Abstract
Background: Checkpoint inhibitors act on exhausted CD8+ T cells and restore their effector function in chronic infections and cancer. The underlying mechanisms of action appear to differ between different types of cancer and are not yet fully understood. Methods: Here, we established a new orthotopic HCC model to study the effects of checkpoint blockade on exhausted CD8+ tumor-infiltrating lymphocytes (TILs). The tumors expressed endogenous levels of HA, which allowed the study of tumor-specific T cells. Results: The induced tumors developed an immune-resistant TME in which few T cells were found. The few recovered CD8+ TILs were mostly terminally exhausted and expressed high levels of PD-1. PD-1/CTLA-4 blockade resulted in a strong increase in the number of CD8+ TILs expressing intermediate amounts of PD-1, also called progenitor-exhausted CD8+ TILs, while terminally exhausted CD8+ TILs were almost absent in the tumors of treated mice. Although transferred naïve tumor-specific T cells did not expand in the tumors of untreated mice, they expanded strongly after treatment and generated progenitor-exhausted but not terminally exhausted CD8+ TILs. Unexpectedly, progenitor-exhausted CD8+ TILs mediated the antitumor response after treatment with minimal changes in their transcriptional profile. Conclusion: In our model, few doses of checkpoint inhibitors during the priming of transferred CD8+ tumor-specific T cells were sufficient to induce tumor remission. Therefore, PD-1/CTLA-4 blockade has an ameliorative effect on the expansion of recently primed CD8+ T cells while preventing their development into terminally exhausted CD8+ TILs in the TME. This finding could have important implications for future T-cell therapies.
Introduction
Liver cancer is the sixth most frequently diagnosed cancer and the fourth most common cause of cancer-related death worldwide. The prognosis for patients with liver cancer remains poor as therapeutic options are limited [1]. Inhibition of receptors that negatively regulate T-cell activation, such as cytotoxic T lymphocyte antigen 4 (CTLA-4) and programmed cell death 1 (PD-1), with monoclonal antibodies (checkpoint blockade) is an efficient alternative for the treatment of tumors that are resistant to traditional therapies. CTLA-4 and PD-1 are normally involved in the maintenance of immune homeostasis, as they regulate inflammatory immune responses to pathogens and tolerance to self-antigens antigens [2]. During chronic infections and cancer, the expression of these receptors is associated with T-cell exhaustion, a state of dysfunctionality defined by poor effector function, low proliferative capacity, and survival after antigen stimulation [2‒5].
Checkpoint blockade is thought to reinvigorate exhausted CD8+ T cells recovering their effector function. Although the mechanisms underlying this process are still not completely understood, previous studies have shown that CD8+ T-cell subsets with varying degrees of exhaustion respond differentially to PD-1 blockade. During chronic infection with murine lymphocytic choriomeningitis virus (LCMV), terminally exhausted CD8+ T cells expressing high levels of PD-1 (CD8+PD-1high) were refractory to blockade of the PD-1 pathway, while progenitor-exhausted CD8+ T cells expressing intermediate amounts of PD-1 (CD8+PD-1int) were responsive [3]. CD8+PD-1int tumor-infiltrating lymphocytes (TILs) in mouse melanoma were also responsive to PD-1 or PD-1/CTLA-4 blockade. Additionally, the duration of response to PD-1/CTLA-4 blockade in patients with melanoma correlated with the frequency of CD8+PD-1int TILs detected in biopsies collected prior to treatment, indicating that this subset responded to the treatment [6]. In contrast, in patients with non-small-cell lung carcinoma, the presence of CD8+PD-1high but not of CD8+PD-1int TILs in biopsies prior to anti-PD-1 therapy predicted response, indicating that CD8+PD-1high TILs may also respond to anti-PD-1 treatment [7]. Therefore, analysis of the effect of checkpoint blockade on CD8+ TIL subsets in different cancers may help to identify predictors of response and improve therapeutic protocols.
Here, we established an orthotopic liver cancer model to study the effect of PD-1/CTLA-4 blockade on exhausted CD8+ TILs. Treatment induced the expansion of CD8+PD-1int TILs but not of CD8+PD1high TILs. Similar expansion was also observed after adoptive transfer of naïve CD8+PD1neg tumor-specific T cells and PD-1/CTLA-4 blockade. Tumor control correlated with the migration of the transferred cells in the tumor microenvironment (TME), expansion, and infiltration of CD8+PD-1int TILs into the tumor. Interestingly, no changes were observed in the transcriptional profile of PD-1int TILs due to the therapy. In conclusion, PD-1/CTLA-4 blockade improves the expansion of recently primed CD8+ T cells, promotes tumor infiltration, and prevents the development of primed cells into CD8+PD-1high TILs in the TME. Additionally, our data demonstrate that PD-1/CTLA-4 blockade can significantly improve the effectiveness of adoptive T-cell therapies (ACTs). These findings are relevant for future clinical interventions.
