The tumor microenvironment of benign and malignant salivary gland tumors

Treatment of salivary gland tumors (SGTs) remains challenging. Little is known about the immune landscape of SGTs. We aimed to characterize the tumor microenvironment in benign and malignant SGTs.

challenge for standardization of treatment.Upfront surgical resection followed by adjuvant radiation for adverse pathologic features is the mainstay of treatment.Unfortunately, locoregional and metastatic disease are common, and systemic therapies have a modest response at best. 1,2The introduction of immunotherapy has begun to transform the landscape of treatment for head and neck cancer, though its role remains unclear in the treatment of SGC. 3,4Elucidating the immune profile of SGCs is essential to understanding and improving response to these new therapies.
Little is known about the tumor microenvironment of SGCs.Most studies have solely focused on the PD-1/PD-L1 pathway, as there are several immune checkpoint inhibitors that target this pathway.Gerdabi et al. performed immunohistochemistry on both benign and malignant salivary gland tumors (SGTs) and found that malignant tumors had increased expression of PD1/PD-L1 compared to benign tumors. 5Furthermore, expression of PD-L1 on tumor cells has been shown to have prognostic significance. 6However, there is a paucity of data analyzing the TME beyond PD-1/PD-L1.One study by Linxweiler and colleagues performed a comprehensive characterization of the immune microenvironment of salivary gland cancers, specifically adenoid cystic carcinoma (ACC), myoepithelial carcinoma (MECA), and salivary duct carcinoma (SDC). 7In this study, the authors performed singlecell RNA sequencing in order to understand the immune infiltration in salivary gland cancers and whole exome sequencing to assist in calculating the tumor mutational burden (TMB).Among their findings, they demonstrated that SDC had much higher levels of immune infiltration and TMB compared to ACC and MECA, which is consistent with prior work. 8These findings suggest that immune checkpoint blockade may be more effective in tumors with high immune infiltration compared to other SGCs.
Only more recently have research technologies become sufficiently mature to survey the diversity of immune cell subsets and their behaviors.Single cell analysis by mass cytometry by time-of-flight (CyTOF) is one of these techniques that uniquely enables investigation and quantification of cell identity and behavior. 9It is a highly multiplexed format for flow cytometry that allows the quantification of 40-50 unique proteins in each of millions of single cells.1][12][13] Utilizing this technology can address a critical gap in our current knowledge in understanding the immune landscape in salivary gland tumors.
There remains an unmet need to further our understanding of this heterogeneous group of cancers in order to improve our multimodal approach to treatment.Thus, the aim of the current study was to characterize the tumor microenvironment in benign and malignant salivary gland tumors utilizing CyTOF.

| Patient samples and collection
Subjects were enrolled using a consecutive sampling approach and provided informed consent under UCSF IRB approved protocols (UCSF IRB# 14-15342) for collection of tumor specimens on the day of their surgery from 01/2017-12/2019.Any patient undergoing surgery for curative intent was eligible for enrollment.Tissue samples were obtained from tumor resection specimens on the day of surgery by UCSF Pathology Assistants.Tissue specimens were placed in ice cold Leibovitz's L-15 medium in a 50-mL conical tube and immediately transported on ice to the laboratory for preparation for either mass cytometry.All clinical data were collected by chart review and stored in a secure Redcap database.

| Tumor processing for mass cytometry
Fresh tumor samples were finely minced and digested in Leibovitz's L-15 medium with 800 U/mL collagenase IV and 0.1 mg/mL DNase I with gentle agitation for 45 min at 37 C.After digestion, cells were filtered through a 70 μm filter into PBS/5 mM EDTA solution, spun down at 500g for 5 min at 4 C, the supernatant aspirated, and resuspended in fresh PBS/EDTA solution and kept on ice.
Tumor cells were then washed with PBS/EDTA and re-suspended 1:1 with PBS/EDTA and 50 μM cisplatin for 60 s at room temperature before quenching 1:1 with PBS/5 mM EDTA/0.5% BSA to determine viability as previously described. 14Cells were centrifuged at 500g for 5 min at 4 C and re-suspended in PBS/EDTA/BSA at a density between 1 Â 10 6 and 10 Â 10 6 cells per mL.Suspensions were fixed for 10 min at room temperature using 1.6% paraformaldehyde (PFA) and frozen at À80 C until ready to be run for CyTOF.

