Characterization of innate lymphoid cell subsets infiltrating melanoma and epithelial ovarian tumors

ABSTRACT

The innate lymphoid cell (ILC) family is composed of heterogeneous innate effector and helper immune cells that preferentially reside in tissues where they promote tissue homeostasis. In cancer, they have been implicated in driving both pro- and anti-tumor responses. This apparent dichotomy highlights the need to better understand differences in the ILC composition and phenotype within different tumor types that could drive seemingly opposite anti-tumor responses. Here, we characterized the frequency and phenotype of various ILC subsets in melanoma metastases and primary epithelial ovarian tumors. We observed high PD-1 expression on ILC subsets isolated from epithelial ovarian tumor samples, while ILC populations in melanoma samples express higher levels of LAG-3. In addition, we found that the frequency of cytotoxic ILCs and NKp46⁺ILC3 in tumors positively correlates with monocytic cells and conventional type 2 dendritic cells, revealing potentially new interconnected immune cell subsets in the tumor microenvironment. Consequently, these observations may have direct relevance to tumor microenvironment composition and how ILC subset may influence anti-tumor immunity.

KEYWORDS:

• Epithelial ovarian cancer
• innate lymphoid cells
• LAG-3
• melanoma
• myeloid cells
• NK cells
• PD-1
• T cells

Introduction:

The tumor microenvironment (TME) is a very complex entity. Highly heterogeneous, it is composed of a wide spectrum of immune and nonimmune cells that interact with cancer cells, directly influencing tumor progression, patient prognosis, and immunotherapy outcomes.^1,2 In melanoma and epithelial ovarian cancer, recent investigations have demonstrated that accumulation of immunosuppressive regulatory T cells,^3,4 tumor-associated macrophages (TAM),^5,6 and myeloid-derived suppressor cells (MDSC)⁷ often correlated with poor prognosis.⁸ In contrast, the accumulation of B cells^9–11, T cells^9,12–14, dendritic cells (DC),¹⁵ or natural killer (NK) cells¹⁴ in the TME are often associated with anti-tumor responses and better patient prognosis. However, these immune cell populations are heterogeneous and are comprised of multiple subsets which may differentially impact clinical outcomes. For example, while regulatory CD8⁺ T cells suppress anti-tumor immune responses,¹⁶ tumor-infiltrating tissue-resident memory CD8⁺ T cells (CD8⁺Trm), a specific subset of CD8⁺ T cells often characterized by the co-expression of the integrin CD103 and the lectin CD69, are regularly associated with favorable patient prognosis in cancer.^17–20 Similarly, conventional type 1 (cDC1) and type 2 (cDC2) dendritic cells, which promote CD8⁺ and CD4⁺ T cell responses, respectively, frequently correlate with enhanced treatment responses.^21,22 In contrast, accumulation of plasmacytoid DCs (pDCs) in tumors is often associated with poor prognosis.²³ Collectively, these studies highlight the heterogeneity and complexity of the tumor immune contexture, which profoundly influences disease outcomes.

More recently, the development of sophisticated mouse models and the use of single-cell RNA sequencing analysis of the TME have revealed important roles of rare and underrepresented immune cell subsets. Strikingly, while innate lymphoid cells (ILCs) may represent only a small fraction of the tumor immune infiltrate, they are associated with patient prognosis and treatment efficacy.^24–27 This heterogeneous family of cells is classified into five broad subsets, namely NK cells, type 1 (ILC1), type 2 (ILC2), and type 3 (ILC3) ILCs, and lymphoid tissue inducer cells, based on their developmental trajectories, transcription factors expression, and effector functions.²⁸ ILCs have drawn considerable interest in the field due to relatively similar transcriptional profiles and effector functions with their adaptive counterparts, T lymphocytes.²⁹ However, unlike T cells, they do not express antigen-specific receptors. The activity of ILCs is rather dictated by the direct integration of signals from the environment, which governs their inflammatory response.³⁰ Therefore, depending on the signals they receive, they can either drive pro- or anti-tumorigenic responses.^24,31 This apparent controversy in studies highlights the need to better understand tumor-infiltrating ILC composition and phenotype. Given that ILC responses depend on their microenvironment and the signals they receive, it is therefore critical to investigate their association with other major immune cell types in the TME³², with the goal of depicting a better picture of the whole TME. While NK cells are relatively well studied, our current understanding of other intratumoral ILC subsets in human melanoma and epithelial ovarian tumors remains limited and is mostly restricted to preclinical studies.^33,34 In addition, recent evidence indicates that, similar to T cells, ILCs express immune checkpoint molecules, such as PD-1^35,36 or LAG-3.^35,37,38 PD-1 is broadly expressed on ILC progenitors^39,40 and all mature ILC subsets^{33,37,41–44}, and has been shown to heavily influence the ILC effector function.^33,44–47 In contrast, LAG-3 expression seems to be restricted to chronically stimulated ILCs.^37,38 For example, chronic stimulation of NK cells induced LAG-3 expression and LAG-3⁺ NK cells displayed reduced IFN-γ expression.³⁸ It is critical to determine the expression of these immune checkpoint molecules on all ILC subsets in cancers such as melanoma and epithelial ovarian tumors as their expression, as reported in other experimental settings, may modulate ILC activity and influence clinical outcomes. Yet, the expression of these molecules on human tumor-infiltrating ILCs remains largely understudied.

Here, we investigate the composition and phenotype of ILC subsets in human melanoma and epithelial ovarian tumor samples, together with intratumoral T cells and myeloid cell subsets. While we found a similar distribution of tumor-infiltrating ILC subsets in melanoma and epithelial ovarian tumors, we observed a higher frequency of CD8⁺ resident memory T cells in ovarian samples compared to melanoma, which was associated with increased frequency of cDC1. On a phenotypic level, we observed higher expression of PD1 on both ILCs and T cells infiltrating epithelial ovarian tumors, while we found increased LAG3 expression on melanoma infiltrating ILCs. In addition, our correlational analyses between ILC, T cell, and myeloid cell subsets have revealed a positive association between the frequency of intratumoral NKp46⁺ ILC3s and cytotoxic ILCs with cDC2s in those tumors.

