Exposure to Melan-A/MART-126-35 tumor epitope specific CD8+T cells reveals immune escape by affecting the ubiquitin-proteasome system (UPS)

Efficient processing of target antigens by the ubiquitin-proteasome-system (UPS) is essential for treatment of cancers by T cell therapies. However, immune escape due to altered expression of IFN-γ-inducible components of the antigen presentation machinery and consequent inefficient processing of HLA-dependent tumor epitopes can be one important reason for failure of such therapies. Here, we show that short-term co-culture of Melan-A/MART-1 tumor antigen-expressing melanoma cells with Melan-A/MART-126-35-specific cytotoxic T lymphocytes (CTL) led to resistance against CTL-induced lysis because of impaired Melan-A/MART-126-35 epitope processing. Interestingly, deregulation of p97/VCP expression, which is an IFN-γ-independent component of the UPS and part of the ER-dependent protein degradation pathway (ERAD), was found to be essentially involved in the observed immune escape. In support, our data demonstrate that re-expression of p97/VCP in Melan-A/MART-126-35 CTL-resistant melanoma cells completely restored immune recognition by Melan-A/MART-126-35 CTL. In conclusion, our experiments show that impaired expression of IFN-γ-independent components of the UPS can exert rapid immune evasion of tumor cells and suggest that tumor antigens processed by distinct UPS degradation pathways should be simultaneously targeted in T cell therapies to restrict the likelihood of immune evasion due to impaired antigen processing.


Selection of melanoma cell clones resistant to CTL-mediated lysis by Melan-A/MART-1 26-35specific CTL.
To identify unknown immune evasion mechanisms allowing melanoma cells to circumvent recognition by specific CTL, the human HLA-A * 0201-positive melanoma cell line UKRV-Mel-15a expressing the transmembrane tumor antigen Melan-A/MART-1 20,24,25 was co-incubated with an HLA-A*0201-restricted Melan-A/MART-1 26-35 -specific CTL clone. Surviving UKRV-Mel-15a cells were exposed to two or three rounds of CTL co-culture selecting for resistant tumor cell clones. To exclude resistance due to HLA or antigen down-regulation, UKRV-Mel-15a cells were analyzed for HLA-A*0201 and Melan-A/MART-1 expression by flow cytometry. As shown in Fig. 1A, three rounds of co-culture resulted in considerable HLA-A*0201 down-regulation. We therefore decided to apply only two rounds of CTL co-culture for the selection of resistant melanoma clones, because under these conditions the HLA-A*0201 and Melan-A/MART-1 expressions of the surviving clones was similar or even higher than those of the parental UKRV-Mel-15a cell line (Fig. 1A).
Down-regulation of constituents of the antigen processing machinery in tumor cells and concomitant impaired epitope generation and presentation is a well-known evasion mechanism described for a variety of different human tumors 22,23,26,27 . However, as mentioned above, the expression of these components can be restored following IFN-γ treatment. Importantly, the expression levels of TAP1 and TAP2 as well as of the three proteasome immunosubunits in UKRV-Mel-15a clones were comparable to those of the non-selected parental UKRV-Mel-15a cells (Fig. S1). Only the clone 30 exhibited some impairment in the inducible expression of the i-subunits, as evidenced by its incapacity to up-regulate β 1i/LMP2, β 2i/MECL and β 5i/LMP7 in response to IFN-γ . This suggests that the observed resistance of the selected UKRV-Mel-15a clones to CTL-mediated lysis was not due to an altered IFN-γ signaling pathway. This notion is supported by the observation that IFN-γ exposure failed to modify the phenotype of the UKRV-Mel-15a clones 6, 18 and 30 (Fig. 1E).
