The clathrin heavy chain isoform CHC22 functions in a novel endosomal sorting step

CHC22 function segregates from CHC17 function and its localization is restricted to endosomes

We established both absolute and relative levels of endogenous expression of the two CHCs in different cell types, in order to understand the apparently discrepant results from studies describing their functional differentiation or compensation (Hood and Royle, 2009; Vassilopoulos et al., 2009). Levels of CHC22 relative to total protein varied between nine cell lines analyzed, derived from different tissues (Fig. 1 A). Myoblast cultures and HeLa cells had the highest expression. Cell lines derived from other human tissues all expressed some CHC22, but considerably less, with the lowest expression in the liver cell line HepG2 and the natural killer cell line NKL. This is consistent with predicted expression based on mRNA levels in different tissues (Kedra et al., 1996; Sirotkin et al., 1996). The murine skeletal muscle cell line C2C12 had no CHC22 protein, as expected from the murine CHC22 pseudogene. The levels of CHC17 were far less variable between cell lines, indicating that the ratios of CHC22 to CHC17 vary between tissues. To determine these ratios from absolute protein levels of each CHC, the antibodies used for immunoblotting were calibrated against known amounts of recombinant protein fragments of each CHC. Dilutions of cell lysates were correlated with these standards based on immunoblotting signals (Fig. 1 B). These experiments revealed that the ratios of CHC22 to CHC17 are ∼1:8 for LHCNM2 and 1:12 in HeLa. Thus, under endogenous conditions, there is 8- to 12-fold more CHC17 in cells than CHC22, even in cells that have the highest amount of CHC22.

To further study the functional relationship between CHCs at normal levels of expression, we used siRNA to individually down-regulate both CHCs in HeLa cells (Fig. 1 C). In nontreated cells (not depicted) and cells treated with control siRNA, immunofluorescence detection of CHC22 revealed a punctate staining pattern with a more uniform intracellular distribution than CHC17, lacking the strong perinuclear crescent typical of CHC17 labeling in the TGN. Minor colocalization between CHC22 and CHC17 was confined to peripheral compartments (Fig. 1 D; Fig. S1 A). In cells depleted of CHC17, CHC22 staining appeared marginally brighter in the perinuclear region (Fig. 1 E; Fig. S1 B) and there was slightly more CHC22 colocalized with AP1 (Fig. S1 E) compared with control-treated cells (Fig. S1 D). This change in distribution may reflect increased membrane recruitment of CHC22 in the absence of CHC17, accounting for the modest increase in CHC22 protein levels detected by immunoblotting after CHC17 depletion (Fig. 1 C). In contrast, CHC17 distribution and levels (Fig. 1 F; Fig. S1 C) were unaffected by depletion of CHC22, as were CLC protein levels (Fig. 1 C), corroborating no detectable biochemical interaction between CHC22 and CLCs or CHC17 (Liu et al., 2001). By comparison, depletion of CHC17 profoundly reduced CLC levels. Together, these depletion studies indicate that under endogenous conditions, CHC22 is outcompeted for localization to the TGN by CHC17.

