Phospholipid composition and a polybasic motif determine D6 PROTEIN KINASE polar association with the plasma membrane and tropic responses

The phytohormone auxin is essential for plant differentiation and growth (Leyser, 2010; Teale et al., 2006). Seemingly all organ development as well as all dynamic tropic growth responses, e.g. phototropism and gravitropism, rely on differential auxin distribution patterns and auxin-induced cellular responses (Adamowski and Friml, 2015; Wang and Estelle, 2014). Auxin levels within the plant are controlled by local auxin metabolism and directional auxin transport, which is achieved by the PIN-FORMED (PIN) efflux and the AUXIN RESISTANT 1/LIKE AUX1 (AUX1/LAX) influx carriers (Adamowski and Friml, 2015). PINs are polarly distributed in the plasma membranes of many cells and their differential distribution results in the formation of local auxin maxima and minima that orchestrate the development of the growing plant body (Adamowski and Friml, 2015; Grunewald and Friml, 2010; Habets and Offringa, 2014).

Studies on directional auxin transport and PIN polar distribution have become attractive systems to study cell polarity in plants (Pan et al., 2015; Sachs, 1991; Yang, 2008). Although it is still unclear whether the targeting of de novo synthesized PINs is polar or apolar, it is well established that the maintenance of PIN polarity requires constitutive recycling in combination with a reduced lateral diffusion in the plasma membrane (Feraru et al., 2011; Kleine-Vehn et al., 2011). PIN polarity is seemingly also critically dependent on the activity of the AGCVIII protein kinase PINOID (PID) and PIN phosphorylation (Friml et al., 2004; Kleine-Vehn et al., 2009; Michniewicz et al., 2007; Offringa and Huang, 2013). Tissue-specific PIN polar distribution patterns can sufficiently explain auxin distribution gradients across plant organs like roots and shoots (de Reuille et al., 2006; Grieneisen et al., 2007; Heisler et al., 2010; Mironova et al., 2010; van Berkel et al., 2013; Wabnik et al., 2011). Additionally, gravity as well as light-dependent changes in PIN polarity correlate with observed changes in auxin distribution required for tropic growth processes (Ding et al., 2011; Kleine-Vehn et al., 2010; Rakusová et al., 2011). Thus, understanding PIN polarity and activity control at the plasma membrane can substantially contribute to the understanding of plant growth.

PIN-mediated auxin transport is critically dependent on phosphorylation by protein kinases (Barbosa and Schwechheimer, 2014). The AGCVIII serine/threonine kinase D6 PROTEIN KINASE (D6PK) but also the proposed PIN polarity regulator PINOID (PID) activate PIN-mediated auxin transport through PIN phosphorylation when tested in Xenopus laevis oocytes (Friml et al., 2004; Zourelidou et al., 2014). Arabidopsis mutants of D6PK and the orthologous D6PK-LIKE genes D6PKL1-D6PKL3 are compromised in directional auxin transport and PIN-dependent growth (Willige et al., 2013; Zourelidou et al., 2009). Conversely, pharmacological interference with D6PK abundance at the plasma membrane impairs auxin transport-dependent tropic responses (Barbosa et al., 2014). Although both D6PK and PINs are polarly localized at the plasma membranes of cells of different plant tissues, their recycling kinetics are dramatically different and obviously mediated by different trafficking pathways (Barbosa et al., 2014; Geldner et al., 2003; Zourelidou et al., 2009). Understanding how D6PK associates with the plasma membrane is thus important for understanding D6PK-mediated auxin transport and plant growth regulation.

Peripheral membrane proteins can be anchored to membranes by diverse mechanisms that include protein-protein and protein-lipid interactions (Cho and Stahelin, 2005). The very low abundant phosphoinositides (PIs) are important players for the recruitment of peripheral proteins to membranes (Balla, 2013; Heilmann, 2016). PIs are mono-, bis- and tris-phosphorylated derivatives of phosphatidylinositol (PtdIns) and PI compositions differ between the plasma membrane and different organellar membranes (Balla, 2013). Membrane-specific PI distribution, a result of the differential distribution and tight regulation of specific PI kinases or phosphatases, allows recruiting specific cognate binding proteins that determine organelle function (Behnia and Munro, 2005). In particular, asymmetric PI distribution in the plasma membrane contributes to the organization of polar domains in many organisms and biological processes (Comer and Parent, 2007; McCaffrey and Macara, 2009; Orlando and Guo, 2009; Shewan et al., 2011). In plants, the polar distribution of PIs, PI kinases, PI binding partners and associated processes are extensively studied in tip growing cells such as pollen tubes and root hairs (Kusano et al., 2008; Stenzel et al., 2008; Ischebeck et al., 2010). Analysis of PtdIns(4,5)P₂ biosensors and respective PIP5Ks (PtdIns4P 5-KINASES) revealed that PtdIns(4,5)P₂ and PIP5Ks are polarly localized in many cell types and required for PIN polarity control (Ischebeck et al., 2013; Tejos et al., 2014). More recently, PtdIns4P was identified as a major PI, conferring negative charges to the plant plasma membrane (Simon et al., 2016).

Peripheral membrane proteins can bind PIs through globular PI species-specific binding domains, e.g. the PtdIns3P-specific FYVE domain or the PtdIns(4,5)P₂-specific PH domain (Hammond and Balla, 2015). Alternatively, peripheral membrane proteins can bind membranes through ionic interactions between stretches of positively charged basic amino acids [lysine (K) or arginine (R)] and the negatively charged lipids, such as PIs or PtdOH (phosphatidic acid) (Brzeska et al., 2010; Heo et al., 2006; Jaillais et al., 2011; McLaughlin and Murray, 2005).

