Abstract
The 3-Phosphoinositide-Dependent Protein Kinase 1 (PDK1) is a conserved and important master regulator of AGC kinases in eukaryotic organisms. pdk1 loss-of-function causes a lethal phenotype in animals and yeast. In contrast, only very mild phenotypic defects have been reported for the pdk1 loss-of-function mutant of the model plant Arabidopsis thaliana (Arabidopsis). The Arabidopsis genome contains two PDK1 genes, hereafter called PDK1 and PDK2. Here we show that the previously reported Arabidopsis pdk1 T-DNA insertion alleles are not true loss-of-function mutants. By using CRISPR/Cas9 technology, we created true loss-of-function pdk1 alleles, and pdk1 pdk2 double mutants carrying these alleles showed multiple growth and development defect, including fused cotyledons, a short primary root, dwarf stature, late flowering, and reduced seed production caused by defects in male fertility. Surprisingly, pdk1 pdk2 mutants did not phenocopy pid mutants, and together with the observations that PDK1 overexpression does not phenocopy the effect of PID overexpression, and that pdk1 pdk2 loss-of-function does not change PID subcellular localization, we conclude that PDK1 is not essential for PID membrane localization or functionality in planta. Nonetheless, most pdk1 pdk2 phenotypes could be correlated with impaired auxin transport. PDK1 is highly expressed in vascular tissues and YFP:PDK1 is relatively abundant at the basal/rootward side of root stele cells, where it colocalizes with PIN auxin efflux carriers, and the AGC1 kinases PAX and D6PK/D6PKLs. Our genetic and phenotypic analysis suggests that PDK1 is likely to control auxin transport as master regulator of these AGC1 kinases in Arabidopsis.
Introduction
Protein phosphorylation by protein kinases is a ubiquitous and crucial posttranslational modification in eukaryotic cells. It is involved in almost all cell activities, such as cell division, cell growth and environmental signaling. The AGC kinase family comprises some of the best-characterized protein serine/threonine kinases in eukaryotic cells, such as the founder members cyclic AMP-dependent protein kinase A (PKA) and calcium-dependent protein kinase C (PKC) (Pearce et al., 2010). These kinases play crucial roles in basal cellular functions in lower (yeast) and higher (human/mice) eukaryotes. For example, protein kinase B (PKB/c-Akt) is important in apoptosis inhibition and insulin signaling (Lawlor and Alessi, 2001), whereas p70 ribosomal protein S6 kinase (S6K) plays an important role in mRNA translational control (Pearce et al., 2010; Bahrami-B et al., 2014). AGC kinases themselves are also phosphorylation substrates that can be activated by serine/threonine phosphorylation in the activation loop (T-loop) or in the C-terminal hydrophobic motif of the kinase domain (H-motif) (Chamoto et al., 2010). The 3-Phosphoinositide-Dependent Protein Kinase 1 (PDK1) is a well-established activator responsible for AGC kinase T-loop phosphorylation (Mora et al., 2004; Chamoto et al., 2010).
PDK1 itself is also a conserved member of AGC kinase family, and typically contains a kinase domain with a PDK1-Interacting Fragment (PIF)-binding pocket at its N-terminus and a PH domain at the C-terminus (Biondi et al., 2000; Frödin et al., 2002). Other AGC kinases have a C-terminal hydrophobic PIF motif, and the interaction with the PIF binding pocket in PDK1 enhances their activation by phosphorylation. The PH domain is essential for PDK1 plasma membrane recruitment and kinase activity in mammals. Binding of the PH domain to the phospholipid phosphatidylinositol (3,4,5)-trisphosphate [PtdIns(3,4,5)P3] at the plasma membrane triggers PDK1 dimer to monomer conversion and phospho-activation (Alessi et al., 1997; Ziemba et al., 2013). PDK1 was originally named PtdIns(3,4,5)P3-dependent protein kinase 1 (Alessi et al., 1997), but the name was changed when PtdIns(3,4)P2, PtdIns3P and PtdIns(4,5)P2 also appeared to bind its PH domain (Currie et al., 1999; Deak et al., 1999). Arabidopsis thaliana PDK1 (AtPDK1) binds to an even broader selection of phospholipids in vitro (Deak et al., 1999). Nevertheless, the two most important phospholipids for mammalian PDK1, PtdIns(3,4,5)P3 and PtdIns(3,4)P2, have not been identified in Arabidopsis thaliana (Arabidopsis) (Heilmann, 2016), and AtPDK1 activity has been reported to be controlled by PtdIns(4,5)P2 and phosphatidic adic (PA) (Anthony et al., 2004). Arabidopsis has two highly homologous PDK1 genes, At5g04510 (AtPDK1.1) and At3g10540 (AtPDK1.2), and for convenience reasons we renamed them to respectively PDK1 and PDK2.
Interestingly, the two yeast PDK1 orthologs Pkh1 and −2, which lack a PH domain, still have the ability to phosphorylate AGC kinases (Casamayor et al., 1999; Niederberger and Schweingruber, 1999; Voordeckers et al., 2011). Physcomitrella patens PDK1 (PpPDK1), which also lacks a PH domain, is able to rescue the lethal phenotype of the yeast pkh1 pkh2 double mutant. This indicates that the PH domain is not required in all eukaryotes or full PDK1 functionality (Dittrich and Devarenne, 2012a). Besides for yeast, complete loss-of-function of PDK1 is also lethal for fruit flies and mice (Lawlor et al., 2002; Rintelen et al., 2002). In plants, several methods have been employed in different species to analyze PDK1 function in planta. Virus-induced gene silencing (VIGS) has been used to knock down tomato PDK1, whereas Tos17 transposon mutagenesis or homologous recombination has been used in rice or Physcomitrella patens, respectively. However, in tomato the claimed cell death phenotype made PDK1 knock-out expression unprovable (Devarenne et al., 2006), and in rice the Tos17 insertion only led to a knock down of PDK1 expression (Matsui et al., 2010; Dittrich and Devarenne, 2012a). Deletion of PDK1 in Physcomitrella patens was not lethal, but pdk1 knock-out mutants showed strong developmental defects (Dittrich and Devarenne, 2012a). In Arabidopsis, three combinations of pdk1 pdk2 T-DNA insertion alleles have been reported to show altered sensitivity to Piriformospora indica induced growth promotion, and a weak developmental defect resulting in reduced silique length (Camehl et al., 2011; Scholz et al., 2019). Inhibition of PDK1 expression in Arabidopsis cell suspensions using RNAi technology delivered no mutant cell phenotype (Anthony et al., 2004).
