MHC Class II Antigen Presentation by Lymphatic Endothelial Cells in Tumors Promotes Intratumoral Regulatory T cell–Suppressive Functions

The tumor microenvironment (TME), which is a major regulator of cancer development, is generally a highly immunosuppressive milieu (1, 2). Regulatory T cells (Treg), which naturally establish and maintain immunologic tolerance and regulate immune homeostasis (3), are potent suppressors of effector T cells and are found at high frequencies in the TME of various types of cancers (4, 5). The different TMEs in distinct cancer types drive intratumoral Treg tumor-specific genetic programs (6, 7).

The development of several solid tumors triggers angiogenesis and lymphangiogenesis, which consist of expansion and remodeling of blood and lymphatic endothelial cells (BEC and LEC), respectively. Tumor-associated blood vessels provide oxygen and nutrients for tumor cells and, together with lymphatic vessels (LV), facilitate tumor metastasis (8). However, the mechanisms by which tumoral LVs promote disease progression remain a matter of debate.

LVs are essential for initiating adaptive immune responses as they transport immune cells and antigens from peripheral tissues to lymph nodes (LN; refs. 9–12). In addition, LECs express a multitude of immune mediators and growth factors affecting immune functions (13). These cells are specialized for their unique microenvironments (13, 14) and modulate immune responses in various ways (15–20). A way LECs affect immune responses is by functioning as antigen-presenting cells (APC). Steady-state murine LN-LECs present endogenously expressed tissue-restricted antigens through MHC class I (MHCI) molecules to induce autoreactive CD8⁺ T deletion (21–23). They also cross-present exogenous antigens to CD8⁺ T cells, promoting their apoptosis (24).

Whether LECs present antigens through MHC class II (MHCII) molecules remains equivocal. In one study, steady-state LECs lack MHCII antigen-presentation machinery (25). However, we have shown that LECs express both endogenous MHCII molecules and MHCII–peptide complexes acquired from dendritic cells (DC) to induce CD4⁺ T-cell dysfunction (26). Furthermore, the genetic abrogation of endogenous MHCII in LN stromal cells induces signs of autoimmunity in aging mice, with a possible role of LECs in impacting Tregs (27). Lastly, the expression of a self-antigen in LECs and FRCs induces conversion of CD4⁺ T cells into Tregs (28).

Tumoral LVs are required for the initiation of antitumor T-cell responses because they facilitate tumor–antigen drainage and DC migration from the tumor bed to the tumor-draining LNs (TdLN) where tumor-specific naïve T cells are activated (29). However, cross-presentation of tumor antigens through MHCI molecules by LECs in TdLNs promotes tumor-specific CD8⁺ T-cell apoptosis, and tumor lymphangiogenesis correlates with tumor-infiltrating Tregs (30). Furthermore, the immunosuppressive molecule PD-L1 is upregulated by LECs following antigen-specific interaction with CD8⁺ T cells in vitro (31). In vivo, tumoral LECs express elevated levels of PD-L1 following exposure to IFNγ in the TME (31, 32). Tumoral LECs seem, however, to be beneficial for the response to immunotherapies such as PD1-blocking antibodies (33).

Endogenous MHCII expression in LECs is IFNγ inducible (26), and this cytokine is present at different levels in several tumors. Therefore, we investigated whether and how LECs functioned as MHCII-restricted APCs in the context of tumor development and how this might affect antitumor immunity and tumor growth. First, we demonstrated that in tumors, LECs upregulated MHCII and MHCII-related antigen-presentation machinery and exhibited an increased ability to capture and process exogenous antigens. Using different mouse models of lymphangiogenic tumors, we showed that abrogating MHCII in LECs altered intratumoral Tregs, resulting in increased T cell–mediated antitumor immunity and impaired tumor growth. Altogether, our study places LECs at the forefront of immunoregulatory players in the TME as they function as tolerogenic APCs that inhibit antitumor immunity.

Female wild-type mice (Charles River), Rag2^−/− (Charles River), OT2 (34), ProxCre^ERT2 (35) crossed with MHCII^f/f (MHCII^ΔProx-1) (27) and IFNγR2^f/f (IFNγR2^ΔProx-1) (36), Foxp3^RFPRORγt^GFP (37), and DEREG (38) mice were used at ages 7 to 12 weeks. All mice had a pure C57BL/6 background and were bred and maintained under specific pathogen–free conditions at the animal facility of Geneva Medical School and at Charles River in France. All procedures were approved and performed in accordance with the guidelines of, and with the approval of, the animal research committee of Geneva. ProxCre^ERT2MHCII^f/f (MHCII^ΔProx-1), MHCII^f/f control (MHCII^WT), IFNγR2^ΔProx-1, and IFNγR2^f/f control (IFNγR2^WT) mice were treated intraperitoneally with tamoxifen (cat. #T5648, Sigma-Aldrich), 10 mg/mouse twice a day for 4 days. All procedures were performed 2 weeks after the last tamoxifen injection.

Bone marrow (BM) chimeric mice were generated as previously described (39). Briefly, BM cells from Foxp3^RFPRORγt^GFP or DEREG mice were recovered from tibia and femurs of donor mice, and 5–7 × 10⁶ cells were injected intravenously into sublethally irradiated recipient MHCII^ΔProx-1 and MHCII^WT mice (two consecutive doses of 500 cGy with 4-hour interval). Reconstitution in BM chimeric Foxp3^RFPRORγt^GFP→MHCII^ΔProx-1 and Foxp3^RFPRORγt^GFP→MHCII^WT mice, as well as BM chimeric DEREG→MHCII^ΔProx-1 and DEREG→MHCII^WT mice was assessed by analyzing blood cells by flow cytometry after 6 to 8 weeks (see “Antibodies, flow cytometry, and cell sorting”).

IFNγ (cat. #315-05-250UG, PeproTech) was injected subcutaneously in both flanks (1 μg/50 μL in each flank). CD25-specific antibody (clone PC-61.5.3, cat. #BE0012, Bio X Cell) or isotype control antibody (clone HRPN, cat. #BE0088, Bio X Cell) was administrated intratumorally (50 μg/mouse) in mice on days 11, 14, and 16 after tumor inoculation. DQ Ovalbumin (cat. #D-12053; Molecular Probes) was injected intratumorally (6 μg/mouse). Diphtheria toxin (cat. #D0564-1 MG; Sigma) was administrated intratumorally (0.2 μg/mouse) on days 15 and 17 after tumor inoculation.

