Expression of Cancer testis Antigens in Tumor-adjacent Normal Liver Predicts Post-resection Recurrence of Hepatocellular Carcinoma

Background: High recurrence rates after resection of hepatocellular cancer (HCC) with curative intent and lack of effective therapy for advanced disease impair clinical outcomes of HCC. Cancer/testis antigens (CTAs) are suitable targets for cancer immunotherapy if selectively expressed in tumor cells. The aims of this study were to establish a panel of CTAs that are frequently and selectively expressed in tumors of HCC-patients, and to investigate whether CTAs might be expressed in tumor-free liver tissues of HCC-patients. Methods: Surgically-resected tumor and paired tumor-free (TFL) tissues of HCC patients (n=100), healthy livers (n=21), and other healthy tissues (n=22 different tissues) were assessed for mRNA expression of 49 carefully selected CTAs by RT-qPCR. Protein expression of 5 CTAs was determined by immunohistochemistry (n=78). Results: Twelve CTAs were expressed at mRNA level in ≥ 10% of HCC-tumor tissues and not in healthy tissues except testis. In tumors, mRNA and protein of ≥ 1 CTA was expressed in 78% and 71% of HCC-patients, respectively. In TFL, CTA mRNA and protein expression was found in 45% and 30% of HCC-patients, respectively. Interestingly, CTA expression in TFL was an independent negative prognostic factor for HCC-recurrence and survival after tumor resection. Conclusions: We established a novel panel of 12 testis-restricted CTAs expressed in tumors of most HCC-patients, that can be safely used for immunotherapeutic targeting of HCC. The increased risk of HCC recurrence in patients with CTA expression in TFL suggests that CTA-expressing (pre-)malignant cells may be a source of HCC recurrence. Therefore, immunotherapeutic targeting of these antigens should be considered as adjuvant therapy to prevent HCC-recurrence after tumor resection. Y-encoded protein 1.

CPI therapy generally shows higher e cacy in patients with high tumor mutational load, extensive tumor CD8 T-cell in ltration, and pre-existing systemic anti-tumor T-cell immunity. (7,8) Immunomodulatory forms of induction chemotherapy or radiotherapy to transform immunologically "cold" tumors, that lack T cell in ltration, into "hot" tumors are being studied to sensitize patients to CPI therapy. (9,10) Also combinations of CPIs with therapeutic vaccination to enhance systemic anti-tumor T-cell immunity are being studied. (11)(12)(13) Therapeutic vaccination has shown potential to elicit systemic anti-tumor T cell responses in advanced HCC, (14,15) and may therefore be an effective method to sensitize HCC patients for CPI therapy. However, such trials have not been conducted in HCC so far.
A prerequisite of a safe and effective vaccine is that the targets are immunogenic and exclusively expressed in tumor cells, to prevent auto-immune side effects. Cancer testis antigens (CTA) are a family of proteins that are highly expressed in immune-privileged germ cells, whereas their expression in other healthy tissues is partially or completely silenced. In addition, CTAs can become aberrantly expressed in cancer cells of various histological subtypes. (16) As CTAs have also shown to be immunogenic, these antigens are considered as suitable shared tumor antigens for cancer immunotherapy, including therapeutic vaccination. (16,17) Based on their expression pro le in adult healthy tissues, they are classi ed into testis-restricted, testis/brain-restricted and testis-selective CTAs, the latter category showing expression in somatic tissues, although often at lower levels. (18) As testis-restricted CTAs are solely expressed in immune-privileged germ cells, they can therefore be safely applied for cancer immunotherapy. (18,19) Although early stage HCC-patients are treated by surgical resection or radiofrequency ablation with curative intent, recurrence rates are high, and currently no therapies to prevent recurrence are available.
Early recurrence likely originates from occult metastases in non-cancerous liver tissue at the time of resection, whereas late recurrences are more likely to represent de novo tumors. (20)(21)(22) To identify patients at risk of HCC recurrence, it remains of great importance to identify such occult (pre-)malignant lesions or cells in the non-cancerous liver tissue at the time of resection. We hypothesized that CTAs might not only be expressed in tumors, but also in occult (pre-)malignant lesions or cells present in noncancerous liver tissues of HCC patients, and that these may be (at least partially) responsible for HCC recurrence after tumor resection. We further reasoned that if this hypothesis is correct, adjuvant therapies to eradicate CTA-expressing (pre-)malignant cells from the remaining liver tissue may be a valuable tool in prevention of HCC recurrence after tumor resection.
The aims of this study were: 1) To establish a panel of CTAs that are frequently expressed in tumors of HCC patients but not in any healthy tissue except testis, for immunotherapeutic purposes such as therapeutic vaccination; 2) To determine whether CTAs are also expressed in non-cancerous liver tissues of HCC-patients, and whether such expression is associated with HCC recurrence after tumor resection.

