Epigenetic priming restores the HLA class-I antigen processing machinery expression in Merkel cell carcinoma

Merkel cell carcinoma (MCC) is a rare and aggressive, yet highly immunogenic skin cancer. The latter is due to its viral or UV-associated carcinogenesis. For tumor progression MCC has to escape the host’s immuno-surveillance, e.g. by loss of HLA class-I expression. Indeed, a reduced HLA class-I expression was observed in MCC tumor tissues and MCC cell lines. This reduced HLA class-I surface expression is caused by an impaired expression of key components of the antigen processing machinery (APM), including LMP2 and LMP7 as well as TAP1 and TAP2. Notably, experimental provisions of HLA class-I binding peptides restored HLA class-I surface expression on MCC cells. Silencing of the HLA class-I APM is due to histone deacetylation as inhibition of histone deacetylases (HDACs) not only induced acetylation of histones in the respective promoter regions but also re-expression of APM components. Thus, HDAC inhibition restored HLA class-I surface expression in vitro and in a mouse xenotransplantation model. In contrast to re-induction of HLA class-I by interferons, HDAC inhibitors did not interfere with the expression of immuno-dominant viral proteins. In summary, restoration of HLA class-I expression on MCC cells by epigenetic priming is an attractive approach to enhance therapies boosting adaptive immune responses.


MCC is characterized by a reduced HLA class-I expression in situ and in vitro.
. Next, we analyzed a series of eight MCC cell lines for their HLA class-I surface expression with an HLA-A, -B and -C detecting antibody (clone W6/32) by flow cytometry revealing differential expression levels of HLA class-I (supplementary Fig. S1). Four MCC cell lines representing almost no (MKL-1), low (BroLi), intermediate (WaGa) or high (MKL-2) HLA class-I surface expression were selected for further analysis (Fig. 1C). The subcellular localization of HLA-A was determined with an HLA-A detecting antibody (clone EP1395Y) by immunofluorescence (Fig. 1D), confirming the respective HLA class-I surface expression determined by flow cytometry for each cell line. Immunofluorescence also captured that despite reduced HLA class-I surface expression of WaGa, BroLi, and MKL-1, HLA-A is still present in their cytoplasm (Fig. 1D).

Reduced HLA class-I cell surface expression on MCC cells is associated with HLA class-I APM component down-regulation.
To elucidate the mechanism(s) leading to the reduced HLA class-I expression in MCC, we analyzed gene expression data from 35 cryopreserved MCC tumors, accessible at the gene expression omnibus (GEO) database (GSE22396), for HLA-A and B2M. Surprisingly, only ~25% of tumors (n = 8) expressed low or intermediate levels of HLA-A mRNA, whereas the majority (~75%, n = 27) contained high amounts of HLA-A specific mRNA ( Fig. 2A). B2M mRNA was also expressed at intermediate to high levels in the majorities of tumors. This discrepancy between HLA-A heavy chain mRNA and HLA-A membrane expression could be due to a lack of MHC complex stabilization by bound peptides. Thus, we next analyzed the mRNA expression of the HLA class-I APM components, TAP1, TAP2, LMP2 and LMP7 in the same data set (GSE22396). To this end, TAP1 and TAP2 mRNAs were expressed at very low levels in all analyzed tumors, and LMP2 and LMP7 mRNAs at low to intermediate levels in ~75% of tumors (n = 27) ( Fig. 2A). To determine whether this expression indeed reflect transcript levels in MCC cells and not cells of the tumor microenvironment, we analyzed the four MCC cell lines with varying degrees of HLA-A surface expression for gene expression of HLA-A, B2M, and the APM components. In line with the results generated from the analysis of their expression in tumor tissue, HLA-A and B2M mRNAs were present at high levels in all MCC cell lines ( Fig. 2B; supplementary Fig. S2), irrespective of the MHC class-I membrane expression ( Fig. 1C and D). However, the three MCC cell lines with reduced MHC class-I membrane expression (BroLi, MKL-1 and WaGa) were characterized by lowered LMP2, LMP7, TAP1 and TAP2, mRNA levels (Fig. 2B). Only MKL-2, a cell line showing high MHC class-I expression, contained high levels of most of the APM-component specific mRNAs (Fig. 2B, supplementary Fig. S2). It should be noted, that TAP2 mRNA expression was also low in MKL-2 cells. To confirm this observation at the protein level, we performed immunoblots of total cell lysates with HLA-A-, β 2 m-, TAP1-, TAP2-, LMP2-and LMP7-specific mAbs (Fig. 2C), revealing that HLA-A and β 2 m were expressed in all analyzed MCC cell lines, while TAP1 and LMP2, LMP7 expression was largely restricted to the MKL-2 cell line (Fig. 2C). In line with the mRNA expression, TAP2 protein was only sparsely expressed in all the analyzed cell lines (Fig. 2C).
