CHC22 Clathrin Diverts GLUT4 from the ER-to-Golgi Intermediate Compartment for Intracellular Sequestration

The Glucose Transporter 4 (GLUT4) in muscle and adipose tissue mediates post-prandial blood glucose clearance by insulin-stimulated GLUT4 translocation to the cell surface from an intracellular GLUT4 storage compartment (GSC). Deregulation of this process establishes insulin resistance and contributes to pathogenesis of type-2 diabetes (T2D). Endocytic pathways targeting GLUT4 to the GSC after insulin-mediated release have been defined extensively in rodent models. Here we map the biosynthetic trafficking pathway leading to initial GSC formation, which is less characterised, and differs between humans and rodents. In human muscle and adipocytes, GSC formation involves the non-canonical isoform of clathrin, CHC22, which does not mediate endocytosis and accumulates at sites of GLUT4 sequestration during insulin resistance. Mice have lost the CLTCL1 gene encoding CHC22, and use conventional CHC17 clathrin for GSC formation by endocytic pathways that do not replace CHC22 function in human GSC formation. Here we report that CHC22 controls GLUT4 transport from the ER-to-Golgi intermediate compartment (ERGIC) in an unconventional secretory pathway, bypassing the Golgi, to form the human GSC. This specialized route for human GLUT4 membrane traffic has relevance for understanding insulin resistance.


CHCs encoded on human chromosome 22 and exhibits unique biochemical
properties and subcellular localization compared to CHC17, indicating a distinct and specific role 6,9 . CHC22 is most highly expressed in muscle cells, though only at 10% of CHC17 levels, and has variable lower expression in other tissues 10 . Depletion of CHC22 from human myotubes and adipocytes abrogates GSC formation and results in loss of insulin-stimulated glucose uptake 6 , indicating that CHC22 mediates pathways to GSC formation that are not replaced by CHC17. Transgenic expression of CHC22 in mouse muscle, which normally expresses only CHC17 clathrin, was found to disrupt GLUT4 trafficking, resulting in expansion of the GSC and over-sequestration of GLUT4 leading to high blood glucose 6 . Although CHC22 interacts with some adaptor molecules that recruit CHC17 to membranes 6,9 and is required for a retrograde transport pathway from endosomes 10 , which has been shown to be important in murine GSC formation 11 , our studies to date indicate that CHC22 must play a role in human GSC formation that is distinct from the role of CHC17 in murine GSC formation.
To further define the role of CHC22 in formation of the human GSC, we developed a model system in which GLUT4 translocation can be easily studied. This was necessitated by limited experimental capacity of available human muscle and adipocyte cell lines and the widely acknowledged problem of detecting endogenous GLUT4 at the cell surface. We produced a stable HeLa cell transfectant (HeLa-GLUT4) expressing human GLUT4 containing a haemagglutinin (HA) tag in the first exofacial loop and a GFP tag at the intracellular carboxyl terminus, a construct that has been extensively not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; characterized and validated for the study of intracellular GLUT4 trafficking in rodent cells 12,13 . HeLa cells were chosen because they have levels of CHC22 comparable to those in human muscle cells 10 and they also express sortilin, another protein important for formation of the GSC 14,15,16 . GLUT4 was sequestered intracellularly in HeLa-GLUT4 cells in the absence of insulin (basal), localized to both peripheral vesicles (arrowheads) and a perinuclear depot (arrows), reminiscent of insulin-releasable vesicles and the vesiculotubular GSC of murine cells, respectively 3,17 (Fig. 1a). Perinuclear GLUT4 colocalized with the GSC marker syntaxin-6 (STX-6) and with CHC22 (Fig. 1b).
