Melanoma-Secreted Lysosomes Trigger Monocyte-Derived Dendritic Cell Apoptosis and Limit Cancer Immunotherapy

Patients with advanced melanoma have very poor prognosis and must undergo a series of excruciating treatments of chemotherapy, oncogene inhibitors, or combined therapy (1). A number of seminal papers have established a direct correlation between a high frequency of cytotoxic T cells in the tumor mass and improved clinical outcomes (2, 3). As a result, many recent and ongoing clinical trials have attempted to harness the cytotoxic capabilities of T cells to eradicate melanoma. Pioneering therapies based on de novo expansion of tumor-infiltrating cytotoxic T cells and their reinfusion to the patient demonstrated the capability of the immune system to mediate tumor rejection (4, 5). Recently, clinical responses were observed in patients with melanoma treated with blocking antibodies directed against suppressive receptors on T cells (6, 7). Nonetheless, a complete curative response is observed in only a fraction of the patients, emphasizing the need to improve immunotherapies (8, 9). Given the high mutational rate in melanoma, it is somewhat surprising that the host does not generate an effective T-cell immunity against the tumor cells (10). The general convention suggests that the capacity of tumor cells to edit their presented antigen and their ability to promote a suppressive microenvironment is what restricts the number of melanoma-infiltrating cytotoxic T cells with a limited T-cell receptor (TCR) repertoire (11–14).

One way to increase the prevalence of tumor-reactive T cells is through exploiting the antigen presentation capacities of dendritic cells (DC) and their ability to activate cytotoxic T cells (15). DC are composed of heterogeneous cell populations that differ in their morphology, membrane receptors, tissue distribution, migratory, and antigen presentation capabilities (16–18). The current understanding of DC ontogeny suggests that they are comprised of four main subsets consisting of plasmacytoid DC (pDC), classical DC type 1 (cDC1), classical DC 2 (cDC2), and monocyte-derived DC (MoDC; ref. 18). Although some DC subsets such as CD103⁺/CD141⁺ or CD8⁺ cDC are superior antigen presenting cells, their prevalence in tumors and blood is limited (19, 20). Vaccination with MHC-I and II neoantigens following resection of stage IIb/c melanoma lesions can prevent tumor recurrence in human patients (21, 22). However, to date, attempts to employ DC to eradicate established melanoma immunity have shown disappointing results in clinical settings (23).

By mapping the molecular mechanism that regulates the processing of tumor-binding antibodies by DC, we recently generated one of the most potent DC-based immunotherapies for cancer in mice (24, 25). Interestingly, although this treatment has proven effective when tumors were under 20 mm², it was almost inert once tumors exceeded an average size of about 40 mm². Examination of most reported immunotherapies in mouse melanoma models (e.g., ret, B16F10 and Braf^V600E/PTEN^−/−) indicated that they are limited to early stage and relatively small tumors (19, 26–30).

We leveraged the unresponsiveness of large tumors to model patients with melanoma who do not respond to immunotherapies. We found that in contrast to the current paradigm, melanoma-infiltrating MoDC do not migrate into sentinel lymph nodes, but rather are required in situ to license the cytotoxic activity of CD8⁺ T cells. We further demonstrated that MoDC undergo apoptosis resulting from their uptake of melanoma-secreted lysosomes, and their absence limits the efficacy of immunotherapies. Although these models do not fully recapitulate the complexity of the human disease, the principle underlying their capacity to evade immune pressure may be applicable to the human disease.

Wild-type (WT) C57BL/6 mice were obtained from Envigo. B6;129S-Gt(ROSA)26Sortm1.1(CAG-COX8A/Dendra2)Dcc/J mice were purchased from Jackson Laboratories. B6.Cg-Ptprca Tg (UBC-PA-GFP)1Mnz/J mice were obtained from Dr. Ziv Shulman (Weizmann Institute, Rehovot, Israel). CD11C-DTR mice were obtained from professor Jakub Abramson (Weizmann Institute). NLRP3 knockout mice were obtained from professor Neta Erez (Tel Aviv University, Tel Aviv, Israel). All mice were housed in an American Association for the Accreditation of Laboratory Animal Care–accredited animal facility and maintained under specific pathogen-free conditions. Animal experiments were approved and conducted in accordance with Tel-Aviv University Laboratory Accreditation #01-16-095. Male and female 8- to 12-week-old mice were used in all experiments.

B16F10 (CRL-6475) and YUMM 1.7 (CRL-3362) cells were purchased from the ATCC, in January 2017, and Ret melanoma cells were a kind gift from professor Neta Erez from Tel Aviv University. Ret authentication was verified using genome sequencing of enzymes in the MAPK pathway. A375 cell line was a kind gift from professor Tamar Geiger from Tel Aviv University. A375 authentication was performed at the Genomics Core Facility of BioRap Technologies and the Rappaport Research Institute in Technion, Israel. Short tandem repeat profiles were determined using the Promega PowerPlex 16 HS Kit. HEK-293FT were purchased from Thermo Fisher Scientific in January 2017. Cells were cultured in DMEM supplemented with 10% heat-inactivated FBS, 2 mmol/L l-glutamine, and 100 μg/mL penicillin/streptomycin (all from Biological Industries). YUMM 1.7 were cultured in DMEM:F12 complete medium. Melan-A cells were purchased from Welcome Trust Functional Genomics Cell Bank in February 2019 and were supplemented with 200 nmol/L of phorbol 12-myristate 13-acetate (Santa Cruz Biotechnology). Primary keratinocytes were a kind gift from Dr. Chen Luxemburg (Tel Aviv University) and were cultured as described in ref. (31). All cells were routinely tested for Mycoplasma (EZ-PCR Mycoplasma Test Kit, Biological Industries).