Materials and Methods
Mice
Tumors were induced in 6- to 8-week-old BALB/cJHanZtm mice. HA-specific T cells were obtained from congenic CD45.1+ BALB-Ptprca-Tg(TCR/C-TCR(HA/PR8/34))CL4/Rlf mice [8]. Albumin-cre x Rosa-26HA F1 mice were utilized for the generation of liver tumor cells. All animals were bred and maintained under pathogen-free conditions at the animal facility of Hannover Medical School, Germany.
Generation of HA-Expressing Murine HCC Cell Lines
Liver tumor cells expressing the influenza hemagglutinin HA peptide were generated by inducing autochthonous liver tumors in Albumin-cre x Rosa-26HA F1 mice [8] by hydrodynamic cotransfection [9] of plasmids carrying oncogenes and Sleeping Beauty transposase (pPGK SB13 Kras g12V and pT/p53-246 sh p53). Under these conditions, HA is expressed at endogenous levels (see online suppl. Methods; for all online suppl. material, see www.karger.com/doi/10.1159/000526899). Tumor cell lines were obtained either after homogenization of tumor-bearing tissues or by overgrowth of tumor cells from a small tumor piece. The obtained cell lines were cultured in DMEM medium containing 10% FCS and penicillin/streptomycin. Clonal tumor cell lines were obtained by a limiting dilution assay, and HA-expressing clones were selected based on their capacity to stimulate HA-specific CD8+ T cells in vitro. The clone Hep-HA-STK-1 was utilized for the experiments.
In vitro Restimulation Assay
Tumor-isolated lymphocytes (1 × 106) were stimulated with irradiated (30 Gy) Hep-HA-STK-1 cells (5 × 104 cells/well) in the presence of suboptimal concentrations of anti-CD3 (0.5 μg/mL) and anti-CD28 antibodies (2 μg/mL) as described by Thommen et al. [7]. To measure degranulation, stimulation was performed in the presence of an anti-CD107a-BV421 antibody (0.25 μg/106 cells) and brefeldin A for 6 h as previously described [10]. Unstimulated cells were incubated for the same time in the presence of anti-CD107a-BV421 antibody and brefeldin A only. After incubation, cells were recovered and stained for CD8 and PD-1. They were then fixed, stained for intracellular TNF-α and IFN-γ, and analyzed by flow cytometry.
Adoptive Cell Transfer
CD8+ T cells expressing an HA-specific TCR were isolated from the spleens of CD45.1+ BALB-Ptprca-Tg(TCR/C-TCR(HA/PR8/34))CL4/Rlf transgenic mice and transferred on day 10 after tumor induction. Isolation was performed using the MojoSort Mouse CD8 T-Cell Isolation Kit (BioLegend) according to manufacturer’s instructions. For transfer, the isolated cells (1 × 106/100 μL in 0.9% NaCl solution) were injected into the tail vein of the mice. The cell purity was controlled by flow cytometry prior to transfer.
Tumor Induction and in vivo Treatments
Hep-HA-STK-1 cells (5 × 105 in 10 μL PBS) were injected into the left liver lobes of BALB/c mice [11]. Mock-operated mice received a 10 μL injection of PBS. Eleven days after tumor induction, tumor-bearing mice were randomized and treated either with 100 μg of α-PD-1 plus 100 μg of α-CTLA-4 antibodies or the same concentrations of the corresponding isotype controls (rat IgG2a plus polyclonal Syrian hamster IgG) by intraperitoneal injections at days 11, 13, and 17. Mock-treated mice received isotype control antibodies. For the histological analysis, untreated mice received PBS. After killing, the length and width of the tumors were measured with a digital caliper. The calculation of the tumor area was performed according to the following formula: area = a/2 × b/2 × π, where a = length (largest tumor diameter) and b = width (perpendicular tumor diameter).
Immunofluorescence and H&E Staining
The liver was perfused with PBS and fixed in 2% formaldehyde. For immunofluorescence, the tissues were equilibrated in 30% sucrose, embedded in OCT medium and frozen. Immunofluorescence microscopy was performed as described before [12, 13]. Briefly, cryosections were rehydrated, blocked, and stained. For sequential staining, a further blocking step was performed between the stains. For the H&E staining, the tissues were embedded in paraffin and sectioned and stained as described before [14, 15]. All images were acquired using an AxioImagerM1 microscope with corresponding Zen2.3 Imaging software (Carl Zeiss). The quantification of the data was done by counting 5 microscopic fields (×20 magnification) from 5 sections per mice and time point.
