Surfaceome Profiling Identifies Basigin-Chaperoned Protein Clients

Abstract

The surface proteome or “surfaceome” is a critical mediator of cellular biology, facilitating cell-to-cell interactions and communication with extracellular biomolecules. Constituents of the surfaceome can serve as biomarkers for changing cell states and as targets for pharmacological intervention. While some pathways of cell surface trafficking are well characterized to allow prediction of surface localization, some non-canonical trafficking mechanisms do not. Basigin (Bsg), a cell surface glycoprotein, has been shown to chaperone protein clients to the cell surface. However, understanding which proteins are served by Bsg is not always straightforward. To accelerate such identification, we applied a surfaceome proximity labeling method that is integrated with quantitative mass spectrometry-based proteomics to discern changes in the surfaceome of hepatic stellate cells that occur in response to the genetic loss of Bsg. Using this strategy, we observed that the loss of Bsg leads to corresponding reductions in the cell surface expression of monocarboxylate transporters MCT1 and MCT4. We also found that these relationships were unique to Bsg and not found in neuroplastin (Nptn), a related family member. These results establish the utility of the surfaceome proximity labeling method to determine clients of cell surface chaperone proteins.

Graphical Abstract

TOC: A surfaceome proximity labeling method allowed for proteomic profiling a cell surface proteins. In Bsg but not Nptn deficient cell lines, we detected the specific reduction of MCT1 and MCT4 at the cell surface, while the rest of the surfaceome remained relatively unchanged. This emphasizes Bsg’s role as a specific chaperone for MCTs within the cell.

Introduction

The total collection of cell surface proteins (CSPs), or “surfaceome,” mediates biological function. CSPs mediate communication with the extracellular environment and can serve as targets for therapeutics^[1]. The surfaceome is dynamic, varying significantly among different cell types and cell states^[2]. Whereas many CSPs contain signal peptide sequences that facilitate prediction of cell surface localization, many others lack such features and are trafficked via non-canonical mechanisms^[3]. Thus, developing an empirical understanding of the surfaceome remains an important, but challenging endeavor. CSPs are expressed in relatively low abundance compared to the rest of the proteome, are dynamically expressed in response to changing environments, and possess hydrophobic transmembrane domains, which can pose challenges for isolation.

Basigin (Bsg; also known as EMMPRINN or CD147 ^[4]) is a ubiquitously expressed cell surface N-linked glycoprotein. Beyond its commonly known roles as a blood group antigen^[5], an inducer of matrix metalloproteinase activity^[4g], or as a driver of liver fibrosis^[6], it is also known to act as a chaperone that escorts proteins from the Golgi apparatus to the cell surface^[7] (Fig. 1A). Bsg and its structurally similar family members, neuroplastin (Nptn) and embigin (Emb), have been known to act as ancillary proteins that facilitate the cell surface localization of monocarboxylate transporters (MCTs)^[8] (Fig. 1B). MCTs comprise the solute carrier family 16 (SLC16) of proteins, of which there are fourteen total members (SLC16A1-14)^[9]. Despite the absence of glycan post-translational modifications, conventional N-terminal peptide signaling motifs, and the relatively low number of residues exposed to the cell surface (8% and 14% for MCT1 and MCT4 respectively^[10]) MCTs are often found on cell surfaces^[11]. Bsg specifically has been implicated in the cell surface localization of the lactate transporters MCT1 and MCT4 in several cell types including pancreatic epithelial carcinoma^[12], colon adenocarcinoma^[13], and lung fibroblast cells^[14] where the removal of Bsg leads to the reduced cell surface expression of these proteins. Although Bsg and Nptn have been shown to be widely distributed amongst various cell types, Emb is less so, with MCT1 association shown in only red blood cells^[8b]. Nptn has also been implicated to be the preferred ancillary protein for MCT2 over other MCTs, although the evidence is mainly circumstantial^[15].

Figure 1. Surfaceome tagging allows surveillance of Basigin (Bsg)-mediated monocarboxylate transporter (MCT) trafficking.

(A) Within the Golgi apparatus, MCT proteins associate with Bsg, which is trafficked to the cell surface via its signal peptide. (B) Cartoon depiction of Bsg and homologous members within the superfamily, neuroplastin (Nptn) and embigin (Emb). Purple bands indicate transmembrane domains. Orange bands indicate signal peptides. Pink dashes indicate sites of putative N-linked glycosylation. (C) Scheme of surfaceome labeling. GGGYC peptide-tagged cholesterol incorporates into the cell surface, and APEX2-LPETG is conjugated via the engineered sortase, eSrtA. Upon addition of biotin-phenol (BP) and H₂O₂, the cell surface-tethered APEX2 generates short-lived (< 1 ms) and reactive biotin radicals that covalently tag proximal (<20 nm) proteins.

