Introduction

Hepatocellular carcinoma (HCC), the major pathological type of liver cancer (comprising 75–85% of cases), is still a leading cause of cancer-related death worldwide [1] and with very limited therapeutic options. Particularly in advanced stage, long-term survival is uncommon [2, 3].

Since cancer survival is generally influenced by different prognostic factors such as cytogenetic abnormalities, immune cell infiltration, and stage [4,5,6], the identification of prognostic factors in HCC may lead to novel and more effective treatments.

Neoantigens are epitopes derived from abnormal proteins encoded by mutant cancer genomes and are considered important targets for cancer immunotherapy because of their immunogenicity and lack of expression in normal tissues. The epitopes are displayed by major histocompatibility complex (MHC) molecules on tumor cells to trigger the activation of cytotoxic T cells [7, 8]. Next generation sequencing technologies and computational analysis have recently made neoantigen discovery possible. Immune cells play a dual role in mediating the anti-tumor immune responses, which can either suppress cancer development by generating cytotoxic T cells or promote cancer through the expression of immunosuppressive factors [9]. Neoantigens have been strongly linked to tumor mutation burden (TMB) and the clinical benefit of checkpoint blockade therapy in melanoma, non-small cell lung cancer (NSCLC) and other cancer [10,11,12,13]. Furthermore, immunogenic personal vaccines based on neoantigens have shown dramatic efficacy in melanoma and glioblastoma, providing tumor-specific immune targets that enhance the efficacy of cancer immunotherapy [14, 15].

Despite recent studies implicating neoantigens in the treatment and prognosis of several cancer types [16,17,18], the relationships between tumor neoantigens, immune cells and prognosis in HCC have not been reported. Here we fill this gap by identify prognostic markers in long-term HCC survivors and evaluate the relationship of neoantigen load, unique neoantigens, immune cell infiltration and prognosis.

Materials and methods

Patient and sample characteristics

Twenty two HCC participants were recruited at Peking Union Medical College Hospital (PUMCH) between March 2019 and July 2019, with ages ranging from 29 to 77 years old (median age: 56 years). All pathology specimens (stages ranged from IIIA to IVA) were reviewed by experienced gastrointestinal pathologists. All patients had received standard therapy with surgery alone or surgery plus chemoradiotherapy. No patients had received immunotherapy treatment prior to surgery. All tumor tissues were surgically resected primary HCC and confirmed by pathological and histological re-review by two expert HCC pathologists before analysis. Informed consent was obtained from all participants. This study was approved by the Ethical Committee of PUMCH and was performed in accordance with the ethical standards of the World Medical Association Declaration of Helsinki.

Whole-exome sequencing and somatic mutation calling

Tumor and matched normal DNA were extracted using the GeneRead DNA FFPE Kit from formalin-fixed paraffin-embedded (FFPE) tissues. Libraries were constructed by the Agilent SureSelect v.4 Kit (Agilent) and sequenced with next-generation sequencing. Genomic DNA was fragmented, end-repaired, adenylated at the 3′ ends, end-connected, amplified, purified, and size-selected in the process of library construction, then was sequenced on the Illumina X10 platform (Illumina Inc., San Diego, CA, USA).

WES data underwent mutation analysis and human genome build hg19 was used as the reference genome. Somatic SNVs and InDels were analyzed via GATK MuTect2. The sequenced reads were realigned to the hg19 by Burrows–Wheeler Aligner BWA-MEM (https://biobwa.sourceforge.net/) to enhance valid SNVs.

RNA sequencing

Tumor RNA was extracted using the RNeasy FFPE Kit (Qiagen) from FFPE tissues. Libraries were constructed using a TruSeq RNA Exome kit and sequenced with NGS. Total RNA was fragmented, reverse transcribed into complementary DNA, base ‘A’ added in the 3′ ends, adapter connected, amplified and purified, and then sequenced on Illumina X10 platform (Illumina Inc., San Diego, CA, USA).

Clean reads were mapped to the hg19 human genome using Bowtie2 (version 2.2.4) software from Tophat2 (version 2.0.10) with default parameters. The program Cufflinks (version 2.2.1) was used to calculate the expression levels of genes in terms of reads per kilobase per million reads (FPKM).

Neoantigen predictions

HLA typing was acquired from normal DNA using OptiType (version 1.3.1). The cancer immunotherapy pipeline—pVACseq, was used to identify neoantigens [19].

Peptide-MHC affinity for half maximal inhibitory concentration (IC50) values were predicted using NetMHC/NetMHCIIpan or PickPocket. IC50 < 500 nM is considered an accepted standard in the field for predicting binders, and IC50 ≤ 150 nM is accepted as a strong binder [20].

