Background Roughly half of all diffuse large B-cell lymphomas (DLBCLs) are infiltrated by large numbers of regulatory T-cells (Tregs). Although the presence of ‘effector’ Tregs in particular is associated with an inferior prognosis in patients on standard rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) immunochemotherapy, the role of this cell type during lymphoma initiation and progression is poorly understood.
Methods Here, we use tissue microarrays containing prospectively collected DLBCL patient specimens, as well as data from publicly available cohorts to explore the mutational landscape of Treg-infiltrated DLBCL. We further take advantage of a model of MYC-driven lymphoma to mechanistically dissect the contribution of Tregs to lymphoma pathogenesis and to develop a strategy of Treg-selective interleukin-2 (IL-2) starvation to improve immune control of MYC-driven lymphoma.
Results We find that all genetic DLBCL subtypes, except for one characterized by co-occurring MYD88/CD79 mutations, are heavily infiltrated by Tregs. Spectral flow cytometry and scRNA-sequencing reveal the robust expression of functional and immunosuppressive markers on Tregs infiltrating MYC-driven lymphomas; notably, we find that intratumoral Tregs arise due to local conversion from naïve CD4+ precursors on tumor contact. Treg ablation in Foxp3iDTR mice, or by antibody-mediated Treg-selective blockade of IL-2 signaling, strongly reduces the lymphoma burden. We identify lymphoma B-cells as a major source of IL-2, and show that the effects of Treg depletion are reversed by the simultaneous depletion of Foxp3-negative CD4+ T-cells, but not CD8+ T-cells or natural killer (NK) cells. The inhibition of ATP hydrolyzation and adenosine production by Tregs at least partly phenocopies the effects of Treg depletion. Treg depletion further synergizes with pro-apoptotic CD40 activation to sustain durable responses.
Conclusion The combined data implicate Tregs as a potential therapeutic target in DLBCL, especially in combination with other immunotherapies.
- Hematologic Neoplasms
- Immune Evation
- Immunologic Surveillance
- Lymphocytes, Tumor-Infiltrating
Data availability statement
Data sharing not applicable as no datasets generated and/or analyzed for this study. All data relevant to the study are included in the article or uploaded as supplemental information.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
- Hematologic Neoplasms
- Immune Evation
- Immunologic Surveillance
- Lymphocytes, Tumor-Infiltrating
WHAT IS ALREADY KNOWN ON THIS TOPIC
Regulatory T-cells (Tregs) are known to infiltrate lymphomas of the diffuse large B-cell (DLBCL) type, but their role in lymphoma growth and progression has not been studied in experimental models, and the prognostic significance of Treg infiltration is controversial.
WHAT THIS STUDY ADDS
We show here that Tregs infiltrating an experimental, MYC-driven lymphoma have characteristics of ‘effector’ Tregs and differ strongly from their normal thymus-derived counterparts in terms of their immunophenotype and transcriptome.
The depletion of Tregs, either in a genetic model or by interleukin-2 (IL-2) starvation, strongly reduces the lymphoma burden, especially when combined with treatments that directly compromise tumor cell viability.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Our study implies that DLBCL that is refractory to standard of care treatments, but infiltrated by large numbers of Tregs, might benefit from experimental Treg-directed therapy, in particular starvation with a Treg-selective IL-2-targeting antibody that is currently in clinical development.
Diffuse large B-cell lymphoma (DLBCL) is an aggressive, highly heterogeneous malignancy derived from mature B-cells that is fatal in a third of patients. The two major advances in DLBCL treatment were the addition of the CD20-specific antibody rituximab to standard chemotherapy two decades ago1 2 and the recent approval of chimeric antigen receptor (CAR) T-cell therapy for a select group of patients with DLBCL.3 DLBCL originates from antigen-exposed B-cells that have undergone the germinal center (GC) reaction.4 Several molecular subtypes can be distinguished based on transcriptional and mutational signatures, copy number alterations and structural variants.5 6 One of the hallmarks of DLBCL arising in immunocompetent patients—in contrast to post-transplant patients on immunosuppressive therapy7—is the mutational inactivation of various genes associated with immune detection and surveillance. Examples include the genes encoding β2-microglobulin and CD58, which are required for cytotoxic T-cell and natural killer (NK) cell recognition and killing of DLBCL cells; both genes are recurrently targeted by deletions, frameshift and other inactivating mutations and their surface expression is compromised in >60% of DLBCL cases.8 Although not directly targeted by mutations, MHCII gene expression is abrogated in DLBCL harboring inactivating mutations in the genes encoding the histone acetyltransferases (HATs) CREB binding protein (CREBBP) and EP300; such mutations occur in 30% of DLBCL9 and prevent surface MHCII expression and detection by CD4+ T-cells through loss of the active (acetylated) histone mark on H3K14, H3K18, and H3K27.10–12 Various genetic aberrations lead to programmed death-ligand 1 (PD-L1) overexpression in one-third of DLBCL cases; these include gene amplifications, transcript stabilization by truncation of the 3′-UTR and translocations of IGH, PIM1, and TP63 to the PDL1 locus.13 14 However, the clinical relevance of PD-L1 overexpression in DLBCL is not clear and trials of PD-1/PD-L1 blockade in DLBCL have delivered variable results, both as frontline therapy in combination with rituximab plus cyclophosphamide, doxorubicin, vincristine, and prednisone (R-CHOP) in PD-L1 positive patients and as salvage therapy in the setting of relapse after CART19/20 therapy.15–18 The lymphoma microenvironment of DLBCL has recently received increasing attention as a result of the availability of fluorescent multiplex immunohistochemistry (IHC) and automated quantification technologies, and certain microenvironmental features appear to predict the outcome of patients under R-CHOP therapy.19–21
We focus here on regulatory T-cells (Tregs) as highly variable, possibly predictive and targetable components of the lymphoma microenvironment. The prognostic value of Treg infiltration varies with the cell of origin DLBCL subtype22 and depends strongly on whether all CD25+ Foxp3+ Tregs, or only ‘effector’ Treg subpopulations were considered in the respective study.23–26 While strong Treg infiltration of the ‘activated B-cell’ (ABC) and ‘not otherwise specified’ (NOS) DLBCL subtypes is negatively prognostic, the opposite is true for Treg infiltration of the ‘germinal center B-cell’ (GCB) DLBCL subtype.22 The presence of Foxp3+ Tregs not further subclassified by additional markers is mostly considered to be beneficial23 25; in contrast, the infiltration of ‘effector’ regulatory T-cells, coexpressing inhibitory markers such as cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) or T-cell immunoglobulin and mucin domain 3 (TIM-3) along with Foxp3, is usually associated with a poor prognosis,24 26 as is a high Treg count in the blood of patients with DLBCL.27 We have previously shown that DLBCL arising on a background of immunosuppression due to solid organ transplantation are generally devoid of Treg infiltration28 and also lack many of the above-mentioned genetic adaptations of immune escape, such as immune-evasive B2M mutations.7 We demonstrate here that Tregs form an integral part of the lymphoma microenvironment not only in human DLBCL but also in a MYC-driven serial transplantation model of the disease, where up to 80% of intratumoral CD4+ T-cells are Foxp3+ Tregs. Intratumoral Tregs differ radically from their normal counterparts in non-tumor-bearing lymph nodes with respect to their immunophenotype and gene expression profile. The selective ablation of Tregs by diphtheria toxin in Foxp3iDTR mice abrogates lymphoma growth, which can be recapitulated by selective starvation of Tregs of their growth factor IL-2. Lymphoma cells are capable of locally converting naïve T-cells into Tregs. Intratumoral Tregs act already at the earliest stages of lymphomagenesis in our experimental model, and serve to locally suppress CD4+ but not CD8+ effector T-cell or NK cell responses.