TCR – Sequencing
RNA isolation was performed with the RNeasy Plus Micro Kit (Qiagen). After isolation, RNAs were converted into cDNAs using the SuperScript III Reverse Transcriptase Kit (Thermo Fisher Scientific). Amplicon libraries of rearranged TRAV12 (Vα8 family) CDR3 regions were generated pars pro toto as previously described [16‒18]. Amplicons were purified after agarose gel electrophoresis using the QIAquick Gel Extraction Kit (Qiagen) and subjected to an index PCR for MiSeq analysis (Nextera Index Kit, Illumina). Index PCR amplicons were purified using the AMPure XP PCR purification kit (Beckman Coulter). The Quant-iTTM PicoGreenTM dsDNA Assay Kit (Thermo Fisher Scientific) was used for DNA quantification. PCR amplicons were processed according to Illumina’s “Denature and Dilute Library guidelines” to generate a 4 nM library. Sequencing was performed according to Illumina MiSeq analysis. FASTQ files and a corresponding quality report (FASTQC tool, Babraham Institute) were generated. The sequences were annotated using IMGT/HighV-Quest [19]. TCR clones were identified based on identic CDR3 sequences, and clone frequencies were calculated using in-house bash and R scripts (https://figshare.com/s/c7f85bcbca2be425bd25). The Shannon indices (HS) were calculated using the following formula:
RNA – Sequencing and Analysis
RNA was isolated as above described. The quality and integrity of the RNA were controlled with a 2,100 Bioanalyzer (Agilent Technologies). Only RNA samples with an RIN ≥8 were used. The samples were purified with the Dynabeads mRNA DIRECT Micro Kit (Thermo Fisher Scientific), and a library was prepared using the NEBNext Ultra II Directional RNA Library Prep Kit (New England BioLabs). Sequencing was performed on a NovaSeq 600 using a NovaSeq 6000 S1 Reagent Kit (Illumina). FASTQ files and a corresponding quality report (Babraham Institute) were generated. The quality and adapter trimming of the FASTQ files were performed using Trim Galore! Wrapper (Babraham Institute), and reads <20 bp were removed. The trimmed sequences were aligned to the reference genome using the STAR tool [20]. Feature counts were performed using the R package Rsubread [21]. Only genes that occurred at least five times in two of the samples were selected. Gene annotation and log2 transformation as well as normalization beyond the 50th percentile were performed by using the R packages bioMaRT [22, 23] and the quartile normalization of edgeR [24], respectively. Differential gene expression was depicted using Qlucore Omics Explorer v3.5 (Qlucore). For gene ontology (GO) analyses, the Generic GO Term Finder was used [25].
Flow Cytometry
Lymphocytes were isolated from the tumors and liver lobes as described in the online supplementary methods to obtain a cell suspension. Cells were pelleted, resuspended (106 cells/100 μL) in PBE containing 20% rat serum and 4% FcR blocking reagent (Miltenyi Biotec), and incubated on ice for 20 min. After blocking nonspecific binding sites, cells were stained with the appropriate concentration of labeled antibodies for 20 min. After staining, cells were washed with PBE, then incubated with DAPI for 1 min, washed again, resuspended in PBE, and measured in a flow cytometer. For intracellular staining, cells were fixed with Fixation Buffer after surface staining and permeabilized with Permeabilization Wash Buffer (Biolegend) prior to staining. Dead cells were stained with the Fixable Viability Dye eFluorTM 780 (Thermo Fisher Scientific) prior to fixation. For sorting, lymphocytes isolated from 5 to 10 individual tumor-bearing liver lobes were pooled, washed, and resuspended in PBE at a concentration of 106 cells/100 μL. Blocking of nonspecific binding sites and staining were performed as above described above. The analysis was done in an LSRII (BD). Sorting was performed using a FACSAria Fusion, a FACSAria Ilu (both from BD), or a MoFloXDP sorter (Beckman Coulter). Analysis was performed using FACSDiva software (BD).
Statistics
Analysis was performed using GraphPad Prism versions 5.01 and 9 (GraphPad Software, La Jolla, CA, USA). All statistical analyses and p values are described in the figure legends. p < 0.05 was considered significant. For a box-whisker plot, the box represents the interquartile range value, the line in the box represents the median, and the whisker represents 1.5 × interquartile range values (Tukey).
Results
New Liver Cancer Model and Effect of PD-1/CTLA-4 Blockade on Tumor Growth
Liver tumor cells expressing endogenous levels of an influenza hemagglutinin HA-peptide (Hep-HA-STK-1 cells) were gained from autochthonous liver tumors induced in Albumin-cre x Rosa-26HA F1 mice. To induce localized orthotopic tumors, Hep-HA-STK-1 cells were injected into the liver parenchyma (Fig. 1a). Monitoring 11 days after tumor induction showed the presence of visible tumors in the liver (Fig. 1b). In addition, an accumulation of exhausted CD8+PD-1high TILs was observed in tumor-bearing liver lobes (Fig. 1c). Histological analysis demonstrated that the tumors expressed PD-L1 and Glypcan-3 (online suppl. Fig. 1a). Unless otherwise mentioned, therapy response was assessed on day 21 in mice that received either PD-1/CTLA-4 blockade or isotype control antibodies on days 11, 13, and 17 (Fig. 1d). In some experiments, mice received 1 × 106 CD8+CD45.1+ HA-specific T cells intravenously on day 10 after tumor induction. Despite high intragroup variability, the tumor size was significantly reduced in mice receiving PD-1/CTLA-4 blockade (Fig. 1e, f). Adoptive T-cell transfer (ACT) without checkpoint blockade had no significant effect on tumor growth. In contrast, PD-1/CTLA-4 blockade caused a comparable reduction of tumor mass independent of the transfer of tumor-specific CD8+ T cells (Fig. 1f).