Consistent with CHC22 showing no coimmunoprecipitation with the endocytic adaptor AP2 (Liu et al., 2001; Vassilopoulos et al., 2009), CHC22 was not observed at the plasma membrane, even in the absence of CHC17, and depletion of CHC22 had no effect on endocytosis of transferrin by HeLa cells (Fig. 1, D–F). These results correlated with no observed effect on EGF uptake after CHC22 depletion from LHCNM2 myoblasts (Vassilopoulos et al., 2009). In contrast, CHC17 depletion strongly inhibited transferrin endocytosis (Fig. 1, D–F), as described previously (Hinrichsen et al., 2003; Motley et al., 2003). However, it was recently demonstrated that supertransfected CHC22 rescued endocytosis defects caused by CHC17 depletion (Hood and Royle, 2009). The GFP-CHC22 construct used in that study was driven by a CMV promotor, resulting in high expression. The compensation by CHC22 in the absence of CHC17 raised the question of whether higher-than-natural levels of CHC22 could compete for endogenous CHC17. To address this, we reproduced the CMV promoter-driven GFP-CHC22 construct, making sure it was full length and not missing the apparently labile exon 9 (Hood and Royle, 2009). After transient transfection of HeLa cells, colocalization between CHC17 and CHC22 was assessed by antibody detection of both endogenous and artificially introduced protein, comparing nontransfected (Fig. 2 A) with transfected cells (Fig. 2 B) in the same culture. High levels of GFP-CHC22 caused increased (three- to fourfold) colocalization of the two CHCs (Fig. 2 C). Thus, when CHC22 is expressed in excess, it can localize to the same cellular regions as CHC17. It is notable that neither the location nor intensity of CHC17 staining was altered by the presence of GFP-CHC22, suggesting that the introduced protein could not compete with endogenous CHC17 when present. Furthermore, CHC22 does not compensate for CHC17 function if absent and CHC22 is expressed at its relatively low endogenous levels (Fig. 1, D–F), so the apparently redundant function of the two CHCs occurs only if CHC17 is absent and CHC22 is highly overexpressed. Similar considerations apply to CHC22 substitution for CHC17 function in mitotic spindle dynamics, as we did not detect CHC22 by antibody staining on spindles in either control cells or cells depleted of CHC17 (Fig. S1, F and G).

Lack of localization of CHC22 with TGN (perinuclear AP1, TGN46; Fig. S1 D, Fig. S2, A and B) and plasma membrane markers (AP2; Fig. S2 C) prompted us to study CHC22 in the endosomal system where subpopulations of AP1 and AP3 are present (Dell’Angelica et al., 1997; Meyer et al., 2000). CHC22 did not localize with early endosomal antigen EEA1 or recycling endosomal Rab11 (Fig. 3, A and B; Fig. S2, G and H), but partially overlapped with the late endosome marker Rab9 (Fig. 3 C; quantified in Fig. S2, G and H). Lesser overlap was seen with AP3 (Fig. S2 E) and sorting nexins SNX5 (Fig. S2 D) and SNX1 (Fig. S2 F), which are present in endosomal sorting domains. Treatment of cells with Brefeldin A disrupted the characteristic TGN localization of CHC17 (Wong and Brodsky, 1992) but did not affect overall CHC22 distribution or CHC17 distribution in endosomal regions (Fig. S3). Hence, endogenous CHC22 is confined to endosomes. We therefore tested a role for CHC22 in distinct endosomal sorting pathways defined by model cargo molecules.

CHC17 and CHC22 act independently in endosomal sorting

Previous studies using dominant-negative constructs of the two CHCs representing the C-terminal third (Hub) of each protein suggested that they are both involved in CI-MPR localization. The CHC22 Hub caused dispersion of the CI-MPR from its normal perinuclear focus (Liu et al., 2001) and the CHC17 Hub induced a tighter perinuclear distribution than normal (Bennett et al., 2001). In the present study, similar effects were observed after treatment of HeLa cells with siRNAs targeting each CHC (Fig. 4). In comparison to the loose perinuclear concentration of CI-MPR in late endosomes in control cells (Fig. 4, A and C; Fig. 5, A and C), CHC17 depletion induced a slightly tighter perinuclear concentration, as reported previously (Hinrichsen et al., 2003; Fig. 4 D, Fig. 5 D). In contrast, after CHC22 depletion, CI-MPR was quantitatively more dispersed throughout the cytosol in small vesicular structures (Fig. 4, B, E and F; Fig. 5, B and E), as assessed by measuring levels of CI-MPR peripheral to the Golgi region (see illustrative Fig. S4). Expression of an siRNA-resistant (knock-down proof [KDP]), FLAG-tagged CHC22 construct completely reversed the effects induced by siRNA targeting of CHC22 without affecting control-treated or CHC17-depleted cells (Fig. 4, C–F), further indicating that the two CHCs act at separate stages of CI-MPR transport. CHC-depleted cells were tested for secretion of the lysosomal hydrolase hexosaminidase (Fig. 4 G), a hallmark of mis-sorting of CI-MPR at the TGN (Riederer et al., 1994). CHC17 and CHC22 loss resulted in a respective 70% or 22% increase of secreted hexosaminidase, with the considerably stronger effect of CHC17 corroborating its role in sorting CI-MPR at the TGN. Simultaneous depletion of CHCs appeared to increase hexosaminidase secretion in an additive fashion, though not statistically significant compared with CHC17 depletion alone, supporting roles for each CHC at distinct steps of CI-MPR traffic.

