Chs3, the catalytic subunit of chitin synthase III in Saccharomyces cerevisiae, is a complex polytopic membrane protein whose plasma membrane expression is tightly controlled: export from the ER requires interaction with Chs7; exit from the Golgi is dependent on the exomer complex, and precise bud neck localization relies on endocytosis. Moreover, Chs3 is efficiently recycled from endosomes to the TGN in an AP-1-dependent manner. Here we show that the export of Chs3 requires the cargo receptor Erv14, in a step that is independent of Chs7. Chs3 oligomerized in the ER through its N-terminal cytosolic region. However, the truncated Δ126Chs3 was still exported by Erv14, but was sent back from the Golgi to the ER in a COPI- and Rer1-dependent manner. A subset of the oligomerization-deficient Chs3 proteins evaded Golgi quality control and reached the plasma membrane, where they were enzymatically active but poorly endocytosed. This resulted in high CSIII levels, but calcofluor white resistance, explained by the reduced intercalation of calcofluor white between nascent chitin fibres. Our data show that the oligomerization of Chs3 through its N-terminus is essential for proper protein trafficking and chitin synthesis and is therefore monitored intracellularly.
Chitin is the second most abundant biopolymer and is an essential component of the cell wall of most fungi. Chitin is synthesized into the periplasmic space by chitin synthases (CS). The yeast Saccharomyces cerevisiae contains three different CSs with distinct biological functions: CSII is involved in the synthesis of the chitin disk that separates mother and daughter cells (Shaw et al., 1991), while CSI has only repair functions (Cabib et al., 1989). Finally, Chs3, which is the catalytic subunit of CSIII (Valdivieso et al., 1991), is responsible for the synthesis of approximately 90% of cellular chitin (Shaw et al., 1991), most of which is assembled in a chitin ring that guarantees cellular integrity at the neck during cell division (Roncero and Sanchez, 2010). Several mutants with defects in CSIII activity have been isolated and their further characterization has indicated that the corresponding Chs proteins are required for the intracellular transport of Chs3 to the plasma membrane (PM) (see Roncero, 2002).
After synthesis and folding in the endoplasmic reticulum (ER), Chs3 depends on its export chaperone Chs7 for ER exit (Trilla et al., 1999), similar to Shr3 for the amino acid permease Gap1 and Pho86 for the phosphate transporter Pho84 (Dancourt and Barlowe, 2010). These chaperones presumably prevent aggregation of the corresponding proteins through their TM domains (Kota and Ljungdahl, 2005). However, the precise mechanism(s) of chaperone–cargo interaction and ER exit remains unknown (Dancourt and Barlowe, 2010). After transit through the Golgi, Chs3 is sorted in a Chs5- and Chs6-dependent manner (two subunits of a complex also known as exomer) into distinct secretory vesicles, ensuring polarized transport to the PM in a cell cycle-dependent manner (Sanchatjate and Schekman, 2006; Trautwein et al., 2006; Wang et al., 2006; Reyes et al., 2007; Zanolari et al., 2011). Interestingly, the requirement for Chs5 or Chs6 can be bypassed by mutations in the endosome-TGN retrograde transport, allowing a partial rerouting of Chs3 to the PM (Valdivia et al., 2002). Chs3 interacts with Chs4, which is required for chitin synthase activation and anchoring of Chs3 to the neck; failure of which results in the immediate endocytosis of Chs3 (Reyes et al., 2007; Sacristan et al., 2012). Upon endocytosis, Chs3 is efficiently sorted at the endosomal compartment, preventing its transport to the vacuole (Ziman et al., 1996) in an AP-1 and Ent5-dependent manner (Valdivia et al., 2002; Copic et al., 2007). These complex sorting and recycling mechanisms allow the rapid accumulation of Chs3 at the PM under stressful conditions (Valdivia and Schekman, 2003). While Chs4 and Chs7 are exclusively devoted to CSIII regulation, Chs5 and Chs6 are involved in the transport of at least one other chitin-unrelated cargo (Barfield et al., 2009). The above studies have provided the basis for the use of Chs3 as a model for studying different aspects of the intracellular transport of PM proteins.
Although many of the proteins that ensure the timely and spatial control of Chs3 trafficking have been identified, the signals of Chs3 that are required for the complex trafficking pattern remain largely elusive. Two recent reports have deciphered the requirement of the N-terminal and C-terminal domains of Chs3 for the correct sorting of the protein by AP-1 and the exomer respectively (Rockenbauch et al., 2012; Starr et al., 2012). However, such analysis has been hampered by the fact that Chs3 is a polytopic membrane protein for which a number of topologies have been described (http://www.uniprot.org/uniprot/P29465; Banks et al., 2005; Meissner et al., 2010). It appears to be clear, however, that four globular domains exist. Nevertheless, depending on the algorithm used a variable number of TM domains can be proposed. Any topological model advanced needs to be reconciled with the multiple post-translational modifications reported for this protein. Chs3 is glycosylated (Cos et al., 1998), phosphorylated (Li et al., 2007; Albuquerque et al., 2008) and ubiquitinated (Peng et al., 2003). However, the function of these post-translational modifications remains elusive. In contrast, Chs3 from Candida albicans has been shown to depend on a single phosphorylation site for proper PM localization (Lenardon et al., 2010). Finally, Chs3 is also palmitoylated at the ER, a modification required for its exit from the ER (Lam et al., 2006).
In an attempt to decipher the available clues about CSIII regulation, we have addressed the topological organization of Chs3 obtaining a model other than previously described ones. More importantly, we demonstrate that the oligomerization of Chs3 is a crucial process for the exit of Chs3 from the ER and for the correct functionality of CSIII at the PM.