For melanoma tumor studies, 2 × 10⁵ of B16F10 or YUMM1.7, or 5 × 10⁵ ret cells were suspended in 50 μL of medium and were injected s.c. to C57BL/6 mice above the right flank. The size of growing tumors was measured twice a week using calipers. When tumors reached 200 mm², the mice were sacrificed due to ethical considerations.

All tissue preparations were performed simultaneously from each individual mouse (after euthanasia by CO₂ inhalation). Peripheral blood (PB) was collected through perfusion of the right atrium of the heart (Hank's Balanced Salt Solution, HBSS, with 10 mmol/L EDTA) into a sodium heparin-coated vacuum tube. Mononuclear cells were collected following centrifugation on Histopaque-1077 Hybri-Max (Sigma Aldrich) density gradient medium.

Spleen and lymph node were homogenized in 2% FBS and 5 mmol/L EDTA supplemented HBSS through 70 μm strainer (Thermo Fisher Scientific). Tumors were digested in RPMI 1640 with 2 mg/mL collagenase IV, 2,000 U/mL DNase I (both from Sigma Aldrich), and 50 ng/mL GMCSF (Peprotech) for 40 minutes at 37°C with magnetic stirrers (200 rpm). CD11b⁺ cells were isolated using magnetic CD11b MicroBeads and column (both from Miltenyi Biotec), incubated for 2 hours in complete RPMI 1640 in 37°C, and loosely adhered cells were sorted by FACSAriaII as FSC^lowSSC^low cells.

For T-cell isolation, cells were incubated with anti-CD4 or anti-CD8 magnetic beads (MojoSort Nanobeads, BioLegend) according to the manufacturer's instruction. T cells were cultured in RPMI 1640 supplemented with 1% Pen-Strep, 10% heat-inactivated FBS, 1% Sodium pyruvate, 1% MEM-Eagle nonessential amino acids, 1% Insulin-Transferrin-Selenium (PeproTech), and 50 μmol/L β-mercaptoethanol (Sigma-Aldrich) and were supplemented with 1,000 IU/mL recombinant murine IL2 (PeproTech) on tissue culture plates precoated with 0.5 μg/mL anti-CD3 antibodies (clone 17A2, Biolegend).

Lentivirus was prepared as described (32), briefly, HEK-293FT cells were transfected with pLVX plasmids containing wasabi, or tdTomato under EF1 promoter together with psPAX2 (a gift from Didier Trono, Addgene plasmid # 12260) and pCMV-VSV-G (a gift from Bob Weinberg, Addgene plasmid # 8454). Media-containing viruses were collected after 24 and 48 hours. For infection, B16 cells were incubated with viruses and 100 μg/mL polybrene (Sigma Aldrich) for 30 minutes followed by 30-minute centrifugation before medium was replaced. After 3 days, cells that expressed wasabi, TRP1-GFP, or tdTomato were sorted by FACSAriaII.

Retroviruses were prepared as described (32). Briefly, Platinum E cells (a gift from professor Cyril Cohen, Bar-Ilan University) were transfected with 2:1 molar ratio of pMIGII and PCL-Eco plasmids (both were kindly provided by professor Dario A.A. Vignali). After 48 hours, viruses were centrifuged for 1 hour at 100,000 rcf and suspended in 1 mL media overnight at 4°C. Splenic CD8⁺ T cells were cultured on a plate precoated with anti-CD3 (0.5 μg/mL) with IL2 (1,000 IU/mL). Next, 0.3 mL of concentrated retroviruses were added to every 2 × 10⁶ CD8⁺ T cells with 10 μg/mL polybrene. Cells were incubated for 30 minutes and centrifuged at 500 rcf for 1 hour. Medium was replaced, and T cells were cultured for additional 3 days.

Mice with 8–10-day B16F10 tumors (average 25 mm² in size) were injected with 1 × 10⁶ gp100-TCR expressing CD8⁺ T cells (26) per mouse after sublethally irradiation (600 rad), followed by i.p. injections of 300,000 IU of IL2 (PeproTech) for 4 consecutive days. To increase DC viability, prior to CD8⁺ T-cell transfer in late-stage tumors, nonirradiated recipient mice were injected i.p. with 80 μg/mouse anti-CD40 and with 1 μg/mouse of recombinant GMCSF.