Deregulation of ERAD components in resistant melanoma cell clones. To determine which of the UPS components and its associated pathways was responsible for the down-regulation of Melan-A/MART-1 [26][27][28][29][30][31][32][33][34][35] epitope presentation; UKRV-Mel-15a clones were subjected to microarray gene expression analysis focusing on more than 1000 components of the antigen processing machinery. Interestingly, all three selected UKRV-Mel-15a clones displayed considerable alterations in the expression levels of a set of mRNAs encoding p97/VCP, VIMP, HERP and Derlin-1 (Table 1) and thus proteins whose functions are related to or directly connected with the ER-dependent degradation (ERAD) pathway 28 . Because cellular protein levels strongly depend on the rate of synthesis and their half-lives, expression levels of mRNAs and corresponding detectable protein levels do not necessarily correlate. Therefore, we next analyzed the expression levels of the affected ERAD-associated proteins by Western blot. As shown in Fig. 2A,B, all three resistant UKRV-Mel-15a clones revealed a strong down-regulation of the ERAD components p97/VCP, which is essential for the extraction of poly-ubiquitinated substrates from the ER 29 , and VIMP, which is a seleno-protein associated with p97/VCP as well as with the membrane protein Derlin-1 17 . In contrast, slightly increased expression levels were found for HERP and Derlin-1 ( Fig. 2A,B). Importantly, there was no substantial loss of the Melan-A/MART-1 antigen expression in the UKRV-Mel-15a  30 (Fig. 3A). Therefore, for proteasome-dependent antigen processing, the Melan-A/MART-1 protein must be retro-translocated into the cytosol. We next aimed to evaluate the impact of VIMP and HERP on Melan-A/MART-1 26-35 epitope generation independently of the Melan-A/MART-1 endogenous expression level. To this end, both of these ERAD components were knocked down by siRNA in the Ma-Mel-91-melanoma cell line (Melan-A/MART-1 − /HLA-A * 0201 + ) which was subsequently engineered to express the Melan-A/MART-1 antigen. Strikingly and as illustrated in Fig. 3B, gene silencing of either VIMP or HERP resulted in a substantial reduced presentation of the Melan-A/ MART-1 26-35 epitope. Because p97/VCP expression was also found to be significantly reduced in our selected UKRV-Mel-15a clones 6, 18 and 30, we next tested its involvement in Melan-A/MART-1 26-35 epitope generation by expressing a dominant negative mutant of p97/VCP (p97QQ), which abrogates cellular p97/VCP function 16 . Interestingly, although the expression of p97QQ had virtually no effect on cell viability (data not shown), it almost completely abrogated Melan-A/MART-1 26-35 epitope generation, whose levels, under these conditions, were quite similar to those observed when proteasomes were inhibited with epoxomicin ( Fig. 3C). Accordingly, CHX-based chase experiments show that p97/VCP silencing resulted in a significant stabilization of the Melan-A/ MART-1 protein (Fig. 3D). It is to note, that neither the down-regulation of VIMP or HERP by siRNA nor the plasmid-mediated transfection of p97/VCP or p97QQ substantially affected the expression level of the Melan-A/ MART-1 full-length protein (Fig. S2).

Melan-A/MART-1 26-35 epitope recognition is rescued by overexpression p97/VCP. The obser-
vation, that the inhibition and/or gene silencing of VIMP and p97/VCP almost completely abrogated Melan-A/ MART-1 26-35 epitope presentation and that both proteins were found to be down-regulated in CTL-resistant UKRV-Mel-15a clones suggests that both or one of them may be causative for the observed immune evasion. Therefore, we next sought to determine whether the immune escape of the UKRV-Mel-15a clones was reversible by restoring p97/VCP or VIMP expression.
To this end, we performed MHC class I presentation rescue experiments by transfecting the UKRV-Mel-15a clone 18 with expression vectors encoding full-length p97/VCP or VIMP proteins. As shown in Fig. 4A,B, re-expression of p97/VCP restored stimulation of Melan-A/MART-1 26-35 -specific T cells to levels comparable to those detected upon exposure to the parental UKVR-Mel-15a cell line, demonstrating that the observed down-regulated expression of p97/VPC was indeed responsible for the observed immune escape of the UKRV-Mel-15a Mel-15 clones. Accordingly, RNAi-mediated down-regulation of p97/VCP in the parental UKVR-Mel-15a cell line was accompanied by severe impairment of Melan-A/MART-1 26-35 presentation (Fig. S3), further confirming that p97/VCP ensures the supply of Melan-A/MART-1 [26][27][28][29][30][31][32][33][34][35] antigenic peptides under normal conditions. This also suggests that the observed emergence of resistant clones following Melan-A/MART-1 26-35 CTL exposure is most likely driven by the spontaneous loss of p97/VCP expression in the heterogeneous UKRV-Mel-15a parental melanoma cell population or a selective enrichment of cells expressing low levels of p97/VCP. In contrast, over-expression of the p97/VCP recruiting membrane protein VIMP failed to rescue the Melan-A/MART-1 26-35 epitope presentation (Fig. 5A,B). This, however, is in agreement with a previous report showing that the over-expression of VIMP in COS cells alters ER morphology which may cause dysfunction of the ERAD pathway 17 .