In contrast to the lack of a clear in planta role for PDK1, all Arabidopsis AGC kinases phosphorylated by PDK1 in vitro, including PINOID (PID), Oxidative Signal-Inducible1(OXI1), UNICORN (UCN) and most AGC1 family members, do play key roles in plant development and defense (Anthony et al., 2004, 2006; Zegzouti et al., 2006a, 2006b; Devarenne et al., 2006; Camehl et al., 2011; Enugutti et al., 2012; Gray et al., 2013; Scholz et al., 2019). PID phosphorylates PIN auxin efflux carries to control their polarity and thereby direct the auxin flux (Christensen et al., 2000; Benjamins et al., 2001; Friml et al., 2004; Kleine-Vehn et al., 2009; Dhonukshe et al., 2010; Huang et al., 2010). OXI1 plays a dual role in regulating both root hair growth and the basal immune response against virulent pathogen infection (Anthony et al., 2004; Rentel et al., 2004; Anthony et al., 2006; Petersen et al., 2009; Matsui et al., 2010; Camehl et al., 2011). UCN was recently shown to be a phosphorylation target of PDK1 in vitro, but genetic evidence suggests that UCN negatively regulates PDK1 at the post-transcriptional level to control planar growth (Scholz et al., 2019). The other established PDK1 targets all belong to the AGC1 protein kinases family (Galván-Ampudia and Offringa, 2007; Rademacher and Offringa, 2012), which has been well-characterized during the past decade. The D6 protein kinases (D6PKs, including D6PK/AGC1.1, D6PKL1/AGC1.2, D6PKL2/PK5, D6PKL3/PK7), PROTEIN KINASE ASSOCIATED WITH BRX (PAX/AGC1.3), PAX LIKE (PAXL/AGC1.4) and AGC1-12 all have been shown to phosphorylate PIN proteins to enhance auxin transport activity (Zourelidou et al., 2009; Willige et al., 2013; Barbosa et al., 2014, 2018; Haga et al., 2018; Marhava et al., 2018). The tomato ortholog of PAX, also known as AvrPto-dependent Pto-interacting protein 3 (Adi3), negatively controls plant cell death caused by pathogen attack (Devarenne et al., 2006; Gray et al., 2013). AGC1.5 and AGC1.7 control polar growth of pollen tubes by phosphorylating RopGEFs. (Zhang et al., 2009; Li et al., 2018), and ROOT HAIR SPECIFIC 3 (RSH3/AGC1.6) specifically regulates root hair morphology (Won et al., 2009). The disproportion between the in vivo data on the functions of the different Arabidopsis AGC kinases that are established in vitro phosphorylation targets of AtPDK1, and the small role that AtPDK1 itself seems to play in development based on the pdk1 pdk2 double mutant phenotype, made us reinvestigate the published data on AtPDK1.
PID has been reported as one of the prime targets of PDK1 (Zegzouti et al., 2006a), but the pdk1 pdk2 double mutant lacks the typical pid loss-of-function phenotypes. We therefore generated Arabidopsis lines overexpressing PDK1 (PDK1ox), and found that seedlings of these lines lacked the strong phenotypes observed in seedlings overexpressing PID (PIDox). These results suggest that either PDK1 requires activation and that this is not triggered in the PDK1ox seedlings, or that it is not rate limiting for PID activity. Next, we re-analyzed the published pdk1 and pdk2 T-DNA insertion alleles. Based on RT-PCR experiments, the two pdk2 alleles appeared to represent true loss-of-function mutants. However, functional PDK1 mRNA was still detectable in the three pdk1 alleles, explaining the lack of strong phenotypes in the pdk1 pdk2 double mutant combinations. Using CRISPR/Cas9, we generated several true pdk1 loss-of-function mutant alleles, which when combined with the pdk2 T-DNA insertion allele did display strong growth and developmental defects. The mutant phenotypes indicate a pleiotropic, but PID-independent role for PDK1 in plant development as regulator of auxin transport.
Results
PDK1ox and PIDox seedlings do not share phenotypes
The key defects caused by PIDox in Arabidopsis are agravitropic seedling growth and collapse of the main root meristems as a result of redirected polarity of PIN-mediated auxin transport (Figure 1G,H, J) (Benjamins et al., 2001; Friml et al., 2004). In view of the model that PDK1 regulates PID kinase activity (Zegzouti et al., 2006a), we expected PDK1ox to cause similar phenotypes as PIDox. More than thirty independent p35S::YFP:PDK1 or p35S::PDK1 transgenic lines were selected and T2 seedlings grown on vertical agar plates showed normal gravitropic growth. Five single locus homozygous T3 lines with different PDK1 overexpression levels were subsequently selected for further phenotype observation and quantification (Figure 1I). All of the representative PDK1ox lines showed normal gravitropic seedling growth and no collapse of the main root meristem was observed (Figure 1). Roots of p35S::YFP:PDK1#5.4 and p35S::YFP:PDK1#9.6 seedlings were even slightly longer than wild-type roots (Figure 1H), however, this phenotype did not clearly correlate with the PDK1 overexpression level (Figure 1I). Also mature PDK1ox plants developed and flowered like wild-type Arabidopsis plants.
The above results suggest that PDK1 is not rate limiting for endogenous PID activity. However, we cannot exclude that the PDK1 kinase itself requires signaling to be activated, and that therefore its overexpression does not lead to additional phenotypes under normal growth conditions.
CRISPR/Cas9-generated mutant alleles indicate a central role for PDK1 in development
To obtain further indications for the proposed role of PDK1 as upstream regulator of PID, we re-assessed the previously described pdk loss-of-function mutant alleles. Three pdk1 and two pdk2 T-DNA insertion alleles have been reported to be loss-of-function mutants (Camehl et al., 2011; Scholz et al., 2019). Neither pdk1 nor pdk2 single mutants showed any noticeable phenotype, and different double mutant combinations of the pdk1 and pdk2 alleles only showed a mild reduction in silique length and plant height (Camehl et al., 2011; Scholz et al., 2019).
