Multiple Immune-Suppressive Mechanisms in Fibrolamellar Carcinoma

Fibrolamellar carcinoma (FLC) is a liver cancer that was formally categorized as a subset of hepatocellular carcinoma (HCC) but now recognized as an entirely different cancer with unique clinicopathologic and molecular features (1, 2). Unlike classic HCC, FLC affects adolescents and young adults without chronic liver disease. FLC is associated with fewer chromosomal changes and less genomic heterogeneity than HCC and is characterized by the presence of a single genetic deletion in chromosome 19 and formation of a fusion of DNAJB1 and PRKACA transcripts in 79% to 100% of the cases (1–3). The most effective treatment for FLC is complete surgical resection. However, outcomes for patients with FLC are poor, with an overall survival of 30% at 5 years (4, 5), and currently, no standardized systemic treatment recommendation exists for patients with recurrent or unresectable disease. Because of this, novel and effective therapies for patients with FLC are needed.

Data on systemic treatment of FLC are limited. One retrospective study demonstrated that the combination of 5-fluorouracil and subcutaneous recombinant interferon alfa (IFNα) has clinical activity for FLC (6). The potential clinical activity of IFNα in FLC suggests that immune evasion may be an important feature of FLC oncogenesis and, therefore, may be an opportunity for therapeutic intervention. Little is known, however, about the immune microenvironment and the mechanisms of immune evasion in FLC.

Immune-checkpoint molecules such as programmed death-1 (PD-1), and its ligand (PD-L1), modulate cellular immunity and can be co-opted by cancers to sustain immune tolerance. The development of immune-checkpoint inhibitors has had a profound impact on the clinical management of many cancers, including liver cancer (7). To date, immunotherapy is not routinely administered in FLC because of the rarity of the disease and the uncertainty in its treatment response. The direct analysis of the cancer immune microenvironment has also emerged as an important step in understanding the relationship between the patients' immune systems and their cancer. For example, PD-L1 expression on tumor cells has been identified among multiple tumor types that respond to PD-1 pathway inhibition, including melanoma, lung, and renal carcinoma (8, 9). PD-L1 expression on tumor-infiltrating immune cells has also been associated with higher clinical response rate to inhibitors of the PD-1 pathway (10–12).

To date, the expression of immune-checkpoint molecules has not previously been assessed in FLC. The objective of the current study was to define the expression of checkpoint molecules PD-1, PD-L1, IDO, B7-H3, and LAG-3 on tumors and tumor immune infiltrates in a cohort of patients with FLC. The findings provide insights about the immune microenvironment in FLC.

The study was approved by the Johns Hopkins University Institutional Review Board and conducted in accordance with the ethical standards of the Declaration of Helsinki. All samples were obtained in accordance with the Health Insurance Privacy and Accountability Act. Specimens from 25 unique patients with FLC were obtained between 2000 and 2016 from the Johns Hopkins Hospital surgical pathology archives (n = 22) and the StoreMyTumor tumor bank (http://www.StoreMyTumor.com; n = 3). From these 25 patients, 32 specimens were studied (18 patients with one specimen on file, and 8 patients with multiple specimens). Hematoxylin and eosin–stained slides were reviewed by two pathologists (F. Bhaijee and R. Anders) to confirm the diagnosis of FLC, and a single representative block was chosen for additional analysis. Clinical information such as age, sex, and previous treatment history was extracted from the electronic medical record for the 22 patients from Johns Hopkins Hospital (Table 1).

