Characterization of immune exhaustion and suppression in the tumor microenvironment of splenic marginal zone lymphoma

The role of the tumor microenvironment (TME) and intratumoral T cells in splenic marginal zone lymphoma (sMZL) is largely unknown. In the present study, we evaluated 36 sMZL spleen specimens by single cell analysis to gain a better understanding of the TME in sMZL. Using mass cytometry (CyTOF), we observed that the TME in sMZL is distinct from that of control non-malignant reactive spleen (rSP). We found that the number of TFH cells varied greatly in sMZL, ICOS+ TFH cells were more abundant in sMZL than rSP, and TFH cells positively correlated with increased numbers of memory B cells. Treg cell analysis revealed that TIGIT+ Treg cells are enriched in sMZL and correlate with suppression of TH17 and TH22 cells. Intratumoral CD8+ T cells were comprised of subsets of short-lived, exhausted and late-stage differentiated cells, thereby functionally impaired. We observed that T-cell exhaustion was present in sMZL and TIM-3 expression on PD-1low cells identified cells with severe immune dysfunction. Gene expression profiling by CITE-seq analysis validated this finding. Taken together, our data suggest that the TME as a whole, and T-cell population specifically, are heterogenous in sMZL and immune exhaustion is one of the major factors impairing T-cell function.


INTRODUCTION
Despite the heterogeneous biology of different lymphoma subtypes, it is well recognized that the tumor microenvironment (TME) plays an important role in the support and survival of lymphomas by affecting antitumor immunity through modulation of the interaction between the lymphoma cells and immune cells.
Over the past few years, this understanding has dramatically changed the field of immunological treatment for lymphoma with the introduction of checkpoint inhibitors and T-cell based therapies, such as chimeric antigen receptor T cells and bispecific antibodies, all of which rely on the immunological fitness of the T cells. Checkpoint inhibitors are antibodies that block inhibitory receptors in order to restore the function of effector T-cells in the TME. Remarkable results have been reported in some lymphoma types, particularly classical Hodgkin lymphoma, with objective response rates (ORRs) of approximately 70% [1,2]. However, responses vary substantially between different histologies and remain suboptimal in low grade non-Hodgkin lymphomas, such as follicular lymphoma (FL) and marginal zone lymphoma (MZL), with ORRs of around 10% [3][4][5]. The reason for this substantial difference in clinical benefit, however, is not well understood.
One possible reason for differences in response to immunological therapy in indolent lymphomas may lie in the composition of the TME. The TME of low grade lymphomas comprises different subsets of T cells with diverse biological functions. For example, CD4 + CD25 + T reg cells possess inhibitory functions affecting other T-cell subsets, while PD-1 high T FH cells support lymphoma cell survival and growth. IL-17-producing cells are a subset of T H cells that potentially enhance inflammation in the TME, whereas T reg cells exert an immunosuppressive role [6,7]. In addition, senescent and anergic T-cell subsets exist in low grade lymphoma and correlate with unfavorable clinical outcomes [8]. Furthermore, exhausted T cells are present in low grade lymphomas with reduced biological functions such as proliferation, cytokine production and cytotoxicity [9][10][11]. While these studies have focused on specific cellular subsets in the TME, there is a paucity of studies that have comprehensively explored the TME to study the intratumormal T cells in MZL, especially splenic MZL (sMZL). One recent study utilized a multi-platform approach to analyze sMZL samples and identified 2 basic types of TMEs that were termed immune-suppressive and immune-silent [12]. The specific cell subtypes that accounted for the subtypes were inferred by deconvolution and results highlighted the additional need to further identify the phenotype of intratumoral immune cells that account for immune exhaustion and dysfunction in sMZL.
T-cell exhaustion describes a type of immune response that stems from chronic antigen stimulation in which T cells expressing variable inhibitory receptors exhibit reduced differentiation, proliferation, and effector function [13][14][15][16]. PD-1 was the primary molecule identified as a marker of exhausted T cells, however, subsequent studies identified additional inhibitory receptors, such as TIM-3, LAG-3, and TIGIT that either induce T-cell exhaustion or render exhausted T-cells more dysfunctional [17][18][19][20][21][22][23]. Our lab has previously shown that immune dysfunction is due in part to the T-cell exhaustion that is present in the TME of indolent non-Hodgkin lymphoma (NHL) and can explain the inferior outcome of a subset of patients [10,11,24,25]. Furthermore, the abovementioned study in sMZL identified upregulation of genes associated with immune checkpoint activation as a feature of the immune-suppressive phenotype [12].
In this study, we performed a comprehensive analysis of a cohort of 36 patients with sMZL by mass cytometry (CyTOF) and CITE-seq. We analyzed T FH , T reg and CD8 cells and determined their phenotype and contribution to immune dysfunction. We specifically focused on exhausted T cells and compared their different subsets, which were defined based on transcriptomic and proteomic data as well as data on their biological function, in order to determine different degrees of immune dysfunction.