A mixed fibroblastic reticular cell (FRC)/LEC culture was performed as previously described (22). In brief, LNs from 1 to 3 mice were dissected and digested with a freshly made enzymatic solution comprising RPMI-1640 (cat. #61870-010, GIBCO) containing 0.8 mg/mL Dispase (cat. #17105-041, GIBCO), 0.2 mg/mL Collagenase P (cat. #11213865001, Roche), and 0.1 mg/mL DNase I (cat. #10104159001, Roche). Tubes containing the digestion mixture were incubated at 37°C in a water bath and gently inverted at 10-minute intervals to ensure the contents were mixed. After 20 minutes, LNs were very gently mixed using a 1-mL pipette. Large fragments were allowed to settle before the supernatant was replaced with fresh digestion mix. Supernatant containing the digested cells was added into 10 mL of cold FACS buffer containing 0.5% BSA (cat. #A1391,0500, AppliChem) and 5 mmol/L EDTA (cat. #A1104,0500, AppliChem) in PBS. These steps were repeated every 10 minutes until all LNs were completely digested. Cells were washed and filtered through a 70-μm cell strainer, quantified using a hemocytometer, and plated in 6-well plates coated with 10 μg/mL human fibronectin (cat. #FCO10-10 MG, Millipore) and Purecol (cat. #50005-100ML, advanced biomatrix) for 30 minutes at 37°C at a concentration of 1 × 10⁷ cells/well. Cell culture media were MEM (cat. #22561-021, Gibco) supplemented with heat-inactivated, 10% batch-tested, low Ig FCS (cat. #10270-106, GIBCO), and penicillin 10,000 U/mL/streptomycin 10,000 μg/mL (cat. #2145449, Gibco). Plates were washed and culture media were renewed every day to remove nonadherent cells. After 5 days, cultures primarily contained LECs and FRCs. OVA protein, OVA AlexaFluor488 (AF488, cat. #O34781), and ovalbumin DQ (cat. #D-12053; all from Molecular Probes; 1 μg/mL) were added to the culture. The frequency of cells having captured OVA-AF488 and containing proteolytic fragments (AF488 from OVA DQ) was assessed at indicated time points by flow cytometry (see “Antibodies, flow cytometry, and cell sorting”) after detaching the cells with Accutase (cat. #00-4555-56, Thermo Fisher) for 5 minutes at 37°C.

All skin LNs (inguinal, axillary, and brachial) were isolated from OT2 mice, scratched and purified using a CD4⁺ T-cell isolation kit (cat. #130-104-454, Miltenyi Biotec), according to the manufacturer's instructions. CD4⁺ T cells were plated in a 24-well plate coated with Ultra-LEAF purified anti-mouse CD3e (17A2, cat. #100238, BioLegend), 1 mg/mL, and Ultra-LEAF purified anti-mouse CD28 (37.51, cat. #102116, BioLegend) 1 mg/mL: 2.5 × 10⁵ cells/well were incubated in complete RPMI: RPMI supplemented with 10% FCS, 100 μmol/L penicillin/streptomycin, 100 mmol/L Sodium pyruvate (cat. #58636, Sigma), 1M Hepes (cat. #15630-056, Gibco), nonessential amino acid 10 × (cat. #11140-035, Gibco) and 50 mmol/L β-mercaptoethanol (cat. #3150-010), containing human TGFβ1 (3 ng/mL; cat. #100-21C-10UG, PeproTech) for 2 days and then transferred to a new plate with IL2 (20 ng/mL; cat. #212-12-20UG, PeproTech) for 3 days. After Treg differentiation, cells were cocultured with LECs (see “Antibodies, flow cytometry, and cell sorting” and “LEC and fibroblastic reticular cell in vitro culture”) at a ratio of 1:1 for 3 days. LECs were treated with IFNγ (2 μg/mL) and loaded with OVA_II peptide (ISQAVHAAHAEINEAGR, Polypeptide; 5 μg/mL) overnight prior to coculture.

For OT2 CD4⁺ T-cell coculture with in vitro LECs, all skin LNs (inguinal, axillary, and brachial) were harvested from OT2 mice, scratched and purified using a CD4⁺ T-cell isolation kit, according to the manufacturer's instructions. LECs were treated with IFNγ (2 μg/mL) and loaded with different doses of OVA_II peptide (5 μg/mL, 250 ng/mL, and 12.5 ng/mL) overnight prior to coculture. OT2 CD4⁺ T cells were cocultured with LECs at a ratio of 1:1 for 7 days under Treg-polarizing conditions (TGFβ at 3 ng/mL and IL2 at 5 ng/mL).

All skin LNs (inguinal, axillary, and brachial) were acquired from OT2 mice, scratched and purified using a CD4⁺ T-cell isolation kit according to the manufacturer's instructions. CD4⁺ T cells were plated with LECs sorted from TdLN and tumor (see “Antibodies, flow cytometry, and cell sorting”) and were cocultured in complete RPMI (see “OT2-induced Treg cocultures with in vitro LECs”) for 3 days. LECs were loaded with OVA_II peptide (5 μg/mL) overnight prior to coculture.

B16-F10 melanoma cells (ATCC) transfected with ovalbumin (B16F10-OVA⁺; gift of B. Huard, Institute for Advanced Biosciences, University of Grenoble-Alpes, Grenoble, France) and B16F10-OVA⁺ cells, engineered to overexpress VEGF-C (B16F10-OVA⁺VEGF-C^hi) as published (30), were maintained in RPMI supplemented with 10% heat-inactivated FCS, 100 μmol/L penicillin–streptomycin, and 50 mmol/L of β-mercaptoethanol. MC38 murine colon adenocarcinoma cells (ATCC) transduced with the lentiviral murine Vegfc cDNA (MC38 VEGF-C^hi) were kindly provided by T. Petrova and maintained in DMEM (cat. #41966-029, Gibco), 10% heat-inactivated FCS, 100 μmol/L penicillin–streptomycin and 50 mmol/L of β-mercaptoethanol. Murine lung adenocarcinoma Kras^−/−p53^−/− tumor cells (40, 41) were kindly provided by T. Jacks' laboratory and maintained in DMEM, 10% heat-inactivated FCS, 100 μmol/L penicillin–streptomycin and 50 mmol/L of β-mercaptoethanol. Cell lines were not authenticated, used by passage 20, and tested negative for Mycoplasma.

BM-derived DCs (BMDC) were generated from BM of C57BL/6 mice. BM cells were recovered from tibia and femurs of mice and cultured, after red cell lysis, for 7 to 9 days in complete RPMI medium (10% heat-inactivated FCS, 50 mmol/L of β-mercaptoethanol, 100 mmol/L sodium pyruvate, and 100 μmol/L of penicillin/streptomycin) supplemented with GM-CSF. GM-SCF supplemented medium is added every 3 days and cells are split after 6 days of culture. OVA protein, OVA AlexaFluor488, and Ovalbumin DQ (all from Molecular Probes; 1 μg/mL) were added to the culture. The frequency of cells having captured OVA-AF488 and containing proteolytic fragments (AF488 from OVA DQ) was assessed at indicated time points by flow cytometry (see “Antibodies, flow cytometry, and cell sorting”).

Mice were anesthetized using isoflurane or a mix of Rompun 2% (Bayer)/Ketamine 10% (Vetoquinol) and their backs were shaved. 0.5 × 10⁶ B16F10-OVA⁺, B16F10-OVA⁺VEGF-C^hi, or MC38-OVA⁺VEGF-C^hi cells were injected in 100 μL of PBS subcutaneously on the back dorsolateral side. Tumor size was monitored every 1 to 2 days using a caliper, and tumor size was calculated by length × width. 0.5 × 10⁶ murine lung adenocarcinoma Kras^−/−p53^−/− tumor cells were injected i.v. and lung tumor nodules were monitored using X-Ray Computed Tomography (Quantum GX microCT imaging system).

Skin, LNs, and tumors were cut into small pieces and digested in RPMI containing 1 mg/mL collagenase IV (cat. #LS004189, Worthington Biochemical Corporation), 40 μg/mL DNase I, and 2% FCS for 30 minutes at 37 °C. Any tissue remaining after 30 minutes was further digested with 1 mg/mL collagenase D (cat. #11088882001, Roche), and 40 μg/mL DNase I and 1% of FCS for 20 minutes at 37°C. The reaction was stopped by addition of 5 mmol/L EDTA and 10% BSA. Samples were further disaggregated through a 70-μm cell strainer and blocked with anti-CD16/32 (cat. #14-0161-86, Invitrogen). Single-cell suspensions were further selected using CD45 microbeads (cat. #120-008-885) and a magnetic bead column separation according to the manufacturer's instructions (Miltenyi Biotec). The negative fraction was used.