Animal experimentation and tissue processing
For the syngeneic serial transplantation model of MYC-overexpressing lymphoma cells, 1×106 cryopreserved splenic tumor cells in 100 μL phosphate buffered saline (PBS) were injected intravenously into 6–8 weeks old mixed-gender mice. Anti-mouse CD8 (clone YTS 169.4; isotype: rat IgG2b), anti-mouse CD4 (clone YTS 191; rat IgG2b), anti-mouse NK1.1 (clone PK136; mouse IgG2a), anti-mouse CD25 (clone PC-61.5.3; rat IgG1), anti-mouse CD73 (clone TY/23; rat IgG2a), anti-mouse TIM3 (clone B8.2C12; rat IgG1), anti-mouse CTLA-4 (clone 9H10; Syrian hamster IgG), anti-mouse GITR agonist (clone DTA-1; rat IgG2b), anti-mouse PD-L1 (clone 10F.9G2; rat IgG2b), anti-mouse CD40 agonist (clone FGK4.5; rat IgG2a), anti-horseradish peroxidase isotype control (clone HRPN; rat IgG1), and anti-keyhole limpet hemocyanin isotype control (clone LTF-2; rat IgG2b) antibodies were purchased from BioXCell. Anti-mouse TIGIT (clone 1B4; mouse IgG1)29 and anti-mouse IL-2 (clone S4B6)30 antibodies were previously described. Mice were administered an initial 300 µg dose intraperitoneally on the day of tumor cell injection, followed by 200 µg injections every fourth day until the study endpoint (day 10–11), unless otherwise specified. Anti-TIGIT and anti-CD40 were administered at 100 µg doses and anti-IL-2 at 50 µg or 100 µg doses. Mice were monitored during the course of the experiment for symptoms of disease including enlarged lymph nodes (LNs), weight loss and hind leg paralysis, which were all criteria for study termination. B6.129(Cg)-Foxp3tm3(DTR/GFP)Ayr/J mice used for the serial transplantation model were depleted of Tregs by intraperitoneal injections of 20 ng/g diphtheria toxin from corynebacterium (Sigma Aldrich, D0664), diluted in 100 µL sterile PBS at the day of tumor cell inoculation and repeated every third day until study endpoint, if not otherwise stated. The mouse strains B6.129(Cg)-Foxp3tm3(DTR/GFP)Ayr/J, B6.129(Cg)-Foxp3tm4(YFP/icre)Ayr/J, B6.Cg-Tg(IghMyc)22Bri/J and B6-Tg(TcraTcrb)425Cbn/J x B6.SJL-Ptprca Pepcb/BoyJ were obtained from the Jackson Laboratories. Diseased mice were euthanized, and their spleens and inguinal and axillary LNs were weighed. Spleens and axillary plus inguinal LNs were passed through a 40 µM cell strainer to produce single-cell suspensions. Cells isolated from spleens, femoral bone marrow and peripheral blood were treated with ACK red blood cell lysis buffer pH 7.2–7.4 (150 mM NH4Cl, 10 mM KHCO3, 0.1 mM Na2EDTA). T cells of single-cell suspensions were subjected to PMA/ionomycin restimulation for 3 hours, followed by intracellular cytokine staining. Mice were injected intraperitoneally with 20 µg Fingolimod (FTY720, Sigma Aldrich, reconstituted in dimethyl sulfoxide (DMSO)), diluted in 0.9% saline, pH 7.4. Control mice were injected with DMSO diluted in 0.9% saline, pH 7.4. Fingolimod was injected daily starting 1 day before tumor cell inoculation and continued every second day until study endpoint. All animal studies were reviewed and approved by the Zurich Cantonal Veterinary Office (licenses 132/2019 and 007/2022 and their amendments, to AM).
Flow cytometric analysis
Cells were stained with the following fluorescently labeled mouse-specific antibodies and used at 1:200 dilution for extracellular staining and at 1:100 dilution for intracellular staining: BV650 CD45 (BioLegend, 30-F11), BV510 TCR (BioLegend, H57-597), BV711 (BioLegend, RM4-5), BUV496 CD4 (BD Bioscience, GK1.5), AF700 CD8 (BioLegend, 53-6.7), PE-Cy7 IFNγ (BioLegend, XMG1.2), PerCp-Cy5.5 IL10 (BioLegend, JES5-16E3), PE-Cy5 CD19 (eBioscience, eBio1D3), BV785 NK1.1 (BioLegend, PK136), FITCb NK1.1 (BioLegend, PK136), FITC IL2 (BioLegend, JES6-5H4), BUV395 CD45 (BD Biosciences, 30-F11), AlexaFluor700 CD8a (BioLegend, 53-6.7), BV570 CD62L (BioLegend, MEL-14), BV605 CTLA4 (BioLegend, UC10-4B9), BV421 Neuropilin-1 (BioLegend, 3E12), BUV737 CD44 (BD Biosciences, IM7), BV480 Ki67 (BD Biosciences, B56), FITC FoxP3 (eBioscience, FJK-16s), BV650 CD25 (BioLegend, PC61), BUV805 CD27 (BD Biosciences, LG.3A10), BV711 PD-1 (BioLegend, 29F.1A12), PB CD73 (BioLegend, TY/11.8), PerCP-Cy5.5 GITR (BioLegend, DTA-1), BUV661 GITR (BD Biosciences, DTA-1), PerCP-eFluor710 CD39 (eBioscience, eBioA1), PE TIM3 (R&D Systems, #215008), APC LAG3 (BioLegend, C9B7W), BV750 ICOS (BioLegend, C398.4A), PE-Dazzle594 TIGIT (BioLegend, 1G9), PE-Cy7 Tbet (BioLegend, 4B10), AlexaFluor647 Tbet (BioLegend, 4B10), PE-Cy7 Fas/CD95 (BD Biosciences, Jo2), APC FasL (BioLegend, MFL3), PerCP-Cy5.5 CD40 (BioLegend, 3/23), APC FasL (BioLegend, MR1), BV785 CD69 (BioLegend, H1.2F3), PE-eFluor610 RORγt (eBioscience, B2D), BV786 CD119/IFNGR1 (BD Biosciences, GR20), eFluor660 TOX (eBioscience, TXRX10), purified CD16/32 (93, 1:100), and Fixable Viability Dye eFluor 780 eBio/Affymetrix (1:1000). For intracellular staining of transcription factors, cells were fixed using the eBioscience fixation/permeabilization kit (Invitrogen, 00-5223-56, 00-5123-43), whereas for intracellular cytokine staining, cells were fixed using the BD cytofix/cytoperm kit (BD Bioscience, 554714). Apoptotic cells were detected by AnnexinV-PE-Cy7 staining (BioLegend, 640951). For high dimensional flow cytometric analysis, data were acquired on a Cytek Aurora, or on a Symphony (BD) flow cytometer and analyzed with the FlowJo software package (TreeStar). FCS files of gated cells of interest were exported from FlowJo and imported into R using the ‘flowCore’ R Bioconductor package V.2.8.0 and down-sampled to 1000 cells per sample. Cofactors for hyperbolic arcsine transformation were determined using the Cytobank Community edition web platform (Kotecha et al, 201031). Imported FCS files were integrated into a SingleCellExperiment object and arcsine transformed using the CATALYST R Bioconductor package V.1.20.1 (https://github.com/HelenaLC/CATALYST). Dimensionality reduction was performed using CATALYST. Cell clusters were identified using an R-based implementation of the PhenoGraph algorithm (https://github.com/JinmiaoChenLab/Rphenograph).32 Pairwise differential abundance of cell clusters between conditions was performed using the edgeR method, and differential state between clusters within conditions and across conditions using the limma method, in the ‘diffcyt’ R Bioconductor package V.1.16.0 (github.com/lmweber/diffcyt). All R functions were executed on R V.4.2 (R Core Team, 2020) and Bioconductor V.3.15.