Improved topological model of Chs3 through analysis of post-translational modifications
Chs3 is used as paradigm to study the temporally and spatially controlled delivery of cargo proteins to the PM (Holthuis et al., 1998; Valdivia et al., 2002; Trautwein et al., 2006), based on the characterization of the effect of different mutations on Chs3 transport. However, there have been no systematic studies addressing the role that the different domains of Chs3 play in this transport, partially owing to the rather complex polytopic nature of Chs3. As a complementary approach to previous studies, we were interested in deciphering the signals intrinsic to Chs3 that govern its transport to the PM and hence we started by elucidating the membrane topology of Chs3. Bioinformatics methods consistently predict four globular domains but varying numbers of trans-membrane domains (Banks et al., 2005; Merzendorfer, 2011) (Fig. 1A). High-throughput mass spectrometric analyses (http://www.uniprot.org/uniprot/P29465) have indicated that domains I and III of Chs3 are phosphorylated. This led to the only topological model proposed so far of domains I, II and III of Chs3 facing the cytosol and domain IV facing the periplasmic space (Meissner et al., 2010). However, the C-terminal domain of Chs3 has been shown to be cytoplasmic (Kim et al., 2006), which is in agreement with the correct processing of this region in a split-ubiquitin system (C. Sacristan, unpubl. obs.). In addition, the proposed topology failed to take into account the known N-glycosylation of Chs3 (Cos et al., 1998). Therefore, the previously proposed model is likely to be incorrect, in particular with respect to the positioning of domain II.
We were unable to detect any cytoplasmic post-translational modifications for region II (results not shown), suggesting that this domain could be exposed to the extracellular space, thus being prone to N-glycosylation. This domain contains three potential N-glycosylation sites at positions 303, 332 and 371. Upon mutating all three asparagines to glutamines, the N-glycosylation of Chs3 was completely abolished (Fig. S1A). This result is in agreement with the recent identification of N-glycosylation at position 332 of Chs3 in a high-throughput analysis (Breidenbach et al., 2012). Therefore, region II must reach the ER lumen and hence face the extracellular space at the PM. Based on our data, we revised the topology model for Chs3 and propose that while regions I, III and IV are cytoplasmic, region II would face the extracellular space (Fig. 1A), being glycosylated at the indicated positions. This model now needs to be conciliated with conclusions based on previous incorrect topological models (Meissner et al., 2010).
The N-terminal domain of Chs3 is required for Chs3 activity and localization
Based on this model, we first focused on determining the role of the post-translational modifications of Chs3. The N-glycosylation-deficient strains containing Asn (N) to Gln (Q) single, double or triple mutations were as sensitive to calcofluor white as the wild type (Fig. S1B) and showed normal calcofluor white staining (Fig. S1C). In addition, the mutated Chs3 proteins localized in a manner that was indistinguishable from that seen for wild-type Chs3, accumulating at the neck and at punctate intracellular structures (Fig. S1D). These results suggest that N-glycosylation is not essential for Chs3 localization and function.
The substitution of several phosphorylated S/T residues in the N-terminal region of Chs3 did not produce apparent phenotypes (Fig. S2). Therefore we next focused on the role of the cytosolic N- and C-terminal regions of Chs3 by creating increasing deletions in both extremes. Interestingly, deletions at the N-terminal (domain I) or C-terminal regions (domain IV) (see supplementary Fig. S3 for details) compromised Chs3 functionality based on the increased resistance to calcofluor white of the corresponding strains (Fig. 1B). The proteins lacking the C-terminal region (Chs3Δ25 or Chs3Δ37, see Fig. S3 for details) elicited full resistance to calcofluor white (Fig. 1B) due to its retention at the Golgi (Rockenbauch et al., 2012).
Much more puzzling were the results obtained by truncating the Chs3 protein at the N-terminal domain I (Chs3Δ63, Chs3Δ126 and Chs3Δ194, see Fig. S3 for details on truncations). All the truncations conferred resistance to calcofluor white, although to different degrees. The Δ194Chs3 protein promoted full resistance to calcofluor white (Fig. 1B) that was associated with very low chitin levels, similar to those of the original chs3Δ mutant (Fig. 1C). However, the Δ63Chs3- and Δ126Chs3-containing strains were partially resistant to calcofluor white (Fig. 1B) but showed normal calcofluor white staining (not shown) and normal levels of chitin (Fig. 1C).
When we addressed the localization of these N-terminal-truncated proteins by GFP tagging, we found that all of them accumulated at the ER (Fig. 1D). The wild-type Chs3 localized to the bud neck in small and large budded cells (32.6% ± 4.7; n = 265 cells). In cells expressing the truncations, an accumulation of Δ63Chs3-GFP or Δ126Chs3-GFP at the ER was noticeable, but a significant part of these proteins still reached the bud neck (bud neck staining: 16.9% ± 2.9; n = 177; and 17.5% ± 3.6; n = 182 cells respectively). However, Δ194Chs3-GFP was almost exclusively retained in the ER (bud neck staining: 0.0% ± 0.0; n = 100) (see supplementary Fig. S5A for a quantitative analysis of the localization). These results are consistent with a dual function of the N-terminal region I: namely, chitin synthase activity and Chs3 localization.
While the absence of functionality of the Δ194Chs3 protein could be explained easily in terms of the collapse of its intracellular transport (see Discussion), the results obtained with the smaller truncations raised two immediate questions: (i) Why do these proteins accumulate at the ER? (ii) Why are these proteins apparently fully functional in chitin synthesis (Fig. 1C) despite their significant retention at the ER? We shall concentrate on the characterization of the Δ126Chs3 protein to address these questions.
Chs3 exits the ER using the receptor Erv14
The accumulation of Δ126Chs3 at the ER could be due to its inefficient exit from the ER. Previous reports have indicated that a critical step in Chs3 exit from the ER is the prevention of the formation of protein aggregates (Kota and Ljungdahl, 2005). In the absence of palmitoylation (pfa4Δ mutant) (Lam et al., 2006) or the Chs7 chaperone (Kota and Ljungdahl, 2005), Chs3 aggregates through its TM domains, preventing its loading into COPII vesicles and hence its exit from the ER. Thus, we wondered whether Δ126Chs3 aggregates in the ER. After DSP treatment Δ126Chs3 still appeared mostly in a monomeric form, similarly to the wild-type protein (Fig. 2A). In contrast, Δ194Chs3 ran mostly as high-weight aggregates, similarly to the wild-type protein in the absence of Chs7 (Fig. 2A). Thus, at least in the case of Δ126Chs3 the formation of aggregates was not the cause of ER retention and its behaviour cannot be functionally linked to Chs7 or palmitoylation.