Serial 5-μm sections from formalin-fixed, paraffin-embedded tissue were cut onto glass slides for each specimen. IHC staining for PD-L1, CD8, Foxp3, PD-1, IDO, LAG3, CD11c, CD68, and B7-H3 was performed by the Tumor Microenvironment Center at Johns Hopkins School of Medicine and the IHC laboratory in Johns Hopkins Hospital, using the reagents listed in Supplementary Table S1. For IDO, automated IDO1 staining was performed on the Leica Bond RX (Leica Biosystems) using the Bond Polymer Refine Kit (DS9800, Leica Biosystems). Slides were baked online using a *Bake and Dewax protocol, followed by high pH antigen retrieval using a *HIER 20 min with ER2 protocol. At room temperature, endogenous peroxidase was blocked using Peroxide Block (Leica Biosystems) for 10 minutes, followed by blocking of nonspecific binding using Protein block (X090930, Agilent Technologies) for 15 minutes. Anti-IDO1 (clone SP260; ab228468, Abcam plc.) was applied for 15 minutes at a concentration of 1.19 μg/mL. Detection was performed by applying anti-rabbit polymer-HRP (Leica Biosystems) for 8 minutes followed by DAB Refine (Leica Biosystems) for 10 minutes. Samples were counterstained using hematoxylin (Leica Biosystems) for 5 minutes. Slides were dehydrated through a graded alcohol series into Histoclear II (508990150, Thermo Fisher Scientific) and coverslipped using Ecomount (5082832, Biocare Medical). Human tonsil specimens were used as positive controls for each staining batch. An IgG control was performed for each individual case as a control for PD-L1 staining.

Scoring for PD-L1, B7-H3, and LAG3 was performed by two separate pathologists (R. Anders and A.J. Layman) blinded to the clinicopathologic features of the cases using standard brightfield microscopy (Nikon Eclipse 50i). The percentage of malignant cells demonstrating positive stain expression was scored at 5% increments, and cases demonstrating ≥ 5% tumor cell expression of a given marker were considered positive (8, 13–15). Immune cells, including tumor-infiltrating lymphocytes and tumor-associated macrophages (TIL/TAM), were scored separately as positive (≥ 5% of cells staining). The scoring of the two pathologists was subsequently averaged into a combined final score.

Slides were scanned at 20 × objective equivalent (0.49 μm/pixel) using the Hamamatsu XR NanoZoomer bright field system (Hamamatsu Photonics, K.K.). For CD8-, Foxp3-, PD-1–, and IDO-stained slides, the tumor–stroma interface was annotated. Specifically, 250 μm from the tumor–stromal border into the tumor center and further into the stroma from this demarcation were considered the “tumor interface” and “nontumor interface” zones, respectively. Additional tumor areas studied included the remaining central aspect of the tumor (referred to here as “tumor” zone), and the areas further into the stroma (away from the tumor) beyond the nontumor interface (referred to here as “nontumor” zone). Greater than 50% of the nontumor area was annotated for analysis (Supplementary Fig. S1). Digitally assisted image analysis (HALO Indica Labs) was used to assess the density (cells/mm²) of cells expressing CD8, PD-1, Foxp3, and IDO in each of the four tumor microenvironment (TME) zones described above.

Categorical variables were reported as whole numbers and proportions, whereas continuous variables were summarized as medians with interquartile range (IQR) as appropriate. Pearson χ² test was used to compare categorical variables (PDL-1 and B7-H3 staining) between patient groups, whereas quintile regression of medians was used to compare continuous data (CD8⁺, Foxp3⁺, IDO⁺, and PD-1⁺ cell densities). For all analyses, a P value of ≤0.05 was used to define statistical significance. All analyses were performed using STATA 14.0 (Stata Corp.).

The pathologic features of 32 FLC specimens are provided in Table 1. Twenty specimens (63%) were from primary tumor, two (6%) from lymph nodes, and ten (31%) were from other metastatic sites. Clinical characteristics were reported for patients from Johns Hopkins Hospital (n = 22). An even sex distribution between males and females was seen, and median patient age was 24. The majority of the patients did not receive any systemic treatment prior to their surgery. Fifty percent of the patients had disease recurrence after surgery. Other detailed treatment history and clinical outcome at the time of data collection are shown in Supplementary Table S2.

As shown in Fig. 1A, 20 of the 32 cases (63%) demonstrated membranous tumor cell expression of PD-L1 in the tumor zone, and 22 cases (69%) had positive PD-L1 expression present on the tumor-infiltrating immune cells. In contrast, only 16% of cases showed PD-L1⁺ nonneoplastic hepatocytes, which was significantly lower than the cases with positive tumor cells (P = 0.007). Forty-one percent of the cases had PD-L1 expression on immune cells in the nontumor zone. Of the cases with PD-L1 expression on TILs and TAMs, 46% of the cases had greater than 25% of the cells stained positive.