Patient samples
Patients with a tissue biopsy that showed MZL on pathologic review and adequate splenic tissue in the lymphoma biospecimen bank were included in this study. All spleen tissues were obtained from splenectomy specimens. Patient must have provided written informed consent to be eligible. All included patients were treated at Mayo Clinic between June 2010 and June 2019. The study was approved by the Institutional Review Board of the Mayo Clinic/Mayo Foundation. Patient characteristics, including age, sex, and health status, are summarized in Supplementary Table 1. Spleen tissue from adult patients without lymphoma was used as controls.
Mononuclear cells were isolated from biopsy specimens of patients or healthy donors using centrifugation over Ficoll Paque solution. T cells were isolated by either microbeads or flow sorter and subjected to primary culture for assays. For functional assays, T cells were stimulated with CD3/ CD28 beads and cultured in RPMI 1640 medium containing 10% fetal bovine serum and 1% penicillin-streptomycin (10,000 U/mL and 10,000 μg/ mL, respectively) in an incubator containing 5% CO 2 at 37°C. T-cells were also stimulated by PMA/ionomycin in the presence of brefeldin for 4 h prior to staining.

Cell isolation and purification
Tissue biopsy specimens from sMZL tumors and rSP were gently minced over a wire mesh screen to obtain a cell suspension. The cell suspension was centrifuged over Ficoll Paque at 500 × g for 15 min to isolate mononuclear cells. CD3 + T-cells were isolated using positive selection with CD3 microbeads kit (StemCell Technologies, Vancouver, Canada).

Mass cytometry (CyTOF)
CyTOF was performed according to the manufacturer's instruction. Briefly, 3 × 10 6 cells were stained with 5 mM Cell-ID™ Cisplatin (Fluidigm, San Francisco, CA) for 5 min and quenched with MaxPal Cell Staining Buffer (Fluidigm) using 5 times the volume of the cell suspension. After centrifugation, cell suspensions (50 μl) were incubated with 5 μL of human Fc-receptor Blocking solution (Biolegend, San Diego, CA) for 10 min and 50 μL of pre-mixed antibody cocktail (Supplementary Table 2) for 30 min. After washing, cells were incubated with 1 ml of cell intercalation solution (125 nM MaxPal Intercalator-Ir into 1 ml MaxPal Fix and Pem Buffer, Fluidigm) overnight at 4°C. Cells were centrifuged with MaxPal Water and pelleted. The pelleted cells were suspended with EQ Calibration Beads (Fluidigm) and cell events were acquired by a CyTOF II instrument (Fluidigm).

CyTOF data analysis
CyTOF data were analyzed using online software Cytobank [26] and R program [27]. All samples were normalized and analyzed simultaneously to account for variability in signal across acquisition times. Assessment of data quality was conducted and summarized in Supplementary Table 3. A high-level gating strategy was applied simultaneously to all CyTOF files ( Supplementary Fig. 1). Files were concatenated to one file for specific analysis and the tSNE platform was used to analyze the data. Lineage markers (CD4, CD8, CD25, CD127, PD-1, CCR7, or CD45RO) were included in  the analysis to distinguish T subsets, T reg , T FH , exhausted cells, naive or  memory T cells. A tSNE plot was generated by t-Distribution Stochastic Neighbor Embedding (tSNE) analysis in order to make pairwise comparisons of cellular phenotypes, to optimally plot similar cells close to each other and to reduce multiple parameters into two dimensions (tSNE1 and tSNE2). Equal events per sample were selected for most analyses. Channel selection depended on cell populations to be clustered. Standard tSNE parameters (2000 iterations, perplexity of 30 and theta of 0.5) were used. For CITRUS (cluster identification, characterization, and regression) analysis, Significance Analysis of Microarrays (SAM) -Correlative and Abundance was selected for Association Models and Cluster Characterization, respectively, with the minimal cluster size being 1%. Patients were divided into two groups based on clinical parameters (for example, patients dead or alive at last follow up).  Table 4) for 30 min at 4°C. Cells were washed three times in cell staining buffer, followed by centrifugation (350 g 5 min at 4°C). After the final wash, cells were resuspended at appropriate cell concentrations (700-1200 cells/µL, viability >90%) in calcium and magnesium-free 1 × PBS (Corning, USA) containing 0.04% BSA (Thermo Fisher Scientific, USA) and run by 10x Genomics applications. The cells were first counted and measured for viability using the Vi-Cell XR Cell Viability Analyzer (Beckman-Coulter). The barcoded Gel Beads were thawed from −80°C and the cDNA master mix was prepared according to the manufacturer's instructions for Chromium Single Cell 5' Library and Gel Bead Kit (10x Genomics). Based on the desired number of cells to be captured for each sample, a volume of live cells was mixed with the cDNA master mix. The cell suspension and master mix, thawed Gel Beads and partitioning oil were added to a Chromium Single Cell A chip. The filled chip was loaded into the Chromium Controller, where each sample was processed and the individual cells within the sample were partitioned into uniquely labeled GEMs (Gel Beads-In-Emulsion). The GEMs were collected from the chip and taken to the bench for reverse transcription, GEM dissolution, and cDNA clean-up. The full-length cDNA was amplified and separated by size selection. The resulting cDNA created a pool of uniquely barcoded molecules used to generate 5' gene expression libraries (GEX). In addition, the supernatant from the cDNA clean-up step contained amplified DNA from cell surface protein feature barcodes. That DNA was further cleaned and used to create cell surface protein libraries. During library construction, standard Illumina sequencing primers and a unique i7 Sample index (10x Genomics) were added to each cDNA and DNA pool (creating gene expression and feature barcodes libraries, respectively). All cDNA and DNA pools and resulting libraries were measured using Qubit High Sensitivity assays (Thermo Fisher Scientific) and Agilent Bioanalyzer High Sensitivity chips (Agilent).