For in vitro Treg suppression assays, CD4⁺CD25^hi Tregs were sorted by flow cytometry from tumors and TdLNs (see “Antibodies, flow cytometry, and cell sorting”) and cultured at different ratios with CFSE (1 μmol/L, cat. #C34554, Life Technologies)-labeled naïve CD4⁺CD25^neg T cells (see “Antibodies, flow cytometry, and cell sorting”) in the presence of BMDCs (see “In vitro BM-derived DC generation and culture”) and CD3 antibodies. T-cell proliferation was assessed by flow cytometry after 3 days of coculture.

Mouse tumors were embedded in paraformaldehyde (PFA; cat. #A3813,1000, AppliChem) 4% for 4 hours, overnight in sucrose 30% and then mounted in OCT medium. Five- to 10-μm–thick sections were cut and fixed with PFA 4% for 20 minutes. After washing and permeabilization, the sections were stained overnight at 4°C using a rabbit anti–Lyve-1 (cat. #103-PA50, Reliatech GmbH). Secondary staining was performed using an Alexafluor546-labeled donkey anti-rabbit (cat. #A10040, Life Technologies) and Alexafluor488-labeled anti-Foxp3 (150D, cat. #320012, BioLegend) or eFluor660-labeled anti-Ki67 (SolA15, cat. #50-5698-82, eBioscience), for 2 hours at room temperature. Sections were mounted with DAPI Fluoromont-G (cat. #0100-20, SouthernBiotech). To analyze lymphatic vascularization and subtumoral distribution of Tregs in B16F10-OVA⁺VEGF-C^hi melanomas, ex vivo imaging of individual tumor slices was performed using an upright spinning disk confocal microscope (Axio Examiner Z1 Advanced Microscope Base, Zeiss) equipped with a confocal scanner unit CSU-X1 A1 (Yokogawa Electric Corporation). The fluorescence was detected with an electron-multiplying charge-coupled device camera (EMCCD, Evolve 512 10 MHz Back Illuminated, Photometrics) and a 10 × /0.3 NA or a 40 × /1.0 NA water immersion objective (W Plan Apochromat, Zeiss) upon visualization using three laser-excitation wavelengths (488 nm, 561 nm, and 640 nm; LaserStack v4 Base, 3i) in combination with appropriate band-pass-emission filters (Semrock). Three-dimensional image stacks were obtained by sequential acquisition of multiple field of views along the z-axis using a motorized XY-stage (ProScan, Prior). SlideBook software (6.0.17, 3i) was used for image acquisition and the creation of maximum projections. The subsequent generation of montage images from contiguous positions was performed using the Fiji grid/collection stitching plugin (42).

Human tumors were acquired from untreated six melanoma patients (ethical authorization GE15-092). The study was conducted in accordance with the Declaration of Helsinki. Samples were collected, stored formalin, and imbedded in paraffin according to standard procedures by the diagnosis service of the Dermatopathology Unit at the Geneva University Hospital. Inclusion criteria: biopsies of melanoma or dysplastic nevus diagnosed before 2010, biopsies of patients with ablation of normal nevus, dysplastic nevus, primary melanoma (grade Ia, Ib, Iia, IIb, IIIa, or IIIb as defined by the l'American Joint Committee on Cancer) or metastatic melanoma (grade IIc, IIIc, or IV as defined by the l'American Joint Committee on Cancer), biopsies available in paraffin in the Dermatopathology Unit repository. Exclusion criteria: uncertainty of beta-blockers or other long-term medications. Paraffin-embedded blocks were deparaffinized and hydrated. Sections were stained overnight with primary antibodies, anti-podoplanin/gp38 (clone #D2-40, cat. #ab77854, Abcam), anti-FOXP3 (clone #SP97, cat. #ab99963, Abcam), or anti-CD3 (cat. #ab16669, Abcam) and subsequently mounted with DAPI Fluoromont-G (cat. #0100-20, SouthernBiotech). Images were acquired with a confocal microscope (LSM 700; Carl Zeiss Inc. and SP5; Leica). Quantification was performed by calculating the LV density (D2-40⁺, % of total area) and the amount of stained Foxp3⁺cells/mm² or stained CD3⁺cells/mm². A square-root transformation on count data was applied to get in line with the assumption of normality for regression analysis.

All antibodies are listed in Supplementary Table S1.

For flow cytometry analysis, single-cell suspensions were incubated with FcBlock (anti-CD16/32 FcgRII-RIII) for 10 minutes, at 4°C and stained with antibodies. Intracellular staining (IFNγ, IL17, Foxp3, Ki67, and Granzyme B) was done using the Intracellular Fixation and Permeabilization buffer set (Thermo Fisher): Fixation/permeabilization concentrate (cat. #00-5123-43) and Diluent (cat. #00-5223-56), Permeabilization Buffer 10 × (cat. #00-8333-56). For IFNγ, IL17, and Granzyme B staining, cells were first restimulated in complete RPMI containing PMA 100 ng/mL (cat. #P1585-1MG, Sigma-Aldrich), ionomycin 1 μg/mL (cat. #I0634-1MG, Sigma-Aldrich), and GolgiPlug solution 1/1,000 (cat. #51-2301KZ, BD Biosciences), and incubated 4 hours at 37°C, 5% CO₂.

Cells were acquired on a Fortessa and analyzed using FlowJo (TreeStar) software.

For LEC flow cytometry sorting from LEC/FRC cultures, cells were stained with mAbs against CD45, gp38, and CD31 and sorted using MoFlowAstrios (Beckman Coulter). For LEC flow cytometry sorting from tumors, single-cell suspensions were stained with mAbs against CD45, gp38, and CD31, and LECs were sorted using a MoFlowAstrios (Beckman Coulter) as CD45⁻, gp38⁺, and CD31⁺ cells. For LEC flow cytometry sorting from TdLN, single-cell suspensions were enriched in CD45⁻ cells using CD45 microbeads, stained with mAbs against CD45, gp38, and CD31, and LECs were sorted using a MoFlowAstrios (Beckman Coulter) as CD45⁻, gp38⁺, and CD31⁺ cells.

For some experiments, Tregs were isolated as RFP⁺ cells from tumors and LNs of Foxp3^RFPRORγt^GFP→MHCII^ΔProx-1 and Foxp3^RFPRORγt^GFP→MHCIIWT BM chimeric mice by flow cytometry cell sorting, after enrichment using a CD4⁺ T-cell depletion kit (negative fraction). For some experiments, Tregs were isolated from tumors and LNs as CD4⁺CD25^hi cells by flow cytometry after enrichment using a CD4⁺ T-cell depletion kit.

Naïve CD4⁺CD25^neg T cells were purified using CD4⁺ T-cell depletion kit by adding to the biotinylated-antibody cocktail a biotinylated anti-CD25 (7D4, cat. #553070, BD Biosciences; negative fraction).