ScRNA-Seq and TCGA RNA-Seq analysis
Axillary and inguinal LNs of 3–4 tumor-bearing and non-tumor-bearing FoxP3cre-YFP mice, respectively, were isolated as described above. Single cell suspensions were stained with the following antibodies for Treg sorting on the BD FACS Aria III (BD Biosciences, Switzerland): PB CD45 (BioLegend, 30-F11), PerCP-Cy5.5 CD4 (BioLegend, RM4-4), fixable viability dye eFluor780 (eBioscience) and anti-mouse CD16/32 (BioLegend, 98). About 100,000 live CD4+ FoxP3/YFP+ Tregs per sample were sorted and collected in IMDM (Gibco, 12440-053) containing 20% FCS (Gibco, 10270-106) and 1% PenStrep (Gibco, 15140-122). After cell sorting, cells were resuspended in 50 µL MgCl2 and CaCl2 free 1X PBS (Gibco, 10010-015) and incubated with 0.5 µL anti-mouse CD16/32 for 10 min on ice. For hash-tagging, cells were incubated with 1.5 µg TotalSeqTM-C0301 anti-mouse Hashtag 1 (BioLegend, M1/42, 30-F11) and TotalSeqTM-C0302 anti-mouse Hashtag 2 (BioLegend, M1/42, 30-F11), respectively, for 30 min on ice. Samples were washed three times with 3.5 mL MgCl2 and CaCl2 free PBS and subjected to scRNA-Seq at the Functional Genomics Center Zürich. Samples were prepared for sequencing using the 10X Genomics BCR/TCR (VDJ) with Feature Barcoding Single Cell reagent kit, following manufacturer’s protocol. Sequencing was performed on the Illumina NovaSeq 6000. Demultiplexing was performed using the Illumina bcl2fastq Conversion Software. Gene expression sequencing generated 351 million reads, VDJ sequencing 20 million reads, and Feature Barcode sequencing 66 million reads. Initial analysis of scRNA-Seq data was performed using the SUSHI framework which encompassed the following steps: Read quality was inspected using FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/); cell identification, read mapping and gene quantification, and sample demultiplexing was performed using 10X Genomics Cell Ranger V.6.0.2 with the GENCODE mouse genome build GRCm39 release M2633; cell clonotypes were identified using cellRanger’s ‘vdj’ function, and the Seurat V.126.96.36.1990334 Bioconductor R package was used to demultiplex the cells, identify cell doublets and dropouts based on hashtags, normalize the gene expression counts (log normalize, scale factor of 10,000), find variable genes (vst, 3000 genes), scale and center the dataset, variance-transform the counts using ‘SCTransform’, run PCA, TSNE and UMAP dimensional reduction (1–20 dimensions), find neighbors (pca, 1–20 dimensions) and clusters (multiple resolutions from 0.2 to 1, with 0.6 the resolution used), and find all marker genes for each cluster (Wilcox). The code used for these steps is available on the ‘ezRun’ Github page (https://github.com/uzh/ezRun/blob/master/R/app-SCFeatBarcoding.R). VDJ-based cell clonotype data were integrated into the Seurat object using the ‘scRepertoire’ R Bioconductor package (www.bioconductor.org/packages/release/bioc/html/scRepertoire.html), and clonotypes were clustered based on TRA amino acid sequence at 80% similarity. All R functions were executed on R V.4.2 (R Core Team, 2020) and Bioconductor V.3.15.
Gene-level counts of 481 samples were downloaded from the TCGA project NCICCR-DLBCL. Counts were generated by the STAR aligner and the human genome build GRCh38.p0. Clinical data associated with each sample were also downloaded from TCGA. Count files were imported into R V.4.0.2 for downstream analysis (including library normalization and transformation) in the DESeq2 package V.1.28.1. Samples that had a genetic subtype of ‘other’ were excluded from downstream analysis, leaving a final count of 220 samples. The DESeq2 model incorporated only the genetic subtype classification. Variance stabilized transformed counts were used to generate boxplots of target gene expression as a function of genetic subtype. Global and pairwise Kruskal-Wallis one-way analysis of variance (ANOVA) was used to generate p values of expression across and between genetic subtypes. Whole exome sequencing (WXS) BAM files of 472 patients were downloaded from the TCGA project NCICCR-DLBCL. BAM files were generated using the BWA aligner and the human genome build GRCh38.p0, and included marking of duplicate reads and base quality score recalibration following GATK best practice recommendations. On downloading, each BAM file was filtered using a BED file of a 258-gene pane, sorted, and indexed, using SAMtools.35 This panel was derived from the genes with variants as published.6 Due to a lack of matched normals, somatic variants were called using Mutect2 from GATK V.188.8.131.52 in tumor-only mode. Variants were then annotated using Funcotator from GATK. MAF files generated from Funcotator were imported into R using the R package Maftools V.2.4.05. Samples with a genetic subtype of ‘other’ were excluded from downstream analysis, leaving a final count of 206 samples. Variants that had fewer than 10 reads supporting the variant allele, and variants that had an allele frequency of greater than 2% in the gnomAD exome database, were discarded. Oncoplots, co-oncoplots, and forest plots were generated using the respective Maftools function. For the co-oncoplots, only the 25% of samples with the highest and 25% with the lowest respective gene expression were directly compared.