CITE-seq analysis
Gene expression libraries (GEX) were sequenced at a minimum of 50,000 fragment reads per cell and feature barcodes libraries were sequenced at 5000 fragment reads per cell. Sequencing steps followed Illumina's standard protocol using the Illumina cBot and HiSeq 3000/4000 PE Cluster Kit. For gene expression libraries, the flow cells were sequenced as 100 × 2 paired end reads on an Illumina HiSeq 4000 using HiSeq 3000/ 4000 sequencing kit and HCS v3.3.52 collection software. For feature barcodes libraries, the flow cells were sequenced as 100 × 2 paired end reads on an Illumina HiSeq 4000. Base-calling was performed using Illumina's RTA version 2.7.3.

CITE-seq data analysis
Reads were aligned to the human reference sequence GRCh38. We used Cell Ranger multi pipeline to analyze FASTQ data derived from Gene Expression data (GEX) that contains the sequence data from the clusters that pass filter on a flow cell and feature barcode (antibody) library from the same GEM Well. We used the Seurat package (v4.1.3) to perform integrated analyses of single cells.

Hyperion panel staining
Hyperion staining was performed at the Pathology Research Core (Mayo Clinic, Rochester, MN). Slides were baked in a 60°C oven for 60 min. Slides were loaded into a Leica Bond RX stainer to perform dewax/baking, heat induced antigen retrieval and blocking steps. An Epitope Retrieval 2 solution (EDTA based, Leica) was used for 20 min. Slides were blocked in Superblock solution (Thermo Fisher) for 30 min, followed by a few washes in PBS-TB. Slides were removed from the stainer and placed in PBS-TB while preparing an antibody cocktail followed by manual staining. Metal-labeled antibodies (Supplementary Table 5) were either bought directly from Fluidigm or conjugated by the Immune Monitoring Core (IMC). Antibodies were spun down at 13,000 × g for 2 min prior to preparing the antibody cocktail, then diluted in PBS-TB according to Supplementary Table 5 and incubated overnight at 4°C in a humidity chamber. Prior to adding the antibody cocktail to each slide, the tissue was encircled using a PAP pen. After overnight incubation, slides were washed three times for 5 min each in PBS-TB and incubated in Cell ID Intercalator-Ir (Fluidigm) at 1:400 for 30 min. Slides were washed in PBS-TB three times for 5 min each, followed by a rinse in Milli-Q water for 5 min. Slides were allowed to air dry and were sent to the IMC lab for imaging of region of interest (ROI). The merged images were viewed and generated using the MCD™ Viewer (Fluidigm).

Statistical analysis
Statistical analysis was performed in GraphPad Prism 7.02. Student's t test was used to compare the distributions of continuous variables when the normal distribution assumption was adequate. For variables without normal distribution, we used non-parametric Wilcoxon rank-sum test. For matched-paired data, paired t test or Wilcoxon sign-rank tests were used. OS was measured from the date of diagnosis until death from any cause and calculated using Kaplan-Meier analysis. Univariate associations between individual clinical features and survival were determined with the log-rank test. Due to the exploratory nature of this study, multiple comparison adjustment was not performed. In all cases, p < 0.05 was used to declare statistical significance.

RESULTS
The overall immune content of spleens in patients with sMZL differs from rSP To define different subsets of immune cells in sMZL, we developed and validated a mass cytometry (CyTOF) panel which integrates T-cell markers of lineage, differentiation, activation, suppression and cytokine production. These phenotypic markers have been validated by our group and others [25,28] in studying immune subsets in the TME with a focus on exhaustion and allow segregation of previously described T-cell subsets (Supplementary Table 1). Our database includes a large number of prospectively enrolled sMZL patients with biopsies at diagnosis and stratified by type of treatment. Data are available regarding event free survival (EFS) at 24 months, as well as overall survival (OS). We thus analyzed viably cryopreserved single-cell suspensions from 36 spleen biopsies from sMZL patients obtained at diagnosis as well as from 6 non-malignant spleen tissues (rSP).