Two pdk2 alleles, pdk2-1 and pdk2-4, were confirmed to be true knock-out mutants by RT-PCR analysis (Figure 2A, B). However, in contrast to published data, the pdk1-c allele appeared to produce a full length mRNA (Figure 2A, B) (Camehl et al., 2011; Scholz et al., 2019), whereas the pdk1-a and pdk1-b alleles produced a partial or mutated mRNA (Figure 2A, B, Figure S1), leading to the production of a PDK1 kinase lacking its PH domain (Figure S1). Previous studies have suggested that a PH domain may not be essential for PDK1 function in plants (Dittrich and Devarenne, 2012b). When we tested the kinase activity of PDK1 lacking a PH domain in vitro, it showed very high autophosphorylation activity (Figure 2C). Based on these findings, we concluded that the three published pdk1 T-DNA insertion alleles are not likely to be true loss-of-function mutants.
In order to obtain true pdk1 alleles for studying PDK1 biological function, we designed guide RNAs against the 3rd and 7th exon, and were able to obtain five CRISPR/Cas9-generated mutants with frame shifts in the PDK1 open-reading frame (Figure 2D, E). Like the pdk1 T-DNA insertion alleles, these new pdk1 mutant alleles did not result in significant morphological differences from wild type. However, when combined with the pdk2-1 or pdk2-4 alleles, all double mutant combinations showed the same striking dwarf phenotype (Figure 2F-J). Complementation analysis using either p35S::PDK1, p35S::YFP:PDK1 or pPDK1::YFP:PDK1 showed that the dwarf phenotype was caused by pdk loss-of-function (Figure 2J, Figure S2). For all three constructs, several lines were obtained that showed complete rescue of the pdk1-13 pdk2-4 double mutant phenotype (Figure S2, Figure S3A). The results show that PDK1 and PDK2 act redundantly and have a much more important role in plant growth and development than was previously reported.
pdk loss-of-function leads to many developmental defects, but not to a pid phenocopy
Besides the decreased rosette diameter and reduced final plant height (Figure 2G, H), pdk1 pdk2 double mutant plants flowered much later and showed strong reproductive defects (Figure 2I, J; Figure 3). The number of double homozygous F2 progeny obtained was much lower (1 in 47.7 ± 2.6) than the Mendelian ratio (1 in 16). Also F2 plants with the pdk1(-/-) pdk2(+/-) or pdk1(+/-) pdk2(-/-) genotype produced homozygous progeny at a much lower frequency than the expected 1 in 4 ratio (Table S1). Seed production of the homozygous pdk1 pdk2 mutants (1.5 ± 0.21 per silique for pdk1-13 pdk2-4) was significantly reduced compared to wild type (65.9 ± 0.61 per silique). Mutant plants developed very short siliques (Figure 3A), a phenotype that has previously been reported for Arabidopsis plants that are both male and female sterile (Huang et al., 2016). These results implied that pdk1 pdk2 loss-of-function causes gametophyte and/or embryo development defects in Arabidopsis. Reciprocal crosses between wild-type and pdk1-13 pdk2-4 double mutant plants revealed both male-related and female-related reduced fertility. However, since the cross Col-0♀ x pdk1-13 pdk2-4♂ produced fewer seeds than the reciprocal cross pdk1-13 pdk2-4♀ x Col-0♂ (Figure S4A), it is likely that male gametophyte development is more strongly impaired by pdk loss-of-function than female gametophyte development. Alexander staining showed that pollen grain development in the pdk1 pdk2 double mutant was not aborted, but that anther dehiscence was the major cause of the male fertility problems (Figure 3B-E, I). In addition, in vitro germination of pdk1 pdk2 pollen resulted in strangely shaped pollen tubes as a result of aberrant tip growth (Figure 3F, G). After 18-hour incubation on pollen germination medium, pdk1-13 pdk2-4 pollen tube growth arrested with a bulb-like structure, and as a result they remained much shorter than wild type pollen tubes (Figure 3G, H). The ovules of double mutant plants did not show noticeable morphological alterations (Figure S4B,C), which is in line with the predominant effect of pdk loss-of-function on male fertility.
In contrast to the fertility problems, pdk1 pdk2 double mutants developed relatively normal flowers that showed no clear patterning defects. Flowers did show early stigma exposure due to impaired sepal growth, and slightly reduced filament elongation (Figure 3I). The short inflorescences seemed not the result of reduced internode elongation, but were most likely caused by early inflorescence meristem arrest (Hensel et al., 1994) (Figure 2J). The lack of phenotypic resemblance between pdk1-13 pdk2-4 and pid-14 inflorescences and flowers (Figure 2J) suggests that PDK1 is not essential for full PID function during inflorescence development. Moreover, expression of a PID:YFP fusion in pdk1-13 pdk2-4 protoplasts showed that PDK1 activity is not necessary for the predominant localization of PID at the plasma membrane (Figure 2K). Based on these results and the overexpression data we conclude that, in contrast to what has previously been suggested (Zegzouti et al., 2006a, 2006b), PDK1 is not a key regulator of PID activity.
Alternative splicing produces a functional cytosolic PDK1 isoform lacking a PH domain
The PDK2 gene produces a single transcript, whereas transcription of PDK1 results in at least six different mature transcripts, due to alternative splicing events at the 5th, 7th and 9th intron (https://www.araport.org/). These transcripts can be translated into five different protein isoforms, which we named respectively PDK1S0, PDK1S1, PDK1S2 and PDK1S3 (Figure 4A). We checked the abundance of each mature transcript using semi-quantitative RT-PCR followed by restriction digestion. The full-length PDK1 transcript was most abundant, and the short PDK1 transcripts producing isoforms lacking part of the kinase domain (PDK1S1, PDK1S2, and PDK1S3) were also present at high levels, while the transcript producing the PDK1S0 isoform with a complete kinase domain, was the least abundant (Figure 4B).