After 15 minutes of insulin treatment, GLUT4 was detected at the cell surface with a monoclonal antibody to the HA tag (anti-HA) (Fig. 1a). Additionally, insulin exposure induced phosphorylation of AKT and its substrate AS160, two modifications required for insulin-stimulated GLUT4 translocation 2 (Fig.   1c). Insulin-mediated GLUT4 translocation in the HeLa-GLUT4 cells was quantifiable using a fluorescence activated cell sorter (FACS) to measure the mean fluorescence intensity of surface GLUT4 relative to the total GFP-GLUT4. GLUT4 expression at the plasma membrane increased 2-fold after 15 minutes of insulin treatment (Fig. 1d).
In rodent models (3T3-L1 mouse adipocytes and L6 rat myotubes), GLUT4 released by insulin to the plasma membrane can return by endocytosis to the GSC within 30 minutes 17,18 . Replicating this analysis in the HeLa-GLUT4 model, we labelled surface GLUT4 with anti-HA antibody following insulin stimulation and observed that internalised GLUT4 accessed a perinuclear compartment overlapping with STX-6 with similar kinetics (Supplementary Fig.   not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; 1a, b). Internalised GLUT4 overlapped to the same extent with CHC22 ( Fig.   1e, f) after 10 and 30 minutes of re-uptake. The GSC formed in the HeLa-GLUT4 cells also had properties of the human GSC, in that siRNA-mediated depletion of CHC22 induced dispersion of the GSC and inhibited insulinstimulated GLUT4 translocation (Fig. 1g, h), as previously observed for human muscle cells 10 . Taken together, our results show that HeLa cells stably expressing HA-GLUT4-GFP depend on CHC22 for GLUT4 sequestration and form an insulin-sensitive GSC, which is accessed by GLUT4 recaptured from the cell surface. Thus, the model recapitulates features of the GSC observed for both mouse and human cells and can be used for analysis of pathways involved in GLUT4 trafficking in human cells. We note that during the course of this work, other laboratories developed and validated similar models of insulin-dependent GLUT4 translocation in HeLa cells 19,20 .
Our previous studies identified a function for CHC22 in retrograde transport from endosomes, but only limited overlap between CHC22 and the endosomal marker Rab 9 was observed 10 . Suspecting that CHC22 might operate in an additional pathway leading to GSC formation, we analysed both the HeLa-GLUT4 cells and differentiated human myotubes (LHCNM2) for overlap between CHC22 and markers of the secretory pathway, as the biosynthetic origins of the GSC are not fully characterised. For this study, we used a new polyclonal antibody specific for CHC22 and not reactive with CHC17 that had stronger activity in immunofluorescence staining than our previously produced anti-CHC22 antibodies 10,21 . In both HeLa-GLUT4 and differentiated LHCNM2 cells, we observed significant co-localization of not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; CHC22 and two markers of the ER-to-Golgi Intermediate Compartment (ERGIC), namely p115 and ERGIC-53 ( Fig. 2a-d), while there was no significant co-localization of CHC22 with endoplasmic reticulum (ER) markers β-COP and calreticulin ( Supplementary Fig. 2a-c) in either cell type. GLUT4 internalised after insulin stimulation also showed time-dependent colocalization with ERGIC-53, indicating that the GSC, where recaptured GLUT4 accumulates, is closely associated with the ERGIC (Supplementary Fig. 2d, e). Considered together, these data are consistent with previous mapping of CHC22 activity to retrograde transport from late endosomes 10 , and suggest a further role for CHC22 possibly at the ERGIC or the cis-Golgi. To map CHC22 relative to these compartments of the secretory pathway and the GSC, we treated HeLa-GLUT4 and LHCNM2 myotubes