HCC patients and tissues
A total of 100 archived surgically-resected fresh frozen tumor tissue samples and paired tumor-free liver (TFL) tissue samples obtained at a distance of > 2 cm from the tumors, as well as 76 formalin-xed para n-embedded (FFPE) paired tumor and TFL tissues, of HCC patients were collected after surgery or retrieved from the archives of the Department of Pathology, Erasmus Medical Center Rotterdam and the Dutch nationwide pathology archives (PALGA), respectively. The included HCC patients underwent hepatic resection (n = 97 and n = 73 for fresh frozen and FFPE samples, respectively) or liver transplantation (n = 3 for both fresh frozen and FFPE samples) for HCC in our center between February 1995 and September 2017, and diagnosis of HCC was con rmed by pathological examination. Medical records were reviewed for clinicopathological variables and date of rst recurrence, HCC-speci c death and last follow-up. This study was approved by the local ethics committees and adhered to the 1975 Declaration of Helsinki.
Further details of these and other included tissues can be found in the supplementary materials and methods.

Selection of CTAs
A literature search to identify CTAs reported to be expressed in HCC was conducted in PubMed on October 4th, 2018. A summary of this search is provided in Fig. 1A and the query in the Supplementary data. Papers written in English that described CTA expression in HCC patients and/or HCC cell lines were included. In addition, the CTA database (http://www.cta.lncc.br/) was consulted to nd additional CTAs expressed in HCC and one relevant paper was added. (23) Quantitative real-time PCR RNA was isolated from the frozen tissues and RT-qPCR was performed. The sequences, Tm-values and product lengths of the used primers are provided in Supplementary Table S1, and detailed methods can be found in the supplementary data le.

Immunohistochemistry
Protein expression was determined by immunohistochemistry (IHC) on tissue microarrays (TMA), that contained three 1 mm cores of each tumor and TFL tissue, as described in the supplementary data le as is the immunohistochemical staining method. The stained TMAs were scored blindly by two researchers, based on intensity of the staining (0 ). If less than 5 positive cells per core were observed, the core was scored as 0, and cores smaller than 50% of the original surface were excluded. The score per core was the product of the intensity and the percentage of positive cells (A = 0.1, B = 0.3, C = 0.7 and D = 1). The nal score was the average score of the three cores.

Statistical analysis
All statistical analyses were performed using Graphpad (Version 8.2.1 for Windows, San Diego, CA) and R Statistical software (Version 3.6.1 for Windows, Foundation for Statistical Computing, Vienna, Austria).
The correlation analysis was performed in RStudio with the 'corplot' package, using Pearson's correlation coe cient. For creating heatmaps, RStudio was used with the 'gplots' and 'pheatmap' packages. Survival analysis was performed by the Kaplan-Meier method and the Cox proportional hazards model. Time to event was calculated from the day of surgery. Used statistical tests are indicated in the gures. P-values < 0.05 were considered signi cant.