Given the important roles of MCTs in regulating the transport of cell nutrients and waste, and their upregulation in cancers^[16], understanding how the presence or absence of ancillary proteins may affect their expression and cell surface localization is key to understanding cellular metabolism.^{[7a, 8b, 8c, 16a, 17]} However, it is currently unclear whether Bsg can serve as an ancillary trafficking protein towards other lesser known MCTs or to other proteins outside of the SLC16 family. Furthermore, it is poorly understood whether homologous family members Nptn or Emb can play compensatory roles for Bsg in its absence.

We recently developed an approach to identify CSPs using cell surface engineering^[18] and a “baitless” proximity tagging^[19] strategy (Fig. 1C). In this method, an engineered ascorbate peroxidase enzyme (APEX2) is localized to the cell surface via a synthetic GGGYC peptide-modified cholesterol tether and an engineered sortase enzyme (eSrtA)^[20] to facilitate the biotinylation of nearby electron-rich residues^[19]. These biotin handles can be used to enrich the proteins for subsequent identification by quantitative mass spectrometry (MS)-based proteomics. Importantly, this method does not rely on the presence of specific post-translational modifications (PTMs) to reduce biases in the enrichment method. To facilitate multiplexed analysis, we used isobaric tandem mass tags (TMT) to tag tryptic peptides with isobaric tags that can be used to generate distinct signals at the MS3 stage^[21]. This allows for the quantitative comparison of up to 16 multiplexed samples injected within a single run of the mass spectrometer.

Here, as a demonstration of the potential of baitless proximity tagging, we applied this surfaceome profiling approach to agnostically surveil changes in surfaceome composition and expedite the determination of protein clients for Bsg-related escorting activities in LX-2 human hepatic stellate cells, the cell types responsible for the initiation of fibrosis in the liver. Using a Bsg knockdown cell line, we discovered that the genetic loss of Bsg leads to reduced cell surface levels of the specific MCT family members, MCT1 (SLC16A1) and MCT4 (SLC16A3). We observed that this reduction was limited to these two proteins across the surfaceome, and that the loss of Nptn did not significantly impact the surfaceome composition of these cells. Our results suggest that Bsg exquisitely serves as a chaperone protein to MCT1 and MCT4 in LX-2 cells.

Results

Using unenriched TMT proteomics, we determined that wild-type (WT) LX-2 hepatic stellate cells display notably higher levels of Bsg and Nptn relative to Emb (Fig. S1). Hence, we focused our subsequent efforts towards the study of these two homologous proteins. To assess the roles of Bsg and Nptn in regulating CSP compositions, we used CRISPR/Cas9 to generate stable knockdowns of Bsg (Bsg^MUT), Nptn (Nptn^MUT) in LX-2 cells. We also generated a double Bsg/Nptn (Bsg^MUT/Nptn^MUT) mutant to evaluate potential compensatory roles. By western blotting, we observed 94%, 88%, and 95%/36% reductions of Bsg, Nptn, and Bsg/Nptn in whole cell lysates of each of the mutant cell lines, respectively, compared to WT cells (Fig. 2A, Fig. S2). Interestingly, we observed a slight yet significant increase in the expression of Bsg in the Nptn^MUT (13%) cell line and of Nptn in Bsg^MUT cells (34%). We also performed RT-qPCR analysis to assess changes in Bsg, Nptn, MCT1, and MCT4 transcript levels. As expected, CRISPR-Cas9 editing resulted in notable reduction of transcript levels of Bsg, Nptn, or Bsg and Nptn in corresponding mutant cell lines (Fig. 2B). There were no significant differences in MCT1 or MCT4 transcript levels in the mutant cells relative to WT cells.

Figure 2. Characterization of Bsg in WT or mutant LX-2 human hepatic stellate cells.

(A) Western blotting (left panel) and densitometry (right panel) of wild type (WT) LX-2 cells and mutant cell lines reveals knockdown of Bsg and Nptn in the Bsg^MUT, Nptn^MUT and Bsg^MUT/Nptn^MUT cell lines. Bsg expression was slightly upregulated in the Nptn^MUT cell lines, and conversely, Nptn was slightly upregulated in the Bsg^MUT cell lines. Densitometry was normalized using GAPDH signals, and the horizontal dashed line marks levels of Bsg and Nptn in WT cells. Error bars indicate standard deviation of two independent replicates. (B) RT-qPCR data measuring the abundance of Bsg, Nptn, MCT1, and MCT4 transcripts within WT and mutant cells. Asterisks indicated significant differences in abundance from WT levels as determined by Welch’s t-test. As expected, Bsg transcript levels were significantly decreased in Bsg^MUT and Bsg^MUT/Nptn^MUT cell lines, and Nptn levels were significantly decreased in Nptn^MUT and Bsg^MUT/Nptn^MUT cell lines. MCT1 and MCT4 transcript levels were not significantly perturbed by genetic manipulation.

