Caspase-1 from Human Myeloid Derived Suppressor Cells Can Promote T cell–Independent Tumor Proliferation

Study approval

Our study was performed in compliance with approved Johns Hopkins Hospital and Vanderbilt University Medical Center Institutional Review Board protocols in accordance with the Belmont Report. All subjects provided written informed consent prior to participation, and the specimens were de-identified prior to analysis. Included were HNSCC patients undergoing surgery for curative intent. Excluded were those with autoimmune diseases and chronic steroid use. Fresh tumor specimens (typically 1cm³) were harvested from non-margin areas and processed in PBS within 2 hours of resection.

Reagents and Cells

Cell Proliferation Dye eFluor 450 ™ (eBioscience) was used for cell proliferation assays, according to the manufacturer’s protocol. Anti-human CD11b-PE (clone ICRF44), anti-human HLA-DR (clone, G46–6)-FITC, anti-human CD14–PerCP-Cy5.5 (clone, mφP9), anti-mouse Gr-1–Alexa Fluor 700 (Biolegend, clone: RB6–8cC5), anti-mouse CD11b–PE-Cy7(clone: M1/70) and Ly-6C-PE (clone AL-21) were obtained from BD Pharmingen. Cell lines used included human head and neck cancer cell lines Cal27, SCC-1, and SCC-25, and human monocytic cell line THP-1. All cell lines were tested free for Mycoplasma. Cell lines were authenticated with GenePrint10 (Promega) through JHH Genetic Resource Core Facility and stored in −70deg C in aliquots when not in use. Cells were cultured with RPMI-1640 medium, with 10% FCS, 10% L-glutamine, 10% Na-pyruvate, 10% Penicillin/Streptomycin and 0.1% beta-ME. All reagents were from Gibco®. When co-cultured with sorted MDSC, the FCS concentration was 1–2%. CD14⁺ monocytes from healthy non-age matched volunteers were used for controls. Murine melanoma cell line B16, colon-cancer cell line CT26, and pancreatic cancer cell line Panc02 were obtained from ATCC. MOC1 murine head and neck cancer line was obtained from Ravi Uppaluri (Dana Farber).

MDSC isolation and flow cytometry analysis

Human MDSCs, characterized as CD11b⁺CD14⁺HLA-DR^low/– were sorted from freshly obtained peripheral blood or tumor tissue from HNSCC patients undergoing surgical treatment at Johns Hopkins Hospital, Johns Hopkins Bayview Medical Center, or Vanderbilt University Medical Center. Peripheral blood mononuclear cells (PBMCs) were separated by Ficoll gradient prior to sorting with antibodies against human (h)CD11b, hCD14, and hHLA-DR conjugates, which were stained at 4°C for 30 minutes in dark at 1:50–200 dilution in 50μl FACS buffer prior to sorting or flow cytometry. Cells were sorted with a MoFlo MLS sorter (Beckman Coulter), as previously described (15). The purity of sorted cells > 90%. All flow data were analyzed under Flowjo software. Human MDSCs were cultured in 2% FCS media.

Murine MDSCs, characterized as Gr-1⁺CD11b⁺ or CD11b⁺Ly6C^high, were sorted from splenocytes of 1–5 C57BL/6 mice (Jackson) whose tumor (B16, Panc02, MOC1) measured 0.125 cm³. Splenocytes were harvested under sterile condition in RPMI culture media with 10% FCS, filtered with 40μm cell strainer, and RBC lysed with ACK buffer treated for 5 minutes. Fresh splenocytes were stained with antibodies under 4 °C in dark for 30min prior to sorting, and these MDSC were transferred intravenously as indicated in the in vivo tumor models below.