Total RNA was isolated using the RNeasy Plus Micro Kit (cat. #74034, Qiagen) from sorted LECs and Tregs (see “Antibodies, flow cytometry, and cell sorting”). cDNA was synthesized using the PrimeScript RT Reagent Kit (cat. #RR037A) followed by a preamplification of our genes of interest using the TaqMan Preamp Master Mix (cat. #4488593, Applied Biosystems). PCRs were performed with the ABI Prism 7900 HT detection system and PowerUp SYBR Green Master Mix (cat. #A25743, Applied Biosystems). Results (two replicates) were normalized with gapdh and l32, normalization factors, and fold changes were calculated according to the GeNorm method (43).

Primers were as follows:

• h2-dm b1–2: fw TAGACGTCCCCGTAGGAAGG, rev CACAGAACGAGAGCGCCA

• i-ab: fw CTGTGGTGGTGCTGATGG, rev CGTTGGTGAAGTAGCACTCG

• cd74: fw CGCCTAGACAAGCTGACCAT, rev AACGTTCTTCACAGGCCCAA

• ctl4: fw GCTTCCTAGATTACCCCT TCT GC, rev CGGGCATGGTTCTGGATCA

• pd1: fw CGTCCCTCAGTCAAGAGGAG, rev GTCCCTAGAAGTGCCCAACA

• ccr8: fw TTCCTCTACTTAGGGAGACAAATGC, rev CATCCAGGGTGGAAGAATGG

• icos: fw TCT AGA CTT GCA GGT GTG ACC, rev CAG GGG AAC TAG TCC ATG CG

• zap70: fw GCATGCGCAAGAAGCAGATT, rev GGGCCTCTCGCATCATCTC

• cd3: fw ATGCGGTGGAACACTTTCTGG, rev GCACGTCAACTCTACACTGGT

• gapdh: fw CCCGTAGACAAAATCGTGAAG, rev AGGTCAATGAAGGGGTCGTTG

• l32: fw GAAACTGGCGGAAACCCA, rev GGATCTGGCCCTTGAACCTT.

For VEGF-C qPCR, an RT2 qPCR Primer Assay was used (Qiagen; cat. #330001 PPM03061F; Reference position: 449).

Flow cytometry–isolated Tregs (see “Antibodies, flow cytometry, and cell sorting”) were collected in RNAprotect Cell Reagent (cat. #76526, Qiagen). RNA was isolated using an RNeasy Plus Micro Kit (cat. #74034, Qiagen), and three to four replicates per condition were used. RNA integrity and quantity were assessed with a Bioanalyzer (Agilent Technologies). cDNA libraries were constructed by the Genomic platform of the University of Geneva as follows: SMART-Seq v4 ultra low input RNA kit (cat. #634891, Clontech) was used for the reverse transcription and cDNA amplification according to the manufacturer's specifications, starting with 1 ng of total RNA. cDNA (200 pg) was used for library preparation using the Nextera XT Kit (cat. #FC-131-1096, Illumina). Library molarity and quality were assessed with the Qubit and Tapestation using a DNA high-sensitivity chip (Agilent Technologies). Pools of 10 libraries were loaded for clustering on a single read Illumina Flow cell. Reads of 50 bases were generated using the TruSeq SBS chemistry on an Illumina HiSeq 4000 sequencer. FastQ reads were mapped to the ENSEMBL reference genome (GRCm38.89) using STAR version 2.4.0j (https://github.com/alexdobin/STAR) with standard settings, except that any reads mapping to more than one location in the genome (ambiguous reads) were discarded (m. 1).

A unique gene model was used to quantify reads per gene. Briefly, the model considers all annotated exons of all annotated protein coding isoforms of a gene to create a unique gene where the genomic regions of all exons are considered coming from the same RNA molecule and merged. All reads overlapping the exons of each unique gene model were reported using featureCounts version 1.4.6-p1 (http://bioinf.wehi.edu.au/featureCounts/). Gene expression was reported as raw counts and in parallel normalized in reads per kilobase million (RPKM) in order to filter out genes with low expression value (1 RPKM) before calling for differentially expressed genes. Library size normalizations and differential gene-expression calculations were performed using the package edgeR (http://bioconductor.org/packages/release/bioc/html/edgeR.html) designed for the R software (http://www.R-project.org/). Only genes having a significant fold change (Benjamini–Hochberg corrected P < 0.05) were considered for the rest of the RNA-sequencing (RNA-seq) analysis.

Gene ontology term and Kyoto Encyclopedia of Genes and Genomes (KEGG) metabolic pathways enrichment were performed using homemade scripts for the R software.

The distance between each pair of samples is the root-mean-square deviation for the top 500 most variable genes. Distances on the plot can be interpreted as leading log₂ fold change, meaning the typical (root-mean-square) log₂ fold change between the samples for the genes that distinguish those samples.

A gene set enrichment analysis (GSEA) yielded a list of upregulated pathways [pathways with FDR lower than 0.05 and abs(NES) (normalized enrichment score) higher than 1] that was then sorted for immunologic relevance: hematopoietic cell lineage, Th1, Th2, Th17 differentiation pathway, cytokine–cytokine receptor interaction, inflammatory response pathway, T-cell receptor (TCR) signaling pathway, NF-κB signaling pathway, cytokines and inflammatory response, IL17 signaling pathway, type II IFN signaling, chemokine signaling, TNF signaling, Jak–Stat signaling, and PD-L1 and PD1 checkpoint pathway in cancer. Genes with a low enrichment score and a fold change lower than 0.05 were excluded. Heat maps display expression levels in ln(1 + RPKM) (reads per kilobase per million) in false colors. For heat maps with z-score: a z-score was applied on each row, to highlight high fold changes.

GEO depository number: GSE168609 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE168609).

Statistical significance was assessed by the two-tailed unpaired Student t test, unpaired, nonparametric t test, and ANOVA using Prism 5.0 software (GraphPad Software).

The data underlying this article are accessible at https://doi.org/10.26037/yareta:3mbcwbt77nakdgbyfkclgxzqvi.

In the B16F10-OVA⁺VEGF-C^hi (B16-OVA⁺VC^hi) lymphangiogenic mouse melanoma model (30), the frequency of LECs (defined as CD45^negCD31⁺gp38⁺) in tumors was increased compared with control parental B16F10-OVA⁺ (B16-OVA⁺) tumors (Fig. 1A). In contrast, the frequency of BECs was similar in both lymphangiogenic and nonlymphangiogenic tumors (Fig. 1A). In addition, VEGF-C overexpression in tumors affected the size of primary tumors, which was increased in lymphangiogenic tumors compared with nonlymphangiogenic tumors (Fig. 1B). Both tumors grew similarly when transplanted into Rag2^−/− immunodeficient mice (Supplementary Fig. S1), indicating that LECs in the TME promote immunosuppression by regulating antitumor adaptive immunity.

We investigated whether LECs function as MHCII-restricted APCs in the B16-OVA⁺VC^hi tumor context. Compared with steady-state skin LECs and LN-LECs, tumoral LECs expressed elevated levels of MHCII mRNA (Fig. 1C). At the protein level, LECs in tumors exhibited enhanced MHCII surface expression compared with skin LECs, TdLN LECs, and nondraining LN (NdLN) LECs (Fig. 1D).

CIITA is the master regulator for MHCII expression (44). LECs express the IFNγ-inducible promoter of CIITA (26). MHCII expression was strongly reduced in tumoral LECs from IFNγR2^ΔProx-1 mice, which lack expression of IFNγR2 in LECs, compared with IFNγR2^WT controls (Fig. 1E), indicating that IFNγ in the TME was promoting upregulation of MHCII by LECs.