We initially screened the WT LX-2 cells for compatibility with the surfaceome profiling approach. In our previous work, we determined that optimization of biotin-phenol incubation periods in each cell line was crucial in order to select CSPs over intracellular proteins^[22]. Thus, we varied biotin-phenol incubation times (1-5 min) during live cell labeling, and the resulting cell lysates were processed by ultracentrifugation to separate membrane-bound and intracellular components. Western blotting was then used to probe for the presence of the biotin-tag. We observed the highest level of localized cell surface biotinylation with 1 minute of biotin-phenol incubation (Fig. S3). Using these optimized conditions, we confirmed that the surfaceome profiling approach results in biotinylation signals that primarily localize to the cell surface by confocal microscopy (Fig. 3A, Fig. S4). Consistent with the requirement for all components of the surfaceome profiling method, omission of the cholesterol lipid, APEX2-LPETG, or eSrtA during the labeling reaction led to reduced cell surface and overall biotinylation signals.

Figure 3. Surfaceome tagging of WT and mutant LX-2 cells.

(A) Confocal microscopy images of WT LX-2 cells labeled by the surfaceome profiling technique. Biotin signals (red) were mostly located on cell surfaces that surround nuclei (blue). The absence of a labeling component (e.g. lipid, APEX2-LPETG, or eSrtA) resulted in abrogation of biotinylation signals, compared to the full reaction (Rxn). (B) Volcano plots depicting enrichment of surfaceome labeled proteins from WT LX-2 and mutant cell lines over negative controls (conditions lacking APEX2-LPETG). Proteins with ≥2 unique peptides that were significantly enriched (p < 0.05) by the surfaceome capture method and annotated as a cell surface protein (CSP) in UniProt are present in the upper right quadrant. Representative canonical CSPs are highlighted.

We proceeded to perform quantitative MS-based proteomics to evaluate the changes in surfaceome content among the different cell lines. After filtering the resulting datasets for proteins with ≥ 2 unique peptides and those classified as CSPs using Uniprot (see SI methods), we identified 121, 116, 131, and 114 significantly enriched CSPs in the WT, Bsg^MUT, Nptn^MUT, and Bsg^MUT/Nptn^MUT conditions respectively. (Fig. 3B) MCT1 and MCT4 were among the most highly enriched proteins within the WT dataset, and canonical membrane and extracellular matrix protein families including integrins (ITGB1, ITGA5, ITGA3), collagens (COL1A1, COL5A1), cadherins (CDH13), and annexins (ANXA5, ANXA2, ANXA1) were identified in both WT and mutant datasets.

We then compared the enriched CSPs among the WT and mutant cell lines. The overall composition of the surfaceome remained relatively unchanged in these cell lines with a few exceptions. As expected, reduced Bsg and Nptn signals were observed in their respective mutant cell lines relative to WT cells (Fig. 4A). We also observed that MCT1 and MCT4 were less enriched in the Bsg^MUT and Bsg^MUT/Nptn^MUT conditions (Fig. 4A; Fig. 4B), suggesting that MCT1 and MCT4 were less abundant at the surfaces of these cells. This observation, however, was not replicated in Nptn^MUT cells, suggesting that Nptn does not fulfill the same role as Bsg in chaperoning MCT1 and MCT4 to the cell surface.

Figure 4. Comparison of CSPs in WT and mutant LX-2 cell lines.

(A) Representative volcano plots of relative enrichment of CSPs between surfaceome labeled WT and mutant LX-2 cell lines across biological duplicates. Areas close to the vertical dashed line mark enrichment between cell lines, and the areas above the horizonal dashed line mark p < 0.05. Proteins within the upper right quadrant were significantly enriched in the WT over compared condition. Orange proteins were consistently identified across replicates. The most significantly enriched and prioritized proteins for study are enclosed within a dashed orange box. (B) Average raw TMT abundance values of Bsg and Nptn in their respective cell lines, as well as MCT1 and MCT4 across all mutant cell lines. Bsg and Nptn were enriched significantly in WT cells over respective mutant cell lines. MCT1 and MCT4 were significantly enriched in WT over Bsg^MUT cells, but not over Nptn^MUT cells. Error bars indicate standard deviation. Asterisks indicate statistical significance (p < 0.05) as determined by multiple unpaired t-test.

