Immunomodulatory Properties of DNA Hypomethylating Agents: Selecting the Optimal Epigenetic Partner for Cancer Immunotherapy

DNA hypomethylating agents (DHAs) play a well-acknowledged role in potentiating the immunogenicity and the immune recognition of neoplastic cells. This immunomodulatory activity of DHAs is linked to their ability to induce or to up-regulate on neoplastic cells the expression of a variety of immune molecules that play a crucial role in host-tumor immune interactions. To further investigate the clinical potential of diverse epigenetic compounds when combined with immunotherapeutic strategies, we have now compared the tumor immunomodulatory properties of the first generation DHAs, azacytidine (AZA) and decitabine (DAC) and of the next generation DHA, guadecitabine. To this end, human melanoma and hematological cancer cells were treated in vitro with 1 μM guadecitabine, DAC or AZA and then studied by molecular and flow cytometry analyses for changes in their baseline expression of selected immune molecules involved in different mechanism(s) of immune recognition. Results demonstrated a stronger DNA hypomethylating activity of guadecitabine and DAC, compared to AZA that associated with stronger immunomodulatory activities. Indeed, the mRNA expression of cancer testis antigens, immune-checkpoint blocking molecules, immunostimulatory cytokines, involved in NK and T cell signaling and recruiting, and of genes involved in interferon pathway was higher after guadecitabine and DAC compared to AZA treatment. Moreover, a stronger up-regulation of the constitutive expression of HLA class I antigens and of Intercellular Adhesion Molecule-1 was observed with guadecitabine and DAC compared to AZA. Guadecitabine and DAC seem to represent the optimal combination partners to improve the therapeutic efficacy of immunotherapeutic agents in combination/sequencing clinical studies.


INTRODUCTION
Epigenetic events are emerging as a hallmark of cancer development and progression, impairing immunogenicity and immune recognition of cancer cells, possibly favoring their escape from the host's immune recognition (Sigalotti et al., 2005;Maio et al., 2015). One of the most widely studied epigenetic modifications in cancer is the aberrant methylation of DNA. It could occur through both global DNA hypomethylation, leading to genomic instability and possibly increasing the frequency of mutations and chromosomal abnormalities (Howard et al., 2008;Pogribny, 2010), and through the hypermethylation of specific genes leading to the impairment of the corresponding protein expression, mainly catalyzed by DNA methyltransferase (DNMT) enzymes (Sigalotti et al., 2014). The plasticity of epigenetic phenomena suggested the feasibility of their targeting by epigenetic drugs, such as DNA hypomethylating agents (DHAs), that can restore the physiologic epigenetic pattern by targeting DNMT enzymes (Maio et al., 2015). The most studied DHAs are nucleoside analogs of cytidine in which the cytosine ring has been modified to give them the DNMT inhibitory activity (Yoo and Jones, 2006). They include the first generation DHAs, azacytidine (AZA) and decitabine (DAC), FDA approved for the treatment of myelodysplastic syndromes (MDS) and acute myeloid leukemia (AML) (Saba, 2007;Cataldo et al., 2009), and the next generation DHA, guadecitabine. The latter is a dinucleotide of decitabine and deoxyguanosine designed to protect its active metabolite, DAC, from cytidine deaminase degradation resulting in a higher stability and a better tolerability of DAC in cancer patients (Yoo et al., 2007;Issa et al., 2015;Jueliger et al., 2016).
We have extensively demonstrated an epigenetic remodeling of cancer by DAC and guadecitabine as a result of the up-regulation and induction of different immune molecules and antigens involved in the immunogenicity and/or immune recognition of cancer cells of different histotype. Among them, HLA class I antigens, the co-stimulatory molecule ICAM-1, and tumor-associated antigens (TAA), such as the cancer testis antigens (CTAs) NY-ESO-1 and MAGE-A3 that are considered suitable therapeutic targets due to their high immunogenic potential (Coral et al., 1999(Coral et al., , 2013. The functional role of these phenotypic changes is demonstrated by the significant improvement of tumor cells recognition by CTA-specific cytotoxic T lymphocytes (CTL) (Sigalotti et al., 2004) and by the induction of anti-CTA humoral immune response in vivo (Coral et al., 2006).
Besides an activity on genes directly involved in the recognition of tumor cell by T lymphocytes, transcriptional changes induced in tumors by DHAs affected also several genes involved in the viral defense pathway, leading to an "indirect" activation of anti-tumor immune response through the modulation of interferon (IFN) signaling (Chiappinelli et al., 2017).
Innate immune cells play an important role in inhibiting cancer progression by complementing the effector activities of T cells; it has been demonstrated that these cells could exploit the action of epigenetic drugs by increasing tumor cell recognition and immune-mediated cell lysis. In this context, several studies reported that DAC-mediated hypomethylation could restore the NK group 2D ligands (NKG2DLs) [e.g., MHC class I-related chains (MIC) A and B] expression in tumors that represent an activating and a costimulatory signal for NK and T cells, respectively (Vasu et al., 2016;Zhang et al., 2016).
Although during the last years the pleiotropic immunomodulatory properties of different DHAs are consolidating, to the best of our knowledge no study investigated the differences among their activity. With the aim to optimize the therapeutic efficacy of DHAs in clinical setting and to identify the best epigenetic partner to be combined with cancer immunotherapy, we performed a comparative study of the immunomodulatory properties of the clinically approved DHAs (i.e., AZA and DAC) and of the next generation DHA guadecitabine, mainly focusing on the expression of different genes involved in different mechanism(s) of anti-tumor immunity.

