DC-HIL/Gpnmb Is a Negative Regulator of Tumor Response to Immune Checkpoint Inhibitors Free

Antigen-specific T cells are activated by a net-stimulatory balance arising from competition between costimulatory and coinhibitory signals transduced by binding of receptors and ligands expressed on the surface of T cells and antigen-presenting cells or suppressor cells (1). The coinhibitory molecules, also termed “immune checkpoints,” include CTLA4/CD80 and PD1/PDL1 (2).

Many cancers that overexpress immune checkpoint molecules exude a net-inhibitory balance that can dampen natural anticancer immunity (3). Immune checkpoint inhibitors (ICI) were developed to circumvent this problem, blocking delivery of coinhibitory signals and thus amplifying antitumor responses (4). Among these ICI, anti-PD1/PDL1 mAb (e.g., nivolumab, pembrolizumab, and atezolizumab) have produced improved objective response rates, including 10% to 20% in cases of non–small cell lung carcinoma (NSCLC; ref. 5), 12% to 28% for melanoma (6, 7), approximately 43% for bladder cancer (8), and 10% to 20% for renal cell cancer (9). Despite these unprecedented benefits, the majority of metastatic cancers have initial resistance to ICI therapy (10). Thus, the development of reliable biomarkers that identify patients who will not respond to ICI therapy before treatment or even during the course of treatment could bring a great benefit to cancer treatment.

We identified a new immune checkpoint pathway comprising the DC-HIL [also known as Gpnmb (11)] receptor and its T-cell ligand, heparan sulfate-attached syndecan-4 (SD4; refs. 12, 13). DC-HIL is expressed constitutively by some cancer cells (including melanoma, breast and lung cancers; refs. 14, 15) and can be induced on myeloid-derived suppressor cells (MDSC) that proliferate exponentially in blood of patients with various malignancies (16). Patients with metastatic NSCLC produced 0.3% DC-HIL-expressing MDSC among peripheral blood mononuclear cells (PBMC), which was significantly higher than healthy controls (0.05%). DC-HIL–expressing cells release the soluble form (sDC-HIL) by cleaving off the extracellular domain by ADAM10 proteinase (17). Both DC-HIL and PDL1 receptors are expressed on tumor cells and mediate cancer-induced immunosuppression (18), but their interactions with suppressor cells differ: PDL1 mediates regulatory T-cell (Treg) function but is less important to MDSC (12, 16). In contrast, DC-HIL is the critical mediator of MDSC function and insignificant to Treg (12). Moreover, their respective T-cell ligands (SD4 and PD1) reside in disparate loci and signal independently of each other: PD1 associates with the CD3ζ chain of the T-cell receptor complex (19), whereas SD4 is separate from this complex (20). Therefore, DC-HIL and PDL1 pathways regulate T-cell response by distinct mechanisms.

To evaluate the hypothesis that DC-HIL counteracts the augmented T-cell response by ICI treatment, we examined whether patients with advanced NSCLC express sDC-HIL in the blood and whether the high levels associate with poor response to ICI therapy. We found high blood sDC-HIL levels to correlate strongly with resistance to ICI therapy in patients with advanced NSCLC. To study mechanisms by which DC-HIL attenuates response to ICI, we searched mouse tumor lines resistant to anti-PDL1 mAb but sensitive to anti-DC-HIL mAb. Among four tumor lines examined, B16 melanoma was distinct in its contrasting responses to the two treatments, such that cancer regression followed anti-DC-HIL mAb infusion, whereas only a marginal effect was achieved by anti-PDL1 treatment (18, 21). However, similar experiments using mice devoid of the DC-HIL gene (DC-HIL KO) resulted in regression of melanoma in response to anti-PDL1 mAb, indicating that resistance to anti-PDL1 treatment in wild-type mice involved expression of host DC-HIL. These findings were also confirmed with mouse LL2 lung cancer. Our studies suggest that host DC-HIL expression may be a biomarker to predict tumor response of patients with advanced NSCLC to ICI therapy.