Proliferative index analysis

Sorted MDSCs were cultured in a 37°C incubator with 5% CO₂ in 1–2% FCS until use, Thawed tumor cell lines (human HNSCC cell lines-Cal27, SCC-1, SCC-25 [ATCC, Johns Hopkins Tissue Core] or murine cancer cell lines CT26 or B16) were passaged once, and they were pre-stained with cell proliferation dye eFlour450 (eBioscience) at a concentration of 5 μM at room temperature for 10 minutes and washed with PBS three times. Tumor cells and sorted MDSCs were cocultured in round-bottom 96-well plates for 3 days. 1–5×10⁵ MDSCs/well were typically used for these assays, with MDSC:tumor cell ratio ranging from 5:1 to 10:1. Stained tumor cells cocultured with peripheral monocytes from healthy individuals were used as controls at the same cell ratios.

Percentage of proliferating tumor cells was measured using the eFlour450 fluorescence signal detected by flow cytometry. Proliferative index was also measured using Ki67 staining of tumor cells. After co-incubation with MDSCs for 3–4 days, tumor cells were harvested and fixed with 5 mL 75% ethanol (added dropwise while vortexing) for 3 hours at −20°C, and the fixed tumor cells were stained with FITC-conjugated mouse anti-human Ki-67 (BD Pharmingen, room temperature, 30 minutes in dark). Immediately prior to flow cytometry analysis, 10 μL of propidium iodide (PI) staining solution was added to the cells to assess viability. Flow cytometry was used to analyze the stained cells.

Annexin V-FITC (ab14085; Abcam) staining to check for tumor cell apoptosis was assayed according to manufacturer’s instructions. MDSCs’ immunosuppression activity was assessed in control MDSCs that were not co-cultured with tumor cells at days 3–4 to ensure these cells were functionally T-cell immunosuppressive (see Supplementary Fig. S1). Briefly, matched CD4+ T-cells from the donors were stimulated with CD3/CD28 with or without sorted mMDSC from HNSCC patients. Either T-cell IFNγ production or CFSE dye dilutional assays were used to confirm the T-cell suppressive activity of mMDSC.

For cytokine incubation with eFlour450 stained Cal27 cells, we used the following concentrations – rhIL1b (500 pg/mL; R&D), IL18 (500 ng/mL; MBL Medical & Biological Lab), IL10 (5 ng/mL; R&D), IL6 (5 ng/mL), IL36 (50 ng/mL; R&D), IL33 (5 ng/mL; R&D), TGFβ (5 ng/mL; R&D), TNFα (100 pg/mL; R&D). 1×10⁵ Cal-27 cells were cultured in each well in 96-well-plate, flat bottom. 4–5 days after the incubation, cells were then collected to run flow cytometry for the proliferation index as described above.

Cytokine array

A Bio-Plex Pro Assays kit (Bio-Rad) was used to measure cytokines on the Bio-Plex system. The supernatant from sorted human MDSCs from peripheral blood or tumor tissue were collected 24 hours after the different treatment conditions. The assay was performed on a 96-well-plate. All samples, including standards, were assayed in duplicate or triplicate wells. Standard curve was generated using the Bio-Plex software, and all the values were within the effective detecting limits.

ELISA

200 μl of the supernatant from each well of cultured human MDSC under different treatments were collected at different time points for measurement of IL1β and IL18 using ELISA kits without dilutions (eBioscience, BMS618/2, MBL: 7620). 5×10⁵ murine MDSCs were activated with 50–100 ng/mL of LPS (Calbiochem) for 24 hours. Each assay had duplicated wells for the standard curve and samples. Concentration was calculated from the standard curve using the recombinant cytokines, per the manufacturer’s instructions.