Material and methods

Patient samples: ethics

All human tissues and blood were obtained through protocols approved by the institutional review board. Surgical specimens were obtained from the UHN Biospecimen Program. Written informed consent was obtained from all donors. Research has been conducted in accordance with the principles stated in the Declaration of Helsinki.

Tumors

Tumor specimens were obtained from patients with melanoma or epithelial ovarian cancer undergoing standard-of-care surgical procedures. Patient characteristics are summarized in Supplementary Table S1. Fresh tumor samples were either processed into single-cell suspension by enzymatic digestion or mechanically separated into 0.3 mm³ tumor fragments. Enzymatic digestion was performed using Gentle MACs^TM Dissociation (Miltenyi Biotec) and digestion media (1 mg/mL collagenase, 10 µg/mL pulmozyme, 2 mmol/L L-glutamine, 100 µg/mL Amphotericin B in RPMI 1640). Single-cell suspensions or tumor fragments were cryopreserved in human serum containing 10% DMSO in liquid nitrogen tanks and thawed prior to analysis.

Blood

Peripheral blood mononuclear cells (PBMCs) were isolated from the blood of healthy donors using Ficoll. Briefly, blood collected in sodium heparinized tubes was diluted in PBS (1:1 ratio) before being layered on top of Ficoll. After centrifugation, the interface between Ficoll and plasma corresponding to PBMCs were collected and washed twice in PBS. Cells were then counted, frozen at –80 degrees in FCS + 10% DMSO, and transferred into liquid nitrogen tanks until further analysis by flow-cytometry.

Flow cytometry

PBMCs and viably frozen enzymatically digested tumor samples were thawed in complete media which consists of RPMI1640 (Gibco) supplemented with 10% FCS, 2 mM L-glutamine (Gibco), 100 U/mL penicillin (Gibco), 100 mg/mL streptomycin (Gibco), and 5 mM β-mercaptoethanol (Sigma). After centrifugation and resuspension in complete media, cells were counted and stained with a T cell, myeloid cell, or ILC staining panels at 4 degrees. Priority was given to the ILC staining panel when a sample contained less than 2 million cells. Cells were stained with Fc block (BD Biosciences, 1:200) in FACS Buffer, consisting of PBS supplemented with 2% FCS, for 20 min. Cells were washed and then stained with surface antibodies together with the viability dye for 30 min. Staining panels, antibodies, clones, and dilutions are detailed in Supplementary Table S2. Intracellular staining was then performed using the Foxp3 transcription factor staining buffer set. Cells were first permeabilized for 30 min, washed in 1X permeabilization buffer, and then stained with intracellular antibodies for 30 min. Samples were acquired on a five-laser BD Fortessa flow cytometer, and data were analyzed with the FlowJo software (version 10.8).

TCGA analysis

We used the TCGA RNA-seq of melanoma and ovarian data to validate the positive correlation between cDC1 and CD8⁺ T_RM cells, tumor associated macrophages (TAMs) and NK cells, and pDCs and cDC2. The upper-quartile normalized and log-transformed FPKM matrices in TCGA-SKCM (skin cutaneous melanoma) and TCGA-OV (ovarian) were downloaded from Xenabrowser (https://xenabrowser.net/) as of June 21, 2023, and we extracted CLEC9A (ENSG00000197992.5), CD8A (ENSG00000153563.14), MRC1 (ENSG00000260314.2), NCR1 (ENSG00000189430.11), CD1C (ENSG00000158481.11), and CLEC4C (ENSG00000198178.9) expression across samples with the GENCODE annotation (v22). Samples with zero expression values for the corresponding genes were removed from the correlation analyses. Pearson correlation coefficient (r) was carried out across tumor types using the cor.test function in the R software (v4.2.2).

Statistical analysis

Data analyses and representations were performed with either R software or Prism (GraphPad version 9.0). For more than two groups, statistical analyses were performed using ANOVA followed by Tukey’s multiple-comparison test or pairwise comparisons with Bonferroni adjustments. Otherwise, for two groups, statistical analyses were performed using the unpaired Student’s t-test. The results are shown as the mean ± SEM correlations between two variables were assessed using non-parametric Spearman correlation tests and adjusted p-values were reported. p-values were two-sided with 95% confidence intervals and considered significant at p < 0.05.

Results

Melanoma and epithelial ovarian tumor samples have been known to have different degrees of sensitivity to immune checkpoint inhibitors. Thus, we analyzed in detail the frequency and phenotype of various ILC populations in 10 human melanoma and 10 human epithelial ovarian tumor samples (Supplementary Table S1). We used peripheral blood from four healthy donors as control for baseline expression levels of the investigated parameters. In addition, we monitored the frequency and phenotype of several T cell and myeloid cell subsets with the aim to correlate ILC infiltration to the TME composition.

