mutant (Bowman et al., 2017;Zhang et al., 2019). These findings imply strongly that the HL domain of the canonical PIN family members experienced evolutionary selection, and therefore led to their divergent cellular localizations and functions during plant evolution. Furthermore, the TMDs also contribute to the polar localization of PIN proteins; replacement of the TMDs of PIN2 with those of PIN5 resulted in enhanced lateral localization when compared with the original PIN2 protein (Figs 3c, S7c).
In the green alga Klebsormidum, only one PIN gene (KfPIN) has been identified, and the protein it encodes is PM-localized and also transports auxin with high specificity (Skokan et al., 2019). In the representatives of early diverging land plants, Marchantia and other bryophytes, the PIN proteins have diverged into two clades: the canonical PINs localized in the PM (e.g. Physcomitrella PINA/ B/C), and noncanonical PINs localized to the ER (e.g. Physcomitrella PIND) (Viaene et al., 2014). In the higher plant Arabidopsis, however, in addition to these two PIN clades, there is another particular PIN protein, PIN6, that shows dual localization to both the PM and the ER (Simon et al., 2016). The different localization between chimeric X2 protein (PM-localized) and X3 protein (ER-localized) implies that the N-TMD of PIN6 is able to partially match with the C-TMD of PIN2 (Fig. 2d), while the C-TMD of PIN6 is fully unmatched with the C-TMD of PIN2 (Fig. 2e). However, the C2 and C3 chimeras showed the same subcellular localization at the ER (Fig. 3d,e), indicating that the N-and C-TMDs between PIN2 and PIN5 are mutually unmatched. These results suggest strongly that, despite the fact that PIN6 is a noncanonical PIN, it probably shares much closer common descent with canonical PINs such as PIN2, rather than the ER-localized noncanonical PIN proteins (e.g. PIN5), which is further supported by our domain swapping results that the TMDs of PIN5 and PIN6 are fully unmatched, leading to the ER localization of these chimeric PIN proteins XC1 and XC2 (Fig. 4c,d).

New Phytologist
Moreover, in contrast to the PM-localized X1, the chimera X2, localized to the PM, failed to similarly partly revert the pin2 agravitropic root phenotype (Fig. 2c,d), suggesting that pairwise matching of N-and C-TMDs might contribute to PIN auxin transport activity, in addition to the contribution to the subcellular localization of PIN proteins.
Recently, evidence has accumulated to demonstrate the existence of intramolecular coevolution within proteins and intermolecular coevolution among interacting proteins. Specifically, changes in the amino acid residues of one protein domain or one protein would impose evolutionary pressure on the other domains or proteins with which it interacts, thus resulting in corresponding residue changes in order to maintain or re-establish the domain-domain or protein-protein interaction (Yeang & Haussler, 2007;Lynch & Hagner, 2015;Cong et al., 2019). Here, our results imply that the intramolecular coevolution also took place in PIN proteins, providing additional insight into the evolution of membrane proteins.
It has been proposed that the differential cellular polarity and subcellular localization of PIN proteins relies on cargo recognition during protein sorting (Adamowski & Friml, 2015;Zwiewka et al., 2019). However, it remains to be addressed whether the subcellular localization determined by this pairwise matching of N-/C-TMDs results from such failed cargo recognition processes or could be due to protein misfolding, given that the misfolded protein may fail to be sorted to the PM and therefore be retained in the ER.