MicroRNA-155 Expression Is Enhanced by T-cell Receptor Stimulation Strength and Correlates with Improved Tumor Control in Melanoma

High numbers of tumor-infiltrating memory CD8⁺ T cells are associated with better clinical outcome in different types of human cancers (1). Nonetheless, tumors manage to evade immune control through multiple mechanisms (2). Therefore, efforts are being made to develop new cancer immunotherapies aiming at boosting the antitumor immune response. For instance, reinfusion of ex vivo–expanded autologous tumor-reactive T cells has shown encouraging results in malignant melanoma patients (3, 4). However, trafficking, survival, and persistence of reinfused T cells need to be optimized, and genetic engineering of T cells is an approach currently explored to improve T-cell qualities. In particular, genetic manipulation of specific microRNAs may allow improvement of T-cell fitness (5).

microRNAs are ∼22nt long noncoding RNAs that posttranscriptionally regulate protein expression. miRNAs can target hundreds of mRNAs and regulate different biological processes, including immune cell differentiation (6). Work from our group showed that miRNAs, and in particular miR-155, are regulated during human CD8⁺ T-cell differentiation and thus may play a role in CD8⁺ T-cell function (7). Studies using miR-155-deficient murine T cells have shown that miR-155 is required for effector CD8⁺ T-cell responses against viruses and tumors (8, 9). Furthermore, overexpression of miR-155 enhanced murine CD8⁺ T-cell antitumor responses (8), suggesting that miR-155 influences fitness of tumor-infiltrating CD8⁺ T cells (CD8⁺ TIL). However, little is known about the key parameters that regulate miR-155 expression and the role of miR-155 in CD8⁺ T cells from cancer patients.

In this study, we measured miR-155 expression in CD8⁺ T cells isolated from tumor-infiltrated lymph nodes (TILN) and tumors from melanoma patients and murine models and compared them with expression in CD8⁺ T cells from non-tumor–infiltrated tissues, peripheral blood (PB), or murine lymphoid tissues. We observed that sustained antigen recognition leads to affinity-dependent miR-155 upregulation within the tumor that correlates with increased CD8⁺ T-cell infiltration in TILNs and tumor control in mice. Moreover, low expression of miR-155 targets in melanoma tumors is associated with prolonged overall survival.

C57BL/6 (B6) mice were obtained from Envigo, OT-1 mice from The Jackson Laboratory (stock no 003831), OT-3 mice from D. Zehn (10), and P14 mice from Prof. Annette Oxenius (Institute of Microbiology, ETH Zurich, Switzerland). Mice were females, 7 to 10 weeks old, and maintained in conventional facilities of the University of Lausanne. This study was approved by the Veterinary Authority of the Swiss Canton Vaud (authorization n°1850), an Institutional Animal Care and Use Committee, and performed in accordance with Swiss ethical guidelines.

B16.N4 and B16.T4 cell lines were generated as previously described (11). All cell lines were tested for Mycoplasma by PCR analysis with specific primers for gpo-1 (5′-ACTCCTACGGGAGGCAGCAGTA-3′) and msgo (5′-TGCACCATCTGTCACTCTGTT-AACCTC-3′) before engraftment. Cells underwent a maximum of two passages from thawing to the time of engraftment.

A total of 10⁵ B16.OVA or 2 × 10⁵ B16.N4 and 2 × 10⁵ B16.T4 cells were subcutaneously (s.c.) engrafted on each flank of B6 mice. After 6 days, CD45.1 10⁵ OT-1 or 10⁶ OT-3 T cells were intravenously (i.v.) transferred. One day later, mice were s.c. vaccinated with 10 μg SIINFEKL (N4) peptide (Protein and Peptide Chemistry Facility, UNIL) and 50 μg CpG (ODN 1826, U133-L01A; Trilink Biotechnologies).

B16.N4 and B16.T4 tumor–bearing mice received 200 μg of anti–PD-1 (clone RMP-1-14, rat IgG2a, BE0146, Bio X Cell) or 2A3 isotype control (Rat IgG2a, BE0089, Bio X Cell) on days 10, 13, and 16 after tumor engraftment.

Spleens and tumors were harvested 14 and 21 days after tumor engraftment. Spleens were mashed through a 70-μm diameter filter (041-352350, Milian), and red blood cells were lysed with RBC lysis buffer (158904, Qiagen). Tumors were dissociated with the Tumor Dissociation Kit (130-096-730, Miltenyi Biotec) following the manufacturer's instructions.

P14 cells were isolated from spleens of P14 mice with EasySep Mouse T-cell isolation kit (Stem cell, 19851A) and activated with 1:1 anti-CD3/CD28 Dynabeads (Thermo Fisher, 11452D) in mouse T-cell media [10% fetal calf serum (FCS, Dominique Dutscher, S1810-500), penicillin (50 IU/mL)/streptomycin (50 μg/mL; Gibco, 15070-063), 0.05 mmol/L ß-mercaptoethanol (Gibco, 31350-010), 10 mmol/L HEPES (Amimed, 5-31F00-H), 2 mmol/L l-glutamine (Amimed, 5-10K100-H), 1 mmol/L sodium pyruvate (Gibco, 11360-039), and 1× nonessential amino acids (Thermo Fisher, 11140-035) in RPMI-1640 with Glutamax (Gibco, 61870-010)] with recombinant human—(rh)IL2 (50 U/mL, Glaxo-IMB). Two days after activation, P14 cells were transduced with MDHI-SCR-GFP or MDHI-miR-155-GFP retroviral vectors (provided by Dr. Gattinoni). One day later, media were replaced by IL2 (10 U/mL), rhIL7 (10 ng/mL, PeproTech, 200-07), and rhIL15 (10 ng/mL, PeproTech, 200-15) containing mouse T-cell media. Five days after activation, media with IL7/IL15 (10 ng/mL) was added, and 2 days later, GFP⁺ P14-SCR/miR-155 cells were sorted by flow cytometry using the Aria II or Aria III flow cytometers (BD Biosciences). Sorted cells were washed 2× with PBS and transferred to mice.