Mice were injected intratumorally with 80 μg anti-CD40 (clone FGK4.5; BioXCell), 0.2 μg TNFα (BioLegend), and 100 μg/mouse anti-TRP1antibody (clone TA99; BioXCell). CpG ODN1826 and PolyI:C were injected for 3 consecutive days (50 and 100 μg, respectively, both from InvivoGen). For high-dose experiment, 0.5 μg TNFα, 200 μg anti-CD40, 400 μg anti-TRP1, 100 μg CpG ODN1826, and 200 μg PolyI:C were injected. Checkpoint blockade was conducted as described by Twyman-Saint Victor and colleagues (33) with slight modifications. Briefly, tumor-bearing mice were anesthetized with i.p. injection of 100 mg/kg ketamine plus 10 mg/kg xylazine and placed on a tray covered with 4 mm thick lead jig exposing only the rear right leg approximately 30 cm from the beam source. Mice were then irradiated at single dose of 15 Gy using Philips industrial X-ray machine (1.1 Gy/min). After 24 hours mice were injected i.p. with 100 μg of blocking antibodies against CTLA-4 (clone 9D9, BioXCell) and PD-1 (RMP1-14, BioXCell). In some experiments, T-cell egression from lymphoid organs was blocked in mice bearing palpable tumors using daily i.p. injections of one mg/kg FTY720 (Sigma Aldrich) prior to injection of anti-TRP1 plus anti-CD40 and TNFα immunotherapy.

To generate CD11c-DTR BM chimeric mice, lethally irradiated (9 Gy using γ-irradiator Biobeam7000, ¹³⁷Cs, 2.6 Gy/min) mice were injected i.v. with 5 × 10⁶ bone marrow cells and were challenged with 2.5 × 10⁵ B16F10 cells 3 weeks later. For CD11c⁺ cell depletion, mice were injected i.p. with 4 ng/g body weight of diphtheria toxin (D0564; Sigma Aldrich) 6 days after tumor challenge. For monocyte depletion, mice were injected with 100 μg of anti-CCL2 (clone-2H5, BioLegend). After 24 hours, mice were treated with 1 × 10⁶ splenic CD8⁺ T cell expressing gp100-reactive TCR and with 300,000 IU IL2, injected i.p. daily.

Tumor-challenged mice were injected i.p. every day with 1 mg of 5-bromo-2-deoxyuridine (BrdU) in 200 μL PBS. At several time points, mice were euthanized, and single-cell suspensions were prepared from tumors and draining lymph nodes. Cells were then stained for lineage markers followed by intracellular staining with FITC-conjugated anti-BrdU antibody according to the manufacturer's instructions (BD Pharmingen) and analyzed by flow cytometry.

PA (photo-activated)-GFP and mito-Dendra2 mice were injected s.c. with 2 × 10⁵ B16F10 cells above the right and left flanks and allowed to grow for 8 days. Tumors were illuminated for 5 minutes with 405 nm wavelength using 20 mW led diode, which was custom built by Thorlabs Inc. After 24 hours, DC from the blood, lymph nodes, spleen, and tumor were isolated and analyzed by flow cytometry.

For frozen sections, prepared as described (32). For anti-mouse staining, the antibodies used were MHCII (clone M5/114.15.2), CD11c (clone N418), CD11b (cloneM1/70), and anti-TCRβ (H57-597), all from BioLegend. Cleaved caspase-3 (clone 5A1E) from Cell Signaling Technologies and LYVE-1 (clone ALY7) from Invitrogen were used. Nuclei were counterstained with Hoechst 33342 (Thermo Fisher Scientific). For anti-human staining, the antibodies used were HLA-DR (clone L243) and CD11c (clone 3.9), both from BioLegend.

B16-Wasabi, CD8⁺ T cells, and CD11b⁺ DC were cocultured on glass-bottom confocal plates (Cellvis) in T-cell medium without IL2 and incubated overnight under standard conditions. For functional assays, DC were incubated for 30 minutes with 1:50 diluted FITC-labeled latex beads (Sigma Aldrich). DC cells were stained with LysoSensor (L7533, Thermo Fisher Scientific). For visualization of lysosome trafficking, B16F10 cells were stained with lysotracker (L7526, Thermo Fisher Scientific) for 30 minutes in 37°C and washed twice before the addition of 50 × 10⁴ MoDC. Images were collected using a Zeiss LSM800 confocal laser scanning microscope and analyzed using ZEN software (Carl Zeiss Microscopy).

Human CD8⁺ T cells, MoDC, and tumor cells were isolated from human melanoma tissue sample, as well as CD8⁺ T cells from PB. Mouse CD8⁺ T cells were isolated from various organs of tumor-bearing mice. CD8⁺ T cells were cocultured with tumor cells (10,000 cells per well) with or without CD11b⁺ DC at a ratio of 1:5:2 (T:E:DC) in a round bottom 96-well plate. After 24 hours, cells were incubated for 1 hour with biotin-conjugated anti-CD107a (BioLegend) at a 1:100 dilution. Cells were further stained with (BV421)-Streptavidin, (Alexa Fluor 488)-CD45, (Brilliant Violet 605)-CD8, and Annexin V for 15 minutes and for 2 minutes with propidium iodide (PI) solution (all from Biolegend) on ice, and staining levels were analyzed by flow cytometry (CytoFLEX, Beckman Coulter).

For isolation of tumor-secreted lysosomes, B16F10, TRP1-GFP–expressing B16F10, Melan-A, and A375 cell lines were cultured for 48 hours before DMEM supplemented with 10% heat-inactivated FBS centrifuged at 140,000 rpm for 1 hour to deplete bovine-derived exosomes. For isolation of T-cell–derived lysosomes, 2 × 10⁷ CD8⁺ T cells were isolated from spleen or PB and plated in 10 cm plate precoated with anti-CD3 antibodies for 48 hours. Supernatants were collected and concentrated using Optima XPN ultracentrifuge at 140,000 rpm for 1 hour, and the pellet containing either exosomes or lysosomes was resuspended in 50 μL PBS.