In order to test the functionality of the different isoforms, we expressed the corresponding cDNAs in yeast (S. cerevisiae) strain INA106-3B, in which the PKH2 gene copy has been replaced by LEU2, and the PKH1 gene copy has been mutated so that strain INA106-3B is able to grow normally at 25°C but not at 35°C. As expected based on previous experiments, expression of the full length PDK1 or PDK2 cDNAs allowed this strain to grow at 35°C (Dittrich and Devarenne, 2012a) (Figure 4C). In contrast, expression of the cDNAs producing the PDK1S1, PDK1S2 or PDK1S3 isoforms did not allow growth at the restrictive temperature, suggesting that any deletion of the conserved kinase domain renders PDK1 non-functional (Figure 4C). This is in line with loss-of-function observed for the new Arabidopsis alleles pdk1-11, −13, −14, −31 and -32, which all express partial PDK1 proteins having a small or bigger deletion of the C-terminal part of the kinase domain (Figure 2D, E, F). Interestingly, expression of the PDK1S0 did permit INA106-3B to grow at 35°C (Figure 4C). The yeast data were confirmed by 35S promoter-driven expression in the Arabidopsis pdk1 pdk2 loss-of-function mutant background. p35S::PDK1 provided full rescue of the vegetative growth phenotypes of the Arabidopsis pdk1 pdk2 mutant, and some p35S::PDK1S0 lines showed the same level of rescue (Figure S3A). In contrast, expression of PDK1S1 and PDK1S2 did not result in any rescue (Figure S3A). Expression of a YFP:PDK1S0 fusion under control of the PDK1 promoter in the pdk1-14 pdk2-4 mutant background also completely rescued the mutant vegetative growth phenotypes (Figure 4D). However, pPDK1::YFP:PDK1S0 pdk1-13 pdk2-4 plants developed shorter siliques carrying fewer seeds compared to wild-type or pPDK1::YFP:PDK1 pdk1-14 pdk2-4 plants (Figure 4E). Interestingly, a similar silique phenotype has also been described for the Arabidopsis pdk1-b pdk2-1 double mutant, and according to our own analysis the T-DNA insertion in the pdk1-b allele leads to the production of a shorter PDK1 protein with an intact kinase domain but lacking the PH domain (Camehl et al., 2011) (Figure 2A, B).
These results corroborate the conclusions from the complementation experiments in yeast that a full-length kinase domain is essential for PDK1 function, but that surprisingly the PH domain is not essential for PDK1 function during Arabidopsis vegetative growth. Since the PH domain is responsible for lipid binding, we checked the PDK1 promoter driven YFP:PDK1 and YFP:PDK1S0 localization in root columella cells, where PDK1 is highly expressed. YFP:PDK1 localized both on the plasma membrane and in the cytoplasm, whereas YFP:PDK1S0 was only found in the cytoplasm (Figure S3B, C). The functional relevance of the latter low abundant cytosolic isoform remains unclear. Our findings do suggest, however, that PH domain-dependent plasma membrane association of PDK1 is only essential during specific developmental processes.
PDK1 and PDK2 are broadly expressed during development
Since the pdk1 pdk2 mutant shows many defects in development and growth, we analysed the spatio-temporal expression pattern of the two PDK genes to uncover their tissue-specific functions. For this purpose we generated Arabidopsis (Col-0) lines carrying the pPDK1::turboGFP:GUS (pPDK1-GG) or pPDK2::turboGFP:GUS (pPDK2-GG) construct and used a pdk1-14 pdk2-4 mutant line carrying the complementing pPDK1::YFP:PDK1 construct. PDK1 appeared to be strongly expressed in (pro)vascular tissues from the early globular embryo stage on, and in the columella root cap (Figure 5, Figure S5). The gene also showed more general expression in young hypocotyls, cotyledons, leaves and floral organs, and in growing siliques. The expression pattern of PDK2 was very comparable to that of PDK1 (Figure 5 D, E, H, K, L), except that no expression was observed in the root apex (Figure 5D) or in embryos (data not shown).
As previously observed for PDK2:EGFP (Scholz et al., 2019), YFP:PDK1 did not localize in the nucleus, but was mainly found in the cytoplasm or associated with the plasma membrane (Figure 5 C, F; Figure S3B;). In (pro)vascular cells in heart stage embryos and roots, YFP:PDK1 showed predominant basal (rootward) localization (Figure 5 C, F; Figure S5), just like PIN1 and the PDK1 targets PAX and D6PK (Gälweiler et al., 1998; Zourelidou et al., 2009; Marhava et al., 2018).
Auxin transport is impaired in pdk1-13 pdk2-4 mutant
Even though the PDK1 overexpression and loss-of-function phenotypes did not point to an important role for PDK1 in PID function, the pdk1-13 pdk2-4 mutant seedling phenotypes did suggest involvement of PDK1 in the regulation of auxin response or -transport (Figure 6A-E). Mutant primary roots elongated normally up to two days after germination, but after that their growth rate declined (Figure 6 A,B), and roots started to oscillate randomly with a large amplitude, resulting in curved short roots (Figure S6). Of 199 7-day-old seedlings, 18.1% of the primary roots grew into the air. On a total of 460 pdk1-13 pdk2-4 mutant seedlings, 58% showed fused or single dark green cotyledons, and the remaining 42% developed two cotyledons with short petioles positioned at an abnormal angle (< 180°) (Figure 6C-E). The cotyledon phenotypes and short agravitropic roots are usually observed in auxin response or -transport mutants or transport inhibitor-treated seedlings. By combining the pdk1-13 pdk2-4 double mutant with the or pDR5::GUS auxin response reporter, we observed that auxin response was absent or strongly decreased in the root stele and confined to the root tip, while an enhanced auxin response was observed at the mutant cotyledon edges and in the fragmented cotyledon veins (Figure 6F, G, I, K). This highly resembled the DR5::GUS expression of 7-day-old seedlings grown on medium supplemented with the auxin transport inhibitor naphthylphtalamic acid (NPA) (Figure 6H, J) (Sabatini et al., 1999; Bao et al., 2004). Moreover, the increase in DR5::GUS expression in cotyledons corroborated that pdk1-13 pdk2-4 mutants are defective in auxin transport, rather than in auxin biosynthesis or -signaling. Short time treatment of wild-type and pdk1-13 pdk2-4 seedlings with IAA and subsequent qPCR analysis showed that the auxin inducible expression of the IAA5, GH3.3 and SAUR16 genes was not impaired, confirming that the mutants are not defective in auxin response (Figure 6L). Instead, pdk1-13 pdk2-4 mutant seedlings were hypersensitive to NPA treatment compared to wild type (Figure 6M). Moreover, the auxin transport capability of pdk1-13 pdk2-4 inflorescence stems was significantly reduced compared to that of wild-type stems (Figure 6N). Together, the above data point toward a role for PDK1 in enhancing polar auxin transport.