Naïve CD45.1 P14 T cells (10³), P14-SCR-GFP⁺ (5 × 10³), or P14 miR-155-GFP⁺ cells (5 × 10³) were i.v. transferred to CD45.2 B6 mice. One day later, mice were i.v. infected with 2 × 10⁶ pfu of wild-type (WT) LCMV cl13 or 1:3 mixture of WT cl13 and A3 cl13 strains (mixed infection). Both LCMV strains were provided by D. Zehn.

Blood 1/5 diluted in cDMEM (10% FCS (Dominique Dutscher, S1810-500), penicillin (50 IU/mL)/streptomycin (50 μg/mL) (Gibco, 15070-063), 0.05 mmol/L ß-mercaptoethanol (Gibco, 31350-010), 10 mmol/L HEPES (Amimed, 5-31F00-H), in DMEM with Glutamax (Gibco, 31966-021)) was mixed with Vero cells (1.6 × 10⁵). After a 3-hour incubation at 37°C, 2× DMEM methylcellulose was overlaid, and cells were further cultured at 37°C. Seventy-two hours later, cells were fixed with 4% paraformaldehyde (PFA) and permeabilized with 1× Triton. Cells were stained with rat anti-LCMV nucleoprotein (clone VL-4; BE0106, Bio X Cell) for 1 hour at room temperature and goat anti-rat-HRP (112-035-167, Jackson ImmunoResearch) for another 1 hour at room temperature. 0.4mg/mL O-Phenylenediamine dihydrochloride, 0.4mg/mL urea hydrogen peroxide in 0.05mol/L phosphate-citrate, pH 5.0 (OPD; P9187, Sigma-Aldrich) was added for 15′ at room temperature to reveal the plaques.

Splenocytes were stimulated with 5 μg/mL KAVYNFATM peptide or 10 ng/mL phorbol 12-myristate 13-acetate (PMA) (P1585, Sigma-Aldrich) + 500 ng/mL ionomycin (I0634, Sigma-Aldrich) as positive control for 30 minutes at 37°C in cDMEM. Unstimulated splenocytes were used as negative controls. BD GolgiPlug and GolgiStop (51-2301KZ and 554724, BD Biosciences) was added to the cells and incubated for another 4 hours at 37°C before cell staining.

Blood from healthy donors (HD) was drawn at the Blood Transfusion Center, Lausanne, under the approval of the Lausanne University Hospital's Institute Review Board and peripheral blood mononuclear cells (PBMC) were isolated by density gradient centrifugation on Lymphoprep (1114547, Axis-Shield PoC AS).

Total CD8⁺ T cells from HDs PB leucocytes were obtained by magnetic bead enrichment with the human CD8⁺ T-cell isolation kit (130-096-495, Miltenyi Biotec) or CD8 Dynabeads (11147D, Life Technologies) following the manufacturer's instructions.

Liquid nitrogen–stored PB lymphocytes, noninfiltrated lymph nodes (NLN), TILNs, and tumors of malignant melanoma patients were thawed in DNase I (10 μg/mL, 18047019; Thermo Fisher) cRPMI (10% FCS, antibiotic mixture [penicillin (5 mg/mL), streptomycin (5 mg/mL), neomycin (10 mg/mL, Invitrogen)], 0.05 mmol/L 2ß-mercaptoethanol (Invitrogen) in RPMI medium (Invitrogen).

Written informed consent in accordance with the Declaration of Helsinki was obtained from all healthy subjects and patients and the use of patients' samples was approved by the Commission Cantonal d'ethique de la recherche sur l'etre humain (CER-VD; BIL, Dermatology Biobank, protocol 87/06 or LUD00-018). Melanoma patients' clinical data are found in Supplementary Table S1.

Sorted naïve (N) CD8⁺ T cells (1.3 × 10⁵) from HDs were activated with 1:2 anti-CD3/anti-CD28 coated Dynabeads (111.31D, Thermo Fisher) in IL2 (50 ng/mL), human T-cell media [8% human serum, penicillin (100 U/mL, Gibco), streptomycin (100 μg/mL, Gibco), kanamycin (0.1 mg/mL, Gibco), 2 mmol/L l-glutamine (Gibco), 1% in volume nonessential amino acids (Gibco), 1 mmol/L Na pyruvate (Gibco), 0.05 mmol/L ß-mercaptoethanol (Sigma) in RPMI medium (Gibco)].

A total of 50,000 HLA-A2/Melan-A ^MART-1_26-35 (EAAGIGILTV)-specific CD8⁺ T-cell clones were in vitro stimulated with HLA-A2/Melan-A^MART1 _{26-35 (A27L)} (ELAGIGILTV) multimers (1 μg/mL) in human T-cell media for 24 hours.