Isolated B16F10 and Melan A melanosomes were adsorbed on formvar/carbon-coated grids and stained with 2% aqueous uranyl acetate. Samples were examined using JEM 1400plus transmission electron microscopy.

CD11b⁺ DC isolated from tumor, draining lymph node (DLN), or PB were cultured in 96 wells (10,000 cells per well) in full DMEM medium supplemented with 50 ng/mL GMCSF. For apoptotic activity, PB CD11b⁺ DC were pretreated for 1 hour with 25 μmol/L caspase-1 inhibitor VX-765 (CAS 273404-37-8, Santa Cruz Biotechnology) or 20 μmol/L caspase-3 inhibitor (Z-VAD(OMe)-FMK, CAS 187389-52-2, Santa Cruz Biotechnology) and incubated with: 50 mg/mL in DMSO synthetic melanin (M0418, Sigma Aldrich), or with B16F10 exosomes that were untreated, heat-denatured (56°C, 1 hour) or lysosomes depleted (biotin-anti-LAMP-1 and streptavidin-magnetic beads; MojoSort Nanobeads, BioLegend). Cells were visualized by confocal microscope or assayed for caspase activity using Caspase-Glo 3/7 Assay (Promega) in Synergy H1M plate reader (BioTek).

Cells were analyzed using flow cytometry (CytoFLEX, Beckman Coulter) and sorted by FACS (BD FACSAria III, BD Biosciences). Datasets were analyzed using FlowJo software (Tree Star). The following mice antigen were used: (Brilliant Violet 421)-CD45 (clone 30-F11), (Alexa Fluor 647)-MHCII (clone M5/114.15.2), (Alexa Fluor 594, Brilliant Violet 605)-CD11c (clone N418), (Per-CP)-CD11b (cloneM1/70), (Alexa Fluor 488)-CD86 (clone GL-1), (PE-Cy7)-CD103 (clone 2E7), (APC-Cy7)-Ly6C (clone HK1.4), (Brilliant Violet 421)-βTCR (clone), (PE)-CD4 (clone RM4-4), (Brilliant Violet 605)-CD8 (clone 53-6.7), (Alexa Fluor 488)-CD45 (clone 30-F11), (APC)-NKp46 (clone 29AI.4), (PE/Cy7)-CD19 (clone 6D5), (PE)-B220 (clone RA3-6B2), (Alexa Fluor 488)-CCR7 (clone 4B12), and (PE)-MERTK (clone 2B10C42).

For human-specific antigens, the following antibodies were used: (Alexa Fluor 488) CD3 (HIT3a), (Alexa Fluor 594) CD4 (RPA-T4), (Allophycocyanin) CD19 (HIB19), (Alexa Fluor 647) CD8 (HIT8a), (Brilliant Violet 650) CD11c (3.9), (Alexa Fluor 647) CD16 (3G8), (PerCp/Cy5.5) CD32 (FUN-2), (Brilliant Violet 421) CD64 (10.1), (Allophycocyanin/Cy7) CD45RO (UCHL1), and (Phycoerythrin/Cy7) CD45RA (HI100). All Abs were purchased from BioLegend. Cells were suspended in FACS buffer consisting of HBSS with 2% FCS and 0.05 mmol/L EDTA.

Note that 474 melanoma expression profiles from The Cancer Genome Atlas (TCGA) RNA sequencing dataset were downloaded from UCSC XENA's web site (http://xena.ucsc.edu/) together with their associated clinical information. We used the PROMO software suite for importing, preprocessing, analyzing, and visualizing the data (http://acgt.cs.tau.ac.il/promo/). Six inconsistent samples were removed based on their phenotype labels. The k-means clustering algorithm was applied to the normalized expression levels of 16 dendritic genes, to partition the 468 remaining melanoma samples into four sample groups. The four sample groups are characterized by different levels of the dendritic expression signature. Survival analysis using a Kaplan–Meier plot and the log-rank test showed that the four clusters represent different 5-year survival risks.

Kaplan–Meier plots of patients with melanoma based on expression of CD1c (https://www.proteinatlas.org/ENSG00000158481-CD1C/pathology/melanoma) and CD11c (https://www.proteinatlas.org/ENSG00000140678-ITGAX/pathology/melanoma) proteins were calculated from data at the Human Protein Atlas repository (http://www.proteinatlas.org).

Each experiment was performed 3 times. Each experimental group consisted of at least 3 mice. For time course experiments, significance was calculated using the nonparametric two-way ANOVA with Tukey correction for multiple hypotheses. In some cases, Bonferroni–Sidak post-test was performed after two-way ANOVA. For two groups' analysis, one-way ANOVA with Dunn test was performed. The results were analyzed by Prism (GraphPad Software). All statistical analyses were performed in Prism (GraphPad Software, Inc.).