with Brefeldin A (BFA) ( Supplementary Fig. 3a, 3b). We previously showed that BFA does not cause CHC22 to dissociate from intracellular membranes 9 and others have shown that the GSC is not affected by BFA 22 . After BFA treatment of either cell type, CHC22 co-distributed with the ERGIC marker p115 and segregated away from the persisting GSC, confirming the BFA-resistance of the GSC and supporting the association of CHC22 with ERGIC membrane. This result further suggests that CHC22 is not a final component of the stable GSC, although involved in GSC formation 6 . not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; To establish whether CHC22 has functional activity where it localizes with ERGIC markers, we tested whether CHC22 plays a role in the life-cycle of the facultative intracellular pathogen Legionella pneumophila (Lp), which hijacks membrane from the early secretory pathway to create an Lp-containing vacuole (LCV), in which it can replicate. Upon infection, Lp secretes ~300 effector proteins, through a type IV secretion system, some of which enable recruitment of ER proteins Sec22b, Rab1 and Arf1 to the LCV 23 , which are needed for its formation. The mature LCV retains ER-like properties, including the lumenal proteins calnexin, BiP and calreticulin and does not acquire Golgi markers [23][24][25] . Given that the LCV is derived from the ER, and that CHC22 localizes to a pathway emerging from the ER, we tested whether CHC22 associates with membranes involved in LCV formation by transient expression of GFP-tagged CHC22 or GFP-tagged CHC17 in A549 human lung adenocarcinoma cells prior to Lp infection. CHC22, but not CHC17, was associated with the membranes surrounding the LCV (Fig. 3a, b). CHC22 did not localize with the isogenic avirulent Lp mutant ΔdotA, which enters cells but lacks a functional secretion system and cannot deploy effectors to induce a replication-supportive vacuole (Fig. 3a, b).
To address whether CHC22 is involved in transfer of membrane to the LCV, we treated A549 cells with siRNA targeting CHC22 prior to infection. The resulting CHC22 down-regulation significantly compromised recruitment of Sec22b to the bacterial vacuole at 1 hour post-infection (Fig. 3c, d), suggesting defective vacuole maturation. This was confirmed by assessing bacterial replication eight hours after infection with WT or ΔdotA Lp strains not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; following CHC22 or CHC17 silencing. CHC22 silencing reduced the proportion of vacuoles containing >4 Lp by more than 9-fold, while CHC17 silencing reduced vacuoles with >4 Lp by approximately 2-fold (Fig. 3e), suggesting a need for CHC22 to form a replicative vacuole and a possible role for CHC17 during bacterial uptake. The latter conclusion is supported by the observation that CHC22 down-regulation increased the percentage of vacuoles with 1 Lp, showing that bacterial entry was still occurring, but vacuoles with 1 Lp comprised a smaller percentage in cells silenced for CHC17 infected with an equivalent number of bacteria. Infection of cells transfected with siRNA targeting CHC22 with an equivalent number of avirulent ΔdotA Lp showed that these bacteria could also enter cells, with 94% vacuoles observed harbouring only 1 Lp (Fig. 3e). Together these data suggest that Lp hijacks CHC22 to acquire host early secretory membrane needed for maturation of a replication-competent LCV, and that Lp effectors might interact with CHC22 or its partners. We did not see recruitment of GLUT4 to the LCV (Supplementary Fig. 4a, b), indicating that the bacteriainduced pathway does not contain all the components needed for formation of a GSC. However, these data suggest CHC22 has an active role in membrane traffic emerging from the ER.
