Characterization of immune infiltration in sarcomatoid hepatocellular carcinoma

Typical pathological features of sHCC

As shown in Figure 1A, sHCC was characterized by pleomorphic spindle cells. Observation indicated that the sarcomatoid components consisted of long and short spindle-shaped cells, marked by elongated nuclei with conspicuous nucleoli and spindle-shaped eosinophilic cytoplasm.

Expressions of PD-L1, B7-H3, and IDO in sHCC

PD-L1 expression was significantly higher in tumor tissue than in peritumoral tissue (Figure 1B). Of the 31 samples analyzed, 10 (32%) exhibited PD-L1 expression on the tumor cell membrane, whereas only 2 (6.5%) showed PD-L1 expression in the nonneoplastic hepatocytes. In tumor tissue, PD-L1 expression was significantly higher in sarcomatoid components than in conventional HCC components (p < 0.05). Notably, it was difficult to visualize PD-L1 expression in stromal cells under our immunohistochemistry assay (Supplementary Figure 1). Therefore, our study work started to focus on PD-L1 expression in tumor cells only.

The expression of B7-H3, as a member of the B7 family [11], was also evaluated in sHCC. Across all samples analyzed, B7-H3 expression was higher in tumor tissue than in peritumoral tissue. In tumor tissue, sarcomatoid components exhibited significantly higher B7-H3 scores than conventional HCC components (p < 0.05).

Beside B7-H3, another major inhibitor of immune responses is IDO, and its expression in tumors is associated with a worse prognosis [12]. In sHCC tissue, peritumoral tissue exhibited lower IDO⁺ cell density (27.6 ± 2.2 [mean ± standard deviation] cells/mm²) than tumor tissue. However, IDO expression was not significantly different between sarcomatoid components and conventional HCC components (48.9 ± 2.8 vs. 43.6 ± 3.9 cells/mm²).

CD8⁺, FOXP3⁺, CD68⁺, PD-1⁺, and LAG-3⁺ cell densities in sHCC

Because T cells play a critical role in tumor immunity, cytotoxic T cells (CD8⁺) and regulatory T cells (FOXP3⁺) were measured in peritumoral, conventional HCC, and sarcomatoid regions for each sample. The average CD8⁺ cell density was significantly higher in peritumoral tissue (169.8 ± 11.2 cells/mm²) than in tumor tissue (Figure 2). Additionally, conventional HCC components exhibited a significantly higher density of CD8⁺ cells than sarcomatoid components (97.0 ± 9.5 vs. 66.3 ± 7.5 cells/mm²). The average FOXP3⁺ cell density was lower in peritumoral tissue than in tumor tissue. In tumors, FOXP3⁺ cells had higher density in sarcomatoid components than in conventional HCC components.

Tumor-associated macrophages (TAMs) are key components of immune cells in tumor microenvironment [13–15]. However, no significant difference in density of CD68⁺ TAMs was found among peritumoral tissue, conventional HCC components, and sarcomatoid components (75.1 ± 6.6, 82.5 ± 6.8, and 90.19 ± 5.1 cells/mm², respectively).

PD-1 and LAG-3 are two major immune checkpoint proteins to induce T-cell inactivation. Similar to CD8⁺ T cells, PD-1⁺ lymphocytes were found to have higher concentration in peritumoral tissue (76.8 ± 13.8 cells/mm²) than in either sarcomatoid or conventional HCC components. Although the PD-1⁺ cell density was lower in sarcomatoid components than in conventional HCC components (21.7 ± 4.4 vs. 36.5 ± 6.7 cells/mm², respectively), this difference was still not statistically significant. The density of LAG-3⁺ tumor-infiltrating lymphocytes was lower in peritumoral tissue than in tumor tissue. In tumors, the density of LAG-3⁺ tumor-infiltrating lymphocytes was higher in sarcomatoid components than in conventional HCC components.

Association between CD8⁺ T cells and expressions of immune checkpoint markers in sHCC

Given that PD-L1 and B7-H3 are associated with both T-cell activation and infiltration in tumors, levels of CD8⁺ T-cell densities were compared in PD-L1⁺ sHCC and PD-L1⁻ sHCC samples. As shown in Figure 3A, PD-L1⁺ sHCC samples had a significantly higher CD8⁺ T-cell density than PD-L1⁻ sHCC samples (104.5 ± 14.2 cells/mm² vs. 70.7 ± 5.8 cells/mm², respectively). Additionally, levels of CD8⁺ T-cell densities were also compared in B7-H3⁺ sHCC and B7-H3⁻ sHCC samples. Although B7-H3⁺ sHCC samples had higher CD8⁺ T-cell density than B7-H3⁻ sHCC samples (85.7 ± 6.9 cells/mm² vs. 54.1 ± 11.4 cells/mm², respectively), the difference was still not statistically significant. Furthermore, a positive linear relationship was observed between IDO⁺ cell density and CD8⁺ T-cell density in tumors (R = 0.09); however, the difference was not statistically significant (p = 0.10).

Relationships among immune checkpoint markers in sHCC

Next, PD-1⁺ and LAG-3⁺ cells were compared in PD-L1⁺ and PD-L1⁻ sHCC samples. As shown in Figure 3B, 3C, PD-1⁺ and LAG-3⁺ cell densities were higher in PD-L1⁺ tumor samples than in PD-L1⁻ tumor samples (PD-1⁺: 47.3 ± 11.7 vs. 24.0 ± 4.5 cells/mm²; LAG-3⁺: 54.8 ± 10.7 vs. 29.1 ± 6.7 cells/mm²). Additionally, B7-H3⁺ tumor samples exhibited higher PD-1⁺ and LAG-3⁺ cell densities than B7-H3⁻ tumor samples. However, the difference was also not statistically significant (PD-1⁺: 33.09 ± 5.6 vs. 20.6 ± 7.3 cells/mm²; LAG-3⁺: 38.3 ± 6.4 vs. 31.0 ± 13.6 cells/mm²).