Down-regulation of p97/VCP is also an immune escape mechanism in Ma-Mel-63a cells. The data obtained so far provide strong evidence that impaired expression of p97/VCP and subsequent dysfunction of the ERAD pathway was causative for the observed immune escape of CTL challenged UKVR-Mel-15a cells. Nevertheless, the possibility remained that this may represent an isolated case due to unknown specific features of the UKVR-Mel-15a cells. Therefore, we performed a second independent experimental analysis using the Melan-A/MART-1 + , HLA-A*0201 melanoma cell line Ma-Mel-63a. As shown in Fig. 6A, Ma-Mel-63a cells   expressed and efficiently presented the Melan-A/MART-1 [26][27][28][29][30][31][32][33][34][35] tumor epitope, as revealed by TNF-α -release assays. Importantly, overexpression of p97/VCP did not further improve the Melan-A/MART-1 26-35 epitope presentation, suggesting that the ERAD pathway was fully functional in these cells. This notion was also supported by immunoblotting experiments revealing normal expression levels of the ERAD associated components (data not shown). Following the same experimental conditions, Ma-Mel-63a cells were challenged twice in short-term co-culture with Melan-A/MART-1 26-35 -specific CTL. The Ma-Mel-63a clones resistant to CTL mediated lysis were further cultivated and assessed for their capacity of presenting the Melan-A/MART-1 26-35 epitope. Of the 33 Ma-Mel-63a clones, 3 clones (i.e. clones 9, 15 and 26) exhibited substantial impaired recognition by Melan-A/MART-1 [26][27][28][29][30][31][32][33][34][35] CTL when compared with the unchallenged parental Ma-Mel-63a cell line (Fig. 6B). Strikingly, all three clones also displayed reduced p97/VCP expression (Fig. 6C). Of these, clone 26 could not be further analyzed because it lost its endogenous Melan-A/MART-1 expression (data not shown). The Ma-Mel-63a clone 15 was finally selected for p97/VCP rescue experiments, because it revealed an almost unaltered expression level of the ERAD components VIMP, HERP or Derlin-1 as well as a normal expression of the IFN-inducible i-subunits and TAP1/2 (Figs 7A and S4). As shown in Fig. 7B,C, overexpression of p97/VCP almost completely restored Melan-A/MART-1 26-35 epitope presentation. These data show that impaired expression of the ERAD component p97/VCP upon immune challenge with Melan-A/MART-1 26-35 -specific T-cells is an universal event that allows melanoma cells to escape CTL-mediated cell death.

Discussion
There is mounting clinical evidence that T-lymphocytes play a central role in the regression of melanoma in vivo and that adoptive transfer of cultured T cells from tumor infiltrating lymphocytes may result in tumor regression 31 . However, loss of MHC class I expression has been reported to render tumors resistant to adoptive T-cell therapy 32 . In fact, analysis of tumor cells also revealed a down-regulation of the IFN-γ -inducible antigen processing machinery [33][34][35][36][37] . It is not clear how frequent immune escape of tumors upon T cell pressure results in the irreversible loss of IFN-γ sensitivity. Studies of tumor cell lines showed that, in most cases, the expression of down-regulated components involved in MHC class I antigen presentation was restored by IFN-γ treatment. Nevertheless, immune evasion of tumor cells upon T cell pressure still remains a problem and very little is known about escape mechanism connected with antigen processing. Our experiments show that co-cultivation of the  (Figs 1 and 6). However, immunoblot analyses failed to provide evidence that the expression of any of the known IFN-γ inducible components of the antigen processing and presentation machinery was significantly modified (Figs S1 and S4) and therefore did not explain the observed immune escape of the resistant UKRV-Mel-15a and Ma-Mel-63a clones.