Mutated peptides with a binding affinity of IC50 < 500 nM was regarded as candidate neoantigens, and mutated peptides (IC50 ≤ 150 nM) were regarded as strong-quality neoantigens [15, 18, 21]. Neoantigen expression was confirmed in any such neoantigens with RNA-Seq counts ≥ 1 [4, 21].

Assessment of immune infiltration and immune cytolytic activity

Immune score

Based on the gene expression (FPKM), an R package—ESTIMATE [22] (https://sourceforge.net/projects/estimateproject/), was used to estimate the fraction of stromal and immune cells in tumor samples. “Immune score” aimed to represent the infiltration of immune cells in tumor tissue.

MCP-counter score

An MCP-counter [23] method (https://github.com/ebecht/MCPcounter) was used to infer the absolute intratumor cell abundance of ten cell types (eight immune and two stromal cell populations) in each tumor tissue.

CYT score

Based on the notion that effective natural anti-tumor immunity requires a cytolytic immune response, we quantified immune cytolytic activity using an expression metric of effector molecules that mediate cytolysis. Immune cytolytic activity (CYT score) was calculated as the geometric mean of GZMA and PRF1 (as expressed in TPM) [24].

Immunohistochemistry

Antibodies against CD3, CD8, CD45RO, and FOXP3 were purchased from Abcam. Immunohistochemistry (IHC) was performed according to the manufacturer’s instructions. Four 5-μm sections were cut from each case. After dewaxing, slides were boiled with 1 mM EDTA pH 8.0 followed by 15 min at a sub-boiling temperature. Slides were washed with phosphate-buffered saline at three times for 5 min each, subsequently quenched in 3% hydrogen peroxide for 15 min, and then blocked with 10% goat serum for 10 min. Slides were incubated overnight at 4 °C with the primary antibody diluent (1:100 anti-CD3 antibody, ab16669. 1:100 anti-CD8 antibody, ab4055. 1:150 anti-CD45RO antibody, ab23 and 1:200 anti-FOXP3 antibody, ab20034). The slides were then incubated with biotinylated secondary antibody per the manufacturer’s recommendation for 30 min. Antibody binding was visualized with DAB.

Statistical analysis

Statistical analysis was performed using SPSS 21.0 software (IBM Corporation, New York). The distribution test was analyzed by Shapiro–Wilk test. Parametric data were calculated by t test and one-way analysis of variance. The Mann–Whitney test and Kruskal–Wallis test were used for non-parametric data. Survival was summarized by a Kaplan–Meier survival curve. The probability value was calculated by two-sided tests. p < 0.0500 was considered statistically significant. Figures were drawn using GraphPad Prism 8.0 software (San Diego, USA).

Results

Patient characteristics

A total of 22 patients with HCC were enrolled in the study, and their clinicopathological characteristics were summarized in supplementary Table 1. The median age of the included patients was 56 (IQR: 53–64) years, and 86.4% were males. No patients received pre-operative chemotherapy. 72.7% of patients had cirrhosis. The Child–Pugh score of 45.5% patients ws A, and that of 54.5% was B. Overall, 36.4% of patients were at TNM stage II, 50.0% were at stage IIIA, 9.1% were at stage IIIB, and 4.5% were at stage IIIC. The Barcelona Clinic Liver Cancer (BCLC) score of 4.5% of patients were 0, 18.2% were A1, 4.5% were A2, 9.1% were A4, 50.0% were B, and 13.6% were C. 9.1% of patients underwent hepatic segmentectomy, 72.7% underwent combined segmentectomy, and 13.6% underwent cholecystectomy. Further, 63.6% of the patients received radical resection. 27.3% of patients received no post-operative therapy, 68.1% received simple interventional therapy, and 4.5% received interventional therapy and targeted drug.

The median OS was 51.00 (IQR: 36.50–61.75) months. 63.6% of patients at last follow-up remained alive. A total of 60.0% of patients reported recurrence.

Overall survival comparison and gene mutation landscape in HCC patients

A cohort of 22 HCC patients (median OS, 51 months) undergoing treatment at Peking Union Medical College Hospital (PUMCH, Beijing, China) was analyzed for this study. Long-term survivors (n = 13, median OS, 58 months) and short-term survivors (n = 9, median OS, 35 months) group had a significant difference on survival with p = 0.0009 (Fig. 1a).

