T cells and macrophages infiltrate the FLC tumor microenvironment
In order to understand the immunosuppressive effects of the FLC TME, we first used mIHC to examine the prevalence and distribution of immune cells in the FLC TME (Table 1, patients 5-10; Fig. 1A). In the tumor compartment, we found that carcinoma cells comprised approximately 50% of all cells, with CD8+ T cells representing 1.5%, CD4+ T cells 5.6%, and macrophages 3.9% (Fig 1B). Overall, we observed a CD4+:CD8+ ratio of 3.7:1.0, with regulatory T cells (Tregs; CD4+FOXP3+) accounting for 8.2% of CD4+ cells.
To quantify the spatial distribution of these immune cell types, we divided the tumor compartment into NTL, tumor interface, and intra-tumoral regions. We observed significantly more CD8+ T cells per mm2 in the tumor-NTL interface compartment compared to the tumor itself (p=0.03), and a trend toward increased CD8+ T cell density in NTL compared to the tumor (Fig. 1C). CD4+ T cell densitywas significantly higher in the NTL and interface compartments than in the tumor (p<0.001 and p=0.005, respectively). No significant difference was found in Treg or macrophage density in the tumor compared to the NTL or interface (Fig. 1C). However, the ratio of CD8+ T cells to Tregs was significantly lower in the tumor compared to NTL (p=0.01) and trended towards being lower in tumor compared to the interface (p=0.09, Fig. 1D). Overall, these findings suggest that T cells in the FLC TME are physically excluded from the tumor compartment and may be subject to suppressive signaling.
We next examined the spatial distribution of immune cells within FLC tumors histologically, dividing the intra-tumoral region into nests of carcinoma cells intermixed with fibrous stromal bands. Similar mIHC examination demonstrated that the stromal compartment had significantly higher densities of CD8+ T cells (p=0.03), CD4+ T cells (p=0.006), and Tregs (p=0.02). Macrophage density was similar between compartments (Fig 1E). These data suggest that T cell exclusion from the carcinoma cells in FLC is accomplished by sequestration of tumor infiltrating lymphocytes in the fibrous stromal bands away from tumor cells.
Lymphocyte populations within FLC have suppressed cytotoxic gene expression
Having defined the spatial distribution of immune cells in FLC, we used gene expression profiling of FLC and paired NTL to infer subtype and activation states of TILs (patients 1-4 in Table 1). We compared these data to those generated from classical HCCs and matched NTL. Using NanoString proprietary scores based on RNA expression, we found no significant difference in CD8+ T cell number between FLC and paired NTL, nor between FLC and HCC (Fig 1F) [37]. There were significantly lower T cell function, cytokine, and chemokine scores in FLC compared with NTL (Fig. 1G), which suggests less cytotoxic activity. Similar trends were observed when evaluating individual gene expression for cell lineage markers as well as pro-inflammatory cytokines such as interferon gamma (IFN-γ) and tumor necrosis factor (TNF) (Suppl. Fig. S1A, S2E). These data suggest that T cells in FLC are either naïve or functionally suppressed.
Surprisingly, few genes in this dataset were upregulated in FLC (Suppl Fig. S1F). Arginase 2 (ARG2), a mitochondrial enzyme involved in the urea cycle, was induced in FLC compared to paired NTL (p=0.007) and HCC (p=0.02). Decreased L-arginine levels have been shown to impair CD8+ T cell cytotoxic function [38]. Both IL32 and IL34 were more highly expressed in FLC than paired NTL (p=0.02 and p=0.006, respectively), and IL34 was also more highly expressed in FLC compared to HCC (p=0.005). While less is known about IL-34, except that it can increase monocyte activity, IL-32 has been implicated in fibrosis and cancers associated with chronic inflammation, including HCC [39,40]. Platelet-derived growth factor receptor β (PDGFRB) gene expression was significantly higher in FLC compared to paired NTL (p=0.01) and HCC (p=0.04). In the liver, the PDGFRB is highly expressed by fibroblasts and smooth muscle cells, and its expression increases with inflammation and fibrogenesis [41]. Similar inductions of ARG2, IL34, and PDGFRB expression were not seen in HCC compared to paired NTL, suggesting this pattern is unique to FLC, although exact mechanisms are unclear.