CD56^dimCD16^– NK cells that infiltrate epithelial ovarian tumors express low levels of PD-1

NK cells are considered the prototypical member of the ILC family and, by far, are the most studied ILC subset in cancer.²⁵ While we have a good understanding of NK cell receptors, less is known about the immune checkpoint expression on these cells. Total live NK cells were gated as lineage (CD3, TCRαβ, TCRγδ, CD14, CD19, CD20, CD34, CD123, CD303, FCεRI)^–CD127^–CD56⁺ within the total lymphocyte gate (Figure 1a). Three NK cell subsets were defined, namely CD56^dimCD16^–, CD56^dimCD16⁺ and CD56^brightCD16^–, based on the level of expression of CD56 and CD16 (Figure 1a). We observed an enrichment of the CD56^dimCD16^– NK cell subset in tumors compared with peripheral blood, at the expense of the CD56^dimCD16⁺ population (Figure 1b). These findings could be a result of a potential downregulation of CD16 expression, identifying activated NK cells.^48,49 Within each NK cell subset, we assessed the expression of NK-cell related markers and immune checkpoint molecules which included CD94, CD161, NKp46, CD25, ICOS, TIGIT, PD-1, and LAG3 (Figure 1c–e). Unexpectedly, no striking differences were observed between tumor types or with circulating cells (Figure 1c–e). Of note, CD56^brightCD16^– NK cells express higher levels of CD94 and NKp46 (Figure 1d), as previously reported.⁵⁰ Interestingly, while relatively low expression, we observed increased PD-1 expression at the surface of ovarian tumor-infiltrating CD56^dimCD16^– NK cells compared with NK cells found in melanoma tumors or in circulation (Figure 1e). Collectively, although no major changes in this population exist between melanoma and epithelial ovarian samples, there is evidence that the CD56^dimCD16^– subset expresses slightly increased levels of PD-1 in the epithelial ovarian TME.

Figure 1. Accumulation of CD56^dimCD16^– NK cells expressing PD-1 in ovarian tumors. a. Representative flow-cytometric dot plots showing the gating strategy used to identify circulating (PBMCs from healthy donors; left panels) and tumor-infiltrating (melanoma and epithelial ovarian tumors; middle and right panels, respectively) NK cell subsets. Cells were first gated in CD45⁺SSC-A^low and NK cells identified as lineage (Lin; CD3, TCRαβ, TCRγδ, CD14, CD19, CD20, CD34, CD123, CD303, FCεRI)^–CD127^–CD56⁺. Subsequently, three populations were identified based on CD56 and CD16 expression levels, namely CD56^brightCD16^– NK cells, CD56^dimCD16^– NK cells, and CD56^dimCD16⁺ NK cells. b. Frequency of circulating and tumor-infiltrating (from left to right) CD56^brightCD16^– NK cells, CD56^dimCD16⁺ NK cells, and CD56^dimCD16^– NK cells. c. Heatmap showing the log₂ fold change in geometrical mean fluorescent intensity (gMFI) of indicated parameters between circulating and tumor-infiltrating NK cell subsets. Circulating cells from healthy donors set as the reference. An asterisk denotes statistical significance between circulating and tumor-infiltrating groups. d. Bar graphs showing the gMFI of CD94, CD161 and NKp46 (from top to bottom) expression on circulating and tumor-infiltrating NK cell subsets. E. Representative flow-cytometric histograms (left), gMFI (top right) and frequency (bottom right) of PD-1 expression on circulating and tumor-infiltrating NK cell subsets. B, D, E. Data show mean±SEM. Each dot represents one sample. Statistical analyses were performed using one-way ANOVA followed by a Tukey’s multiple comparisons test. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

Melanoma and ovarian tumors displayed reduced frequencies of ILC2 and ILC3 compared to peripheral blood

We then investigated the frequencies and phenotype of other ILC subsets in melanoma and epithelial ovarian tumor samples. Total live ILCs were gated as lineage (CD3, TCRαβ, TCRγδ, CD14, CD19, CD20, CD34, CD123, CD303, FcεRI)^– CD127⁺ within the total lymphocyte gate (Figure 2a). Five distinct ILC populations were subsequently identified based on CD56, CD94, CD117, CRTH2, and NKp46 expression. The ILC subsets that were identified include a CD94⁺CD56⁺ subset, resembling the recently described cytotoxic ILC3s,⁵¹ and three helper ILC populations. The helper ILC populations include ILC1 (CD94^–CD56^–CRTH2^–CD117^–), ILC2 (CD94^–CD56^–CRTH2⁺CD117^±) and ILC progenitors (ILCp; CD94^–CD56^–CD117⁺CRTH2^–).⁵² ILCp is comprised of (a) multi-potent cells (NKp46^–), having the ability to give rise to all ILC subsets, and (b) a committed lineage restricted ILC3, characterized by NKp46⁵³ expression (Figure 2a). While no reduction in total ILCs was found in tumors compared with peripheral blood, we observed significant changes in the frequency of individual ILC subsets. Decreased CD94⁺CD56⁺ ILCs were detected in tumor samples (Figure 2b). Remarkably, we found reduced intratumoral ILC2 associated with increased ILCp (Figure 2c), specifically the NKp46^–ILCp subset, at the expense of committed NKp46⁺ILC3s (Figure 2 d), compared with peripheral blood cells. While we found significant changes between the frequency of peripheral ILCs isolated from healthy donors and tumor-infiltrating cells, we did not observe any significant differences in tumor-infiltrating non-NK cell ILC subsets between melanoma and epithelial ovarian tumor samples.

Figure 2. Enrichment in ILCps at the expense of cytotoxic ILCs, ILC2 and ILC3. a. Representative flow-cytometric dot plots showing the gating strategy used to identify circulating (PBMCs from healthy donors; left panels) and tumor-infiltrating (melanoma and epithelial ovarian tumors; middle and right panels, respectively) ILC subsets. Cells were first gated in CD45⁺SSC-A^low and ILCs identified as lineage (Lin; CD3, TCRαβ, TCRγδ, CD14, CD19, CD20, CD34, CD123, CD303, FCεRI)^–CD127⁺. Cytotoxic ILCs, ILC1, ILC2, CD117⁺CRTH2^–, NKp46-ILCp and ILC3 were identified as CD56⁺CD94⁺, CD56^–CD94^–CD117^–CRTH2^–, CD56^–CD94^–CD117^±CRTH2⁺, CD56^–CD94^–CD117⁺CRTH2^–, CD56^–CD94^–CD117⁺CRTH2^–NKp46^–, and CD56^–CD94^–CD117⁺CRTH2^–NKp46⁺, respectively. B-D. Frequency of circulating and tumor-infiltrating (b, from left to right) ILCs and cytotoxic ILCs, (c) ILC1, ILC2, and CD117⁺CRTH2^–, and (d) NKp46^–ILCp and ILC3. B-E. Data show mean±SEM. Each dot represents one sample. Statistical analyses were performed using one-way ANOVA followed by a Tukey’s multiple comparisons test. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

