Golgi complex–plasma membrane trafficking directed by an autonomous, tribasic Golgi export signal

The polybasic motif is required for p14 localization to the plasma membrane

All of the FAST proteins contain a membrane-proximal cluster of three to eight basic residues in their cytosolic C-terminal endodomains. Previous substitution analysis of the PBM of the p10 and p15 FAST proteins indicated that this motif is required for cell–cell fusion (Shmulevitz et al., 2003; Dawe et al., 2005), although the basis for this loss of fusion activity was not determined. The seven-residue p14 PBM resides four residues downstream of the TMD and comprises two tribasic clusters separated by an acidic residue (KRRERRR). To examine the importance of the p14 PBM, we replaced the two tribasic clusters by Ala, either individually (constructs p14-KRR and p14-RRR) or in combination (construct p14-PB; Figure 1A). These constructs were transfected into QM5 cells and examined for syncytiogenesis by microscopic examination of Giemsa-stained monolayers (Figure 1B), and the extent of syncytium formation was determined by quantifying the numbers of syncytial nuclei in random microscopic fields (Figure 1C). The p14-KRR and p14-RRR constructs showed slightly reduced levels of syncytium formation relative to authentic p14, whereas the p14-PB construct containing Ala substitutions of all six basic residues was devoid of cell–cell fusion activity. All three of the Ala constructs were expressed at approximately equivalent levels as authentic p14 (Figure 1D), although p14-PB showed a trend to slightly lower (∼20–40% decrease) expression levels in multiple Western blots. These modest and variable difference in p14-PB expression did not correlate with the consistent loss of syncytium formation, indicating that the p14 PBM is essential for fusion activity independent of effects on p14 expression.

To determine the basis for the syncytiogenic defect in the p14-PB construct, we assessed cell surface expression of p14-PB by fluorescence-activated cell sorting (FACS) analysis using an antiserum specific for the p14 ectodomain to stain live cells. The p14 FAST protein induces very rapid and robust syncytium formation in QM5 cells, which commences 4 h posttransfection (hpt) and encompasses the entire monolayer by 12 hpt (Salsman et al., 2005). The p14 antiserum was therefore added to cells 3 hpt to inhibit syncytium formation, allowing extended incubations to increase the detection and quantification of cell surface fluorescence, as previously described (Barry and Duncan, 2009; Barry et al., 2010). Replacement of three of the six basic residues in the PBM (constructs p14-KRR and p14-RRR) led to no significant reduction in p14 plasma membrane localization (Figure 2C), indicating that the slight reduction in syncytiogenesis induced by these constructs (Figure 1C) reflects an effect on fusion activity, not trafficking, of p14. In contrast, Ala replacement of all six basic residues in the PBM abrogated syncytium formation (Figure 1) and decreased p14-PB cell surface expression by >95% (Figure 2, A and C). FACS analysis using permeabilized cells to detect total expression levels revealed <20% decrease in p14-PB expression (Figure 2, B and D), similar to Western blotting results (Figure 1D), implying that the dramatic inhibition of p14-PB plasma membrane localization was not due to effects on protein expression. Because basic residues on the cytosolic side of a TMD influence protein topology—the so-called “positive inside rule” (Nilsson et al., 2005)—flow cytometry was also performed using antiserum raised against full-length p14. This antiserum has high affinity for epitopes in the C-terminal p14 cytosolic domain (Clancy and Duncan, 2009) but still failed to detect p14-PB on the cell surface (Figure 2E). Basic residues in the p14 PBM therefore do not affect p14 topology in the membrane but are required for p14 trafficking to the plasma membrane.