We also investigated the expression of B7-H3, another member of the B7-family closely related to PD-L1 and known to modulate T cells in FLC (Fig. 1B). It is established that B7-H3 is expressed in numerous solid tumors including lung, stomach, prostate, and HCC (16–18). Overall, expression of B7-H3 in the entire TME of FLC was observed. Forty percent of the cases showed B7-H3⁺ tumor cells in the tumor zone, which was similar to the adjacent nontumor tissue (50%). Ninety-one percent of the cases had B7-H3⁺ TILs and TAMs, and 79% showed B7-H3⁺ nontumor immune cells (Fig. 1C). As seen with PD-L1 staining on primary FLC samples, almost all immune cells, including TILs and TAMs, were located in the sinusoids of the liver, which are known to be enriched with immune cells such as macrophages (Kupffer cells) and liver-associated natural killer cells (19, 20). Comparison of the primary tumors to metastatic sites did not show any significant difference in PD-L1 and B7-H3 expression in tumor cells or TILs and TAMs (Fig. 1D).

Indoleamine 2,3-dioxygenase (IDO) is another major inhibitor of immune responses (10, 21), and its expression in tumor cells and tumor-associated immune cells is associated with a worse prognosis (22, 23). In the FLC samples, IDO expression was concentrated at the tumor interface (Fig. 2A), and its cell density was almost 3-fold greater (44.8 cells/mm²) compared with other zones (Fig. 2B). However, no significant difference in IDO expression was observed between the tumor versus nontumor interface, as well as the nontumor zones (15.4, 9.7, and 15.9 cells/mm², respectively).

The expression of LAG3 was also investigated in this cohort in the first 22 samples. In contrast to other checkpoint inhibitor expression in our FLC samples, only 10% of the samples (2 of 22 samples) had less than 1% of lymphocyte expression in the tumor area, as shown in Supplementary Table S3.

Given that T-cell activation is critical to the development of the adaptive immune resistance associated with PD-L1 expression in tumors (13), we quantified standard T-cell subsets using antibodies to CD8⁺, as well as Foxp3⁺ [representative of regulatory T cells (Tregs)] in the tumor, peritumoral interface, and nontumor regions. Overall, the median density of CD8⁺ T lymphocytes was significantly higher in the tumor interface zone compared with the tumor zone (P = 0.037), and the CD8⁺ density was similar in the tumor zone versus the nontumor zone (71 cells/mm² and 111 cells/mm², respectively; P = 0.557). This was consistent with our previous finding in HCC, where the mean CD8⁺ T-lymphocyte density was highest at the tumor interface and lowest in the tumor region (24). Foxp3⁺ cell density was lower compared with CD8⁺ cells (Fig. 3), but no statistically significant differences in Foxp3⁺ cell density were observed among the tumor regions.

Similar to the distribution of CD8⁺ cells, a high concentration of PD-1⁺ lymphocytes in the interface area was seen compared with the tumor and nontumor areas (Fig. 3), and cell densities were similar for the tumor and nontumor side of the interface for PD-1. No differences in CD8⁺ and PD-1⁺ cell densities were observed between tumor and nontumor regions (P = 0.598).

Given the association of B7-H3 and PD-L1 to T-cell activation and infiltration in tumors (13, 25, 26), CD8⁺ T-cell densities in B7-H3⁺ tumor samples were compared with B7-H3⁻ tumor samples. The median CD8⁺ T-lymphocyte density in B7-H3⁺ tumor samples was more than 5-fold greater than in B7-H3⁻ samples (245 vs. 33.1 cells/mm², P < 0.001; Fig. 4A). These data suggested that despite similar B7-H3 expression between tumor and nontumor zones (Fig. 1), there was a correlation between T-cell activation and expression of B7H3 in tumor area. A similar correlation was seen between CD8⁺ and PD-L1⁺ tumor samples, although it did not meet statistical significance (Fig. 4A). The median CD8⁺ T-lymphocyte density was 128.3 cells/mm² in samples with PD-L1⁺ tumor cells versus 34.9 cells/mm² in samples with PD-L1⁻ tumor cells (P = 0.184). A similar pattern was seen in adjacent nontumor regions, where CD8⁺ T-lymphocyte density (312.2 cells/mm²) was almost 3-fold higher in samples with PD-L1⁺ nontumor hepatocytes versus the samples with PD-L1⁻ cells (124.2 cells/mm², P = 0.089). A linear relationship between IDO⁺ cell density in the tumor zone with CD8⁺ density (P < 0.001) was noted, with a similar positive trend in the interface region (P = 0.289; Fig. 4B), also suggesting a local interaction between T cells and its ligand, IDO, in the FLC microenvironment.