Monoclonal Antibodies and Reagents
PE Mouse anti-human ICAM-1 clone 84H10 monoclonal antibody (mAb) was purchased from Beckman Coulter; alexafluor 488 mouse anti-human HLA class I clone W6/32 mAb was purchased from Biolegend; guadecitabine was kindly provided by Astex Pharmaceuticals, Inc. (Pleasanton, CA, United States); DAC was purchased from Abcam and AZA from Sigma Chemical Co.

In vitro Tumor Cells Treatment With DHAs
Human melanoma (1 × 10 6 ) and hematological cancer (1,2 × 10 6 ) cell lines were seeded in T75 tissue culture flasks and treated 24 h later with 1 µM dose of guadecitabine or DAC (Coral et al., 2013), compared to an equimolar dose of AZA every 12 h for 2 days (4 pulses). At the end of the treatment (day 6th), cell lines were collected and analyzed. Control cultures were treated under similar experimental conditions without drugs.

Quantitative Real-Time Methylation Specific PCR (qMSP) Analysis
Genomic DNA (500 ng) extracted from melanoma cell lines, using QIAmp DNA Blood mini Kit (Qiagen, Hilden, Germany), was subjected to modification with sodium bisulfite using the EZ DNA Methylation-Gold Kit (Zymo Research, CA, United States). Primers for the analysis of the methylation status of LINE-1 were designed using the free on-line software MethPrimer (Li and Dahiya, 2002), and are the follows: LINE-1 Unmethylated F: 5 -TGTGTGTGAGTTGAAGTAGGGT-3 , Unmethylated R: 5 -ACCCAATTTTCCAAATACAACCATCA-3 ; LINE-1 Methylated F: 5 -CGCGAGTCGAAGTAGGGC-3 , Methylated R: 5 -ACCCGATTTTCCAAATACGACCG-3 . SYBR green qMSP reactions were performed with methylated-or unmethylated-specific primer pairs on 2 µl of bisulfite-modified genomic DNA. The copy number of methylated or unmethylated sequences for LINE-1 gene was established by extrapolation from the standard curves. The percentage (%) of methylation was defined as ratio between methylated molecules and the sum of methylated and unmethylated molecules and data were reported as % of LINE-1 demethylation ± standard deviation (SD) of treated vs. untreated cells. CpG Methyltransferase (New England BioLabs, Ipswich, MA, United States) and RepliG mini Kit (Qiagen, Hilden, Germany) were used to obtained positive (CTRL +) and negative (CTRL −) methylation control, respectively.