T-cell avidity of Melan-A^MART-1_26-35 CD8⁺ T-cell clones was measured by NTAmer staining as previously described (12). Briefly, CD8⁺ T-cell clones were stained in 50 μL of 0.2% BSA, 5 mmol/L EDTA in PBS for 45 minutes at 4°C with HLA-A*0201 NTAmers loaded with Melan-A ^MART-1_{26-35 (A27L)} peptide containing Cy5-labeled monomers multimerized through streptavidin–PE. After washing at 4°C, cells were resuspended in 2% FBS in PBS, and dissociation kinetics was assessed at 4°C using LSR-II flow cytometer (BD Biosciences). Upon 30 seconds of acquisition (baseline), imidazole (100 mmol/L) was added and Cy5 fluorescence was measured during 5 to 10 minutes.

Human samples were stained with anti-human: CD3-AlexaFluor700 (cl.UCHT1, BioLegend) CD4-PE (cl.RPA-T4, BD Biosciences), CD8-APC-Alexa750 (cl.B9.11, Beckman Coulter), CCR7-BrilliantViolet421 (cl.G043H7, BioLegend), CD45RA-ECD (cl.2H4LDH11LDB9, Beckman Coulter), and PD1-BV421 (cl.EH12.2H7, BioLegend).

Murine samples were stained with CD3-A700 (cl.17A2, eBioscience), CD8-PE-TexasRed (cl.MCD0817, Life Technologies), CD45.1-FITC (cl.A20.1, FACS facility UNIL), CD45.2-APCeF780 (cl.104, eBioscience), PD-1-APC (cl29F.1A12, eBioscience), CD44-PacificBlue (cl.IM781, FACS facility UNIL), and CD62L-PE-Cy5 (cl.MEL-14, eBioscience) after H2-D^b/LCMV gp276-286 (SGVENPGGYCL) multimer staining when necessary. For cytokine staining, cells were fixed with fixation buffer (420801, BioLegend) and permeabilized with intracellular staining permeabilization wash buffer (421002, BioLegend). Cells were then stained with IFNγ-PerCp-Cy5.5 (cl.XM61.2, eBioscience), TNFα-Pacific Blue (cl. MP6-XT22, BioLegend), IL2-PE-Cy7 (cl. JES6-SH4, eBioscience), and granzyme B-PE-TexasRed (GRB17, Molecular Probe).

LIVE/DEAD Aqua fluorescent reactive dye (Invitrogen) was used for dead cell discrimination.

All stainings were performed in darkness at 4°C for 20′ in 2% FCS, 2 mmol/L EDTA in PBS (sorting buffer), except the staining with Vivid, which was performed in PBS and intracellular staining in intracellular staining permeabilization wash buffer (421002, BioLegend). Multimer staining was performed in darkness at 4°C for 60′ in sorting buffer.

Samples were analyzed with LSRI-II flow cytometer (BD Biosciences) or sorted with Aria III (BD Biosciences) flow cytometry cell sorter.

Total RNA was extracted with the mirVana kit (AM1561, Ambion). Mature human miRNA-155 (hsa-miR-155), miRNA-21 (hsa-miR-21), and RNU44 or mature mouse miR-155 (mmu-miR-155) and snoRNA202 were individually reverse transcribed (30 minutes at 16°C, 30 minutes at 42°C, and 5 minutes at 85°C) with TaqMan RT MicroRNA Kit (4366597, Applied Biosystems) following manufacturer's instructions.

cDNA was amplified using TaqMan Fast Universal PCR Master Mix, No AmpErase UNG (4352042, Thermo Fisher) and hsa-miR-155, hsa-miR-21, RNU44, mmu-miR-155, and snoRNA202-specific TaqMan primers (002623, 000397, 001094, 002571, and 001232; Applied Biosystems) in MicroAmp Fast Optical 96-Well Reaction Plates (4346906; Applied Biosystems) on a 7500 Fast Real-Time PCR System (Applied Biosystems; 40 cycles of 15 seconds at 95°C and 60 seconds at 60°C). RNU44 Ct of human samples and snoRNA202 Ct of mouse samples was subtracted to hsa-miR-155 and hsa-miR-21 Ct and mmu-miR155 Ct, respectively, to calculate relative expression (ΔCt). Expression fold change relative to naïve CD8⁺ T cells from spleen or blood for murine and human samples, respectively, was calculated using the following formula:

2^{−(ΔCt sample−ΔCt reference)}. All statistical tests were performed in non-normalized ΔCt data.

Melanoma patients' formalin-fixed paraffin-embedded (FFPE) tissue sections were incubated for 90 minutes at 45°C in standard hybridization solution with 100 nmol/L of dual terminal tag miR-155 (5′ FAM-tT+TA+AT+GCT+AAT+CGT+GAT+AG+GG+GTt-3′ FAM) and U6 (5′-biotin-tCGTGTATCCTTGCGCAGGGGCCATGCTAATCTTCTCTGTt-3′biotin) probes and washed 3 times with 0.5 × SSC for 10 minutes at hybridization temperature. Probe signal was revealed by consecutive rounds of horseradish peroxidase–mediated deposition of tyramide-conjugated fluorescein (miR-155 probe) and tyramide-conjugated rhodamine (U6 probe) using standard protocol on an automated Dako Autostainer Link 48 (Dako North America, Inc.). Multiplex IHC assay for CD8 and Ki-67 was performed on consecutive tissues to ISH assay as described (1). Briefly, slides were pretreated with standard program with citrate buffer pH = 6 for heat-induced epitope retrieval on Dako PT Link. Primary antibody signal [1:200 dilution of mouse anti-CD8 (4B11), Bio-Rad, MCA1817; 1:400 dilution of rabbit anti-Ki-67 (SP6), Spring Bioscience, M3064] was revealed with appropriate combination of HRP-conjugated anti-host species goat secondary antibodies and tyramide-conjugated dyes as above. All tissue slides were counterstained with DAPI. Images were acquired and multispectrally separated using the Vectra Automated Quantitative Pathology Imaging System (PerkinElmer). PerkinElmer InForm software package was used for DAPI-based cell segmentation, for quantitating expression of miR-155 with H-score function, and for enumerating Ki-67⁺ and CD8⁺ cells with cell phenotyping algorithm.