Caspase-1 staining and inhibition

Fluorescent-labeled inhibitors of caspase-1 (FLICA) covalently bind to active caspase enzymes in cells. Caspase-1–specific FLICA 660 Assay kit (ImmunoChemistry) was used to stain the sorted human MDSCs at 37 °C for 1 hour, then washed 2 times with buffer provided in the kit, and counterstained with DAPI, per the assay protocol. FLICA stained cells (far red) were analyzed with a Hamamatsu Photonics C9100–02 EMCCD camera using the following filter set: Chroma #41137 – Excitation – HQ620/60; Chroma #39187 – Dichroic – Q660LP, and Chroma #40075 – Emission – HQ700/75. The objective used was an Olympus 40x NA 1.4, oil immersion. For caspase-1 inhibition, the caspase-1–specific inhibitor, Z-WEHD-FMK (BD Pharmingen ™), was added to cultured MDSCs at 80–100 μM for 24 hours (for cytokines analysis) or for 5 days (for cell proliferation assays).

Transwell system

HTS Transwell^® 96-well plates with 1.0 μm polyester membranes were used to determine the effect of MDSCs on tumor cells without cell-cell contact. 10⁵ sorted MDSCs were cultured in the upper chamber, with 10⁵ eFluor450 stained SCC-25 or Cal27 tumor cells in the bottom chamber. Cell proliferative index was determined on day 4–6.

In vivo tumor model

Caspase-1^–/– mice (Jackson Laboratory (B6N.129S2-Casp-1^tm1flv or NOD.129S2(B6)-Casp1tm1Sesh Casp4del/LtJ)) were housed in Johns Hopkins animal facility under institutional guidelines after animal IACUC protocol was approved. To generate chimeric mice, 5×10⁶ bone marrow cells (BMC) from caspase^–/– mice or wildtype C57BL/6 mice were adoptively transferred to lethally irradiated (10Gy) wildtype C57BL/6 mice. 8 weeks after bone marrow transfer, engraftment was confirmed by the presence of CD45⁺ PBMCs. After engraftment confirmation, B16, MOC1, or Panc-02 tumor cells (1–5 × 10⁵/mouse) were subcutaneously injected into the chimeric mice. Tumors were measured daily or every 2 days with calipers, and mice were sacrificed when the tumors measured 2 cm in the largest diameter. Depleting anti-CD3 (100 μg/mouse from InVivoMab, Lot number: 4773–2/0914 or clone 17A2 from Bio X Cell) were injected intraperitoneally twice a week in some cohorts. 2–5×10⁶ isolated Gr-1⁺CD11b⁺ MDSCs (Miltenyi, Cat#130–094-538) harvested from the spleen of B16-or Panc02-bearing wildtype mice were injected intravenously into caspase-1^–/– or wildtype mice with CD3 depletion on Day7, Day14 and Day21 after tumor injection. In some groups, splenic MDSCs were treated with racemic thalidomide (Sigma, Cat#1652500, 50 μg/mL) in RPMI supplemented with 10% FBS at 37°C, 5% CO₂ for 2–3 hours and then washed with PBS prior to adoptive transfer. Tumor sizes were measured with calipers daily, and tumor volumes were estimated using the formula V (cm³) = 3.14 × [largest diameter × (perpendicular diameter)²]/6.

SiRNA and qPCR

Pre-made inducible lentiviral particles suppressing the MyD88 gene (GenTarget Inc, #LVP 207) were used for siRNA targeting in Cal27 cells. Cal-27 cells at 65–70% confluency were transduced with lentiviral construct particles. The RFP reporter gene was used to sort transduced cells by flow cytometry. RFP⁺ cells were labeled as Cal27-MyD88 KO cells after validation. Total RNA was extracted and reverse transcribed to cDNA with High Capacity RNA-to-cDNA kit (Applied Biosystems). Concentrations of cDNA were measured with Nanodrop 1000 (Thermo Scientific). A Taqman^® gene expression assay that targets NFkB (Assay ID: Hs00765730_m1) was used to probe for NFκB expression in both the Cal27 and Cal27-MyD88 KO cells. The 18S rRNA was used as the internal control. 25 ng cDNA was used as the template, and 20μL reactions were carried out in 96-well plates using the Taqman Universal Master Mix II (2x) as the reaction cocktail. Double delta Ct (ΔΔCt method) analysis on NFkB transcripts was performed to provide its relative quantitation between the two cell lines (38). Assay was carried on the Applied Biosystems Veriti 9902 96-well thermal cycle PCR system. Thermal cycling parameters: 95°C for 10 minutes, denature at 95°C for 15seconds, and anneal/extend at 60°C for 1 minute, and then repeated for 40 cycles.