Patient characteristics
The clinicopathological characteristics of the 100 HCC patients, 35 cirrhotic patients and 21 healthy controls analyzed for mRNA expression are listed in Table 1. The majority of HCC-patients are Caucasian (82%), 34% of patients had cirrhosis, and 33% of patients had no underlying liver disease. Using a query to identify publications on CTAs expressed in HCC tissue, 281 publication records were obtained through the PubMed search and one relevant paper(23) was added. After removal of non-English publications, 270 publications were screened on title and abstract, of which 231 papers were excluded. Full texts were screened of the remaining 39 studies, which all met the inclusion criteria (Fig. 1A). In these 39 studies, expression of 73 different CTAs in HCC was reported; mRNA expression of 51, protein expression of 1 and both mRNA and protein expression of 21 CTAs (Supplementary Table S3). In addition, the CTA database (http://www.cta.lncc.br/) was consulted, which resulted in identi cation of 34 other CTAs expressed in HCC; 27 by mRNA, 4 by protein and 3 by protein and mRNA expression. Furthermore, 38 CTAs identi ed by the CTA database had already been identi ed in the literature search (Fig. 1B). Consecutively, to exclude expression of these 107 CTAs in healthy tissues, studies using nextgeneration sequencing to quantify mRNA expression levels in samples obtained from a large array of healthy tissues and organs, provided by the FANTOM consortium,(24) Human Protein Atlas (HPA) consortium (25) and genome-based tissue expression (GTEx) consortium, summarized on www.proteinatlas.org, and the genome-wide analysis of CTA mRNA expression by Hofmann, et al. (18) were consulted, which led to the exclusion of 47 CTAs due to expression in non-germline tissues (Fig. 1B).
To verify the absence of expression in healthy adult non-germline tissues, the expression of the remaining 60 CTAs was rst determined in 21 healthy liver tissues by RT-qPCR. For 11 CTAs it was not feasible to design speci c primers, due to high sequence homology with other genes. Of the 49 CTAs tested, 23 were expressed in healthy livers, with prevalence rates varying from 14-100%, and therefore excluded from further analysis. Twenty-four CTAs showed undetectable mRNA expression levels in healthy livers. Two CTAs (MAGEC1 and RING nger protein 17 [RNF17]) were each found to be expressed in 1 out of 21 tested healthy livers (with very low relative expression levels of 0.005 and 0.002 respectively), and therefore not excluded ( Fig. 1C and Supplementary Table S4). These 26 CTAs were selected for further study.
A panel of 12 CTAs is expressed in more than 10% of HCC tumors and not in healthy tissues The mRNA expression of these 26 CTAs was determined in 100 paired HCC tumors and TFL and in 35 non-malignant cirrhotic liver tissues. Thirteen CTAs were expressed in tumors of > 10% of HCC patients at variable expression levels ( Table 2, Fig. 2A and Supplementary Table S5) and selected for further study.
To verify the absence of these 13 CTAs in healthy adult non-germline tissues, mRNA expression was determined in 23 types of healthy adult tissues other than liver (Fig. 2B). Most tissues did not express any CTA, except for ovary which expressed ve CTAs. Four CTAs were expressed at very low relative expression levels in ovary (MAGEB2 0.002, cancer/testis antigen family 47 member A1 [CT47A1] 0.002, MAGEC1 0.003 and MAGEC2 0.002). However, RNF17 had a higher relative expression level (0.097) and was also expressed in other tissues (thyroid, adrenal gland, bladder, brain, throat, trachea, ovary and thymus), and was therefore excluded from further analysis. Table 2 mRNA expression of CTAs in HCC-patients. 1 Percentage of hepatocellular carcinomas (HCC) expressing mRNA of the CTA -meaning a Ct-value < 35 and relative expression > 0.001 (n = 100); 2 Mean relative expression (relative to the geometric mean of the 3 household genes-GUSB, HPRT1, PMM1) level in HCCs expressing the CTA and range; 3 Mean relative expression of the CTA in HCC expressing the CTA, relative to the relative mean expression in 3 testis tissues; 4 Percentage of paired tumor-free liver (TFL) tissues expressing mRNA of the CTA (n = 100); 5 Mean relative expression level in TFLs expressing the CTA and range; 6 Mean relative expression of the CTA in TFL expressing the CTA, relative to the relative mean expression in 3 testis tissues; 7 Percentage of non-cancerous/non-dysplastic cirrhotic liver tissues expressing the CTA (n = 35); *% in male  Thus, based on mRNA expression data, we identi ed a panel of 12 CTAs prevalently expressed in tumors of HCC-patients, but not in healthy adult tissues except testis. Seventy-eight percent of tumors expressed at least one of these 12 CTAs, 59% expressed at least 2 CTAs, 50% expressed at least 3 CTAs, and 40% 4 or more CTAs (Fig. 2C and Supplementary Figure S1).