IHC and immunofluorescence microscopy of DLBCL patient cohorts and murine tissues
The two DLBCL patient cohorts used here are described in Juskevicius et al36 and in the website (https://www.eahp-sh2022.com/wp-content/uploads/2022/09/Lymphoma-symposium.pdf). Tissue microarrays (TMAs) were stained for FoxP3 using the clone SP97 (Abcam, ab99963) at 1:50 dilution and for CD40 using the clone CL1673 (Novus Biological, NBP2-34488) at a 1:800 dilution. Cases containing ≥5% FoxP3+tumor-infiltrating T-cells were considered positive22 and CD40 was scored as follows: 0, negative; 1, weakly positive in <50% of tumor cells; 2, either weakly positive in >50% or moderately positive in 10%–50% of tumor cells; 3, either moderately positive in >50% or strongly positive in tumor cells. For co-immunofluorescence microscopy, patient cryosections were thawed at room temperature (RT), rehydrated in PBS and incubated in blocking buffer (1:2 PBS, 1:4 normal goat serum (Vector Laboratories, S-1000), 1:4 7.5% bovine serum albumin (BSA) in PBS, 1% ChromePure IgG (Jackson Immunology, 154669)) for 30 min. After washing, the following primary anti-human antibodies (1:100 in blocking solution) were added: rabbit FoxP3 (Sigma-Aldrich, HPA045943), rat Ki67 (Invitrogen, SolA15), and incubated over night at 4°C. Slides were incubated with the following secondary antibodies diluted 1:400 in 7.5% BSA in PBS for 2 hours at RT in the dark: goat anti-rat IgG AlexaFluor488 (Invitrogen, A11006) and goat anti-rabbit IgG AlexaFluor568 (Invitrogen, A11011). After washing, DAPI (Thermo Fisher, 62248, 1:1000 in PBS) was added for 5 min, slides rinsed, dried at RT and 20 µL mounting medium (Invitrogen, Fluoromount-GTM, 00-4958-02) and cover slips (epredia) were added. Images were recorded using a Leica DM6 B Microscope. Informed consent was obtained from all subjects, and all experiments using human material conformed to the principles set out in the WMA Declaration of Helsinki and the Department of Health and Human Services Belmont Report. The Ethics Committee of Northwestern and Central Switzerland and the Ethics Committee of both Basel Cantons approved the work with patient material presented here (EKNZ 2014-25 and 282/08, respectively).
Spleens harvested from tumor-bearing and non-tumor-bearing mice were fixed for 24 hours in 4% PFA (Laborchem international, LC-6470.5) at RT. After paraffin-embedding, tissues were fixed to the slides by baking at 52°C for 3 hours, then dewaxed with Xylol (Sigma-Aldrich, STBJ5892), rehydrated using an ethanol dilution series (100% EtOH, 90% EtOH, 75% EtOH, H2O, PBS). Slides were subjected to antigen retrieval by boiling for 2 hours in a water bath in 1 × citrate buffer (Sigma, C9999-1, 10× stock, pH 6), followed by blocking in 25% normal goat serum (Vector Laboratories, S-1000), 25% PBS, 25%–7.5% BSA and 1% ChromePure IgG (Jackson Immunology, 154669) for 1 hour. Slides were incubated overnight with the following primary anti-mouse antibodies (1:100 in blocking solution): rabbit FoxP3 (Cell Signaling Technology, D608R) and rat Ki67 (Invitrogen, SolA15). After washing, the following secondary antibodies (1:400 in 7.5% BSA) were added for 2 hours in the dark: goat anti-rat IgG Alexa Fluor 488 (Invitrogen, A11006), goat anti-rabbit IgG Alexa Fluor 568 (Invitrogen, A11011). After washing, DAPI (Thermo Fisher, 62248, 1:1000 in PBS) was added for 5 min, followed by mounting.
Cell culture experimentation
We used a panel of 13 DLBCL cell lines including six of the GCB subtype (SU-DHL4, SU-DHL5, SU-DHL6, SU-DHL8, SU-DHL10, and SU-DHL16), 6 of the ABC subtype (HBL-1, OCI-Ly3, OCI-Ly10, SU-DHL2, RIVA and U-2932) and 1 unclassified DLBCL cell line (RC-K8) that have been previously described.10 37 38 All cell lines were routinely tested and negative for mycoplasma. Cell line authentication was performed according to the guidelines of the International Cell Line Authentication Committee using short tandem repeat profiling; the result of this effort was described recently for all cell lines used here, along with their sources.39 Cell lines were maintained at 37°C and 5% CO2 in IMDM (OCI-Ly3, OCI-Ly10, RIVA) or RPMI (all others) supplemented with 10% (OCI-Ly3, RIVA) or 20% (all others) heat-inactivated fetal calf serum (FCS, Gibco) and 100 U/mL penicillin and 100 µg/mL streptomycin. Primary murine MYC-overexpressing lymphoma cells were cultured in high-glucose Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% FCS, 50 µM β-mercaptoethanol (Sigma-Aldrich), 100 U/mL penicillin, 100 µg/mL streptomycin (Gibco), and 100 µM L-asparagine (Sigma-Aldrich) at 37°C and 10% CO2.
RNA was extracted using the NucleoSpin RNA kit (Macherey-Nagel) according to the manufacturer’s instructions. One microgram of total RNA was reverse-transcribed using SuperScript III reverse transcriptase (Invitrogen). TaqMan gene expression assays (Thermo Fisher Scientific) were performed for the following genes: human CD40 (Hs00386848_m1), human IL-2 (Hs00174114_m1) and human ACTB (HS01060665_g1). The samples were analyzed on a LightCycler 480 instrument (Roche) and target mRNA abundance was calculated relative to ACTB.
All statistical analyses were performed using GraphPad Prism software. Graphs represent means plus SD of at least two independent experiments. Statistical analysis was performed using ordinary one-way ANOVA in case of three or more groups or unpaired two-tailed Mann-Whitney test for a comparison of two groups.