TCGA bioinformatics analysis

We compared RNA-Seq by Expectation Maximization (RSEM)-transformed caspase-1 gene expression with progression-free survival (Fig. 5A) and The Cancer Genome Atlas (TCGA)-defined subtype (Fig. 5B). The expression data was from the Broad Firehose (January 28^th, 2016; https://gdac.broadinstitute.org/), and the survival and subtype data were retrieved from the TCGA Network (41); for each calculation, all samples with available data were used. The Cox proportional hazards ratio, 95% CI (confidence interval), and Wald P value were calculated using the coxph function from the R Survival package (Therneau, “A package for survival analysis in S,” Jan 1999), where patients were partitioned into two categories using a caspase-1 gene expression Z-score of 1.0 (i.e., patients with a Z-score <1.0 in one category, and patients with a Z-score ≥ in the second category; Fig. 5A). Significant association between caspase-1 gene expression and the basal vs. non-basal TCGA subtypes was calculated as the Wilcoxon rank-sum P value (Fig. 5B).

MDSC gene expression analysis by microarray

Sorted monocytic (CD11b⁺/HLA-DR^low/CD14⁺) and granulocytic (g)(CD11b⁺/HLA-DR^low/CD15⁺ MDSC from peripheral blood of patients with advanced HNSCC as described (see above) were placed into Trizol (TRIzol^®Reagent), and RNA isolation was done using the RNeasy Mini Kit (QIAGEN, www.qiagen.com). RNA quality was confirmed by analysis on the Agilient 2100 Bioanalyzer. Biotin-labeled targets were generated from 20 ng total human RNA using the NuGEN WT-Ovation Pico assay, as per the manufacturer’s recommended protocols. Affymetrix Human Gene 1.0 ST microarrays (Affymetrix, Santa Clara, CA) were hybridized for 16 hours at 45° C with rotation (60rpm) as described in Affymetrix’s GeneChip Expression Wash, Stain, and Scan User Manual (www.affymetrix.com). Fluorescent signals were determined using Affymetrix’ GeneChip Scanner 3000 7G at default parameters described in the manufacturer’s GeneChip Expression Analysis Technical Manual. Images were analyzed using the Affymetrix Command Console version 3 (AGCC v3.0), and processed into CEL files at the manufacturer’s default settings. RMA normalized log2 signal values were extracted and summarized for gene-level analysis with the Partek Genomics Suite v6.6 (Partek Inc., St. Louis MO, USA). In some cases, the gene expression data for gMDSCs and mMDSCs were compared by one-way ANOVA using Partek, and the results were exported for examination and further evaluation to Spotfire DecisionSite (TIBCO Software Inc., Palo Alto CA, USA).

Statistics

The statistical analysis of the results was performed using Prism 6 software or Microsoft Excel v16 software. The ANOVA analysis was used to compare the differences among multiple groups. Two groups were compared using Fischer’s exact test or student’s t-test with a two-tailed distribution and calculated at 95% confidence. Differences were considered statistically significant at P<0.05. Kaplan-Meier curves were compared using the log-rank (Mantel Cox) test. When present, error bars reflect standard error of measurement (SEM).