The punctate distribution of CI-MPR in CHC22-depleted cells suggested that it might be trapped in some stage of the endosomal pathway. Furthermore, CI-MPR did not colocalize with CHC17 in CHC22-depleted cells, indicating that CI-MPR was not present in the TGN upon loss of CHC22 (Fig. 5 B). CI-MPR sorting in the endosomal pathway requires CHC17 and retromer for retrograde exit from EEA1-positive endosomes (Arighi et al., 2004; Saint-Pol et al., 2004; Seaman, 2004). Consequently, CI-MPR showed increased colocalization with EEA1 after depletion of CHC17 or SNX1 (Fig. 5, D and F; and quantified in Fig. 5 G). In contrast, CI-MPR did not display increased association with EEA1 after CHC22 depletion (Fig. 5, E and G). EEA1 itself displayed little overall change in CHC22-depleted cells, although a tendency for tubulation of EEA1⁺ structures was noted. Thus, in endosomal sorting of CI-MPR, CHC22 appears to act downstream of early endosome exit and separately from CHC17.

In contrast to CI-MPR cycling, TGN46 traffics rapidly from early endosomes to the TGN via recycling endosomes (Banting et al., 1998; Ganley et al., 2008; Nokes et al., 2008). This results in steady-state perinuclear accumulation in control cells (Fig. 6 A). TGN46 localization was not affected in CHC22-depleted cells (Fig. 6 C), whereas loss of CHC17 resulted in diminished and dispersed TGN46 labeling. This phenotype suggested retention of TGN46 in early endosomes and subsequent degradation in cells depleted of CHC17, correlating with reduced protein levels in these cells (Fig. 6 D). Thus, dependence of TGN46 traffic on CHC17 and independence from CHC22 function further distinguishes trafficking steps mediated by the two CHCs.

Shiga toxin B-subunit (STxB) serves as a model cargo for early endosome-to-TGN trafficking. Upon clathrin-independent entry into cells, STxB transits EEA1-positive early endosomes where it is sequentially sorted in CHC17- and retromer-dependent steps, before it is targeted to the TGN (Saint-Pol et al., 2004; Bujny et al., 2007; Popoff et al., 2007; Johannes and Popoff, 2008). In cells depleted of either CHC or SNX1, or in cells treated with control siRNA, STxB internalized for 15 min was colocalized with EEA1 (Fig. S5, A–D), indicating no defect in plasma membrane uptake. After 60 min of uptake, STxB showed perinuclear localization in control-treated cells (Fig. 7 A, Fig. S5 E), reproducing described kinetics for TGN arrival (Mallard et al., 1998). In cells depleted for either CHC or SNX1, STxB was more disperse, as though its retrograde transport to the TGN was impaired (Fig. 7, B–D, Fig. S5, F–H). Quantification of proximity to GM130 showed that peripheral STxB localization was more than doubled compared with its normal distribution in all three depletion conditions (Fig. 7 E). However, in CHC22-depleted cells, STxB was less colocalized with EEA1 (Fig. 7, C and F) than in cells depleted for CHC17 or SNX1 (Fig. 7, B, D, and F). Increased colocalization between STxB and EEA1 in CHC17- or SNX1-depleted cells was as expected, based on previous studies (Saint-Pol et al., 2004; Popoff et al., 2007). Thus, depletion of CHC22 did not affect STxB exit from the early endosome, where STxB was localized after 15 min of uptake (Fig. S5, A–D) and is sorted by CHC17 and SNX1. Instead, CHC22 depletion interfered with a later sorting step needed for endosome-to-TGN traffic of STxB, separating it from CHC17 function in this pathway, as well as in CI-MPR transport.