The polybasic motif is not required for p14 export from the ER

Basic residues in the cytosolic tails of some membrane proteins can promote ER export (Dong et al., 2012). To determine whether the absence of the p14-PB construct in the plasma membrane reflected an inability to traffic from ER to the Golgi complex, we introduced a V9T substitution into p14-PB. Previous results using endoglycosidase H (endo H) in the presence and absence of the N-linked glycosylation inhibitor tunicamycin demonstrated this substitution creates a functional glycosylation signal in p14 (Corcoran and Duncan, 2004). FACS analysis of cell surface fluorescence indicated that this glycosylation signal does not inhibit, and might actually slightly enhance, p14(V9T) plasma membrane localization (Figure 3A). Western blots of lysates from QM5 cells transfected with p14(V9T) detected two p14 polypeptides (Figure 3B). Treatment of cell lysates with N-glycosidase F (PNGase F), an amidase that cleaves both high-mannose and complex N-linked carbohydrates from glycoproteins, confirmed that the slower-migrating species was glycosylated p14 (Figure 3B). The slower-migrating species was also partially resistant to treatment with endo H, indicative of processing in the Golgi complex to an endo H–resistant complex oligosaccharide. A similar situation was observed for p14-PB(V9T), with the presence of a slower-migrating, partially endo H–resistant polypeptide (Figure 3B). Absence of the PBM in p14-PB(V9T) led to no significant difference in the proportion of endo H–resistant p14-PB(V9T) relative to p14(V9T) (Figure 3C), indicating that the PBM is not required for p14 ER–Golgi trafficking.

The p14 PBM functions as a Golgi export signal

The endo H and surface expression results indicated that p14-PB traffics to the Golgi complex but fails to traffic to the plasma membrane, suggesting that the PBM might function as a Golgi export signal. To test this hypothesis, we examined the influence of the PBM on p14 subcellular localization by immunofluorescence microscopy, using Vero cells transfected with p14(V9T) constructs. The p14(V9T) construct correctly traffics to the plasma membrane (Figure 3A) but is fusion dead due to the Val–Thr substitution, which allowed longer incubations without the need to inhibit syncytium formation using high concentrations of α-p14 antiserum. The flatter, larger, and more uniform morphology of Vero cells provided better visualization of cytosolic compartments than in QM5 cells and provided a means to assess whether the PBM exerts any cell-specific trafficking effects. The p14(V9T) construct displayed disseminated intracellular fluorescence throughout the cytoplasm that radiated out to the cell periphery, with little indication of p14 concentration in the Golgi complex (Figure 4, top row). In contrast, p14-PB(V9T) showed a very faint, diffuse cytoplasmic staining pattern with no evidence of staining at the cell periphery (Figure 4). The p14-PB(V9T) construct strongly colocalized with phosphatidylinositol-4 kinase-IIIβ (PI4KIIIβ), a Golgi complex marker, and with TGN46; this colocalization was obvious, reproducible between experiments, and apparent in almost all cells expressing p14-PB(V9T). Absence of the PBM therefore results in p14 accumulation in the Golgi complex and TGN and inability of p14 to traffic to the plasma membrane.

Accumulation of p14-PB in the Golgi complex could reflect either failure to exit the Golgi complex or rapid retrieval of p14 from the plasma membrane to the Golgi complex via the endocytic pathway. To distinguish between these possibilities, we inhibited dynamin-dependent endocytic events using dynasore, a dynamin inhibitor (Kirchhausen et al., 2008). We used the nonfusogenic p14(G2A) construct, which displays the same subcellular localization pattern as authentic p14 (Corcoran and Duncan, 2004), instead of the nonfusogenic p14(V9T) construct to ensure that results were reproducible irrespective of the p14 construct. The transferrin receptor, a prototype marker for endocytic recycling, was used as a positive control for dynamin inhibition. Treatment of vector-transfected cells with dynasore inhibited endocytic recycling of the transferrin receptor, as indicated by a dramatic increase in cell surface fluorescence when using a fluorescent transferrin-binding assay (Figure 5, A and D). Treatment of cells with dynasore marginally increased cell surface expression of p14(G2A) (Figure 5C), implying that p14 stably localizes to the plasma membrane and is not subject to rapid endocytic recycling. More important, dynasore did not promote detection of p14-PB(G2A) on the cell surface (Figure 5, B and C). To ensure that dynamin-independent endocytosis did not contribute to the absence of p14-PB plasma membrane localization, we treated cells with anti-p14 antiserum at 37°C for 30 min before fixation, permeabilization, and labeling with a fluorescent secondary antibody. This antibody internalization assay detected only low levels of p14-PB in both premeabilized and nonpermeabilized cells, with no evidence of endocytic antibody uptake or concentration in the Golgi complex (Figure 5E). Endocytic events are thus unlikely to influence Golgi accumulation of p14-PB.