External supply of MCPyV-derived peptide epitopes stabilizes HLA class-I surface expression by MCC cells. Most MCCs analyzed by us expressed high levels of both HLA heavy and B2M light chain mRNA whereas the respective expression of APM components was low. Since empty β 2 m-HLA class-I heavy chain complexes are rapidly removed from the cell surface and subsequently degraded 20,21 , reduced HLA class-I surface expression by MCC cells is likely to be due to deficient antigen processing. Indeed, addition of either a pool of HLA class-I-restricted T-cell epitopes derived from influenza virus or MCPyV increased the HLA class-I surface expression of MKL-1, BroLi and WaGa cell lines (Fig. 2D,E). The used MCPyV peptides have been described before 11 and were selected according to the HLA-A type of the tested MCC cell line (Supplementary Table S3).
In cells with deficient antigen processing such as the TAP-deficient T2 cell line, an excess of externally provided peptides binding to the respective HLA molecules stabilizes HLA class-I complexes in an APM-independent manner 22 . Thus, reduced HLA class-I cell surface expression in MCC is at least in part due to a diminished availability of peptides for HLA class-I stabilization.
Transcription of APM genes in MCC cell lines is epigenetically silenced by histone hypoacetylation. Expression of APM genes can be silenced by DNA methylation and/or histone hypoacetylation at the MHC gene cluster 19 . Analyses of GSE22396 revealed that a high abundance of histone deacetylase (HDAC) mRNAs (sum of HDAC 1-10) was negatively correlated with the mRNAs encoding LMP2, LMP7, TAP1 and TAP2, (R squared = 0.32, p = 0.0006; supplementary Fig. S3A). Since no such correlation was observed for methyltransferases (supplementary Fig. S3B), and treatment of MCC cell lines even with high doses (5 µM) of two distinct methyltransferase inhibitors (5-azacytidine and RG108) had no effect on HLA class-I surface expression (supplementary Fig. S3C), we focused on the relevance of histone hypoacetylation.
Global histone acetylation of MKL-2, WaGa, MKL-1 and BroLi was determined before and after HDAC inhibition by immunoblot (Fig. 3A). As the culture of these MCC cell lines for 24 hours in clinically relevant concentrations of vorinostat (V), i.e. the plasma level of 1.25 µM observed under current treatment regimens 23 , resulted only in a modest induction of global histone acetylation, we combined V with mithramycin A (MA). MA is a clinically approved DNA-binding cytotoxic antibiotic synergizing with HDAC inhibitors by (i) transcriptionally inhibiting the compensatory induction of certain HDACs 24 and (ii) by preventing the formation of SP1/HDAC inhibitory complexes at the promoters' GC box 25 . Notably, the combined treatment of MCC cell lines with V and MA resulted in a marked induction of global histone acetylation. Next, histone H3K9 promoter acetylation of LMP2, LMP7, TAP1 and TAP2 was analyzed in untreated or V plus MA treated WaGa cells by chromatin immune precipitation (ChIP). Treatment with V plus MA increased histone H3K9 acetylation in all analyzed promoter regions (Fig. 3B). mRNA expression of APM components LMP2, LMP7, TAP1 and TAP2 was measured by RT-qPCR before and after HDAC inhibition in MKL-2, WaGa, MKL-1 and BroLi cells (Fig. 3C). The combined treatment with V and MA strongly induced mRNA expression of the analyzed APM component encoding genes in all the tested MCC cell lines. As suggested by the increased HLA class-I membrane expression in response to the external supply of HLA-binding peptides, HDAC inhibition-induced expression of APM components resulted in an increased HLA class-I surface expression as demonstrated by immunofluorescence (Fig. 3D) and flow cytometry ( Fig. 3E and F). Confocal microscopy also visualized, that HDAC inhibition resulted in a cytoplasmic conglomeration of HLA-A protein, most likely in cellular compartments where MHC-class-I complexes are assembled, loaded and transported to the cell surface.
Surprisingly, MA treatment alone had a stronger effect on APM gene expression than V alone. However, only HDAC inhibition by V plus MA had a marked synergistic effect on HLA class-I surface expression ( Fig. 3E and F).  It is important to note, that the combined treatment did not interfere with the expression of MCPyV-encoded proteins, which represent the source of immunogenic epitopes in MCPyV + MCC (Fig. 3G).