Histological analysis revealed that only few CD4+ and CD8+ T cells were found in the tumors of untreated mice. These were predominantly located at the invasive margin and in the outer areas of the tumors, indicating that T cells did not effectively infiltrate the tumor. Impairment of T-cell infiltration, also called T-cell exclusion, is an immune evasion mechanism frequently observed in solid tumors [26] and has been previously observed in experimental HCC [27]. In contrast, residual tumor areas present in treated mice were heavily infiltrated with T cells (Fig. 1g, h). Tertiary lymphoid organs were observed in both groups but were much more frequent in treated animals (online suppl. Fig. 1b). Therefore, PD-1/CTLA-4 blockade induced a strong antitumor response.
PD-1/CTLA-4 Blockade Results in Activation and Proliferation of T Cells
To characterize T cells infiltrating the TME after PD-1/CTLA-4 blockade, we analyzed tumor sections from untreated and treated mice killed at different times after tumor induction and treatment. While effector CD8+ and CD4+Foxp3− T cells were found almost exclusively in the tumors of treated mice, CD4+Foxp3+ regulatory T cells (Tregs) were present in the tumors of both mouse groups (Fig. 2a). The number of both CD8+ and CD4+ T cells significantly increased over time in treated animals. These increases occurred at comparable rates so that no significant changes in the CD8+/CD4+ T-cell ratio were observed between untreated and treated animals (Fig. 2b, upper panels). While the number of CD4+Foxp3− effector T cells significantly increased after therapy, the numbers of CD4+Foxp3+ Tregs remained constant. This led to a significant increase in the CD4+Foxp3−/CD4+Foxp3+ T-cell ratio in treated animals (Fig. 2b, lower panels). Consequently, tumor remission was not due to depletion of Tregs by the anti-CTLA-4 antibody as described by others [28], but the expansion of effector T cells shifted the ratio of effectors to regulators in favor of a proinflammatory response.
The few CD8+ TILs found in the tumors of untreated mice were CD44− and PD-1+. In contrast, the majority of the CD8+ TILs in the tumors of the treated mice were CD44+ and PD-1− (Fig. 2c). PD-1/CTLA-4 blockade induced a stable increase in the numbers and proportions of these cells (Fig. 2d), indicating that treatment led to priming and activation of naïve CD8+ T cells in the TME. In murine LCMV infection and human HCC, CD8+PD-1int cells express large amounts of CD44. In contrast, terminally exhausted CD8+PD-1high cells express low levels of CD44 [3, 29]. Given the detection limit of the immunofluorescence, it is likely that only cells expressing high levels of PD-1 and CD44 were recognized as positive. Therefore, CD8+ T cells identified as PD-1+CD44− most likely correspond to CD8+PD-1highCD44low TILs, whereas cells identified as PD-1− correspond to CD8+PD-1negCD44+ and CD8+PD-1intCD44+ TILs.
Treatment also led to a transient increase in the expression of the proliferation marker Ki67, which was observed on day 14 but not on day 21 (Fig. 2e, f). Co-staining of Ki67 and PD-1 revealed that both CD8+PD-1− and CD8+PD-1+ TILs proliferated in response to the treatment, although CD8+PD-1+ TILs showed a high proliferative rate (Fig. 2f). Since immunofluorescence staining did not discriminate between the CD8+PD-1 TIL subsets, we analyzed the coexpression of CD8, Ki67, and PD-1 by flow cytometry. Both CD8+PD-1high and CD8+PD1int TILs expressed Ki67. However, in agreement with previously published data [6, 7], the majority of the CD8+Ki67+ TILs were PD-1high (Fig. 2g, upper panel). In addition, the percentages of Ki67+ cells belonging to the different PD-1 subsets were similar despite treatment (Fig. 2g, lower panel). Although a high percentage of CD8+PD1high TILs expressing Ki67 was detected in the tumor of treated mice on day 14, the absolute number of CD8+PD-1+ TILs detected by immunofluorescence remained constant over time (Fig. 2d, lower left panel). This is surprising since cells expressing high levels of PD-1 are recognized as PD-1+ by immunofluorescence. This result indicated that CD8+PD-1high-proliferating cells either left the tumor or died after stimulation. The latter possibility is consistent with the observation that CD8+PD-1high TILs in experimental melanoma proliferate more efficiently than CD8+PD-1int TILs but undergo a limited number of division cycles and have a high death rate after TCR stimulation [6].
PD-1/CTLA4 Blockade Results in the Expansion of a Few Tumor-Specific T-Cell Clones
Blockade of both CTLA-4 and PD-1 pathways leads to expansion of the CD8+ T-cell pool. While PD-1 blockade leads to an increase in the clonality of the TIL repertoire, CTLA-4 blockade broadens the entire peripheral TCR repertoire [30‒32]. Broadening of the peripheral TCR repertoire results from the polyclonal expansion of nontumor-specific TCR clones and correlates more with the toxicity of the therapy than with the antitumor response [30, 31]. As demonstrated in Figure 3, the frequencies of the most common clonotypes were much higher in the tumors than in the spleens regardless of treatment, demonstrating that PD-1/CTLA-4 blockade did not expand the peripheral TCR repertoire.
