Arabidopsis ROCK1 transports UDP-GlcNAc/UDP-GalNAc and regulates ER protein quality control and cytokinin activity

Significance

Nucleotide sugars are donor substrates for the formation of glycan modifications, which are important for the function of many macromolecules such as proteins and lipids. Although most of the glycosylation reactions occur in the endoplasmic reticulum (ER) and Golgi of eukaryotic cells, nucleotide sugar activation occurs in the cytosol and specific transporters must carry these molecules across the membrane. We identified REPRESSOR OF CYTOKININ DEFICIENCY 1 (ROCK1) as an ER-localized transporter of UDP-GlcNAc and UDP-GalNAc in plants. In contrast to animals, nothing is known about the function of the two respective sugar residues in the plant ER. We demonstrate that ROCK1-mediated transport plays a role in the ER-associated protein quality control and loss of ROCK1 enhances cytokinin responses by suppressing the activity of cytokinin-degrading CKX proteins.

Abstract

The formation of glycoconjugates depends on nucleotide sugars, which serve as donor substrates for glycosyltransferases in the lumen of Golgi vesicles and the endoplasmic reticulum (ER). Import of nucleotide sugars from the cytosol is an important prerequisite for these reactions and is mediated by nucleotide sugar transporters. Here, we report the identification of REPRESSOR OF CYTOKININ DEFICIENCY 1 (ROCK1, At5g65000) as an ER-localized facilitator of UDP-N-acetylglucosamine (UDP-GlcNAc) and UDP-N-acetylgalactosamine (UDP-GalNAc) transport in Arabidopsis thaliana. Mutant alleles of ROCK1 suppress phenotypes inferred by a reduced concentration of the plant hormone cytokinin. This suppression is caused by the loss of activity of cytokinin-degrading enzymes, cytokinin oxidases/dehydrogenases (CKXs). Cytokinin plays an essential role in regulating shoot apical meristem (SAM) activity and shoot architecture. We show that rock1 enhances SAM activity and organ formation rate, demonstrating an important role of ROCK1 in regulating the cytokinin signal in the meristematic cells through modulating activity of CKX proteins. Intriguingly, genetic and molecular analysis indicated that N-glycosylation of CKX1 was not affected by the lack of ROCK1-mediated supply of UDP-GlcNAc. In contrast, we show that CKX1 stability is regulated in a proteasome-dependent manner and that ROCK1 regulates the CKX1 level. The increased unfolded protein response in rock1 plants and suppression of phenotypes caused by the defective brassinosteroid receptor bri1-9 strongly suggest that the ROCK1 activity is an important part of the ER quality control system, which determines the fate of aberrant proteins in the secretory pathway.

The biosynthesis of glycans and glycoconjugates (e.g., glycoproteins or glycolipids) requires glycosyltransferases residing in the Golgi apparatus and endoplasmic reticulum (ER). Their activity depends on the presence of activated monosaccharide donor substrates, nucleotide sugars. About 30 different nucleotide sugars have been detected in plants, most of which are synthesized in the cytosol and required to be selectively transported over the compartmental membrane (1). This transport is mediated by nucleotide sugar transporters (NSTs), which generally function as antiporters transporting nucleotide sugars usually in exchange to the corresponding nucleoside monophosphate across the membranes of ER and Golgi (2). They belong to the NST/triose-phosphate translocator family consisting of 40 members in Arabidopsis (3). Transported substrates have been previously identified for 13 NSTs in Arabidopsis, which include UDP-glucose (UDP-Glc), UDP-galactose (UDP-Gal), GDP-mannose (GDP-Man), and CMP-sialic acid (4–8). However, molecular mechanisms underlying transport of other nucleotide sugars in plants are not understood. Interestingly, for some nucleotide sugars, like for example UDP-GalNAc, which was shown to accumulate in plant tissues (9), no target molecule carrying the corresponding sugar moiety has been identified. Hence, the cellular function of several nucleotide sugars is completely unknown in plants (1).