MYC-expressing lymphomas are infiltrated by large numbers of activated Tregs that express high levels of immunosuppressive markers
To characterize the Tregs infiltrating the lymphoma microenvironment in a convenient, rapid and standardized model, we took advantage of MYC-expressing lymphoma cells from Eµ-MYC mice, which, on transplantation into syngeneic immunocompetent C57BL/6 recipients, form tumors in the LNs, spleen and bone marrow within 10–12 days post-transplant; tumor cells can be readily identified flow cytometrically by their CD19, intermediate CD45 and Ki67 expression.28 MYC-driven lymphomas in the LNs (and spleens, data not shown) are highly enriched for Foxp3+ Tregs relative to control LNs (figure 1A). The extent of Treg infiltration correlates strongly with the tumor burden (figure 1B). Tregs infiltrate the tumor mass and are found in close proximity to tumor B-cells (figure 1C). Spectral flow cytometry using a panel of 14 antibodies specific for relevant Treg markers identified five Treg clusters that were predominantly found in normal LNs and consisted of naïve cells (CD62L+, CD44-) with low to intermediate expression of functional/immunosuppressive markers (CD39, CD73, GITR) (figure 1D–G, online supplemental figure S1A,B); the immunophenotype of normal LN Tregs described here is consistent with the published markers of thymus-derived Tregs in murine secondary lymphoid organs.40 In contrast, Tregs within tumors showed downregulation of CD62L and enhanced expression of the immunosuppressive markers CD39, CD73 and GITR, of neuropilin-1 (Nrp-1) and of the interleukin-2 (IL-2) receptor α-subunit CD25 (figure 1D–G, online supplemental figure S1B), all of which are attributes of effector Tregs.41 The largest changes in the expression between control and tumor-infiltrating Tregs were observed in CD62L, Nrp-1, CD25 and CD73/CD39, two enzymes that convert immunostimulatory ATP into immunosuppressive adenosine (online supplemental figure S1B). Only one, numerically minor cluster of (tumor-infiltrating) Tregs was positive for Ki67, indicating active proliferation (figure 1G). Tumor-infiltrating Tregs appeared in tumors of wild-type as well as T-cell-receptor (TCR)-transgenic OTII recipients, in which the TCRs of all CD4+ T-cells, including Tregs, are specific for the chicken ovalbumin peptide 323–339; this result indicates that Treg recruitment and/or local expansion is not dependent on binding to cognate antigen (figure 1H). Regular orally administered doses of the sphingosine-1-phosphate analog fingolimod, which prevents lymphocyte egress from lymphoid tissues, did not prevent the appearance of effector Tregs within the transplanted lymphoma, indicating that these cells arise as a consequence of local differentiation or expansion, rather than recruitment from the circulation or from distant lymphoid organs (figure 1I). Indeed, the adoptive transfer of >95% pure, Foxp3/YFP-, naïve CD45.2+ CD4+ T-cells sorted from Foxp3YFP reporter mice into tumor-bearing CD45.1+ congenic recipients led to their efficient conversion into Foxp3/YFP+ Tregs in tumor-bearing organs, which was not seen in non-tumor-bearing control mice (figure 1J). Furthermore, co-culturing of sorted, highly pure (>95%) tumor cells with naïve Tregs isolated from non-tumor-bearing animals’ LNs for 3 days resulted in their increased expression of Ki67, CD44, CD27, Tbet, GITR, CD39 and other markers characteristic of tumor-infiltrating effector Tregs (figure 1K). Interestingly, co-culturing not only affected the immunophenotype of the Tregs, but also boosted the proliferation of the tumor cells, as assessed in a total of eight independent co-cultures (figure 1L). The combined data suggest that naïve T-cells differentiate into Foxp3+ Tregs and acquire properties of effector Tregs in the lymphoma microenvironment, a process that appears to be tumor antigen-independent, locally driven and proportional to the tumor burden.
Single cell RNA-sequencing uncovers the transcriptional signature of lymphoma-infiltrating Tregs
To address how lymphoma-infiltrating Tregs differ from their normal LN counterparts using a complementary technique, we took advantage of Foxp3GFP reporter mice to sort and single cell-sequence 20,000 Foxp3GFP-positive Tregs per condition. Dimensionality reduction followed by Seurat clustering revealed nine Treg clusters that differed strongly in their relative abundance in tumor-bearing versus control LNs (figure 2A). Three clusters were exclusively present in lymphoma-bearing LNs, and the remaining six were vastly more abundant in the normal LNs (figure 2A). Hierarchical clustering of the top 80 most differentially expressed transcripts allowed the segregation of the nine populations (figure 2B) and confirmed several of the differentially regulated markers that were identified by flow cytometry, in addition to uncovering new ones. Several transcripts were highly intratumoral Treg-specific (among them the IL-7 receptor-encoded by Il7r), or at least predominantly expressed by tumor-infiltrating Tregs (Tigit, Icos, Pdcd1-encoding PD-1-, Ifngr and Ctla4) (figure 2B,C). Other transcripts (one example being Stat1) were exclusively expressed by control Treg clusters or were expressed by both control and tumor-infiltrating Treg clusters (Bcl2, Stat4; figure 2B,C). The combined data suggest substantial differences in Treg gene expression as a result of lymphoma infiltration, of which several could be confirmed also at the protein level (figure 2D). We further performed a TCR repertoire analysis on tumor-infiltrating versus control Tregs; this analysis uncovered hundreds of T-cell receptor α-chain (TRA) clusters in each sample of pooled Tregs from three to four mice per condition (309 TRA clusters in the control Tregs, 189 in the tumor Tregs; cut-off of >85% amino acid sequence similarity used for assignment to TRA cluster), which together accounted for ~2/3 of all Tregs. A total of 849 cells (control Tregs, vs 463 tumor Treg cells) were assigned to the most numerically dominant cluster; TRA-based Treg clusters did not segregate with the Treg clusters identified by scRNA sequencing of non-TCR transcripts, but rather mapped more or less randomly to the Seurat clusters shown in figure 2A,E. The top 20 most abundant TRA clusters were found at similar frequencies among control and tumor-infiltrating Tregs (figure 2F). The combined data indicate that Tregs infiltrating MYC-driven lymphomas are not subject to clonal expansion, but show a very distinct (non-TCR) gene expression profile that clearly distinguishes them from normal LN Tregs.
Tregs infiltrate some but not all subtypes of human DLBCL
To assess whether the level of Treg infiltration observed in our model is reflective of Treg infiltration in human DLBCL cases, we stained 105 prospectively collected specimens from newly diagnosed patients with DLBCL in Switzerland (two independent cohorts, one unpublished and one published36) for FoxP3 and subjected their tumor DNA to amplicon sequencing of the 68 most recurrently mutated DLBCL genes. We found FoxP3-positive Tregs in close proximity to tumor cells, with some tumor cell/Treg pairs forming what appeared to be direct contacts (figure 3A). Roughly one half of cases exhibited evidence of substantial (>5%) FoxP3+ Treg infiltration, whereas the other half contained negligible numbers of Tregs (figure 3B,C). GCB-DLBCL and ABC-DLBCL—distinguished using the Hans classification—were similarly likely to be infiltrated by Tregs (figure 3C). A comparison of the spectrum of mutations in DLBCL driver genes revealed that several mutations were over-represented in the Treg-infiltrated cases; among these were SOCS1, ACTB and EZH2 (online supplemental figure S2A-C). Other mutations were over-represented in the Treg-negative cases; among these were CD79B, KMT2D and DTX1 (online supplemental figure S2A-C). The parallel analysis of Treg infiltration of 206 DLBCL patient datasets publicly available through TCGA using cibersort42 demonstrated that the ‘MCD’ genetic subtype (defined by co-occurring MYD88 and CD79 mutations) had the lowest Treg infiltration and FOXP3 transcript levels, whereas the other three genetic subtypes introduced by Schmitz et al all had high Treg infiltration (figure 3D,E); over-represented mutations in the Treg-infiltrated cases included the TNFRSF14 variant, whereas Tregneg/low cases were more likely to carry mutations in MYD88, PIM1 and CD79B (figure 3F, online supplemental figure S2D). The combined data suggest that DLBCL exhibiting the characteristic ‘MCD’ subtype mutations are largely devoid of Treg infiltration; the Treg-infiltrated cases are less uniform in their mutational spectrum and include cases with canonical GCB-DLBCL mutations but also cases of the ‘N1’ genetic subtype.