Quantitative Real-Time RT-PCR Analysis
Total RNA was extracted by using Trizol reagent (Invitrogen, Milan, Italy) according to the manufacturer's instruction. RNA extracted was digested with RNAse-free DNAse (Roche Diagnostics, Milan, Italy). Synthesis of cDNA was performed on 2 µg of total RNA using M-MLV reverse transcriptase (Invitrogen, Milan, Italy) and random hexamer primers (Promega, Milan, Italy), according to the manufacturer's instructions. cDNA standards were obtained by RT-PCR amplification of the specific mRNAs and quantitated by NanoDrop2000 Spectrophotometer (Thermo Scientific, Massachusetts, United States). Quantitative real time RT-PCR were performed on 20 ng retrotranscribed total RNA in a final volume of 20 µl SYBR Green Master Mix (Applied Biosystems, Foster City, CA, United States) utilizing the 7500 Fast Real-Time PCR System (Applied Biosystems, Foster City, CA, United States) and software. The copy number of specific antigen and of the reference gene β-actin was established in each sample by extrapolation of the standard curve. The number of selected antigen cDNA molecules in each sample was then normalized to the number of cDNA molecules of β-actin. Gene expression was considered: (i) positive if numbers of gene/β-actin molecules were ≥ 1E-04; (ii) up-regulated if its positive expression was increased at least twice [Fold Change (FC) ≥2]. Data analyzed by multiparametric Dunn's test with p < 0.05 were considered statistically significant. The primers used for the quantitative real-time RT-PCR analyses are listed in Supplementary Table 1.

RT-PCR Analysis
RT-PCR reactions, using oligonucleotide primer sequences and PCR amplification programs specific for CTA family genes (i.e., MAGE-A2, -A4, -A10, GAGE1-2, SSX1-2, and SSX1-5), were performed as previously described (Sigalotti et al., 2004). The integrity of RNA and random primers-synthesized cDNA was confirmed by the amplification of all cDNA samples with β-actinspecific primers (Sigalotti et al., 2004). Five microliters of each RT-PCR sample were run on a 2% agarose gel, stained with green gel plus (Fisher Molecular Biology, Rome, Italy) and visualized by Gel doc XR (Bio-Rad Laboratories, Hercules, CA, United States). The primer sequences used for the quantitative RT-PCR are listed in Supplementary Table 2.

Comparative Analysis of the Demethylating Activity of Different DHAs in Human Cancer Cell Lines
To compare the demethylating activity of the different investigated DHAs in cancer cells, qMSP analysis was performed to measure the extent of LINE-1 methylation repetitive elements, chosen as a surrogate of the overall genomic DNA methylation, in 14 melanoma and in 10 hematological tumor cell lines treated with 1 µM guadecitabine, DAC or AZA.
Statistical analysis performed on data obtained from all (n = 14) investigated melanoma cells showed significant differences in levels of NY-ESO-1 (p < 0.0001) and MAGE-A1 (p < 0.05) expression detected after treatment with guadecitabine and DAC, but not with AZA, vs. untreated cells (Supplementary Tables 3, 5). No significant changes were observed for MAGE-A3 expression in all DHAs-treated melanoma cells (Supplementary Table 4). Data from all (n = 10) investigated hematological cancer cells showed significant (p < 0.05) differences in levels of NY-ESO-1 and MAGE-A1 expression after treatment with all DHAs, compared to untreated cells (Supplementary Tables 6, 8). Statistically significant (p < 0.005) differences in levels of MAGE-A3 expression were observed only in guadecitabine-and DAC-treated, compared to untreated hematological cancer cells (Supplementary Table 7).