Protein glycosylation can have an influence on protein folding and stability, interaction with proteins and ligands, or enzymatic activity (10–12). Protein N-glycosylation starts within the ER lumen with the transfer of a cytosol-derived core glycan on the nitrogen of an asparagine residue followed by its transformation into a high-mannose glycan (13). After protein transport into the Golgi apparatus, N-glycans can be further modified to hybrid, complex or paucimannosidic N-glycans. The initial committed step in this process is the addition of a GlcNAc residue to appropriately trimmed glycans by N-acetylglucosaminyltransferase I (GnT-I) (14). Whereas there are many examples for luminal protein O-glycosylation on serine and threonine residues of mammalian proteins (15), the only luminal O-glycosyltransferase described so far in plants adds glycans to hydroxyproline residues of proteins found in the cell wall (16).

Glycosylation is essential for protein folding and maturation in the ER, which is equipped with a quality control (ERQC) system that safeguards correct folding and assembly of secretory and membrane proteins in eukaryotic cells and that eliminates improperly folded proteins (17). Under stress conditions, misfolded proteins can accumulate in the ER, which leads to ER stress. This triggers an unfolded protein response (UPR) that alleviates ER stress through enhanced expression of genes encoding components of the protein folding machinery or the ER-associated degradation (ERAD) system (18).

Growth and development of plants is controlled by phytohormones. Among them, a group of N⁶-substituted adenine derivatives called cytokinins (CKs) play an important role in many developmental processes, like regulation of cell proliferation and differentiation in root and shoot apical meristems (SAM) (19–21). The catabolic degradation of CKs is mediated by CK oxidase/dehydrogenase (CKX) enzymes encoded by seven genes in Arabidopsis (22) and their overexpression causes a CK deficiency characterized by complex morphological changes such as a smaller SAM, a dwarfed shoot, and enhanced root growth (19). ckx3,5 Arabidopsis plants develop more active generative shoot meristems, larger flower organs, and more ovules (23). This demonstrates an important role of regulated CK degradation in defining the CK status of a given tissue. Most CKXs are probably N-glycoproteins (22, 24) and the role of glycosylation in fine tuning CKX activity has been discussed (25); however, the hypothesis has not been confirmed and no direct analysis of in planta glycosylation of individual CKX proteins has been reported.

Here we describe a suppressor allele of CK deficiency named repressor of cytokinin deficiency 1 (rock1). Transport assays revealed ROCK1 to be the first transporter of UDP-GlcNAc and UDP-GalNAc identified in plants. We present data demonstrating that ROCK1 plays an important role in regulating CK responses and activity of the generative SAM through modulating the activity of CKX proteins. We provide evidence suggesting that the activity of ROCK1, in supplying the substrate(s) for a yet to be described protein modification, is important for ERQC in plants.

Results

rock1 Decreases the CKX Activity.

To identify new molecular components required for the proper activity of the CK system, we carried out a genetic screen for suppressor alleles of the CK deficiency syndrome displayed by 35S:CKX1 plants. The isolated mutant line rock1 was characterized by restored rosette size, leaf and flower number, flowering time and, to a lesser extent, root growth (Fig. 1A and SI Appendix, Fig. S1). Genetic analysis showed that rock1-1 is a recessive second-site allele (SI Appendix, Table S1) not affecting 35S:CKX1 transgene expression (SI Appendix, Fig. S1C).

Fig. 1.

rock1 suppresses the cytokinin deficiency phenotype by repressing CKX activity. (A) Suppression of the 35S:CKX1 shoot phenotype by rock1-1 mutation in 4-wk-old plants. (B) Relative transcript abundance of A-type ARR genes in shoots of soil-grown seedlings 10 d after germination (dag) measured by quantitative real-time PCR. Data are means ± SD (n = 4; *P < 0.05, t test). (C) Effect of rock1-1 on shoot development in plants expressing 35S:CKX2 or 35S:CKX3. The shoot fresh weight of soil-grown plants was determined 17 dag (means ± SD, n ≥ 15). Significant differences to wild type were determined by t test (*P < 0.05). (D) CKX activity measured in total protein extracts. Activity is expressed relative to wild type. Values are means ± SD (n ≥ 3). Significant differences to the respective CKX overexpression line were determined by t test (*P < 0.05).