The use of melanoma patients FFPE sections was approved by the CER-VD (ID 2017-00174).

TCGA-normalized RNA-seq data from 103 primary melanoma tumors were downloaded using TCGAbiolinks R package (v2.6.12) in October 2018. Kaplan–Meier survival curves were generated using Survival and TCGAbiolinks R Packages. Differences between survival curves were evaluated using log-rank Kaplan–Meier test, where samples/patients were split and compared based on normalized gene expression (log₂ TPM) into lower (<33%) and higher (>66%) groups.

We defined a transcriptomic signature of miR-155 targets from the set of target genes downregulated upon miR-155 overexpression from ref. 13. After mapping mouse to human orthologues, the gene set was composed of IRF2BP2, NRP1, SATB1, AP3D1, SSH2, CDON, S1PR1, DYRK2, PANK1, FOSL2, FAM91A1 and SLC38A2. A signature score was then computed as the average of the log₂-normalized expression values of each gene.

In a similar way, we calculated a signature of CD8⁺ T-cell infiltration, using the gene set CD2, CD3G, CD3E, CD3D, CD8A and CD8B.

Flow-cytometry data were analyzed with FlowJo (TreeStar). Graphs and statistical analysis were made with Prism (GraphPad Software).

Specific statistical analyses are described in figure captions and Supplementary Table S2. Overall, normality of data distribution was analyzed by the Shapiro–Wilk normality test. Comparison between two unpaired groups was performed by parametric Student t test or nonparametric Mann–Whitney test. For multiple comparison, a parametric one-way ANOVA or nonparametric Kruskal–Wallis test was performed followed by Tukey multiple comparison test or Dunn multiple comparison test, respectively. Simultaneous analysis of two variables among multiple groups was performed by two-way ANOVA or two-way repeated measurements (RM) ANOVA followed by the Tukey multiple comparison test. Long-rank Mantel–Cox test was applied for survival curves and Pearson correlation coefficients were calculated for correlation analysis. P values are coded as *, P < 0.05; **, P < 0.01; ***, P < 0.001, and ****, P < 0.0001 in figures.

The statistical program OpenEpi was used to evaluate the size of each group in view of the expected mean and SEM taken from previous similar experiments. For the monitoring of antiviral immune responses, 5 mice per group are sufficient for statistical significance, whereas for antitumor therapy, a minimum of 7 mice per group are required, in view of the large variability of tumor growth between mice. Mice were grouped so that tumor size mean and SD were similar among groups before the first treatment.

Cell-intrinsic expression of miR-155 is required for efficient CD8⁺ T-cell–mediated antitumor responses, as demonstrated using genetic deletion or enforced expression of this microRNA in adoptive T-cell transfer mouse models (14). Whether this may also be the case in naturally occurring antitumor antigen-specific T-cell responses remains to be addressed. To directly determine miR-155 expression dynamics in T cells in vivo, we analyzed miR-155 expression in tumor-infiltrating CD8⁺ T cells (Fig. 1A). Fourteen days after tumor engraftment, endogenous effector CD8⁺ T cells and transferred effector OT-1 cells showed higher miR-155 expression in B16.OVA tumors than in spleens (Fig. 1B). This higher miR-155 expression was still detected in CD8⁺ TILs on day 21 (Fig. 1B). Thus, in the tumor microenvironment, effector CD8⁺ T cells upregulate and maintain high miR-155 expression.

OT-1 cells and other tumor-specific CD8⁺ T cells that infiltrate the B16-OVA tumor are chronically exposed to tumor antigens. To assess whether chronic antigen exposure could influence, or even determine, increased miR-155 expression in CD8⁺ TILs, we measured miR-155 expression in CD8⁺ T cells in a standard model of chronic antigen stimulation. This is provided by the viral infection using the clone-13 (cl13) strain of LCMV (Fig. 1C). At 7 days after infection, both transferred P14-transgenic and endogenous gp276-specific CD8⁺ T cells showed 20 and 15 times higher miR-155 expression, respectively, than naïve CD8⁺ T cells (Fig. 1D), similarly to our observations in tumor-infiltrating CD8⁺ T cells (Fig. 1B). Such high miR-155 expression was also observed 3 weeks after infection (Fig. 1D) as shown in a previous report (13), indicating that indeed chronic antigen exposure is consistently associated with increased miR-155 expression in antigen-specific CD8⁺ T cells.