Tumor-infiltrating cytotoxic ILC3s exhibit a reduction of NK-cell-related molecule expression

Cytotoxic ILC3s were first characterized in tonsil and peripheral tissues.⁵¹ In line with the reported phenotype, we found high levels of CD56, CD161, and NKp46 expression, intermediate levels of CD117, and low levels of ICOS on peripheral cytotoxic ILC3s (Lin^–CD127⁺CD56⁺CD94⁺) isolated from healthy donors (Figure 3a–g). In contrast to what was previously reported,⁵¹ we found that approximately 40% of circulating cells express CD16 (Figure 3e). In melanoma and epithelial ovarian tumors, we observed profound reduction of CD94, CD56, and NKp46 expression compared with circulating cytotoxic ILC3s isolated from healthy donors (Figure 3a–d). In addition, we found reduced CD16 expression on tumor-infiltrating cytotoxic ILC3s, in agreement with the original report⁵¹ (Figure 3e). We did not observe any noticeable differences in the level of expression of CD25, CD161, TIGIT, PD-1, LAG3, CD117, between circulating and tumor-infiltrating cytotoxic ILCs, apart from a slight decrease in ICOS expression on ILC3s from melanoma compared with epithelial ovarian tumors (Figure 3f).

Figure 3. Tumor-infiltrating cytotoxic ILCs displayed reduced NK-cell related markers expression. a. Representative flow-cytometric dot plots showing the frequency of cytotoxic ILCs according to their tissue of origin. b. Geometrical mean fluorescent intensity (gMFI) of CD94 and CD56 expression on cytotoxic ILCs isolated from healthy donors, melanoma or ovarian carcinoma patients. c. Heatmap showing the log₂ fold change in gMFI of indicated parameters between circulating and tumor-infiltrating cytotoxic ILCs. Circulating ILCs from healthy donors set as the reference. Asterisk denotes statistical significance between circulating and tumor-infiltrating groups. d. Representative histograms (left) and gMFI (right) of NKp46 expression. e. Representative histograms (left), gMFI (middle), and frequency (right) of CD16 positive cytotoxic ILCs according to their tissue of origin. f. Representative histograms (left) and frequency (right) of CD25, TIGIT, LAG3, TIM3, CD161, PD-1, CD117, ICOS expression on cytotoxic ILCs. FMO controls are shown in light gray. B,D-F. Data show mean+SEM. Each dot represents one sample. Statistical analyses were performed using one-way ANOVA followed by a Tukey’s multiple comparisons test. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

To date, there are no markers that accurately define human ILC1, and their identification essentially relies on the absence of expression of other ILC-specific molecules. While a relatively low proportion of circulating ILC1 express CD161, we found that almost half of the population expresses this receptor in tumors (Supplementary Figure S1A). Conversely, CD127 expression was found to be slightly reduced on tumor-infiltrating cells (Supplementary Figure S1B). Although not significant, we observed a trend toward increased expression of ICOS, TIGIT, PD-1, and LAG3 expression on tumor-infiltrating ILC1 compared with cells isolated from the blood of healthy donors (Supplementary Figure S1C and D). These findings suggest that tumor-infiltrating ILC1 harbors a phenotype compatible with increased cellular activity, characterized by enhanced expression of stimulatory and inhibitor molecules as well as decreased of IL-7 receptor expression.

Tumor-infiltrating ILC2, ILC3, and NKp46^–ILCp express PD-1 and LAG3

The ILC2 subset is heterogeneous⁵⁴ and is composed of two main populations according to CD117 expression (Figure 4a).⁵⁵ CD117⁺ ILC2 are considered more immature, having the capacity to transdifferentiate into ILC3, secreting IL-17A and IFNγ upon type 3 stimulation.⁵⁶ In contrast, CD117^–ILC2 produce higher levels of type 2 cytokines through increased expression of GATA3, denoting a more mature ILC2 population.⁵⁵ In melanoma and epithelial ovarian tumors, we found heterogenous CD117 expression on ILC2 between individuals (Figure 4b), indicating the presence of both subsets. As environmental signals have been known to affect these cells,^55–57 we hypothesized that ILC2s would harbor distinct characteristics between melanoma and epithelial ovarian tumors. Hence, we assessed CRTH2, CD127, CD161, CD25, ICOS, TIGIT, PD-1, and LAG3 expressions on CD117⁺ and CD117^–ILC2 subsets, separately (Figure 4c–e). While no reduction in CRTH2 expression was observed, we found a strong diminution of CD127 and CD25 expression on tumor-infiltrating ILC2, irrespective of the tumor type or of the ILC2 subset studied (Figure 4c–e). Although not significant, we observed a trend toward increased ICOS expression on CD117⁺ and CD117^– ILC2 in comparison with their blood counterparts. Interestingly, increased PD-1 expression was observed on CD117⁺ ILC2 infiltrating epithelial ovarian tumors, whereas CD117^– ILC2 in melanoma upregulated LAG3 expression in comparison with circulating cells from healthy donors.