We first examined the immune microenvironment of MZL as a whole by performing t-distribution stochastic neighbor embedding (tSNE) analysis and generated viSNE plots to visualize different types of immune cells, including T-cells, NK cells, macrophages and B-cells (Fig. 1A). We noted that the immune content of the TME was highly variable between different patients and significantly different compared to that of control samples. Specifically, while CD19 + B cells were significantly upregulated in sMZL, the numbers of T cells as well as NK cells and macrophages were significantly lower in sMZL patients than controls (Fig. 1B). Clustering analysis revealed that some subsets were more abundant in sMZL than rSP, or vice versa (Fig. 1C, D).
We next determined T-cell differentiation status by measuring the frequency of naïve and memory T-cell sunsets in sMZL and rSP. Overall, the vast majority of T cells in sMZL were of memory type and accounted for 91.7% of CD3 + T cells (Fig. 1E). The number of T cells with a central memory phenotype (CD45RA -CCR7 + , T CM ) was significantly higher in sMZL than rSP. In contrast, sMZL tend to have less terminally-differentiated T cells (CD45RA + CCR7 -, T EMRA ) when compared to rSP. A subset of T EMRA bearing a phenotype of KLRG1 + 4-1BB + TIGIT + cells were substantially higher in rSP than sMZL (Fig. 1F). In sMZL, CD8 + T cells contained a significant higher percentage of T EMRA than CD4 + T cells (Fig. 1G). Importantly, patients with higher levels of T CM and lower levels of T EMRA as well as lower total CD8/CD4 ratio were more likely to achieve event-free survival at 24 months (EFS24) [29] (Fig. 1H). A lower CD8/CD4 ratio was also associated with better OS in sMZL (Fig. 1I).
ICOS + T FH cells are expanded and positively correlate with memory B cells in sMZL Follicular T helper (T FH ) cells are a subset of CD4 + T cells that play an important role in B-cell differentiation in sMZL [30]. Therefore, we wanted to further explore this subset by determining its phenotype and clinical significance. Previous studies have shown that T FH cells are PD-1 high in contrast to exhausted cells, which tent to be PD-1 low [10,24,30]. In sMZL, PD-1 high cells were variably present with a frequency ranging from almost zero to being the predominant subset within the CD4 + T-cell population ( Fig. 2A). in contrast, PD-1 high cells were negligible among CD8 + cells (Supplementary Fig. 2A). Among 26 sMZL patients, the number of PD-1 high cells accounted for 15.3% (range 0.5-51%) of CD4 + cells ( Fig. 2A). Phenotypically, CD4 + PD-1 high cells were characterized by a higher expression of TIGIT, ICOS, BTLA and CD69, canonical markers for T FH cells (Fig. 2B). These PD-1 high cells expressed reduced levels of CD127, CCR7, CD26 and CD45RA (Fig. 2B), suggesting that these cells constitute an advanced differentiation stage.
When measuring whether the number of PD-1 high T FH cells differed between sMZL and rSP, we found no significant difference between patients and controls. We then focused on ICOS + T FH cells, a main T FH subset in sMZL, and found that ICOS + T FH cells were significantly enriched in sMZL when compared to rSP (Fig. 2C). CITRUS analysis of CD4 + T cells, an algorithm used to identify statistically significant stratifying biological signatures within populations of individuals, revealed 4 clusters (Fig. 2D(i)) that expressed high level of PD-1 and ICOS, as well as TIGIT, BTLA, and CD69, suggestive of a T FH phenotype (Fig. 2D(ii)). These clusters were significantly more abundant in sMZL when compared to rSP (Fig. 2D(iii)), confirming the above finding. We next analyzed the transcriptome of ICOS + T FH cells in sMZL. Among 19 clusters identified in mononuclear cells from biopsy specimens of sMZL, cluster 0 showed high PD-1, ICOS and CXCR5 expression, representing ICOS + T FH cells (Fig. 2E, F). A volcano plot (Fig. 2G) revealed a number of differentially expressed genes (DEG) that were up-or down-regulated in cluster 0 compared to other clusters. The DEG included genes of heat-shock protein members HSAPA1A and HSPA6, highly conserved proteins that play a role in cellular repair and protective mechanisms. The relevance of this finding is unclear and further investigation is warranted in the future.
Given the variable frequency of PD-1 high CD4 + T cells whose primary role is to promote B cell growth, we tested whether a correlation between the number of lymphoma cells and PD-1 high CD4 + T cells existed. While we did not find a correlation between the number of total lymphoma cells and PD-1 high CD4 + T cells, ICOS + T FH cells positively correlated with CD27 + memory B cells, indicating that ICOS + T FH cells may be involved in lymphoma B cell differentiation (Fig. 2H). Of note, increased numbers of ICOS + T FH cells tended to correlate with a better OS (p = 0.05, Fig. 2I).
TIGIT + Treg cells are enriched in MZL and correlate with a suppression of T H 17 and T H 22 cells CD4 + CD25 + regulatory T (T reg ) cells play a crucial role in the regulation of tumor immunity that is clinically relevant. In the present study, we assessed the phenotype of this T reg subset and its potential impact on T helper (T H ) cells. T reg cells represented a major CD4 + subpopulation in the TME of sMZL and accounted for 22.3% (range 4.9-48.5%, n = 26) of CD4 + T cells (Fig. 3A). Increased expression of genes of canonical molecules, including FOXP3, TNFRSF18 (GITR) and IKZF2 (Helios) on CD4 + CD25 + cells compared to CD25cells validated the population of T reg cells (Fig. 3B). This subset also expressed high levels of TIGIT, ICOS and CD69 and low levels of PD-1. Of note, CD26 and CD161 were mutually exclusively expressed on T reg cells (Fig. 3C).