An alternative explanation would be that truncated Chs3 proteins fail to be recognized by the COPII export machinery that drives the selective capture and concentration of the secretory cargo into ER-derived vesicles (Dancourt and Barlowe, 2010). To address this possibility, we first attempted to identify the mechanism by which Chs3 is exported from the ER by testing the role of different cargo receptors in the ER export of Chs3. As shown in Fig. 2B, Chs3 was retained at the ER in the erv14Δ mutant but not in mutants lacking other types of ER cargo receptors. Consistent with this, the erv14Δ mutant is resistant to calcofluor white while the rest of the cargo receptor mutants are not (Fig. 2C). In addition, Erv14 specifically co-immunoprecipitated Chs3, but not other unrelated secretory proteins such as the GPI-anchored protein Gas1, a well-known cargo of the p24 receptor (Muñiz et al., 2000). Taken together, these results indicate that Erv14 functions as the ER exit receptor for Chs3.
Once we had identified Erv14 as the cargo receptor for Chs3, we next tested whether Erv14 fails to recognize the Δ126Chs3 protein. Interestingly, both proteins were efficiently co-immunoprecipitated (Fig. 2D). However, the fact that the receptor interacts with Δ126Chs3 does not rule out the possibility that the deleted region harbours a specific sequence for the sorting of Chs3 at the ER. In order to isolate that hypothetical signal we constructed new mutants of Chs3 deleting partial regions within the first 126 amino acids of the protein (Fig. S4). All of the partial truncations behaved similarly to the original ones, showing partial resistance to calcofluor white and ER retention (Fig. S4A and B). Importantly, the non-overlapping truncations Δ63Chs3 and Δ63–125Chs3 were retained at the ER, suggesting that there is not a discrete signal for the sorting of Chs3 at the ER.
Δ126Chs3-GFP is retrieved from the Golgi to the ER in a COPI- and Rer1-dependent manner
Given that there are no apparent reasons for preventing Δ126Chs3 exit from the ER, its accumulation at this organelle could be explained alternatively in terms of its recycling from the Golgi compartment if Δ126Chs3 were recognized there as a misfolded protein and then be returned back to the ER in COPI retrograde transport vesicles. To test this hypothesis, we blocked retrograde transport from the Golgi to the ER by using a temperature-sensitive mutant in the α-COP subunit of the coatomer of COPI vesicles, ret1-1. Under these conditions, Δ126Chs3 was no longer efficiently retained in the ER but mainly localized to the PM and the vacuole (Fig. 3B, see also supplementary Fig. S5B for a quantitative analysis, and Fig. S7 for additional images). In addition, a mutated version of Chs3 (Chs3KK1125/6RR-GFP) lacking a C-terminal di-lysine motif putatively involved in COPI loading was not retained in the ER and reached the PM efficiently (Fig. 3A and Fig. S5B). Taken together, our data strongly suggest that Δ126Chs3 is a substrate for a Golgi quality control system, resulting in its retrograde transport in COPI vesicles to the ER for refolding.
Rer1 has been demonstrated to mediate the retrieval of type I and polytopic membrane proteins to the ER in both yeast and mammalian cells (review in Dancourt and Barlowe, 2010). Its function is not restricted to the retrieval of ER-resident transmembrane proteins since the surface expression of the acetylcholine receptor in muscle cells is also dependent on Rer1, which retrieves unassembled alpha subunits and sends them back to the ER and prevents their degradation in lysosomes (Valkova et al., 2011). We tested whether the retrieval of Δ126Chs3-GFP was also dependent on Rer1. While the deletion of RER1 essentially had no effect on Chs3-GFP localization, Δ126Chs3-GFP was no longer efficiently retained in the ER and accumulated partially in the vacuole (Fig. 3C; see also supplementary Fig. S5B). Similar results were obtained with all the other truncations, whose ER retention was also relieved in the rer1Δ mutant (Fig. S6). In agreement with this result, Rer1 immunoprecipitated together with Δ126Chs3 more efficiently than with the wild-type Chs3 protein (Fig. S5D). Thus, Rer1 may recognize the incorrectly folded Δ126Chs3 protein and facilitate its inclusion in COPI vesicles destined for the ER.
All the partial N-terminal truncations constructed behaved similarly, showing alleviated retention at the ER in the ret1-1 and rer1Δ mutants (see Fig. S6B). Therefore our results clearly suggest that the retrograde transport from the Golgi is the main cause of the accumulation of Δ126Chs3 and all the other N-terminal truncated proteins at the ER. We have previously shown that Chs7 is synthesized in limited amounts (Trilla et al., 1999), which could lead to a functional shortage of Chs7 after Chs3 accumulation in the strains containing Chs3 truncations. If this were true, overexpression of CHS7 should also alleviate the excessive accumulation of N-terminal truncated proteins at the ER. Indeed, after CHS7 overexpression N-terminal truncated proteins were partially released from the ER (Fig. 3D, see also Figs S5 and S6 for additional truncations). In contrast, the overexpression of CHS7, ret1-1 or rer1Δ did not have any apparent effect on Δ194Chs3 traffic (Fig. 3D, see also Fig. S6), in agreement with its ER aggregation (Fig. 2).