To provide further support for the role of antigen exposure on the regulation of miR-155 expression and exclude the impact of chronic inflammation, P14 cells were exposed directly in vivo, in the course of viral infection, to different doses of the LCMV gp33 epitope within the same inflammatory environment. Mice were infected with either WT cl13 LCMV or a 1:3 mixture of WT and an A3 mutant LCMV cl13 strain, which lacks the gp33 epitope (Fig. 1E). Although miR-155 expression at day 7 were similar in effector P14 cells from WT and mixed virus infection, reduced miR-155 expression was found at day 21 on effector P14 cells from the mixed infection (Fig. 1F). These expression changes did not result from reduced total virus titers in the mixed infection (Supplementary Fig. S1A). In fact, gp276-specific CD8⁺ T cells showed similar miR-155 expression in both WT and mixed infection (Fig. 1G). These results show that in the same inflammatory environment, the amount and frequency of antigen exposure determines miR-155 expression in CD8⁺ T cells.

CD8⁺ T cells from tumors and chronic viral infections display an exhaustion profile characterized by increased expression of inhibitory receptors such as PD-1 and reduced cytokine production (15–18). Chronic antigen stimulation has been proposed as one of the main T-cell exhaustion drivers (19). Indeed, using the mixed infection model, it was shown that in the same inflammatory environment antigen exposure contributes to the acquisition of an exhausted phenotype by CD8⁺ T cells (20). As shown here, we confirmed that P14 cells from the mixed infection show reduced PD-1 expression (Supplementary Fig. S1B) and increased cytokine production upon in vitro peptide restimulation, when compared with P14 cells from WT infection (Supplementary Fig. S1C and S1D). Thus, increased miR-155 expression was accompanied by decreased P14 cell functionality 21 days after WT compared with mixed infection. To determine whether high miR-155 expression and activity may be detrimental and linked to the exhausted phenotype observed in P14 cells after WT infection, we analyzed the functionality of miR-155–overexpressing P14 cells after WT or mixed LCMV infection. To this aim, naïve P14 cells were transduced with MDHI-miR-155-GFP or MDHI-SCR-GFP retroviral vectors. Transduced P14 GFP⁺ cells sorted by flow cytometry were transferred to naïve B6 mice that were later infected with LCMV WT or mixed infection.

qPCR analysis of ex vivo–sorted GFP⁺ cells confirmed that P14-miR-155 cells demonstrated higher miR-155 expression than P14-SCR cells before and during chronic LCMV infection (Fig. 2A). In both infection models, P14-miR-155 cells showed increased expansion by day 7 after infection and were still detectable at day 21 in contrast to P14-SCR cells (Fig. 2B), whereas viral titers were similar in the blood of all mice (Fig. 2C). However, on day 7, PD-1 expression was 1.3 × higher in P14-miR-155 cells (Fig. 2D) as previously reported (13). This could indicate greater activation but also increased exhaustion of P14-miR-155 cells. Thus, we checked their capacity to produce cytokines upon ex vivo peptide restimulation. Although IFNγ production and frequency of IFNγ-producing cells were similar, decreased frequencies of TNFα-producing cells were observed in P14-miR-155 compared with P14-SCR cells (Fig. 2E). However, this was only observed in the WT LCMV cl13 infection, and the amount of TNFα produced per cell was similar (Fig. 2E and F). Thus, miR-155 overexpression did not lead to significant differences in IFNγ and TNFα production. In contrast, increased granzyme B production was observed in P14-miR-155 compared with P14-SCR cells (Fig. 2E and F). Therefore, high miR-155 expression does not appear to be detrimental for CD8⁺ T cells in chronic viral infections; here, CD8⁺ T cells show increased expansion and persistence regardless of the antigen dose.

Antitumor immune responses rely on CD8⁺ T cells, which respond with variable affinity to tumor neoantigens (21, 22) or with rather low affinity to nonmutated tumor-associated antigens (TAAs; ref. 23). To mimic high and low antigen affinities, we compared miR-155 expression of high (OT-1) and low (OT-3) avidity OVA-specific CD8⁺ T cells (10). Although naïve OT-1 and OT-3 cells showed similar miR-155 expression (Fig. 3A), effector OT-3 cells expressed lower miR-155 expression than OT-1 cells in spleen and tumors at 14 and 21 days after tumor engraftment. Nonetheless, like OT-1 cells, OT-3 cells consistently showed higher miR-155 expression in tumors than in spleen (Fig. 3B), indicating that chronic antigen stimulation leads to sustained miR-155 upregulation in CD8⁺ T cells regardless of the affinity for the antigen.

We have reported decreased expansion and effector differentiation of OT-3 cells compared with OT-1 cells upon N4-CpG vaccination in B16.OVA tumor–bearing mice (11). Thus, to exclude differences during peripheral T-cell priming, which could explain miR-155 expression differences between high- and low-avidity T cells, we compared miR-155 expression of OT-1 cells isolated from tumors expressing the high SIINFEKL (N4) or low SIITFEKL (T4) affinity peptide ligand. At days 14 and 21 after tumor engraftment, OT-1 cells showed increased miR-155 expression in B16.N4 tumors and to a lower extent in B16.T4 compared with expression in the spleen (Fig. 3C). OT-1 cells from B16.N4 tumors showed higher miR-155 expression than OT-1 cells from B16.T4 tumors (Fig. 3C), indicating that antigen binding affinity affects miR-155 expression in the tumor microenvironment.