MDSCs with increased inflammasome complex can promote tumor cell proliferation

We sorted monocytic CD11b⁺CD14⁺HLA-DR ^low/– MDSCs (mMDSCs), from head and neck squamous cell cancer (HNSCC) patients, and demonstrated MDSC T-cell suppressive activity (Supplementary Fig. S1), and we assessed relative mRNA expression profiles for the MDSCs via microarray (15). We noted increased expression of inflammasome complex mRNAs that were correlated with increased IL1β and IL18 secretion (Fig. 1A and B). To characterize the functional significance of the inflammasome signaling in human mMDSCs, we modeled the human TME using cocultures of multiple HNSCC tumor cell lines (Cal27, SCC-1, SCC-25) with sorted mMDSCs to examine whether MDSCs can directly regulate the proliferation of tumor cells. When fluorescently labeled cancer cells were cocultured with mMDSCs without T cells, the proliferation index of the tumor cells was significantly enhanced in all HNSCC lines assayed without any changes in apoptosis (Fig. 1C, left panels; Supplementary Fig. S2). Most of the MDSCs sorted from the 21 HNSCC patients showed this ability to promote tumor cell proliferation, whereas monocytes from healthy donors did not alter tumor cell proliferation (Fig. 1C, right upper panel). The pro-tumorigenic MDSC response was dependent on the ratio of MDSCs to tumor cells (Fig. 1C, lower right panel). MDSCs also increased the tumor cells’ expression of Ki-67, a marker of cell proliferation, corroborating the dye dilutional assays (Fig. 1D).

To investigate whether the upregulated inflammasomes in MDSCs were related to their promotion of tumor cell proliferation, we treated human mMDSCs with the caspase-1–specific inhibitor, Z-WEHD-FMK, which reduced the secretion of IL1β and IL18 prior to the co-incubation assay (Fig. 2A). A caspase-1–specific FLICA assay was used independently to corroborate that caspase-1 was active in situ in the sorted human MDSCs (Fig. 2B). When the caspase-1 inhibitor, Z-WEHD-FMK, was added to the MDSCs prior to co-culture, we found that MDSCs decreased their ability to induce tumor cell proliferation (Fig. 2C).

Paracrine mitogenic activity is higher in monocytic MDSCs

When we compared the mitogenic induction between sorted gMDSCs (CD15⁺CD14^–HLA-DR^low/–) to the mMDSCs from the same patients, we noted that mMDSCs had greater ability to promote tumor proliferation (Fig. 3A). When we compared the relative expression profiles between mMDSCs and gMDSCs, mMDSCs had greater expression of inflammasome complex mRNAs than gMDSCs (Fig. 3B). We reasoned that if the promotion of tumor proliferation is dependent on caspase-1/IL1β/IL18 signaling pathway, this phenotype would be a paracrine-dependent mechanism, and cell-cell contact would not be necessary. We performed the tumor proliferation assay in a transwell assay, and we noted that the presence of mMDSCs in the test chamber increased the proliferation index of tumor cells in the other chamber (Fig. 3C, P<0.001). We also showed that supernatants from human mMDSCs phenocopied the enhanced proliferative index of HNSCC cells co-incubated with MDSCs (Fig. 3D, right panel). When we followed the proliferative index of Cal27 with the incubation of recombinant human cytokines expressed by MDSCs, enhanced proliferation was noted when IL1β, IL18, IL6, and IL10 were combined, whereas IL1β and IL18 individually only had about 1.3 fold effect (Supplementary Fig. S3).