| Antibody heavy metal conjugation for mass cytometry
The sources for all mass cytometry antibodies can be found in Table S1, Supporting Information.Antibodies were conjugated to their associated metals with MaxPar X8 labeling reagent kits according to manufacturer instructions, diluted with Candor PBS Antibody Stabilization solution supplemented with 0.02% sodium azide, and filtered through an UltrafreeMC 0.1-mm centrifugation filter (Millipore) before storage at 4 C. Surface and intracellular master antibody cocktails were made and kept at À80 C.

| Mass-tag cellular barcoding for mass cytometry
Prior to antibody staining, mass tag cellular barcoding of prepared samples was performed by incubating cells with distinct combinations of isotopically purified palladium ions chelated by isothiocyanobenzyl-EDTA as previously described. 15After counting, 1 Â 10 6 cells were barcoded with distinct combinations of stable Pd isotopes for 15 min at room temperature on a shaker in Maxpar Barcode Perm Buffer.Cells were washed twice with cell staining media (PBS with 0.5% BSA and 0.02% NaN3), and pooled into a single 15 mL tube for subsequent staining and washing steps.

| Mass cytometry staining
Barcoded cells were stained with Human TruStain FcX Receptor Blocking Solution at 20 mg/mL for 5 min at RT on a shaker.Surface antibody cocktail was then added with a 500 μL final reaction volume for 30 min at RT on a shaker.Following staining, cells were washed twice with cell staining media.Before intracellular staining, cells were permeabilized for 10 min with methanol at 4 C. Methanol was then removed by washing the cells 2 times with cell staining media.The intracellular cocktail was then added to the cells for a 500 μL final reaction volume for 1 h at RT on a shaker.Cells were washed twice in cell staining media to remove unbound antibodies and then stained with 1 mL of 1:4000 191/193Ir Cell-ID Intercalator Solution diluted in PBS with 4% PFA overnight.Before mass cytometry was run, cells were washed once with cell staining media, and twice with Cell Acquisition Solution.

| Mass cytometry data acquisition
Mass cytometry samples were diluted in Cell Acquisition Solution containing bead standards to approximately 1 Â 10 6 cells/mL and then analyzed on a Helios mass cytometer (Fluidigm) equilibrated with Cell Acquisition Solution.A minimum of 10 Â 10 6 cell events were collected for each barcoded set of samples at an event rate of 400-500 events/s.

| Mass cytometry pre-processing and batch normalization
Each group of barcoded samples was run with a control sample (replicates of a single normal human tonsil) to validate staining and for normalization between groups of barcoded samples.Bead standard data normalization and de-barcoding of the pooled samples into their respective conditions was performed using the R package Premessa from the PICI institute available at https://github.com/ParkerICI/premessa.All manually gated live, intact, single cells were downloaded as FCS files from CellEngine (CellCarta, Montreal, Canada).CytoNorm 16 (https:// github.com/saeyslab/CytoNorm)was utilized to correct for batch effects.All markers were used for batch effect normalization.

| Statistical analysis
All statistical tests were performed in R using the nonparametric Wilcoxon rank sum test.For multiple testing corrections, we applied Benjamini-Hochberg correction and statistical differences were declared significant at FDR < 0.1.The R packages dply 24 and reshape 25 were used for data manipulation.Plots were produced with ggplot2. 26| RESULTS

| Patient cohort
A total of 20 patients were included in this study (Table 1).There were 11/20 benign tumors and 9/20 malignant tumors.The majority of benign tumors were pleomorphic adenomas (73%).Of the malignant tumors, 6/9 were high-grade salivary gland cancers, including salivary duct carcinoma (n = 3), high-grade mucoepidermoid carcinoma (n = 1), and adenocarcinoma (n = 2).Of the malignant tumors, 7/9 had persistent and/or recurrent disease.Two of nine patients were lost to follow-up but had no evidence of disease at their last visit.Among those with persistence/ recurrence, three patients were treated with immunotherapy.One patient with adenocarcinoma died within the week following initiation of treatment.However, the remaining two patients (one SDC and one adenocarcinoma) are still alive more than 4 years after diagnosis of their recurrence.