To assess whether total expression or cell surface trafficking of these proteins was lost, we evaluated total MCT1 and MCT4 expression by performing western blotting of total whole cell lysates and of intracellular soluble versus membrane fractions using ultracentrifugation. We used the CSP integrin α_Vβ₃ as a control cell surface-localized protein in our experiments^[23]. Upon ultracentrifugation, we observed the preferential separation of MCT1 and MCT4 exclusively into the membrane fraction in all cell lines regardless of genetic manipulation, similar to α_Vβ₃ (Fig. 5A; Fig. S5; Fig. S6). However, we observed an overall reduction in MCT1 and MCT4 expression in whole cell lysates as well as the membrane fraction of Bsg^MUT and Bsg^MUT/Nptn^MUT cell lines relative to WT cells. The loss of total MCT1 and MCT4 protein levels is not due to changes in transcript levels, and are likely due to engoenous degradation mechanisms related to Bsg^[24]. We did not observe changes of MCT1 and MCT4 signals in the Nptn^MUT cells, further supporting the observation that Nptn does not perform the same ancillary function that Bsg does in MCT1 or MCT4 trafficking (Fig. 5B).

Figure 5. MCT expression and localization in WT and mutant LX-2 cell lines.

(A) Western blotting of protein signals in LX-2 whole cell lysates (L), as well as cytosolic soluble (S) and membrane (M) fractions separated by ultracentrifugation. The control cell surface protein integrin α_Vβ₃ exhibits preferential localization to M over S fractions. MCT1 and MCT4 signals are consistently and preferentially located in M over S fractions. (B) MCT1 and MCT4 signals in the lysate or membrane fractions were normalized to stain-free gel image intensities (Fig. S5). (C) Sequence alignment of Bsg and Nptn transmembrane regions. Sequence mismatches detected by ClustalW algorithm are highlighted in red. Homology of the transmembrane regions putatively responsible for MCT binding was 60%, whereas overall protein homology was 39%. Residues at both the N- and C- termini regions of the transmembrane domain are mismatched (red), potentially preventing interactions with MCT1 and MCT4.

Discussion and Conclusions

Previous work has demonstrated that Bsg can serve as an ancillary protein for MCT1 and MCT4 in pancreatic epithelial adenocarcinoma, colon adenocarcinoma, and lung fibroblast cells^[8a] by forming tightly bound heterocomplexes observed via immunoprecipitation, confocal microscopy, or chemical crosslinking^{[8a, 8c]}. These associations were dependent primarily on the transmembrane region of Bsg, which shares some (60%) homology with the Nptn transmembrane region (Fig. 5C). In some instances, Nptn was found to compensate for the loss of Bsg (in trafficking PMCA).^[8e] While these low throughput experiments provided insight into specific Bsg- or Nptn-mediated interactions, it remained unclear whether Bsg or Nptn could serve as general protein chaperones for other CSPs.

By implementing a radical-mediated baitless proximity tagging technique to enrich and identify CSPs, we detected the reduced enrichment of MCT1 and MCT4 in LX-2 hepatic stellate cells with reduced Bsg but not Nptn expression. Notably, the lack of glycan modifications on MCT1 and MCT4 render them invisible by glycan-based enrichment methods. We observed that these changes in cell surface enrichment do not correlate with changes in transcript levels, suggesting that changes in cell surface localization of MCT1 and MCT4 result from post-translational processing^[8a]. Overall, these results suggest that the chaperone functions of Bsg in these cells are non-redundant with Nptn, and that most of these activities are directed towards MCT1 and MCT4 isoforms of the SLC16A family. This non-redundancy is likely due to the key differences in the specific flanking residues within the transmembrane regions (Fig. 5C), as previous work has shown that the formation of the Bsg-MCT1 complex is dependent primarily on the N- and C- termini of the transmembrane region of Bsg^[17a]. Alignment of the transmembrane regions of Bsg and Nptn using the ClustalW algorithm^[25] (Fig. 5C, Fig. S7) indicates 60% homology between Bsg and Nptn within the transmembrane region and 39% homology overall, however residues at both the N- and C- termini are mismatched, potentially preventing interactions. While we focused primarily on MCT1 and MCT4 as the most significantly reduced proteins (enclosed within the orange box Fig. 4A), we also observed the loss of other proteins at the cell surface within the Bsg^MUT/Nptn^MUT cell line. Presumably, these are proteins that can be chaperoned by either Bsg or Nptn.

The surfaceome is a complex environment of proteins involved in a plethora of biological functions. Whereas most CSPs undergo trafficking to the cell surface through canonical signal-peptide based pathways, MCT1 and MCT4 rely on the expression of the glycoprotein Bsg for proper translocation. Our data suggests that Bsg exhibits exquisite specificity for MCT1 and MCT4, and that this chaperoning activity is specific to Bsg. Our future work will assess changes in metabolic activity in the mutant hepatic stellate cells, given that MCTs are responsible for the transport of short-chain fatty acids such as lactate and pyruvate through the plasma membrane^[26], which relies on proper localization. Disruption of the Bsg-MCT4 interaction has been shown to inhibit tumor progression in glioblastoma cells^[27], hinting at the importance the role of Bsg as a chaperone. This function is of particular interest given the implications in pathologically relevant pathways, especially in the activation of hepatic stellate cells.