Supporting the notion that LECs in B16-OVA⁺VC^hi tumors could function as MHCII-restricted APCs, invariant chain (Cd74) mRNA was increased in tumor LECs compared with skin LECs, although expressed at lower levels compared with LN-LECs (Fig. 1F). In addition, levels of H2-DM mRNA were elevated in LECs from tumors compared with LECs from skin or LNs (Fig. 1F).

LEC proliferation is a prerequisite for antigen acquisition (45). When cultured in vitro, LECs were highly proliferative (Supplementary Fig. S2A) and were superior to BMDCs at capturing exogenous OVA-AF488 antigen (Supplementary Fig. S2B). When incubated with pH-sensitive OVA DQ, which releases fluorescent fragments upon proteolytic digestion, LECs were fluorescent after both 4 and 12 hours, demonstrating that in vitro LECs were capable of targeting exogenous antigens to the late endosomal compartment (Supplementary Fig. S2B). In vivo, LECs in B16-OVA⁺VC^hi tumors proliferated significantly more than skin LECs (Fig. 1G; Supplementary Fig. S2C), likely because of elevated VEGF-C levels in the TME. TdLN LEC expansion was more modest, but still significantly greater than NdLN LEC expansion (Fig. 1G). Intratumoral OVA DQ injection resulted, after 4 hours, in a robust antigen capture and access to late endosomal compartments in LECs from tumors and, to a lesser extent, in LECs from TdLNs (Fig. 1H). These results show that tumoral LECs capture and process exogenous antigens in lymphangiogenic tumors.

We also compared LECs from B16-OVA⁺VC^hi and poorly lymphangiogenic parental B16-OVA⁺ tumors. Neither MHCII nor PD-L1 expression by LECs was altered by overexpression of VEGFC in tumors (Supplementary Fig. S3A). However, LECs proliferated more in B16-OVA⁺VC^hi compared with parental B16-OVA⁺ tumors (60% compared with 40% of proliferating LECs, respectively; Supplementary Fig. S3A). In addition, LECs from both B16-OVA⁺ and B16-OVA⁺VC^hi tumors were able to capture and process OVA DQ exogenous protein (Supplementary Fig. S3B). Altogether, our data indicate that LECs ability to express MHCII, proliferate, and capture and process exogenous Ags in tumors is not dependent on VEGF-C overexpression.

To determine whether the MHCII-restricted antigen-presentation machinery in LECs is functional, we exposed OVA-specific CD4⁺ cultured in Treg-polarizing condition to IFNγ treated–, OVA_II peptide–loaded LECs. LECs efficiently promoted OT2-induced Treg (OT2 iTreg) cell expansion and accumulation, whereas they induced an abortive proliferation of non-iTreg OT2 T cells, with increased T-cell death (Supplementary Fig. S3C). Similar results were obtained when naïve OT-2 cells were cocultured with LECs under Treg-polarizing conditions, in an antigen dose-dependent manner (Supplementary Fig. S3D). We confirmed the results obtained with cultured LECs using LECs sorted from B16-OVA⁺VC^hi tumors or TdLNs. TdLN LECs did not induce OT2 proliferation, whereas tumor LECs promoted the proliferation and/or differentiation of Foxp3⁺, but not Foxp3⁻, OT2 cells (Fig. 1I).

Next, we crossed Prox-1-Cre^ERT2 mice with MHCII^fl/fl mice (MHCII^ΔProx-1 mice) to allow selective deletion of MHCII in LECs upon tamoxifen treatment (27). MHCII^fl/fl mice were used as controls (MHCII^WT). Tamoxifen administration led to abrogation of IFNγ-induced MHCII upregulation in LECs but not BECs and FRCs (Supplementary Fig. S4). In addition, when B16-OVA⁺VC^hi tumor cells were injected into tamoxifen-treated MHCII^ΔProx-1 and MHCII^WT mice, MHCII expression was abrogated only in tumoral LECs from MHCII^ΔProx-1 but not tumoral LECs from MHCII^WT mice, whereas BECs were unaffected (Fig. 2A). We also investigated whether deletion of MHCII affected other LEC characteristics but could not find any differences in frequency, density, or expression of PD-L1 between LECs infiltrating tumors in MHCII^ΔProx-1 and MHCII^WT mice (Supplementary Fig. S5A).

Analysis of B16-OVA⁺VC^hi tumor growth showed a significant decrease in the size of tumors developing in MHCII^ΔProx-1 mice compared with MHCII^WT mice at day 16 [mean tumor size at day 16, 109.4 ± 36.2 mm² and 203.1 ± 7.2 mm², and area under the curve (AUC) 767.7 and 451.9, respectively; Fig. 2B]. There was no difference in the numbers of IFNγ-producing CD4⁺ T cells and IFNγ-producing CD8⁺ T cells in the TdLNs at day 10 (Fig. 2C), suggesting that MHCII abrogation in LECs does not affect the priming of antitumor T cells. In contrast, the densities of tumor-infiltrating IFNγ-producing CD4⁺ effector T cells and IFNγ- and Granzyme B–producing effector CD8⁺ T cells were significantly increased in tumors from MHCII^ΔProx-1 mice compared with MHCII^WT mice (Fig. 2D). Numbers of tumor-infiltrating immune cells DCs, myeloid-derived suppressor cells, and natural killer (NK) cells were not affected by the absence of MHCII in LECs, but neutrophils were more abundant in tumors in which LECs did not express MHCII (Supplementary Fig. S5B).

We also saw a difference in the growth of parental B16-OVA⁺ tumors in when MHCII expression was absent in LECs, although the effect was less pronounced compared with the lymphangiogenic B16-OVA⁺VC^hi tumors (Supplementary Fig. S6). In addition, there was a tendency toward increased effector T-cell densities (CD8⁺IFNγ⁺ and CD4⁺IFNγ⁺ T cells) in tumors from MHCII^ΔProx-1 compared with MHCII^WT control mice (Supplementary Fig. S6). Therefore, in both lymphangiogenic and low lymphangiogenic tumors, MHCII expression by LECs promoted tumor growth and attenuated intratumoral effector T-cell densities.

We validated our findings using an MC38 colon adenocarcinoma cells overexpressing VEGF-C (MC38-VEGF-C^hi, referred to as MC38-VC^hi) that did not express OVA. In vitro VEGF-C mRNA levels in MC38-VC^hi cell cultures were increased around 10-fold compared with the MC38 parental cell line, but were 10-fold less than in B16-OVA⁺VC^hi cells (Supplementary Fig. S7A). We detected LECs in MC38-VC^hi tumors, although the frequency was much lower compared with B16-OVA⁺VC^hi tumors (Supplementary Fig. S7B). MC38-VC^hi tumors grew significantly faster when transplanted into C57BL/6 mice compared with parental MC38 tumors (Supplementary Fig. S7C). Therefore, although LECs are less numerous in MC38-VC^hi compared with B16-OVA⁺VC^hi tumors (Supplementary Fig. S7B), lymphangiogenesis still promoted tumor growth. Tamoxifen treatment abrogated expression of MHCII in tumoral LECs in MC-38-VC^hi tumors transplanted into MHCII^ΔProx-1 mice (Fig. 3A), and did not affect tumoral LEC frequency, density, or expression of PD-L1 (Supplementary Fig. S7B). Furthermore, in MC38-VC^hi tumor–bearing C57BL/6 mice, LECs in tumors and TdLNs proliferated extensively and significantly more compared with LECs in ndLNs and skin (Supplementary Fig. S7D). Although LECs were less abundant in MC38-VC^hi tumors compared with B16-OVA⁺VC^hi tumors, growth of both tumors was significantly reduced in MHCII^ΔProx-1 mice compared with MHCII^WT mice (Figs. 2B and 3B). Effector T-cell responses were unaffected in TdLNs (Fig. 3C). However, the densities of tumor-infiltrating effector CD4⁺IFNγ⁺ T cells and CD8⁺IFNγ⁺Granzyme B⁺ T cells were increased in MHCII^ΔProx-1 mice compared with MHCII^WT mice (Fig. 3D), recapitulating what we observed in B16-OVA⁺VC^hi tumors (Fig. 2D).