Fig. 1
figure 1

Overall survival comparison and gene mutation landscape in HCC patients. a Overall survival of 22 HCC cohort patients. b Gene mutation landscape of 22 HCC cohort patients. c Mutation frequency comparison between TCGA-HCC and top 10 mutated genes in HCC01 ~ HCC22. d Mutated gene comparison between HCC01 ~ HCC22 and TCGA-HCC

Full size image

We performed exome sequencing on genomic DNA of 22 HCC tumor and matched normal tissue. After filtering the exome sequencing data, we observed a mean of 124 somatic mutations per tumor (range 40–300). In the exome sequencing data analysis, we focused on genes that were mutated in at least two patients. This resulted in a list of 341 genes (supplementary Table 2) and most of these genes (272/341, 79.77%) were mutated in two tumors. Top 10 mutated genes of HCC were TP53, MUC4, TTN, CTNNB1, LRP1B, MUC16, ALB, CSMD2, FLG and RYR2 (Fig. 1b). The most frequently mutated gene was TP53, which was mutated in 12 tumors (12/22, 54.55%).

We compared our mutated genes with HCC patients in TCGA database and found 2061 genes were in common (Fig. 1d). Our top 10 mutated genes were all included in TCGA-HCC and had a higher mutation frequency (Fig. 1c), which may be caused by our limited patients.

Immune infiltration comparison between long term and short term

Tumors underwent RNA-seq and normalized data were analyzed to quantify immune infiltration using Immune score and MCP-Counter score. We compared the immune infiltration observed in two groups. Long-term patients had a significant higher neutrophils infiltration than Short term, p = 0.0011 (Fig. 2b). Neutrophils infiltrating tumor tissues play functions in the anti-tumor immune responses and restrict cancer growth by expressing anti-tumor and cytotoxic mediators [25,26,27]. Immune score and other cells’ infiltrated MCP-Counter score showed no statistically significant difference (Fig. 2a, b).

Fig. 2
figure 2

Immune infiltration difference with Long term and Short term. a Immune score comparison between Long term and Short term. b 10 kinds of cells’ infiltrated MCP-Counter score comparison between Long term and Short term

Full size image

The effect of TMB, neoantigen load on patient survival

Several reports have shown that TMB and neoantigen load were associated with prognosis in lung cancer and melanoma [28,29,30]. To investigate whether the same was true for HCC patients, we analyzed the relationship between TMB, neoantigen load and prognosis. TMB ranged from 0.58 to 4.28 (median = 1.73) in these patients. The total neoantigen load with IC50 < 500 nM ranged from 32 to 521 (median = 241). High-quality neoantigen load with IC50 < 150 nM ranged from 2 to 181 (median = 78).

First, TMB and neoantigen load were compared based on Long term and Short term. But there was no significant difference (Fig. 3a–d). Then all the patients were divided into two groups separately based on the median TMB and neoantigen load. There was no significant correlation between TMB, neoantigen load, and survival in HCC patients (Fig. 4a–c). It means that TMB and neoantigen load may have no contribution to survival for HCC patients in our project.

Fig. 3
figure 3

TMB, neoantigen load comparison between Long term and Short term. a TMB comparison between Long term and Short term, p = 0.8898. b Neoantigen number per patient in HCC01 ~ HCC22. c, d Neoantigen load comparison between Long term and Short term. p = 0.8459 for total neoantigen load and p = 0.6509 for high-quality neoantigen load with IC50 < 150 nM

Full size image
Fig. 4
figure 4

Correlation between TMB, neoantigen load and prognosis. a TMB and prognosis, p = 0.6531. b Total neoantigen load and prognosis, p = 0.2782. c High-quality neoantigen load (IC50 < 150 nM) and prognosis, p = 0.5892

Full size image

TP53 neoantigen predicts better prognosis in HCC patients

Some neoantigen-specific immunity gained during tumor outgrowth could be associated with decreased relapse and prolonged survival, like MUC16 neoantigen in long-term survivors of pancreatic cancer [18].

To identify neoantigens that can be used to treat or predict prognosis in HCC, we analyzed and found that patients who carried with TP53 neoantigen had a significant longer overall survival than those who did not carry TP53 neoantigen, p = 0.0371 (Fig. 5a). However, they showed no statistically significant difference on TMB and neoantigen load (Fig. 5b–d).

Fig. 5
figure 5

OS, TMB and neoantigen load comparison between patients with the TP53 neoantigen and without TP53 neoantigen. a Overall survival comparison. b TMB load comparison. c, d Neoantigen load comparison. ‘TP53 neoantigen’ group (n = 9) include patients carrying TP53 neoantigen, and ‘No TP53 neoantigen’ group (n = 13) include patients without TP53 neoantigen

Full size image

Besides, immune infiltration and immune cytolytic activity (CYT score) were estimated between ‘TP53 neoantigen’ group and ‘No TP53 neoantigen’ group. ‘TP53 neoantigen’ group had a significantly higher Immune score (Fig. 6a, p = 0.0441), higher cytotoxic lymphocytes infiltration (Fig. 6b, p = 0.0428), and higher CYT score (Fig. 6c, p = 0.0388). The immunohistochemistry results also confirmed that patients in ‘TP53 neoantigen’ group had more immune cell infiltration (Fig. 7).