While there was no difference in the frequency of total T reg cells between patients and controls, we observed that TIGIT + T reg cells were significantly upregulated in sMZL when compared to rSP (Fig. 4D). Phenotypically, TIGIT + T reg cells generally differed from TIGIT -T reg cells in sMZL as UMAP plots appeared profoundly different (Fig. 3E). Particularly, TIGIT + T reg cells expressed higher level of PD-1, ICOS and BTLA than TIGIT -T reg cells (Fig. 3F). In contrast, TIGIT -T reg cells expressed higher levels of CCR7, CD127, and CD26 than TIGIT + T reg cells, suggestive of a more advanced differentiation stage.
Given the essential role of T reg cells in suppressing T H cells, we determined the correlation between T reg cells and T H subsets in sMZL. We identified T H 1, T H 2, T H 17, and T H 22 cells using a combination of surface markers (Supplementary Fig. 2B) and measured the frequency of each T H subset in sMZL and rSP. As shown in Fig. 4G, the percentage of T H 17 and T H 22 was significantly lower in sMZL than rSP while the percentage of TIGIT + T reg cells increased. Consistent with this finding, the number of TIGIT + T reg cells negatively correlated with the number of T H 17 and T H 22 cells (Fig. 4H). These results suggest that TIGIT + T reg cells may suppress T H 17 and T H 22 cells in sMZL.
CD8 + T cells in sMZL display a phenotype of late-stage differentiation CD8 + T cells play a critical role in anti-tumor immunity. In sMZL, CD8 + T cells accounted for 5.6% of CD45 + leukocytes or 32.7% of total T cells, which was slightly lower in sMZL than rSP (Fig. 4A). Intratumoral CD8 + T cells in sMZL were phenotypically heterogenous and consisted of cells of various differentiation stage. The majority of CD8 + T cells (68.1%) were memory type (CD45RA − CCR7 + T CM and CD45RA − CCR7 − T EM ) whereas naïve cells (CD45RA + CCR7 + T N ) accounted for approximately 5.3% of CD8 + T cells (Fig. 4B). Of note, terminally differentiated cells (CD45RA + CCR7 − T EMRA ) were present with substantial numbers and accounted for approximately 25.7% of CD8 + T cells (Fig. 4B). We next performed a clustering analysis to identify the cell subsets that comprise the CD8 + T population in both sMZL and rSP (Fig. 4C). We used exhaustion and differentiation markers including PD-1, TIGIT, KLRG1, CD57, and CD28/CD27 to define these clusters and observed that the major CD8 + clusters were CD27/CD28 deficient (CD27 − /CD28 − ) short-lived effector cells (SLECs, KLRG1 + CD127 − , 30.4%), exhausted (T EXH , PD-1 + TIGIT + ,) and effector cells (T EFF , PD-1 -CXCR3 + , 28.6%) that expressed CD57 and downregulated CD27 and CD28. While the overall phenotype of CD8 + T cells was not markedly different between sMZL and rSP, certain clusters were more abundant in sMZL (Fig. 4D). Specifically, CD57 + T EXH with a phenotype of T EMRA were more abundant in sMZL than rSP. In contrast, T N and memory precursor effector cells (MPECs) were enriched in rSP (Fig. 4E). Consistent with the phenotypical of findings, it appeared that CD8 + T cells from sMZL were functionally suppressed as their ability to produce cytokines was significantly decreased compared to CD8 + T cells from rSP (Fig. 4F). Importantly, intratumoral CD8 + T cells expressing CCR7, a marker of early differentiation (T N and T CM ) correlated with a favorable OS in sMZL. In contrast, CD8 + T cells expressing KLRG1 and lacking CD28, markers of late differentiation, were associated with inferior outcomes (Fig. 4G). Taken together, these results suggest that the majority of intratumoral CD8 + T cells exhibit phenotypes of an advanced differentiation stage and correlate with poor prognosis.
Phenotypically, these 4 subsets were distinct (Fig. 5D). DP cells expressed the highest levels of ICOS, TIGIT, LAG-3 and CD57 among these 4 subsets (Fig. 5E). In addition, DP cells expressed higher levels of Eomes and TOX when compared to TIM-3 + SP (Fig. 5F). These results suggest that DP cells are more immunologically exhausted than the other three subsets. Consistent with this finding, DP cells generated by an in vitro model showed reduced proliferation and cytokine production (Fig. 5G). The in vitro model was built upon our previous finding that IL-12 induces T-cell exhaustion by upregulating TIM-3 expression on T cells [10]. We stimulated cells in the presence of IL-12 to induce expression of exhausted T cells. As shown in Fig. 5G, T cells subjected to repeated TCR stimulation led to higher increase in DP whereas additional IL-12 treatment led to further expansion of DP cells (Fig. 5G). We then measured cell function and found that cell proliferation (Fig. 5H) and cytokine production (Fig. 5I) was reduced in IL-12-treated cells, confirming that DP cells are profoundly exhausted.