Chs3 oligomerizes through its N-terminal domain
Although, we have convincingly shown above that the N-terminal truncated forms of Chs3 are efficiently retrieved from the Golgi, we were puzzled by the reason for such retrieval. The most obvious reason, namely incorrect folding, cannot have been involved because the truncated proteins were functional. One evident possibility would be that by deleting the N-terminal region, we would be interfering with the insertion, and probably the topology, of Chs3 in the ER. To test this possibility, we checked the glycosylation state of the N-terminal truncations Δ63Chs3 and Δ126Chs3 as described above. However, we were unable to detect any difference in the glycosylation pattern between the wild-type and mutant proteins (not shown), indicating that truncated proteins are efficiently recognized by the N-glycosylation machinery and hence that they are correctly inserted into the ER membrane.
It has been described that Rer1 participates in the retrieval of a number of protein complexes incorrectly assembled at the ER (Sato et al., 2004; Valkova et al., 2011), and it has been proposed that Chs3 may form oligomers, the N-terminal half of the protein probably being involved in this oligomerization (DeMarini et al., 1997). To follow up these initial findings, we performed an in-depth yeast two-hybrid analysis. We used a C-terminal-truncated version of Chs3 that contained only the first 855 amino acids (Chs31–855), since inclusion of the C-terminal region, which contains multiple transmembrane domains, abolished two-hybrid interactions. This truncated protein was able to interact with itself and with Chs4 (Fig. S8), as previously reported (DeMarini et al., 1997; Meissner et al., 2010). Deletion of the first 210 amino acids of Chs3 abolished the interaction with Chs3 but not with Chs4 (Fig. S8). In addition, the expression of only the first 230 amino acids of Chs3 (Chs31–230) was sufficient for interaction with Chs31–855, but not with Chs4 (Fig. S8), whose interaction site has been assigned roughly to the middle part of Chs3 (Meissner et al., 2010). These results indicate that the N-terminal domain I was required for the self-association of Chs3. To corroborate our findings, we resorted to the more quantitative β-galactosidase assay (Fig. 4A). The creation of N-terminal deletions demonstrated that the first 63 aa of Chs3 were only partially involved in self-recognition because of the modest reduction in β-galactosidase activity (Fig. 4A). However, when interaction with Chs31–855 was assayed, the protein lacking the first 126 aa (Chs3126–855) afforded β-galactosidase levels similar to those of the negative control. From these results, we conclude that the self-association domain of Chs3 resides between its first 126 residues. However, these results did not demonstrate directly that the interaction occurred through the N-terminal region of Chs3. To confirm this, we expressed the first 171 amino acids of Chs3 tagged with 6xHIS or GST tags in Escherichia coli. In vitro the Chs31–171-GST protein fragment was able to recruit the Chs31–171-6xHIS fragment (Fig. 4B), but this recruitment did not occur when the truncated Chs3125–171-GST was used as the bait, similarly to what we observed in case of the C-terminal region of Chs3 (Chs3CT).
To show that this self-association of Chs3 also occurred in vivo, we expressed two versions of Chs3 in the same cell: one GFP-tagged and the other 3xHA-tagged. In event of self-association, it would be possible to co-immunoprecipitate GFP-tagged Chs3 with anti-HA antibodies. GFP-tagged Chs3 co-precipitated with Chs3-3xHA, confirming that Chs3 could at least form dimers, and – potentially – also higher-order structures (Fig. 4C). When we used the different N-terminal deletions of Chs3-3xHA we always observed reduced levels of Chs3-GFP co-immunoprecipitation (Fig. S9A), but the deletion of the first 126 amino acids of Chs3 (Δ126Chs3-3xHA) co-immunoprecipitated only trace amounts of Chs3-GFP (Fig. 4C, see also Fig. S9A). Taken together, our results are consistent with Chs3 oligomerization through its N-terminal region, although we cannot establish precisely the interaction region. However, the partial co-immunoprecipitation of Δ63Chs3 and Δ63–125Chs3, together with their limited resistance to calcofluor white suggested that this domain extends to both sides of position 63, and only the deletion of the whole region in Δ126Chs3 abrogates this interaction almost completely. In addition, oligomerization should occur early on Chs3 intracellular trafficking since co-immunoprecipitation was detected in different chsΔ mutants (see Fig. S9B), including chs7Δ (see Fig. S9C), where Chs3 is completely retained in the ER.
We next analysed the degree of oligomerization of Chs3 by using BN-PAGE. After digitonin solubilization of the samples, Chs3 ran as three discrete bands (Fig. 4D), suggesting that Chs3 associates at least in three different oligomerization states. However, Δ126Chs3 migrated mostly as a unique band around the size of the faster migrating band detected for the wild type, indicating that this truncated protein oligomerized poorly, in agreement with the IP data. In order to stabilize the different oligomeric states of Chs3, we later treated the digitonin-solubilized samples with the DSP cross-linker. After cross-linking, wild-type Chs3 ran as two discrete bands with an estimated MW of 284 and 137 KD, fully compatible with the dimeric and monomeric forms of the protein based on predicted sizes. Δ126Chs3 also ran as two bands showing higher mobility than wild-type ones, but the faster migrating was the most abundant, contrary what occurs for the wild type. The reasons for the higher mobility of Chs3 after DSP are unclear, but cross-linking very likely affects the binding properties of this protein to Coomassie, increasing its electrophoretic mobility (Wittig et al., 2010). These results strongly support Chs3 oligomerization, mostly in a dimeric form, and that Δ126Chs3 oligomerization is severely reduced in clear agreement with the direct involvement of the N-terminal domain of Chs3 in self-association. In addition, Chs3 showed partial dimerization in the chs7Δ mutant (Fig. S9D), confirming that Chs3 oligomerization occurred at the ER.
