Rab6 is required for rapid, cisternal-specific, intra-Golgi cargo transport

Rab6, the most abundant Golgi associated small GTPase, consists of 2 equally common isoforms, Rab6A and Rab6A′, that differ in 3 amino acids and localize to trans Golgi cisternae. The two isoforms are largely redundant in function and hence are often referred to generically as Rab6. Rab6 loss-of-function inhibits retrograde Golgi trafficking, induces an increase in Golgi cisternal number in HeLa cells and delays the cell surface appearance of the anterograde cargo protein, VSVG. We hypothesized that these effects are linked and might be explained by a cisternal-specific delay in cargo transport. In pulse chase experiments using a deconvolved, confocal line scanning approach to score the distribution of the tsO45 mutant of VSVG protein in Rab6 depleted cells, we found that anterograde transport at 32 °C, permissive conditions, through the Golgi apparatus was locally delayed, almost tenfold, between medial and trans Golgi cisterna. Cis to medial transport was nearly normal as was trans Golgi to TGN transport. TGN exit was unaffected by Rab6 depletion. These effects were the same with either of two siRNAs. Similar intra-Golgi transport delays were seen at 37 °C with RUSH VSVG or a RUSH GPI-anchored construct using a biotin pulse to release the marker proteins from the ER. Using 3D-SIM, a super resolution approach, we found that RUSH VSVG transport was delayed pre-trans Golgi. These visual approaches suggest a selective slowing of anterograde transport relative to 3 different marker proteins downstream of the trans Golgi. Using a biochemical approach, we found that the onset of VSVG endoglycosidase H resistance in Rab6 depleted cells was delayed. Depletion of neither Rab6A or Rab6A′ isoforms alone had any effect on anterograde transport through the Golgi suggesting that Rab6A and Rab6A′ act coordinately. Delayed cargo transport conditions correlate strongly with a proliferation of Golgi cisternae observed in earlier electron microscopy. Our results strongly indicate that Rab6 is selectively required for rapid anterograde transport from the medial to trans Golgi. We suggest that the observed correlation with localized cisternal proliferation fits best with a cisternal progression model of Golgi function.


Results
Rab6 is required for the rapid, intra-Golgi transport of tsO45G protein. Rab6 loss-of-function results in a post-ER exit delay in anterograde cargo transport, e.g., VSV-G protein, from the ER to the cell surface in which the initial steps of ER to Golgi transport appear to be normal 17,18 . Based on the literature, much of the delay could be due to post-Golgi effects; others have shown previously that Rab6 regulates post-Golgi vesicle transport through an effect on kinesin recruitment 19 . To test whether there is a significant intra-Golgi delay in cargo transport in Rab6-depleted cells, we investigated initially the qualitative effects of Rab6 knockdown on transit for a model cargo protein, i.e., tsO45-GFP, through the Golgi apparatus. In these experiments, we used siRNA directed against coding sequences shared between Rab6A and Rab6A′ and hence both isoforms were depleted equally. Temperature was used to control tsO45G-GFP release from the ER with the protein being ER retained at 39.5 °C and released from the ER at 32 °C. As shown in Fig. 1 (ortho slice view of confocal image stack, see other Figures for XY view), accumulation of tsO45G-GFP in p115-, a cis Golgi marker, positive structures were indistinguishable for Control and Rab6 siRNA treated HeLa cells; after a 20-min time chase, tsO45G-GFP co-localized with cis-Golgi marker p115 in both control and Rab6 KD cells. At late chase times (40 min), however, tsO45G-GFP in Control cells showed no correspondence with the Golgi exit marker, TGN46, while tsO45G-GFP in Rab6-depleted cells showed extensive correspondence with the TGN marker. We conclude that there is a significant intra-Golgi delay in cargo transport when Rab6 is depleted.
Validation of a confocal line scan approach to mapping intra-Golgi cargo protein distribution. We used a confocal line scan approach 22 to determine quantitatively the distribution of Golgi marker proteins relative to the model cargo protein, tsO45G-GFP. Work in this laboratory suggests that Rab6 plays a major role in Golgi cisternal organization. At the electron microscopy level, when Rab6 is depleted in HeLa cells, several Golgi stacks combine together to form a continuous, longer Golgi ribbon with additional cisternae and one or two dilated cisternae at trans side of the Golgi stack 18,21 (Fig. S-1b, present work). However, when cells were gently fixed with paraformaldehyde as done for immunofluorescence analysis and then high pressure frozen, we found that the Golgi cisternae in both control and Rab6 depleted cells were dilated ( Fig. S-1). Quantitatively, the distance across the Golgi stack was expanded by a factor of 2-3, a factor sufficient to make the quantitative analysis of distribution of a cargo marker, tsO45G-GFP, relative to resident Golgi proteins resolvable by deconvolved, spinning disk confocal microscopy as shown below.