PDK1 regulates PIN-mediated auxin efflux probably through the AGC1 kinases
Several in vitro PDK1 phosphorylation substrates are AGC kinases that have been reported to regulate auxin transport by direct phosphorylation of PIN auxin efflux carriers (Zegzouti et al., 2006b; Willige et al., 2013; Marhava et al., 2018; Haga et al., 2018). The fact that their loss-of-function mutants share phenotypic defects with the pdk1-13 pdk2-4 mutant, such as the short root of the pax mutant (Marhava et al., 2018), or the fused cotyledons of the d6pk012 triple mutant (Zourelidou et al., 2009), hinted that these AGC kinases might indeed be regulated by PDK1 in planta. The stronger pdk1-13 pdk2-4 mutant phenotype suggested that PDK1 has many more phosphorylation substrates.
In order to investigate whether PIN proteins themselves are PDK1 phosphorylation substrates, we deduced based on published in vitro phosphorylation data, that PDK1 prefers to phosphorylate the second serine residue in the RSXSFVG motif (X represents any amino acids) that is part of the activation segment of the AGC kinases (Zegzouti et al., 2006a, 2006b). Analysis of the large central hydrophilic loop (HL) of the 5 Arabidopsis PIN1-type PIN proteins identified several RXXS motifs. However, in vitro phosphorylation assays using GST-tagged PDK1 (GST-PDK1) and GST-tagged versions of the HL of PIN1, PIN2, PIN3 or PIN7 (GST-PIN1/2/3/7HL) only showed phosphorylation of the PIN2HL (Figure 7A). Interestingly, PIN2HL S1,2,3A, in which the PID phosphorylation sites are substituted by alanines, was also phosphorylated by PDK1 at same level as the wild-type PIN2HL (Figure 7A). PDK1 must therefore phosphorylate one or more other serine residues that are unique to the PIN2HL. However, PIN2 is not co-expressed with PDK1, and the PIN proteins that are co-expressed with PDK1 in the root stele or columella cells (PIN1, PIN3 and PIN7) are not phosphorylated by PDK1 in vitro. Moreover, no noticeable alteration in PIN1/3/7 protein polarity was observed in pdk1-13 pdk2-4 mutant roots (Figure 7C-J). The PIN2:GFP abundance was slightly decreased in pdk1-13 pdk2-4 mutant root tips (Figure 7I, J, L), but this might be an indirect effect of pdk loss-of-function on auxin distribution in the root tip, as we measured a slight increase in GFP intensity in DR5::GFP pdk1-13 pdk2-4 mutant versus wild-type root tips (Figure 7B, K). Our results suggest that PDK1 regulates auxin transport, most likely by activating one or more AGC kinases, such as PAX and D6PK, which subsequently regulate auxin efflux activity by direct phosphorylation of the PINs.
Discussion
PDK1 is a well-established key regulator of AGC kinases in animals and yeast, and its importance in these organisms is demonstrated by the lethality caused by loss-of-function mutations in the genes encoding for this protein kinase (Casamayor et al., 1999; Rintelen et al., 2002; Lawlor et al., 2002; Mora et al., 2004). Also in the model plant Arabidopsis, PDK1 has been shown to phosphorylate several AGC kinases in vitro (Zegzouti et al., 2006a, 2006b), However, the previously reported impact of loss-of-function of the two gene copies PDK1 and PDK2 on Arabidopsis development was only limited (Camehl et al., 2011; Scholz et al., 2019). In this study, we found that the published T-DNA insertion alleles of the Arabidopsis PDK1 gene copy are not loss-of-function mutants. Here we generated several CRISPR/Cas9-based true loss-of-function pdk1 alleles that, when combined with the available pdk2 loss-of-function mutant alleles, did lead to strong developmental defects. Different from animals and yeast though, and more similar to the situation in Physcomitrella Patens, Arabidopsis pdk1 pdk2 loss-of-function mutants are viable, indicating that the substrate preference of plant PDK1 has changed from that in other eukaryotes, and that it has lost its involvement in signaling pathways that are essential for cell survival.
PDK1 is not essential for PID activity controlling inflorescence and cotyledon development
By carefully recording the pdk1 pdk2 mutant phenotypes, we analysed the genetic relation between PDK1 and its reported in vitro substrates, of which PID was the key candidate (Zegzouti et al., 2006a). Loss-of-function of both PDK genes leads to fused cotyledons, short wavy roots, dwarf stature and reduced fertility resulting in short siliques. To our surprise, pdk1 pdk2 does not share the three cotyledon, pin inflorescence and aberrant flower phenotypes that are typical for pid loss-of-function mutants, implying that PDK1 is not essential for PID function in these tissues. PID is an auto-activating kinase in vitro and might act independent of upstream activating kinases (Christensen et al., 2000; Benjamins et al., 2001), or other kinases than PDK1 might be involved in hyper-activating PID during embryo, inflorescence and flower development. The latter seems most likely based on the observation that flower, leaf and shoot extracts can hyperactivate PID in vitro (Zegzouti et al., 2006a). A physical interaction between PID and PDK1 through the PIF domain, as suggested by Zegzouti et al. (Zegzouti et al., 2006a), has never been proven, and was purely based on in vitro phosphorylation data. Here we show unequivocally that PID does not require PDK1 for its association with the PM, which corroborates the finding that this is mediated by an arginine-rich loop in the kinase domain of PID (Simon et al., 2016). All data are in line with the observation that a PID:GUS fusion lacking the PIF domain can still complement pid loss-of-function mutants (Benjamins et al., 2001). Although we cannot fully exclude that PDK1 and PID do have a functional interaction, our results at least indicate that this interaction is not essential for the majority of the PID activities in plant development.