D6PK associates also with root hair tips following root hair initiation (Stanislas et al., 2015). PINs, the established D6PK substrates, are reportedly absent from root hair tips and D6PK might have other phosphorylation substrates and PIN-independent roles at this plasma membrane region (Stanislas et al., 2015). The association of D6PK with the root hair initiation site is dependent on PIP5K3 and, reciprocally, D6PK can bind phospholipids in vitro (Stanislas et al., 2015). A basic sequence stretch noted in several AGCVIII kinases is a good candidate region for phospholipid interactions (Zegzouti et al., 2006).

Here, we show that D6PK is able to bind phospholipids with polyacidic headgroups and that phospholipid composition critically determines D6PK plasma membrane recruitment and polarity. We further identify a polybasic K-rich motif within the D6PK middle domain that is required for D6PK phospholipid binding and membrane association as well as for proper PIN phosphorylation and PIN-dependent tropisms. Additionally, we find that K/R-rich motifs are essential for membrane and phospholipid-binding in other, as yet, uncharacterized AGCVIII kinases. Thus, phospholipid-dependent recruitment of AGCVIII kinases through ionic interactions between the polyacidic phospholipid headgroups and polybasic motifs emerge as general control mechanisms of protein-membrane binding, and in the case of D6PK for auxin transport control.

The molecular determinants of the membrane association and polar distribution of D6PK at the basal plasma membrane of root cells are unknown. In lipid overlay assays, D6PK bound to phospholipids with polyacidic headgroups, namely phosphatidic acid (PtdOH) as well as the PIs PtdIns3P, PtdIns4P, PtdIns5P, PtdIns(3,4)P₂, PtdIns(3,5)P₂, PtdIns(4,5)P₂ and PtdIns(3,4,5)P₃ (Fig. 1A). D6PK binding was strongest to PtdIns3P when using TRIS-buffered saline and to PtdOH when using PBS (phosphate-buffered saline) of comparable ionic strengths (188 mM and 181 mM) (Fig. 1A) (Stanislas et al., 2015). D6PK did not bind to phospholipids with neutral headgroups, namely PtdCho (phosphatidylcholine) and PtdEtn (phosphatidylethanolamine), or to phospholipids with monoacidic headgroups such as PtdIns and phosphatidylserine (PtdSer; Fig. 1A). In a lipid bilayer context using liposomes of mixed phospholipid compositions, D6PK bound to liposomes containing the polyacidic phospholipids PtdIns4P, PtdIns(4,5)P₂ or PtdOH, but not to liposomes containing only PtdEtn and PtdCho, phospholipids with neutral headgroups (Fig. 1B).

PI biosynthesis begins with PtdIns and involves several phosphorylation and dephosphorylation steps by PI kinases and phosphatases (Meijer and Munnik, 2003). PtdOH, conversely, is produced either from PIs by phosphoinositide phospholipase C (PI-PLC) and subsequent action of diacylglycerol kinase (DGK) or from PtdCho or PtdEtn by phospholipase D (PLD; Fig. 2A) (Meijer and Munnik, 2003). To address whether the membrane phospholipid composition determined the basal localization of YFP-tagged D6PK (YFP:D6PK) in root cells, we examined YFP:D6PK localization and distribution after 30 min treatments with established PI and PtdOH biosynthesis inhibitors (Fig. 2). For comparisons and to establish the effectiveness of the inhibitor treatments, we also analysed the localization of phosphoinositide-specific biosensors, PIN2:GFP trafficking and the uptake of the lipophylic dye and endocytosis tracer FM4-64 (Fig. S1). Interestingly, the PI3-KINASE inhibitors LY294002 and Wortmannin (WM) led to a significant depletion of YFP:D6PK from the plasma membrane (Baggiolini et al., 1987; Simon et al., 2016; Vlahos et al., 1994). Our WM and LY294002 treatments efficiently altered the localization of the PdtIns3P sensor (P18Y) but not of the PdtIns4P (FAPP1) nor PdtIns(4,5)P₂ (P24Y) sensors. With regard to PIN2:GFP, they led to PIN2:GFP accumulation in endosomes, as previously reported for WM treatments, but had no visible effects in overall FM4-64 uptake (Fig. S1) (Jaillais et al., 2006). Conversely, the PI4-KINASE inhibitor PAO caused a strong plasma membrane dissociation of YFP:D6PK (Fig. 2B,C) (Hammond et al., 2012; Simon et al., 2016). As expected, PAO treatment specifically led to the internalization of the PdtIns4P sensor (FAPP1) but, interestingly, also blocked FM4-64 uptake as well as the accumulation of PIN2:GFP in BFA compartments, suggesting that PAO treatment might block endocytosis (Fig. S1). Strikingly, short-term treatments with U73122 (+), an inhibitor of the PI-PLC-mediated hydrolysis of PIs, but not treatments with its inactive analogue U73343 (−), led to a depolarization of YFP:D6PK at the plasma membrane, concomitant with the appearance of the protein in endosomes. Effects on the seemingly polar PdtIns(4,5)P₂ biosensor or on polar PIN2:GFP could not be detected (Fig. 2B,C) (Bleasdale et al., 1990). It is likely that the increased plasma membrane dissociation by the U73122 (+) treatment resulted from the apolar redistribution and consequent reduction of YFP:D6PK signals at the basal plasma membrane where we performed our measurements (Fig. 2B,C). Surprisingly, the subcellular fractionation of protein extracts prepared from roots expressing YFP:D6PK from the 35S CaMV promoter indicated that the observed plasma membrane dissociation after WM, LY294002, PAO, and U73122 treatments was accompanied by an obvious increased solubilization of YFP:D6PK only in the case of the PAO treatment (Fig. 2B-E). For all treatments, we observed similar effects on YFP:D6PK localization, regardless of whether the fusion was expressed from the D6PK or the 35S CaMV promoter (Fig. 2; Fig. S1).