CTAs are expressed in tumor-free liver tissues of HCC patients
Despite the TFL being located at least 2 cm away from the tumor and being classi ed as tumor-free by a pathologist, all 12 CTAs were expressed in these tumor-free liver tissues of HCC patients, although at signi cantly lower levels ( Table 2 and Fig. 2A). Forty-ve percent of patients expressed at least one CTA in TFL (Supplementary Figure S1). The CTAs most frequently expressed in TFL were MAGEA1 (13% of patients), MAGEC1 (32%) and MAGEC2 (19%). The latter two were also found to be expressed in approximately 25% of cirrhotic liver tissues of HCC-patients without liver cancer, suggesting that their expression may be activated during early (pre-)malignant transformations in the liver. Interestingly, when a particular CTA was detected in TFL, it was often also present in the tumor (Supplementary Figure S2); 85% of patients that expressed any CTA in TFL, also had CTA expression in tumor. For example, LIHCC-064 expressed 7 CTAs in tumor, of which 5 were also expressed in TFL, suggesting that CTA-expressing cells in TFL were derived from the primary tumor.
CTA mRNA expression in tumor correlates with vascular invasion and PAGE1 is associated with HBV To assess whether CTA expression in tumors was associated with certain etiologies or clinical factors, a clustering analysis of the patients based on the relative expression of the CTAs was performed (Supplementary Figure S3B). Patients with vascular invasion clustered, and had higher relative CTA expression levels and a higher number of CTAs expressed in their tumors compared to patients without vascular invasion in general. The majority of tumors that did not express any CTA, did not display vascular invasion (17/22 patients, 77%).
To study this in more detail, we also analyzed the CTA expression per etiology or known clinicopathological risk factors. Hepatitis C infection, serum alpha-fetoprotein (AFP)-level (≤ 400 ng/ml vs > 400 ng/ml) and differentiation grade were not associated with CTA expression in tumors (Supplementary Figure S4). However, tumors in cirrhotic livers had signi cantly less frequent expression of MAGEC1, whereas HBV-related tumors expressed PAGE1 more frequently. In line with the clustering analysis, tumors with vascular invasion signi cantly expressed MAGEA9, MAGEC1 and SLCO6A1 more frequently, and the number of CTAs expressed per tumor was also signi cantly higher (Fig. 2E).  Figure S5). (26) CT47A1, PAGE1, MAGEA9, MAGEC2 and MAGEA1 were detected at protein level in tumor tissues (CT47A1 in 14%, PAGE1 in 23%, MAGEA9 in 11%, MAGEC2 in 59% and MAGEA1 in 34% of tumors; Fig. 3). Seventyone percent of HCC tumor tissues expressed at least one of these CTA on protein level (Fig. 3C). MAGEA9 was not expressed in any TFL tissue, while we observed expression of CT47A1, PAGE1, MAGEC2 and MAGEA1 in hepatocytes in 1%, 3%, 17% and 9% of TFL tissues, respectively, but at signi cantly lower expression levels than in tumors (Fig. 3B). Thirty percent of patients expressed at least one protein in their TFL tissue. Most protein expression was focal, as illustrated by the observation that in most patients only part of the tumor cores included in the TMA showed protein expression (Supplementary Figure S6).
CTA protein expression in tumors showed similar associations with etiological and clinicopathological factors as CTA mRNA expression in tumors, but signi cance was not reached (Supplementary Figure S7).
HCC tumors with vascular invasion tended to have more CTA protein expressed than those without and PAGE1 protein tended to be more prevalently expressed in HBV-related tumors.
In conclusion, the CTAs that were studied for protein expression, also showed protein expression in tumors and, except MAGEA9, also in TFL.
CTA expression in TFL is correlated with HCC recurrence and HCC-speci c survival after surgical resection Early recurrence, de ned as HCC recurrence within 2 years, was observed in 64% of patients with CTA expression in TFL versus 40% in those without. Two-year HCC-speci c survival rates were 71% and 89% in patients with and without CTA expression in TFL, respectively. CTA protein expression (Fig. 4C) was associated with poor postsurgical outcome as well (Fig. 4D). In multivariate analysis both CTA mRNA and protein expression in TFL were independent prognostic factors for HCC recurrence (hazard ratio [HR] 2.48 and 2.47 for mRNA and protein expression, respectively) and HCC-speci c survival (HR 2.32 and 4.99, respectively; Table 3 and Supplementary Table S8). In conclusion, we found that CTA expression in TFL is an independent negative prognostic factor of both HCC recurrence and HCC-speci c survival. This may indicate that occult CTA-expressing (pre-)malignant cells are present in the remaining liver tissue after tumor resection and that these cells are at least partially responsible for HCC recurrence after surgery.