We also noticed that the variability in miR-155 expression among OT-1 cells from B16.T4 tumors was higher than in B16.N4 tumors (SD −ΔCt 1.18 vs. 0.40). miR-155 expression in OT-1 cells isolated from B16.T4 tumors negatively correlated with tumor volume (Fig. 3D). Because antigen recognition is one of the main drivers of miR-155 upregulation, higher miR-155 expression in tumor-infiltrating OT-1 cells could simply reflect the activation state of the cells. However, when we analyzed expression of miR-21, another miRNA upregulated upon T-cell activation (7, 24), we did not observe any correlation with tumor volume (Fig. 3E), suggesting that miR-155 expression may be an indicator of CD8⁺ T-cell fitness in low-affinity antigen–expressing tumors. Indeed, we found that high miR-155 expression was associated with increased frequencies of IFNγ⁺TNFα⁺ OT-1 cells in B16.T4 tumors (Fig. 3F). Such association was not observed in B16.N4 tumors (Fig. 3F). However, miR-155 expression and B16.N4 tumor control also did not correlate (Fig. 3D).

It has been shown that blockade of the inhibitory receptor PD-1 enhances CD8⁺ T-cell functionality (25). When we measured miR-155 expression in tumor-infiltrating OT-1 cells from anti–PD-1–treated mice, we observed increased miR-155 expression in B16.N4 tumors and a tendency in B16.T4 tumors (Fig. 3G). In contrast, miR-21 expression was similar between mice treated or not with anti–PD-1 (Fig. 3H). Anti–PD-1–treated mice showed improved tumor control due to enhanced OT-1 cell infiltration and functionality (11). Thus, miR-155 upregulation upon PD-1 blockade also suggests that increased miR-155 expression correlates with enhanced CD8⁺ T-cell fitness.

Our data obtained in the B16 mouse melanoma tumor model showed that miR-155 is involved in CD8⁺ T-cell–mediated antitumor response (8). Thus, we aimed to determine whether these findings could be translated into cancer patients by analyzing miR-155 expression in CD8⁺ T cells from melanoma patients.

We previously reported that miR-155 expression is upregulated in effector memory (EM) compared with naïve (N) CD8⁺ T cells from the blood of HDs (7). miR-155 expression in PB CD8⁺ T-cell subsets from melanoma patients was similar to those of HDs, whereby EM and CD45RA⁺ EM (EMRA) CD8⁺ T cells showed increased miR-155 expression compared with N and central memory (CM) CD8⁺ T cells (Fig. 4A).

Higher miR-155 expression in differentiated CD8⁺ T cells compared with N CD8⁺ T cells suggests that miR-155 activity is regulated during human CD8⁺ T-cell differentiation. Indeed, in vitro activation of N CD8⁺ T cells from 4 HDs with anti-CD3/anti-CD28 coated beads showed miR-155 upregulation until day 2. Thereafter, although there is substantial variability among donors, miR-155 expression tends to decrease, but activated cells tend to retain higher miR-155 expression than resting N CD8⁺ T cells (Fig. 4B).

Increased miR-155, but not miR-21, expression was observed in PD-1⁺ compared with PD-1⁻ EM CD8⁺ T cells from PB of melanoma patients (Fig. 4C). This suggests that miR-155 reflects the activation status of human CD8⁺ T cells, as PD-1 is a TCR-driven early T-cell activation marker (26).

miR-155 expression in bulk CD8⁺ T cells from PB of melanoma patients may not reflect miR-155 expression of tumor-specific CD8⁺ T cells, as they are found in relatively low frequencies in the blood (27). We evaluated miR-155 expression in HLA-A2/Melan-A_26-35–specific CD8⁺ T-cell clones from PB of metastatic melanoma patients upon stimulation with multimers containing the high-affinity ELAGIGLTV (ELA) analog peptide. Twenty-four hours after multimer stimulation, miR-155 was upregulated in Melan-A_26-35 specific CD8⁺ T cells, and expression levels positively correlated with their avidity for the ligand measured with HLA-A2/ELA NTAmers, as previously described (Fig. 4D; ref. 12). Thus, in line with our observations in murine CD8⁺ T cells, TCR stimulation of human tumor-specific CD8⁺ T cells leads to avidity-dependent miR-155 upregulation.

As tumor-specific murine OT-1 cells showed high miR-155 expression, we analyzed miR-155 expression in melanoma patients' ex vivo–sorted CD8⁺ T-cell subsets from tumors, TILNs, and NLNs. EM CD8⁺ T cells, and to a lesser extent CM CD8⁺ T cells, had higher miR-155 expression in TILNs and tumors than in PB (Fig. 5A and B). In contrast, miR-155 expression was similar between PB and NLNs CD8⁺ T-cell subsets (Fig. 5C). In addition, EM CD8⁺ T cells in TILNs showed higher miR-155 expression than in NLNs (Fig. 5D), suggesting that the presence of cancer cells plays a role in miR-155 upregulation in EM CD8⁺ T cells in tumor-infiltrated tissues and tumor lesions.

miR-155 expression in EM CD8⁺ T cells from TILN and tumor samples widely varied among patients. We observed a positive correlation between miR-155 expression in EM CD8⁺ T cells from TILNs and the percentage of EM CD8⁺ T cells in TILNs (Fig. 5E), suggesting again that responsive CD8⁺ T cells display high miR-155 expression. Due to the limited availability of fresh tumors samples from patients (n = 4), we were not able to assess whether the same association is observed in tumors.