As YFP:D6PK strongly bound to PtdOH, we also tested the effects of the modulation of PtdOH abundance with R59022, an inhibitor of DGK­- and PI-PLC-dependent PtdOH synthesis (de Chaffoy de Courcelles et al., 1985; Potocký et al., 2014). R59022 led to the occasional accumulation of YFP:D6PK on endosomes, detectable only when YFP:D6PK was expressed from the D6PK promoter (Fig. 2B; Fig. S1B). Interestingly, PIN2:GFP and FM4-64 also accumulated in R59022-induced compartments (Fig. S1). At the same time, inhibition of PLD-dependent PtdOH production by 1-butanol resulted in the partial internalization of YFP:D6PK, an apparent reduction in FM4-64 uptake as previously reported, but no detectable effects on PIN2:GFP trafficking (Fig. 2B,C; Fig. S1) (Li and Xue, 2007). Conversely, anticipated increases in PtdOH abundance as a result of propanolol treatments had neither an obvious effect on YFP:D6PK localization nor on PIN2:GFP trafficking or FM4-64 uptake (Fig. 2B,C; Fig. S1). Thus, the intracellular levels of PtdOH seemingly regulate YFP:D6PK localization more subtly than PI levels, and likely in a differential manner, depending on whether the PtdOH originates from the PI-PLC or the PLD pathway. Fractionation experiments indicated that none of the PtdOH inhibitor treatments resulted in the dissociation of YFP:D6PK from cellular membranes (Fig. 2D,E). Taken together, these findings suggested that PtdIns3P, PtdIns(4,5)P₂, and PtdOH, but most importantly PtdIns4P, might be essential for YFP:D6PK plasma membrane localization and also membrane association. By contrast, PtdIns(4,5)P₂ or one of its derivatives might critically determine YFP:D6PK polar distribution.

PIP5K1 and PIP5K2, two Arabidopsis PIP5Ks isoforms that convert PtdIns4P to PtdIns(4,5)P₂, preferentially localize to apical and basal domains in root epidermal cells (Ischebeck et al., 2013). As U73122 treatments indicated that PtdIns(4,5)P₂ might control the D6PK membrane association and polarity in root epidermal cells, we introduced the YFP:D6PK transgene into the pip5k1 pip5k2 double mutant. Similarly to the reported internalization of YFP:D6PK in root hair cells of pip5k3 mutants, we noted reduced YFP:D6PK plasma membrane association in the root epidermal cells of the pip5k1 pip5k2 double mutant (Fig. 2F,G) (Stanislas et al., 2015). Thus, the reported alterations of PtdIns(4,5)P₂ levels in pip5k1 pip5k2 mutants might exert an influence on the plasma membrane targeting of D6PK (Ischebeck et al., 2013; Tejos et al., 2014).

AGCVIII kinases have an insertion between kinase subdomains VII and VIII that has been implicated in the targeting of different AGCVIII kinases (Rademacher and Offringa, 2012). In the case of D6PK and the orthologous D6PK-LIKE 1 to 3 (D6PKL1-D6PKL3), this insertion is between 101 and 104 amino acids long and designated here as the middle (MID) domain (Fig. 3A). A role of the insertion domain for the specific localization of different AGCVIII kinases has previously been proposed (Ek-Ramos et al., 2010; Zegzouti et al., 2006). In addition, the kinase domains are flanked by N- and C-terminal domains of 117 to 199 amino acids and 45 to 57 amino acids, respectively (Fig. 3A) (Rademacher and Offringa, 2012). To examine the importance of the three domains for D6PK localization, we analysed YFP:D6PK variants lacking these three domains in YFP:D6PKΔN, YFP:D6PKΔMID and YFP:D6PKΔC lines. Conversely, we analysed fusions between YFP and the three isolated domains in YFP:D6PK_N; YFP:D6PK_MID and YFP:D6PK_C lines. The deletion of any of the three domains resulted in the loss of the plasma membrane association of D6PK (Fig. 3B; Fig. S2). Regardless, each of these deletion variants remained associated with membranes, possibly endomembranes (Fig. 3C). Specifically in the case of YFP:D6PKΔMID, an endosomal localization was apparent as judged by its co-localization with the membrane lipid dye FM4-64 (Fig. 3B; Fig. S2). In turn, YFP:D6PK_N and YFP:D6PK_C were neither associated with the plasma membrane nor with membrane fractions after subcellular fractionation (Fig. 3B,C). The MID domain, however, was sufficient to recruit YFP:D6PK_MID to the plasma membrane in a polar manner, albeit with a reduced polarity when compared with YFP:D6PK, and the majority of the protein was identified in the membrane fraction after subcellular fractionation (Fig. 3B-D). Thus, the MID domain is sufficient to mediate the plasma membrane association of D6PK, and to some extent also its polar distribution, but all three domains are required for D6PK plasma membrane association.

The rapid recycling of D6PK from the plasma membrane can be blocked by Brefeldin A (BFA), an inhibitor of the ARF-GEF GNOM (Barbosa et al., 2014; Geldner et al., 2003; Zourelidou et al., 2009). Whereas wild-type YFP:D6PK was observed in intracellular compartments after BFA treatments, no obvious accumulation in BFA-compartments was observed with any of the deletion variants, indicating that their intracellular trafficking might be distinct from that of the BFA-sensitive YFP:D6PK (Fig. S2). Although YFP:D6PK_MID localized to the plasma membrane in a strongly polar manner, the protein was partially insensitive to the effects of BFA (Fig. 3B; Fig. S2). Thus, although the MID domain is sufficient for polar plasma membrane targeting, it displays different recycling and polarity properties from YFP:D6PK.

D6PK is an auto-phosphorylating kinase that trans-phosphorylates PINs (Zourelidou et al., 2014). In vitro kinase assays with the PIN1 cytosolic loop showed that D6PKΔN or D6PKΔMID retained the ability for auto-phosphorylation but lost the ability for PIN1 trans-phosphorylation (Fig. 3E). In turn, D6PKΔC had no detectable auto- or trans-phosphorylation activity (Fig. 3E). This suggested that the N-terminus and the MID domain might be required for PIN substrate recognition or trans-phosphorylation, whereas the C-terminus might be required for D6PK kinase activity. Interestingly, D6PK auto-phosphorylation was increased in D6PKΔN and D6PKΔMID when compared with D6PK, suggesting that the N-terminus and the MID domain might have an inhibitory effect on D6PK kinase activity.