Consistent with our data using B16-OVA⁺VC^hi and MC38-VC^hi lymphangiogenic tumors, we found that tumor incidence was reduced in MHCII^ΔProx-1 mice compared with MHCII^WT mice injected with Kras^−/−p53^−/− murine lung adenocarcinoma cells (40, 41), which spontaneously develop lymphangiogenic tumor nodules in lungs (Fig. 3E–G).

Given that we saw that the numbers of tumor-infiltrating CD4⁺ and CD8⁺ effector T cells were increased in MHCII^ΔProx-1 mice and that LECs promotes Treg proliferation in vitro and ex vivo (Supplementary Fig. S3; Fig. 1I), we investigated whether an active mechanism of suppression by Tregs was lost in the absence of MHCII expression by LECs. We first found that Tregs were more abundant in LV-rich compared with LV-free areas of B16-OVA⁺VC^hi tumors and were enriched in proximity to LVs (Supplementary Fig. S8). In thick tissue sections, T cells (CD3⁺) were located proximal to LVs, and T-cell population contained a detectable fraction of Foxp3⁺ Tregs in close proximity to LVs (Fig. 4A). Intensity quantification showed that the Foxp3⁺ signal was largely nuclear overlapped with DAPI, and existed outside and inside LVs (Supplementary Fig. S9). Consistent with these data, in sections of human melanomas, both CD3⁺ T-cell and Foxp3⁺ Treg numbers were enriched in LV-rich compared with LV-free areas (Fig. 4B and C). Foxp3⁺ Treg numbers and LV density exhibited a more significant positive correlation (Fig. 4C), demonstrating that Tregs are particularly enriched in LV-rich areas of human melanoma.

In mice, Foxp3⁺CD25^hi Treg frequencies were increased in B16-OVA⁺VC^hi tumors compared with parental B16-OVA⁺ tumors at day 11 (Fig. 5A). In addition, tumoral LEC and tumor-infiltrating Treg densities positively correlated, with a more significant association at day 10 compared with day 14 after tumor inoculation (Fig. 5B). Treg and CD8⁺Granzyme B⁺ effector T-cell densities negatively correlated, especially at day 14 (Fig. 5C). CD4⁺ effector T-cell and LEC densities also positively correlated, but to a lesser extent compared with the positive correlation between LECs and Tregs (Fig. 5D and E). No positive association was observed between the densities of LECs and IFNγ- or TNFα-producing CD8⁺ T cells (Fig. 5F and G). There was a negative correlation between the densities of LECs and Granzyme B⁺ effector CD8⁺ T cells, especially at day 14 (Fig. 5H). These data indicate that TA-lymphangiogenesis promotes Treg accumulation, whereas it dampens CD8⁺ effector T-cell infiltration in tumors, suggesting that tumoral LECs support an immunosuppressive environment by impacting Tregs.

To test our hypothesis that IFNγ produced by immune cells in tumors promotes MHCII upregulation by LECs, enhancing their ability to function as MHCII-restricted APCs and positively impacting the Treg compartment, we assessed whether Tregs were altered when B16-OVA⁺VC^hi tumors developed in mice lacking MHCII expression in LECs. We found significantly reduced frequency and proliferation of Foxp3⁺CD25^hi Tregs in tumors, but not TdLNs, in MHCII^ΔProx-1 mice compared with MHCII^WT mice (Fig. 6A). In parental B16-OVA⁺ tumors, Treg frequencies tended to be decreased, although not significantly, in MHCII^ΔProx-1 compared with MHCII^WT mice (Supplementary Fig. S6). Together with the trend (not significant) toward increased CD4⁺ and CD8⁺ effector T cells (Supplementary Fig. S6B), the Treg/Teff ratio seems to be disrupted enough in MHCII^ΔProx-1 mice to significantly reduce the growth of B16-OVA⁺ tumors compared with MHCII^WT mice (Supplementary Fig. S6A).

To properly isolate Tregs for RNA sequencing based on their specific marker Foxp3, we generated BM chimeric mice. Tamoxifen-treated MHCII^ΔProx-1 and MHCII^WT mice were irradiated and reconstituted with BM cells from Foxp3^RFPRORγt^GFP mice. BM chimeric Foxp3^RFPRORγt^GFP→MHCII^WT and Foxp3^RFPRORγt^GFP→MHCII^ΔProx-1 mice were injected with B16-OVA⁺VC^hi tumors. At day 12, RFP⁺Tregs were purified from tumors and TdLNs (Supplementary Fig. S10A). Two-dimensional scaling analysis revealed that Tregs from MHCII^WT mice clustered separately when sorted from either TdLNs or tumors (Fig. 6B), suggesting that they exhibit a distinct phenotype. Detailed analysis showed that in MHCII^WT mice, genes implicated in Treg-suppressive functions were upregulated in tumoral Tregs compared with TdLN Tregs. The tumor Tregs showed a distinct signature, with an upregulation of genes such as Ccr8, Ccr, Il12rb1, Irf4, Il1r2, and Il21r and genes associated with TCR signaling (Fig. 6C; Supplementary S10B). These observations are consistent with previous studies showing that upon migration from LNs to tumors, Tregs are exposed to a transcriptional program that is the combination of tissue adaptation and intratumoral signature (6, 7, 46).

TdLN Tregs from Foxp3^RFPRORγt^GFP→MHCII^ΔProx-1 and Foxp3^RFPRORγt^GFP→MHCII^WT mice demonstrated similar gene-expression patterns, with minor changes in the above-mentioned genes (Fig. 6C; Supplementary S10B). These observations confirm that, as suggested by comparable effector T-cell responses in TdLNs in MHCII^ΔProx-1 and MHCII^WT mice (Fig. 2C), Tregs in TdLNs were not affected by the loss of MHCII expression in LECs. In contrast, Tregs isolated from tumors of Foxp3^RFPRORγt^GFP→MHCII^ΔProx-1 clustered separately from Tregs from Foxp3^RFPRORγt^GFP→MHCII^WT tumors and did not upregulate genes crucial for Treg-suppressive function (Fig. 6B and C; Supplementary S10B). Lta and Ltb genes, which were upregulated in Tregs in tumors compared with Tregs from TdLNs in control mice, were downregulated in tumor Tregs from Foxp3^RFPRORγt^GFP→MHCII^ΔProx-1 mice (Fig. 6C; Supplementary S10B).