Examination of STxB, CI-MPR, and EEA1 together (Fig. 8) confirmed that both cargoes were able to transit from early endosomes after CHC22 depletion and were colocalized in a compartment (Fig. 8, C and E) with less EEA1 (Fig. 5 G and Fig. 7 F) compared with their localization after depletion of CHC17 or SNX1. After SNX1 depletion, both cargoes were colocalized with each other to the same extent as seen with CHC22 depletion (Fig. 8, D and E), but were increasingly localized with EEA1 (Fig. 5 G and Fig. 7 F), reflecting a common dependence on SNX1 for early endosome exit. CHC17 depletion, on the other hand, differentially affected the two cargoes. CI-MPR and STxB were less colocalized after CHC17 depletion than after CHC22 or SNX1 depletion (Fig. 8, B and E), and there was more overlap between STxB and EEA1 than between CI-MPR and EEA1 in CHC17-depleted cells (Fig. 7 F and Fig. 5 G). This can be explained by a requirement for CHC17 in CI-MPR exit from the TGN that has been previously demonstrated (Hinrichsen et al., 2003). Thus, CHC22 function is required after early endosome exit of either cargo and before the divergence of STxB from CI-MPR on their respective routes to the TGN from endosomes. In complete distinction, CHC17 function is critical for CI-MPR export from the TGN, sorting of STxB (and some CI-MPR, e.g., Fig. 5 G) out of early endosomes, and recycling of TGN46 from endosomes to the TGN (Fig. 6). The CHC22-dependent step that commonly affects CI-MPR and STxB before their known retrograde divergence subdivides retrograde sorting, defining a novel stage in the endosomal transport pathway.

CHC22 plays a role in endosomal sorting in human myoblasts

Having localized CHC22 function to a distinct sorting step in the endosomal pathway in nonmuscle cells, it was of interest to determine if CHC22 plays the same role in myoblasts. In doing so, the issue of whether CHC22-dependent endosomal sorting is an important step in GLUT4 sequestration in the GSC is addressed (Hou and Pessin, 2007; Huang and Czech, 2007; Vassilopoulos et al., 2009). The human myoblast cell line LHCNM2 has a 1.5-fold higher ratio of CHC22 expression relative to CHC17 than HeLa (Fig. 1 B), and differentiates into myotubes, like the commonly studied rodent myoblast lines C2C12 and L6 (Zhu et al., 2007). Depletion of CHC22 or CHC17 from LHCNM2-derived myotubes (Fig. 9, A–D) had similar effects to those observed for HeLa cells (Fig. 5). In myotubes depleted of CHC22, CI-MPR was dispersed compared with its perinuclear focus around myonuclei in control-treated myotubes. This strong CI-MPR association with myonuclei made it hard to quantify a qualitatively denser appearance upon CHC17 depletion. However CI-MPR labeling was clearly different in myotubes depleted for each CHC and, as seen for HeLa cells, a slight increase in CHC22 protein level was observed in CHC17-depleted cells (Fig. 9 E). CI-MPR-levels were decreased to 83% and 61% of control cells after depletion of CHC17 or CHC22, respectively (Fig. 9, E and F), confirming a role for both CHCs in influencing endosomal sorting of CI-MPR in myotubes. To address whether CHCs function at different steps of CI-MPR sorting, as in HeLa cells, processing of the CI-MPR–transported lysosomal hydrolase, Cathepsin D, was analyzed. Upon exiting the TGN as an immature protein (pro, 53 kD) bound to CI-MPR, Cathepsin D enters the endocytic pathway where it is proteolytically processed to an intermediate (pre) form (47 kD) and then into a mature form (31 kD). Levels of these Cathepsin D forms may be used to assess trafficking toward the lysosome (Mardones et al., 2007). After CHC22 depletion, the amount of intermediate (pre) form was increased (Fig. 9, E and G), indicative of Cathepsin D accumulation with CI-MPR in an intermediate endocytic compartment. This effect was not observed upon CHC17 depletion, consistent with the limited effect of CHC17 depletion on CI-MPR in muscle cells. Thus, CHC22 plays a role in endosomal sorting in muscle cells that is distinct from functions of CHC17, and is implicated in the same functional pathway as was mapped for CHC22 in HeLa cells.