In melanoma and ovarian cancer patients, PD-L1 expression on myeloid cells, including myeloid dendritic cells (DC) and macrophages, has a critical function in T-cell immunity and in response to immunotherapy (27, 28). To investigate the role of myeloid cells in FLC, we selected a subset of the FLC specimens with more than >25% PD-L1 expression in TILs and TAMs and stained for DCs and macrophages with CD11c and CD68 antibodies, respectively. Phenotypically, these myeloid-derived cells appeared to be a major component of PD-L1⁺ nontumor cells (Fig. 5), with more CD11c⁺ cells within the tumor-infiltrating immune cells than CD68⁺ cells, suggesting an important role of PD-L1–expressing antigen-presenting cells (APC) in immune microenvironment of FLC.

We compared the PD-1 and IDO cell densities in tumors with PD-L1 expression to those without the PD-L1 expression (Fig. 6A). The expression of PD-1 and IDO was higher in PD-L1–positive tumors. However, they did not meet statistical difference (Fig. 6A). In B7-H3⁺ tumors, statistically higher PD-1⁺ and IDO⁺ cell densities were observed in comparison with B7-H3⁻ tumors (87.6 vs. 15.7 cells/mm² and 35.2 vs. 9.3 cells/mm² for PD-1⁺ and IDO⁺ cells, respectively; Fig. 6B). PD-L1⁺ tumor samples were more likely to also have B7-H3 expression, with 76% of PD-L1⁺ tumor specimens having B7-H3 expression versus 45% in PD-L1⁻ tumor samples (P = 0.033; Fig. 6C).

In this study, we evaluated the expression of multiple immune-checkpoint molecules in FLC tumor samples, which included PD-1, PD-L1, B7-H3, IDO, and LAG3. We found that FLC is a T-cell inflamed tumor, specifically at the interface region, and has expression of PD-1, PD-L1, B7-H3, and IDO on both tumor cells and immune cells in response to cytotoxic immune responses (i.e., by CD8⁺ T cells), suggesting an adaptive immune resistance. This upregulation of immune checkpoints in FLC is distinctive from other pediatric tumors, which usually present with an inherently low mutation burden with negative to poor PD-L1 expression (29–32).

Almost two thirds of the FLC specimens demonstrated membranous PD-L1 expression on tumor cells and PD-L1 expression on TILs and TAMs. PD-L1⁺ TILs/TAMs were also more robust and concentrated compared with PD-L1⁻ tumors, with close to half of all samples encompassing ≥25% PD-L1⁺ TILs and TAMs. Considering that other cancers with PD-L1⁺ tumor-infiltrating immune cells show high response rates to immunotherapy (11, 12), this may suggest an important prognostic feature of FLC. We found a subset of PD-L1⁺ samples with CD11c and CD68 expression, suggesting the possible role of PD-L1⁺ APCs in antitumor immunity and that there are multiple cell types expressing PD-L1 within the TME that create a PD-L1 immunoprotective barrier.

We also found a high proportion of the samples with B7-H3 expression in both the tumor and adjacent nontumor regions. Although it is still unclear whether tumor-associated B7-H3 is considered to directly augment CD8⁺ T-cell effector function or play inhibitory roles in T-cell activation (25), we also found that the samples with B7-H3⁺ tumor cells had much higher CD8⁺ T-cell density compared with those without B7-H3 expression. This is in contrast to our previous study where B7-H3 expression in other solid tumors did not show any association with an adaptive-type response (33). A similar correlation was seen in samples with PD-L1⁺ tumor cells and CD8⁺ T-cell density. We also found that B7-H3⁺ tumors are more likely to present with PD-L1/PD-1 expression, as well as IDO expression and immune cell infiltration.