In a search for possible motifs in the MID domain that could confer binding to the plasma membrane, our attention was drawn to a positively charged motif including multiple K and R residues in each of the four D6PK and D6PKL proteins (Fig. 4A) (Hammond and Balla, 2015; Zegzouti et al., 2006). We generated alanine (A) replacement variants of the three most highly conserved Ks (D6PK-3KA) or all six Ks (D6PK-6KA) in D6PK and found that phospholipid binding was partially (D6PK-3KA) or strongly (D6PK-6KA) impaired in phospholipid binding assays or in liposome binding assays with PtdIns4P or PtdIns(4,5)P₂ (Fig. 4B,C). In turn, these mutations did not affect D6PK auto- or PIN1 trans-phosphorylation (Fig. 4D).

In transgenic plants, the YFP:D6PK-3KA and YFP:D6PK-6KA mutations strongly impaired the plasma membrane association of the protein when expressed from a 35S CaMV promoter in a wild-type Col-0 background (Fig. 4E) or from a D6PK promoter fragment in the d6pk d6pkl1 (d6pk01) double mutant (Fig. 4F). The observed intracellular distribution patterns ranged from a partial accumulation in the case of YFP:D6PK-3KA to an almost complete intracellular accumulation of YFP:D6PK-6KA (Fig. 4E-G). In quantitative terms, the effects of the K-to-A mutations on YFP:D6PK plasma membrane association differed depending on the promoter used, where the stronger 35S CaMV promoter led to stronger dissociation indices than the weaker D6PK promoter (Fig. 4G). This variation could arise from limiting plasma membrane recruiting factor(s), such as phospholipids or protein, that might become limiting when YFP:D6PK is abundant and presents lower affinity to membranes. In contrast, the individual KA mutants revealed surprisingly conserved dissociation indices between independent transgenic lines with different protein levels, regardless of the promoter driving their expression. For instance, the four D6PK promoter lines of YFP:D6PK-6KA with variable absolute signal intensities at the plasma membrane displayed strikingly conserved plasma membrane dissociation indices (Fig. 4F,G).

To test whether absolute protein abundance contributed to the variable plasma membrane abundance of the D6PK promoter YFP:D6PK-6KA lines, we analysed total YFP:D6PK abundance from these lines after immunoprecipitations of YFP:D6PK with anti-GFP. We indeed detected a strong accumulation of the YFP:D6PK-6KA protein in lines #13 and #10 with stronger plasma membrane signals, in comparison with the weaker lines #9 and #11 and to YFP:D6PK and YFP:D6PK-3KA lines (Fig. S3A). These differences could not be explained by differences in transcript abundance between these lines (Fig. S3B). We concluded that K residue mutations combined with the resulting localization defects might lead to a disturbance of YFP:D6PK trafficking, protein turnover and ultimately homeostasis.

Owing to the low protein expression levels, we were unable to perform fractionation experiments with lines expressing YFP:D6PK from the D6PK promoter. However, when we performed subcellular fractionation experiments with the 35S CaMV promoter-driven YFP:D6PK lines, we detected a shift in the distribution of the protein between the soluble and the membrane fraction in the case of YFP:D6PK-6KA (Fig. 4H,I). Whereas ∼80% of YFP:D6PK was retained in the membrane fraction, only ∼40% of YFP:D6PK-6KA of the protein resided in the membrane fraction (Fig. 4H,I). In summary, these findings suggest that the K-rich motif of the D6PK MID domain confers the ability to bind to polyacidic phopholipids in vitro and to the plasma membrane and endomembranes in vivo.

YFP:D6PK abundance at the plasma membrane is crucial for PIN phosphorylation and auxin transport activation (Barbosa et al., 2014). d6pk d6pkl1 (d6pk01) or d6pk d6pkl1 d6pkl2 (d6pk012) mutants are impaired in auxin transport-dependent growth responses, including negative gravitropism and phototropism of the hypocotyl. Expression of YFP:D6PK from a D6PK promoter fragment is sufficient to suppress these phenotypes in d6pk012 (Willige et al., 2013). Likewise, YFP:D6PK-3KA, when expressed from the D6PK promoter, suppressed the tropism defects of the d6pk01 mutant efficiently (Fig. 5A,B). However, YFP:D6PK-6KA suppressed d6pk01 defects less efficiently and the suppression efficiency varied between the individual YFP:D6PK-6KA lines in correlation with their differential tendency to associate with the plasma membrane (Fig. 4, Fig. 5A,B). We therefore suggest that their differential ability to associate with the plasma membrane is the cause for their differential effects in the suppression of the tropism defects of d6pk01 mutants.

A further indication for an impaired functionality of the proteins came from our analyses of seedlings overexpressing wild-type and mutated YFP:D6PK. Dark-grown seedlings overexpressing YFP:D6PK from the 35S CaMV promoter have strong tropism defects that can be suppressed by treatments with the auxin transport inhibitor N-1-naphthylphthalamic acid (NPA), suggesting that these phenotypes are caused by auxin transport defects (Barbosa et al., 2014). Similarly, the overexpression of YFP:D6PK-3KA, but not that of YFP:D6PK-6KA, caused defects in hypocotyl tropism and apical hook formation that were comparable with, but milder than, those observed after overexpression of wild-type YFP:D6PK (Fig. 5C,D). As in root epidermal cells, we observed a reduction in the plasma membrane signal in YFP:D6PK-3KA and YFP:D6PK-6KA lines in epidermal, endodermal and vascular cells in the hypocotyl when compared with the lines expressing wild-type YFP:D6PK (Fig. 4D, Fig. 5E). Thus, the severity of the overexpression phenotypes in the hypocotyl decreased after mutation of the K-rich motif in line with their differential defects in plasma membrane association and their defects in auxin transport regulation at the plasma membrane.