The N-terminal region of Chs3 is required for proper endocytic turnover
The results reported so far explain why less Δ126Chs3 reaches the PM, but they do not explain why we observed similar chitin levels in Δ126Chs3 and full-length Chs3 or the increased resistance to calcofluor white (Fig. 1B an 1C). These phenotypes are probably related to the properties of the Chs3 pool reaching the PM. Wild-type Chs3 is confined to the bud neck region at the PM and in an internal reservoir. In contrast, Δ126Chs3 spilled from the bud neck region and was also found throughout the PM (Fig. 5A). In addition, Δ126Chs3 is conspicuously absent from the internal reservoir. These phenotypes were reminiscent of those observed after treatment of wild-type cells with the actin polymerization inhibitor latrunculin A (latA) and in endocytosis mutants (Fig. 5A; Reyes et al., 2007; Zanolari et al., 2011), indicating that the internalization of Δ126Chs3 might occur less efficiently than that of wild-type Chs3. This delay in internalization was confirmed by subcellular gradient fractionation because more Δ126Chs3 co-migrated with the plasma membrane ATPase Pma1 than full-length Chs3 (Fig. 5B, compare lanes 1–3 in both gradients).
Chs4 is required for anchoring Chs3 to the bud neck and failure of this to occur leads to prompt endocytosis of Chs3 (Reyes et al., 2007; Sacristan et al., 2012; Fig. 5C). In contrast, Δ126Chs3 remained localized to the PM in the absence of Chs4 and was absent from intracellular structures, supporting the notion that this mutant protein is endocytosed inefficiently.
However, it remained to be tested whether the longer retention time of Δ126Chs3 at the PM could produce higher CSIII activity, as suggested by the in vivo phenotypes (Fig. 1B and C). First, we measured the total CSIII activity (Fig. 5D; with trypsin) and found no difference between the wild type and Δ126Chs3, indicating that the N-terminal-truncated protein is not more active than the full-length Chs3. Next, we examined the basal level of CSIII activity (Fig. 5D, w/o Trypsin), which reflects the amounts of CSIII confined to the PM (Reyes et al., 2007). Interestingly, under these conditions the activity of Δ126Chs3 was much higher, comparable to the CSIII activity seen in the endocytosis mutant end4Δ. Therefore, our data indicate that the defect in Δ126Chs3 endocytosis would be responsible for the similar chitin levels of cells expressing full-length or truncated Chs3. In agreement with this interpretation, chitin synthesis occurred at the bud neck but also throughout the yeast cell wall in Δ126Chs3 cells (Fig. 6B). However, the overall signal intensity of the calcofluor white staining was not higher in Δ126Chs3-expressing cells in comparison with the wild type.
Although the reduced endocytosis of Δ126Chs3 could be simply linked to its inefficient oligomerization, it is also conceivable that N-terminal deletion could remove specific motifs of Chs3 directly involved in its endocytic recycling. The deletion of the first 126 amino acids might affect the ubiquitination of Chs3, which has been shown to occur at least at K136 (Peng et al., 2003). To test this possibility, we mutated the four proximal lysines to arginines (K118, 119, 123, 136) and the localization of the quadruple mutant was assessed (Fig. 6A). Chs3RRRR-GFP construct was mostly localized correctly; the staining was confined to the bud neck region and to internal structures, similarly to what was observed for the wild type. Cells containing the Chs3RRRR were also slightly hypersensitive to calcofluor white and showed a patchy calcofluor white staining (Fig. 6B and C), phenotypes that could be associated with a higher retention time of Chs3RRRR at the PM. However, neither of these phenotypes was observed in the cells containing Δ126Chs3. Recently, it has been shown that Chs3 contains exomer and AP-1 binding motifs on its first 25 amino acids (Starr et al., 2012). We therefore addressed the localization of Δ126Chs3 in the chs5Δ mutant. The Δ126Chs3 protein was still significantly retained in the TGN reservoir (Fig. 6A), although Δ126Chs3 chs5Δ cells showed a slightly higher sensitivity to calcofluor white than chs5Δ (not shown). These results suggested that Δ126Chs3 still depends on exomer for trafficking but that part of the protein can reach the PM from an alternative route, in agreement with recently reported results (Starr et al., 2012). Thus, together the results obtained indicate that the removal of the known specific motifs in Δ126Chs3 cannot alone account for the reduction in endocytosis of the N-terminal-truncated proteins and their distribution along the PM.
Despite the progress made in recent years regarding the proteins involved in the intricate trafficking behaviour of the chitin synthase Chs3, very little is known about the regions of Chs3 required for proper transport and localization. Here we revisited this issue using an improved topology model for the multi-transmembrane-spanning Chs3. We identified the cytoplasmic N-terminal region I of Chs3 as being important for correct intracellular localization. This region contains a self-interacting motif, which is required both for efficient export from the ER and for endocytosis at the PM. Delayed endocytosis results in higher CSIII activity at the PM, and the spreading of the enzyme and the synthesized chitin along the whole-cell surface. Part of these phenotypes could be linked to the absence of an AP-1 retention motif (DEESLL) described recently (Starr et al., 2012), which would allow the truncated Δ126Chs3 protein to reach the PM not only through the exomer pathway but also through the alternative exocytic pathway (Valdivia et al., 2002). However, deletion of the AP-1 pathway did not abrogate Chs3 accumulation at the TGN reservoir as occurred for the Δ126Chs3 protein, suggesting that the reduced endocytosis of this protein could be linked, at least partially, to the absence of the self-interacting domain located within the first 126 amino acids of the protein.
Interestingly, the absence of this self-interacting domain did not compromise CSIII activity, thus the oligomerization of Chs3 may not be a pre-requisite for Chs3 activation and function at the PM. Our data suggest, however, that the way chitin is deposited and anchored would be affected by the inability of Chs3 N-terminal truncations to oligomerize. Despite the relatively similar chitin synthesis levels of the oligomeric and monomeric forms of Chs3 (the N-terminal truncations), the expression of the monomeric forms led to calcofluor white resistance, as though the calcofluor white were unable to intercalate with the chitin properly (Roncero and Duran, 1985) and thus poison the cell. A role for oligomerization of Chs3 in chitin synthesis levels could perhaps be expected, since the proper synthesis of cellulose – another β(1–4) polymer with a crystalline structure – depends on cellulose synthase oligomerization at the PM of plant cells (Atanassov et al., 2009). Similarly, insect chitin synthase has been proposed to form oligomers (Maue et al., 2009), therefore it would be tempting to speculate that synthase oligomerization might be a general prerequisite for the proper synthesis of biological polymers with a crystalline structure.