Comparative Analysis of HLA Class I Antigens and Co-stimulatory Molecules Expression in Human Cancer Cell Lines Treated With Different DHAs
The immunomodulatory activity of different DHAs on the constitutive expression levels of HLA class I antigens and the costimulatory molecule, ICAM-1, was evaluated in 14 melanoma cell lines and 10 hematological cancer cell lines treated with 1 µM guadecitabine, DAC or AZA, by flow cytometry.
Statistical analysis performed on data obtained from all investigated 14 melanoma and 10 hematological cancer cells showed significant (p < 0.05) differences in MFI of HLA class I and ICAM-1 expression detected after treatment with guadecitabine and DAC, but not with AZA, vs. untreated cells (Tables 2, 3).
No differences in the % of HLA class I antigens and ICAM-1 positive cells were observed after DHAs treatment, in all investigated melanoma and hematological tumor cell lines (data not shown).
Statistical analysis performed on data obtained from all investigated melanoma and hematological cancer cells showed significant (p < 0.05) differences in levels of CTLA-4 and PD-1 expression detected after guadecitabine and DAC treatment, but not after AZA, compared to untreated cells ( Supplementary  Tables 9, 10, 12, 13). Moreover, levels of PD-L1 expression observed only after DAC treatment were significantly different (p < 0.05) in both melanoma and hematological cancer cells vs. untreated cells (Supplementary Tables 11, 14).

Comparative Analysis of the Activity of Different DHAs in TME Immunomodulation
The study of the immunomodulatory effects of different DHAs was expanded by qRT-PCR analysis of changes in the expression of selected immunostimulatory molecules (e.g., CXCL10 and CXCL9, MICA and MICB) in melanoma (n = 14) and hematological tumor (n = 10) cell lines treated with 1 µM guadecitabine, DAC or AZA.
No differences were observed in the number of positive hematological tumor cell lines in which CXCL10, CXCL9, MICA and MICB expression was up-regulated (FC ≥2) by all investigated DHAs treatments (Supplementary Tables 19-22).
Statistical analysis performed on data obtained from all investigated melanoma cells showed significant differences (p < 0.05) in levels of expression of all immunostimulatory molecules after guadecitabine and DAC treatment (Supplementary Tables 13-16), but not after AZA exposure, compared to untreated cells.
Conversely, data from all hematological cancer cells showed significant (p < 0.005) differences in levels of CXCL9 and MICA expression only after guadecitabine and DAC treatment vs. untreated cells; while significant (p < 0.05) differences in levels of CXCL10 and MICB expression were observed only after AZA treatment, compared to untreated cells ( Supplementary  Tables 19-22).

Comparative Analysis of Anti-viral Genes Expression in Melanoma Cell Lines Treated With Different DHAs
To compare the immunomodulatory activity of first and next generation DHAs, relative quantitative real-time RT-PCR analyses for HERVs expression were performed on 14 melanoma and 10 hematological cancer cell lines, respectively, and for ISGs expression on 14 melanoma cell lines, treated with 1 µM guadecitabine, DAC or AZA.