GLUT4 targeting to the GSC in human muscle depends on CHC22 (Vassilopoulos et al., 2009), and the studies reported here map CHC22 function in muscle to retrograde endosomal transport. In mice, which normally do not express CHC22, transgenically introduced CHC22 interacts with adaptors responsible for GLUT4 transport and disrupts murine GLUT4 traffic such that the formation and function of the GSC is impaired. It is conceivable that for CHC22 to function as it does in humans, additional species-specific factors are needed to complete its role in GSC formation (Vassilopoulos et al., 2009). An obvious candidate for such a factor is the SNARE family protein STX10, which plays a role in retrograde transport of CI-MPR from endosomes to the TGN, is enriched in muscle, and is expressed in humans but not mice (Tang et al., 1998; Ganley et al., 2008). The effects of STX10 depletion on GLUT4 traffic in human myoblasts were therefore assessed (Fig. 10) and found to phenocopy CHC22 depletion, causing dispersion and loss of GLUT4 staining (Fig. 10, B–D). As seen previously, STX10 depletion also caused dispersion of the CI-MPR in HeLa cells (Fig. 10 F), but it did not affect retrograde transport of STxB (Fig. 10 G; Mallard et al., 2002; Ganley et al., 2008). Again, phenocopying CHC22 (Vassilopoulos et al., 2009), STX10 depletion slightly reduced total GLUT4 levels in myotubes (Fig. 10, H and I), indicating that GLUT4 was partially degraded when GSC formation was disrupted. Localization of CHC22 was not severely affected in cells depleted of STX10 (Fig. 9 A, Fig. 10 C), whereas CHC22 depletion led to some dispersion of STX10 (Fig. 10, A and D). This suggests that CHC22 acts upstream of STX10 in endosomal sorting, consistent with CHC22 depletion affecting all three cargoes but STX10 depletion affecting only CI-MPR and GLUT4, but not STxB trafficking. Targeting STxB to the TGN must precede that of the other two cargoes, further subdividing retrograde transport from endosomes to TGN into two steps that follow a novel CHC22-dependent step. These data, which establish the joint participation of CHC22 and STX10 in GLUT4 traffic, map endosomal sorting and retrograde endosome-to-TGN transport as critical steps in GSC formation in human tissue.

Expression plasmids, antibodies, recombinant proteins, and siRNAs

Full-length human CHC22 was cloned from LHCNM2 skeletal muscle cells. mRNA was extracted using the RNeasy kit (QIAGEN) and transcribed into cDNA using SuperScript (Invitrogen). CHC22 was cloned into pT7blue cloning vector and subcloned into pEGFP-C1 and p3xFLAG-CMV-10 vectors for expression in HeLa cells. siRNA-resistant FLAG-CHC22-KDP was generated by QuikChange mutagenesis (Agilent Technologies) using primers ­CAACAACTGGG­AGGATTTGGTG­AAATTTCTGCA­GATG and ­CATCTGCAGAA­ATTTCACCAAA­TCCTCCCAGTT­GTTG on CHC22 in p3xFLAG-CMV-10, resulting in three base changes without changing the amino acid sequence.

One monoclonal and two polyclonal antibodies that specifically recognize CHC22 and not CHC17 have been described previously (Liu et al., 2001; Vassilopoulos et al., 2009). Anti-CHC17 antibodies (mouse monoclonals X22 and TD.1) and rabbit polyclonal antibody against clathrin light chains were generated in the Brodsky laboratory (Brodsky, 1985; Acton et al., 1993). Commercial sources of antibodies are as follows: TGN46 (sheep polyclonal; AbD Serotec); EEA1, AP1, SNX1, GM130 (mouse monoclonal; BD); α-tubulin, β-actin, FLAG (mouse monoclonal; Sigma-Aldrich); EEA1 (rabbit polyclonal; Cell Signaling Technology); GM130 (rabbit polyclonal, Sigma-Aldrich); CI-MPR (mouse monoclonal, EMD); Cathepsin D (rabbit polyclonal; EMD); and SNX5 (goat polyclonal; Abcam). Mouse monoclonal anti-Rab9 and rabbit polyclonal anti-STX10 antibodies were gifts from Suzanne Pfeffer (Stanford University School of Medicine, Stanford, CA). Polyclonal anti–CI-MPR was a gift from Linton Traub (University of Pittsburgh, Pittsburgh, PA). The Rab11-GFP construct was a gift from Keith Mostov (University of California, San Francisco, San Francisco, CA). Secondary antibodies for immunofluorescence against mouse IgG, IgG1, IgG2a, rabbit IgG, sheep IgG, and goat IgG were from Invitrogen (Alexa Fluor 488, 555, 568, and 647 conjugates). Secondary antibodies coupled to HRP were from Invitrogen. Recombinant STxB coupled to Cy3 was produced as described previously (Mallard et al., 1998).