The N-terminal truncations of Chs3 were enriched in the ER at steady state; however, they seem to be correctly folded and properly interact with Erv14. The identification of Erv14 as the ER exit receptor of Chs3 is consistent with recent reports suggesting that Erv14 would be the major facilitator for the ER export of membrane-spanning proteins of the late secretory pathway or PM (Castillon et al., 2009; Herzig et al., 2012), including those dependent on ER chaperones, whose mechanism of exit remains poorly understood (Dancourt and Barlowe, 2010). Moreover, our results would be compatible with Erv14 interacting with the TM regions of the cargo (Herzig et al., 2012), since these regions are probably unaffected in the truncated proteins. The function of Erv14 would be also compatible with the role of Chs7 in the exit of Chs3 from the ER. Like Gsf2, Pho86 and Shr3, Chs7 is a member of the family of ER-specific chaperones that prevent improper protein folding at the ER. It has been shown that Shr3 binds to the N-terminal five transmembrane segments of Gap1, also to allow proper folding of the remaining seven TMDs (Kota et al., 2007). In analogy to Shr3, Chs7 would be able to interact normally with the TMDs of wild-type or N-terminal truncated Chs3, allowing its exit from the ER. The physical interaction of Chs3 with Erv14 still occurred in the absence of Chs7 (data not shown), suggesting that the aggregation of Chs3 at the ER would affect its loading into COPII vesicles, hence providing new clues concerning the role of ER chaperones such as Chs7. In agreement with this, the Δ194Chs3 mutant protein was retained in the ER under all conditions tested, probably because of its high level of aggregation at the ER (Fig. 2A). Interestingly, Chs3 was still able to oligomerize in the absence of Chs7 (Fig. S9C and D), reinforcing the general idea that oligomerization and Chs7 contribute independently to the exit of Chs3 from the ER.
Although our results cannot completely exclude any other defects of the truncated proteins in their exit from the ER, we demonstrate that all the truncations, with the exception of Δ194Chs3, did leave the ER efficiently to be sent back to the ER from the Golgi in a COPI-dependent manner, indicating that the truncations must be recognized by a post-ER quality control pathway. The retrieval was also dependent on the Golgi-ER-sorting receptor Rer1, in agreement with the notion of Rer1 interacting more efficiently with the N-terminal-truncated Chs3 versions (see Fig. S5D). It has been proposed that Rer1 recognizes transmembrane domains in different modes (review in Dancourt and Barlowe, 2010). According to this, it is possible that the monomeric form of Chs3 exposes some residues in the structure of a hydrophobic helix, otherwise hidden within the dimer. In addition, wild-type protein also seems to conceal the C-terminal di-lysine motif required for COPI transport, suggesting that the signals required for retrograde transport would only be exposed in the oligomerization-compromised versions of Chs3, promoting the sorting of these proteins into COPI vesicles destined to the ER. Interestingly, the elimination of the di-lysine motif seems to be the most effective way of preventing Δ126Chs3 accumulation, suggesting that it would act as the critical determinant for the retrieval of Chs3 in COPI vesicles. A similar mechanism of concealing signals also acts on the transport of the human immunoglobulin FcεRI receptor (Letourneur et al., 1995), but this has not been reported previously for yeast polytopic proteins that depend on specific ER chaperones. The joint dependence on Rer1 and di-lysines motifs of Chs3 retrograde transport favours the proposed hypothesis of Rer1 acting as a Golgi chaperone (Sato et al., 2003). Taken together, our data support a model in which the oligomerization of Chs3 is monitored intracellularly and that defects in oligomer formation make Chs3 a substrate of a Golgi-associated quality control (GAQC) system based on the retrograde transport chaperone Rer1 and COPI vesicle transport. Whether other cargos that are dependent on transport chaperones for ER exit are also subject to GAQC remains to be elucidated.
In conclusion, although not essential for chitin synthase activity the oligomerization of Chs3 is carefully monitored intracellularly. This surveillance is based on the recognition of protein domains hindered or exposed by the oligomerization, allowing rerouting of the incorrectly dimerized proteins along their intracellular transport.
Strains and growth conditions
The characterization of Chs3 was performed using several S. cerevisiae strains, most of which were created in the W303 background. These included the strains CRM103 (W303 MATa chs3::URA3), CRM892 (W303 chs4::HIS3 chs3::URA3), CRM1348 (W303 chs3::URA3 chs7::natMx4) and MMY1012 (W303 rer1::KanMx4 chs3::LEU2). For some experiments, additional genetic backgrounds were used, such as those of strains CRM921 (BY4741 end4::kanMx4), CRM926 (BY4741 end4::KanMx4 chs3::URA3), MMY981 (MATa ret1-1 chs3::LEU2) and RH7016 (erv14-mCi::SpHIS5 ura3 leu2 his3 trp1) (Castillon et al., 2009).
All strains used contained unique copy of CHS3 and were made by transforming the appropriate chs3Δ strains with the different constructs of Chs3, expressed from their own promoter in centromeric plasmids. Where required, CHS7 was additionally overexpressed from pRS423::CHS7 or pRS424::CHS7, which were introduced into the strains expressing the different constructs of CHS3.
Yeast cells were typically grown at 28°C, except for the end4Δ and ret1-1 mutants, which were grown at 25°C. YEPD (1% Bacto yeast extract, 2% peptone, 2% glucose) or SD medium (2% glucose, 0.7% Difco yeast nitrogen base without amino acids) supplemented with the appropriate amino acids were used as growth medium. When required, the media were supplemented with hygromycin, nourseothricin or geneticin at the standard concentrations.
Calcofluor white resistance was tested by a plate assay on SD media buffered with 50 mM potassium hydrogen phthalate, pH 6.2, as described (Trilla et al., 1999).