Immune profiles in the tumor microenvironment of FLC
To further understand the constituent immune cells in the FLC TME, we catalogued the immune landscape in six FLC samples and four paired NTL with flow cytometry using myeloid, innate(-like) lymphocyte and T cell-centered panels. The myeloid-centered panel distinguishes all canonical dendritic cell (DC) subsets, with parallel enumeration of monocytes/macrophages, T cells, B cells and natural killer (NK) cells [42, 43]. UMAP analysis of the entire live immune compartment stained with the myeloid-centered panel revealed the presence of diverse immune cell populations, which were separated clearly (Fig. 2A) Neutrophils and T cells were found to be the most common populations of the tumor-infiltrating immune cells, both in FLC and NTL (Fig. 2A).
Flow cytometric analysis of FLC showed fewer CD45+ immune cells compared to matched NTL (Fig. 2B). Generally, we found large differences in both the variety and composition of myeloid immune cells (Fig. 2C), innate(-like) lymphocytes (Fig. 2D), and lymphocytes (Fig. 2E) across patients (gating strategies outlined in Suppl. Fig. S2). Regarding myeloid cell population abundance, a trend towards increased monocyte/macrophage frequency was observed in FLC compared to NTL (Fig. 2C). Conventional type 1 DCs ( cDC1s), which have a known role in initiating T cell responses to cancer, were found to be significantly lower in number in FLC compared to NTL, while no differences were observed in the frequencies of neutrophils, plasmatocytoid DCs (pDCs) and conventional type 2 DCs (cDC2s, Fig. 2C) [44-46]. Regarding innate-like lymphocytes, we observed a significant reduction in the proportion of NK cells and a slight decrease in mucosal-associated invariant T (MAIT) cells in FLC compared to NTL (Fig. 2D); however, there was no difference in the abundance of other innate cell populations such as NKT cells, innate lymphoid cells type 1 (ILC1s), ILC2s and ILC3s observed between FLC and NTL (Fig. 2D, Suppl. Fig S3) [47, 48].
In FLC samples, a significantly higher frequency of CD4+ T cells among CD45+ immune cells was observed compared to NTL (Fig. 2E). This finding was consistent with our mIHC analysis in FLC, in which a higher absolute number of CD4+ T cells per mm2 was observed compared to other immune cell types. CD4+ Tregs (p=0.011) were found specifically to be higher in FLC, while there was only a trend towards increased conventional CD4+ T cells (p=0.0896) in FLC (Fig. 2E). No differences were observed in the proportions of CD8+ T cells, gamma-delta (γδ) T cells and B cells between FLC and NTL (Fig. 2E). The increase in CD4+ T cells in FLC resulted in a significant decrease in the CD8+ to CD4+ T cell ratio in FLC compared to NTL (Fig. 2F). Furthermore, we found a decreased ratio of CD8+ T cells to Tregs in FLC compared to NTL (Fig. 2F). The increased prevalence of CD4+ T cells over CD8+ T cells and decreased CD8+ T cell to Treg ratio in FLC compared to NTL is consistent with our spatial mIHC analysis (Fig. 1B). Taken together, these results suggest an immune excluded and immunosuppressive microenvironment in FLC, with a TME that is biased towards CD4+ T cells, and contains fewer cDC1s and NK cells.
The T cell infiltrate in FLC is skewed toward an effector memory phenotype that expresses PD-1 and CTLA-4
We next characterized the phenotype of tumor-infiltrating CD4+ and CD8+ T cells using flow cytometry. We included C-C chemokine receptor type 7 (CCR7) and CD45RA to define naïve (Tnaïve), central memory (Tcm), effector memory (Tem) and terminally differentiated T effector (Teff) cells in FLC and NTL samples (Suppl. Fig. S2). Both the CD4+ and CD8+ T cells were predominantly Tem in FLC, and this proportion was increased in FLC compared to NTL for both cell types, suggesting that the majority of intratumoral T cells are antigen-experienced (Fig. 3A, 3B). The proportion of CD4+ and CD8+ Teff were decreased in FLC compared to NTL, while the proportion of Tcm and Tnaïve cells were unchanged.