The UPS with its more than 1000 protein components plays a central role in the generation of tumor epitopes presented by MHC class I molecules. Many of these may be directly or indirectly involved in the regulated proteasome dependent degradation of the Melan-A/MART-1 protein. Because of this complexity, we focused our analysis in particular on components of the ER associated degradation (ERAD) pathway known to be involved in the degradation of membrane-associated tumor antigens such as tyrosinase 38 . Indeed, our cDNA array analyses of melanoma cells resistant to cytolysis by Melan-A/MART-1 26-35 -specific CTL revealed several of the known ERAD components to be down regulated in these cells. Interestingly, the most reproducible down-regulation was observed for the AAA ATPase p97/VCP (Table 1). Importantly, the down-regulation of p97/VCP was observed in two independent melanoma cell lines, indicating that such effect was not cell line-specific but rather a more general phenomenon in response to CTL exposure (Figs 2A and 6C). The essential and critical role of ERAD in the ubiquitin-dependent degradation of membrane proteins and, as such, in the generation of antigenic peptide derived from membrane proteins is well documented 39,40 . Given the high number of membrane-bound proteins among melanoma-associated antigens, an active contribution of ERAD in tumor surveillance is very likely. For example, a functional ERAD pathway has been reported to be a prerequisite for the breakdown of the tyrosinase antigen 38,41 . This is also true for secretory proteins such as the pro-insulin auto-antigen 42 . Interestingly, the ERAD-mediated degradation of tyrosinase occurs in an EDEM1-dependent fashion, while that of pro-insulin mainly relies on Derlin-2, HRD1 and p97/VCP. The observation that ERAD substrates do not necessarily rely on the same components implies a strong heterogeneity and plasticity of the ERAD pathway. This point is of considerable importance, as it suggests that the loss of one particular ERAD-associated protein in tumors might not necessarily cause immune escape. An exception, however, is p97/VCP which is essentially required for the extraction of substrates from the ER to the cytoplasm, regardless of the composition of the ERAD complexes and/or pathways employed. Thus, any down-regulation of p97/VCP will be accompanied by a decreased MHC class I presentation of all ERAD-dependent antigens. Importantly, the involvement of ERAD in the supply of MHC class I-restricted peptides is not exclusively restricted to hydrophobic antigens. Hence, a major role for ERAD has also been described in the cross-presentation of soluble antigens by dendritic cells (DC) [43][44][45][46] , although the precise mechanisms by which internalized antigens use ERAD to gain access to the cytoplasm remain unclear. Interestingly, cross-presentation is not restricted to DC and may be also used by melanoma cells to (B) Melanoma cells were subjected to protein extraction and whole cell extracts were resolved on SDS-PAGE followed by Western blotting using antibodies specific for p97/VCP and β -actin (loading control), as indicated.
Scientific RepoRts | 6:25208 | DOI: 10.1038/srep25208 present antigens derived from secretory proteins such as the matrix metalloproteinase-2 (MMP2) 560-568 epitope 47 . As expected, a role for ERAD has been suggested in the processing of this antigenic peptide 48 , although the implication of p97/VCP remains to be formally addressed. Our experiments demonstrate that re-expression of p97/VCP in CTL-resistant UKRV-Mel-15a clone 18 and Ma-Mel-63a clone 15 to normal levels was able to almost completely recover Melan-A/MART-1 26-35 epitope presentation. The expression of p97/VCP is not induced by cytokines and it is not clear at the moment why its expression may be so influenced by exposure of melanoma cells to Melan-A/MART-1 [26][27][28][29][30][31][32][33][34][35] CTL. In any case, the down-regulation of ERAD is of general importance, as it will affect not only the presentation of the Melan-A/MART-1 [26][27][28][29][30][31][32][33][34][35] epitope but also the presentation of all other membrane protein-derived tumor epitopes generated by the ERAD pathway. Our data suggest that, in order to minimize or circumvent rapid immune evasion upon T-cell therapy, those tumor epitopes should be chosen and simultaneously targeted by T cells that are derived from separate antigens, which are degraded and processed by distinct degradation pathways. In this context, the option to target antigens whose degradation does not primarily rely on the UPS appears indeed particularly attractive. In view of the UPS, and a fortiori, p97/VCP as a major target whose deregulation might facilitate tumor escape, future research efforts may also address the functional relevance of proteasome-independent MHC class I tumor peptides in immunotherapy. Preparation of RNA, microarray hybridization, and data analysis. Microarrays hybridizations were conducted, as previously described 50 . Briefly, total RNA was extracted, amplified and labelled with biotin (Message AmpII-Biotin enhanced Kit, Ambion). The Human U133 2.0 Plus-Array (Affymetrix) was custom hybridized and evaluated by standard procedures (Signature Diagnostics). Raw data files were deposited on NCBIs Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE75929.