HeLa cells were fixed at various chase times and imaged in three color channels with two being Golgi markers. Image stacks were taken at near full cell depth, deconvolved to further sharpen the images, and then analyzed by plotting the intensity distribution of each channel along lines drawn perpendicular to the Golgi ribbon in Scientific RepoRtS | (2020) 10:16604 | https://doi.org/10.1038/s41598-020-73276-w www.nature.com/scientificreports/ areas of maximal separation between the Golgi markers, i.e., optimal orientation of the ribbon in 3D space (for details, see Methods). In Fig. 2a,b, this approach is illustrated for HeLa cell labeled for two trans Golgi markers, GalT and SialylT, and the TGN marker, TGN46. In the raw Golgi images, the distributions of GalT (green) and SialylT (red) are highly similar and in the merged XY image the two overlap (yellow color). XY position data were then converted to distance in microns to determine the physical distance between markers for Control and Rab6 knockdown cells ( Fig. 2c-e). In total, we used 5 reference markers, p115 (cis), NAGT-1 (medial, myctagged stable transfectant 23 ), GalT/SialylT (trans) and TGN46 (TGN). The position of the cis marker, p115, was arbitrarily set to "0". The measured distance across the Golgi apparatus, cis to TGN, was about 0.6 µm. This distance is the same as that reported by Dejgaard et al. 22 and approximately two to three times that determined by electron microscopy 18 . We attribute this greater distance to the point spread function of light microscopy and to cisternal dilation due to the comparatively weak fixation methods used for immunofluorescence microscopy (Supplemental Fig. 1). We note that the final endogenous protein mapped, Rab6, behaves as a trans Golgi protein (Supplemental Fig. 2). At the resolution of deconvolved, spinning disk, confocal microscopy, the distribution of the five reference markers was unaltered by Rab6 depletion (Fig. 2e).
To validate further the application of a confocal line scan approach to analyzing the intra-Golgi transport of tsO45G-GFP, we compared quantitatively the distribution of tsO4G-GFP to that of p115 and TGN46 at 0 time and 20 min chase, 32 °C, in Control and Rab6-depleted cells. As shown in Fig. 3, at 0 time, tsO45G-GFP protein displayed a web-like cytoplasmic distribution and rim-like staining of the nuclear envelope in both Control and Rab6-depleted cells, exactly the staining distribution expected for ER localization. Quantitatively, the reference Golgi proteins were well separated from one another and there was a time-independent tendency for the tsO45G-GFP distribution to peak with either of the Golgi markers. In striking contrast at 20 min chase, the relative distribution of the two cis and TGN markers was unaltered while that of tsO45G-GFP now corresponded almost exactly with that of p115 in both Control and Rab6-depleted cells. We conclude that the confocal line scan accurately reflects the distribution of tsO45G-GFP protein within the HeLa cell Golgi apparatus and that Rab6-depletion has little to no effect on initial steps in anterograde cargo transport. tsO45G-GFP transport stalls between the medial-and trans-Golgi in Rab6 depleted cells. Next, we tested the hypothesis that Rab6 is preferentially required for anterograde cargo transport between the medial-Golgi and TGN. We reasoned that Rab6 as a trans Golgi Rab protein would when depleted produce phenotypic, cargo transport effects downstream of the normal trans localization of Rab6 within the Golgi apparatus, i.e., a downstream traffic jam. As shown in Fig. 4a versus e, Rab6 depletion had no effect on tsO45G-GFP transport   In Control cells, tsO45G-GFP co-profiled with TGN46 at the end of a 40-min chase while in Rab6 depleted cells there was little, if any, accumulation of the cargo protein at the TGN. In fact, transport from the medial-Golgi to TGN was very slow initially ( Fig. 4i) and full co-profiling of tsO45G-GFP with the trans markers, GalT and SialylT required more than a 45-min chase (Fig. 5). As a final, line scan, kinetic determination, we compared the TGN exit kinetics for tsO45G-GFP in Control and Rab6depleted cells. As shown in Fig. 5g and Table 1, we found that the kinetics were virtually identical, ≤ 6 min, after allowing for a 20-min delay in cargo accumulation in the TGN with Rab6-depletion. In sum, as summarized in Fig. 5f and Table 1, there was a significant, cisternal-specific delay in cargo transport between the medial-and trans-Golgi in Rab6-depleted cells. Similar results were found with two HeLa cell variants ( Fig. S-3) and two dif- Overall, the transport of tsO45G from the ER to the cis-Golgi or medial-Golgi was the same with Rab6 depletion. However, transport of tsO45G was slowed significantly during exit from the medial-Golgi compartment and in its transport to the TGN compartment in Rab6 KD cells, an indication that Rab6 is essential for proper intra-Golgi cargo transport at the medial-to-trans Golgi interface. Software used includes iVision 4.  . S-4). We also performed a box/whisker graphical depiction of a data set ( Fig. S-5). This methodology provides an alternate illustration of a time specific inhibition of cargo transport.