Is PDK1 a master regulator of AGC kinases in Arabidopsis?
If not PID, which and how many other Arabidopsis AGC kinases are potential phosphorylation substrates of PDK1? The developmental defects of the pdk1 pdk2 double mutant together with the altered pDR5 expression pattern and reduction in auxin transport all point toward a role for PDK1 in promoting polar auxin transport. Interestingly, several pdk1 pdk2 mutant phenotypes are also observed for loss-of-function mutants of members of the AGC1 kinase sub-family, for some of which a role as regulator of polar auxin transport is now well established (Zourelidou et al., 2009; Willige et al., 2013; Barbosa et al., 2014; Haga et al., 2018; Marhava et al., 2018). For example, the fused cotyledons, deficient lateral root emergence and agravitropic primary root growth closely resemble phenotypes observed for the d6pk012 triple mutant (Zourelidou et al., 2009). And a short primary root is also observed for the pax mutant (Marhava et al., 2018), and like pdk1 pdk2, the agc1.5 agc1.7 double mutant is defective in pollen tube growth (Zhang et al., 2009). Since the corresponding AGC1 kinases are strongly dependent on PDK1 for their in vitro activation (Zegzouti et al., 2006b), it seems possible that PDK1 might act as a master regulator of these AGC1 kinases. Further experimentation is required, however, for each of these kinases to prove this hypothesis.
Recently, a genetic interaction with PDK1 was reported for a kinase of the AGC2 clade, UNICORN (UCN, AGC2-3), in controlling integument growth (Scholz et al., 2019). According to Scholz and coworkers, UCN acts as repressor of PDK1 function. The absence of UNC activity or PDK1 overexpression leads to uncontrolled integument and petal growth, and the pdk1-c pdk2 T-DNA insertion mutant combination can rescue the ucn loss-of-function mutant defects. Like Scholtz et al., we did not observe defects in ovule development for the new pdk1 pdk2 allelic combinations (Figure S4B, C). It will be interesting to see if our new pdk1-13 pdk2-4 double mutant combination will also lead to restoration of the ucn-1 flower and ovule phenotypes.
Alternative splicing provides possible functional differentiation for PDK1
By studying pdk1 T-DNA insertion alleles and splice variants produced by the PDK1 gene, we revealed that the PH domain is not essential for the general PDK1 function in Arabidopsis. The alternative splicing product PDK1S0, which lacks phospholipid binding ability and membrane localization but still has kinase activity, is able to rescue the thermosensitive growth of the yeast pkh1 pkh2 double mutant strain INA106-3B, and most of the developmental defects of the Arabidopsis pdk1 pdk2 loss-of-function mutant. A similar PDK1 protein variant appeared to be produced in the Arabidopsis pdk1-a and pdk1-b T-DNA insertion alleles, which were initially thought to be complete loss-of-function alleles. This explains the relatively mild flower phenotypes observed for the Arabidopsis pdk1-a pdk2 and pdk1-b pdk2 double mutants (Camehl et al., 2011; Scholz et al., 2019). The reduced growth response of these mutants induced by phosphatidic acid downstream of Piriformospora indica infection (Camehl et al., 2011) suggests a differential function for PDK1S0 and PDK1 in development and stress response, respectively. This is in line with the observation that phosphatidic acid, an important second messengers for stress response, can directly bind and stimulate the activity of full length Arabidopsis PDK1 (Deak et al., 1999; Anthony et al., 2004). In animals, phospholipids are known to bind to PDK1 to induce PDK1 dimer to monomer conversion and activation (Alessi et al., 1997; Ziemba et al., 2013). The functionality of PDK1S0 in most developmental processes questions the importance of the clear basal polarity of full length PDK1 in (pro) vascular cells in the embryo and root tip. Apparently, PDK1 binding to the AGC kinase PIF domain is sufficiently efficient, and does not require prior co-localisation at the PM. In conclusion, alternative splicing of PDK1 transcripts may provide a novel and unique regulation mechanism for balancing growth and defense in Arabidopsis, which differs from animals and yeast.
Materials and methods
Plant lines and growth condition
Arabidopsis thaliana (L.) ecotype Columbia 0 (Col-0) was used as wild-type control for all experiments, since all mutant and transgenic lines are in the Col-0 background. Previously described T-DNA insertion lines SALK_053385 (pdk1.1-1, renamed to pdk1-c), SALK_11325C (pdk1.1a, renamed to pdk1-a), SALK_007800 (pdk1.1b, renamed to pdk1-b), SAIL_62_G04 (pdk1.2-2, renamed to pdk2-4) and SAIL_450_B01 (pdk1.2-3, renamed to pdk2-1) were ordered from the Nottingham Arabidopsis Stock Centre (Camehl et al., 2011; Scholz et al., 2019). The following Arabidopsis lines are also described elsewhere: pDR5::GFP (Ottenschlager et al., 2003), pDR5::GUS (Benjamins et al., 2001), pPIN1::PIN1:GFP (Benkova et al., 2003), pPIN2::PIN2:GFP (Xu and Scheres, 2005), pPIN3::PIN3:GFP (Zádníková et al., 2010), pPIN7::PIN7:GFP (Blilou et al., 2005) and p35S::PID#21 (Benjamins et al., 2001). For lines created in this study, the T-DNA constructs p35S::YFP:PDK1, p35S::PDK1, pPDK1/2::turboGFP:GUS, and pYAO-Cas9-gRNA1/2/3 were transformed into Col-0 using Agrobacterium-mediated floral dip transformation (Clough and Bent, 1998). Homozygous lines with a single T-DNA insertion were selected for further analysis. Of the 80 CRISPR/Cas9 transgenetic alleles obtained, 7 appeared to contain loss-of-function mutations in the 3rd and 7th exon of PDK1. The CRISPR/Cas9 T-DNA construct in the new pdk1 mutant alleles was segregated out during the generation of the pdk1 pdk2 double mutant. Five mutant alleles with open reading frame shifts were used for further analysis (Figure 2C).