Fig. 6
figure 6

Immune infiltration and cytolytic activity difference between patients with the TP53 neoantigen and without TP53 neoantigen. a Immune score comparison. b MCP-Counter score of cytotoxic lymphocytes comparison. c CYT score comparison

Full size image
Fig. 7
figure 7

Immunohistochemistry results between patients with the TP53 neoantigen (HCC19) and without TP53 neoantigen (HCC03)

Full size image

Discussion

To the best of our knowledge, this is the first study that identifies a prognostic marker for neoantigens in long-term HCC survivors and evaluates the relationship of neoantigen load, unique neoantigens, the infiltration of immune cells, and prognosis.

In this study, we predicted neoantigens by WES, RNA-Seq, and HLA affinity in 22 HCC patients and found that OS is not associated with TMB, neoantigen load. We discovered that unique TP53 neoantigen is associated with better prognosis, higher cytotoxic lymphocyte infiltration as well as immune cytolytic activity (CYT score). These results suggest that the TP53 neoantigen could affect prognosis by regulating anti-tumor immune responses among HCC patients.

There are discrepancies regarding the relationship between TMB, neoantigen load, and prognosis in various cancers. Several reports have shown that high TMB and neoantigen load are associated with poor prognosis in NSCLC, head and neck squamous cell carcinoma, adenoid cystic carcinoma, liver hepatocellular carcinoma, and myeloma [4, 31, 32]. On the other hand, other studies reported no relationship between TMB or neoantigen load and prognosis in clear cell renal cell carcinoma [28], which is more in line with our current study of HCC. The exact reasons for such discrepancies require future investigations but may be potentially due to different cancer types, treatment history, and patient sample sizes involved in the different studies.

Cytotoxic lymphocytes (including CD8 + cytotoxic lymphocytes, natural killer-NK cells and lymphokine-activated killer cells, etc.) depend primarily on the perforin or granzyme system to kill their targets [33, 34]. Cytotoxic T cells (CTLs) and NK cells release perforin 1 (PRF1), granzymes, and granulysin, upon their expose to infected or dysfunctional somatic cells. These cytotoxins can ultimately lead target cells to apoptosis by trigger a caspase cascade or cell–surface interaction between the CTL and the infected cell [35,36,37]. The expression of the toxins granzyme A (GZMA) and perforin 1 (PRF1), secreted by effector cytotoxic T cells and natural killer (NK) cells, was recently used as a denominator of the intratumoral immune cytolytic activity (CYT). These levels are significantly elevated upon CD8 + T cell activation as well as during a productive clinical response against immune-checkpoint blockade therapies [38, 39]. Another study has proposed that intrinsic tumor factors, such as mutated neoantigens, can induce local immune infiltrates that include cytolytic effector cells (expressing GZMA/PRF1) that kill tumors [24].

Our study draws a preliminary conclusion that it is not TMB or neoantigen load, but the TP53 specific neoantigen is related to overall survival of HCC patients. Patients carried TP53 neoantigen could activate immune system and trigger higher immune infiltration to lead tumor apoptosis, which could prolong patients’ life. Higher T cell infiltration was observed in TP53 neoantigen group based on the IHC and sequencing data. However, T cell did not correlate with HCC long-term prognosis, which indicates that TP53 may have other effect than the T cell infiltration. Interestingly, in previous study, Cooks et. al [40] found that TP53-mutated colon cancer cells could shed exosomes to alter tumor status with reprograming macrophages. More research is needed to understand the potential mechanisms of TP53 neoantigen.

Because our study population showed heterogeneity in patient clinicopathological characteristics and management, we evaluated the impact of these factors on patients’ survival, TP53, immune score, and CD8 T cell count. No significant correlation has been revealed under the analysis, which implies that management and patients’ status are not related to survival time, neoantigens,and immune microenvironments (supplementary Table 3–6). However, the small patients’ number limited the applicability of our study, and future researches focused on TP53 could broaden the database and further confirmed the clinical impact of TP53.

In conclusion, we propose that TP53 neoantigen ould be used as a biomarker to predict prognosis or provide a novel choice for the treatment in HCC patients someday in the future.