Myeloid caspase-1 directly promotes carcinogenesis

Because the in vitro cell mixing experiments do not replicate the complexity of the TME, we directed our attention towards in vivo characterization of myeloid caspase-1, in the context of a syngeneic tumor model. In order to probe whether MDSC caspase-1 can directly promote tumor proliferation in vivo, we used chimeric C57BL/6 mice adoptively transferred with bone marrow cells (BMCs) from caspase-1 null or wildtype control mice. We initially showed that murine Gr1⁺CD11b⁺ MDSCs expressed IL1β and IL18 (Supplementary Fig. S4), and its coculture with B16 or CT26 tumor cells promoted caspase-1–dependent tumor cell proliferation as we observed with human mMDSCs (Fig. 4A). In chimeric mice with caspase-1 null hematopoietic cells, B16 tumor growth was significantly blunted, regardless of whether CD3⁺ T cells were depleted or not (Fig. 4B, left upper panel). The T cell–independent effect of caspase-1 signaling was more prominent in this subcutaneous tumor injection model. We repeated the experiment with Panc02 pancreatic cancer cells (Fig 4B, right upper panel), as well as MOC1 oral cavity tumor cells (Supplementary Fig. S5) with the same results. When wildtype Gr1⁺CD11b⁺ MDSCs harvested from B16-bearing mice were adoptively transferred into caspase-1 null chimeric mice after tumor injection, the protective effect of absent caspase-1 signaling was reversed (Fig. 4B, left lower panel). To test whether pharmacologic blockade of MDSC inflammasome signaling can specifically phenocopy the caspase-1 null chimeric mice, we treated wildtype MDSCs with caspase-1 inhibiting thalidomide prior to their adoptive transfer into tumor-bearing mice that had caspase-1 deficiency and T cells depletion (40). Although the wildtype MDSCs restored their pro-tumorigenic effect, thalidomide-treated MDSCs failed to increase the pro-tumorigenic tumor growth rate of B16 in chimeric mice (Fig. 4B, lower right panel). Together these results show that caspase-1 activity in MDSCs could directly promote in vivo tumor growth, independent of T cells.

Caspase-1 expression in the TME correlates with worse prognosis

To investigate the role of caspase-1 as a potential prognostic marker in HNSCC patients, we correlated caspase-1 gene expression with survival and subtype using data from a TCGA HNSCC patient cohort (Fig. 5)(41). Elevated caspase-1 expression was significantly associated with decreasing progression-free survival in TCGA HNSCC patients (Fig 5A). Similarly, when we probed for caspase-1 expression in the basal subtype, which has previously been associated with worse clinical outcomes with relative absence of T-cell infiltration (39, 42), caspase-1 expression was significantly higher in the basal subtype than in the non-basal subtypes (Fig. 5B).

MDSCs’ promotion of tumor proliferation is dependent on tumor cell MyD88 signaling

Because both IL1β and IL18 signaling results in activation of tumor MyD88 signaling pathway to induce oncogenic transcription factors NFkB and AP-1 (27, 43), we inquired whether MDSCs’ stimulation of MyD88 signaling is partly responsible for the enhanced mitogenic signaling in HNSCC cells. We suppressed the MyD88 gene expression in Cal27 HNSCC cell line with siRNA that also expressed an RFP reporter gene, and we sorted the RFP-expressing Cal27 cells. MyD88 knockdown in Cal27 cells showed significantly lower NFkB transcripts as indicated by the higher number of NFkB qPCR cycling. (Fig. 6A). When we compared the proliferation index of wildtype Cal27 cells and the MyD88 suppressed Cal27 cells cocultured with MDSCs, the MDSC-dependent promotion of tumor cell proliferation was significantly blunted in the MyD88-suppressed tumor cells in vitro (Fig. 6B). We reasoned, therefore, that tumor cells deficient in MyD88 signaling would not be affected by myeloid caspase-1 signaling in vivo. Thus, we compared the tumor growth rate of B16 cells and B16 MyD88 knockdown cells in wildtype mice and found that MyD88 suppression significantly decreased the growth rate of B16 in vivo (Fig. 6C). However, using chimeric mice adoptively transferred with caspase-1 null BMCs under T cell–depleting conditions, MyD88 signaling in the B16 tumor cells resulted in no differential tumor growth rate in vivo compared to control tumor cells with transduction of scrambled shRNA (Fig 6C). The blunted tumor growth rates attributable to myeloid caspase-1 signaling seen previously in Fig. 4 were not found in these tumors cells lacking MyD88/NFkB oncogenic signals.