Figure 4. Tumor-infiltrating ILC2 express PD-1 and LAG3. a. Representative flow-cytometric dot plots showing the frequency of ILC2 and the proportion of CD117⁺ and CD117^– cells according to their tissue of origin b. Cumulative frequency of circulating and tumor-infiltrating CD117^– and CD117⁺ ILC2 subsets. c. Heatmap showing the log₂ fold change in gMFI of indicated parameters between circulating and tumor-infiltrating ILC2 subsets. Circulating ILC2s from healthy donors set as the reference. Asterisk denotes statistical significance between circulating and tumor-infiltrating groups. d. Bar graphs showing the gMFI of CRTH2, CD127 and CD161 (from left to right) expression on circulating and tumor-infiltrating ILC2 subsets. e. Representative histograms (top) and quantification of the gMFI (middle) and frequency (bottom) of (from left to right) CD25, ICOS, TIGIT, PD-1 and LAG3 expression on CD117^–ILC2 and CD117⁺ILC2. FMO controls are shown in light gray. B,D,E. Data show mean+SEM. Each dot represents one sample. Statistical analyses were performed using one-way ANOVA followed by a Tukey’s multiple comparisons test. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

As ILC2 may arise from committed CD117⁺ precursors that express CD25,⁵² within ILCps, we assessed CD25 expression in NKp46^– ILCps. We found a strong downregulation of CD25 expression in tumor-infiltrating cells compared with peripheral ILCps (Figure 5a), potentially explaining the reduced ILC2 frequency observed in tumors (Figure 2c). Further analysis of the NKp46^– ILCp subset has revealed increased CD161 expression on ovarian-infiltrating cells (Figure 5b), whereas higher levels of LAG3 expression were found on melanoma-infiltrating cells (Figure 5c–d). Although not significant, we observed increased PD-1 expression in NKp46^– ILCp from epithelial ovarian tumors in comparison with PBMCs (Figure 5c–d).

Figure 5. Reduced CD25 expression on tumor-infiltrating NKp46^–ILCps. a-b. Representative histogram (left) and quantification of the gMFI (middle) and frequency (right) of CD25 (a) and CD161 (b) expression on NKp46^–ILCps. c. Heatmap showing the log₂ fold change in gMFI of indicated parameters between circulating and tumor-infiltrating NKp46^–ILCps. Circulating NKp46^–ILCps from healthy donors set as the reference. Asterisk denotes statistical significance between circulating and tumor-infiltrating groups. d. Representative histograms (top) and quantification of the gMFI (middle) and frequency (bottom) of (from left to right) ICOS, TIGIT, PD-1 and LAG3 expression on NKp46^–ILCps. FMO controls are shown in light gray. a,b,d. Data show mean+SEM. Each dot represents one sample. Statistical analyses were performed using one-way ANOVA followed by a Tukey’s multiple comparisons test. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

The analysis of tumor-infiltrating NKp46⁺ILC3 has revealed reduced levels of NKp46 (Figure 6a) and absence of CD25 (Figure 6b) expression when compared with circulating cells. Further phenotypic characterization of these cells has uncovered increased LAG3 expression in melanoma, whereas PD-1 was overexpressed in epithelial ovarian tumors in comparison with circulating NKp46⁺ILC3 isolated from peripheral blood of healthy donors (Figure 6c–d).

Figure 6. NKp46+ILC3s infiltrating melanoma tumors overexpress LAG3. a. Geometrical mean fluorescent intensity (gMFI) of NKp46 expression on NKp46⁺ILC3s isolated from healthy donors, melanoma or ovarian carcinoma patients. b. Representative histogram (left) and quantification of the gMFI (middle) and frequency (right) of CD25 expression on NKp46⁺ILC3s. c. Heatmap showing the log₂ fold change in gMFI of indicated parameters between circulating and tumor-infiltrating NKp46⁺ILC3s. Circulating NKp46⁺ILC3s from healthy donors set as the reference. Asterisk denotes statistical significance between circulating and tumor-infiltrating groups. d. Representative histograms (top) and quantification of the gMFI (middle) and frequency (bottom) of (from left to right) ICOS, TIGIT, PD-1 and LAG3 expression on NKp46⁺ILC3s. FMO controls are shown in light gray. a,b,d. Data show mean+SEM. Each dot represents one sample. Statistical analyses were performed using one-way ANOVA followed by a Tukey’s multiple comparisons test. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

Collectively, these observations highlight strong specificities in the TME that drive common phenotypic traits on to all ILC subsets. Remarkably, tumor-infiltrating ILC in epithelial ovarian tumors expresses higher levels of PD-1 compared with melanoma-infiltrating or circulating cells, while in melanoma, the TME promotes high LAG3 expression.

Increased CD8⁺T_RM cell infiltration is associated with enriched cDC1 in epithelial ovarian tumors

As ILCs have been known to influence other populations of cells, including T cells, B cells, and myeloid cells,^58,59 we sought to analyze the differences of these tumor-infiltrating subsets using high-dimensional flow cytometry. First, frequencies of total live SSC-A^lowCD45⁺ B cells and T cell subsets were analyzed (Supplementary Figure S2A). We did not observe striking differences in the frequency of T cell and B cell subsets between melanoma and epithelial ovarian samples (Figure 7a–c). Interestingly, we observed increased CD8⁺T_RM cell infiltration in epithelial ovarian tumors compared with melanoma lesions (Figure 7d), suggesting potential tumor-intrinsic specificity driving the accumulation of this subset in epithelial ovarian samples. Given the differences observed in the ILC phenotype between epithelial ovarian and melanoma tumors, we then inquired about potential differences in the expression of PD-1, CD39, CD69, CD103, and the proliferation indicator Ki67 between melanoma and ovarian-infiltrating CD8⁺ T cells and CD4⁺ conventional and regulatory T cells (Supplementary Figure S2B-E). Interestingly, we found increased PD-1 expression at the surface of ovarian tumor-infiltrating T cells in comparison with melanoma tumors (Figure 7e), mirroring our observations on ILC subsets. In addition, we observed a trend toward enhanced expression of CD69 and CD103 on T cells infiltrating epithelial ovarian tumors, while CD39 expression and T cell proliferation capabilities were similar between the two tumor types (Supplementary Figure S2C-E). Collectively, these observations suggest that T cells infiltrating epithelial ovarian tumors express higher levels of PD-1 in comparison with melanoma-infiltrating cells.