HDAC inhibition induces histone acetylation and HLA class-I stabilization in vivo. Next, we translated these observations into a MCC xenotransplantation model based on WaGa cells and NOD/SCID mice 26 . Once the induced tumors reached a volume of approximately 100 mm 3 , treatment with i.p. injections of either V plus MA or the respective drug carriers was started. Mice were treated for two weeks at concentrations equivalent to those applied in humans. Forty-eight hours after the last dosage, animals were sacrificed and tumors harvested. In analogy to the in vitro results, treatment of mice with V plus MA significantly increased the transcription of TAP2, LMP2 and LMP7 in xenotransplanted MCC tumors (p < 0.05; Fig. 4A). However, the expression of TAP1 in vivo was not increased by this treatment; this finding may reflect the strong TAP1 induction by xenotransplantation itself. This induction may be due to the fact that NOD/SCID mice are comprised of functional B and T cells, but still harbor functional dendritic cells and NK cells expressing interferon gamma 27 . Immunohistochemistry of the xenotransplants confirmed that these effects were indeed due to HDAC inhibition, as histone H3K9 acetylation was strongly induced by treatment of mice with V plus MA (Fig. 4B). Consequently, the combined treatment enhanced the HLA class-I surface expression on WaGa xenotransplants (Fig. 4B). In accordance to the in vitro results RT-qPCR did not demonstrate any increase in HLA-A and β 2 m mRNA expression in response to treatment with V plus MA, supporting the assumption that HDAC inhibitor mediated HLA class-I induction on MCC cells is largely mediated by an enhanced antigen processing (Fig. 4A). Furthermore, it is important to note that MCPyV large T antigen (LT) protein expression was not altered in MCC xenotransplants by V plus MA treatment (Fig. 4B).

Discussion
Virus-associated malignancies like MCC are highly immunogenic and are consequently subjected to a strong immuno-selective pressure. Thus, these cancers can only progress if efficient immune escape strategies are operative. Indeed, although most patients harbor MCPyV-derived epitope-specific CD8 + T cells circulating in the blood 11,28 , and that in most MCC cases the oncogenic phenotype critically depends on continuous expression of the MCPyV early genes 9, 10 , brisk T-cell infiltrates in MCC tumors are rare 13,29 .
Consequently, immune modulation is a promising therapeutic option for MCC patients, which is currently tested in several clinical trials including: autologous MCPyV-specific CD8 + T-cell transfer, checkpoint blocking antibodies, or cytokine-based therapies using tumor stroma-targeting antibody-IL2 fusion proteins and in vivo electroporation of IL-2-coding plasmids (reviewed in ref. 30). Notably, very promising results on the efficacy of immune checkpoint blocking antibodies have already been reported 6,7 . These therapeutic approaches depend on MCC-specific adaptive immune responses. For cytotoxic T cells a cognate recognition, i.e. interaction of the T-cell receptor with suiTab. peptide/HLA class-I complexes presented on the surface of the target cells, is essential. Thus, it is not surprising that diminished HLA class-I expression on cancer cells is one of the major immune escape mechanisms both spontaneously and in response to immunotherapy [31][32][33][34] . The molecular mechanisms to achieve such an HLA class-I loss are multiple and vary depending both on the type of tumor as well as on the individual patient 35 .