All mice were housed in an animal facility accredited by the American Association for the Accreditation of Laboratory Animal Care and were maintained under specific pathogen-free conditions. Animal experiments were approved and conducted in accordance with Tel Aviv University Laboratory Accreditation (01-17-067 and 01-19-034). The Tel Aviv University Institutional Review Board and Rabin Medical Center–Beilinson board approved the human subject protocols (Helsinki ethical approval 0326-18), and informed consent was obtained from all subjects prior to participation in the study.

Initially, we sought to determine the exact time point at which immunotherapies lose their therapeutic efficacy. To this end, we injected C57BL/6 mice s.c. with B16F10 cells and allowed to grow until they reached a palpable size. At several time points, we injected the mice intratumorally with a combination of TNFα+anti-CD40 and antibody against the melanoma antigen TRP1, which was shown to be highly effective in mouse tumors (25, 34). Mice treated until day ten had a significant tumor regression. This treatment was still effective when tumors were treated on day 12, although less potent compared with earlier days. In contrast, this treatment had no impact on tumor size when tumors were treated on day 14, or later (Fig. 1A). Similar results were also obtained using a cell line isolated from ret transgenic mice. Injection of TNFα+anti-CD40 and anti-TRP1 antibody to mice bearing tumors under 20 mm² induced significant tumor regression, whereas treating mice bearing large tumors was almost inert (Fig. 1B). Importantly, increasing the dosage of the administrated immunotherapy did not improve the responsiveness of late-stage tumors (Supplementary Fig. S1A and S1B).

We next tested whether this characterizes other immunotherapies, or rather is limited to this particular treatment. Mice bearing B16F10 tumors were injected with TLR9 and TLR7 agonists. Again, treatment efficacy was almost completely abrogated once in late-stage tumors (Fig. 1C). Increasing the administrated dosage failed to induce tumor regression in late-stage tumors (Supplementary Fig. S1C). Recently, Twyman-Saint Victor and colleagues have developed a protocol combining high-dose radiation and checkpoint blockade (33). Injection of antibodies against PD-1 and CTLA-4 followed by local tumor radiation induced significant tumor regression in small tumors, but not in late-stage tumors (Fig. 1D).

Next, we tested whether adoptive transfer of tumor-reactive T cells can induce regression of late-stage tumors. To this end, splenic CD8⁺ T cells infected with TCR against the melanoma antigen gp100 were injected, along with high-dose IL2, to sublethally irradiated tumor-bearing mice, according to established protocol (Fig. 1E; ref. 26). Similarly, injection of gp100-reacive T cells induced remarked regression of early-stage tumors, but was almost inert when administrated to mice bearing late-stage tumors (Fig. 1F).

We characterized the changes in the composition of immune cell subsets from early- to late-stage melanoma tumors. We found a drastic reduction over time in the numbers and percentages of total leukocytes infiltrating B16F10 and ret melanoma tumors (Fig. 2A–C). This reduction was found across almost all immune cell populations, but mainly in natural killer (NK) cells, T cells, and DC (Supplementary Fig. S2A–S2B). This reduction was limited to the tumors and was not observed in the blood or DLN (Supplementary Fig. S2C and S2D). Although reduction of NK and CD8⁺ T cells in late-stage tumors is expected to correlate with reduced tumor lysis, why the decrease of tumor-infiltrating DC affects the efficacy of immunotherapy is not clear.

Therefore, we initially set to understand why DC decrease over time. Tumor-infiltrating DC (TIDC) were found to be composed of MoDC and migratory CD103⁺ DC (Supplementary Fig. S2E). Both DC subsets were present in early-stage tumors but almost completely absent in late-stage tumors, when immunotherapies are not effective (Fig. 2D and E; Supplementary Fig. S2F). Although CD103⁺ DC are thought to be better antigen-presenting cells, over 90% of TIDC are MoDC, providing rationale to further elucidate their impact on the efficacy of immunotherapy. The drastic reduction in MoDC numbers in late-stage tumors was further visualized by tissue sections of mouse melanoma tumors. Sections from day-eight tumors were heavily infiltrated with CD11b⁺ MoDC, whereas sections from day 14 contained only negligible amounts of MoDC (Fig. 2F).

The general wisdom suggests that following tumor antigen uptake, MoDC migrates to the DLN, to activate T cells. MoDC isolated from wasabi-labeled tumors were found to contain tumor antigens (Fig. 2G), and histologic sectioning of these tumors indicated that they are mainly localized in the T-cell zone in the DLN (Supplementary Fig. S2G). Consistently, T-cell numbers in the DLN continued to proliferate throughout tumor progression (Supplementary Fig. S2H), and blocking T-cell egression from lymphoid organs using the sphingosine analog FTY720 almost completely abrogates the therapeutic effect of immunotherapy (Fig. 2H). These results suggest the DLN as the primary site of antigen presentation and the main source for tumor-reactive T cells.

Intriguingly, we did not observe any significant increase in the percentages and numbers of MoDC in the DLN or in the blood during tumor progression (Fig. 3A and B; Supplementary Fig. S3A–S3C). In addition, all the tumor lymph vessels are located in the tumor periphery, with no lymph vessels penetrating the center of the tumors, where the majority of MoDC are. All DC adjacent to or inside lymph vessels have expressed low levels of CD11c and Ly6C, suggesting that they are dermal DC or Langerhans cells, and not MoDC (Fig. 3C). We could also detect secreted tumor antigens, such as TRP1, in MoDC found in DLN, suggesting that tumor antigens can be passively drained (Supplementary Fig. S3D and S3E). In addition, MoDC did not express detectable levels of CCR7, which is required for their migration to the DLN (Fig. 3D).