PD-1/TIM-3-defined subsets are transcriptomically different and DP cells exhibit gene expression profiling of exhaustion in sMZL
To gain a better understanding of the transcriptome of PD-1/TIM-3-defined subsets, we determined the gene expression profile of Fig. 1 The overall immune content of spleens in patients with sMZL differs from rSP. A The tSNE plots showing expression and distribution of CD14 + , CD16 + , CD56 + , CD19 + and CD3 + from representative sMZL and rSP biopsy specimens. B Graph showing percentage of CD14 + , CD16 + , CD56 + , CD19 + and CD3 + in CD45 + cells in sMZL (n = 26) and rSP (n = 6) biopsy specimens. C The tSNE plots showing clusters from sMZL and rSP based on gene expression profile from CITE-seq analysis. D Heatmap showing gene expression and frequency of clusters from the tSNE plot. E Dot plots from representative sMZL and rSP biopsy specimens showing expression of CD45RA and CCR7 in CD3 + cells. Graphs showing percentage of T N (CD45RA + CCR7 + ), T CM (CD45RA -CCR7 + ), T EM (CD45RA -CCR7 -) and T EMRA (CD45RA + CCR7 -) in sMZL and rSP. F The tSNE plots of CD3 + T cells showing expression of selected surface markers from concatenated files of sMZL and rSP. G The tSNE plots of CD4 + and CD8 + T cells showing expression of CD45RA and CCR7 from concatenated file of sMZL. Graphs showing percentage of T N , T CM , T EM and T EMRA in CD4 + and CD8 + T cells from sMZL. H graphs showing percentages of T N , T CM , T EM and T EMRA as well as CD3 and CD8/CD4 ratio in patients with EFS24 achieved or failed. I Kaplan-Meier curves of sMZL patients with CD8/CD4 ratio greater or less than 1 for OS of FL patients (n = 36). these 4 subsets from CD3 + T cells using CITEseq analysis. PD-1 high T FH cells were included in the analysis as a control. As shown in Fig. 6A, DP cells expressed more genes related to T-cell exhaustion such as LAG3, KLRG1, PRDM1 (BLIMP1) and TIGIT (Fig. 6B).
In contrast, genes present in T FH cells such as PDCD1, CXCR5, and TOX were highly or exclusively expressed by T FH cells.
We next performed clustering analysis to determine which clusters differed between these subsets. As shown in Fig. 6C, the tSNE plots among DP, PD-1 SP, TIM-3 SP and DN looked different whereas T FH cells were more uniform with fewer clusters compared to the 4 PD-1/TIM-3-defined subsets. Gene expression profiling from cluster 1 marked a traditional T FH phenotype with strong expression of PDCD1, TOX2 and TIGIT. Cluster 3 featuring higher expression of LAG3, KLRG1, and CST7 was more abundant in DP cells (Fig. 6D). Cluster 4 with CCR7 and SLC40A1 phenotype was more abundant in TIM-3 SP and DN than DP and PD-1 SP, suggesting cells with an earlier differentiation stage were more present in TIM-SP/DN than PD-1 SP/DP. In contrast, cluster 0 was more abundant in DP/PD-1 SP than TIM-3 SP/DN and expressed high levels of the gene for CD161, KLRB1, a C-type lectin-like receptor expressed by a subset of cytotoxic T cells. We also compared DEG between each two PD-1/TIM-3-defined subsets and confirmed that each of them exhibited a unique DEG (Fig. 6E). To further validate the gene expression profiles of the TIM-3/PD-1defined subsets, we performed CITE-seq analysis on an independent cohort of an additional 9 sMZL patients. As shown in Supplementary Fig. 3, while the transcriptomic profiling varies among each subset, DP cells expressed more exhausted genes than other subsets, consistent with the findings shown in Fig. 6B.