The p14 PBM functions as a tribasic Golgi export signal

To define more clearly the nature of the p14 PBM Golgi export signal, we created 12 additional substitution constructs in the PBM (Figure 6A). When cell surface expression levels of the 15 PBM substitution constructs were rank ordered, there was striking congruence between the number of basic residues and the extent of plasma membrane localization, largely independent of specific basic residue locations within the PBM (Figure 6B). The addition of one basic residue to p14-PB increased cell surface fluorescence from <5 to ∼20% of p14 levels; two basic residues in any of several locations within the PBM increased cell surface fluorescence to ∼40–60%; three or more basic residues exhibited cell surface fluorescence equivalent to p14 with an authentic PBM containing six basic residues (Figure 6B). A one-way analysis of variance (ANOVA) analysis of three independent experiments indicated no significant difference in cell surface expression levels between constructs sharing the same number of basic residues, but differences in cell surface expression were statistically significant between groups (Figure 6B). The substitution results also indicated that it is the number of basic residues, not the net charge of the PBM, that influences Golgi export, since there was no correlation between presence or absence of the acidic Glu residue that separates the two tribasic motifs in the PBM and p14 plasma membrane localization (Figure 6, A and B).

Steady-state total expression levels varied slightly for the different constructs (Figure 6C), with constructs containing fewer than three basic residues consistently showing ∼20–40% reduction in steady-state levels by Western blot analysis (Supplemental Figure S1). Decreased steady-state expression is to be expected for plasma membrane proteins that fail to traffic out of the Golgi, since their accumulation would presumably lead to a percentage being trafficked to lysosomes for degradation. The colocalization of p14PB with Golgi complex and TGN markers (Figure 4) also suggests that some of the p14PB proteins that fail to exit the TGN might be retrieved via the COPI pathway. More important, the progressive increase in cell surface expression correlated with increasing number of basic residues (Figure 6B), not with variations in overall expression levels (Supplemental Figure S1). One notable exception to this trend was PB+K,R,R, which contained three basic residues but displayed cell surface fluorescence levels of ∼50%, more similar to the levels observed in constructs with two basic charges (Figure 6B). This construct lacked the acidic residue separating the two tribasic motifs, but other constructs lacking this acidic residue were efficiently expressed on the cell surface (e.g., RRE and RER constructs in Figure 6). The basis for this anomaly is unknown, but we note that this was the only construct with three basic residues that did not contain adjacent basic residues. Clustering of basic residues or their spatial arrangement may therefore exert some influence on the function of the PBM as a tribasic Golgi export signal.

The cell surface fluorescence results were confirmed by confocal microscopy of permeabilized cells using representative polybasic mutants in a p14(V9T) backbone. As before, p14-PB(V9T) colocalized extensively with the PI4KIIIβ Golgi marker (Figure 7). Addition of one basic charge (PB+K(V9T)) resulted in a slight increase in diffuse cytoplasmic staining, consistent with FACS results indicating a low level of Golgi export of this construct to the plasma membrane (Figure 6B). Addition of two basic residues further decreased Golgi localization while increasing broadly distributed punctate, cytoplasmic staining (B+K,R(V9T)). Addition of a third basic residue (p14-KRR(V9T)) resulted in little p14 colocalization with the Golgi marker and obvious immunofluorescence staining at the cell periphery, a similar staining pattern as p14(V9T) (Figure 7). Visual results were supported by Pearson's correlation coefficients, which showed a progressive decrease in Golgi colocalization with increasing number of basic residues in the PBM (Figure 7). Efficient p14 trafficking from the Golgi is therefore largely dependent on the presence of three basic residues in the PBM.

Context-dependent effects on the trafficking properties of p14 PBM

Previous studies suggest that the proximity to the TMD of a cytosolic, dibasic ER export signal can influence the function of this trafficking motif (Giraudo and Maccioni, 2003). To determine whether membrane proximity influences the function of the p14 PBM, we appended the p14 PBM to the C-terminus of the p14-PB construct. In this p14-PBextPB construct (Figure 8A), 68 endodomain residues separate the PBM and TMD instead of the four-residue separation in authentic p14. FACS analysis indicated this membrane-distal version of the PBM did not promote trafficking of p14 to the plasma membrane (Figure 8B). To define the nature of the p14-PBextPB trafficking defect, we examined this construct in the p14(V9T) background using the endo H assay. As previously observed, the slower-migrating p14(V9T) polypeptide was completely sensitive to PNGase F but partially resistant to endo H (Figure 8C), indicative of glycosylation and processing in the Golgi complex to a complex oligosaccharide. In contrast, the glycosylated p14-PBextPB(V9T) polypeptide was completely sensitive to endo H treatment (Figure 8C), suggesting an inability to transit to the Golgi complex. Confocal microscopy also revealed disseminated cytoplasmic staining of p14-PBextPB(V9T) reminiscent of ER staining with no evidence of Golgi localization (Figure 8D), consistent with results from the endo H assay. Thus the proximity of the PBM to the TMD influences its function as a trafficking motif.