To understand whether rock1 directly influenced the CK status, the transcript levels of primary CK response genes, A-type Arabidopsis response regulators (ARRs), were analyzed in the suppressor line. The mRNA levels of all analyzed ARR genes were restored almost to those found in wild type (Fig. 1B). Next, we analyzed the impact of rock1 mutation on the endogenous CK content. Because rock1-1 had stronger effects on shoot than on root development, we determined the CK content specifically in seedling shoots and inflorescences of the suppressor line. CK levels in the rock1-1 suppressor line were five- and twofold increased in comparison with shoots and inflorescences of the parental 35S:CKX1 line, respectively (SI Appendix, Tables S2 and S3); however, the restoration of the CK content was not complete.

To gain information about the specificity of rock1 in suppressing CKX overexpression phenotypes, rock1-1 was introgressed into 35S:CKX2, 35S:CKX3 (19) and 35S:CKX7 plants (26). Whereas rock1-1 fully suppressed phenotypes caused by overexpression of CKX2 and CKX3 proteins localizing to the secretory pathway (Fig. 1C), it had no effect on the phenotypes caused by the overexpression of the cytosolic CKX7 isoform (SI Appendix, Fig. S1D). Further genetic analysis revealed that rock1 had only weak or no effect in suppressing shoot phenotypes of mutant plants lacking two or all three CK receptors (27), respectively (SI Appendix, Fig. S1). Similarly, the phenotype of mutants lacking multiple CK biosynthetic isopentenyltransferase (IPT) genes (28) was only partially suppressed by rock1-1 (SI Appendix, Fig. S1G). Interestingly, comparable restoration of ipt3,5,7 growth was induced by the application of a chemical inhibitor of CKX activity (29) (SI Appendix, Fig. S1G).

Together, the extensive genetic analysis indicated that the main molecular targets of rock1 in suppressing CK deficiency are CKX proteins associated with the secretory pathway. To test this hypothesis biochemically, the CKX activity in 35S:CKX1 parental line and rock1 suppressor was compared. Whereas the CKX activity in 35S:CKX1 seedlings was 22-fold higher in comparison with wild type, rock1 reduced the activity to a level only threefold higher than that of wild type (Fig. 1D). Likewise, the enhanced CKX activity in 35S:CKX2 and 35S:CKX3 plants was reduced through rock1 introgression by 64% and 100%, respectively (Fig. 1D), supporting the notion of rock1 affecting CKX proteins.

ROCK1 Encodes an NST Transporting UDP-GlcNAc and UDP-GalNAc.

The rock1-1 mutation was mapped to a 49-kb interval on chromosome 5. Sequencing candidate genes revealed a G-to-A transition in the first exon of the At5g65000 gene leading to a Gly-to-Arg substitution at amino acid position 29 (SI Appendix, Fig. S2). This substitution is in the first predicted transmembrane domain of the previously uncharacterized protein of the NST family (SI Appendix, Fig. S2). A mutation, thin-exine2 (tex2), in the At5g65000 gene was previously linked to defective pollen exine production (30). Introduction of a genomic complementation construct into rock1-1 35S:CKX1 plants resulted in a full recapitulation of 35S:CKX1 phenotypes, confirming that the rock1-1 mutation was causative for the suppression phenotype (SI Appendix, Fig. S2D). This was further corroborated by isolating two T-DNA insertion null alleles, rock1-2 (30) and rock1-3 (SI Appendix, Fig. S2), which displayed similar developmental changes as rock1-1 (see below).