A previous study tracking transiently expressed GFP-tagged GLUT4 suggested that, in 3T3-L1 adipocytes, newly synthesized GLUT4 accesses the GSC from the secretory pathway, before reaching the plasma membrane 26 . Based on CHC22 localization and its role in membrane transport from the early secretory pathway during LCV maturation, we not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; hypothesized that, in humans, CHC22 controls GSC formation from the secretory pathway directly from the ERGIC. The potential existence of this pathway was supported by the earlier demonstration that overexpression of a dominant negative fragment of p115 abrogated the GLUT4 insulin response in 3T3-L1 cells 27 . To investigate the involvement of secretory pathways in GSC formation, we depleted p115, CHC22, CHC17 and GM130 from HeLa-GLUT4 cells using siRNA (Fig. 4a), assessed the morphology of the GSC by microscopy (Fig. 4b, c) and evaluated the impact on insulin-induced GLUT4 translocation by FACS (Fig. 4d). Silencing of p115 or of CHC22 abrogated or reduced perinuclear localization of GLUT4 and increased GLUT4 dispersion in the cell periphery, with down-regulation of each affecting the distribution of the other (Fig. 4b). Insulin-stimulated GLUT4 translocation was lost from cells silenced for p115 or CHC22 (Fig. 4d), while CHC17 silencing had a partial effect on GLUT4 translocation (Fig. 4d), consistent with previous observations for CHC22 and CHC17 down-regulation 6 . Notably, however, siRNA-mediated down-regulation of the cis-Golgi protein GM130, functionally confirmed by reduction of alkaline phosphatase secretion 28 (Fig. 4e), did not abrogate insulin-stimulated GLUT4 translocation ( Fig. 4d) or affect the distribution of perinuclear GLUT4 (Fig. 4c). These results suggest p115 and CHC22 participate in a step of GSC formation that diverts GLUT4 from the ERGIC and bypasses transport through the Golgi.
For another indication of how formation of the GSC depends on CHC22 and p115, these proteins and CHC17 were depleted by siRNA ( Supplementary Fig.   5a-c) and levels of GSC-associated proteins, including insulin-regulated not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; amino peptidase (IRAP), sortilin 14 and Golgi-localized γ ear-containing ARFbinding protein 2 (GGA2) 15,26,29 , were assessed by quantitative immunoblotting (Fig. 5a-c). Analysis of statistically significant changes indicated that CHC22 depletion but not CHC17 depletion reduced cellular levels of GLUT4, sortilin, and GGA2. Depletion of p115 strongly reduced IRAP levels. Depletion of CHC17 increased CHC22, as previously observed 6,10 and also slightly increased GGA2 levels. The differential effects of p115, CHC22 and CHC17 depletion suggest that each of these membrane traffic mediators contribute differently to GSC formation. A biochemical association between p115 and IRAP has already been established 27 and their association occurs whether or not cells have a GSC, as IRAP has a wider cellular expression pattern than GLUT4 30 . Based on the demonstration of lumenal interactions between sortilin and IRAP, sortilin and GLUT4 15 and IRAP and GLUT4 30 , it has been proposed that IRAP, which initially traffics with p115 27 , becomes associated with sortilin and GLUT4 during the formation of the GSC 30 , at which point the GSC becomes segregated from the constitutive secretory pathway. Sortilin is known to recruit the GGA2 adaptor 31 , which has been previously implicated in GLUT4 traffic to the GSC 16,26,29 and preferentially binds to CHC22 over CHC17 6 . Thus, CHC22 could be recruited as a membrane coat that coalesces the players for GSC formation and stabilises their interaction for forming a specialised compartment. Our data on CHC22 localization presented here suggests that the process of initial sorting to the GSC, triggered by the meeting of the p115-IRAP complex with GLUT4 and sortilin, occurs at the ERGIC (Fig 5d), which is consistent with previous studies suggesting direct GLUT4 diversion from the secretory pathway to the not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; GSC 26 in a p115-dependent step 27 . Interestingly, p115 is not recruited to the Lp early replicative vacuole 23 and it has been proposed that one of Lp's secreted effectors can functionally replace p115 to facilitate recruitment of vesicles from the early secretory pathway. CHC22 may also be recruited by the bacterial effector, while GLUT4 apparently is not. The differential dependence of GLUT4 and alkaline phosphatase (and other proteins 32,33 ) on p115 for export suggests the ERGIC is a sorting station for several secretory pathways, some of which, such as autophagy 34 , bypass conventional secretion.