| Immune cell populations in salivary gland tumors
After employing our manual gating strategy to classify immune cell populations which has been previously published, 13 we found that benign and malignant salivary gland tumors have variable immunologic profiles (Figure 1A).Among the benign tumors, Warthins tumors appear immunologically similar across different patient samples, while pleomorphic adenomas demonstrate wide variation.Malignant salivary gland tumors also demonstrate wide variation across all patient samples, though high-grade malignancies appear more similar to one another compared to low-grade malignancies.The composition of different immune cell phenotypes in these salivary gland tumors is highlighted in Figure 1B, with the malignant tumors classified into high-grade and lowgrade salivary gland malignancies.Consistent with the known histologic and cytologic appearance, the Warthins tumors in this study showed a high percentage of lymphocytic infiltration, with the majority of the immune infiltrate consisting of B cells, CD4+ T cells, and CD8+ T cells.Using differential expression analysis, we then compared the frequency of each immune cell population in benign versus malignant salivary gland tumors (Figure 1C).After correction for multiple testing, only conventional type-2 dendritic cells (cDC2) were different between the two populations (Figure 1C,D).We also examined the frequency of CD4+ and CD8+ T cells given their relevance to immunotherapy (Figure 1D).Overall, there were greater percentages of both CD4+ and CD8+ T cells in salivary gland cancers compared to benign salivary gland tumors, though statistical significance was no longer met after correction for multiple hypothesis testing.

| Subpopulation of CD8+ T cells in salivary gland tumors
As CD8+ T cells play an essential role in antitumor immunity and immunotherapy, we decided to perform a more in-depth analysis of the subpopulations of CD8+ T T A B L E 1 Demographic and clinical data for entire cohort (N = 20), and data for malignant tumors only (N = 9).

Cohort characteristic
Total cohort (N = 20)  cells in all salivary gland tumors.Overall, there was a trend towards increased CD8+ T cells among high-grade compared to low-grade malignancies (Figure 2).We clustered 50 397 CD8+ T cells identified from our manual gating strategy, which can be visualized as a Uniform Manifold Approximation and Projection (UMAP) dimensionality reduction plot in Figure 3A.We found that a diverse array of CD8+ T cells are seen among all salivary gland tumors.When comparing benign versus high-grade versus low-grade tumors (Figure 3A), although there was some overlap, cells from high-grade malignancies tended to exhibit unique phenotypes compared to those from low-grade/benign tumors.This can be further visualized by tumor pathology (Figure 3A).Although there is a wide variation of different CD8+ T-cell populations within each tumor histology, interestingly, the high-grade salivary gland malignancies tended to cluster more closely.Specifically, clusters 14 and 19 seemed to be most represented in adenocarcinoma patients.Although cluster 4 appears unique to SDCs, the majority of these cells are from a single patient.We also looked at PD1 expression in CD8+ T-cell clusters and PD-L1 expression in myeloid cells and tumor cells, which also demonstrated wide variation in expression among benign salivary gland tumors, low-grade SGCs, and high-grade SGC (Figure S1).The subsequent heatmap (Figure 3B) allowed us to characterize each cluster.We identified naïve (clusters ).Using differential abundance analysis, we compared the frequency of each cluster between high-grade (n = 5) versus low-grade (n = 2) malignant tumors (Figure S2).With correction for multiple testing, although no statistically significant differences were found, there were some clusters that trended towards being different.In particular, cluster 19 ( p = 0.09 corrected), represents CD39+ CD103+ PD-1+ T cells, consistent with a terminally exhausted phenotype (Figure 3C).The frequency of this exhausted CD8+ T-cell population tended to be greater in high-grade malignancies compared to low-grade ones.These trends suggest that there may be important inflammatory differences in CD8+ T-cell populations which may make certain tumor types more responsive to immunotherapies.

| Subpopulation of CD8+ and CD4+ T cells in salivary duct carcinoma versus other salivary gland malignancies
Given the unique molecular and immunologic features of SDCs seen in prior studies, 7,27,28 we compared CD8+ and CD4+ T cells in SDCs to the other salivary gland malignancies (Figure S4a,b, respectively).Although the SDCs tended to cluster together in the UMAP for both CD8+ T cells (Figure 3A) and CD4+ T cells (Figure 4A), no significant differences or trends were seen in the various subpopulations of T cells, perhaps due to the small sample size.