To identify the subcellular compartment in which ROCK1 functions, we transiently expressed ROCK1 N- and C-terminally fused to GFP under control of the 35S promoter in Nicotiana benthamiana and studied the cellular distribution of the fluorescence signal. The expression of GFP-ROCK1 led to a reticulate GFP signal that colocalized with an ER, but not Golgi, marker (Fig. 2A and SI Appendix, Fig. S3A). In contrast, the ROCK1-GFP fusion clearly colocalized with the Golgi marker (SI Appendix, Fig. S3A), suggesting the possible presence of a C-terminal ER retention/retrieval signal. Indeed, after deleting six C-terminal amino acids in GFP-ROCK1 (GFP-ROCK1^1–319) containing a cluster of five Lys residues, the GFP signal localized mainly in motile dots colocalizing with a Golgi marker and only a very weak ER signal was observed (Fig. 2A). To rule out the possibility that the N-terminal GFP fusion masked an important localization signal, ROCK1 was internally fused with GFP (ROCK1-GFPin) and expressed stably under the ROCK1 promoter in rock1-1 plants. A characteristic net-like GFP signal was detected, indicating that the fusion protein localized to the ER (SI Appendix, Fig. S3B). Again, a C-terminal truncation (ROCK1:ROCK1^1–319-GFPin) relocalized the GFP signal into the Golgi (SI Appendix, Fig. S3C) and only a weak ER signal was detected. Interestingly, both constructs were able to fully complement rock1-1 35S:CKX1 plants (SI Appendix, Fig. S3D). Together, these results revealed that ROCK1 is an ER-resident protein whose localization is largely controlled by its C-terminal di-lysine motif (31).

Fig. 2.

ROCK1 encodes an ER-localized nucleotide sugar transporter. (A) Subcellular localization of ROCK1. 35S:GFP-ROCK1 (Upper) and 35S:GFP-ROCK1^1-319 lacking the putative di-lysine signal (Middle and Lower) were transiently expressed in N. benthamiana leaves and colocalization with marker proteins for ER and Golgi (red) were analyzed. (B) Measurement of ROCK1-mediated uptake of radiolabeled nucleotide sugars into yeast microsomes expressing FLAG-ROCK1 or empty vector. Means ± SEM (n = 3; *P < 0.05, t test). Fuc, fucose; Gal, galactose; GalNAc, N-acetylgalactosamine; Glc, glucose; GlcNAc, N-acetylglucosamine; GlcA, glucuronic acid; Man, mannose; UDP, uridine diphospate; Xyl, xylose.

The molecular function of ROCK1 has so far not been directly studied. Sequence analysis showed that the closest homologs in Arabidopsis are two proteins with unknown function, AT2G43240 and AT4G35335, and CMP-sialic acid transporter AT5G41760 (6) (SI Appendix, Fig. S4) with only low, ∼15%, sequence identity to ROCK1, suggesting that the substrate cannot be inferred from the sequence comparison and, also, that no functional paralogs may exist in Arabidopsis. Consistently, usually a single orthologous sequence was identified in other sequenced plant species. To directly test the transport specificity of ROCK1, a FLAG-tagged ROCK1 protein was expressed in Saccharomyces cerevisiae, which has, with the exception of GDP-Man, a low background for most nucleotide sugar transport activities and is commonly used as a heterologous test system for NSTs (32). ER/Golgi microsomal vesicles isolated from ROCK1 and empty vector control transformed cells were tested in vitro for transport activity with a range of commercially available radiolabeled nucleotide sugars (Fig. 2B). In vesicles expressing ROCK1 the uptake of UDP-GalNAc and UDP-GlcNAc was seven- and threefold increased, respectively, in comparison with the control (Fig. 2B). A low but significant increase was also detected for UDP-Glc. Interestingly, the relative transport of GDP-Man and GDP-Fuc, which is also mediated by the intrinsic yeast GDP-Man transporter (33), was lowered in the ROCK1 microsomes for unknown reasons. Taken together, these data clearly show that ROCK1 functions as a NST transporting UDP-GalNAc and UDP-GlcNAc as main substrates.

ROCK1 Regulates the Activity of the Shoot Apical Meristem.