The p14 PBM directs Golgi export of a heterologous protein

Results indicated the p14 PBM is necessary for efficient Golgi export of p14, but can it also function to promote Golgi export of a heterologous protein? To address this question, we sought to transfer the PBM to the cytoplasmic tail of a protein that resides in the Golgi complex. Ideally, this heterologous protein should lack a Golgi retention signal that could compete with the Golgi export properties of the p14 PBM. A chimeric ERGIC-53 protein fulfilled these criteria. This protein contains the myc-tagged ectodomain of ERGIC-53 but the TMD of CD4 connected through two Arg residues to a 10-residue polyalanine tail (Gut et al., 1998). The TMD and cytoplasmic tail replacements remove signals that normally retain the ERGIC-53 protein in the early secretory pathway, resulting in preferential localization of the chimeric ERGIC-53 protein to the Golgi complex in stably transfected MDCK cells (Gut et al., 1998). We inserted a sequence corresponding to the p14 membrane-proximal region containing the PBM (KYLQKRRERRRQ) in place of the two Arg residues at the TMD/cytoplasmic tail junction in the chimeric ERGIC-53 construct (Figure 9A). We also constructed a similar replacement with a sequence substituting the basic residues in the p14 PBM with Ala residues (KYLQAAAEAAAQ). As a positive control for Golgi export of the ERGIC-53 chimera, a second chimeric version of this protein was used that contains two glycosylation sites in the ectodomain, which generates a glycoprotein that is more efficiently exported from the Golgi complex to the plasma membrane (Gut et al., 1998).

All four of the chimeric ERGIC-53 proteins were expressed in stably transfected QM5 cells at approximately equivalent levels, as observed by Western blotting using α-myc antiserum (Figure 9B). As previously reported (Gut et al., 1998), the chimeric ERGIC-53 protein showed reduced levels of cell surface expression when analyzed by flow cytometry, but addition of the glycosylation signals substantially increased plasma membrane localization (Figure 9C). Although the construct containing the p14 PBM region with Ala substitutions had slightly higher levels of cell surface expression than the parental chimeric ERGIC-53 construct, addition of the authentic p14 PBM resulted in cell surface expression at levels equivalent to, or slightly higher than, those induced by glycosylation of the chimeric ERGIC-53 protein (Figure 9C). Immunofluorescence microscopy of stably transfected Vero cells confirmed the cell surface fluorescence results. The parental chimeric ERGIC-53 protein and the construct containing the mutated p14 PBM both displayed strong colocalization with the PI4KIIIβ Golgi complex marker (Figure 10). In contrast, both glycosylated ERGIC-53 protein and the construct containing the p14 PBM displayed disseminated cytoplasmic fluorescence that radiated out to the cell periphery with limited Golgi colocalization (Figure 10). The p14 PBM can therefore function as a Golgi export signal to direct Golgi export and plasma membrane trafficking of heterologous proteins.

Cells and antibodies

Vero and QM5 cells were maintained at 37˚C and 5% CO₂ in medium 199 with Earle's salts supplemented with 5 or 10% fetal bovine serum (FBS), respectively_. The rabbit polyclonal anti-p14 and anti-14 ectodomain antisera were previously described (Corcoran and Duncan, 2004; Top et al., 2005). Antibodies against actin (Sigma-Aldrich, St. Louis, MO), myc epitope tag (Sigma-Aldrich), PI4KIIIβ (BD Biosciences, Franklin Lakes, NJ), TGN46 (Serotec, Oxford, UK), protein disulfide isomerase (PDI; Enzo Life Sciences, Farmingdale, NY), horseradish peroxidase (HRP)–conjugated goat anti-rabbit (Jackson ImmunoResearch, West Grove, PA), Alexa Fluor 488–conjugated goat anti-mouse or donkey anti-sheep, Alexa Fluor 555–conjugated donkey anti-rabbit, and Alexa Fluor 647–conjugated goat anti-rabbit secondary antibodies (Life Technologies, Carlsbad, CA) were obtained from the indicated suppliers.