Our proposed pathway of GSC formation from the ERGIC (Fig. 5d) is consistent with earlier analysis of GLUT4 glycosylation. The rate that the carbohydrate side chains on newly synthesized GLUT4 become resistant to cleavage by Endoglycosidase H was shown to be considerably delayed (3 days vs 1 hour) compared to that of GLUT1, which is constitutively expressed on the cell surface 35,36 . This could be explained by direct GLUT4 diversion to the GSC from the ERGIC, followed by a process of carbohydrate maturation that depends on secretion and recapture through a retrograde recycling compartment 17 with low-level access to enzymes of the medial and trans-Golgi. Our model is also consistent with the reported ERGIC localization of the TUG protein involved in GLUT4 vesicle retention 37 and potentially with the recent discovery of Sec16A as a target for Rab10 activity 38 . Sec16A seems to participate in GLUT4 membrane traffic independently of its conventional COPII partners 38 . Its presence at ER exit sites and documented interaction not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; with p115 39 suggests that Sec16A could be sequestered from the ERGIC during the pathway of GSC formation suggested here. We propose that for both humans and mice, GSC formation is initiated by direct diversion from the ERGIC to sequester GLUT4 away from the conventional secretory pathway in a step that involves p115 and IRAP. Then, once GLUT4 is released by insulin to the cell surface, it is returned to the GSC by retrograde transport from the endosome-TGN recycling pathway, which can expose GLUT4 to carbohydrate modification. The difference between human and murine GLUT4 traffic is that humans have CHC22 clathrin, which can localize to the ERGIC and thereby capture GGA2-bound sortilin-IRAP-GLUT4 complexes as they coalesce in the ERGIC to enhance GLUT4 delivery to the GSC (Fig. 5d). Thus the murine pathway is likely to be more dependent on the recapture and recycling from the plasma membrane to keep the GSC replenished, while humans have a more robust route to their GSC from the ERGIC. For both murine and human cells, CHC17 clathrin is needed for GLUT4 endocytosis from the plasma membrane and, in murine cells, CHC17 is the likely mediator of the retrograde endosome-TGN recycling pathway that involves AP1 40 and GGA2 41 . Our previous studies of CHC22 suggested it also functions in this recycling pathway because CHC22 depletion affected mannose-6-phosphate receptor (M6PR) traffic 10 , which depends on retrograde transport. However, our demonstration here that CHC22 depletion reduces cellular levels of GGA2 could also account for effects on the M6PR pathway 41,42 so it is not clear if CHC22 plays this second role in human GLUT4 transport in addition to its role in GSC formation from not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; the ERGIC. CHC17 does not localize to the ERGIC and it cannot replace CHC22 function in human GSC formation, indicating that human cells depend more heavily on the ERGIC-derived pathway and cannot maintain a GSC with only the CHC17-dependent endocytosis and recycling pathway, as occurs in murine cells. We reiterate here that CHC22, at endogenous levels of expression, does not have a role in receptor-mediated endocytosis from the plasma membrane 9,10 . The evolutionary retention of CHC22 in humans, following the gene duplication process that generated it, may have been advantageous for tighter control of GLUT4 sequestration in order to maintain high levels of blood glucose for large brain function during starvation. Loss of CHC22 in mice may have been favoured by a need to regularly clear blood glucose generated by a nibbling diet (and no need to feed a large brain). Thus, the extra role of CHC22 in human GSC formation may provide too much GLUT4 sequestration as we transition to a nibbling diet. Insulin resistance is sensitive to levels of available GLUT4 43 , so by accumulating on the expanded GSC, which occurs as a result 6 , CHC22 could potentially exacerbate human insulin resistance. not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
Data expressed as mean ± SEM, N=9, 10,000 cells acquired per experiment.
Merged images in (b) and (c) show red/green overlap in yellow, red/blue overlap in magenta, green/ blue overlap in turquoise, and red/green/blue overlap in white.     not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was . http://dx.doi.org/10.1101/242941 doi: bioRxiv preprint first posted online Jan. 4, 2018; Overlap not peer-reviewed) is the author/funder. All rights reserved. No reuse allowed without permission.