To understand the function of ROCK1 under physiological conditions, we analyzed the rock1 mutations in the absence of the 35S:CKX1 transgene. The most prominent morphological changes were observed during generative growth, which was overall accelerated in rock1 plants. All three rock1 mutants developed enlarged inflorescences (Fig. 3A) and detailed analysis showed that the frequency of flower initiation was increased by 30% in comparison with wild type. Additionally, stem elongation was accelerated by up to 23% (Fig. 3B). Seven weeks after germination, rock1 had generated ∼50% more flowers and siliques on the main stem than did the wild type (Fig. 3B). Continuous flower initiation results from the activity of the inflorescence meristem (IM). Scanning electron microscopy analysis revealed that the IM in rock1 was strongly enlarged and initiated supernumerary flower primordia (Fig. 3C), demonstrating that the enhanced flower formation in rock1 plants was due to increased IM activity and that ROCK1 plays a negative regulatory role in this process. These phenotypic changes were strongly reminiscent of those caused by the loss of the CKX3 and CKX5 genes (23).

Fig. 3.

ROCK1 regulates the activity of the shoot apical meristem. (A) The main inflorescence of the wild-type and rock1-2 plants 5 wk after germination. (B) Number of flowers and siliques (stages 13–18) and height of the main stem of 7-wk-old rock1 mutants and wild type. Values are means ± SD (n ≥ 20; *P < 0.05, t test). (C) Scanning electron micrographs of inflorescence meristems (IM) of 4-wk-old wild-type and rock1-2 plants. (Scale bar, 30 µm.) (D) Activity of the cytokinin reporter construct ARR5:GUS in the shoot meristems of Arabidopsis seedlings 2 dag. Staining performed for 1 h. (E) Histochemical detection of ROCK1:ROCK1-GUS activity in the IM and young flowers. (F) Analysis of cytokinin metabolic profiles in wild-type and rock1 seedlings after feeding with ³H[iP] for 2 h. Values are means ± SD (n = 3; *P < 0.05, t test). iP9G/iP7G, iP-N9/7-glucoside; iPR, iP riboside; iPRP, iPR 5′-phosphate.

Transcript levels of A-type ARR genes were elevated in rock1 shoots (SI Appendix, Fig. S5A) and the activity of the CK reporter ARR5:GUS was increased in the shoot meristem of rock1 plants (Fig. 3D), suggesting that ROCK1 regulates SAM activity through adjusting CK signaling in the meristem. Endogenous CK levels were increased up to 35% in rock1 inflorescences in comparison with the control (SI Appendix, Tables S4 and S5). In accordance with the observed changes in meristem development, the ROCK1:ROCK1-GUS reporter construct revealed that ROCK1 is strongly expressed in the SAM and young flowers (Fig. 3E), supporting a direct role of ROCK1 in regulating shoot meristem development. The reporter construct further revealed expression of ROCK1 in numerous other tissues (SI Appendix, Fig. S6); however, we observed neither changes in root development nor altered responses to exogenous CK in rock1 roots (SI Appendix, Fig. S5), supporting the notion that ROCK1 is more relevant for regulating CK responses in the shoot.

To analyze whether rock1 alters CK responses through regulating CKX activity also under physiological conditions, we performed feeding experiments in which we supplied plants with radiolabeled CK (isopentenyladenine, iP) and followed its metabolic conversion. The level of degradation products of CKX reaction was reduced by 30% in rock1 plants after a 2-h incubation, whereas the fraction containing iP with the corresponding riboside and nucleotide was significantly larger in comparison with wild type (Fig. 3F). This further substantiates a ROCK1 regulatory function in tuning CKX-mediated CK degradation.

ROCK1 Plays an Important Role in ERQC.

