Serum PD-1 Levels Change with Immunotherapy Response but Do Not Predict Prognosis in Patients with Hepatocellular Carcinoma

1. Patients

We enrolled patients with HCC who underwent surgical resection between 2012 and 2017 at Severance Hospital, Seoul, Korea. The diagnosis of HCC was made according to current guidelines [18], and matched tissue and serum samples were obtained from patients. The exclusion criteria were age ≤18 years, history of malignancy other than HCC, and history of liver transplantation. Ten serum samples from a healthy control cohort were retrieved from Yonsei Liver Blood Bank. In addition, serial serum samples were collected from 9 patients with HCC undergoing nivolumab treatment. Patients received 240 mg nivolumab intravenously every 2 weeks. This study was approved by the Independent Institutional Review Board of Severance Hospital (IRB number: 4-2016-1107) and conformed to the ethical guidelines of the 1975 Declaration of Helsinki.

2. Collection of tissue and blood samples

The tissue and blood specimens were obtained from a biobank at Severance Hospital, Seoul, Korea. Primary HCC tissues (n=78) and matching non-neoplastic tissue were collected immediately following surgery and stored at -80℃ until processing and use. On the day of surgery, blood samples were obtained from each patient. The serum was obtained by centrifugation (3,000×g for 10 minutes) and stored at -80℃ until analysis.

3. Immunohistochemistry and Enzyme-Linked Immunosorbent Assay (ELISA) assay

The quantification of PD-L1 expression and tumor-infiltrating lymphocytes (TIL), including cluster of differentiation (CD)3 and CD8, were analyzed by immunohistochemical detection. The tissue samples were immersed overnight in freshly prepared 10% neutral-buffered formalin. Fixed tissue samples were embedded in paraffin and sectioned serially into 6 μm slices. The slices were stained with hematoxylin and eosin. The paraffin sections were deparaffinized in xylene and rehydrated in gradually decreasing ethanol concentrations. The antigen epitopes were then unmasked using sodium citrate buffer (6.0 pH). Subsequently, the sections were incubated overnight at 4℃ with the primary antibody PD-L1 (13684S; Cell Signaling Technology, Inc., Danvers, MA, USA). After primary antibody incubation, the sections were incubated with the appropriate biotinylated secondary antibody (PK-6101; Vector Laboratories Inc, Burlingame, CA, USA) followed by treatment with freshly prepared 3,3'-diaminobenzidine substrate (SK-4100; Vector Laboratories Inc). The sections were lightly counter-stained with hematoxylin and mounted.

The expression levels of PD-L1 in HCC tissues were scored. Scores were determined based on the percentage of positively stained immunoreactive cells and the staining intensity in tumor cells (score 1, no staining; score 2, <10% and weak staining; score 3, 10-50% and moderate staining; and score 4, >50% and strong staining; Fig. 1) [16].

Figure 1.

Immunohistochemical staining of hepatocellular carcinoma (HCC) tissues using anti-PD-L1. Representative staining patterns of HCC with negative, score 1 (A), weak intensity, score 2 (B), moderate intensity, score 3 (C), and strong intensity, score 4 (D) staining of PD-L1 are shown (×400). PD-L1, programmed cell death ligand 1.

The serum samples were assayed by using commercially available ELISA kits for PD-1 and PD-L1 (Invitrogen, Corp., Carlsbad, CA, USA), following the manufacturer’s protocol. Briefly, 96-well plates were incubated with standards at different concentrations, and serum samples were incubated for 2 hours at 37℃. After covering of the biotinylated antibodies and several aspiration/wash cycles, the plates were incubated with horseradish peroxidase-conjugated streptavidin at 37℃ for 1 hour and protected from light. Enzymatic reactions, as indicated by blue coloration on the addition of the substrate solution were developed and the absorbance was measured at 450 nm (A450) in a VersaMax ELISA microplate reader (Molecular Devices, San Jose, CA, USA). The protein levels were calculated according to standard curves. To properly evaluate the results, parallel investigations were made using healthy untreated controls. The minimum detectable level of sPD-L1 was 31.621 ng/mL. The level of soluble sPD-L1 was determined as described above, and the minimum detectable level of sPD-1 was 10.915 ng/mL.

4. Statistical analyses

Continuous data are presented as median values (Q1-Q3), and categorical data are presented as counts and percentages. For continuous data, Student's t -test was used to test for differences between groups. When the assumption of normality was violated, a nonparametric test (Mann-Whitney test or Wilcoxon signed-rank test) was performed. Differences between categorical variables were examined for statistical significance chi-squared test (or Fisher’s exact test, if appropriate). A two-sided P -value<0.05 was considered to indicate statistical significance. Pearson’s chi-squared test and Fisher’ s exact test were used for categorical data. Spearman’s correlation coefficient was used to assess the association between the levels of tPD-L1 and sPD-L1/sPD-1. The trend between sPD-L1/sPD-1 and tPD-L1 levels was tested using the Jonckheere-Terpstra test. Cox proportional-hazard models were performed to estimate the risk factors for overall survival. Cut-off values for sPD-L1/sPD-1 levels were calculated using the Contal and O’Quigley method. All analyses were performed using SAS software (version 9.4; SAS Institute Inc., Cary, NC, USA).