Cloning and vectors

The p14 gene cloned into pcDNA3 and the nonfusogenic p14(V9T) and p14(G2A) point substitution constructs were described previously (Corcoran and Duncan, 2004). These constructs were used as templates to create the indicated point substitutions in the p14 PBM using the QuikChange (Stratagene, La Jolla, CA) method according to the manufacturer's specifications. The p14PBextPB was created using p14-PB as a template for PCR with a reverse primer that added the PBM to the C-terminus of the protein. The myc-tagged chimeric ERGIC-53 construct with or without glycosylation sites (D61E and VNATASA substitutions in the ectodomain) were provided by Karl Matter (University College London, London, United Kingdom) in the pCB6 vector, subcloned into pcDNA3, and used as templates for PCR with primers that replaced the membrane-proximal diarginine sequence in the cytoplasmic tail with the p14 membrane-proximal region containing the PBM (KYLQKRRERRRQ) or Ala substitutions of the PBM (KYLQAAAEAAAQ). All experiments were performed using transient transfections with these vectors, except for the ERGIC-53 chimeras that were used in stably transfected cells.

Transfection and syncytial indexing

Cluster plates containing subconfluent monolayers of QM5 or Vero cells were transfected with expression plasmids using Lipofectamine (Invitrogen) or jetPRIME (PolyPlus Transfection, Illkirch, France), respectively, according to the manufacturer's instructions. At 8 hpt the QM5 monolayers were rinsed with Hank's balanced salt solution (HBSS), fixed with methanol, and stained using Wright–Giemsa stain, and a syncytial index was determined by counting the number of syncytial nuclei (i.e., cells containing >4 nuclei/cell) in five random microscopic fields, as previously described (Corcoran and Duncan, 2004). For stable transfections with the ERGIC-53 constructs, at 24 hpt the growth medium on QM5 or Vero cells was replaced with growth medium containing 1 mg/ml G418 (Life Technologies), and transfected cells were cultured under these selective conditions for 5 d with subculturing in the selective medium as required.

Western blotting

At 8 hpt, QM5 cells were lysed with RIPA buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 1 mM EDTA, 1% NP-40, 0.5% IGEPAL) containing 1 μM protease inhibitors (aprotinin, pepstatin, and leupeptin) for 45 min on ice. Equivalent protein loads, as determined by Lowry assays (Bio-Rad, Hercules, CA), were analyzed by SDS–PAGE (15% acrylamide) and Western blotting using anti-p14 antiserum (1:10,000) or antibodies against actin (1:2500) or the myc epitope tag (1:2500) and a HRP-conjugated goat anti-rabbit secondary antibody (1:10,000). For the endo H assays, cells were lysed at 24 hpt, and before SDS–PAGE lysates were treated at 37˚C for 2 h with endo H or PNGase F according to the manufacturer's specifications (New England BioLabs, Ipswich, MA). Membranes were developed using ECL-Plus reagent (GE Healthcare, Little Chalfont, UK) and imaged on a Typhoon 9410 variable-mode imager (GE Healthcare) or a Kodak 4000-mm Pro CCD imager. Blots were quantified by ImageJ (National Institutes of Health, Bethesda, MD) and results reported as band intensity relative to authentic p14.

Cell surface immunofluorescence flow cytometry

Transfected QM5 cells were cultured for 24 h in growth medium containing a 1:20 dilution of anti-p14 ectodomain antiserum to prevent syncytium formation. Conversely, cells transfected with the nonfusogenic p14(G2A) pr p14(V9T) constructs were cultured for 24 h in normal growth medium. Live cells were then incubated at 4˚C for 30 min in blocking buffer (5% normal goat serum, 1% bovine serum albumin, 0.02% NaN₃ in HBSS) and stained with a 1:1000 dilution of anti-p14 ectodomain, anti-14 full length antiserum, or anti-myc antibody (for the chimeric ERGIC-53 constructs) and a 1:2000 dilution of Alexa 647–conjugated goat anti-rabbit secondary antibody, each for 1 h at 4˚C. Cells were resuspended in phosphate-buffered saline (PBS) containing 10 mM EDTA and fixed with 3.7% formaldehyde, and 10,000 cells were analyzed by flow cytometry (FACSCalibur; BD Biosciences) using De-Novo software. Cells transfected with empty vector were used as negative controls for immunostaining, the fluorescence gate was set to <5%, and the same gate was applied to quantify surface fluorescence of the p14-transfected cells. Background fluorescence from vector-transfected cells was subtracted before mean percentage surface fluorescence was calculated.