Figure 7. Ovarian carcinoma tumors are enriched in PD-1 expressing T cells, CD8⁺T_RM and cDC1. a-c. Frequency of circulating and tumor-infiltrating (a) b cells, (b, from left to right) CD3⁺ T cells, γδ⁺T cells, CD4⁺ T cells, CD8⁺ T cells, and (c) Foxp3⁺ regulatory T cells (tregs). Populations were identified as shown in supplementary figure S2A. d. Representative flow-cytometric color plots (left) and frequency (right) of circulating and tumor-infiltrating CD8⁺T_RM cells. e. Representative flow-cytometric histograms (left) and frequency (right) of PD-1 expression in circulating and tumor-infiltrating (from left to right) CD4⁺ tconv, CD4⁺ tregs and CD8⁺ T cells. f. Frequency of circulating and tumor-infiltrating macrophages (left) and conventional CD103⁺ cDC1 (right), as identified in supplementary figure 3, are shown. g. Correlation between the frequency of intratumor cDC1 and CD8⁺ T_RM cells in melanoma and epithelial ovarian samples. (left) comparison of the frequency of tumor-infiltrating cDC1⁺ relative to tumor-infiltrating CD8⁺ T_RM cells. The linear regression curve was overlaid. Dots are colored according to tumor type with blue and red representing respectively ovarian carcinoma and melanoma samples. (right) scatter plot depicting the expression levels of log2 FPKM(CLEC9A) (X-axis) and log2 FPKM(CD8A) (Y-axis) in melanoma (red) or ovarian (blue) cancer samples. A-F. Data show mean±SEM. Each dot represents one sample. Statistical analyses were performed using (A-F) unpaired Student’s t-test or (G) Spearman correlation test. *p < 0.05; **p < 0.01; ns, not significant.

Next, we investigated the composition of the myeloid cell compartment. We assessed the frequency of several myeloid cell populations (gated in total live SSC-A^highCD45⁺Lin^–(CD3^–CD19^–CD56^–) in blood samples and melanoma and epithelial ovarian tumors using high-dimensional flow-cytometry (Supplementary Figure S3). Unexpectedly, apart from an enrichment of macrophages and cDC1 in epithelial ovarian samples (Figure 7f), we found similar frequencies of monocytes, tumor-associated macrophages, other dendritic cell subsets, and neutrophils between epithelial ovarian and melanoma samples (Supplementary Figure S4A-F). Of note, circulating myeloid cells are mainly composed of monocytes and neutrophils, while macrophages constitute the main subset in tumor samples (Supplementary Figure S3–4 and Figure 1f). Strikingly, we found a positive correlation between the frequencies of tumor-infiltrating cDC1 and CD8⁺T_RM cells (Figure 7g). Utilizing publicly available TCGA datasets, we analyzed RNA-seq profile of 472 melanoma and 392 ovarian samples. The correlation coefficient between CLEC9A, mostly expressed by cDC1, and CD8A, a CD8⁺ T cell marker, was 0.651 (p = 4.9E–96) in 787 samples (Figure 7g). Altogether, these observations indicate that the increased PD-1 expression on T cells was associated with the enrichment of cDC1 and the accumulation of CD8⁺T_RM cells in epithelial ovarian tumors.

Frequencies of intratumor NKp46^–ILC3 and cytotoxic ILCs positively correlate with cDC2 tumor infiltration

Giving the fact that ILCs were previously shown to influence TME composition in preclinical mouse models, we questioned whether similar observations could be made in human melanoma and epithelial ovarian cancer. We generated correlation matrix plots to visualize potential correlations between tumor-infiltrating ILCs and other adaptive or innate immune cell subsets in melanoma and epithelial ovarian samples, combined altogether (Figure 8a) or analyzed separately (Supplementary Figure S5A and B). Positive and negative correlations are indicated in red and blue, respectively, and the size of the circle provides indications on the strength of detected associations between two variables. While tumor specificities were observed (Supplementary Figure S5A and B), the combined analysis has revealed that only a few subsets significantly correlated with each other after correction for multiple testing. Immune subsets that were significantly correlated together include the following: cDC1 and CD8⁺T_RM cells (Figure 7g), NK cells with TAM (Supplementary Figure S5C), and pDCs with monocytic cDC2 and cDC2 (Supplementary Figure S5D). Using the TCGA database, we confirmed these findings by looking at NCR1, MRC1, CLEC4C, CD1C gene expression levels, identifying NK cells, TAMs, PDCs, and cDC2, respectively, and their correlations (Supplementary Figure S5C and D). In regard to other ILC subsets, we found that the frequencies of intratumor cytotoxic ILC3 and NKp46⁺ ILC3 were both positively correlated with cDC2 (Figure 8a–b). These findings suggest potential unilateral or reciprocal communication paths between cDC2 and ILC3, which may influence their accumulation in tumors.

Figure 8. Frequency of tumor-infiltrating cytotoxic ILC3s and NKp46⁺ILC3s correlate with cDC2 infiltration. a. Correlation plot showing Spearman rho between indicated immune cell subsets. Positive and negative correlations are shown in red and blue, respectively, and the size of the circle provide indications on the strength of detected associations between two variables. Asterisk denotes statistical significance between indicated subsets and represents adjusted p-values using Spearman‘s rank correlation coefficient. Blood samples were excluded from this analysis. b. Comparison of the frequency of tumor-infiltrating cDC2⁺ relative to tumor-infiltrating cytotoxic ILCs (left) or NKp46⁺ILC3s (right). The linear regression curves were overlaid. Dots are colored according to tumor type with blue and red representing respectively ovarian carcinoma and melanoma samples. Each dot represents one sample. Statistical analyses were performed using Spearman correlation test. *p < 0.05; **p < 0.01; ***p < 0.001; ns, not significant.