Targeting and nontargeting control siRNAs were synthesized by QIAGEN. DNA sequences targeted were AATTCTCCGAACGTGTCACGT for nontargeting control, AAGCAATGAGCTGTTTGAAGA for CHC17 (Huang et al., 2004), AAGAACAAGACCAAGAGCCAC for SNX1 (Bujny et al., 2007), TATCCCACATGACGAGTGA for STX10 (Ganley et al., 2008), and TCGGGCAAATGTGCCAAGCAA and AACTGGGAGGATCTAGTTAAA for CHC22 (1:1 mixture of siRNAs were used; Vassilopoulos et al., 2009). In experiments shown in Fig. 4 E only the second siRNA against CHC22 was used, and the depletion efficiency was comparable to results using the siRNA mixture, as judged by immunoblotting whole-cell detergent lysates.

Cell culture

HeLa 229, MDA231, A431, HEK 293, HFF, HepG2, JR1, and C2C12 cells were grown in DME supplemented with 10% FBS, 50 U/ml penicillin, 50 µg/ml streptomycin, and 0.625 µg/ml fungizone. NKL natural killer cells were grown in RPMI 1640 supplemented with 10% FBS, 200 U/ml IL2, 50 U/ml penicillin, and 50 µg/ml streptomycin. LHCNM2 human skeletal muscle cells (Zhu et al., 2007) were maintained in tissue culture flasks coated with rat-tail collagen in basal medium (4:1 DME/M199, 0.02 M Hepes, 0.03 µg/ml ZnSO₄, and 1.4 µg/ml vitamin B12) with 15% fetal bovine serum (FBS), 55 ng/ml dexamethasone, 2.5 ng/ml hepatocyte growth factor (Sigma-Aldrich), 50 U/ml penicillin, 50 µg/ml streptomycin, and 0.625 µg/ml fungizone (growth medium). Differentiation was induced by culturing in fusion medium (basal medium with 0.5% FBS, 10 µg/ml insulin, 50 µg/ml apo-transferrin, and 5.5 ng/ml dexamethasone) followed by differentiation medium (basal medium with 0.5% FBS and 55 ng/ml dexamethasone).

Cells were transfected with 10 nM siRNA using HiPerfect (QIAGEN) according to the manufacturer’s instructions. HeLa cells were split, reseeded 24 h after transfection, and retreated with siRNA 24 h later. Cells were used for experimentation 48–72 h after the second round of transfection. LHCNM2 cells were transfected twice, 72 h apart, and used 72 h later. For plasmid transfection Lipofectamine 2000 (Invitrogen) was used according to the manufacturer’s instructions 24 h before analysis of transfected cells.

Internalization assays

For internalization assays cells were grown on coverslips and transfected with siRNA as described above. STxB uptake was performed as described previously (Popoff et al., 2007). In brief, cells were washed twice in cold PBS on ice and incubated on ice in prechilled medium containing 2 µg/ml STxB coupled to Cy3 for 30 min. Cells were then washed twice in cold PBS on ice and taken off ice. Warm media was added and cells were incubated at 37°C for various times before fixation. For transferrin internalization, cells were washed twice in warm PBS and preincubated for 45 min at 37°C in DME supplemented with 0.02M Hepes and 1% BSA (serum-free media [SFM]) to allow for recycling of the intracellular receptor. Cells were then washed twice in cold PBS on ice and incubated on ice for 30 min in prechilled SFM containing 10 µg/ml transferrin coupled to Alexa Fluor 488 (Invitrogen). After washing twice in cold PBS on ice, cells were taken off ice and warm medium was added. Cells were then incubated at 37°C for 10 min before fixation.