Treg depletion reduces lymphoma growth and boosts anti-lymphoma effector T-cell responses
Given the high frequencies of effector Tregs within the lymphoma microenvironment in our experimental model, we set out to functionally assess the consequences of Treg depletion for lymphoma growth and anti-lymphoma immunity. Mice expressing the diphtheria toxin (DT) receptor specifically in the Treg compartment (Foxp3DTR mice) received DT every 3 days starting on the day of lymphoma cell injection; this treatment led to a >90% reduction of Tregs in LNs, blood and other lymphoid and non-lymphoid tissues (online supplemental figure S3A and data not shown). Treg depletion resulted in a strongly reduced lymphoma burden in the LNs and spleen (figure 4A, online supplemental figure S3B); DT administration to wild-type mice had no effect (online supplemental figure S3C). The residual tumor cell population in Treg-depleted Foxp3DTR mice failed to proliferate as determined by Ki67 staining (figure 4B). Continuous Treg depletion extended the lifespan of tumor-bearing mice by at least 3 days past the usual humane endpoint (day 11) of the experimental protocol, and measurably reduced the tumor burden as late as 14 days post injection (online supplemental figure S3D). The reduced lymphoma growth on Treg depletion was accompanied by strongly elevated frequencies of intratumoral CD4+ and CD8+ T-cells and by elevated interferon-γ (IFN-γ) production by both NK and CD4+ T-cells, but not CD8+ T-cells (online supplemental figure S3E-J). Spectral flow cytometry of the CD4+ and CD8+ T-cell compartments of Treg-replete and Treg-depleted tumor-bearing and control LNs further revealed large differences in these T-cell populations, with samples segregating into the three treatment groups (online supplemental figure S3K,L). Whereas CD4+ and CD8+ T-cells in control LNs were naïve (CD62Lhi, CD44neg) and expressed the co-inhibitory receptor CTLA-4, their tumor-resident counterparts had downregulated CD62L and CTLA-4 but for the most part remained negative for CD44; only after Treg depletion, both compartments were fully activated and expanding as judged by their CD44, ICOS and Ki67 expression, and had differentiated into Tbet-positive type I cells; CD8+ T-cells additionally upregulated the activation/exhaustion markers TIM3 and LAG3 (figure 4C–H, online supplemental figure S3M,N). To address whether the emergence of CD8+, CD4+ T-cells or NK cells is functionally relevant for the reduction of lymphoma growth associated with Treg depletion, all three populations were depleted by suitable antibodies alongside the depletion of Tregs by DT in Foxp3DTR mice. Neither CD8+ T-cell nor NK cell depletion reversed the effects of Treg depletion on lymphoma growth, despite a >90% reduction of the targeted populations; in contrast, the depletion of CD4+ T-cells—in addition to Tregs—resulted in the restoration of lymphoma growth (figure 4I, online supplemental figure S3O-Q).
Various monoclonal antibodies that target the immunosuppressive activities of Tregs are currently in clinical development; among the most promising are antibodies blocking TIM3, TIGIT, CD73 and CD25 as well as a GITR agonistic antibody, especially when combined with immune checkpoint inhibitors such as anti-PD-L1.43 To address whether any of the listed antibodies in clinical development would phenocopy the effects of DT-mediated Treg ablation, we administered each antibody, alone and/or in combination with anti-PD-L1 or anti-CTLA-4. Neither the agonistic GITR-specific antibody, nor the TIGIT-blocking or CD25-blocking antibodies reduced the lymphoma burden relative to isotype control antibodies; in contrast, robust effects were obtained with anti-CD73 and anti-TIM3 blocking antibodies, alone and together with PD-L1 neutralization (figure 4J). The reduction of lymphoma growth was accompanied by a reduction of Treg infiltration under all effective treatments (figure 4K), with the latter being quite predictive of treatment success. Whether Treg depletion is a cause or consequence of the reduction in the lymphoma burden cannot be deduced from these data. The combined results suggest that Treg depletion effectively reduces the tumor burden, which can be attributed to reinvigorated CD4+ effector T-cell, but not to CD8+ T-cells or NK cell responses. Further experimental evidence suggests that Tregs suppress CD4+ T-cells via their expression of the ATP hydrolyzing enzymes CD39 and CD73.
Tregs promote early stages of lymphoma engraftment and growth
To study the dynamics of lymphoma engraftment and growth, and of Treg support of these processes in more detail, we first tracked and quantified lymphoma cells and Tregs in the LNs and spleen, and also in the blood and bone marrow. Tumor cells were first detectable in blood, spleen and bone marrow at 6 days post injection, and appeared in the LNs 2 days later (figure 5A, online supplemental figure S4A). Treg frequencies among all CD4+ T-cells differed substantially from one examined tissue to the other (being highest in the bone marrow at 50%, and lowest in the LNs and blood at <10%), and increased Treg frequencies were detected in all tissues on lymphoma cell injection (figure 5B, online supplemental figure S4B). The kinetics of increased Treg frequencies coincided with the kinetics of lymphoma cell appearance in the three tissues (day 8 in the LNs, day 6 in bone marrow and spleen), but not in blood, where only modest Treg increases were visible, and only with a 3-day delay at day 9 (figure 5A,B, online supplemental figure S4A,B). To address at which point in time Treg depletion would have the strongest impact on the lymphoma burden, we initiated DT treatments either 2 days before, or on days 1, 2, 3, 4 or 5, or on day 9 post tumor cell injection, and continued the treatment until the study endpoint. Only the early depletion of Tregs—initiated up until day 3, but not later depletion effectively reduced the lymphoma burden, and lymphoma cell proliferation as assessed by Ki67 staining (figure 5C,D). Early, but not late Treg depletion further also boosted the intratumoral CD4+ and CD8+ effector T-cell responses that coincided with tumor control (online supplemental figure S4C-E). Indeed, a few doses of DT administered up until day 4 (resulting in the restoration of Tregs back to normal levels by the end of the study) were as effective as continuous Treg depletion until the study endpoint at suppressing the lymphoma burden (figure 5E–G). This was accompanied by increased CD4+ T-cell activation, as indicated by elevated levels of IFN-γ, Ki67 and PD-1 (figure 5H). The combined results indicate that Tregs facilitate lymphoma cell engraftment and growth in our experimental model, likely by affecting CD4+ effector T-cell priming, and have a critical role long before elevated Treg levels (and lymphoma cells) appear in affected tissues and the bloodstream.