Discussion

The composition of the tumor microenvironment and the phenotype of tumor-infiltrating immune cells profoundly influence cancer patient prognosis and treatment outcomes.^1,2 In melanoma and epithelial ovarian tumors, the accumulation of CD8⁺ T cells^12,13,60,61 and the presence of tertiary lymphoid structures^9,11 have been associated with improved survival. In contrast, the prevalence of Tregs^3,62 and M2 macrophages^63,64 in the TME was linked to poor prognosis. While these immune cell populations have been extensively characterized in those tumors, we still know very little about the distribution and phenotype of other underrepresented immune cells such as tumor-infiltrating ILCs. Our study provides a framework for an in-depth phenotyping of melanoma and epithelial ovarian TME with the aim to identify key differences in immune cell composition and phenotype. Here, we characterized T cell, myeloid, and ILC infiltrates using three validated flow cytometry panels (Supplementary Figures S2 and 3, and Figures 1 and 2).

The analysis of ILC subsets has revealed key differences in the frequency and phenotype of tumor-infiltrating and circulating cells (Figures 1 and 2). Particularly, we found that CD56^dimCD16^– NK cells accumulate in the TME of melanoma and epithelial ovarian tumors (Figure 1B). Previous studies have similarly reported an expansion of this subset following activation⁶⁵ and their accumulation in melanoma in response to the DC vaccine.⁶⁶ In non-NK cell ILCs, we observed reduced frequencies of tumor-infiltrating ILC2, cytotoxic ILC3 and NKp46⁺ ILC3 compared to circulating cells of healthy donors, while the frequency of NKp46^– ILCps was increased (Figure 2). Although Vazquez et al.⁶⁷ have previously identified distinct ILC clusters from the ascites from patients with ovarian cancers, our study is the first characterization of tumor-infiltrating ILCs in primary epithelial ovarian cancer. In regard to the TME of melanoma, previous studies have revealed contrasting evidence in regard to whether tumor-infiltrating ILCs subsets differ in frequencies compared to circulating ILCs from healthy donors.^42,68 These contrasting results may be due to differences in patient cohorts, disease stages, and previous line of treatment. Moreover, the phenotype of circulatory ILCs from patients with different cancer types has been relatively unexplored. Future work should identify how disease burden and different therapies may influence the proportion and phenotype of circulatory ILCs and matched ILCs from the TME of cancer patients.

Furthermore, we found that increased cDC1 frequency was associated with enhanced CD8⁺T_RM cell infiltration in epithelial ovarian cancer compared to melanoma (Figure 7). In agreement with our observation, preclinical studies using breast and head and neck mouse cancer cell lines, as well as humanized mouse models transplanted with human breast cancer samples, have shown that DC activation and reprogramming toward cDC1 in the TME leads to CD8⁺ T_RM generation and accumulation in tumors. Although these changes have also been associated with anti-tumor immune responses and disease control,^69,70 because of the small sample size of our cohorts, we were not able to test whether the accumulation of cDC1 and CD8⁺T_RM cell frequencies were associated with patient prognosis. Previous studies have demonstrated that the clinical efficacy of anti-PD1 therapies was associated with (i) PD-1 expression in the TME,^71,72 (ii) tumor infiltration by cDC1,^73,74 and (iii) CD8⁺ T_RM cells.^75,76 Because recent clinical trials have found that patients with epithelial ovarian cancer do not respond well to anti-PD-1/PD-L1 therapies,^77–82 we initially hypothesized that PD-1 expression on tumor-infiltrating T cells would be lowly or not expressed in patients with epithelial ovarian cancer. Surprisingly, our flow cytometric data revealed that tumor-infiltrating CD4⁺ and CD8⁺ T cells from epithelial ovarian tumors expressed high levels of PD-1 (Figure 7E). One possibility which could explain why patients with epithelial ovarian cancer do not respond to anti-PD-1 therapies despite the presence of PD-1⁺ T cells is that immunosuppressive cells such as Tregs may inhibit T cells and prevent an effective anti-tumor response in these patients. Using preclinical mouse models, we previously observed that resistance to anti-PD-1 antibody was associated with the presence of Tregs and high PD-1 expression on tumor-infiltrating CD8⁺T cells.⁸³ Reinforcing this notion, we have also found that tumor-infiltrating Tregs from epithelial ovarian cancer had an elevated expression of PD-1, Foxp3, CD25, ICOS, and 4-1BB associated with increased suppressive capabilities in comparison to tumor-infiltrating Tregs from patients with melanoma.⁸⁴ Corroborated findings from other groups also indicate that immunosuppressive Tregs not only impact primary resistances in non-responsive tumors such as in epithelial ovarian cancer but also influence anti-PD-1 Ab treatment efficacy in melanoma and other cancer types.^{74, 85–87} Collectively, our observations, corroborated by others, reinforce the importance of considering Tregs targeting in combination with ICB treatment to restore anti-tumor immunity and fully unleash the anti-tumor potential of effector T cells.⁸⁸