To monitor migration patterns of MoDC, we used mice expressing the green-to-red photoconvertible fluorescent protein Dendra2. Mice were injected s.c. with B16F10 cells, and tumors were allowed to grow for 8 days. We then performed a small surgery on the skin adjacent to the tumor, and tumor was illuminated using a 405 nm diode LED array (Fig. 3E). After 24 hours, blood, spleen, tumor, right inguinal DLN, and left axillary non-DLN were removed, and the green–red conversion in MoDC was analyzed by flow cytometry. We could not detect MoDC expressing red fluorescence protein in any of the tested organ, but the tumors (Fig. 3F and G). In contrast, B cells, NKT cells, and neutrophils expressing red fluorescence protein were found in the blood and lymphoid tissues across the body (Supplementary Fig. S3F). One caveat to this method is the relatively rapid decaying of red fluorescence signal, as immediately after illumination, about 80% of immune cells expressed red fluorescence, whereas after 24 hours, about 40% remained positive (Supplementary Fig. S3G). Therefore, we also used mice expressing photoactivatable GFP protein (PA-GFP), which is more stable compared with Dendra2, but has a much lower fluorescence intensity. We could detect CD103⁺ DC and monocytes, but not in mature MoDC, expressing PA-GFP in DLN of tumor-bearing mice (Fig. 3H and I).

When isolating MoDC from tumors, we noticed that MoDC from late-stage tumors, but not early-stage tumors, exhibit rapid cell death when cultured in vitro (Fig. 4A). Caspase-3/7 activity assay indicated that these cells undergo apoptosis (Fig. 4B). In contrast, MoDC isolated from the DLN and blood of late-stage tumor-bearing mice were viable and comparable with MoDC from early-stage tumors and naïve mice (Fig. 4A and B). Although MoDC from late-stage tumors have completely lost their ability to phagocytize latex beads, the phagocytotic capacity of MoDC from early-stage tumors was comparable with that of naïve mice (Supplementary Fig. S4A). Consistently, higher levels of cleaved caspase-3 were observed in histologic sections of late-stage tumors, in comparison with early-stage sections (Fig. 4C).

We also noticed that most tumor-infiltrating MoDC, even in early-stage tumors, contain a high density of melanosomes, indicative of exosomes uptake (Supplementary Fig. S4B). Therefore, we isolated exosomes from the supernatants of B16F10 cells and incubated them with MoDC from the blood of tumor-bearing mice. As a control, we used synthetic melanin, or melanosomes isolated from immortalized melanocytes (Fig. 4D). Incubation with exosomes from B16F10, but not from normal melanocytes or with melanin, induced massive apoptosis of MoDC (Fig. 4E; Supplementary Fig. S4C). Induction of apoptosis by exosomes was limited to MoDC, as incubation of primary keratinocytes and splenic T cells with melanoma-secreted exosomes did not induce their apoptosis (Supplementary Fig. S4D). Analysis of B16 mutanome indicated that none of the enzymes that synthesize melanin are mutated in B16F10 (35). Furthermore, the examination of B16F10 melanosomes under electron microscopy indicated that they are intact and similar to the melanosomes secreted from normal melanocytes (Supplementary Fig. S4E), suggesting that other exosomes may be involved. In addition to melanosomes, B16F10 secrete also high levels of lysosomes (Supplementary Fig. S4F). Selective labeling of lysosomes in B16F10 indicated that they are efficiently taken up by MoDC (Supplementary Fig. S4G). Depleting the lysosomes from melanoma-derived exosomes prior to incubation with MoDC almost completely prevented apoptosis (Fig. 4F and G).

Denaturation of melanoma-secreted exosomes at 56°C for 30 minutes prior to their incubation with MoDC completely blocked their capacity to induce apoptosis (Fig. 4F and G). Incubation of MoDC with a pan-caspase inhibitor—or with VX-765, a selective inhibitor of caspases 1 and 4—significantly inhibited apoptosis in these cells (Fig. 4F and G; Supplementary Fig. S4H). Consistently, apoptosis induced by lysosomes was almost completely abrogated in MoDC obtained from mice deficient in Nod-like receptor protein 3 (NLRP3 KO; Fig. 4H and I). Lysosomes isolated from splenic T cells induce apoptosis in MoDC at levels similar to that induced by melanoma-secreted lysosomes (Fig. 4H), suggesting that this is not a result of a special feature of melanoma-derived lysosomes.

Next, we tested the effectivity of immunotherapy in late-stage melanoma that does not secrete lysosomes. Screening analysis indicated that YUMM1.7 melanoma cells (36) do not secrete lysosomes (Supplementary Fig. S4H). Thus, mice bearing large tumors were treated with combination of TNFα+anti-CD40 and anti-TRP1 antibodies. Indeed, these tumors contained viable MoDC, and tumors underwent regression upon administration of immunotherapy (Fig. 4J).