DISCUSSION
The demonstrated immunomodulatory activity of DHAs, which improves immunogenicity and immune recognition of cancer cells, results in priming and sensitizing the host immune response to immunotherapies. In light of these considerations, several clinical studies are investigating the combination of DHAs with different IC blocking mAbs in tumors of different histotype. To optimize the therapeutic efficacy of these new promising combination therapeutic strategies, we performed a comparative study of the immunomodulatory properties of selected clinically approved DHAs, AZA and DAC, and of the next generation DHA, guadecitabine, on human melanoma and hematological cancer cell lines, to identify the best epigenetic partner to be combined with immunotherapy.
The first evidence emerging from our results is a different hypomethylating effect of investigated DHAs on tumors of different histotypes. The highest LINE-1 global demethylation is achieved with guadecitabine in both melanoma and hematological cancer cell lines. A different hypomethylating effect between DHAs was already discussed in AML cells (Flotho et al., 2009;Hollenbach et al., 2010;Srivastava et al., 2014), suggesting the distinction of investigated DHAs as non-equivalent agents.
Different data support the role of epigenetic compounds in facilitating immunological targeting of cancer cells due to their ability to modulate different molecules and pathways involved in the interplay between tumor cells and the immune system (Sigalotti et al., 2014). Based on this evidence, we demonstrate the higher immunomodulatory activity of guadecitabine or DAC, compared to AZA, in reverting the CTA-negative phenotype without differences among histotypes analyzed. In particular, the higher levels of CTAs expression observed after guadecitabine or DAC treatment vs. AZA, represent an important benefit for immune recognition of cancer cells, as CTAs are able to induce both humoral and cell-mediated immune responses (Sigalotti et al., 2005), thus representing ideal targets for tumor immunotherapeutic approaches. This immunomodulatory property of guadecitabine or DAC could render tumor cells more susceptible to vaccination-stimulated CTA-specific immune responses, and more generally to CD8+ T cell specific recognition. A stronger immunomodulatory effect by guadecitabine or DAC vs. AZA treatment is observed also in the up-regulation of both HLA class I antigens, playing a central role in the presentation of TAA peptides to CTL, and of the co-stimulatory molecule, ICAM-1, allowing an increased recognition of cancer cells and promoting the activation of T cells.
Noteworthy, in addition to the above reported effects on adaptive immunity by DHAs, recent data indicated that epigenetic drugs may be exploited to allow the tumor cells eradication by innate immune system (Kima et al., 2014;Sigalotti et al., 2014). In this context, the expression of the NKG2DL MICB on melanoma cells induced only by guadecitabine and DAC treatment, could contribute to the immune recognition of transformed cells and accordingly, to their apoptosis.
Anti-tumor immunity within the TME can be supported by immune-stimulatory cytokines, such as CXCL9 and CXCL10, involved in the recruitment of immunological infiltrates at tumor site. A positive modulation of these pro-inflammatory Th1 cytokines by DHAs was previously described in ovarian cancer (Peng et al., 2015) and in epithelial cancer cell lines (Wolff et al., 2017;Lai et al., 2018); in this respect, our results underline the strongest effect of guadecitabine and DAC compared to AZA in the modulation of these cytokines, in both investigated tumor histotypes, suggesting their major contribute to the development of anti-tumor immune response.
Epigenetic activation of immune response has been recently demonstrated also through the IFN pathway signaling, upstream of antigen processing and presentation genes machinery (Chiappinelli et al., 2017). In detail, DAC and AZA primed ISGs expression in ovarian and colon cancer cells through the activation of double strand RNA derived from HERVs. Our study confirms these data in melanoma and hematological cancer cells, but also demonstrates that guadecitabine and DAC, compared to AZA, up-regulate a higher "viral mimicry" state that could eventually increase immune response. In addition, DNA demethylation offers the possibility to restore and/or to up-regulate the immunogenic potential of cancer cells, making them better targets for immunotherapeutic approaches. In this context, an important way in which DHAs may sensitize tumor cells to IC blocking therapy is through the up-regulation of immune tolerance ligands (Wrangle et al., 2013). Targeting of CTLA-4 or PD-1/PD-L1 molecules has profoundly improved the clinical management of advanced disease in a wide range of solid malignancies (Hu-Lieskovan and Ribas, 2017). In line with these evidence, we demonstrate that guadecitabine or DAC, compared to AZA, strongly upregulate the IC mRNA expression in all investigated tumor histotypes.
The translational relevance of the immunomodulatory activities of DHAs in cancer is sustained by the results from our previous studies, in a syngeneic mouse tumor model, demonstrating how guadecitabine or DAC were able to sensitize tumor cells to the anti-tumor activity of CTLA-4 blockade, inducing a significantly stronger tumor growth reduction compared to treatment with single agents (Covre et al., 2015a,b). The immunologic aspect of the anti-tumor effects induced by DHAs in combination with IC blocking therapy was demonstrated by the highest degree of CD3 infiltrating T cells, including both CD8+ and CD4+ T cells, detected in tumors from mice treated with the combination regimen (Covre et al., 2015b).
Comprehensively, this study shows that guadecitabine has similar immunomodulatory effects to DAC and both these compounds work better compared to AZA, identifying these two drugs as optimal partners to potentiate the anti-tumor activity of different immunotherapeutic approaches, not only in solid but also in hematological tumors. The higher resistance of guadecitabine to degradation by cytidine deaminase, supports its promising clinical activity and acceptable safety profile, by prolonging its in vivo exposure (Roboz et al., 2018). Along this line, the ongoing NIBIT-M4 clinical study, testing the immunologic and clinical efficacy of guadecitabine combined with the anti-CTLA-4 mAb, ipilimumab, in metastatic melanoma patients (Di Giacomo et al., 2018), will provide further support to the therapeutic potential of epigenetically based immunotherapy.