Prognostic factors

Our studied patient population had 1-, 3-, and 5-year OS rates of 67.7%, 58.0%, and 51.6%, respectively. The 1-, 3-, and 5-year DFS rates were 67.7%, 54.8%, and 40.1%, respectively.

Results of the Kaplan-Meier analysis revealed that 1-, 3-, and 5-year OS rates in sHCC patients with PD-L1⁺ expression (50.0%, 20.0%, and 20.0%, respectively) were significantly lower than those in sHCC patients with PD-L1⁻ expression (76.2%, 71.4%, and 66.7%, respectively) (Figure 4A). Moreover, the 1-, 3-, and 5-year DFS rates in sHCC patients with PD-L1⁺ expression (40.0%, 30.0%, and 30.0%, respectively) were significantly lower than those in sHCC patients with PD-L1⁻ expression (80.9%, 61.9%, and 52.3%, respectively) (Figure 4B). By contrast, the 1-, 3-, and 5-year OS rates in sHCC patients with a high density of CD8⁺ cells (86.7%, 73.3%, and 66.7%, respectively) were significantly higher than those in sHCC patients with a low density of CD8⁺ cells (50.0%, 37.5%, and 37.5%, respectively) (Figure 4C). Additionally, the 1-, 3-, and 5-year DFS rates in sHCC patients with a high density of CD8⁺ cells (80.0%, 60.0%, and 53.3%, respectively) were significantly higher than those in sHCC patients with a low density of CD8⁺ cells (56.3%, 37.5%, and 37.5%, respectively) (Figure 4D).

Results of the univariate analysis indicated that the following factors did not predict both OS and DFS: age, sex, hepatitis B virus, liver cirrhosis, alpha-fetoprotein concentration, alanine aminotransferase concentration, tumor number, and tumor encapsulation. For DFS, the tumor size, PD-L1 expression, and CD8⁺ cell density were the predictors. Similarly, these three predictors, as well as vascular invasion, were associated with OS (Table 1). For the multivariate Cox proportional hazards analyses, the covariates were clinicopathologic features that showed significance in the univariate analysis (Table 2). Vascular invasion remained association with OS (p = 0.028). Additionally, both PD-L1 expression and CD8⁺ cell density were independent prognostic factors for both improved OS (p = 0.012 and p = 0.014, respectively) and DFS (p = 0.016 and p = 0.009, respectively).

Patient information and follow-up details

Between March 2013 and April 2017, 53 sHCC tissue samples were collected from patients undergoing surgical resection at the Zhongshan Hospital, Fudan University (Shanghai, China). Patients with unavailable tumor and peritumoral samples or incomplete clinicopathologic and follow-up data were excluded from the study. In total, our study included 31 patients with sHCC. The clinical information of these patients is listed in the Supplementary Table 1.

Hematoxylin and eosin staining and immunohistochemical staining (Hep Par-1, arginase-1, cytokeratin, and vimentin) were reviewed by two pathologists to confirm the pathological diagnosis of sHCC. All tissue samples from 31 sHCC patients were mixed with peritumoral, conventional HCC, and sHCC components. All patients gave their written informed consent for their samples to be used in our study. Surveillance of postsurgical patients was described in a previous study [31].

Immunohistochemical staining assay and evaluation

Immunohistochemical staining was conducted on whole-slide serial tissue sections from FFPE surgical resection specimens to determine the expression levels for PD-L1, B7-H3, IDO, CD8, FOXP3, CD68, PD-1, and LAG-3, according to standardized institutional protocols as previously described [32]. The reagents used for this evaluation are listed in Supplementary Table 2.

Briefly, one representative FFPE block from each tumor specimen (n = 31) was selected with tumor and peritumoral liver tissue present on the same block. Consecutive 5-μm-thick sections were sliced from each block and placed on cleaned glass slides. After deparaffinization/rehydration, FPEE sections were subjected to endogenous peroxidase blocking, heat-induced antigen retrieval, antibody incubation at specific dilutions, and signal detection.

Two pathologists who were blinded to the clinical outcomes independently assessed the immunostaining results. A cell membrane staining algorithm was applied to analyze the levels of PD-L1 and B7-H3 expression in tumor cells. This expression was quantified using a four-value staining intensity score (0, none; 1, weak; 2, moderate; 3, strong) and the extension (percentage) of expression. H-score was calculated by multiplying the intensity and reactivity extension values (H-score range, 0–300) [33]. For both PD-L1 and B7-H3, tumor cells with expression levels greater than 5% were considered to be positive. For the profiling analysis of CD8, FOXP3, CD68, PD-1, LAG-3, and IDO, cell density was measured by counting the number of positive cells expressing these markers in three randomly selected fields. For each marker, the average of these values was expressed as the number of cells per square millimeters.

Statistical analyses

The mean expression levels of PD-L1, B7-H3, IDO, CD8, FOXP3, CD68, LAG-3, and PD-1 in various tissue compartments (peritumoral, conventional HCC component, and sarcomatoid component) were compared using Wilcoxon signed-rank tests. The statistical analyses were conducted using STATA 14.0 (Stata Corp.). A correlation analysis was performed using Spearman linear correlation analysis. The Kaplan-Meier method was used to determine OS and DFS, and the log-rank test was applied to analyze the differences. Univariate and multivariate analyses were conducted using the Cox proportional hazards regression model. A p-value less than 0.05 was statistically significant.