Discussion
We established a novel panel of 12 CTAs, that are each expressed in at least 10% of HCC tumors, while none of them are expressed in healthy tissues except immune-privileged testis. Based on mRNA analysis, approximately 80% of our HCC-patients expressed one or more of these antigens in their tumor tissues, whereas protein expression of ve of these CTAs was detected in tumors of approximately 70% of these patients. This CTA-panel can therefore be safely applied for immunotherapeutic purposes in the majority of Western HCC-patients. In addition, we found that 45% of HCC-patients expressed one or more of the 12 CTAs of our panel in their histologically tumor-free liver tissue, and that expression in TFL was associated with more HCC recurrence and worse patient survival. These data suggest that occult CTA-expressing (pre-)malignant cells may remain present in non-cancerous liver tissue after tumor resection, and that these cells might be at least partially responsible for HCC recurrence after surgery.
CTA expression in tumors of HCC-patients has been studied before, however, as demonstrated by the results of our literature study (Supplementary Table S3), most studies investigated only a few CTAs, determined either RNA or protein expression but not both, and most importantly, did not exclude CTAs expressed in healthy tissues ( Fig. 1B and C, Supplementary Tables S4). As far as we are aware, the present study is the most comprehensive investigation of CTA-expression in tumor and paired TFL tissues of HCC-patients performed. Moreover, we determined both RNA and protein expression and excluded any CTA that showed up to 4-log lower RNA expression compared to the mean of 3 reference genes in any healthy tissue, in order to prevent therapy-induced auto-immunity in future clinical applications. As 59% of HCC-patients expressed at least 2 CTAs, 50% expressed at least 3 CTAs, and 40% expressed 4 or more CTAs in their tumors, our CTA-panel enables therapeutic targeting of multiple CTAs in most HCC-patients, which is important to prevent escape of tumor cells that do not express a particular CTA from therapy-induced immunity.
Currently, somatic mutation-derived neo-antigens are considered to be the most promising candidate antigens for therapeutic vaccination in cancer patients. (28)(29)(30)  Expression of CTAs in tumor-free liver tissues of HCC patients has been sparsely investigated before (Supplementary Table S3). Our study is the rst to analyze CTA expression in tumor-free liver tissues of HCC-patients both at RNA and protein level. Moreover, this study is the rst to analyze in HCC, or in any other type of cancer, whether CTA expression in non-cancerous tumor-surrounding tissues is associated with post-operative HCC recurrence and patient survival. To our surprise, we observed RNA expression of one or more of the 12 CTAs of our panel in histologically tumor-free liver tissues of 45% of our HCCpatients, while protein expression of one or more of 4 of these CTAs was detected in non-cancerous liver tissues of 40% of patients. Most notably, we found that expression in TFL was associated with faster and more HCC recurrence as well as worse patient survival after tumor resection. Aufhauser et al. (38) hypothesized that early HCC recurrence (< 2 years) after tumor resection in HCC patients is the consequence of occult multi-focality present at the time of tumor resection, but failed to nd markers to identify such occult metastases. The 2-year recurrence rate in our cohort was signi cantly higher in patients with CTA-expression in TFL compared to patients without CTA-expression in TFL. Moreover, CTA mRNA expression pro les in TFL were similar to those in the corresponding tumors, and our preliminary immunohistochemical data show that CTA-expressing cells in TFL were either single cells or small foci.
Based on these observations, we hypothesize that CTA-expressing cells in TFL of patients with early HCC recurrence represent intra-hepatic metastases. This hypothesis is supported by a study performed in colorectal cancer patients with liver metastasis. In TFL, they detected low frequencies of somatic mutations that were also observed in matched tumor samples, despite appearing normal histologically. Since these mutations were not found in the matched blood samples, it was hypothesized that either We acknowledge several limitations of this study. First, since the etiologies of HCC differ geographically, this CTA-panel might not be applicable to non-Western HCC-populations. Secondly, protein expression of CAGE1, MAGEB2 and TSPY needs to be con rmed. Thirdly, the reported association between CTA expression in TFL and cancer recurrence has to validated in another cohort. Finally, future research is required to investigate whether CTA-expressing cells in TFL are really (pre-)malignant cells that can give rise to cancer recurrence.

Conclusions
We established a panel of 12 testis-restricted CTAs that are expressed in almost 80% of HCC patients. In addition, we demonstrated expression of these CTAs in tumor-free liver tissues of 45% of HCC-patients.
The negative association between expression of these CTAs in TFL and HCC-recurrence and survival, combined with immunohistochemical data, suggest that CTA-expressing cells remain present in the liver after tumor resection and are at least partially responsible for HCC recurrence. Therefore,

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.