To provide a non-imaging approach, we assayed for effect of Rab6 depletion on the conversion of tsO45G protein to an endoglycosidase H (EndoH) resistant form. This N-linked oligosaccharide processing step occurs in the medial-Golgi apparatus with mannosidase II removal of the last 2 mannose residues outside the N-glycan core. As shown in Fig. 6, at a 20-min chase time, tsO45G protein was fully sensitive to EndoH digestion in both Control and Rab6 knockdown HeLa cells. Over a 35 min or a 60-min chase, the protein in the case of Control  Table 1 and Fig. 1. Table 1. Effect of Rab6 knockdown on the rate of tsO45G-GFP transport through the Golgi apparatus (32 °C). *p < 0.001. Assignments of Golgi regions were based on the quantitative mapping shown in Fig. 2E. Rates were calculated from Fig. 5G, H data. Based upon the line scan of the micrographs, fluorescence peaks were assigned distance value. Rates were then then determined from the slope of the respective line segment in Fig. 5H and expressed as rates in Table 1 www.nature.com/scientificreports/ cells became progressively resistant to EndoH digestion. However, in Rab6 depleted cells, tsO45G processing was much slower with a small fraction of the protein resistant at a 35 min chase and only ~ half of the protein resistant at 60 min. Even at a 120 min chase time, when in essence all the protein is cell surface, there was still a significant portion of EndoH sensitive tsO45G protein in the Rab depleted cells (Fig. 6b, c). In separate experiments and consistent with these results, siRab6 knockdown HeLa cells displayed a significant increase in GSII cell surface lectin staining (Fig. S-6). GSII as a lectin binds to exposed N-acetylglucosamine residues.

Rab6 depletion inhibits intra-Golgi transport of VSVG-GFP and GPI anchored proteins at 37 °C.
To further validate the Rab6 requirement for rapid intra-Golgi transport, we used RUSH constructs expressing VSVG-GFP or GPI-GFP to test the transport of cargo proteins at 37 °C. As shown in Fig. 7a, before treating with biotin, VSVG-GFP was distributed in the ER in both control and Rab6 knockdown cells. After addition of biotin, VSVG-GFP was released from the ER resident hook protein and started to move from ER to Golgi apparatus and plasma membrane 22 . The kinetics of VSVG-GFP transport from ER to the juxtanuclear Golgi apparatus was rapid while the transport of VSVG-GFP from Golgi apparatus to plasma membrane was significantly delayed in Rab6 depleted cells. The clearance of a second RUSH cargo protein, GFP-GPI from Golgi apparatus to plasma membrane was also slower in Rab6 depleted cells than control cells (Fig. 7b). The above results were confirmed using super-resolution, 3D-SIM microscopy. At 0 min, RUSH VSVG-GFP was accumulated in the ER in both control cells and Rab6 knockdown cells. At the end of 40-min chase, surface accumulation of VSVG-GFP was observed in control cells, while in Rab6 knockdown cells, VSVG-GFP was still distributed in the Golgi apparatus and showed at this time point little to no overlap with the distribution of GalT (red), indicative of RUSH VSVG-GFP transport that was stalled before trans-Golgi (Fig. 8). These results provide additional evidence for a cisternal-specific involvement of Rab6 in rapid intra-Golgi transport.