PIN3 has a role in the control of hypocotyl tropisms and is a direct target of D6PK phosphorylation (Willige et al., 2013; Zourelidou et al., 2014). PIN3 phosphorylation is reduced in d6pk01 and d6pk012 (Willige et al., 2013) and BFA treatments, which lead to a dissociation of D6PK from the plasma membrane and impair PIN3 phosphorylation in the wild type (Barbosa et al., 2014). We accordingly expected that the differential accumulation of YFP:D6PK and its mutant variants at the plasma membrane should result in a differential PIN3 phosphorylation. In immunoblots with anti-PIN3, PIN3 can be separated into a slow- and a fast-migrating form, which are better distinguished in intensity profiles (Fig. 6A,B). It was shown that the more slowly migrating form corresponds to a phosphorylated form of PIN3 according to several criteria (Willige et al., 2013). Densitometric analysis revealed that phosphorylated PIN3 was reduced by about 40% in d6pk01 when compared with the wild type or the d6pk012 mutant complemented with the wild-type YFP:D6PK transgene. Whereas the complementation of the phosphorylation defect was still efficient with YFP:D6PK-3KA, the degree of complementation was variable between the different YFP:D6PK-6KA lines. Here, the PIN3 phosphorylation defects were partially suppressed in the YFP:D6PK-6KA lines (#13 and #10, Fig. 6), in line with suppression of d6pk01 tropism defects by YFP:D6PK-6KA in these lines and the fact that they accumulated YFP:D6PK-6KA at the plasma membrane. In turn, PIN3 phosphorylation was not normalized in the YFP:D6PK-6KA lines that could not complement the d6pk01 phenotype and that had reduced YFP:D6PK-6KA plasma membrane abundance (#9 and #11, Fig. 6 and Fig. 5). Thus, the differential complementation efficiency of the different YFP:D6PK-6KA lines on PIN3 phosphorylation correlated with their efficiency to complement d6pk01 tropism defects and also with their abundance at the plasma membrane (Fig. 4F,G, Fig. 5).

D6PK and the three D6PKLs belong to the AGC1 subgroup of the 23-member AGCVIII kinase family of Arabidopsis (Rademacher and Offringa, 2012). Besides a number of previously uncharacterized kinases, this family also includes several plasma membrane-resident kinases: the proposed PIN polarity regulators PID, WAG1 and WAG2 (AGC3 subgroup); the blue light receptors PHOTOTROPIN1 (phot1) and phot2 (AGC4 subgroup); and UNICORN (UCN; AGC2 subgroup) (Dhonukshe et al., 2010; Enugutti et al., 2012; Kong et al., 2006; Sakamoto and Briggs, 2002) (Fig. 7). Although all kinases have an insertion corresponding to the D6PK MID domain, the amino acid sequence conservation in this insertion domain and over the K-rich motif of D6PK is rather limited (Fig. 7A). Recently, studies on a biochemically diverse group of proteins from different organisms demonstrated that protein domains required for membrane association could be reliably predicted based on their basic and hydrophobic score (BH score) above a critical threshold value of 0.6 (Bailey and Prehoda, 2015; Brzeska et al., 2010). Accordingly, the biophysical properties, determined by, for example, the BH score rather than amino acid conservation, could control protein-membrane associations. Indeed, when we applied the BH score prediction to D6PK, we identified the K-rich motif with a BH score (0.79) above the critical threshold value (Fig. 7; Fig. S4). Importantly, the 3KA and 6KA mutations reduced the BH score (<0.6), thus supporting the predictive power of this value (Fig. 7; Fig. S4).

We subsequently surveyed the 23 members of the Arabidopsis AGCVIII kinase family for their BH profile and found a peak with BH scores ≥0.6 in 18 Arabidopsis proteins at the end of the MID domain that coincided, in its relative position, with the D6PK K-rich motif (Fig. 7; Fig. S4). We next tested whether the K/R-enriched motifs were indeed required for plasma membrane association and phospholipid binding in hitherto uncharacterized protein kinases. To this end, we examined the localization of representative members of individual AGC1 family subclades, namely AGC1.3 (BH=0.78), AGC1.7 (0.68), AGC1.8 (0.94) and AGC1.12 (0.8), in their wild-type form and after mutation of the K/R-enriched motifs in protoplasts (Fig. 7B). Conversely, we also tested the binding of wild-type and mutant proteins in lipid overlay assays. In each but one case, the protein kinases associated with the plasma membrane in a polybasic motif-dependent manner (Fig. 7C). A noteworthy exception was AGC1.7, which had the lowest BH score of all proteins tested (0.68) and did not associate with the plasma membrane. All four kinases bound to polyacidic phospholipids with comparable binding specificities in their wild-type form but with visibly reduced affinity in their mutant form, for most of the phospholipids, with the exception of AGC1.8 (Fig. 7C). Thus, the K/R-rich motifs with high BH scores are, in most cases, required for plasma membrane binding and good candidate motifs for phospholipid binding.

We have shown that proper phospholipid composition is required for the polar recruitment of D6PK to the plasma membrane and that a K-rich motif in the D6PK MID domain mediates protein-phospholipid interactions. We propose that phospholipid metabolism determines the plasma membrane composition and polar distribution of negatively charged phospholipids that recruit D6PK via its positively charged K-rich motif (Fig. 8). In line with this, we find that mutations of the K-rich motif impair D6PK membrane association and auxin transport-dependent tropic responses that correlate with defects in the phosphorylation of the auxin efflux carrier PIN3.