Interestingly, elevated PD-1 expression was not only found on T cells but also on tumor-infiltrating ILC subsets from epithelial ovarian cancers. More specifically, PD-1 was expressed at low levels on CD56^dimCD16^– NK cells (Figure 1), intermediate levels on ILC2 (Figure 4), and high levels on NKp46^–ILCps (Figure 5) and NKp46⁺ILC3s (Figure 6), suggesting the presence of common factors in epithelial ovarian tumors that drive, or sustain, PD-1 expression across different lymphoid cell subsets. Around 10–15% of CD56^dimCD16^– NK cells express PD-1, whereas low or no PD-1 expression was observed on melanoma-infiltrating or circulating NK cells (Figure 1E). Although conflicting results exist with regards to PD-1 expression on NK cells,^{35,36,89–92} a recent report suggested that activated NK cells could acquire PD-1 expression through trogocytosis,⁹³ a mechanism by which a part of the membrane is exchanged between two cell types that interact together, reconciling seemingly paradoxical observations. In contrast to NK cells, previous reports unanimously found PD-1 expression on ILC2 and the PD-1 expression level is further upregulated on activated or tumor-infiltrating cells.^{33,41,43,45,94} In line with these previous observations, we found increased PD-1 expression on tumor-infiltrating ILC2, particularly the CD117⁺ subset (Figure 4). Furthermore, we extended these results to epithelial ovarian tumor-infiltrating NKp46^–ILCps (Figure 5) and NKp46⁺ILC3s (Figure 6). Differences between previous studies, including our own work, reporting high PD-1 expression on ILCs in melanoma^33,94 and our current findings might be explained by the specific characteristics of our melanoma cohort. Indeed, most of the patients included in this study have previously received, but not responded to, immunotherapies, including anti-PD-1 antibodies (Supplementary Table S1), which may have impacted the level of expression of this checkpoint molecule on the surface of lymphoid cells. Interrogating the impact of the blockade of PD-1 signaling on ILCs, Cristiani and colleagues⁴² recently found that anti-PD-1 treatment in melanoma patients restored PD-1-expressing CD117^–ILC2 proliferation as well as promoted CD117⁺ILC2-derived IL-13 expression and CD117⁺ILCs derived TNFα production. Most importantly, their analysis suggests that high frequency of circulating CD117^–ILC2 before treatment was negatively correlated with patient survival.⁴² Collectively, these data indicate that ILCs represent important, yet ill-defined, immune cell populations that play a role in mediating anti-tumor responses in anti-PD-1 therapy in cancer patients. Future studies are therefore warranted to precisely determine the impact of anti-PD-1 blocking antibody on the effector function of both human circulating and tumor-infiltrating ILCs, as well as to better define the prognostic value and impact of these innate immune cells in treatment efficacy.

Besides PD-1 expression, we also found high LAG-3 expression on melanoma-infiltrating ILC2, (Figure 4), NKp46^–ILCps (Figure 5) and NKp46⁺ILC3s (Figure 6). Although preliminary, these findings suggest that these ILC subsets might play a role in immune checkpoint blockade treatment responses, particularly in regard to the efficacy of the combination of anti-PD-1 + anti-LAG-3 antibodies in melanoma.⁹⁵ While many subsets have been previously reported to express LAG-3,⁹⁶ to the best of our knowledge, this is the first time that LAG-3 expression was described on tumor-infiltrating ILCs other than NK cells or ILC1.^37,97 It now remains to determine how anti-LAG-3 antibody treatment impacts circulating and tumor-infiltrating ILC frequencies, phenotypes, and functions. Importantly, potential correlations between baseline ILC frequencies and clinical responses following anti-LAG-3 antibodies remain to be assessed.

This current study was designed as an exploratory analysis of ILC infiltration in melanoma and epithelial ovarian tumors, with the main objective to characterize differences in phenotype and TME composition between the two tumor types. Although, caution must be applied given the heterogeneity and the small sample size of the cohorts, this study is the first step toward enhancing our understanding of TME composition related to ILC infiltration, revealing particular phenotypic traits. In conclusion, this study has revealed key changes in tumor-infiltrating ILC subsets and unraveled melanoma and epithelial ovarian tumor-specific features. Notably, we found high PD-1 expression on lymphoid cells infiltrating epithelial ovarian tumors, while melanoma-infiltrating ILCs displayed high levels of LAG-3 expression. Further investigations involving larger cohorts are thus needed to confirm these findings and determine the prognostic value and potential clinical utility of these innate immune cells.

Authors’ contributions

DCC, MG, KW, AS performed the experiments, participated in data analysis and interpretation, reviewed and edited the manuscript; SDS, MQB, BAC, PAS, MOB, AE, BXW, LN provided patient samples and clinical information, reviewed and edited the manuscript; SM participated in the bioinformatics analysis and interpretation, reviewed and edited the manuscript. PSO participated in data analysis and interpretation, reviewed and edited the manuscript, provided funding and supervision; NJ performed data analysis and interpretation, provided funding and supervision, wrote the initial draft, reviewed and edited the manuscript.

Acknowledgment

We thank the members of the Ohashi and Jacquelot laboratories for their insightful comments and discussion. We thank Dr. Heewon Seo for his help with the TCGA analysis and correlation plots. We thank Dr. Danny Ghazarian for his help with sample collection. We thank the Princess Margaret Cancer Centre flow facility for their technical support.

Disclosure statement

Dr. Pamela Ohashi is on Scientific Advisory Boards for Providence Therapeutics, Treadwell Therapeutics, Tikva Allocell and Rondo Therapeutics. Inc. Dr. Ohashi holds a sponsored research agreement with Providence Therapeutics. No potential conflict of interest was reported by the other authors.

Data availability statement

The authors confirm that the data supporting the findings of this study are available within the article and its supplementary materials.

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/2162402X.2024.2349347

Additional information

Funding

This work was supported by a CIHR foundation award (CIHR FDN #143220 to PSO), a CHIR operating grant (to PSO), the Wolfond Immunotherapy Fund (to PSO), the Alberta Cancer Foundation/Arnie Charbonneau Cancer Institute laboratory start-up package (to NJ), a Canadian Cancer Society Emerging Scholar Research Grant (grant #708072 to NJ), and the Dr. Robert C. Westbury Fund for Melanoma Research (to NJ). The results published here are in part based upon data generated by the TCGA Research Network (https://www.cancer.gov/tcga/).