Next we aimed to analyze the molecular mechanism underlying the regulation of CKX activity and to understand the function of ROCK1-transported substrates in this process. Whereas there is virtually no cellular activity requiring UDP-GalNAc known in plants, UDP-GlcNAc is a substrate of GnT-I in a step converting high mannose to hybrid and complex N-glycans. We tested CKX1 glycosylation and the nature of linked N-glycans. Total proteins from Arabidopsis plants expressing N-terminally myc-tagged CKX1 (myc-CKX1) from the 35S promoter were extracted and subjected to treatment with peptide N-glycosidase F (PNGase F) removing all N-linked oligosaccharides except those carrying core α1,3-fucose. Immunoblot analysis revealed an electrophoretic mobility shift of myc-CKX1 (Fig. 4A), showing that the protein contains N-linked oligosaccharides. Treatment with endoglycosidase H (EndoH), which is unable to cleave complex N-glycans, resulted in a similar mobility shift of myc-CKX1, suggesting that CKX1 contains mainly high-mannose N-glycans. Interestingly, ROCK1 is not substantially involved in CKX1 N-glycosylation as myc-CKX1 extracted from rock1-1 plants showed no obvious difference in mass compared with myc-CKX1 from wild type and was comparably affected by PNGase F and EndoH treatment (Fig. 4A). Similarly, rock1 did not influence the overall protein modification with complex N-glycans as indicated by the immunoblot analysis with antibodies against complex N-glycans (SI Appendix, Fig. S7).

Fig. 4.

ROCK1 regulates ERQC and CKX protein abundance. (A) N-glycosylation status of myc-CKX1 protein in wild-type and rock1 seedlings. Protein extracts were treated with PNGase F or EndoH and the size of myc-CKX1 was compared with mock-treated control by SDS/PAGE and immunoblot with anti-myc antibody. (B) The loss of GnT-I activity in cgl1 mutants has no influence on the shoot phenotype of 35S:CKX1 plants (21 dag). (C) The level of myc-CKX1 is decreased in rock1 compared with wild type. Total protein extracts were analyzed by immunoblot with anti-myc antibody. Coomassie blue staining of Rubisco large subunit (RbcL) was used as loading control. Relative densitometric analysis of the myc-CKX1 signal is shown. Values are means ± SEM (n = 10; **P < 0.01, t test). (D) Analysis of myc-CKX1 stability. Protein extracts were prepared from 7-d-old seedlings treated with mock or 100 μM CHX for indicated times and analyzed by immunoblotting. (E) myc-CKX1 is degraded in proteasome-dependent manner. Seedlings were treated with 100 μM MG132 and myc-CKX1 analyzed by immunoblotting and densitometry. Means ± SEM (n = 4; **P < 0.01, t test). (F) Quantitative real-time PCR analysis of UPR genes in shoots of soil-grown seedlings 9 dag. Means ± SD are shown (n = 4; *P < 0.05, t test). (G) The rock1 mutation suppresses the brassinosteroid deficiency shoot phenotype of bri1-9 mutant plants (26 dag).

To test unequivocally whether CKX1 activity is dependent on hybrid or complex N-glycans, the complex glycans less 1 (cgl1) mutation of GnT-I was introgressed into 35S:CKX1 plants. As Fig. 4B shows, cgl1-2 had no effect on the 35S:CKX1 phenotype, indicating that CKX1 function is independent of GnT-I activity and further supporting the idea that ROCK1 does not provide UDP-GlcNAc for this reaction.