In in vitro binding assays, D6PK could bind to polyacidic PIs and PtdOH in a K-rich motif-dependent manner. In vivo manipulation of PI or PtdOH metabolism led to the intracellular accumulation of YFP:D6PK and impaired its polar plasma membrane distribution but only manipulation of PtdIns4P synthesis affected YFP:D6PK solubility. At the same time, K-rich motif mutations in D6PK led to increased solubilization of the protein. This suggests that the intact K-rich domain is required for interactions with polyacidic phospholipids in different membranes and that the specific phospholipid composition in a given membrane might determine D6PK recruitment to different membranes. The D6PK interaction with multiple polyacidic phospholipids through ionic interactions resembles previously reported mechanisms for protein-phospholipid ionic interactions and is distinct from interactions mediated between phospholipids and globular protein domains (Hammond and Balla, 2015; Li et al., 2014).

PtdOH biosensors have indicated that PtdOH is enriched at the apical and sub-apical plasma membranes in pollen tubes (Potocký et al., 2014; Simon et al., 2014). Interestingly, the same manipulations that affected the behaviour of a PtdOH biosensor in pollen tubes also affected the intracellular distribution of D6PK in root cells, i.e. the inhibition of PLD by 1-butanol resulted in the internalization of both proteins and inhibition of DGK by R59022 resulted in the accumulation of both proteins in endosomal compartments. PtdOH is thus a good candidate phospholipid for anchoring D6PK at the plasma membrane.

PI-kinases and phosphatases determine the abundance of different PI species in different membranes; PI biosensors indicate that PtdIns4P and PtdIns(4,5)P₂ reside in the plasma membrane whereas PtdIns3P resides in the late endosomes and the tonoplast. In turn, genetic evidence indicates that PtdIns(3,5)P₂ might be produced at the tonoplast and late endosomes (Heilmann and Heilmann, 2015; Hirano et al., 2015; Novakova et al., 2014; Simon et al., 2014, 2016; Vermeer et al., 2006, 2009). The cellular distribution of PtdIns5P in plants is unknown and neither PtdIns(3,4)P₂ nor PtdIns(3,4,5)P₃ have as yet been detected in plants (Heilmann, 2009; Munnik and Nielsen, 2011). In our experiments, treatments with WM and LY294002, which are expected to result in the depletion of PtdIns3P (WM and LY294002) and PtdIns4P (only WM), caused D6PK internalization but did not affect D6PK membrane association. As PtdIns3P has not yet been detected at the plasma membrane, the effects at least of WM but not of LY294002 on D6PK plasma membrane localization could be caused by the plasma membrane depletion of PtdIns4P or other indirect effects on membrane or protein trafficking. A recent study supports this hypothesis by demonstrating that PtdIns4P is the predominant phospholipid in plant plasma membranes and that the AGCVIII kinase PID localizes to the plasma membrane in a PtdIns4P-dependent manner (Simon et al., 2016).

The U73122-mediated inhibition of PtdIns(4,5)P₂ hydrolysis led to a quick depolarization of D6PK. This resembled the U73122-induced depolarization of a tip-localized PtdIns(4,5)P₂ biosensor in pollen tubes, which was observed after inhibition of PI-PLCs with subapical lateral localization (Helling et al., 2006; Zhao et al., 2010). The observed depolarization of YFP:D6PK suggests that PtdIns(4,5)P₂ is also polarized in root epidermal cells. There is, however, still some controversy as to whether the PtdIns(4,5)P₂ biosensor has an apicobasal localization in these cells (Ischebeck et al., 2013; Simon et al., 2016; Tejos et al., 2014). Regardless, it is tempting to speculate that U73122 treatment blocks a lateral PI-PLC function in the root epidermis, which could then lead to a depolarization of PtdIns(4,5)P₂ and D6PK, and that the current PtdIns(4,5)P₂ biosensor fails to recapitulate the PI dynamics. Future dissection of the polarity functions of PI4P 5-kinases and PI-PLCs in the root epidermis will be required to examine this hypothesis.

As almost all of our (short-term) manipulations of PI and PtdOH biosynthesis had an impact on D6PK localization, with the strongest effects following the PAO-induced depletion of PtdIns4P, we suggest that not only one particular phospholipid but rather a combination of PtdOH and PIs might be responsible for D6PK plasma membrane recruitment. The retention of a weak but detectable plasma membrane association in the D6PK-3KA and -6KA variants or with a deletion of the MID domain suggests that there are other determinants of membrane association within D6PK. Alternatively, regions other than the D6PK MID domain might mediate additional interactions with membrane proteins or phospholipids, or yield lipid modifications required for D6PK affinity and specificity to membranes. Such a coincident detection has been often described for the recruitment of peripheral membrane proteins (Carlton and Cullen, 2005; Hammond and Balla, 2015; Stahelin et al., 2014). Importantly, we can exclude that kinase activity might control D6PK localization, as we have previously shown that inactive YFP:D6PK maintains similar polarity and recycling kinetics to the wild-type protein (Barbosa et al., 2014).

Despite limited sequence similarity in the insertion domain, many AGCVIII protein kinases have a high BH score, specifically in the region corresponding to the D6PK K-rich motif. We thus propose, and have in part tested, that these other kinases are recruited to the plasma membrane (and possibly internal membranes) through similar mechanisms to those studied in detail for D6PK. While this study was in progress, an analysis of the mechanisms targeting the AGCVIII kinase PID to the plasma membrane identified the corresponding motif in PID to be required for PID plasma membrane association (Simon et al., 2016). Although the BH score might allow prediction of membrane associations, slight differences in motif length and features of nearby residues, e.g. hydrophobicity or phosphorylation, might affect their properties with regard to phospholipid specificity, membrane localization or trafficking as reported for other proteins (Schmick et al., 2014; Yeung et al., 2008). For example, the K/R-rich motif in the middle domain of the D6PK-related ADI3 kinase from tomato determines its dual targeting between nucleus and endomembranes (Ek-Ramos et al., 2010, 2014; Heo et al., 2006; Maures et al., 2011). Similar mechanisms might control the distribution of D6PK between the plasma membrane and endomembranes, or even the nucleus, and the existence of such mechanisms might also explain the differential localization of AGC1.7 in our protoplast studies.