For complementation analysis of PDK1 isoforms, the T-DNA constructs p35S::YFP:PDK1 or p35S::PDK1FL/S1/S2 were transformed into the pdk1-13(+/-) pdk2-4(-/-) mutant background, p35S::YFP:PDKS0 or p35S::PDK1S0 were transformed into the pdk1-13(-/-) pdk2-1(+/-) mutant background, or pPDK1::YFP:PDK1FL and S0 were transformed into the pdk1-14(-/-) pdk2-4(+/-) or pdk1-13(+/-) pdk2-4(-/-) mutant background, respectively. The genotype of the pdk1 pdk2 mutant background was confirmed by PCR before floral dip transformation. All genotyping primers are summarized in Table S2.
Plants were grown on soil at 21 ℃, 16 hr photoperiod, and 70% relative humidity. For seedling growth, seeds were surface-sterilized by 1 minute in 70% ethanol, 10 minutes in 1% chlorine followed by five washes with sterile water. Sterilized seeds were kept in the dark at 4 °C for 2 days for vernalization and germinated on vertical plates with 0.5× Murashige and Skoog (1/2 MS) medium (Duchefa) containing 0.05% MES, 0.8% agar and 1% sucrose at 22 °C and 16 hr photoperiod.
RNA extraction and (q)RT-PCR
Total RNA was extracted from 5-day-old seedlings using a NucleoSpin RNA Plant kit (Macherey Nagel, #740949). Reverse transcription (RT) was performed using a RevertAid Reverse Transcription Kit (Thermo Scientific™, #K1691). For qRT-PCR on auxin induced genes, RNA was isolated from 5-day-old Col-0 and pdk1-13 pdk2-4 seedlings treated for 1 hour with 10 μM IAA. Gene expression was normalized to the reference gene PP2A-3 (AT2G42500) using the ΔΔCt method. For analysis of the pdk1 and pdk2 T-DNA alleles, RT-PCR was performed with DreamTaq DNA Polymerases (Thermo Scientific™). (q)RT-PCR primers are listed in Table S2. For detection of the PDK1 splice variants, RT-PCR was performed for 40 cycles using the forward (FP) and reverse (FL, S0, S1, and S2) primers (Figure 5B), as listed in Table S2. PCR reactions with primer pair FP and (FL)R were digested with BstZ17I, NsiI and SspI to detect PDK1FL, with primer pair FP and (S0)R with BstZ17I and NsiI to detect PDK1S0, and with primer pair FP and (S1)R with BstZ17I to detect PDK1S1. 0.1 μL of the enzymes BstZ17I, NsiI and/or SspI (Thermo Scientific™) was directly added to the 20 μL PCR reaction and reactions were incubated at 37 ℃ overnight before gel electrophoresis. Detection of PDK1S2 and PDK1S3 with primer pair FP and (S2)R did not require restriction enzymes digestion. qRT-PCR was performed in the CFX96 Touch™ Real-Time PCR Detection System (Bio-Rad) using TB Green Premix Ex Taq II (Tli RNase H Plus) (Takara, #RR820B).
Cloning procedures
To generate the Promoter::turboGFP:GUS fusions, a SacI-TurboGFP-PacI fragment was cloned from pICSL80005 into pMDC163, resulting in pMDC163(gateway)-TurboGFP:GUS. PDK1 and PDK2 promoter regions of approximately 2.0 kb including the first six codons were amplified from Col-0 genomic DNA using the primers listed in Table S2, and cloned in pDONR207 by LR recombination. The resulting fragments were subsequently fused in-frame with the turboGFP:gusA reporter gene in pMDC163(gateway)-TurboGFP:GUS by BP recombination. (Invitrogen, Gateway BP/LR Clonase II Enzyme Mix, #11789020 and #12538120).
PDK1 splice variants were amplified from cDNA of 5-day-old seedlings using the respect primers (Table S2), after which restriction enzymes described in the RT-PCR section were employed. Fragments were cloned in pDONR207 by BP recombination, and subsequently transferred to pART7-35S::YFP:gateway by LR recombination, resulting in pART7-35S::YFP:PDK1FL/S0/S1/S2. Expression cassettes were excised with NotI and cloned into NotI digested pART27, resulting in pART27-35S::YFP:PDK1FL/S0. The same entry vectors and LR recombination were used to generate pMDC32-35S::PDK1FL/S0/S1/S2 and pGEX-PDK1FL/S0. pGEX-PIN1HL, pGEX-PIN2HL and p35S:PID:YFP have been described previously (Galván-Ampudia and Offringa, 2007; Huang et al., 2010; Dhonukshe et al., 2010). PIN3HL and PIN7HL were amplified from Col-0 cDNA using primers listed in Table S2 and cloned into pGEX also using Gateway cloning technology to obtain pGEX-PIN3HL and pGEX-PIN7HL.
To generate pPDK1::YFP:PDK1FL/S0 fusions, the 2.0Kb PDK1 promoter region was introduced into pART27-35S::YFP:PDK1S0 by replacing the 35S sequence using restriction enzymes BstXI and KpnI. pART27-pPDK1::YFP:PDK1S0 and pDONR207 were mixed with BP clonase to obtain pART27-pPDK1::YFP:gateway. PDK1FL was then recombined into pART27-pPDK1::YFP:gateway by LR reaction to obtain pART27- pPDK1::YFP:PDK1FL.
To obtain the p416GPD-PDK constructs for expression in yeast, BamHI-PDK1FL/S0/S1/S2-EcoRI and BamHI-PDK2-XhoI fragments were amplified from pDONR207-PDK1FL/S0/S1/S2 and 5-day-old seedling cDNA, respectively, using primers listed in Table S2. Fragments were digested with the appropriate restriction enzymes and ligated into vector p416GPD.
The pCambia-pYAO-Cas9-gRNA1/2/3 plasmids for CRISPR/Cas9 mediated mutagenesis were obtained by ligating the EcoRI-(Cas9+terminator)-AvrII fragment from pDE-Cas9 (Fauser et al., 2014) into pCambia1300 digested with EcoRI and XbaI. The EcoRI and SalI sites in the resulting pCambia-Cas9 plasmid were used to clone the EcoRI-YAO promoter-EcoRI (Yan et al., 2015) and XhoI-gateway-XhoI fragments amplified from respectively Arabidopsis Col-0 genomic DNA and the pART7-35S::YFP:gateway plasmid. Regions producing guide RNAs (Table S2) designed to target respectively the 3rd, 6th or 7th exon of PDK1 were ligated into pEn-Chimera (Fauser et al., 2014), and introduced behind the YAO promoter in pCambia-pYAO-Cas9-gateway by LR recombination.