Treg-specific IL-2 starvation reduces Treg infiltration and lymphoma growth
As IL-2 is a key cytokine driving Treg differentiation by promoting Foxp3 expression in naïve T-cells,44 we asked whether Treg-specific IL-2 starvation might phenocopy the effects of Treg ablation in Foxp3DTR mice, and whether tumor B-cells might directly act as a source of IL-2. Indeed, we found that MYC-expressing lymphoma cells produce large amounts of IL-2, which is not the case for normal B-cells in the same LNs (figure 6A). Human DLBCL cell lines vary in their IL-2 expression, with some expressing enormous amounts, others modest amounts and yet others no IL-2 whatsoever (online supplemental figure S5A). To specifically starve Tregs, but not effector T-cells of IL-2, we took advantage of an antibody that selectively prevents IL-2 binding by the α-chain of the IL-2 receptor, CD25, which is expressed by Tregs at much higher levels than by effector T-cells.45 Anti-IL-2 antibody treatment reduced Treg frequencies without affecting CD4+ or CD8+ effector T-cell frequencies, and dose-dependently reduced the lymphoma burden (figure 6B and C, online supplemental figure S5B); CD25 expression was indeed much higher on tumor-infiltrating Tregs than on intratumoral CD4+ and CD8+ T-cells (online supplemental figure S5C). The small residual Treg population that escaped starvation did not immunophenotypically differ from tumor-infiltrating Tregs in isotype-treated tumor-bearing mice as determined by spectral flow cytometry (figure 6D, online supplemental figure S5D). Anti-IL-2 treatment had no effect on the tumor burden in T-cell (and Treg)-deficient RAGγc-/- mice (figure 6E), suggesting that the beneficial effects of IL-2 neutralization are probably not tumor cell-intrinsic. Interestingly, IL-2 neutralization had no effect on the quantity or immunophenotype of normal Tregs in control mice, the surface marker expression and frequencies of which were unchanged; indeed, Tregs in non-tumor-bearing LNs expressed much less CD25 than their tumor-infiltrating counterparts (figure 6F,G, online supplemental figure S5E,F). The combined results provide experimental evidence for the efficacy of Treg-specific IL-2 starvation in controlling lymphoma growth in this MYC-driven model.
IL-2 neutralization synergizes with CD40 activation in promoting durable responses
As Treg starvation alone is unlikely to confer long-term survival benefits in our aggressive lymphoma model, we asked whether anti-IL-2 therapy would synergize with other immunotherapies in promoting durable responses beyond the usual 10-day timeframe of our model. As we found aberrantly low expression of the co-stimulatory surface marker CD40 on tumor B-cells relative to normal B-cells (figure 7A), and detected a strong increase in CD40 expression on tumor B-cells but not normal B-cells on Treg depletion (figure 7B), we set out to determine whether targeting CD40 using an agonistic antibody would reduce the lymphoma burden, with and without additional Treg depletion by IL-2 neutralization. Twice-weekly injection of a CD40-specific agonistic antibody effectively reduced the lymphoma burden and lymphoma cell proliferation (figure 7C,D). In fact, a single anti-CD40 injection administered on the day of tumor cell injection was sufficient to prevent lymphoma growth until the humane endpoint was reached for all isotype antibody-treated mice (figure 7C,D). As CD40 activation led to a strong Fas upregulation on MYC-expressing lymphoma cells (online supplemental figure S6A), and simultaneously to Fas ligand (FasL) upregulation on NK cells (online supplemental figure S6B), we asked whether the tumoricidal activity of the CD40 agonistic antibody required the cytotoxic activities of T-cells or NK cells, or of FasL. CD40 activation was as effective in RAG2γc-/- mice-lacking T-lymphocytes and B-lymphocytes and NK cells—as in wild-type C57BL/6 mice at reducing lymphoma growth (figure 7E) and could not be counteracted by FasL blockade (online supplemental figure S6C). A single dose of CD40 agonist administered 1 day prior to the study endpoint was sufficient to reduce the lymphoma burden (figure 7E) and allowed us to determine that tumor cells became annexin-V-positive under treatment, indicating apoptotic cell death induced by CD40 activation (figure 7F). Normal B-cells in the same lymph nodes did not undergo increased apoptosis on CD40 activation, and even appeared to proliferate more (online supplemental figure S6D,E).
To address how common CD40 expression is among patients with DLBCL, we stained 110 samples from the two prospectively collected patient cohorts mentioned earlier for CD40. CD40 expression varied radically across patients, with some tumors expressing no CD40 whatsoever and others expressing copious amounts (figure 7G). Roughly one-fourth of cases expressed high, moderate, low levels of, or no CD40, respectively (figure 7H). No bias towards ABC-DLBCL or GCB-DLBCL subtype was observed (figure 7H). A similar variability of CD40 expression was observed in a panel of DLBCL cell lines (online supplemental figure S6F). Treg infiltration and CD40 expression appeared to be independent variables in the 94 patients with DLBCL for whom both CD40 and FoxP3 immunohistochemical stainings were evaluable (figure 7I).
Although highly efficient in the short term at reducing the lymphoma burden, CD40 activation alone did not sustain durable responses, neither if the treatment was continued for another 2 weeks, nor if it was discontinued at the usual study endpoint of 10 days (figure 7J,K). However, when CD40 activation was combined with IL-2 neutralization, with both treatments beginning on the day of tumor cell injection, all mice on the dual treatment were symptom-free and had a minimal or undetectable tumor burden at the study endpoint (figure 7J,K). Treg infiltration was reflective of the tumor burden, and lowest in the absence of tumor and under IL-2 neutralization (figure 7L). The combined results implicate CD40 as a promising target enabling direct killing by apoptosis of a subset of CD40-expressing DLBCL; the response to CD40-directed therapy is further improved by Treg depletion.
We show here that (1) Tregs are abundant, highly immunosuppressive cells in the lymphoma microenvironment in our experimental model of MYC-driven B-cell lymphoma, and in several genetic subtypes of DLBCL, that (2) Tregs are induced in the lymphoma microenvironment and sustained by tumor-derived IL-2 and that (3) their DT-driven or antibody-driven ablation strongly reduces the lymphoma burden, especially when combined with a direct cytotoxic immunotherapy such as CD40 stimulation. Several of the functional and immunosuppressive effector Treg markers identified here in the lymphoma context by spectral flow cytometry and scRNA-seq have previously been assigned to Tregs infiltrating the tumor microenvironment of solid organ malignancies, and are well known to maintain Treg functional stability and/or to enhance intratumoral Treg suppressive activities.46 Examples include Nrp-1, which marks Tregs associated with poor prognosis in human melanoma and head and neck squamous cell carcinoma; its loss specifically in Tregs leads to control of B16 melanoma.47 48 TIGIT and CTLA-4 are examples of inhibitory receptors that are not only overexpressed on tumor-infiltrating relative to naïve Tregs in tumor-bearing mice and patients, but also contribute functionally to Treg suppressive activity and may be targeted to boost tumor rejection,29 46 49 although not in our hands when using a TIGIT-blocking antibody. The effector Tregs in our experimental model appear to arise via an intermediate state that is characterized by loss of CD62L but lacks expression of CD44 or any of the immunosuppressive markers listed above; our adoptive transfer experiments suggest that Tregs are converted locally in the lymphoma microenvironment from CD4+ precursors, rather than being recruited as fully functional, mature cells.