The protein immunoblot analysis revealed that the level of myc-CKX1 was consistently lower in rock1-1 compared with wild type (Fig. 4C), suggesting that CKX1 protein abundance might be controlled by ROCK1. To test CKX1 turnover, we analyzed myc-CKX1 levels in the presence of the translation inhibitor cycloheximide (CHX). As shown in Fig. 4D, myc-CKX1 was relatively unstable, with a half-life of ∼4 h. The turnover of myc-CKX1 in rock1 was comparable to wild type (Fig. 4D). Interestingly, treatment with MG132, a widely used inhibitor of the proteasome, increased the level of myc-CKX1 in wild type (Fig. 4E), indicating that CKX1 protein, which has been shown to localize to the ER/secretory system (19), is degraded by a proteasome-dependent ERAD mechanism. Intriguingly, inhibition of the proteasome in the rock1 background strongly stabilized myc-CKX1 levels (Fig. 4E), suggesting that the lower myc-CKX1 steady-state levels in rock1 were caused by increased ERAD. Reduced levels of myc-CKX1 could thus indicate inefficient protein processing and folding. This was supported by the analysis of the UPR status through measuring the expression level of typical ER stress response genes, encoding components of the ER protein-folding machinery. Fig. 4F shows that the steady-state transcript levels of the binding protein 1 (BiP1), calnexin 1 (CNX1), and calreticulin 2 (CRT2) genes were significantly increased by up to twofold in rock1 plants in comparison with wild type, demonstrating that UPR was constitutively enhanced and further suggesting defects in ERQC caused by the rock1 mutation. To address this notion experimentally, we used a mutant allele of the brassinosteroid receptor gene, brassinosteroid insensitive 1-9 (bri1-9). The gene product is functionally competent as a hormone receptor but is retained by the ERQC system and degraded by ERAD, causing severe dwarfing of this receptor mutant (34) (Fig. 4G). Introgression of rock1-2 into bri1-9 led to a strong suppression of the dwarf bri1-9 phenotype (Fig. 4G), indicating that bri1-9 leaked from its ER retention machinery, which became compromised in rock1 in a similar fashion as described for other suppressor genes of bri1-9 (35). This was confirmed by the detection of an EndoH-resistant, complex N-glycan–carrying form of bri1-9 in rock1 (SI Appendix, Fig. S9). Hence, our data indicate that ROCK1 is a very important component of the protein folding machinery and/or ERQC in plants.

Discussion

We could show that Arabidopsis ROCK1 is a NST with main transport activity toward UDP-GlcNAc and UDP-GalNAc. Multiple lines of evidence show that ROCK1 is a positive regulator of CKX activity in plants, which raises the question of how transport of these substrates is functionally connected to regulation of CKX proteins.

In the lumen of the plant secretory pathway, UDP-GlcNAc is used for the formation of hybrid or complex N-glycans on secreted and membrane proteins (14) and it has been suggested that N-glycosylation is relevant for regulating CKX activity (25). Our genetic data showed that CKX1 activity is independent of the presence of complex N-glycans, which indicates that the reduced GlcNAc availability for GnT-I in the Golgi was not the cause for the reduced CKX1 activity in rock1. In the same line of evidence, the overall protein N-glycosylation pattern was not apparently affected in rock1, suggesting also that a ROCK1-independent transport route for UDP-GlcNAc into the Golgi exists. The notion that ROCK1-mediated transport of UDP-GlcNAc does not serve the complex N-glycan processing in the Golgi was also supported by localization of ROCK1 in the ER membrane.

Beside the N-linked oligosaccharides, various proteins bearing a single GlcNAc residue attached to the Asn of the canonical Asn-X-Ser/Thr sequon have been described in fungi (36) and recently also in animals (37) and plants (38). Interestingly, Kim et al. proposed that such N-GlcNAc modifications originate by a mechanism that is, at least in part, different from the cleavage of N-linked oligosaccharides (38). The corresponding glycosyltransferase is, however, currently unknown. Additionally, single O-linked GlcNAc is a well-studied reversible posttranslational modification of Ser/Thr residues on cytosolic and nuclear proteins of higher eukaryotes (39). Interestingly, an analogous modification has recently been identified on extracellular domains of Drosophila Notch protein and the corresponding glycosyltransferase localizes to the ER (40).