The plasmids used in this work are listed in supplementary Table S1. All CHS3 mutants were generated by site-directed mutagenesis, as described (Trilla et al., 1999), using plasmid pRS315::CHS3-GFP as template. The N-terminal truncations were generated using long synthetic oligonucleotides, which created the deletion loops upon hybridization with the template. All constructs were confirmed by sequencing. The HA-tagged and untagged versions of the truncated forms of CHS3 were constructed by replacing the N-terminal regions of pRS315::CHS3-3xHA and pRS315::CHS3 with the corresponding regions from the truncations using the SpeI site from the polylinker and the internal HindIII site of CHS3. In all cases the tagged proteins behaved identically that their untagged versions, as was previously shown fro the original constructs (Trilla et al., 1999).
CHS7 was cloned as an EcoRI–SacI fragment from pRS426::CHS7 (Trilla et al., 1999) into the pRS423 and pRS424 plasmids.
Plasmids expressing the different N-terminal fragments of Chs3 were generated by amplifying the corresponding fragments by PCR and cloning them into the vector pETGEXCT (Sharrocks, 1994) using NcoI/SacI restriction sites to give pETGEXCT-CHS31–171 and pETGEXCT-CHS3126–171. The C-terminal tail comprised the last 55 amino acids of CHS3 and was cloned between the EcoRI and XhoI restriction sites of pGEX-6P-1.
Protein extracts and immunoblotting
Total-cell lysates from a 30 ml logarithmic culture were prepared by glass-bead lysis (Fast prep FP120, BIO101) in 150 μl of lysis buffer (50 mM Tris-HCl, pH 8, 0.1% Triton, 150 mM NaCl) with 1× protease inhibition cocktail (1 mM PMSF, 1 μg ml−1 aprotinin, 1 μg ml−1 leupeptin). Cell debris was removed by centrifugation (5 min, 1000 g, 4°C) and the supernatant was boiled for 5 min with 4× SDS loading buffer. One hundred micrograms of protein was subjected to 7.5% SDS-PAGE separation and transferred to PVDF membranes (Trilla et al., 1999). After blocking with 5% skim milk, the membranes were incubated with anti-HA 12CA5 (Roche) or anti-GFP JL-8 monoclonal antibodies (Living colours, Clontech). Rabbit polyclonal antibody against Pma1p (Serrano et al., 1986) was used as a marker for PM fractions in subcellular fractionation experiments. Other blots were developed with anti-Gas1 (Nuoffer et al., 1991) or anti-6xHIS (H-15, Santa Cruz) antibodies. Blots were developed using the ECL kit (GE Healthcare).
Subcellular fractionations were performed on discontinuous sucrose gradients as previously described (Reyes et al., 2007).
The direct interaction between Chs3 molecules was determined by co-immunoprecipitation as follows: cells from 100 ml of logarithmically growing cultures were harvested by centrifugation and washed with cold water. Cells were lysed as described above in 200 μl of lysis buffer (50 mM Tris-HCl, pH 8.0, 0.1% Triton, 150 mM NaCl) containing 1× protease inhibitor cocktail (1 mM PMSF, 1 μg ml−1 aprotinin, 1 μg ml−1 leupeptin), and the lysates were cleared of cell debris (5 min, 16 000 g, 4°C). Two milligrams of total protein from cell lysates was brought up to 400 μl with IP buffer (50 mM Tris-HCl, pH 8.0, 0.1% Triton, 150 mM NaCl, 2 mg ml−1 BSA). This suspension was incubated with anti-HA HA1.1 (Covance) at 1:75 dilution at 4°C for 2 h, and then mixed with 50 μl of 0.1 mg ml−1 of Protein A Sepharose (GE Healthcare) and incubated for a further 2 h. Alternatively, the polyclonal α-GFP (Invitrogen) was used for co-immunoprecipitation at 1:100 dilution. The beads were washed three times with lysis buffer and boiled with 4× SDS loading buffer for 5 min. Twenty microlitres of the samples was separated by 7.5% SDS-PAGE, and the proteins were visualized by immunoblotting, as indicated above.
The interaction of Chs3-3xHA with Erv14-mCi or Rer1-GFP (Sato et al., 2001) at the ER was tested in co-immunoprecipitation experiments as follows: cells containing the indicated constructs were lysed in 800 μl of TNE (50 mM Tris-HCL pH 7.5, 150 mM NaCl, 5 mM EDTA) supplemented with 1× protease inhibitor cocktail. Lysates were then centrifuged at 1000 g for 5 min at 4°C, followed by a second centrifugation at 13 000 g for 15 min at 4°C. The ER fraction pelleted was resuspended in 1 ml of TNE and membranes were solubilized by adding 225 μl of 5% TNED (TNE with 5% of digitonin) and incubating for 1 h at 4°C. Non-solubilized membranes were removed by ultracentrifugation at 100 000 g for 1 h at 4°C. One hundred microlitres of GFP-Trap (Chromtek) beads at 20% in TNE was added to 900 μl of the supernatant and incubated for 3 h at 4°C. The beads were washed twice with 1000 μl of TNED 1% and a further two times with 0.2% TNED. Finally, the beads were dried and bound proteins were eluted in 50 μl of sample buffer and analysed by immunoblotting.
Protein purification and pull-down assays
All the protein fragments used for the pull-down assays were expressed in Rossetta E. coli cells by the addition of 0.5 mM IPTG and growth at 37°C for 4 h. Cultures were washed once with ice-cold water and stored at −80°C. The N-terminus domain of Chs3 tagged with 6xHis was expressed from pCR-T7/Ct-TOPO-CHS31–171-6xHis (Trautwein et al., 2006) and purified under native conditions using Ni-NTA resin (Qiagen). The final elution step was done with 250 mM Imidazol. GST-tagged fragments were purified from E. coli cells lysed in PBS/5% Glycerol. Extracts were then incubated with GSH agarose (GE Healthcare) and eluted with 40 mM GSH. Purified proteins were later dialysed against PBS/5% Glycerol for the subsequent pull-down assay, which was carried out as follows. GST fusions were immobilized on GSH-agarose and incubated for 2 h at 4°C with purified Chs3-6xHis in 20 mM HEPES pH 6.8 and 150 mM NaCl. Finally, the coupled resin was washed three times with 20 mM HEPES pH 6.8 and 150 mM NaCl, and then resuspended in 100 μl SDS sample buffer, followed by incubation at 95°C for 5 min. Samples were analysed by Coomassie staining and Western blot.