We next addressed why the absence of tumor-infiltrating MoDC affects the outcomes of immunotherapies. Although the DLN is a major site in which tumor-reactive T cells are generated, incubation of CD8⁺ T cells isolated from the blood, or DLN of B16F10 tumor-bearing mice with B16F10 cells did not result in release of lytic granules and tumor cell lysis (Fig. 5A–C). In sharp contrast, incubation of B16F10 cells with day eight tumor-infiltrating CD8⁺ T cells resulted in significant lysis of B16F10 cells and secretion of lytic granules from T cells (Fig. 5A–C). Addition of MoDC isolated from the tumor, but not from the blood or the DLN of tumor-bearing mice, induced the release of cytotoxic granules from CD8⁺ T cells and tumor cell lysis (Fig. 5D–F). Consistent with their loss of function, MoDC from late-stage tumors failed to promote secretion of lytic granules and tumor cells lysis by CD8⁺ T cells (Supplementary Fig. S5A).

We next tested whether the signals that induce CD8⁺ T-cell activation are secreted by MoDC, or rather are cell-contact dependent. Thus, CD8⁺ T cells from the DLN of tumor-bearing mice were incubated with B16F10 cells in the presence of the isolated DC or with their supernatants. Significant tumor cell lysis and secretion of lytic granules over baseline levels were observed only upon addition of tumor-infiltrating MoDC, but not their supernatants, indicating that T-cell activation required cell contact (Fig. 5G and H; Supplementary Data; Fig. 5B). CD8⁺ T-cell activation was not reduced upon blocking of MHCI-TCR interactions (Fig. 5I).

Because MoDC express higher levels of CD86 and CD40 (Fig. 5J and K), we tested whether stimulation of MoDC with classical adjuvants is sufficient to activate CD8⁺ T cells. Thus, PB MoDC were isolated and activated with LPS or ant-CD40 and added to a coculture of CD8⁺ T cells and tumor cells. Similarly to MoDC from tumors, stimulated blood MoDC activated CD8⁺ T cells to secrete lytic granules and to kill tumor cells (Fig. 5L and M).

To test whether tumor MoDC are indeed required for T-cell immunity, we generated CD11c-DTR BM chimeric mice and challenged them with B16F10 cells. Once tumor reached a palpable size, mice were injected with diphtheria toxin to systemically deplete DC followed by adoptive transfer of gp100-reactive T cells. Although injection of T cells to tumor-bearing control mice induced tumor regression, their administration to DC-depleted mice was almost inert (Fig. 6A). Because depletion of CD11c is not restricted to MoDC, we next tested if blocking monocyte migration will affect T-cell immunotherapy. Mice bearing palpable B16F10 tumors were treated with anti-CCL2 antibodies followed by adoptive transfer of gp100-reactive T cells, or intertumoral injections of TNFα+anti-CD40 and anti-TRP1 antibodies. In both experiments, blocking monocyte migration significantly reduced the efficacy of the administrated immunotherapies (Fig. 6B and C).

We also tested whether increasing the viability and activity of MoDC in late-stage tumors will improve the efficacy of tumor-reactive CD8⁺ T cells. Mice were injected with B16F10 and tumors were allowed to grow for at least 10 days, until they reached an average size of 50 mm², when adoptive transfer of CD8⁺ T cells is no longer effective. Mice were then treated with gp100-reactive CD8⁺ T cells alone or in combination with anti-CD40 and GMCSF, which we found to increase MoDC numbers and viability in tumors (Fig. 6D and E). The effect of CD8⁺ T cells alone or in combination with GMCSF was benign and did not reduce tumor burden. In sharp contrast, injection of T cells in combination with anti-CD40, with or without GMCSF, induced significant tumor regression (Fig. 6F).

We next sought to determine if these finding are also applicable to human melanoma patients. Initially, we assessed whether a gene expression signature of activated MoDC correlates with improved survival of patients with melanoma. We performed an unsupervised analysis of gene expression profiles of samples of 468 patients with melanoma obtained from TCGA repository (37) using PROMO software tool (http://acgt.cs.tau.ac.il/promo/). Patient's profiles were clustered into four groups, based on the expression levels of 16 genes associated with activated MoDC (Fig. 7A). Kaplan–Meier plot of these four groups indicated a direct correlation between the activated MoDC and improved survival (Fig. 7B). Although the gene signature of activated MoDC may correlate with other immune subsets, this analysis highlights their prevalence as a predictive prognostic marker. In addition, dataset analysis of patients with melanoma from the Human Protein Atlas repository indicated that proteins predominantly expressed on myeloid DC correlate with improved survival of patients with melanoma (Supplementary Fig. S6A and S6B).

To test if MoDC are needed to license the cytotoxic activity of tumor-reactive T cells, tumor cells from a patient with melanoma were incubated with autologous T cells isolated from the tumor and PB. Tumor-infiltrating T cells induced apoptosis in about 15% of tumor cell, whereas blood T cells were almost inert. Addition of autologous tumor-infiltrating MoDC (SSC^lo/MHCII^hi/CD11c⁺/CD14^lo-int) to the blood T cells promoted tumor cells' killing comparable with that observed with tumor-infiltrating T cells (Fig. 7C and D).

Histologic sections of primary melanoma from patient resistant to PD-1 treatment showed high levels of cleaved caspase-3 in the majority of tumor-infiltrating DC (Fig. 7E). Finally, consistent with our results in mice, exosomes secreted from melanoma cell line were found to induced apoptosis in MoDC isolated from healthy donors (Fig. 7F and G).