All primers used for cloning are summarized in Table S2.
General phenotypic analysis and physiological experiments
NPA treated (stock in DMSO,1/104 dilution) or normally grown seedlings, potted plants, siliques and inflorescences were photographed with a Nikon D5300 camera at the indicated time. For imaging of inflorescences, the top part of the inflorescence was cut from 15 cm high plants. For Figure 6A, seedlings were transferred to and aligned on a black plate before imaging. Primary root length, rosette diameter and silique length were measured with ImageJ (Fiji). Plant height was measured directly using a ruler. Root tips, opened siliques, flowers, details of floral organs and cotyledons were imaged using a Leica MZ16FA stereomicroscope equipped a with Leica DFC420C camera. All measurements based on photos were performed in ImageJ and analyzed and plotted into graphs in GraphPad Prism 5.
Phenotypic analysis of reproductive organs
To examine pollen vitality, anthers were collected from flowers just before opening into 70μL Alexander staining buffer [10% ethanol, 0.01% (w/v) Malachite green, 25% glycerol, 5% (w/v) phenol, 5% (w/v) chloral hydrate, 0.05% (w/v) fuchsin acid, 0.005% (w/v) OrangeG and 1.5% glacial acetic acid] on a microscopy slide, covered with cover slip, and incubated at 55 ℃ for 1 hr before imaging. For pollen tube growth, pollen of just opened flowers were transferred to a dialysis membrane placed on solid pollen germination medium [18% Sucrose, 0.01% Boric acid, 1 mM CaCl2, 1 mM Ca (NO3)2, 1 mM MgSO4 and 0.5% agarose] and incubated at 22 ℃ for 18 hrs. Ovules were cleared in chloral hydrate solution (chloral hydrate: glycerol: water = 4:2:1 by weight) for 4 hrs. Stained or germinated pollen and cleared ovules were imaged using a Zeiss Axioplan 2 microscope with DIC optics and Zeiss AxioCam MRc 5 digital color camera. Pollen tube length was measured with ImageJ (Fiji).
Protoplast isolation and transformation
Protoplasts were isolated and transformed as previously described, but with some modifications to the protocol (Schirawski et al., 2000). Protoplasts were isolated from 4-week-old rosette leaves instead of from cell suspensions, and we used a 40% PEG4000 solution and 15 μg pART7-35S::PID:YFP for each transformation.
Auxin transport measurements
Auxin transport assays were carried out as previously reported, with some modifications (Zourelidou et al., 2009). Four 2.5 cm inflorescence stem segments from the basal part of 15cm inflorescence stems were placed in inverted orientation into 30 μL auxin transport buffer (0.5 nM IAA, 1% sucrose, 5 mM MES, pH 5.5) with or without 50 μM NPA for 1 hour, then transferred to 30 μL auxin transport buffer with or without 50 μM NPA containing 200 nM radiolabeled [3H]IAA (Scopus Research BV, Veenendaal, The Netherlands), allowed to incubate for 30 minutes and subsequently transferred to 30 μL auxin transport buffer without radiolabeled [3H]IAA and incubated for another 4 hrs. Segments were cut into 5 mm pieces, the bottom piece (0-0.5 cm) was discarded and the remaining pieces were placed separately into 5 mL Ultima Gold™ (PerkinElmer, # 6013329) for overnight maceration. The [3H]IAA was quantified using a PerkinElmer Tri-Carb 2810TR low activity liquid scintillation analyzer.
GUS staining and microscopy
Fresh seedlings and plant organs were directly soaked into GUS staining buffer [10 mM EDTA, 50 mM sodium phosphate (pH 7.0), 0.1% (v/v) Triton X-100, 0.5 mM K3Fe(CN)6, 0.5 mM K4Fe(CN)6, 1 mg/ml 5-bromo-4-chloro-3-indolyl-D-glucuronide] under vacuum for 15 min and incubated at 37 ℃ for 18 hrs. Subsequently, samples were cleared in 70% (v/v) ethanol at room temperature before imaging with Leica MZ16FA or Leica MZ12 equipped with Leica DFC420C or DC500 camera respectively.
To visualize YFP:PDK1 in embryos and roots or PID:YFP in protoplasts, a Zeiss LSM5 Exciter/AxioImager equipped with a 514 nm laser and a 530-560 nm band pass filter was used on. GFP signals in roots of 5-day-old seedlings were visualized by optionally staining with 10 μg/mL propidium iodide (PI) for 5 min on slides, and observing the samples with a Zeiss LSM5 Exciter/AxioImager equipped with a 488 nm laser and a 505–530 nm band pass filter to detect GFP fluorescence, or a 650 nm long pass filter to detect PI fluorescence. All images were captured with a 40× oil immersion objective (NA = 1.2). Images were optimized in Adobe Photoshop cc2018 and assembled into figures using Adobe Illustrator cc2017. DR5::GFP total intensity was measured from three-dimensional reconstruction of the root tips with ImageJ (Fiji). Apical PIN2:GFP abundance was also measured with ImageJ (Fiji) by drawing a free-hand line along the center of the apical PM of epidermal cells.
in vitro phosphorylation and yeast complementation
In vitro phosphorylation and yeast complementation experiments were performed as previously described (Huang et al., 2010; Dittrich and Devarenne, 2012a)
Acknowledgments
We thank Christian Hardtke and Claus Schwechheimer for providing pax paxl and d6pk012 mutant seeds, respectively. We thank Timothy Devarenne for providing yeast strain p416GPD, Sylvia de Pater for providing the CRISPR/Cas9 plasmids, Xiao Men for sharing preliminary data on PIN2HL phosphorylation by PDK1, Kees Boot for help with the auxin transport assay, and Gerda Lamers and Joost Willemse for help with microscopy. We are grateful to Nick Surtel, Ward de Winter and Jan Vink for their help with plant growth and media preparation. This project was supported by the China Scholarship Council.