Rab6A and Rab6A′ collectively support Intra-Golgi anterograde cargo transport. Previous work in our laboratory showed that individual depletion of Rab6A or Rab6A′ by selective siRNA treatment produced Golgi cisternal elongation and vesicle accumulation, albeit at half the level of a Rab6A/A′ knockdown and significantly with little-to-no increase in Golgi cisternae number 21 . Based on these results, we hypothesized that the Rab6A or Rab6A′ individually would have little to no effect on RUSH VSVG-GFP or GFP-GPI transport. As shown in Fig. 9a, the transport of VSVG-GFP from ER to the Golgi apparatus and plasma membrane was little changed with Rab6A or Rab6A′ knockdown. At 40 min, VSVG-GFP was primarily in post-Golgi vesicles. However, with co-depletion of Rab6A and Rab6A′, the intra-Golgi cargo transport was delayed. As shown in Fig. 9b, at 0 min, GFP-GPI was accumulated in the ER. At the end of a 40-min or 60-min chase, Golgi and significant surface accumulation of GFP-GPI was observed in Control cells, while GFP-GPI retention in the Golgi apparatus of Rab6A and Rab6A′ co-depleted cells was high. These results indicate that both Rab6A and Rab6A′ contribute to regulate intra-Golgi transport. Whether that contribution is by distinct individual roles or a result of mass action requires further experimentation.

Discussion
Important roles for Rab GTPases in Golgi apparatus structure/function have been apparent for almost 3 decades with Rab6, the most abundant mammalian Golgi associated Rab protein 5 , being a particularly striking and challenging example. Rab6 has multiple effectors, at least 15; multiple roles in Golgi associated membrane trafficking; a role in Golgi associated Rab cascades; and is a determinant of Golgi structure 2-4 . A ying/yang relationship between gain-of-function and loss-of-function phenotypes make Rab6 a rare case among the roughly 20 Golgi associated Rab proteins. Overexpression of Rab6 redistributes Golgi resident proteins into the ER 13,14 while Rab6 depletion results in delayed anterograde transport and stabilization of Golgi structure with the generation of extended Golgi cisternal stacks and increased cisternal number 17,18,20 . Bhave et al. 24 have suggested that increased cisternal persistence could generate increased cisternal number. However, whether Rab6 regulates anterograde cargo transport through the Golgi apparatus in a local, cisternal-specific manner consistent with increased cisternal persistence, i.e., traffic jam consequences, was unknown before our studies. Our analysis demonstrated that the intra-Golgi transport of both tsO45G-GFP (32 °C) and RUSH VSVG-GFP or RUSH GFP-GPI (37 °C) were delayed within the Golgi apparatus. Under permissive conditions, all were transported normally from ER to Golgi apparatus. Detailed line scan analysis of deconvolved, confocal images containing cis, medial, trans and TGN markers indicated normal kinetics for tsO45G-GFP transport from cis to medial Golgi and normal TGN exit kinetics. Strikingly, medial Golgi exit kinetics were slowed more than tenfold. Similar line scan results were obtained using 2 different validated Rab6 siRNAs. siRNAs directed specifically against Rab6A or Rab6A′ had no effect on cargo transport despite substantial protein knockdown. Under these conditions, there is no increase in Golgi cisternal number 21 . Biochemical analyses revealed altered functionality as indicated by time-correlated acquisition of EndoH resistance of tsO45G. Furthermore, by 3D-SIM analysis, a super resolution technique, the delay in RUSH VSVG-GFP transport was found to be pre-trans Golgi apparatus. In brief, multiple analytical approaches demonstrate that Rab6 is required for rapid, intra-Golgi, cargo transport between the medial and trans-Golgi cisternae, a step immediately downstream of the normal trans-Golgi localization of the protein and the site of Rab6-dependent vesicle accumulation 18 .
These results support a model for Rab6(A/A′) functionality in which anterograde cargo trafficking phenotype is separated spatially from the normal intra-Golgi apparatus distribution of the Rab6A/A′ proteins, the two very closely related Rab6 isoforms. Multiple lines of evidence show Rab6 to be a trans-Golgi protein 8,22 (present work). Here, by line scan, deconvolved confocal, immunofluorescence analysis, we demonstrated Rab6 to coincide with the two trans-Golgi marker proteins, GalT and SialylT. As might be expect, Rab6-dependent, Golgi vesicle accumulation is also trans-Golgi specific 18 Table 1 and Fig. 1.