Using qRT-PCR analysis on FoxP3^RFP+ Tregs sorted from tumors and TdLNs of FoxP3^RFP+→MHCII^WT and FoxP3^RFP+→MHCII^ΔProx-1 BM chimeric mice, we confirmed that mRNA levels of Pd1, Icos, Cccr8, Ctla4, Cd3, and Zap70 were increased in tumoral Tregs compared with TdLN Tregs in control mice and significantly decreased in tumoral Tregs from FoxP3^RFP+→MHCII^ΔProx-1 BM chimeras compared with tumoral Tregs from controls (Fig. 6D). Flow cytometry further confirmed that key markers implicated in Treg-suppressive functions were upregulated in tumor-infiltrating Tregs compared with TdLN Tregs in control animals (Fig. 7A). In addition, CD103, which is also implicated in Treg-suppressive functions, was upregulated in tumor Tregs (Fig. 7A). Apart from PD1, which was similarly expressed by tumoral Tregs from both groups, the expression levels of these markers remained generally lower in tumoral Tregs from MHCII^ΔProx-1 mice compared with tumoral Tregs from MHCII^WT mice; some were significantly reduced (ICOS, CD25, and CTLA-4), whereas others showed a trend for reduced expression (CD103 and CCR8; Fig. 7A). Altogether, our data indicate that the absence of MHCII expression by tumoral LECs locally alters the phenotype of Tregs, which express lower levels of markers important for their tumor signature, TCR activation, and suppressive function.

We investigated whether alterations of the tumoral Treg phenotype in the absence of MHCII expression in LECs affected tumor growth by depleting tumoral Tregs in B16-OVA⁺VC^hi tumor–bearing MHCII^ΔProx-1 and MHCII^WT mice using intratumoral injection of a CD25-depleting antibody. Treg depletion efficacy was variable between tumors (Supplementary Fig. S11A). Tumor growth in mice injected with isotype control antibody recapitulated what was previously observed, with MHCII^ΔProx-1 mice developing smaller tumors compared with MHCII^WT mice (Supplementary Fig. S11B). Treg depletion did not affect tumor growth in MHCII^ΔProx-1 mice, but the size of the tumors in MHCII^WT mice treated with CD25-depleting antibody was reduced to the size observed in MHCII^ΔProx-1 mice (Supplementary Fig. S11B), indicating that the tumor size difference observed in nondepleted animals was due to the Treg population. This also suggested that Treg-mediated suppression was impaired in the absence of MHCII expression in tumoral LECs. Consistent with these data, densities of tumor-infiltrating IFNγ-producing CD4⁺ effector T cells and IFNγ- and Granzyme B–producing CD8⁺ effector T cells were similar in MHCII^ΔProx-1 and MHCII^WT mice depleted of Tregs (Supplementary Fig. S11C).

To further validate the importance of Tregs, we used a second approach to specifically delete Tregs. BM cells from DEREG (pFoxP3 DTR-eGFP) mice were transferred into MHCII^WT and MHCII^ΔProx-1 mice, so that upon diphtheria toxin injection (DT) Tregs would be deleted (Supplementary Fig. S12). Tumoral Treg depletion was more homogeneous compared with the protocol using CD25-depleting antibody, with reduction in tumoral Tregs of ∼70% 18 hours after intratumoral DT injection (Fig. 7B). Treg frequencies in TdLNs were not affected by intratumoral DT injection (Fig. 7B). Untreated mice exhibited the tumor size difference we expected, whereas intratumoral DT injection reduced tumor growth to a similar extent in both groups (Fig. 7C). This demonstrated that the difference in tumor growth between untreated MHCII^ΔProx-1 and MHCII^WT mice was due to Tregs. Consistent with these data, densities of tumor-infiltrating IFNγ-producing CD4⁺ effector T cells and IFNγ- and Granzyme B–producing CD8⁺ effector T cells were increased in DT-treated MHCII^WT injected mice compared with the noninjected mice (Fig. 7D). In MHCII^ΔProx-1 mice, Treg depletion did not affect the elevated effector T-cell densities in tumors. Altogether, these experiments indicate that differences in effector T-cell responses seen in nondepleted Treg MHCII^ΔProx-1 and MHCII^WT animals were a consequence of impaired Treg-suppressive functions and not due to a direct effect on CD4⁺ effector T cells.

To further assess whether Treg-suppressive functions were intrinsically altered in tumors in the absence of MHCII expression in LECs, Tregs isolated from tumors and TdLNs of MHCII^ΔProx-1 and MHCII^WT mice were cultured at different ratios with CFSE-labeled naïve CD4⁺ T cells in the presence of anti-CD3 (Fig. 7E). Tregs from TdLNs in MHCII^ΔProx-1 and MHCII^WT mice exhibited similar ability to suppress naïve CD4⁺ T-cell proliferation (Fig. 7E), further confirming that a lack of MHCII in LECs in LNs does not affect the Treg population and does not alter their immunosuppressive functions. Tregs from MHCII^WT tumors, however, were much more potent at suppressing CD4⁺ T-cell proliferation compared with LN Tregs (Fig. 7E). Tregs from tumors in MHCII^ΔProx-1 mice were much less efficient at inhibiting CFSE-labeled CD4⁺ T-cell proliferation compared with Tregs from MHCII^WT tumors (Fig. 7E), demonstrating that whereas the function of Tregs was not affected in LNs of MHCII^ΔProx-1 mice, MHCII expression by LECs was required locally in tumors for Tregs to fully exhibit suppressive functions.

LVs developing in the intratumoral and peritumoral zone influence tumor growth in many ways. First, circulating VEGF-C levels (47), as well as lymphangiogenesis at tumor distal sites (48, 49), correlate with metastasis and bad prognosis. In addition, LECs lining the LVs affect primary tumor growth by modulating antitumor immunity. Although LECs are essential in the initiation of tumor-specific T-cell responses (50), tumoral LECs actively contribute to an immunosuppressive TME (30, 32, 50). Studies suggest, however, that immunotherapeutic approaches can overcome tumoral LEC immunosuppressive functions and that lymphangiogenesis may be beneficial for immunotherapies (33, 51).

LECs also play a critical role in the maintenance of peripheral self-reactive T-cell tolerance, acting as a brake against autoimmune attacks. Indeed, they endogenously express peripheral tissue–restricted antigens and present them through MHCI molecules to induce CD8⁺ T-cell deletional tolerance (21, 23, 25, 52). Although LEC expression of MHCII is low at steady state (26), research suggests that MHCII⁺ LECs function as a brake to prevent autoimmunity in elderly mice by promoting Treg-suppressive functions (27). However, whether LECs shape CD4⁺ T-cell responses as MHCII-restricted APCs, in particular during tumor development, is not known.

Using lymphangiogenic version of mouse melanoma (B16-OVA⁺VC^hi), we have shown here that tumoral LECs have increased levels of MHCII expression compared with skin LECs or LN-LECs. In mice, MHCII expression by LECs is controlled by promoter IV of CIITA, the master regulator for MHCII expression (26), which is inducible by IFNγ (44). Using mice in which IFNγR was genetically abrogated in LECs, we showed that tumoral LECs upregulate MHCII, H2-DM, and CD74 in response to IFNγ, suggesting these cells could be capable of presenting antigens through MHCII. In agreement with previous observations showing that LEC proliferation is a prerequisite for their capacity to capture antigens (45), we found that LECs in tumors were highly proliferative, exhibited an enhanced ability to capture exogenous antigens, and could process internalized antigens. This contrasts with the prior study (45), in which LECs are shown to be incompetent in processing and presenting peptides through MHCII in a vaccination context. Discrepancies between that study and our data could be explained by the extensive proliferation of LECs in tumors, compared with lower frequencies of proliferation (<20%) in LN-LECs following vaccination (45). Therefore, the amount of antigen acquired by LECs in tumors might be increased compared with the vaccination setting. In our study, LECs isolated from tumors and antigen-loaded promoted antigen-specific Treg proliferation. Altogether, our data suggest that tumoral LECs can act as MHCII-restricted APCs.