Single-cell suspensions from fresh or frozen tumor and TILNs are rarely available. Therefore, to analyze a larger cohort of patients in a validation study, we developed an automated miR-155 in situ hybridization (ISH) on FFPE tumor and TILN biospecimens. We observed evidence of miR-155 expression in CD8⁺ T cells and in other cells of small diameter, probably other lymphocyte subsets. Small-diameter cells revealed the highest miR-155 expression, whereas larger cells, such as cancer cells, showed low miR-155 expression (Supplementary Fig. S2A and S2B), in concordance with miR-155 qPCR expression data on sorted cells (Supplementary Fig. S2C). Most miR-155⁺ cells were in the peritumoral region (Supplementary Fig. S2D), and miR-155 H-score was variable among different areas of the same section as well as among tumors (Supplementary Fig. S2E). Of note, we could also detect a small subset of in situ proliferating Ki-67⁺ CD8⁺ T cells in some tumors (Supplementary Fig. S2A, S2B, and S2D). However, currently available techniques and reagents did not allow the development of a codetection assay for miR-155 and CD8 on the same tissue section to unequivocally establish their coexpression and assess the relationship between miR-155 expression and CD8⁺ T-cell infiltration in tumors.

The correlation between miR-155 expression and EM CD8⁺ T-cell frequencies in TILNs suggests that miR-155 could be a good prognostic marker in melanoma. Assessment of prognostic markers, however, requires analysis of large patient cohorts. Our small cohort of 19 patients had only a tendency for prolonged overall survival (OS) in patients with high miR-155 expression in EM CD8⁺ T cells from TILNs (Fig. 6A). We opted to analyze a publicly available larger cohort of patients. As there are currently no large data sets of microRNA expression profiles in tumors from melanoma patients, we examined the melanoma primary tumor RNA-seq data set of TCGA (n = 103) for a correlation between miR-155 target expression and OS. We first analyzed the expression of miR-155 targets known to influence effector CD8⁺ T-cell responses: SOCS1, PTPN2 and SHIP1 (INPP5D, refs. 8, 28). Only low PTPN2 expression correlated with prolonged OS (Supplementary Fig. S3A). Stelekati and colleagues identified a set of genes that were downregulated in miR-155–overexpressing mouse CD8⁺ T cells in chronic LCMV cl13 infection (13). As the benefit of miR-155 in CD8⁺ T-cell responses is likely due to multiple mRNA targeting (9, 29), we generated a miR-155 target gene signature with the human gene orthologues of the genes identified by Stelekati and colleagues (IRF2BP2, NRP1, SATB1, AP3D1, SSH2, CDON, S1PR1, DYRK2, PANK1, FOSL2, FAM91A1, SLC38A23; ref. 13). Although expression of individual targets did not significantly correlate with OS (Supplementary Fig. S3B), patients with low expression of the whole miR-155 target signature (defined as “MIR155SIG1 Low”) had prolonged OS compared with patients with high miR-155 target signature (defined as “MIR155SIG1 High”; Supplementary Fig. S3C). The correlation was more significant than with CD8⁺ T-cell signature (Supplementary Fig. S3D). To assess miR-155 target signature while controlling for CD8⁺ T-cell infiltration, we removed the 33% of patients (n = 34) with lower CD8⁺ T-cell signature. Among the remaining 66% (n = 69), there was no longer a benefit associated with CD8⁺ T-cell signature (hence, effectively controlling for T-cell infiltration; Supplementary Fig. S3E). However, in this segment, the association of PTPN2 with survival remained, and the one with miR-155 target signature became more significant (P = 0.002, log-rank test; Fig. 6B and C). As a control, removing the 33% of the patients with highest CD8⁺ T-cell signature resulted in nonsignificant association of miR-155 signature and survival. Altogether, low miR-155 target signature, most likely due to high miR-155 expression, in tumors of melanoma patients with CD8⁺ T-cell infiltration/signature is associated with significantly prolonged OS.

In this study, we found that miR-155 expression in CD8⁺ T cells reflects the strength of in situ antigen stimulation, independent of the inflammatory environment. TCR signaling leads to the activation of NFκB and AP-1 (30), which have been shown to bind to miR-155 promoter and induce its expression (31, 32). Additionally, IRF4, a transcription factor induced in an affinity-dependent manner in T cells (33), has also been shown to induce miR-155 expression (34) and is highly expressed in CD8⁺ T cells from LCMV chronic infection compared with cells from acute infection (35). Sustained activity of NFκB, AP-1, and IRF4 most likely maintains high miR-155 expression upon continuous antigen stimulation in CD8⁺ T cells. Increased activity of NFκB and AP-1 is probably also involved in the enhanced miR-155 expression observed in CD8⁺ TILs upon PD-1 blockade. Upon interaction with its ligand PD-L1, PD-1 recruits Shp2 that in turn dephosphorylates CD28 and TCR signaling molecules (36, 37). Therefore, PD-1 blockade prevents the inhibition of TCR and CD28 signaling that ultimately leads to the activation of NFκB and AP-1 (30, 38, 39). Follow-up studies may discern which of these or whether both are necessary to maintain high miR-155 expression in CD8⁺ TILs.

As miR-155 expression reflects in situ antigen stimulation, we could consider that high miR-155 expression is a hallmark of responsive CD8⁺ T cells. This would explain the correlation between miR-155 expression and CD8⁺ T-cell frequencies in TILNs of melanoma patients. Our results are in line with previous observations in HIV-infected individuals in whom miR-155 expression in PB CD8⁺ T cells correlates with increased CD8⁺ T-cell numbers as well as their activation state measured by PD-1 and CD38 expression (40).