Scientific RepoRtS | (2020) 10:16604 | https://doi.org/10.1038/s41598-020-73276-w www.nature.com/scientificreports/ at the medial Golgi by shuttling back and forth between the medial and trans Golgi with a set of rate constants in which the Rab appears to be a trans Golgi resident protein.
In support of this model, one might take as an example mannose 6-phosphate receptors which bind ligand within the Golgi apparatus but appear at steady-state to be late endosomal/multivesicular body resident proteins. In a shuttle model, one is still left with the problem of explaining the apparent strong linkage between anterograde transport rate and Golgi cisternal number. We favor a cisternal persistence model. Rab43 provides another recent example of a Rab protein that affects intra-Golgi anterograde cargo transport events at the medial Golgi level 25,26 . In this case, the primary effect is proposed to be upon the routing of cargo proteins to different subregions within a Golgi cisternal compartment. Here routing was shown to be dependent on the sequence of the cytoplasmic domain of the transmembrane cargo protein. In brief, we give weight to the increased cisternal number observed in Rab6 knockdown HeLa cells as a basis for favoring a cisternal maturation/progression model versus another Golgi models [27][28][29][30][31][32][33] .
In conclusion, we present evidence that Rab6, a trans-Golgi protein, is a key determinant of intra-Golgi, anterograde cargo transport rate at the level of medial-to-trans Golgi transport. Our data indicate an increase in medial Golgi persistence time and as such suggest a delayed cisternal progression explanation for the observed increase in Golgi cisternal number in Rab6 depleted HeLa cells 18,21 . To the best of our knowledge, these are the  . For expression of tsO45G protein, HeLa cells were cultured at 39.5 °C to allow accumulation of protein in the ER overnight, and then transferred to media that has been prewarmed to 32 °C, conditions permissive for anterograde transport of tsO45G to the cell surface 18 .
All siRNAs used for this work were manufactured by Dharmacon RNA Technologies, a division of Thermo Fisher. siRNAs were transfected at a final concentration of 100 nM, using Oligofectamine Transfection Reagent (Invitrogen) in the absence of FBS according to previously described protocols. In brief, for visual analysis, 70,000 cells were seeded per 35 mm tissue culture dish containing 12 mm diameter coverslips (Fisher Scientific) one day prior to siRNA transfection (Day 0). Successive cycles of siRNA transfections were performed on Day 1 and Day 2. For light microscopy experiments, cells were fixed with 3.4% paraformaldehyde or cold methanol (− 20 °C) 96 h (4 days) post initial transfection. For Western blotting, cell lysates were prepared from individual tissue culture dishes for subsequent quantitative assessment of protein depletion 16 . Electron microscope (EM) analysis of Golgi organization. HeLa cells were cultured on sapphire discs as previously described 21 and transfected with Control or siRNA directed against Rab6 as above. 96 h posttransfection, cells were either immediately processed for high-pressure freezing (HPF) and freeze substitution (FS) as previously described 21 or were paraformaldehyde fixed, saponin permeabilized per the immunofluorescence procedure described below prior to subsequent HPF/FS and EM. All EM was done as described 21 .
VSVG transfection and chase. WT and NAGT1 HeLa cells (70,000) were seeded on coverslips in 35 mm dishes 1 day prior to siRNA transfections. Transfections were carried out as previously described 16 , however on Day 4 after incubations with either siControl or siRab6 A/A′, HeLa cells were transfected with plasmids encoding GFP-tagged tsO45 mutant VSVG protein using FuGENE HD transfection reagent (Promega). Posttransfection cells were incubated overnight at 39.5 °C, a nonpermissive temperature for VSVG transport from the ER to the Golgi apparatus. To chase, cells were shifted to 32 °C, permissive temperature for VSVG transport, in the presence of cycloheximide to prevent further protein synthesis 18 . Cells were incubated at 32 °C for various chase times, then fixed with 3.4% formaldehyde and stained with various Golgi markers. Quantitative results are the mean of three independent experiments. RUSH expression and chase. Wild type HeLa cells were treated with the corresponding siRNA as described above. Typically, 48 h after the first transfection cycle, cells were transfected with RUSH plasmid encoding VSVG-GFP or GPI-GFP using FuGENE HD transfection reagent (Promega) according to the manufacturer's protocol. Expression of RUSH constructs in HeLa cells and chase were performed as described previously with minor modifications 35 (plasmids were a gift of Franck Perez, Institut Curie). In brief, transfected cells were incubated overnight at 37 °C. 1 × 10 −7 M avidin (Sigma-Aldrich) was added for efficient retention of cargo proteins in the ER. Cells were then incubated in the presence of 40 μM biotin (Sigma-Aldrich) to release the cargo proteins. Cells were incubated at 37 °C for various chase times in the presence of cycloheximide to prevent further protein synthesis. Cells were then fixed with formaldehyde. Confocal image stacks were taken as described above.