Tregs were previously shown22 26 and confirmed again here to infiltrate DLBCL at considerable densities. Cibersort analysis of TCGA data and our own in situ observations indicated that the MCD genetic subtype was less infiltrated than other genetic subtypes6; indeed, signature mutations such as the MYD88L265P, PIM1 and activating CD79B mutations were significantly overrepresented in the Treg-negative cases. This observation could be confirmed for CD79B in our independent Swiss cohorts, whereas the stratification by cell-of-origin (COO) was not helpful in predicting Treg infiltration, as the two genetic ABC-DLBCL subtypes (N1, MCD) have strongly divergent Treg infiltration that masks potential differences between the GCB-DLBCL and ABC-DLBCL COO subtypes.
The presence of effector Tregs, characterized by co-expression of CTLA-4 or TIM-3 with Foxp3, in the only two studies to date taking markers other than Foxp3 into account24 26 was of strong negative prognostic importance. Indeed, ablation of Tregs, either in a genetic, DT-driven model or by selective Treg starvation of their growth factor IL-2, substantially and consistently reduced the tumor burden in all examined organs in our MYC-driven model. The tumor-promoting effect of Tregs could be attributed to their effects on CD4+, but not CD8+ T-cells or NK cells. Interestingly, our own earlier immunohistochemical analysis comparing Treg infiltration of DLBCL arising in immunocompetent versus immunodeficient patients provides circumstantial evidence for an exclusive role of Tregs in the immunocompetent patient, as we found Tregs to be almost completely absent in post-organ transplant DLBCL patients on immunosuppressive therapy.28
IL-2 acts at various stages of Treg differentiation, driving Foxp3 expression during Treg development, and promoting the long-term survival and suppressive activity of mature Tregs.44 Selectively interfering with IL-2 signaling in Tregs (but not other IL-2-consuming T-cells) by preventing IL-2 binding to the main ligand-binding subunit of IL-2R in Tregs, CD25, phenocopied the effects of DT-driven Treg depletion. The human variant of the antibody we used here is in clinical development as an IL-2/anti-IL-2 antibody complex fusion (not to be confused with the naked antibody administered in our study), which directs IL-2 to CD8+ T-cells resulting in their efficient expansion and in tumor immune control in various solid cancer models.45 50 The crystal structure of the human version of the antibody has revealed that it occupies the CD25-binding epitope of IL-2 and precisely overlaps with CD25, which explains why it prevents binding of IL-2 to the trimeric IL-2R on Tregs (consisting of CD25, CD122 and the common gamma chain γc) in our hands, while—in complex with IL-2—selectively targeting the dimeric IL-2R (consisting of CD122 and γc) on CD8+ T-cells.45 As anti-IL-2 treatment efficiently depleted tumor-infiltrating CD25hi effector Tregs in our study, but left the naïve Treg compartment in control mice largely intact and immunophenotypically unaltered, it represents an attractive alternative to other Treg-targeting therapeutic strategies that are potentially less tumor-Treg-specific.43 High level IL-2 production is a fairly common adaptation in DLBCL as 4 of the 13 examined cell lines secreted copious amounts of the cytokine when in culture.
IL-2 starvation and Treg depletion by DT were effective in short-term studies, but—at least as monotherapy—failed to extend the survival of our tumor-bearing mice by more than a few days. As CD40 expression by lymphoma B-cells was aberrantly low relative to normal B-cells, but could be restored by Treg depletion, we investigated whether activation of CD40 by an agonistic antibody might synergize with Treg depletion in eradicating tumors and generating durable responses. This was indeed the case as mice on anti-IL-2/anti-CD40 combination treatment survived for 2 weeks past the usual humane endpoint, with only minimal evidence of residual tumor cells at the study endpoint.
CD40-directed agonistic antibodies were not initially developed to directly target, and kill, CD40-expressing tumor cells but rather to increase the number and quality of tumor-infiltrating T-cells and thereby, response effectiveness, in CD40-negative solid organ malignancies.51–53 In such settings of CD40-negative malignancies, CD40 on antigen-presenting cells (APCs) plays a central role in stimulating immune synapses, including during T-cell priming, when its interaction with the CD40 ligand (CD40L) licenses DCs to activate antigen-specific T-cells.54 The situation is complicated in B-cell malignancies by varying extents of tumor-intrinsic CD40 expression; chronic lymphocytic leukaemia, Burkitt lymphoma and DLBCL cells have all been reported to express CD40 and to respond to CD40 activation with activation-induced cell death (AICD) in vitro and/or in experimental models.55–57 Interestingly, normal and neoplastic B-cells respond differently to CD40 agonistic antibodies and soluble CD40L, whereas CD40 stimulation promotes differentiation and growth of normal B-cells in vitro (confirmed here also in vivo), malignant B-cells typically undergo AICD when exposed to CD40 agonistic antibodies or soluble CD40L,57 58 or are sensitized to other cell death-inducing agents.55 An active CD40 signaling pathway was found to be predictive of treatment response in a phase I trial of dacetuzumab, a CD40-specific antibody with partial agonistic activity57 and CD40 expression has generally emerged as a favorable prognostic marker in patients on CHOP or R-CHOP.59–61 Our experimental data do not completely rule out tumor-cell extrinsic effects of CD40 targeting, but the comparable levels of tumor reduction in WT and RAGγc-/- mice suggest that direct killing by CD40 activation is the dominant mechanism of tumor control in our model.
Our combined data from both experimental mouse studies and patient data provide a strong rationale for targeting Tregs in highly Treg-infiltrated DLBCL, especially in combination with cytotoxic immunotherapy directed at CD40.
Data availability statement
Data sharing not applicable as no datasets generated and/or analyzed for this study. All data relevant to the study are included in the article or uploaded as supplemental information.
Patient consent for publication
This study involves human participants and was approved by Ethics Committee of Northwestern and Central Switzerland (number: EKNZ 2014-252) and Ethics Committee of both Basel Cantons (number: 282/08). Participants gave informed consent to participate in the study before taking part.
The authors would like to thank all members of the Müller and Tzankov labs for helpful comments and discussions.
Correction notice This article has been corrected since it was first published online. Under the methods section, the sentence 'anti-mouse IL-2 (clone NARA-1; rat IgG2a) antibodies were recently described' has been updated to 'anti-mouse IL-2 (clone S4B6) antibodies were previously described'. The accompanying citation has been amended to Boyman et al. 2006.
Contributors Authorship declaration: KS designed, performed and analyzed experiments and co-wrote the manuscript. PL performed all bioinformatic analyses of high-dimensional flow cytometry, scRNAseq and TCGA data. DW and DS helped with experiments. UK, NJ and OB provided critical tools and reagents as well as advice. AT stained and analyzed patient samples and AM supervised the study, co-wrote the manuscript and acts as a guarantor.
Funding This work was supported by the Swiss Cancer League (KFS-5228- 02-2021 to AM and AT) and the Clinical Research Priority Program “Precision Oncology” of the University of Zurich. Additional support was provided by the Forschungskredit of the University of Zurich (to KS) and the Comprehensive Cancer Center Zurich (grants to PL).
Disclaimer The sponsors had no role in the study design, data analysis or any other part of the research and manuscript submission. The authors declare no competing financial interests.
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.