Spatial profiling identifies different T cell subsets residing in various areas in sMZL tissue
The precise location of immune cells is very important for their functional impact on tumor cells and on prognosis. In this study, we performed multiplex immunohistochemistry (IHC) and imaging mass cytometry (IMC) with Hyperion (Fluidigm) to gain a better understanding of the topography of intratumoral T cells in sMZL. For IHC staining, two representative patient specimens were selected from the above CyTOF cohort. As shown in Fig. 7A, T cells were variably represented in sMZL with one case showing modest CD3 staining and the other case having abundant T cells. This is consistent with the finding from CyTOF that T cells variably infiltrate the involved spleens. Further staining, using ICOS and CXCL13 as markers, confirmed that T FH cells reside within the follicles (white pulp) (Fig. 7B). For IMC, we stained a sMZL splenic tissue with 26 antibodies. As shown in Fig. 7C, we selected a white pulp area including follicular region to explore the topography of immune cells. CD11c, a marker for follicular dendritic cells, identified cells specifically located in the follicle with strong HLA-DR expression. B cell marker CD20 staining, identifying B-cells, was seen throughout the region of interest. While CD14 + monocytes/macrophages were rarely seen, cells positive for CD68 staining were present in substantially numbers. We observed that CD4 + T cells were located in the center of the follicle (white pulp), whereas CD8 + T cells resided in the marginal area and red pulp. Consistent with this, Foxp3, a marker of CD4 + T reg cells, was positive in intrafollicular cells whereas granzyme B (GzmB) staining, produced by CD8 + T cells and CD16 + NK cells, was seen in extrafollicular cells. While the majority of cells were CD45RO + , a substantial number of cells around marginal area were positive for CD45RA, suggesting a unique spatitial profile of the naïve population. Taken together, these results reveal variable topography of immune cells in sMZL and indicate that different T cells preferentially reside in distinct areas.

DISCUSSION
The TME plays a crucial role in immune responses to tumor cells, which affects the clinical outcomes of patients with cancer. However, the specific characteristics of the immune dysfunction are largely unknown and data on how ithe TME content affects patient outcomes are largely lacking [25,31]. In the present study, we explored the content of TME utilizing single cell analysis and found that the immune microenvironment was distinct between sMZL and rSP. Specifically, the frequency of some subpopulations such as TIGIT + T reg , ICOS + T FH , CD8 + CD57 + T EXH and MPECs differed between sMZL and rSP.
Analysis of the TME as a whole identified differences between the immune content of sMZL and rSP tissues. Specifically, biopsy samples from sMZL patients contained lower numbers of total T-cells, monocytes/macrophages and NK cells, but higher numbers of lymphoma B cells when compared to rSP. Clustering analysis confirmed this finding. The variability between different patients with sMZL appears to be of clinical significance, as the TME of patients who had disease progression within 2 years of diagnosis contained lower CD8/CD4 ratios and higher levels of Temra. Significant variability was also noted with regards to expression of several activation and inhibitory markers and was associated with decreased cell function. Similar heterogeneity has been previously described in the TME of other indolent NHL, such as FL [25,32]. Using gene expression profiling, a previous study characterized the TME of sMZL as immune-suppressive or immune-silent [12]. The immune-suppressive and immune-silent patterns correlated with tumor-infiltrating lymphocytes/macrophage recruitment as well as immune checkpoint activation. The immune-silent cases had a TME where a B-cell signature of tumor origin dominated. Our clustering analysis found that 5 and 3 (of 10) clusters demonstrated abundant T cells and B cells, respectively, supporting this characterization.
In sMZL, the frequency of T FH cells varied significantly among different patients and a substantial number of tumors from patients did not contain T FH cells. Previous studies have found that the T FH frequency is low in spleens and 10 times lower than that in tonsils, consistent with previous findings [30,33,34]. However, the frequency of ICOS + T FH cells was higher in sMZL than rSP, and ICOS + T FH clusters by CITRUS analysis were also more abundant in sMZL than rSP. The number of ICOS + T FH cells positively correlated with memory B cells, suggesting an involvement of ICOS + T FH cells in the regulation of lymphoma cell differentiation, an essential function for T FH cells. A prior study found that T FH cells are more often found near Ki-67 + proliferating B cells [35], which supports our finding.
CD4 + CD25 + T reg cells that expressed concanical markers such as Foxp3, GITR and Helios represented a major population and accounted for approximately 20% of CD4 + T cells in sMZL. T reg cells also expressed high level of TIGIT, ICOS and CD69 as well as PD-1, CD26 and CD161; the latter two markers have been comprehensively studied in sMZL [36]. In the present study, we found that TIGIT + T reg cells are enriched in sMZL when compared to rSP and have a very distinctive phenotype compared to TIGIT -T reg cells. The finding that TIGIT + T reg cells negatively correlate with T H 17 and T H 22 cells suggests that TIGIT + T reg cells are functional and suppressive. This imbalance between T reg cells and T H 17 or T H 22 has also been observed in other hematologic malignancies [7,[37][38][39].
When analyzing intratumoral CD8 + T cells in sMZL, we found that the subset of terminally differentiated memory cells are substantially increased while naïve T cells are almost negligible within the CD8 + compartment. Consistent with this, clustering analysis found that CD8 + T cells consist of subsets with short-lived, exhausted and of advanced differentiation stage cells, with impaired function. While there was no difference in the frequency of total CD8 + cells between sMZL and rSP, subsets of CD8 + T cells, such as SLEC and T EMRA , were more abundant in sMZL than rSP. Together, these results suggest the presence of terminal CD8 + T-cell differentiation in sMZL. This distinct phenotype of CD8 + cells may be trigged by the inflammatory TME in sMZL. We observed that CD8 + T cells in an advanced stage of differentiation are associated with an inferior outcome and failed to achieve EFS24, a finding that we also observed in follicular lymphoma [25]. In addition, it has been shown that a low CD4/CD8 ratio correlates with overall and event-free survival in gastric diffuse large B-cell lymphoma [40], AIDS-related lymphoma [41], findings consistent with our observation in sMZL.