Lastly, we also demonstrated the importance of the peritumoral interface area in IDO expression, as this area contained the highest concentration of IDO⁺ tumor and immune cells, which correlated with CD8⁺ T-cell density. Similar to the PD-L1 and PD-1 axis (34), IDO is also driven by IFNγ secreted by tumor antigen–specific T cells. It is possible that IDO⁺ tumor cells at the interface between the tumor periphery and the surrounding tissue may form an IDO immune-protective barrier around the tumor to counteract the activated tumor-specific T-cell response. Further investigation is needed to evaluate the immune signatures and identify the critical players that recruit or drive IDO expression at this region of the TME.

The current study had several limitations. Specifically, this was a single-center study that included a small sample size of patients with limited follow-up data. As such, we were unable to correlate the immune-checkpoint expression with survival and, therefore, cannot comment on the prognostic value of these markers. However, considering the rarity of FLC, the current study represents a relatively large cohort of FLC patients. Also, we did not have the multiplex capability in this study to perform multiple color immunofluorescence for coexpression analysis for the immune-specific markers.

There is a long-standing, unmet need for effective systemic therapy in patients with FLC. Data from the study provide further basis for targeting these different immune-checkpoint axes in FLC, including combinational treatments. High frequency of PD-L1, IDO, and B7-H3 expression on TILs and TAMs may play an important role in immunosuppression in the TME, which could be targeted for inhibition. Combination of PD-L1 inhibition with either IDO inhibitors or B7-H3–targeted therapy is already in early clinical trials for other solid cancers (35–37). Although it will be important to interrogate larger tumor cohorts to confirm our findings, the upregulation of multiple immune checkpoints in our cohort provides a basis for investigating immune-checkpoint blockade as a novel therapeutic strategy for patients with FLC.

N.J. Llosa reports receiving a commercial research grant from and is a consultant/advisory board member for Bristol-Myers Squibb. M. Yarchoan reports receiving other commercial research support from Bristol-Myers Squibb, Merck, and Exelixis, Inc., and is a consultant/advisory board member for Exelixis, Inc. E.M. Jaffee reports receiving commercial research grants from AduroBiotech, Bristol-Myers Squibb, Amgen, and Corvus; has ownership interest in Aduro Biotech; and is a consultant/advisory board member for Genocea, CSTONE, DragonFly, and Adaptive Biotech. J.M. Taube reports receiving a commercial research grant from Bristol-Myers Squibb and is a consultant/advisory board member for Bristol-Myers Squibb, AstraZeneca, Amgen, and Merck. R. Anders is a consultant/advisory board member for Bristol-Myers Squibb and Merck. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.K. Kim, F. Gani, A.J. Layman, F. Bhaijee, M. Yarchoan, E.M. Jaffee, H.C. Simons, R.A. Anders

Development of methodology: A.K. Kim, F. Gani, A.J. Layman, Q. Zhu, E.L. Engle, R.A. Anders

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.K. Kim, F. Gani, A.J. Layman, E.L. Engle, F. Bhaijee, M.B. Goggins, T.M. Pawlik, R.A. Anders

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.K. Kim, F. Gani, S. Besharati, Q. Zhu, F. Succaria, T.M. Pawlik, E.M. Jaffee, J.M. Taube, R.A. Anders

Writing, review, and/or revision of the manuscript: A.K. Kim, F. Gani, A.J. Layman, Q. Zhu, F. Bhaijee, N.J. Llosa, T.M. Pawlik, M. Yarchoan, E.M. Jaffee, H.C. Simons, J.M. Taube, R.A. Anders

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.K. Kim, S. Besharati, R.A. Anders

Study supervision: R.A. Anders

This study was supported by Cancer Research Institute/Fibrolamellar Cancer Foundation Grant.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.