| DISCUSSION
In this study, we describe the immune landscape of benign and malignant salivary gland cancers at the single-cell level using mass cytometry by time-of-flight.We demonstrated that there is wide variation in immune cell phenotype among all benign and malignant SGTs, with a trend towards greater frequency of CD8+ and CD4+ T cells among malignant tumors.We found that, when comparing high-grade versus low-grade SGCs, malignant tumors trended towards a greater number of exhausted CD8+ CD39+ CD103+ PD-1+ T cells and greater number of FoxP3+ CD4+ Tregs.They also exhibited a trend towards increased CD69+ CD45RA-CCR7À CD4+ T cells that could potentially represent a subset of FoxP3À Tregs that have been identified in several murine cell lines, 29 though the functional capacity of these cells would need to be further investigated to draw strong conclusions about their activity.
Most of the literature has described the immune infiltration in salivary gland cancers utilizing immunohistochemistry (IHC).Although the immune environment is diverse, the majority of studies have focused on PD-L1 expression on tumor cells or tumor infiltrating lymphocytes or PD-1 on tumor-infiltrating lymphocytes. 6,28,30his has been of particular interest, as it is a therapeutic target for immunotherapies.There is no standard cut-off for positive versus negative expression of PD-L1 by IHC, and several different scoring systems have been used including the Tumor progression score (TPS), which only evaluates staining on viable tumor cells, the Combined positive score (CPS), which includes PD-L1 staining of tumor cells, lymphocytes and macrophages, and the Immune cell score, which evaluates PD-L1 scoring on immune cells. 4,6,31,32Thus, the expression of PD-L1 in SGCs has been highly variable ranging from 17% to 53%, but studies have consistently shown that SGCs expressing PD-L1 are associated with poorer clinical outcomes. 6,31,33nterestingly, one study examined PD-1 expression of tumor-infiltrating lymphocytes and found that higher expression did not reflect disease severity. 34Although this was not the focus of our current study, we saw a wide variation in CD8+ T-cell PD-1 expression among benign and malignant SGC cells, though this did appear to be more common in high-grade malignancies.Similarly, we saw a wide variation in myeloid and tumor cell PD-L1 expression, though the overall expression levels also appeared to be more common in high-grade malignancies.Other studies have also used IHC to evaluate immune cell infiltration into SGCs, and have observed differences among various malignancies.One study of 94 patients with SGCs demonstrated that adenocarcinoma had an elevated number of CD3+ T cells, as well as increased TP and CP scores, compared to adenoid cystic carcinoma (ACC), mucoepidermoid carcinoma, or acinic cell carcinoma. 32Characterizing these differences in the immune environment may provide important information for targeted therapy with immune checkpoint inhibitors, especially given the high heterogeneity in salivary gland cancers.
There are far fewer studies that have utilized methodologies beyond IHC to characterize the immune landscape for salivary gland tumors.Among the prior work that has been done, our findings have several parallels regarding the tumor microenvironment in salivary gland tumors.A study by Haghshenas and colleagues evaluated CD4+ T-cell populations in benign versus malignant SGTs using flow cytometry. 35They demonstrated a higher frequency of CD4+ CD25+ Foxp3+ Tregs in blood samples of patients with malignant tumors compared to both benign tumors and controls.The authors suggest this may contribute to development and progression of malignant salivary gland tumors.The data from our study supports the idea that an increased frequency of Tregs may contribute to a more aggressive pathology, as we demonstrate a trend towards greater number of Tregs in high-grade malignancies.Given these patterns, furthering our understanding of CD4+ T-cell subsets warrants investigation, as it may have implications for response to immunotherapy or other targeted therapies.
The majority of the high-grade malignancies in the current study were either SDCs or adenocarcinomas.Although we cannot directly compare our cohort to other studies given the heterogeneity of pathologies represented and differences in technology employed (CyTOF in this study), our data are consistent with prior work that salivary gland malignancies are immunologically diverse, and that certain histologic types of high-grade salivary gland malignancies tend to have higher levels of T-cell dysfunction compared to other histologies and are immunologically unique. 7,36For example, Linxweiler et al. demonstrated that SDCs were highly T-cell infiltrated tumors, while adenoid cystic carcinomas (ACCs) were characterized by a more immune-depleted environment, and myoepithelial carcinoma was intermediate. 7hese differences may help to explain why there is inconsistency in which patients benefit from immunotherapies. 7Further, given the rarity of SGCs, it has been difficult to compare specific response rates of immunotherapy among the varying histologic subtypes, which is true in our study as only patients with high-grade malignancies received immunotherapy.Interestingly, in a recent phase II clinical trial, Vos and colleagues treated patients with recurrent/metastatic SGCs with nivolumab (anti-PD-1) and ipilimumab (anti-CTLA-4). 27The authors found that the overall response rate by RECIST criteria was 6% in patients with ACC versus 16% in patients with non-ACC SGC, and specifically was the highest at 25% in patients with SDCs.Pretreatment tumor biopsies were analyzed using RNA-seq, and demonstrated that at baseline, SDCs tended to have the highest degree of immune infiltration by multiple immune signatures, such Immu-neScore and immune cytolytic activity (CYT), 37,38 while ACCs tended to have the lowest, with MECA, mucoepidermoid carcinoma and other histologies were somewhere in between.The authors demonstrates that non-ACC tumors clustered in greater T-cell infiltrated subgroups as well.Although our cohort had different SGC subtypes compared to these studies, our data also demonstrated that SDCs, for example, cluster closely together, suggesting they have a unique tumor microenvironment.Taken together with prior studies, this suggests that the immune landscape in higher-grade salivary gland cancers may make them more responsive to immune checkpoint blockade.
Our study was limited by a small sample size, especially given the wide variation in salivary gland tumor histology.With correction for multiple hypothesis testing, we were only able to discuss trends in our data between high-grade and low-grade salivary gland cancers.Despite these limitations, we were able to hypothesize possible immunologic reasons why high-grade malignancies may be more clinically aggressive.This data can inform future studies, especially as new immunologic technologies such as multiplexed beam ion beam imaging (MIBI) can utilize formalin-fixed tissue specimens, 13,39,40 which could allow for larger studies of rare cancers.
In conclusion, this study characterizes the immunologic landscape for both benign and malignant salivary gland tumors utilizing CyTOF, a technique that enables analysis at the single-cell level.Our data shows that there is wide variability in the immune profile of salivary gland tumors.When grouped into high-grade versus low-grade salivary gland malignancies, there are important trends within the T-cell subpopulations that suggest these two groups likely have different, unique immunologic features.These data provide new hypotheses about this rare, heterogeneous group of cancers, and may help guide future investigations to better understand the immunologic landscape, which has implications for how to employ immunotherapies.