One of the main focuses in this study was to assess T-cell exhaustion, given that approximately half of T cells in sMZL express low levels of PD-1, a hallmark for T-cell exhaustion [42]. Some PD-1 low T cells coexpress TIM-3, forming 4 subsets: (DP, PD-1 + SP, TIM-3 + SP, and DN). Expression of TIM-3 resulted in a phenotypic and functional shift of PD-1 low T cells toward a more exhausted phenotype. Gene expression profiling analysis supported this finding, as more genes that can serve as exhaustion markers, such as LAG3, KLRG1, PRDM1, and TIGIT, were expressed in DP compared to other subsets. Previous studies have also found that coexpression of TIM-3 and PD-1 induces more exhaustion [17,20].
As a lymphoma subtype, the sMZL TME exhibits both a similarity to and distinctness from other lymphoma subtypes. Three major models that divide the range of TMEs in B-cell lymphoma subtypes have been proposed and represent the extremes of the spectrum of TME [43,44]. The first, "re-education", is typified by follicular lymphoma (FL), in which malignant cells retain a substantial degree of dependence on the microenvironment for survival and proliferation signals. The second, 'recruitment', is observed in classical Hodgkin lymphoma (cHL) in which the infrequent Reed-Sternberg cells are surrounded by an extensive support milieu of nonmalignant cells that is distinct Fig. 5 Coexpression of TIM-3 and PD-1 results in severe exhaustion in T cells. A Dot plots showing PD-1 and TIM-3 expression on CD3 + , CD4 + and CD8 + T cells from a representative sMZL biopsy. The PD-1 low subset was defined as cells expressing PD-1 at a level between high and negative. Graphs showing percentage of PD-1 low and TIM-3 + in CD3 + , CD4 + and CD8 + cells. B Dot plots showing coexpression of PD-1 and TIM-3 on CD4 + and CD8 + T cells from a representative sMZL biopsy. The gating is to define PD-1 high , PD-1 low and PD-1 neg . Graphs showing percentage of TIM-3 + cells in PD-1 high , PD-1 low and PD-1 neg cells in sMZL. C Dot plots showing coexpression of PD-1 and TIM-3 to define 4 subsets: DP, PD-1 SP, TIM-3 SP and DN. Graphs showing percentage of each PD-1/TIM-3-defined subset in CD8 + cells in sMZL. D The tSNE plots showing a global visualization of 4 subsets from a concatenated file of 26 sMZL. E Bar graphs showing percentage of selected markers in PD-1/TIM-3-defined subsets in sMZL. F Graphs showing percentge of T-bet + , Helios + , TOX + and Eomes + cells in CD8 + , TIM-3 SP and DP in sMZL, n = 8. The expression of T-bet, Helios, TOX and Eomes were measured by CyTOF analysis. G-I Density dot plots showing coexpression of PD-1 and TIM-3 on resting (R) or activated T cells with anti-CD3/CD28 Ab (A) in the absence or presence of IL-12 (A + IL-12) for. Cell proliferation was measured by CFSE staining (H) and granule production was measured by intracellular staining (I). from the composition of normal lymphoid tissue. The third, "effacement", is seen in Burkitt lymphoma (BL), wherein the malignant cells harbor genetic aberrations, such as translocation of MYC, that impart strong cell-autonomous survival and proliferation signals, removing dependence on the microenvironment for these stimuli. The TME from other lymphoma subtypes may align with these 3 patterns and the TME in sMZL appears similar to the "re-education" subtype.
In summary, through mass cytometry analysis and in vitro experiments, we analyzed several T-cell subsets (T FH , T reg and CD8) to determine their phenotype and biological function in sMZL. We conducted a detailed characterization of T-cell subsets and found that TIM-3-expressing PD-1 low T cells were exhausted. Our findings clearly observed a profoundly exhausted immune phenotype in the TME of sMZL, and future studies are needed to validate these findings with different large cohorts and to determine whether this phenotype can be reversed or prevented so as to improve the potential efficacy of immunological therapies. Given that PD-1 high T cells are highly represented in sMZL, targeting immune checkpoints may be a strategy that may be beneficial in selected Fig. 7 Spatial profiling identifies different T cell subsets residing in various areas in sMZL tissue. A Multiplex IHC staining of CD20 and CD3 from two representative cases of sMZL spleen tissues. B Multiplex IHC staining of ICOS and CXCL13 from one representative case of sMZL spleen tissue. C The staining images from imaging mass cytometry (IMC) with Hyperion. The whole slide was stained with 26 metalconjugated antibodies and the regions of interest were selected for Hyperion to scan. The images were viewed and emerged using the MCD viewer (Fluidigm).
patients. In addition, given that T cells with an advanced differentiation stage are associated with an inferior outcome, a strategy to increase T cells in an early differentiation stage may show therapeutic potential in sMZL patients.