Little information exists on the expression pattern of targetable immune checkpoint molecules on T cells in FLC [20]. Therefore, we analyzed the expression of PD-1 and cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) on T cells in FLC and NTL. We observed heterogenous expression of PD-1 and CTLA-4 on conventional CD4+ T cells, CD4+ Tregs and CD8+ T cells in FLC as well as NTL (Fig. 3C). CD4+ Tregs infiltrating FLC expressed significantly more PD-1 and CTLA-4 than Tregs present in NTL (Fig. 3C). In patients 12 and 14, we observed an increase in conventional CD4+ T cells and CD8+ T cells expressing PD-1 in FLC compared to NTL, while for patients 8 and 13 the PD-1 expression was comparable between FLC and NTL (Fig. 3C). In three of four patients, an increase in conventional CD4+ T cells expressing CTLA-4 was observed, while CTLA-4 was rarely detected on CD8+ T cells, except for CD8+ T cells infiltrating the FLC of patient 14.
Together, these data demonstrate that the T cell infiltrate in FLC consists mainly of antigen-experienced effector memory CD8+ and CD4+ T cells, and that naïve T cells are largely absent. Furthermore, we found expression of immune checkpoint molecules in FLC, suggesting T cell exhaustion and the development of regulatory mechanisms that play a role in suppressing local antitumor immune function.
FLC tumor infiltrating TCRs are relatively unexpanded yet highly conserved
Baseline intratumoral T cell receptor (TCR) clonality has previously been shown to positively correlate with prognosis and immunotherapy response across a variety of tumor types [49, 50]. The DNAJ-PKAc fusion protein represents a potential foreign antigen and target for either TCR selection of different rearrangement events. However, the various observed immunosuppressive features within the FLC TME could limit clonal expansion of various TCRs, including those recognizing tumor neoantigens. To better understand the function of TILs in FLC patients, we first performed high-throughput sequencing of the TCRß chain from four FLC primary tumors with paired NTL samples (Table 1, patients 2, 8-10) [26, 51]. Identifying between 5,866 and 144,221 productive rearrangements per sample, we defined a TCR clonotype as the β chain complementarity-determining region 3 (CDR3β) amino acid sequence. We found that while clone sizes for both FLC and paired NTL followed a characteristic power-law distribution, TCRs in NTL samples appeared to have more frequently undergone clonal expansion than those clonotypes within the tumors (Fig 4A).
To further quantify repertoire dynamics within FLC, we calculated the clonality of each sample as one plus the normalized Shannon entropy. In brief, clonality ranges between zero and one with a decreasing clonality score corresponding to increasing repertoire diversity (e.g., a clonality of one represents a repertoire composed of only a single clone and thereby minimum diversity). We found that the clonality of NTL samples was higher than their paired primary tumors in three out of four patients (Fig 4B), though the observed increase in NTL repertoire clonality was not statistically significant with this limited sample size (p=0.37). Furthermore, there was a strong inverse correlation between the clonality of FLC samples and the densities of CD8+, CD4+, and Tregs in the carcinoma compartment based on mIHC (Suppl. Fig S4A). This finding clearly demonstrates that large clonal expansions do note explain the relatively high TIL frequencies in certain FLCs. Together, these findings are suggestive of an immunosuppressive TME that inhibits clonal expansion of TILs in FLC.