In line with recent reports, we demonstrate here that the impaired HLA class-I expression is a frequent immune escape mechanism of MCC, which however, appears not to be genetically fixed as it can be therapeutically overcome 16 . Actually, we demonstrate here that the common underlying molecular mechanism of HLA class-I loss observed in the MCC cell lines investigated by us is epigenetic silencing of the antigen presentation machinery (APM): The transporter subunits TAP1 and TAP2 as well as the immunoproteasome subunits LMP2 and LMP7 are simultaneously down-regulated by histone hypoacetylation. While genetic aberrations in single components of the APM have been reported 18 , the synchronized down-regulation of several APM components appears to be an even more efficient immune escape mechanism 36 . Viral infection of cells is frequently associated with an inflammatory response. Thus, infected cells are exposed to cytokines such as interferon gamma, which results in the replacement of certain catalytic subunits of the proteasome by LMP2 and LMP7 to allow the formation of the immunoproteasome 37 . The immunoproteasome not only generates a broader spectrum of peptides for presentation by HLA class-I complexes, but is also much more efficient in processing viral proteins, which are poorly processed by the regular proteasome 38,39 . Thus, epigenetic silencing of LMP2 and LMP7 is a particularly suiTab. means of immune evasion for cancers with a viral carcinogenesis such as MCC. Indeed, viruses can change the host epigenome in order to escape immune responses (reviewed in ref. 40). In this regard, viral infection can alter the host epigenome directly via the interaction of viral components with the epigenetic machinery of the infected cell 41,42 , or indirectly due to the increased immune selective pressure caused by the expression of viral antigens. It has to be elucidated which scenario is operative in MCC. served as nuclear stain. Single cell images in the upper right corner of the overview image were acquired using confocal microscopy. (E,F) HLA class-I cell surface expression was determined by flow cytometry using a HLA-ABC specific antibody as exemplified for WaGa (E); the results for all cell lines analyzed are depicted as the geometric mean fluorescence intensity (gMFI) of HLA class-I (HLA-ABC) staining, + SEM in three independent experiments (F). Statistical analysis was performed using the Friedman test as indicated. (G) Viral protein expression was determined by immunoblot of whole cell lysates using an antibody specific for LT (clone CM2B4) depicting the different truncated LTs characteristically for MCC cells. β-tubulin served as loading control. In any case, this highly effective immune escape mechanism may be reversed pharmacologically: Epigenetically silenced genes of the HLA class-I locus are effectively re-induced by HDAC inhibition [43][44][45] . To this end, treatment of MCC cells with the HDAC inhibitor vorinostat in combination with the Sp1 inhibitor mithramycin A resulted in the transcription of the APM components followed by HLA class-I surface expression. Notably, we observed this effect at clinically relevant concentrations and could translate these findings into a pre-clinical in vivo model. Consequently, epigenetic immunological priming could be readily used to overcome primary or secondary resistance to PD-1/PD-L1 blocking antibodies 6,7 . Indeed, the combination of different HDAC inhibitors with immune checkpoint blocking antibodies is currently tested in solid tumors, including melanoma and non-small cell lung cancer (Supplementary Table S4). This notion is particularly attractive for MCC, since HDAC inhibition also induces the expression of activating and co-activating molecules such as NKG2D ligands 46 . An alternative means to up-regulate HLA class-I surface expression are interferons. We have previously reported that both type I and II interferons are potent inducers of HLA class-I surface expression in MCC 11,16 . However, interferons reduce the expression of the MCPyV encoded sT and LT transforming early genes 26 , i.e. the source which most of the immuno-dominant T cell epitopes are derived from 11 . In line with this, treatment of two MCC patients with MCPyV positive primary tumors and metastasis with interferon alpha had no therapeutic effect 47 .
We here demonstrate that epigenetic priming by HDAC inhibition appears to be an effective means to improve viral antigen processing and surface presentation in the majority of MCCs. MCC patients characterized by primary or secondary resistance to immunotherapy are likely to be rendered responsive by HDAC inhibition. As demonstrated here, testing for HLA class-I expression together with the H3K9 acetylation status of tumor lesions by immunohistochemistry, which can be readily performed in FFPE samples, may prove to be a valuable predictive biomarker of response to immunotherapy [48][49][50] . As mentioned above, there are currently several clinical trials ongoing in which the combination of HDAC inhibition and immunotherapy is tested as treatment of solid tumors (Supplementary Table S4); based on the presented data, MCC should be one of these cancers.

MCC tumor samples. A total of 56 formalin-fixed and paraffin-embedded (FFPE) tumor samples from 40
unselected MCC patients presenting at the Department of Dermatology, Medical University of Graz, were used for analysis. All tumor samples were excised for diagnostic or therapeutic reasons between 1992 and 2013, and were confirmed to be Merkel cell carcinoma according to established histological and immunohistochemical diagnostic guidelines. FFPE specimens included primary lesions, recurrent tumors, as well as metastases to the skin or lymph nodes. All studies on human material were approved by the institutional review board of the Medical University of Graz (24-295 ex 11/12). Informed consent has been obtained and the methods were carried out in accordance with the approved guidelines.