T cells sorted from the tumors 21 days after induction were highly clonal, with four clones accounting for 60% of the CD8+ TIL repertoire of untreated mice. The treatment led to further clonal expansion, as 60% of the CD8+ TIL repertoire of treated mice corresponded to a single clone (Fig. 3a, b). Although the TCR clonality of the CD8+ TILs clearly increased after treatment, the differences in the calculated Shannon indices were not significant between untreated and treated mice (p = 0.11, Fig. 3c). Nevertheless, the distribution of the cumulative frequencies of the 500 most frequent clones indicated a rather clonal TCR repertoire after treatment (Fig. 3d). In conclusion, the treatment increased the clonality of the CD8+ TIL repertoire without leading to a broadening of the peripheral CD8+ T-cell repertoire.
PD-1/CTLA-4 Blockade Supports Priming of Naïve CD8+ T Cells in the TME and Induces the Local Expansion of CD8+PD-1int TILs
To clarify which CD8+ TIL subsets expanded in the tumors after therapy, we analyzed the phenotype of the CD8+ TILs isolated from untreated and treated mice on day 21 after tumor induction by flow cytometry. The fate of naive tumor-specific T cells (CD8+CD45.1+ HA-specific T cells) adoptively transferred 10 days after tumor induction was also investigated.
As expected, the treatment resulted in an expansion of CD8+ TILs. Similar results were observed in the rest of the liver but not in the tumor-draining lymph nodes or the spleen (Fig. 4a, upper panels), confirming the localized activation of T cells. Transferred naïve CD8+CD45.1+ HA-specific T cells behaved similar to endogenous CD8+CD45.2+ T cells (Fig. 4a, lower panels).
Flow cytometric analysis confirmed that PD-1/CTLA-4 blockade led to sustained expansion of CD8+PD-1int but not of CD8+PD-1high TILs (Fig. 4b, c, upper panel). In addition, a significant expansion of transferred CD45.1+CD8+PD-1neg TILs was also observed (Fig. 4c, lower panel). In general, transferred CD8+PD-1neg and CD8+PD-1int TILs expanded more efficiently than their endogenous counterparts after treatment, as the number of CD45.1+CD8+PD-1neg and CD45.1+CD8+PD-1int TILs increased more than 200-fold compared to mock-operated mice. In contrast, endogenous CD45.2+CD8+PD-1neg TILs did not expand, and the number of CD45.2+CD8+PD-1int TILs was only 7-fold higher in comparison to mock-operated mice. In summary, the therapy induced the preferential expansion of tumor-specific T-cell clones. Importantly, treatment enabled the priming of transferred naïve tumor-specific CD8+PD-1neg T cells, resulting in their expansion and differentiation into CD8+PD-1int cells.
CD8+PD-1high TILs in Liver Cancer Shown a T-Cell Exhaustion Signature
To determine the transcriptional signature of CD8+ T cells naturally occurring in tumors, we analyzed the mRNA profiles of sorted CD8+PD-1neg, CD8+PD-1int, and CD8+PD-1high TILs of untreated mice at day 21 (gating strategy and purity, online suppl. Fig. 2a). As demonstrated in Figure 5a, the three CD8+ TIL subsets clearly segregated from each other. Multigroup comparison analysis identified 151 differentially expressed transcripts between the three CD8+ T-cell subsets (online suppl. Table 1). Two main clusters were identified corresponding to transcripts that were up- or downregulated in CD8+PD-1high TILs compared to the other two subsets (Fig. 5b). GO enrichment analysis revealed that processes involved in cell death, regulation of signal transduction, and regulation of response to stimuli were upregulated, while processes involved in cell proliferation and immune response were downregulated in CD8+PD-1high TILs (Fig. 5c; online suppl. Table 2).
The transcriptional signature of CD8+PD-1high TILs differed more strongly from that of CD8+PD-1neg TILs, as 122 transcripts were differentially expressed between these two subsets. In addition, 27 transcripts were exclusively up- or downregulated in this subset. In contrast, only 76 transcripts were differentially expressed between CD8+PD-1high and CD8+PD-1int TILs. CD8+PD-1int TILs and CD8+PD-1neg TILs were more similar, as only 46 transcripts were differentially expressed between these two subsets (Fig. 5d). Transcripts related to CD8+ T-cell exhaustion and dysfunctionality, such as Pdcd1 (PD-1), Havcr2 (TIM-3), Tnfrsf9 (4-1BB), Tigit, Csf1, Dusp4, Klri2, and Iftm1 [6, 7, 33‒35], were significantly upregulated in CD8+PD-1high TILs (Fig. 5e, f; online suppl. Table 2). Some transcripts frequently expressed by exhausted CD8+ T cells, such as Lag-3, Entpd1 (CD39), Tnfrsf18 (GITR), and CD200, tended to be upregulated, but their calculated q values were greater than 0.05 (Fig. 5g). Transcripts for PD-1 and TIM-3 were also upregulated in CD8+PD-1int TILs compared to CD8+PD-1neg TILs; however, their expression levels were significantly lower than those in CD8+PD-1high TILs, most likely reflecting the activated status of these cells (Fig. 5h). Transcripts for CCR7, L-Selectin (Sell), Lef1, and S1P1 (S1pr1) were downregulated in CD8+PD-1high TILs, indicating that these cells were activated, while CD8+PD-1neg cells most likely represented naïve T cells (Fig. 5e, f; online suppl. Table 3). In conclusion, CD8+PD-1high TILs show similarities to exhausted T cells found in both chronic LCMV infection and cancer models [3, 4, 6].