High miR-155 expression in EM CD8⁺ T cells from melanoma patient TILNs could also indicate that those EM CD8⁺ T cells recognize the antigen with higher affinity than EM CD8⁺ T cells from patients with low miR-155 expression. In this regard, we reported (11) that high-affinity compared with low-affinity stimulation in the tumor leads to increased tumor control due to increased CD8⁺ T-cell expansion/survival and cytotoxicity.

Increased survival of CD8⁺ T cells with high miR-155 expression may also explain the positive correlation between miR-155 expression and EM CD8⁺ T-cell infiltration in TILNs from melanoma patients. Our previous report in miR-155–deficient mice showed that the absence of miR-155 impaired CD8⁺ T-cell proliferation and survival in acute viral infections (8). Another study also showed that miR-155 promotes CD8⁺ T-cell persistence in chronic infections (13). It was subsequently shown that enhanced tumor control by miR-155–overexpressing CD8⁺ T cells resulted from SOCS1 and PTPN2 targeting, which increased responsiveness to IL7 and IL15 homeostatic cytokines (28), essential for survival and homeostatic proliferation of memory T cells (41). Consistent with this molecular mechanism, the silencing of SOCS1 recapitulated the enhanced tumor control of miR-155–overexpressing CD8⁺ T cells in B16 tumor–bearing mice (8). The miR-155–SOCS1 regulation axis was also shown to be essential for CD8⁺ T-cell maintenance in chronic viral infections (42). However, although we did observe an association between low PTPN2 expression in tumors and survival of melanoma patients, we did not observe any association with SOCS1 expression. The fact that we analyzed RNA-seq data of whole tumors containing all types of cells (tumor, immune, and stromal cells) prevents us from discerning the contribution of CD8⁺ T cells to the overall gene-expression profile. Nonetheless, we only found an association between prolonged survival and miR-155 target signature in patients with high CD8⁺ T-cell signature/infiltration, indicating that the miR-155 target signature may indeed arise from CD8⁺ TILs. Thus, we believe that these results are encouraging for follow-up studies using single-cell RNA-seq (sc-RNA-seq) data to specifically assess the effect of miR-155 target signature in CD8⁺ TILs.

The association between high miR-155 expression and increased number of cytokine-producing cells in B16.T4 tumors as well as B16.N4 tumors upon PD-1 blockade is in line with previous reports in which miR-155 overexpression led to increased cytokine production in acute infections (9) and tumors (28). Stelekati and colleagues, however, did not observe such association in chronic viral infections (13). Instead, the benefit of miR-155 overexpression arose from increased T-cell persistence rather than functionality. The miR-155 target signature, which in our analysis was associated with increased OS in melanoma patients, was derived from this latter study. Therefore, the mechanism(s) by which high miR-155 expression promotes effector CD8⁺ T cells may be context dependent (42), whereby increased functionality, survival, and proliferation may be differentially involved.

Altogether, we hypothesize that miR-155 expression may be a marker of CD8⁺ T-cell responsiveness to antigen in the tumor microenvironment. Future analyses of large patient cohorts scRNA-seq data may allow specific assessment of the miR-155 target signature in CD8⁺ TILs and reveal its potential as prognostic biomarker in cancer patients.

P. Romero is Editor-in-Chief of Journal for Immunotherapy of Cancer, reports receiving a commercial research grant from Roche pRED–Zurich, has received honoraria from the speakers bureau of Bristol-Myers Squibb, AstraZeneca, and Roche and is a consultant/advisory board member for Immatics Biotechnologies, NexImmune, and Transgene. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A. Martinez-Usatorre, C. Jandus, P. Romero

Development of methodology: A. Martinez-Usatorre, L.F. Sempere, S.J. Carmona, D. Zehn

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Martinez-Usatorre, L.F. Sempere, L. Carretero-Iglesia, N. Rufer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Martinez-Usatorre, S.J. Carmona, L. Carretero-Iglesia, G. Monnot, P. Romero

Writing, review, and/or revision of the manuscript: A. Martinez-Usatorre, L.F. Sempere, A. Donda, C. Jandus, P. Romero

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Monnot, D.E. Speiser

Study supervision: A. Donda, C. Jandus, P. Romero

The authors are grateful to the patients for their dedicated collaboration and to healthy donors for their blood. We thank Hélène Maby-El Hajjami, Laurène Cagnon, and Samia Abed Maillard for assistance with patients' clinical data, and Daniel Utzschneider and Francesca Alfei for advice with the LCMV cl13 model. We thank Leyder Lozano, Candice Stoudmann, and Silvia Ferreira for technical help as well as Romain Bedel for technical assistance in the FACS. We also thank David Gfeller for his support, the VARI Pathology and Biorepository Core for assistance with multiplex assays, and the VARI Confocal Microscopy and Quantitative Imaging Core for assistance with multispectral image scanning and analysis. This work was funded in part by grants from the Swiss National Science Foundation to P. Romero (Sinergia CRSII3_141879, Sinergia CRSII3_160708, and 31003A_156469), from the ERC (PROTECTC) to D. Zehn, and research funds from Van Andel Research Institute and the Phi Beta Psi Sorority to L.F. Sempere.

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.