Endoglycosidase H treatment
EndoH treatment was performed as previously described (Cos et al., 1998). Briefly, 20 ml of exponentially growing cells was processed as described above for protein immunoprecipitation with anti-HA. After incubation of the protein extracts with Protein A Sepharose, the beads were washed twice with 50 mM citrate buffer, pH 5.5, and were resuspended in 200 μl of 50 mM citrate buffer, pH 5.5. One hundred microlitres of this suspension was boiled with 4× SDS loading buffer and the remaining 100 μl was treated overnight at 37°C with 10 mU EndoH (Boehringer). The beads were washed twice with lysis buffer and boiled in 50 μl of 4× SDS loading buffer, and 20 μl of the samples was loaded on a 6.5% SDS-PAGE gel and subjected to immunoblot analysis.
Cross-linking was carried out according to Kota (Kota and Ljungdahl, 2005). Twenty millilitres of culture was harvested, washed in cold water, and resuspended in 150 μl of PBS, pH 7.4. After breakage of the cells with glass beads and the removal of cell debris by centrifugation (5 min, 16 000 g, 4°C), 10 μg of protein was treated with DSP cross-linker (Pierce) at concentrations ranging from 0.01 to 0.1 mM for 30 min at 22°C. Samples were treated with 30 mM Tris-HCl, pH 7.5, for 30 min at 22°C to neutralize free reactive groups. Control samples were additionally treated with 40 mM DTT for 30 min at 37°C. Finally, all samples were boiled for 5 min with loading buffer prepared without β-mercaptoethanol. Fifteen microlitres of the samples was separated on a 6.5% SDS-PAGE gel.
Blue native gel electrophoresis
Cells from a 50 ml culture were resuspended in 500 μl of lysis buffer (50 mM Tris-HCl, pH 7.5, 5 mM EDTA) and were lysed with glass beads. Cell debris was removed by centrifugation (10 min at 3000 g, 4°C) and membranes were pelleted by high-speed centrifugation (21 000 g, 20 min 4°C). The supernatant was removed and the membranes were resuspended in 100 μl of 15% glycerol. Ten micrograms of protein was solubilized in native PAGE 4× Sample buffer (Invitrogen) with 1% Digitonin and samples were incubated for 30 min at 4°C. Samples were centrifuged at 15 000 g for 5 min at 4°C to remove non-solubilized membranes. Some samples were later treated with DSP 2 mM for 30 min at 22°C. The supernatants were supplemented with 0.25% Coomassie G-250 (Invitrogen) and loaded on 3–12% NativePAGE from Invitrogen. Proteins were transferred to a PVDF membrane and visualized as described above for SDS-PAGE.
Yeast two-hybrid analysis
The CHS4 coding region and the indicated truncations of CHS3 were amplified by PCR and cloned in-frame into the pAS2 and the pACT2 plasmids (Harper et al., 1993) using the NcoI and SmaI restriction sites. The pACT2 and pAS2 constructs were transformed into strain Y190 and the interaction was assessed by the determination of β-galactosidase levels. Qualitative determination of β-galactosidase activities was carried out by transferring colonies onto a nitrocellulose filter, and lysing them by freezing in liquid nitrogen. The filters were then placed on a plate containing Z-buffer and X-Gal to analyse enzyme activity by the appearance of blue colour. Quantitative measurements of β-galactosidase activity were performed directly in cell extracts using the substrate o-nitrophenyl-β-d-galactopyranoside (ONPG) in Z buffer, as described (Rose et al., 1990).
Imaging was carried out using a Leica RX150 epifluorescence microscope equipped with an Orca-ER digital camera. Images were processed using the ImageJ software (NIH) and mounted with Adobe Photoshop CS. All images in the series were acquired under identical conditions and processed in parallel to maintain relative intensities.
For chitin staining, cells were grown in the presence of 50 μg ml−1 for 2 h. Yeast cells expressing GFP-tagged proteins were grown to early logarithmic phase in SD supplemented with 0.2% adenine and observed directly under the microscope. Where indicated, cells where incubated with the actin-depolymerizing drug Latrunculin A (15 μM) for at least 30 min. The vacuole was labelled by incubating the cells in the presence of FM4-64 (30 μM) for 1 h.
Chitin content and chitin synthase activity
Chitin levels were measured as described (Trilla et al., 1999). In brief, chitin was digested with chitinase (Sigma-Aldrich) and the N-acetyl-d-glucosamine (GlcNAc) released from the reaction was determined colorimetrically.
For the determination of chitin synthase activity, cell membranes were prepared as described (Choi and Cabib, 1994). The enzymatic assay was performed with (total activity) and without (basal activity) trypsin treatment and in the presence of Ni2+ and Co2+, allowing the specific measurement of CSIII activity. Activity was expressed as mU h−1 mg−1 protein.
We thank R. Valle and V. Casquero for technical assistance, and M. Trautwein for his generous help with the MS analysis. C.S. is grateful to P.O. Ljungdahl for materials and helpful discussions. We acknowledge support of C.S. by a FPU fellowship from the MEC. A.R. was supported by a EMBO short-term fellowship and by the JCyL Grant GR231, and J.M.-L. by an University of Seville fellowship. Work in the Spang Lab was supported by the University of Basel and the Swiss National Science Foundation. This research was supported by the Spanish CICYT Grants BIO2007-60779 and BFU2010-18632 to C.R. and BFU2008-04119 and BFU2011-24513 to M.M., and by the Junta de Andalucia Grant P09-CVI-4503 to M.M.