PD-1/CTLA-4 Blockade Only Modestly Affects the Transcriptional Signatures of the Responding CD8+ TIL Subtypes
To assess how checkpoint inhibition affects the different subsets of CD8+ TILs, we compared their transcriptional profiles in untreated and treated mice. Due to the low number of CD8+PD-1high TILs in the tumors of the treated mice, we could not sort this subset for analysis (online suppl. Fig. 2b). Multigroup comparison analysis of the five analyzed subsets identified 376 significant differentially expressed transcripts (Fig. 6a and online suppl. Table 4). Eleven clusters of genes displaying distinct expression profiles between CD8+PD-1high TILs and at least one of the other CD8+ TIL subsets were identified. Four of these clusters (C1-C4) contained transcripts that were significantly enriched in the GO analysis (Fig. 6b; online suppl. Table 5). The expression profiles of transcripts belonging to C2 and C4 were very similar between untreated CD8+PD-1high and treated CD8+PD-1int TILs. Transcripts in C2 identified upregulated processes involved in the regulation of response to stimulus, cell communication, cell proliferation, immunological response, signal transduction, and apoptosis. Transcripts in C4 identified downregulated processes involved in T-cell receptor rearrangement, T-cell activation and differentiation, cell migration, and proliferation. Surprisingly, when the CD8+PD-1high subset was excluded from the analysis, only a few significant differences in the transcriptional signatures of the remaining CD8+ TIL subsets were observed between untreated and treated mice (Fig. 6c). In addition, no significant differences in the expression of transcripts for effector molecules and cytokines were observed between CD8+PD-1int TILs of untreated and CD8+PD-1int TILs of treated mice, indicating that these cells had comparable functionalities.
We additionally tried to identify genes potentially involved in the abolishment of T-cell exhaustion after therapy by comparing untreated CD8+PD-1high TILs with the other two CD8+ subsets of untreated and treated mice. As shown in Figure 6d, untreated PD-1high TILs were more similar to treated PD-1neg + PD-1int TILs than to untreated PD-1neg + PD-1int TILs (37 vs. 118 exclusively regulated transcripts). As expected, transcripts related to T-cell exhaustion were upregulated in PD-1high TILs compared to untreated PD-1neg + PD-1int TILs (Fig. 6e; online suppl. Table 6). In contrast, Spp1 (osteopontin), Sprr1a (Cornifin-A), Serpine1, Vcan (Versican), Krt8 (Cytokeratin8), and Tm4sf1 were upregulated in PD-1high TILs compared to treated PD-1neg + PD-1int TILs (Fig. 6f; online suppl. Table 2). Osteopontin, also called Eta-1 (early T-lymphocyte activator-1), is a T-cell cytokine that is expressed following TCR signaling. The expression of osteopontin in activated CD8+ T cells is regulated by T-bet, and its expression is essential for the development of effector Tc1 CD8+ T cells. Osteopontin is also highly expressed in a subset of senescent CD4+PD-1+ memory T cells, which hardly proliferate after TCR stimulation [36]. Furthermore, expression of osteopontin has been recently demonstrated for CD8+PD1+ TILs isolated from mammary carcinoma of mice feed with high-fat diet [37]. Therefore, high osteopontin expression could be a common feature of dysfunctional T cells. The other differentially expressed transcripts were atypical for T cells, and their functions remain to be clarified.
Finally, to compare the transcriptional signatures of endogenous (CD45.2+) and tumor-specific (CD45.1+) CD8+PD-1int TILs, we transferred CD8+CD45.1+ HA-specific T cells in tumor-bearing mice 10 days after tumor induction and treated them subsequently as described. Endogenous CD8+CD45.2+ and transferred CD8+CD45.1 PD-1int TILs were sorted on day 21 and used for mRNA analysis. Due to the very low numbers of CD45.1+CD8+PD-1int TILs in the tumors of untreated mice, we compared only the three successfully sorted TIL subtypes. Only minor differences were observed between the analyzed TIL subtypes, independent of their origin or treatment, indicating that endogenous and transferred tumor-specific TILs had similar functions (Fig. 6g).