Although immunotherapies for melanomas have been proven effective even in disseminated and metastatic disease (4, 9), only a small subset of patients experience a complete curative response (8). Therefore, a huge scientific effort has been invested into identifying the factors that limit treatment efficacy. Several meta-analysis studies of clinical samples have attempted to identify a gene signature that will predict responsiveness of patients with cancer to immunotherapies (3, 38). The net parameters that determine whether a patient will respond may depend on the driver mutations, mutational load, antigen presentation machinery, and neoantigen burden (39, 40), as well as on immune parameters in the tumor microenvironment, such as the type, prevalence, cytokine signature, and diversity of TCR of infiltrating T cells (3, 38, 41, 42).

Here, we leveraged the unresponsiveness of large melanoma tumors in mice to immunotherapies, to identify immune elements that govern the host immune response to melanoma. We found that tumor-infiltrating MoDC undergo apoptosis as melanoma progresses; and in their absence, T-cell–based therapies will not work. Examination of clinical biopsies indicated a positive correlation between tumor-infiltrating myeloid and MoDC, and favorable clinical outcomes in cutaneous melanomas (20, 43–45). Substantial reduction of DC in the epidermis along tumor progression has been reported starting three decades ago in patients with melanoma (43, 44, 46, 47). Esche and colleagues were the first to report that tumor-derived factors induce apoptosis in murine and human DC (48). Attempts to identify which melanoma-derived factor induces DC apoptosis suggested that GM3 and GD3 gangliosides induce significant DC apoptosis in culture through activation of reactive-oxygen species (49, 50), alternatively, MUC2 secreted from tumor cells have been suggested to promote MoDC apoptosis through ligation to Siglec-3 (51). Notably, this is the first time that melanoma-derived lysosomes are suggested to be the main mechanism that induces apoptosis in MoDC through induction of inflammasome and caspase-1. It remains unclear why uptake of lysosomes by tumor-associated macrophages is not followed by apoptosis.

Most studies attributed the positive correlation between tumor-infiltrating DC and favorable prognosis to the capacity of DC to migrate into sentinel lymph nodes, where they prime tumor-reactive T cells (19, 20, 52–55). We found that MoDC are different than cDC and epidermal DC, as they do not migrate to lymphoid organs, but rather remain in tumors and provide stimulatory signals to tumor-infiltrating T cells. Consistent with the seminal work of Randolph (56), which found that monocytes engulfing latex beads migrate to DLN and mature to become DC, we found monocyte, but not mature MoDC, containing tumor antigens in lymphoid organs. Numerous clinical trials have attempted to pulse ex vivo autologous MoDC with tumor antigens and reinfused them into the patient in order to elicit T-cell immunity (57, 58). Although routes of injections play an important role in facilitating this treatment (59), our work provides another explanation as to why injection of MoDC did not meet expectations by activating T cells in in vitro assays.

The exact nature of the licensing signal that tumor-infiltrating MoDC provide to T cells remains unclear. Marigo and colleagues have demonstrated that tumor-infiltrating Tip DC, which is a subset of MoDC, activate the infiltrating T cells through cross-linking of CD40 ligand (60). Nonetheless, because CD40L lacks a signaling chain, it is more likely that they activate the CD40 of Tip-DC and that the licensing signaling is secondary to CD40–CD40L interactions. Indeed, the activation of MoDC with agonistic anti-CD40, which blocks subsequent interactions with CD40L on T cells, induces significant tumor cell lysis by T cells.

Overall, this work redefines the role of MoDC in melanomas, and potentially in other tumors; it suggests that their main role is not to present tumor antigens in sentinel lymph nodes, but rather to provide in situ licensing signals to infiltrating T cells. Improving DC viability in patients who do not respond to immunotherapy may increase treatment efficacy substantially.

No potential conflicts of interest were disclosed.

Conception and design: N. Santana-Magal, P. Rider, Y. Carmi

Development of methodology: N. Santana-Magal, I.L. Linde, Y. Carmi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): N. Santana-Magal, A. Gleiberman, O. Zlotnik, I.L. Linde, N.E. Reticker-Flynn, Y. Carmi

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. Gleiberman, D. Rasoulouniriana, L. Tal, D. Netanely, R. Shamir, R. Blau, I.L. Linde, P. Rider, Y. Carmi

Writing, review, and/or revision of the manuscript: N. Santana-Magal, L. Farhat-Younis, A. Gutwillig, A. Gleiberman, L. Tal, M. Feinmesser, P. Rider, Y. Carmi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Santana-Magal, L. Farhat-Younis, A. Gutwillig, M. Feinmesser, O. Zlotnik, P. Rider, Y. Carmi

Study supervision: P. Rider

Other (technical help in performing studies and analyzing data): D. Rasoulouniriana

Other (surgeon involved in tumor tissue retrieval, selection, and preservation): H. Gutman

This work was funded by the Swiss Bridge Foundation, the Israel Cancer Association (Grant Number 20180041), and the Israel Science Foundation (Grant Number 2262/18).

The authors would like to deeply thank Rabin Medical Center institutional Tissue Bank and especially Dr. Adva Levi-Barda for their invaluable support of this research.

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.