3 F I G U R E 1
n o c a rc in o m a -8 0 2 6 A d e n o c a rc in o m a -8 3 8 6 A d e n o c a rc in o m a -9 9 8 4 M u c o e p .c a rc in o m a -1 c a rc in o m a -9 5 5 7 M y o e p it h e li a l c a rc in o m a -8 7 4 Legend on next page.

3
Legend on next page.

F
I G U R E 3 Overview of CD8+ T-cell subpopulations.(A) UMAP of CD8+ T-cell clusters, colored by cluster (top left panel), benign tumors versus high-grade versus low-grade malignancy (top right panel), or pathology (bottom left panel).(B) Heatmap of markers used for CD8+ T-cell clustering.Scaled median expression per marker is shown for cluster annotation.(C) Cluster 19 abundance (as a percentage of CD8+ T cells) in high-grade versus low-grade salivary gland malignancies.p-values obtained by Wilcoxon rank-sum test (before Benjamini-Hochberg correction).[Color figure can be viewed at wileyonlinelibrary.com] cell clusters in high-grade vs. low-grade salivary gland malignancies F I G U R E 4 Legend on next page.

F
I G U R E 4 Overview of CD4+ T-cell populations.(A) UMAP of CD4+ T-cell clusters, colored by cluster (top left panel), benign tumors versus high-grade versus low-grade malignancy (top right panel), or pathology (bottom left panel).(B) Heatmap of markers used for CD4+ T-cell clustering.Scaled median expression per marker is shown for cluster annotation.(C) Cluster 10 and 12 abundances (as a percentage of CD4+ T cells) in high-grade versus low-grade salivary gland malignancies.p-values obtained by Wilcoxon rank-sum test (before Benjamini-Hochberg correction).[Color figure can be viewed at wileyonlinelibrary.com]