In summary, the treatment seems not significantly altering the effector capacity of the CD8+ TILs, making it unlikely that PD-1/CTLA-4 blockade reinvigorated exhausted CD8+ T cells. To prove this hypothesis, we examined the ability of these cells to produce TNF-α and IFN-γ after in vitro restimulation with Hep-HA-STK-1 cells plus anti-CD3 and anti-CD28 antibodies. Consistent with the mRNA data, a comparable proportion of CD8+ TILs able to produce TNF-α were found in the tumors of untreated and treated mice (online suppl. Fig. 4a, c). Moreover, comparable proportions of cells belonging to the three CD8+ TIL subsets produced TNF-α, regardless of therapy (online suppl. Fig. 4b, d). However, the total number of CD8+ TILs that produced TNF-α+ after restimulation was significantly increased in treated animals (online suppl. Fig. 4e). Remarkably, this increase was due to an increase in the number of TNF-expressing CD8+PD-1int TILs (online suppl. Fig. 4, f). In addition, part of these cells coexpressed IFN-γ and membrane-bound CD107a (online suppl. Fig. 5 a, b), indicating that these cells but not CD8+PD-1neg TILs are able to degranulate and exert the effector function after therapy.
Discussion
In the present work, we aimed to investigate the effect of PD-1/CTLA-4 blockade on exhausted CD8+ TILs in liver cancer. For this purpose, we established a model in which the induced tumors developed an inflamed TME, as evidenced by the presence of tertiary lymphoid organs and the accumulation of immune cells surrounding the tumor nodes. In agreement with previous studies, we observed three subtypes of CD8+ T cells (PD-1neg, PD-1int, and PD-1high) in the TME, although terminally exhausted CD8+PD-1high TILs predominated [6, 7].
Which subset of CD8+ T cells responds to checkpoint blockade remains controversial. In all cases, however, the response to therapy correlated with proliferation. Previous studies demonstrated that CD8+PD-1high T cells express higher levels of the proliferation marker Ki67 than CD8+PD-1int cells. However, despite a robust proliferation after in vitro stimulation, they only modestly proliferate in vivo. In contrast, CD8+PD-1int cells strongly expanded after adoptive transfer. In the majority of the studies, checkpoint blockade increased the expansion of CD8+PD-1int but not of CD8+PD-1high T cells [3, 4, 6]. In line with these observations, almost all CD8+PD-1high TILs in our model expressed Ki67 after checkpoint blockade, but no expansion of this subset was observed. Therefore, the expression of Ki67 seems to correlate with the ability of individual cells to divide but not with their ability to maintain proliferation and survive after TCR stimulation, which is essential for maintenance of the antitumor response.
In non-small-cell lung carcinoma, ex vivo-isolated CD8+PD-1high TILs are dysfunctional [7]. Although they can regain their effector function after in vitro cultivation, it is not clear whether they can do the same in vivo. In contrast, in experimental melanoma, CD8+PD-1high TILs show a higher cytotoxic capacity in vitro than CD8+PD-1int TILs. However, due to their limited proliferative capacity, they show a low ability to control tumor growth after adoptive transfer. In addition, transferred CD8+PD-1high cells cannot differentiate into CD8+PD-1int cells. In contrast, transferred CD8+PD-1int TILs differentiated into highly cytotoxic CD8+PD-1high TILs in vivo. Since increased numbers of CD8+PD-1high TILs were observed in the tumor of mice that received CD8+PD-1int TILs and checkpoint blockade, the authors proposed that PD-1 and PD-1/CTLA-4 blockades induce not only the expansion of CD8+PD-1int TILs but also their differentiation into CD8+PD-1high TILs [6].
In our model, only CD8+PD-1int TILs expanded in response to PD-1/CTLA-4 blockade. CD8+PD-1high TILs disappeared after treatment, showing that CD8+PD-1int TILs did not differentiate into CD8+PD-1high TILs. In addition, we did not observe the upregulation of transcripts for Granzyme A and B and effector cytokines in CD8+PD-1high compared to CD8+PD-1int TILs, as described for experimental melanoma [6]. As demonstrated by immunofluorescence, treatment increased the number of activated CD8+ TILs expressing CD44. Consistent with this observation, the total number of CD8+ TILs producing TNF-α and IFN-γ and exhibiting surface expression of CD107a after in vitro restimulation was also increased in treated mice. Interestingly, although CD8+PD-1neg and CD8+PD-1int cells were capable of producing cytokines, translocation of CD107a was virtually restricted to the PD-1int subset. It is therefore likely that in our model, CD8+PD-1int TILs perform the effector function. Since the transferred tumor-specific CD8+ T cells in our model responded to therapy with strong expansion of CD8+PD-1neg and CD8+PD-1int but not CD8+PD-1high TILs, we believe that treatment promotes the recruitment of naïve T cells into the TME, leading to their priming, activation, and proliferation. Due to activation, the expression of inhibitory receptors increases, resulting in the emergence of CD8+PD-1int cells. This hypothesis is consistent with previously published data, demonstrating that the response to PD-1 blockade is mediated by T-cell clones that have recently entered the tumor [38].