In summary, we propose a model for D6PK plasma membrane anchoring and polarity maintenance where both the phospholipid composition of the plasma membrane and the D6PK surface charge at the K-rich motif of the MID domain critically determine D6PK plasma membrane steady-state levels (Fig. 8). This model provides a basis for the involvement of phospholipids in the control of auxin transport activity, and possibly directionality, by D6PK plasma membrane abundance and distribution. The findings reported here for D6PK might also be relevant for the other, as yet uncharacterized, AGC family kinases or other membrane-associated proteins with a high BH score.

The following previously published Arabidopsis thaliana mutants and transgenic lines in the Columbia-0 background were used in this study: d6pk d6pkl1 double mutants (d6pk01) (Zourelidou et al., 2009), pip5k1 pip5k2 double mutants (Ischebeck et al., 2013), 35S:YFP:D6PK in the wild-type background and D6PK:YFP:D6PK in the d6pk d6pkl1 d6pkl2 background (D6PK d6pk012) (Willige et al., 2013; Zourelidou et al., 2009), PIN2:PIN2:GFP (PIN2:GFP) (Abas et al., 2006), as well as the phosphoinositide marker lines P18Y (2xFYVE^HRS) for PtdIns3P, YFP-PH_FAPP1 for PtdIns4P, and P24Y (2xPH^PLC) for PtdIns(4,5)P₂ (Simon et al., 2014; Vermeer et al., 2009).

Clones of D6PK protein domain variants were obtained by PCR-amplification of Gateway-compatible PCR fragments from pDONR201 (Life Technologies) containing the D6PK coding sequence (Zourelidou et al., 2009) with the following primers: IB2 and IB3 for YFP:D6PKΔN (amino acids 109 to 499); IB1 and IB4 for YFP:D6PK_N (amino acids 1 to 108); IB1 and IB5 for YFP:D6PKΔC (amino acids 1 to 446); IB2 and IB6 for YFP:D6PK_C (amino acids 447 to 499); IB7 and IB8 for YFP:D6PK_MID (amino acids 255 to 335). The PCR fragments were then inserted into the entry vector pDONR201. YFP:D6PKΔMID, with a deletion of amino acids 255 to 355, was obtained by SacI digestion and re-ligation of a pDONR-D6PK clone, which had been mutagenized with the primers IB9 and IB10 to introduce SacI sites flanking the MID domain. To generate the D6PK-3KA variant, lysines K315, K317 and K318 were replaced by alanines in pDONR-D6PK using site-directed mutagenesis with the primer IB70. For the D6PK-6KA mutant variant, lysines K312, K314 and K320 were mutated in D6PK-3KA with IB62. All primer sequences are listed in Table S1.

Inserts of the pDONR201 entry clones with the above-described deletions or mutations were introduced into the destination vectors pDEST15 (Life Technologies) to obtain glutathione-S-transferase (GST) N-terminal fusion constructs for recombinant expression in bacteria or into pEXTAG-YFP-GW to generate constructs for the expression of N-terminal YFP-fusions under control of the 35S CaMV promoter in plants. D6PK mutant variants were also excised as XhoI/NotI fragments from the pEXTAG-YFP-GW vectors and inserted into a pGREEN0229 vector derivative containing a D6PK promoter sequence and a 35S CaMV terminator sequence (Zourelidou et al., 2014). All pEXTAG-YFP-GW overexpression constructs were transformed by floral dip into the wild type (Col-0) and all pGREEN0229 constructs were transformed into d6pk01 for complementation analysis (Clough and Bent, 1998).

The open-reading frames of AGC1.3, AGC1.7, AGC1.8 and AGC1.12 were amplified by PCR and inserted via pDONR201 into the destination vectors pDEST15 (Life Technologies) to obtain GST N-terminal fusion constructs for recombinant expression in bacteria or into pEXTAG-YFP-GW to generate constructs for the expression of N-terminal YFP fusions under control of the 35S CaMV promoter in plants. Polybasic motifs were mutated using overlap extension PCRs with complementary mutated primers (Hansson et al., 2008). In the case of AGC1.8, this PCR was performed in two rounds due to the extended length of the polybasic motif. Protoplast transformation was performed as previously described (Katsiarimpa et al., 2011). Confocal analysis was carried out 14 h after transformation. All primer sequences are listed in Table S1.

Unless otherwise stated, seedlings for microscopic analyses were grown in continuous white light (110 μmol/m^–2 s^–1) at 21°C for 5-7 days on ½ MS medium (2.15 g/l Murashige and Skoog salts, 0.5 g/l 2-N-morpholino ethane-sulfonic acid, 8 g/l agar, pH 5.8). The following inhibitors were dissolved in DMSO and used as previously described: Wortmaninn (Applichem) (Jaillais et al., 2006); LY294002 (Sigma) (Simon et al., 2016); PAO (Sigma) (Simon et al., 2016); U73343 and U73122 (Sigma) (Andreeva et al., 2010; Zhao et al., 2010); R59022 (Sigma) (Li and Xue, 2007; Potocký et al., 2014); propanolol (Sigma) (Li and Xue, 2007; Potocký et al., 2014); 1- and 2-butanol (Sigma) (Li and Xue, 2007; Potocký et al., 2014); Brefeldin A (Fischer Scientific) (Barbosa et al., 2014). All inhibitor treatments were controlled by solvent control treatments performed in parallel. FM4-64 (2 µM) staining was performed as described in the figure legends. All confocal images were acquired with an Olympus FV1000 confocal microscope with high-sensitivity GaAsP detectors using identical acquisition settings between samples (Olympus). Image analyses and quantification were performed using Fiji and as further specified in the results sections and figure legends (Schindelin et al., 2012).