ROCK1 was identified to be a multispecific NST with high transport affinity also for UDP-GalNAc. Transport of UDP-GalNAc has recently been described in tobacco (41) and it could be that a ROCK1 homolog is responsible for this activity. However, despite the occurrence and transport of UDP-GalNAc in plants, the function of the corresponding sugar moiety remains obscure. Although plants obviously lack the machinery to produce the, typically mammalian, mucin-type O-glycosylation of proteins (41), several earlier reports described GalNAc as part of glycoproteins in higher plants and green algae (42–44). It will be important to investigate whether any of the above discussed rare glycosylations reactions involving GlcNAc or GalNAc occur within the plant ER. A direct comparison of the glycan composition of CKX proteins isolated from wild-type and rock1 plants could be instrumental in identifying these modifications.

rock1 was identified due to its capacity to suppress CK deficiency caused by CKX1 overexpression. We could show that CKX1 is an ERAD substrate and that rock1 mainly affects CKX1 by reducing its abundance. It cannot be currently excluded that rock1 influences the enzymatic activity and/or subcellular localization of CKX1 as well. This, together with the fact that the myc-CKX1 levels were restored in rock1 plants upon inhibiting the proteasomal degradation, suggests the protein folding capacity or fidelity is compromised in the rock1 mutant, resulting in enhanced ERAD of CKX1. The hypothesis, that ROCK1 plays an important role during quality monitoring of secretory proteins in the ER was underpinned by the complete phenotypic suppression of bri1-9 by rock1, which was not an indirect CK effect (SI Appendix, Fig. S9). It has been shown previously that, similarly to rock1, the loss of other components of the ERQC system leads to leakage of the bri1-9 protein from the ER and restoration of brassinosteroid responses (35, 45). The exact mechanism of ROCK1 interaction with the ERQC and ERAD pathway is currently unclear. It needs to be investigated whether some of the above discussed glycan modifications serve as a signal for protein folding, quality control, and ERAD. It is also important to keep in mind that for the correct activity of the ERQC machinery an import of UDP-Glc into the ER lumen is required to supply UDP-glucose:glycoprotein glucosyltransferase (UGGT) (17). This enzyme adds glucose to the branched N-linked oligosaccharide present in unfolded proteins, to form Glc₁Man₉GlcNAc₂. The terminal glucose present in this structure is bound by chaperones in the calnexin/calreticulin cycle controlling the folding status of glycoproteins in the ER. The loss of UGGT compromises the ERQC in Arabidopsis, which leads to bri1-9 suppression (34). In our assay system, yeast microsomes expressing ROCK1 showed only a weak increment in the transport of UDP-Glc, which could be due to relatively high background transport activity in control microsomes. This raises the question of whether this transport also contributes to the total UDP-Glc influx into the ER in planta and, hence, whether ROCK1 can function, at least partially, in supplying UDP-Glc for UGGT. This hypothesis is, however, not supported by the work of Reyes et al. (7), who has proposed that AtUTr1 and AtUTr3 are the main, if not the only, NSTs involved in the import of UDP-Glc into the Arabidopsis ER. In contrast to these two transporter genes, ROCK1 expression is not induced by ER stress (SI Appendix, Fig. S10).

The activity of the reproductive SAM was strongly increased in rock1 to an extent similar to that caused by mutation of CKX3 and CKX5 (23). This result is very consistent with the proposed role of ROCK1 as a positive regulator of CKX proteins, which was further supported by the increased CK content in the rock1 inflorescences. Thus, the modification of CKX activity by ROCK1 adds an additional layer of fine tuning of meristem activity and organ formation. It remains to be determined whether ROCK1 activity modulates CK responses also in other tissues. For example, ROCK1 was earlier identified in a screen for genes involved in pollen exine development and designated as THIN-EXINE2 (TEX2) (30). The pollen-localized activity of ROCK1 reporter shown in this work is consistent with the tex2 mutant pollen phenotype. It has been shown that CK signaling is involved in regulation of pollen development (46) and it will be interesting to analyze whether CKX isoforms expressed in pollen (19) are regulated by ROCK1. Alternatively, it could be that GlcNAc/GalNAc is directly required for the biosynthesis of sporopollenin, which is a major component of the exine layer, or the polysaccharide-containing exine precursor, primexine (47).

Materials and Methods

For the plant material, the gene mapping, the cloning of the constructs, the molecular and microscopic analyses, the transport assay, and the cytokinin measurements, see SI Materials and Methods.