In conclusion, our data suggest that the amount of effector cells in the tumor plays a key role in tumor remission. Therefore, antitumor response after therapy depends on the proliferation of CD8+ TILs. There is increasing evidence that exhausted T cells exhibit metabolic insufficiencies that impedes proliferation after activation, and it is known that PD-1 blockade can lead to metabolic reprogramming that likely restores proliferation [39, 40]. Our GO analysis showed that several metabolic pathways were altered in CD8+PD-1high TILs compared with the other PD-1 subsets from untreated and treated mice. Therefore, it would be possible that metabolic reprogramming supports the expansion of CD8+PD-1int TILs in our model.
Besides proliferation, infiltration of activated CD8+ TILs into the tumor is a critical step required for tumor remission. Interestingly, we observed a slight but significant increase in the concentration of CCL2 in tumor lysates of treated mice compared to untreated mice (online suppl. Fig. 3). Elevated concentrations of CCL-2 after PD-1 blockade correlated with tumor infiltration by CD8+ T cells in a model of metastatic breast cancer [41]. In addition, nitration of CCL2 by reactive nitrogen species in the inflamed TME led to trapping of T cells in the stroma surrounding the tumor. Inhibition of CCL2 nitration resulted in tumor infiltration [42]. Nitration of CCL2 is not recognized by anti-CCL2 antibodies, so that nitration inhibition increases the CCL2 levels detected in the tumors. It is therefore plausible that checkpoint blockade inhibited the nitration of CCL2, resulting in the elevated levels of CCL2 observed in our model. In this case, CCL2 would be the crucial molecule involved in T-cell exclusion.
One limitation associated with checkpoint inhibitor treatment is that the presence of preexisting tumor-specific T cells is required for therapeutic success [43]. Genetically engineered T cells displaying chimeric antigen receptors have been successfully used in the treatment of B-cell malignancies [44, 45]. Unfortunately, the same success has not been observed in the treatment of solid tumors [44]. The failure of CAR-T-cell therapy in the treatment of solid tumors has been associated with poor migration to the tumor site, lack of tumor infiltration, and cell survival after transfer [46, 47]. In our liver cancer model, few doses of checkpoint inhibitors subsequent to T-cell transfer were sufficient to overcome these problems. This finding could have important implications for future therapies.
Acknowledgments
We would like to acknowledge the assistance of the Cell Sorting Core Facility of the Hannover Medical School and Prof. Dr. Christine Falk for providing analysis software and for constructive discussion of the data.
Statement of Ethics
Animal care and experiments were performed in accordance with institutional and national guidelines. All animal experiments were performed according to protocols approved by the Animal Welfare Commission of the Hannover Medical School and Local Ethics Animal Review Board (#33.12-42502-04-15/1779; April 20, 2015 and #33.19-42502-04-18/3063; May 10, 2019; Lower Saxony State Office for Consumer Protection and Food Safety, Oldenburg, Germany).
Conflict of Interest Statement
The authors have no conflicting interests.
Funding Sources
This work was supported by the Wilhelm Sander-Stiftung (Grant No. 2017.070.1 to Elmar Jaeckel and Ana C. Davalos-Misslitz), the Government of Canada’s New Frontiers in Research Fund (NFRF), NFRFT-2020-00787, and Deutsche Forschungsgemeinschaft (BU2722/2-3 to Laura Elisa Buitrago-Molina and HA6880/2-1 to Matthias Hardtke-Wolenski).
Author Contributions
Conceptualization: Elmar Jaeckel and Ana C. Davalos-Misslitz; methodology: Ana C. Davalos-Misslitz, Elmar Jaeckel, and Matthias Hardtke-Wolenski; investigation: Sandra Bufe, Artur Zimmermann, Sarina Ravens, Robert Geffers, and Ana C. Davalos-Misslitz; formal analysis: Sandra Bufe, Ana C. Davalos-Misslitz, Sarina Ravens, Robert Geffers, and Steven R. Talbot; validation: Elmar Jaeckel, Ana C. Davalos-Misslitz, Immo Prinz; resources, Norman Woller, Florian Kühnel, Michael Peter Manns, and Heiner Wedemeyer; visualization: Sandra Bufe, Ana C. Davalos-Misslitz, Laura Elisa Buitrago-Molina, and Fatih Noyan; writing – original draft: Ana C. Davalos-Misslitz and Sandra Bufe; writing – review and editing: Matthias Hardtke-Wolenski, Laura Elisa Buitrago-Molina, Sarina Ravens, Immo Prinz, Norman Woller, Fatih Noyan, Florian Kühnel, Steven R. Talbot, Michael Peter Manns, Heiner Wedemeyer, and Elmar Jaeckel; supervision: Ana C. Davalos-Misslitz, Elmar Jaeckel, and Matthias Hardtke-Wolenski; and funding acquisition: Elmar Jaeckel, Ana C. Davalos-Misslitz, Laura Elisa Buitrago-Molina, and Matthias Hardtke-Wolenski.
Data Availability Statement
All data generated or analyzed during this study are included in this article and its supplementary material files. Further enquiries can be directed to the corresponding author.
References
Additional information
Elmar Jaeckel and Ana C. Davalos-Misslitz contributed equally to this work.