Hexosaminidase secretion

Hexosaminidase secretion was measured using a protocol modified from one described previously (Riederer et al., 1994). Cells treated with siRNA as described above were washed in PBS and incubated in DME supplemented with 10 mM mannose-6-phosphate for 8 h at 37°C. Secreted hexosaminidase activity was assayed in 0.5 ml medium by addition of 120 µl 5x hexosaminidase substrate buffer (0.5 M sodium acetate, pH 4.4, 0.5% Triton X-100, and 5 mM p-nitrophenyl-N-acetyl-β-d-glucosaminide). After shaking at 37°C for 1 h, reactions were stopped by addition of 0.6 ml stop buffer (0.5 M glycine and 0.5 M Na₂CO₃, pH 10.0). Release of p-nitrophenol was measured spectrophotometrically at 405 nm. Intracellular hexosaminidase was measured by washing the cells twice in cold scraping buffer (10 mM phosphate buffer, pH 6.0, and 0.15 M NaCl) and then scraping them into scraping buffer supplemented with 0.5% Triton X-100. After centrifugation at 100,000 g for 15 min at 4°C, 50 µl of the resulting supernatant was assayed in 570 µl 1x hexosaminidase substrate buffer. Secreted hexosaminidase was determined as the percentage of the total hexosaminidase for each sample (extracellular plus intracellular). Secreted hexosaminidase of samples treated with specific siRNA was then compared with that of samples treated with control siRNA and ratios were calculated. P-values were calculated using unpaired two-tailed Student’s t test.

Immunoblotting

Protein samples were separated by SDS-PAGE (4–12% acrylamide gel; Invitrogen), transferred to nitrocellulose (Millipore) and labeled with primary antibodies at 1–5 µg/ml and peroxidase-conjugated secondary antibodies. The presence of antibody-labeled proteins in samples was detected using Western Lightning Chemiluminescence Reagent (GE Healthcare). Rainbow molecular weight markers (GE Healthcare) were used to assess transfer efficiency and judge protein sizes. The migration positions of molecular weight marker proteins are indicated in kilodaltons (kD) at the right of each blot presented. Quantification was performed using Quantity One software (Bio-Rad Laboratories). To calibrate antibody blotting signals for quantification of CHC levels, dilutions of cell lysates were analyzed on the same gel as dilutions of recombinant protein fragments representing portions of CHC22 (amino acids 1521–1640) and CHC17 (1–1074) containing the blotting antibody epitopes. Gels were transferred to nitrocellulose and immunoblotted sequentially with polyclonal antibody against CHC22 and monoclonal antibody against CHC17, stripping in between. The dilutions of lysates and recombinant fragments analyzed were empirically selected to generate signals in a linear range. The generated antibody signals were plotted against protein concentration of material analyzed so that the CHC content of each lysate could be determined by comparison to the recombinant protein signals. The molar ratios of CHC22 to CHC17 in the cell lysates were then calculated, taking into account fragment length. All calculations and graphs were performed and generated in Microsoft Excel and GraphPad Prism software. P-values were calculated using unpaired two-tailed Student’s t test.