For lipid binding assays, GST fusion proteins were recombinantly expressed in Escherichia coli BL21(DE3) and purified using Glutathione Sepharose 4B (GE Healthcare). Free GST protein was expressed from pGEX-6P-1 (GE Healthcare) as a negative control. Lipid overlay assays using PIP-strips were performed following manufacturer's instructions (Echelon). In brief, membranes were blocked overnight at 4°C in a blocking buffer with 4% BSA either in TBS-T (0.1% Tween) or PBS-T (0.1% Tween). Purified proteins (0.5 μg/ml blocking buffer) were incubated with PIP-strip membranes for 1 h at room temperature, washed three times for 10 min with TBS-T or PBS-T, then incubated for 1 h with anti-GST (1:2000 in blocking buffer; GE Healthcare), washed three times and incubated for 1 h at room temperature with anti-goat peroxidase conjugate antibodies (1:8000 in blocking buffer; Sigma, A5420). After three final washes, detection of bound proteins was performed using Pierce ECL Plus Western Blotting Substrate (Thermo Scientific) in a Fujifilm LAS 4000 Mini (Fuji).

Liposome binding assays were performed as previously described (Julkowska et al., 2013). Liposomes were prepared from 400 nmol of total lipids at the following molar ratios: PtdCho:PtdEtn, 1:1; PtdCho:PtdEtn:PtdIns4P, 2:2:1; PtdCho:PtdEtn:PtdIns(4,5)P₂, 2:2:1; PtdCho:PtdEtn:PtdOH, 2:2:1. All lipids were obtained from Avanti Polar Lipids. Proteins were separated by 10% SDS-PAGE and blotted to Protran nitrocellulose membranes (GE Healthcare). GST-tagged D6PK and D6PK-6KA were detected with goat anti-GST (GE Healthcare, 27457701V) and visualized using a rabbit anti-goat antiserum conjugated to an alkaline phosphatase reporter (Sigma).

In vitro kinase assays were performed with purified recombinant kinases using GST:PIN1 CL (cytosolic loop) as a phosphorylation substrate (Zourelidou et al., 2009). Equal protein amounts (∼0.5 µg) were incubated for 60 min at 28°C in a buffer [25 mM Tris-HCl pH 7.5, 5 mM MgCl₂, 0.2 mM EDTA, 50 μM ATP, 1× Complete Protease Inhibitor Cocktail (Roche)] supplemented with 10 μCi [γ-³²P]-ATP [370 MBq, specific activity 185 TBq (Hartmann Analytic)]. Reactions were stopped by adding 5× Laemmli buffer, samples were then boiled and separated in two SDS-PAGE gels, one for autoradiography and one for staining with Coomassie Brilliant Blue as a loading control.

Seedlings were grown for 2.5 days on ½ MS in the dark. Two hours before the experiment, agravitropically growing seedlings were straightened under green safe light. For phototropism response experiments, etiolated seedlings were transferred to a FloraLED chamber (CLF Plant Climatics) and illuminated for 4 h with 5 μmol m^–2 s^–1 unilateral blue light. For the hypocotyl negative gravitropism response experiments, plates were reoriented by 90° and kept in the dark for an additional 20 h. Plates were scanned and bending angles were quantified using Fiji (Schindelin et al., 2012).

For the detection of YFP-tagged D6PK and its variants, roots of seven­-day-old seedlings were homogenized with protein extraction buffer (PEB; 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.1 mM MG132, 0.1 mM PMSF, 1% (v/v) Protease Inhibitor Cocktail (Sigma), and PhosStop phosphatase inhibitor cocktail (Roche)]. For fractionation experiments, extracts were pre-cleared by centrifugation at 10,000 g and the supernatant (T) was subsequently ultra-centrifuged at 100,000 g to yield the soluble (S) and membrane (M) fractions. Membrane fractions were resuspended in the same volume as soluble fractions with PEB. Samples were denatured with 1× Laemmli buffer and heated at 42°C for 15 min. For immunoblotting, the equivalent of 10 µg total protein was separated on a 10% SDS-PAGE and transferred to nitrocellulose membrane (GE Healthcare) by semi-dry blotting. Immunoblots were detected with rabbit anti-GFP (1:3000; Fischer Scientific, A11122) and rabbit anti-UGP (1:2000; Agrisera, AS05086) and secondary antibody goat anti-rabbit HRP (1:2000; Sigma, A9169). Chemiluminescence detection was performed with a Fujifilm LAS 4000 Mini, total band and band profile intensities were quantified using the Fujifilm Multi Gauge v3.0 software.

For immunoprecipitation, protein was extracted in PEB with 0.5% Triton X-100 from seven-day-old seedlings and subsequently diluted to reduce the Triton X-100 concentration to 0.01%. YFP:D6PK was then immunoprecipitated from 3 mg protein extract with 15 µl GFP-Trap_MA slurry (Chromotek) for 1 h at 4°C. Subsequent wash steps followed the manufacturer's protocol. Beads were resuspended in 40 µl 2× Laemmli buffer and 10 µl were used for SDS-PAGE and immunoblotting as described above. A 5 µg aliquot of total protein extract (input) was used for SDS-PAGE and immunoblotting with anti-UGP as a loading control.

For PIN3 immunoblots, protein extraction, quantification and profile analysis was performed using four-day-old dark-grown seedlings as previously described (Barbosa et al., 2014; Willige et al., 2013).

RNA extraction and qRT-PCR analysis were performed as previously described (Lutz et al., 2015). Primer sequences are provided in Table S1.

The authors would like to thank Benjamin Weller and Lanassa Bassukas (Technische Universität München) for helpful discussions throughout this study. Jutta Elgner is thanked for her excellent technical assistance. Lanassa Bassukas, Lena Frank and Andreas Scaia (Technische Universität München) are thanked for generating primary cDNA constructs for several AGC1 kinases.