Intracellular immunofluorescence flow cytometry

QM5 cells were transfected with nonfusogenic p14(G2A) or p14-PB(G2A) for 24 h in normal culture medium. Cells were then resuspended in PBS with 10 mM EDTA and fixed with 1% paraformaldehyde in PBS for 15 min at room temperature. Cells were pelleted, washed with PBS, and permeabilized in blocking buffer containing 0.1% saponin for 30 min at room temperature. Cells were then labeled with 1:1000 dilution of anti-p14 antiserum and 1:2000 dilution of Alexa 647–conjugated goat anti-rabbit secondary antibody, each for 1 h at room temperature in blocking buffer containing 0.1% saponin. Cells were resuspended in PBS and analyzed by flow cytometry as indicated for cell surface immunofluorescence.

Intracellular immunofluorescence microscopy

At 24 hpt, transfected Vero cells cultured on glass coverslips (Thermo Scientific, Hampton, NH) were fixed in 3.7% formaldehyde for 20 min at room temperature, permeabilized for 20 min at room temperature with 0.1% Triton X-100, blocked for 30 min in blocking buffer, and stained with rabbit anti-p14 antiserum (1:200) or rabbit anti-myc antibodies (1:1000) or with 1:1000 dilutions of mouse monoclonal antibodies against a Golgi marker (PI4KIIIβ), TGN marker (TGN46), or ER marker (PDI) and with 1:1000 dilutions of Alexa 488–conjugated goat anti-mouse or Alexa 647–conjugated goat anti-rabbit secondary antibodies. Coverslips were mounted on glass slides using fluorescence mounting medium (Dako, Glostrup, Denmark) or Prolong gold antifade reagent (Life Technologies) and then visualized and photographed using a Zeiss LSM 510 META confocal microscope or a Zeiss Axioplan II MOT and AxioCam HRC Color Camera. Images were acquired with either 63 or 100× objective. Images were analyzed by ImageJ to generate fluorescence intensity graphs. Pearson's r was determined for 5–10 cells on unadjusted images using the Fiji version of ImageJ (Schindelin et al., 2012). If required, background was corrected using the rolling ball background plug-in, and colocalization thresholds were set using the Coloc_2 plug-in (Costes et al., 2004).

Endocytosis inhibition assay

Cells were transfected with nonfusogenic p14(G2A) or p14-PB(G2A), and at 24 hpt, cells were incubated with 80 μM dynasore (Santa Cruz Biotechnology, Dallas, TX) or 0.8% dimethyl sulfoxide (DMSO; Sigma-Aldrich) at 37˚C for 1 h in serum-free growth medium. Cells were then treated and analyzed for p14 surface expression as described in surface expression assay. For analyzing dynasore functionality, vector-transfected cells were serum starved for 30 min and then incubated with 80 μM dynasore or 0.8% DMSO at 37˚C for 30 min. Alexa 555–conjugated transferrin (Molecular Probes) was then added (20 μg/ml) to cells for 20 min at 4˚C. Cells were then harvested and analyzed as described in surface expression assay.

Antibody internalization assay

Vero cells cultured on glass coverslips were transfected with nonfusogenic p14(G2A) or p14-PB(G2A) and incubated for 24 h with normal growth medium. Cells were then incubated with blocking buffer and subsequently with 1:200 dilution of anti-p14 antiserum in blocking buffer, each for 30 min at 37C. Cells were fixed with 4% paraformaldehyde for 15 min at room temperature and then processed further as nonpermeabilized cells or after permeabilization with 0.1% Triton X-100 in PBS for 15 min at room temperature using Alexa Fluor 488–conjugated goat anti-rabbit antibody, as described for 1 h at room temperature. Coverslips were then mounted and analyzed as indicated in intracellular immunofluorescence staining.

Statistical analysis

Statistical analysis and sample comparison were performed using Prism software (GraphPad, San Diego, CA). Groups of two samples were analyzed with a paired two-tailed t test, and groups of more than two samples were analyzed using ANOVA with a Tukey post test.