Immunofluorescence. Before immunofluorescence staining, MCC cells were attached to coverslips by poly-L-lysine coating. In short, coverslips were coated with 0.01% of poly-L-lysin (Sigma) for 10 minutes, and placed after one wash with PBS in a 12-well plate. 1 ml of cell suspension was added per well and incubated for 30 minutes at 37 °C in 5% CO 2 . After excess media was removed, the cells were washed with PBS once and incubated with 5 µg/ml wheat germ agglutinin (WGA, conjugated with Alexa Fluor 647) for 30 minutes at room temperature to stain cellular membranes. After three 5 minute washing steps with PBS, cells were fixed with ice cold acetone for 10 minutes, followed by three more washes with PBS and blocking with 1% bovine serum albumin (BSA) in PBS for one hour. Samples then were incubated with an antibody specific for HLA-A (Supplementary  Table S1, clone EP1395Y, Abcam) diluted 1:250 in blocking solution over night at 4 °C. After three washes with PBS a dylight 488 conjugated goat-anti-rabbit IgG (Thermo Scientific, Waltham, MA, USA) was added at a 1:200 dilution in blocking reagent for 1 hour at room temperature. Cells were washed three more times and nuclei were stained with DAPI. Cells were washed with PBS three more times and embedded with ProLong Dimond Antifade Mountant (Invitrogen, Carlsbad, CA, USA) and incubated at room temperature in the dark for 24 h before imaging. For spheroid imaging a Zeiss Axio Observer.Z1 with Apotome.2 system was used. Subcellular imaging of individual cells was performed with a Leica SP5 inverse confocal laser scanning microscope.
Quantitative real time polymerase chain reactions. mRNA of in vitro propagated cells or cryopreserved xenografts was isolated, transcribed into cDNA and RT-qPCR was performed as described before 46 . RPLP0 served as endogenous control using specific primers and Taqman probe, listed in Supplementary Table S2. The mRNA expression of HLA-A, B2M, TAP1, TAP2, LMP2, LMP7 was detected using SYBR green assays with specific primers listed in Supplementary Table S2. Relative quantification was calculated by the ΔΔCt method.
HLA class-I stabilization assay. MCC cell lines were incubated with 10 µM of a flu-peptide-pool (PANATecs, Heilbronn, Germany) or a peptide mixture of MCPyV LT-, sT-and VP1-derived epitopes known to match the respective HLA types of the used cell lines (Supplementary Table S3) 11 . The flu-peptide-pool served as a well established positive control for HLA class-I stabilization 54 . As negative control, an irrelevant peptide with no HLA binding motif was used. Incubation was performed until optimal HLA class-I stabilization was observed, i.e. 24 hours for WaGa and MKL-1 and 48 hours for BroLi. Quantification of HLA class-I expression by flow cytometry was performed as described above.
Chromatin immune precipitation (ChIP). ChIP assays were performed using the SimpleChiP ® Enzymatic Chromatin IP Kit with agarose beads (Cell Signaling). In brief, proteins were cross-linked to DNA with 1.5% formaldehyde for 10 minutes. Nuclear membranes were broken up using the Bioruptor sonicator (Diagenode, Seraing, Belgium) set to high, for 2 cycles of 10 minutes (30 seconds on/30 seconds off). Afterwards, antibodies against histone H3 (clone D2B12, Cell Signaling), acetyl-histone H3 (Lys9) (AcH3K9) (clone C5B11, Cell Signaling), or normal rabbit IgG (cat. no. 2729, Cell Signaling) were used for immune precipitation. The precipitated DNA was analyzed by qRT-PCR using primers specific for the TAP1, TAP2, LMP2 (PSMB9) or LMP7 (PSMB8) promoter region; primer sequences are given in Supplementary Table S2. The percentage of acetylated histones H3 (AcH3K9) was normalized to total histones H3 and calculated using following equation: = * − . %AcH3 over total H3 100 2(C H3 C AcH3K9) T T Xenotransplantation experiments. Six-week-old female NOD.CB17/Prkdc scid mice were obtained from Charles River Laboratories (Erkrath, Germany) and housed under specific pathogen-free conditions. Tumors were induced by s.c. injection as described before 26 . Twenty-four days after tumor cell inoculation at a volume of approximately 100 mm 3 treatment was started. Mice were divided into four groups of five ensuring an equal overall tumor burden. Mice were treated with 60 mg/kg vorinostat in combination with 0.2 mg/kg bodyweight or the drug carrier 45% Polyethylene glycol (PEG-400, Sigma) intra-peritoneal injection as describe befor 46 . Mice were treated five consecutive days per week for two weeks. IHC was performed as described in the Immunohistochemistry section. Anti-acetyl-histone H3 (Lys9) (AcH3K9) (cell signaling, cat. no. 9671) was diluted 1:50 and the LT specific antibody clone CM2B4 was diluted 1:200 (Supplementary Table S1). All animal studies were approved by the Austrian ministry of education and science according to the regulations for animal experimentation (BMWF-66.010/0151-II/3b/2012).