1. Baseline characteristics

A total of 72 patients with HCC were included in this study. The baseline characteristics of the study population are summarized in Table 1. The median age was 58.5 years and male sex predominated (80.6%). The etiologies of liver disease were hepatitis B (n=53, 73.6%), nonalcoholic steatohepatitis (n=13, 18.1%), hepatitis C (n=4, 5.6%), and others (n=2, 2.8%). The median alanine aminotransferase and platelet levels were 27.0 IU/mL and 147 × 10³/μL, respectively. The median prothrombin time (internationalized ratio) was 1.0 and the median value of alpha-fetoprotein (AFP) was 12.6 ng/mL. The tPD-L1 score was 1 in two (2.8%) patients, 2 in 16 (22.2%) patients, 3 in 27 (37.5%) patients, and 4 in 27 (37.5%) patients.

Table 1.

Baseline characteristics of the study population

2. Levels of sPD-L1 and sPD-1

The median sPD-L1 level was 25.72 pg/mL (Q1-Q3, 18.51-67.02 pg/mL). The optimal cut-off values for the prediction of survival and HCC recurrence were 19.18 pg/mL and 34.03 pg/mL, respectively. To define the sPD-L1 reference levels, sPD-L1 concentrations were determined from the serum of healthy blood donors. The median sPD-L1 level from 10 healthy donors (three male and seven female) with a median age of 28.5 years (Q1-Q3, 25.5-30.8 years) was 1.88 pg/mL (Q1-Q3, 0.53-4.09 pg/mL).

The median sPD-1 level was 341.44 pg/mL (interquartile range, 152.59-584.51 pg/mL). The optimal cut-off values for the prediction of survival and HCC recurrence were 279.10 pg/mL and 719.64 pg/mL, respectively. The median value from healthy donors was 28.5 pg/mL (Q1-Q3, 25.5-30.8 pg/mL). No difference was found between sexes. A positive correlation was detected between tPD-L1 and sPD-1 in patients with HCC (R² =0.430, P <0.001). The median sPD-1 level increased linearly with the tPD-L1 score (P =0.002; Fig. 2).

Figure 2.

Linear trend between tPD-L1 and sPD-1 levels. tPD-L1, PD-L1 levels in tissue; sPD-1, PD-1 in serum; PD-1, programmed deathligand 1.

3. Liver-related mortality and recurrence of HCC

During the follow-up period (median, 16.1 months), liver-related mortality occurred in eight (10.3%) patients. No significant difference was observed in the baseline characteristics, except for the tPD-L1 score in patients with and without liver-related mortality (Table 2). The tPD-L1 scores were significantly higher in patients who died than in those who survived (tPD-L1 score ≥3, 100% vs. 71.9%, P =0.035).

Table 2.

Comparison of patients without and with liver-related mortality

HCC recurred in eight (10.3%) patients during the follow-up. No difference was observed between patients with and without HCC recurrence (Table 3). In addition, no significant association was observed between the sPD-L1/sPD-1 level and recurrence of HCC (data not shown).

Table 3.

Comparison of patients without and with hepatocellular carcinoma recurrence

4. Immunohistochemical analysis of TIL

An immunohistochemical assessment of TILs was performed in an attempt to evaluate the contribution of relevant cells in cancer immune landscape, including cytotoxic, helper, and regulatory T-cells and NK cells. Considering the close relationship between CD3+ cell intensity and tPD-L1 expression, we found lower PD-L1 expression together with negative CD3+ infiltration in tumor and higher PD-L1 expression accompanied with positive CD3+ infiltration. A significant correlation was detected between PD-L1 expression and CD8+ lymphocytes and was the same as the correlation between PD-L1 expression and CD3+ lymphocytes (Supplementary Fig. 1).

5. Independent factors for survival in patients with HCC

A multivariate analysis showed that higher sPD-L1 levels (cut-off value, ≥19.2 pg/mL) tended to be associated with liver-related mortality (hazard ratio [HR], 6.866; 95% confidence interval, 0.804-58.659), but this association did not reach statistical significance (P =0.078; Table 4).

Table 4.

Independent factors for survival in patients with hepatocellular carcinoma

6. Changes of sPD-1 level in patients with advanced HCC treated with nivolumab

Serially collected serum samples from a total of nine patients with advanced HCC who received nivolumab treatment were included for analysis (Supplementary Table 1). The median age was 60.0 years, and male sex predominated (88.9%). All patients had extrahepatic metastasis. The mean number of nivolumab administrations was 9.4 (range, 6-13). The changes in PD-1 are shown in Fig. 3. Of the patients who responded to nivolumab treatment, the best responses during the treatment were categorized as partial response (PR; n=2, 22.2%), stable disease (SD; n=3, 33.3%), and progressive disease (PD; n=4, 44.5%). Except for one of the five favorable therapeutic responses, a reduction of sPD-1 by ≥50% compared with baseline was observed immediately after nivolumab administration. One of the favorable therapeutic responses showed a serial decrease after 9 doses but eventually showed PD. The final responses were PR and SD in one patient each after the administration of 12 doses, and these patients ultimately developed PD. However, no change in the sPD-L1 level that matched the therapeutic response was observed (Supplementary Fig. 2).

Figure 3.

P1-P9 refers to each patient case. Change of sPD-1 levels according to nivolumab treatment. sPD-1, PD-1 in serum; PD-1, programmed death-ligand 1.