FLC is unique in that it arises in healthy liver parenchyma due to expression of the DNAJ-PKAc fusion protein [2]. We therefore theorized that this unique fusion protein could represent an immunogenic neoantigen shared across individuals, which could in turn generate shared TCR responses despite the lack of large clonal expansions. To test this hypothesis, we calculated Morisita’s overlap index (MOI) to evaluate the pairwise proportion of rearrangements coding for the same CDR3β amino acid sequence in FLC. Here, a value of 0 indicates no shared sequences, while a value of 1 indicates that all sequences are shared between two samples. Calculating the MOI for all pairwise combinations of either NTL or FLC tumor samples, we indeed found that the tumor samples had higher rates of TCR sharing between individuals (Fig 4C). Similar to previous findings in healthy peripheral blood samples, these shared TCR sequences typically had higher generation probabilities and were more frequently known to recognize common viral epitopes relative to sequences unique to a single individual (Suppl Fig S4B) [52-54]. However, as these known TCR sequences are present in both FLC tumor and NTL samples, these observations do not explain the increased conservation of TCR sequences among FLC tumors, and none have been directly linked to the DNAJ-PKAc fusion protein.
To further delineate the importance of conserved TCR sequences across FLC, we downloaded TCRβ sequencing data from six other solid tumor types in the immuneAccess database [27-33]. Our seven FLC primary tumor samples (including three additional samples without paired NTLs) had a lower average clonality but significantly higher MOI than melanoma (ML), pancreatic ductal adenocarcinoma (PDA), colorectal adenocarcinoma (CRC), breast carcinoma (BRCA), and ovarian carcinoma (OV) (Fig 4D-E). Merkel cell carcinomas (MC), the majority of which were positive for the common Merkel cell polyomavirus T-antigen, displayed a similar pattern of relatively low clonality but high repertoire overlap across individuals. Together, these results highlight a high degree of TCR sequence conservation across FLC tumors. These findings could be consistent with an enrichment of FLC TILs with shared TCRs recognizing common neoantigens despite the immunosuppressive TME limiting the expansion and in vivo efficacy of these tumor-specific clonotypes. We thus hypothesized that reversing the suppressive TME in FLC could enable these putatively neoantigen specific TILs to effectively kill tumor cells.
Blockade of PD-1 and IL-10 can reactivate TILs to kill FLC cells in human TSC
Using ex vivo TSC from five patients, we tested the hypothesis that immunotherapy has the potential to rescue endogenous anti-tumor immune responses in FLC (Table 1, Patient 10-14) [34, 35]. To augment the effectiveness of ICI in the immunosuppressive nature of the FLC TME, we decided to target macrophage pathways due to the trend towards increased in monocyte/ macrophage presence in FLC compared to NTL. Due to our prior identification of CD8+ T cell-mediated tumor cell death after IL-10 blockade in colorectal liver metastases, we evaluated PD-1 and IL-10 blockade alone or in combination in tumor slices to target T cell and macrophage-mediated suppression, respectively [55]. FLC slice cultures were treated with IgG control mAb, anti-PD-1 mAb, anti-IL-10 mAb, or anti-PD-1 plus anti-IL-10 mAbs for 6 days and tumor cell apoptosis was measured using cleaved caspase 3 (cC3) IHC (Fig. 5A). The average percentage of tumor cells expressing cC3 was 32.2% (SD 16.3) for IgG, 43.5% (SD 9.3) for anti-PD-1, 49.1% (SD 12.5) for anti-IL-10, and 56.9% (SD 14.2) for anti-PD-1 and anti-IL-10 combination treatment. Notably, the combination of anti-PD-1 and anti-IL-10 induced significantly more tumor cell apoptosis than IgG control (p=0.03), while the difference between either of the individual treatment groups compared to IgG control (p=0.30 for anti-PD-1 and p=0.09 for anti-IL-10) did not reach statistical significance (Fig 5B). Generally, more tumor cell apoptosis was observed when FLC slice cultures were treated with anti-IL-10 mAb alone compared to anti-PD1 mAb alone, but TSCs from patient 11 were an outlier in that more cC3 positivity was observed in response to PD-1 blockade alone compared to IL-10 blockade alone (Suppl Fig S5). Interestingly, that patient also had the lowest percentage Tregs. Conversely TSCs from patient 13, who had the highest percentage of monocytes/ macrophages in FLC by flow cytometry, had a large increase in apoptosis for anti-IL-10 alone compared to IgG control. Together, these results indicate TILs present in FLC can be reactivated to kill tumor cells following PD-1 and IL-10 blockade.