Immunofluorescence and image analysis

Cells grown on coverslips were washed in warm PBS, fixed in warm PFA (3.7% in PBS, 10 min), then washed (2x, 5 min, PBS), permeabilized (4 min, 3% BSA, and 0.5% Triton X-100 in PBS), and blocked in blocking solution (3% BSA in PBS, 30 min). Antibody labeling was performed by inversion of coverslips on 20 µl blocking solution with primary or secondary antibodies (1–5 µg/ml), DAPI, or TO-PRO3 on parafilm and washing with PBS. Samples were mounted in the Prolong Antifade kit (Invitrogen). Images presented in Figs. 3 C; Fig. 5, A and B; Fig. S1, F and G; and Fig. S2, B and E were acquired by confocal laser scanning microscopy using a microscope system (EZ-C1Si; Nikon) equipped with a Plan-Apo 100x, 1.40 NA oil objective and 405-, 488-, and 561-nm laser lines. All other samples were analyzed using a microscope system (TCS SP5; Leica) with a 63x HCX Plan-Apo CS oil objective using 488-, 543-, and 633-nm laser lines. Imaging was performed at room temperature using Nikon immersion oil and Leica Type F immersion oil. DAPI, Alexa 488, Alexa 555, Alexa 568, Alexa 647, and TO-PRO3 fluorescence was sequentially excited using lasers with wavelengths of 405 (DAPI), 488 (Alexa 488), 543 (Alexa 555), 561 (Alexa 568), and 633 nm (Alexa 647, TO-PRO3). Images (1024 × 1024 pixels) were saved as TIFF files in NIS Elements software (Nikon), MetaMorph software (MDS Analytical Technologies), and Leica LAS AF software, and input levels were adjusted in Adobe Photoshop. Labeling detected in individual channels is presented in black and white in individual panels. 3x enlarged merged images of boxed areas are presented in color with labeling colors being color coded as described in accompanying figure legends. Overlap of red and green staining is yellow, overlap of red and blue staining is magenta, overlap of green and blue staining is turquoise, and overlap of all three channels is white.

Image quantification was performed using National Institutes of Health’s ImageJ (http://rsb.info.nih.gov/ij). Colocalization of markers on vesicles in the periphery of double-labeled cells was performed after excluding the dense concentration of vesicles in the Golgi region because measuring individual vesicles is problematic in this area. Exclusion was the same for all cells and defined by a circular area of 72.2 µm² centered at the most dense point of the Golgi. For each cell, individual marker fluorescence was measured in separate channels, and signals were adjusted to their dynamic ranges. For each marker, only particles larger than 0.094 µm² (16 pixels) were automatically selected and counted. These automatically selected vesicles from individual labelings were then overlaid and colocalization was highlighted using the Colocalization Highlighter plugin (Pierre Bourdoncle, Institut Jacques Monod, Service Imagerie, Paris, France). The colocalized signal was again subjected to automatic particle count using the same automatic size exclusion as used to select individually labeled vesicles. The quotient of colocalized particles over total particles for one marker was calculated for individual cells and averaged.

Degree of overlap of markers in individual cells was determined by Pearsons’ correlation coefficients and pixel shifts (van Steensel et al., 1996) to estimate colocalization as well as its specificity. Measurements were performed on total fluorescence of single cells using the JACOP plugin (Bolte and Cordelières, 2006).

To quantify peripheral and perinuclear fluorescence of CI-MPR and STxB, individual cells were analyzed using the radial profile plugin (Paul Baggethun; Hehnly et al., 2006; see also illustrative Fig. S4). Fluorescent intensities within individual cells were measured as a function of distance from the centerpoint in a circular area of 2,927.7 µm² (white circle in Fig. S4). This area was selected to include all the signals detected in all cells analyzed. The centerpoint for each cell was manually set as the center of the Golgi defined by the marker GM130. Data series were imported into GraphPad Prism software and the sum of all points were calculated using the Area under Curve function representing total fluorescence. Data points enclosed in an area of 182.5 µm² around the centerpoint (red circle in Fig. S4) were defined as perinuclear with peripheral signal representing the difference between total and perinuclear signal. The boundary of the perinuclear area was set to include most of the Golgi staining for all cells analyzed. All calculations and graphs were performed and generated in Microsoft Excel and GraphPad Prism software. P-values were calculated using unpaired two-tailed Student’s t test. Figures were prepared in Adobe Photoshop and Illustrator software. Some data for this study were acquired at the Nikon Imaging Center at the Mission Bay campus of UCSF.

Online supplemental material

Fig. S1 shows minimal colocalization between endogenous CHC22 and CHC17, and a modest recruitment of CHC22 to the TGN but not to mitotic spindles in cells depleted of CHC17. Fig. S2 compares CHC22 localization to various cellular markers and includes quantification for data in Fig. 3. Fig. S3 demonstrates that CHC22 distribution is not affected by Brefeldin A treatment. Fig. S4 illustrates the method used for measuring perinuclear vs. peripheral CI-MPR fluorescence. Fig. S5 shows control experiments for STxB uptake assays. Online supplemental material is available at http://www.jcb.org/cgi/content/full/jcb.200908057/DC1.