Fluorescence microscopy and image processing. To assess the effect of protein knockdown on Golgi phenotype, HeLa cells (wild type or stably expressing NAGT1myc) were grown on coverslips and transfected with either Control or Rab6 A/A′ siRNAs. 96 h post siRNA transfections, cells were fixed with 3.4% formaldehyde and stained with Rab6 antibody 16 . Widefield microscopy and spinning disk confocal imaging were performed with a Zeiss Axiovert 200 M microscope (Carl Zeiss) fitted with high numerical aperture (NA) objectives: 20X/0.8 NA, 63X/1.40 NA and 100X/1.40 NA and a CARV II spinning disk confocal accessory (BD Bioimaging) mounted to the side port of the microscope 16 . The microscope was controlled with iVision Mac (BioVision). For the analysis of VSV-G fluorescence distribution relative to known cis, medial and trans Golgi markers (see below), images were first deconvolved using Huygens Professional 4.3.0p3 software to sharpen the confocal images. Reported results are the average across 30 or more cells per condition and time point. Line scan analysis of VSV-G distribution relative to Golgi markers was done as described below.

Scientific RepoRtS
| (2020) 10:16604 | https://doi.org/10.1038/s41598-020-73276-w www.nature.com/scientificreports/ Confocal line scan analysis and microscopy. The distributions of several Golgi proteins, including cis localized p115, medial localized NAGT1, trans localized GalT and SialylT, trans Golgi network localized TGN46, and VSV-G-GFP were analyzed from peak intensities along a line track drawn perpendicular to the Golgi ribbon. In brief, ribbon-like Golgi apparatus structures in a plane giving the brightest Golgi staining and maximal separation of Golgi markers were chosen (5-7 areas per Golgi, ≥ 30 cells per condition). The pixel intensities along the line track for these markers were plotted versus distance in microns using ImageJ software and the distances between peak intensities were determined for each time point. Spinning disk confocal image stacks for Golgi apparatus visualization were collected through the entire depth of the cell and either compressed into a single plane using maximum intensity projection (MIP), or as in most cases the best single plane was chosen for confocal line scanning. Images were color joined using iVision Mac software. Deconvolution to sharpen further the confocal fluorescent images was done using Huygens software (SVI). Line scanning was done in the ImageJ freeware program, and images were prepared for publication with Adobe Photoshop and Illustrator software (Adobe Systems).
3D-SIM. 3D-SIM microscopy was performed with Zeiss ELYRA PS.1 super-resolution system. Image stacks were collected with 100x/1.46 oil objective and processed with Zen software (Zeiss). The images were taken at the same exposure time. The image stacks were compressed into a single plane using "3D" processing mode in Zen software.

SDS-PAGE and western blot analysis.
SDS-PAGE and western blotting were performed to determine knockdown of protein level for each experiment using commercial antibodies and reagents. Protein samples were analyzed by electrophoresis on either 8% or 10% polyacrylamide gels followed by transfer to nitrocellulose membrane. The blots were incubated first with primary antibodies, and then with a secondary IgG antibody conjugated with IRDye 800 (light sensitive). The blots were scanned and analyzed with an Odyssey Infrared Imaging System (LICOR Biosciences).

Endoglycosidase (EndoH) assays. Control and siRab6 treated HeLa cells transfected with tsO45G-GFP
for 18 h at 39.5ºC were shifted to permissive temperature, 32ºC, for various times. Cells were detergent lysed with 1% Triton X-100 in isotonic buffer as described by Whitt et al. 25 . Lysates were denatured, made acidic with citrate buffer following the manufacturer's instructions (New England BioLabs), and incubated in the presence or absence of EndoH prior to immunoblotting with anti-GFP antibodies.