Lymphangiogenic tumor growth was significantly dampened in mice in which MHCII had been abrogated in LECs, supporting a protumorigenic role for MHCII⁺ LECs. The fact that MHCII molecules were upregulated by LECs in tumors, and not so much by LECs in TdLNs, could reflect either different levels of LEC sensitivity to IFNγ or different ranges of IFNγ concentrations in the two different organs. In both cases, LECs should function as MHCII-restricted APCs locally in tumors. Consistent with this hypothesis, we found that antitumor T-cell effector responses were significantly enhanced in tumors, but not affected in TdLNs, of mice in which LECs lacked MHCII. We recapitulated key results in poorly lymphangiogenic parental B16-OVA⁺ tumors—tumoral LECs expressed MHCII, and captured and processed exogenous antigens—suggesting that the MHCII-restricted antigen-presenting ability of tumoral LECs was not dependent on VEGF-C overexpression. The impact of a lack of MHCII in LECs on antitumor T-cell responses was, however, less pronounced in B16-OVA⁺ compared with B16-OVA⁺VC^hi tumors.

Our results indicate that differences in tumor-specific effector T-cell responses in MHCII^ΔProx-1 and MHCII^WT mice were due to alterations of the intratumoral Treg population. First, in both murine and human melanoma, LV densities positively correlated with the number of Tregs infiltrating the tumors. Second, both Treg frequencies and proliferation were reduced in tumors in MHCII^ΔProx-1 mice. Lastly, intratumoral Treg depletion abolished the differences in tumor growth between MHCII^ΔProx-1 and MHCII^WT mice. In addition, CD4⁺ and CD8⁺ effector T-cell densities in tumors from Treg-depleted MHCII^WT mice increased to reach levels comparable to MHCII^ΔProx-1 mice. However, intratumoral Treg depletion in MHCII^ΔProx-1 mice did not affect effector T-cell densities, suggesting that Tregs were already dysfunctional in the absence of MHCII expression in LECs. RNA-seq data demonstrated that Treg transcriptomics were profoundly affected in tumors, but not in TdLNs, by the loss of MHCII in LECs. The tumor-specific Treg gene signature, which is conserved across species and tumor types (53), was upregulated in MHCII^WT mice in intratumoral Tregs compared with TdLN Tregs. However, this signature, together with genes important in Treg adaptation and differentiation (7, 54, 55), and TCR signaling were not upregulated in intratumoral Tregs from MHCII^ΔProx-1 mice. Genes important for Treg-suppressive functions were downregulated in intratumoral Tregs from MHCII^ΔProx-1 mice, and this was confirmed at surface protein levels, except for PD1. A posttranscriptional regulatory pathway controls PD1 expression (56), which likely explains why we observed differences between mRNA and protein expression levels. ICOS, CD25, CTLA-4, and CD103 protein expression, when taken individually, were significantly but not dramatically reduced in tumors from MHCII^ΔProx-1 compared with MHCII^WT mice. We hypothesize that the lack of MHCII expression by LECs in tumors induces a broad spectrum of changes in Treg phenotype. Therefore, it is likely that impairments of Treg functions observed in MHCII^ΔProx–1 mice do not result from changes in one or the other marker, but more from a global alteration of the Treg phenotype. Altogether, we show that absence of MHCII in tumoral LECs results in altered TCR signaling in Tregs, which is likely due to impaired Ag presentation by LECs, and consequently a global alteration of the Treg-suppressive phenotype.

With respect to the mechanisms explaining why MHCII⁺ LECs in tumors would selectively alter Tregs and not other CD4⁺ T-cell populations, two scenarios can be envisioned. Either MHCII-mediated antigen presentation by tumoral LECs is necessary to maintain a tumor-specific gene signature in already differentiated Tregs or some naïve CD4⁺ T cells, which can be found and locally primed in tumors (57, 58), differentiate into Tregs upon LEC encounter. The fact that PD1 expression by Tregs is not affected by the loss of MHCII in LECs at the protein level might support the first hypothesis, that LECs, which express higher levels of PD-L1 in the TME, preferentially establish antigen-specific interactions with already committed/activated PD1^hi Tregs. Alternatively, the cross-talk between Tregs and LECs could be promoted through LTα1β2–LtβR pathway, which promotes Treg–LV interactions and suppressive functions of Tregs (59). Consistent with this idea, we found that Lta and Ltb were upregulated by intratumoral Tregs compared with TdLN Tregs. Whatever the mechanism, Tregs isolated from tumors in MHCII^ΔProx-1 mice exhibited an impaired ability to suppress the proliferation of naïve T cells ex vivo compared with Tregs from MHCII^WT tumors.

B cells, macrophages/monocytes, NK cells, and DCs infiltrating tumors did not seem affected by the loss of MHCII expression by LECs, although we did not precisely characterize their activation state. In contrast, neutrophil numbers were elevated in MHCII^ΔProx-1 tumors. This could be because of the alteration of the Treg compartment, since it has been reported that Tregs limit neutrophil responses in skin (60). Further work will decipher whether the phenotype of tumor-infiltrating neutrophils is altered following MHCII abrogation in LECs and determine whether it is a consequence of Treg alterations.

Altogether, our study broadens the scope of LECs acting as immunomodulatory APCs during tumor development. We show that tumoral LECs target antigens to the MHCII-restricted antigen-presentation machinery to promote immunosuppression. Together with PD-L1 (32), MHCII are upregulated by tumoral LEC in the presence of IFNγ. Therefore, it is tempting to speculate that both PD-L1– and MHCII-mediated LEC immunosuppression will be enhanced in IFNγ-rich immunogenic tumors, providing a way for lymphangiogenic tumors to evade antitumor immunity.

No disclosures were reported.

A.O. Gkountidi: Data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. L. Garnier: Data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. J. Dubrot: Conceptualization, data curation, formal analysis, investigation, visualization, methodology, writing–review and editing. J. Angelillo: Data curation, formal analysis, investigation. G. Harlé: Data curation, formal analysis, investigation, methodology. D. Brighouse: Formal analysis, investigation. L.J. Wrobel: Investigation, visualization, methodology. R. Pick: Formal analysis, investigation. C. Scheiermann: Resources, data curation, investigation, visualization, methodology, writing–review and editing. M.A. Swartz: Conceptualization, resources, funding acquisition, investigation, visualization. S. Hugues: Conceptualization, resources, data curation, supervision, funding acquisition, investigation, visualization, writing–original draft, project administration, writing–review and editing.

The authors thank J.P. Aubry-Lachainaye, C. Gameiro, and G. Schneiter for excellent assistance in flow cytometry. They also thank S. Lemeille for RNA-seq data analysis, T. Petrova for valuable discussions and for providing MC38 VEGF-C^hi, D. Pejoski and A. Marti Lindez for valuable discussions and/or technical help, and H. Boehncke for providing human melanoma samples. This work was supported by the Leenaards Foundation (S. Hugues and M.A. Swartz), the Swiss Cancer Research foundation (KFS-3950-08-2016-R, to S. Hugues), and the Swiss National Science Foundation (310030_16654, 310030_185255, to S. Hugues).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.