Patients with advanced NSCLC (n = 76, Supplementary Table S1) who had ICI treatment between 2015 and 2017 at Harold C. Simons Comprehensive Cancer Center at The University of Texas Southwestern Medical Center, Dallas, TX, were included in this study (22). Patients were treated with standard of care ICI monotherapy (atezolizumab, nivolumab, or pembrolizumab) every 2 to 3 weeks (23, 24). Response to treatment was determined at every 6 to 12 weeks by RECIST version 1.1 (25). Each patient provided informed consent before enrollment. The Institutional Review Board of UT Southwestern approved this study (STU-082015-53). Peripheral blood samples were collected from all patients before treatment (week 0) and some patients provided samples at 2 and 6 weeks posttreatment. Plasma was obtained by centrifugation (at 2,000 g for 10 minutes) and supernatant collected and stored at −80°C. The study was conducted in accordance with the amended Declaration of Helsinki and the International Conference on Harmonization Guidelines.

Leukocyte marker Ab and antimouse PDL1 mAb (10F.9G2) were purchased from eBioscience and Bio-X-Cell, respectively. We generated UTX103 rabbit antimouse DC-HIL mAb (14) and the chimeric UTX103 IgG (UTX-m16) was constructed by replacing the C-region of UTX103 rabbit IgG with that of mouse IgG₁. These Abs were produced by transient transfection of suspension cultures with the IgG genes using ExpiCHO systems in serum-free media (Thermo Fisher Scientific) and purified by Protein A-agarose (Invitrogen). Purified preparations comprise very low endotoxin level (<0.05 EU/mL), determined by Pierce Chromogenic Endotoxin Quant Kit. Female C57BL/6 mice (∼8-week-old) and pmel-1 TCR transgenic mice were purchased from Harlan Breeders and The Jackson Laboratory, respectively. Animals were housed in the pathogen-free facility of the Institutional Animal Care Use Center of UT Southwestern Medical Center. All animal protocols were approved by the Center. Mouse tumor lines were purchased from the American Type Culture Collection, Old Town Manassas, VA, and maintained in DMEM supplemented with 10% FCS. Human NSCLC cells were obtained from Hamon Center for Therapeutic Oncology Research, UT Southwestern, Dallas, TX.

For leukocyte isolation and culture, PBMC from blood (∼15 mL) of patients with NSCLC (n = 3) were divided into two parts: one was stained with fluorescently labeled anti-HLA-DR and anti-CD14 and analyzed by flow cytometry for receptor expression; the other was treated sequentially with anti-CD14 Ab-coated magnetic beads (for monocytes) and anti-CD15-beads (granulocytes). The rest contained B and T cells. These fractions were cultured for 2 days with GM-CSF (100 ng/mL) or IL2 (20 unit/mL) and assayed for mRNA expression and sDC-HIL levels in culture supernatant.

An aliquot (100 μL) of 5× PBS-diluted plasma samples was applied to ELISA wells in triplicate (human GPNMB R&D Systems). For measuring mouse sDC-HIL levels in blood and culture supernatant, 1:100-diluted or undiluted samples were used. To allow reliable and unbiased method comparison, all tests were done in a single-blind manner by an independent laboratory technician.

pGuide-it-ZsGreen vector (Clontech Laboratories Inc) containing DC-HIL-targeted gRNA (CCGGCCGAAGACCAGCCACGTAAT) and CAS9 was transfected into B16 melanoma cells with Lipofectamine LTX with PLUS Reagent (Life Technologies) in Opti-MEM I (Life Technologies). After transfection, DC-HIL-deleted B16 cells were enriched by cell sorting using BD FACS Aria (BD Biosciences) and confirmed by immunoblotting for null DC-HIL expression.

Tumor cells were harvested from the growing cultures by ethylene diamine tetra-acetic acid (EDTA), washed extensively with PBS, and injected intravenously into mice [(2–5) × 10⁵ cells/mouse] via tail vein. Six days postinjection, all mice were given intraperitoneal injection of mAb (200 μg/mouse) every 2, 3, or 4 days at the total of six injections. Two days after the last injection (day 18), mice were scored for metastatic indices, lung weight, number of foci, and melanin content in the tissue (18). For other tumor cells, GFP-transfected cells were injected intravenously into mice, and lung-metastatic tumor cells were counted by flow cytometry: Lung was perfused with PBS and single-cell suspension isolated by digesting with collagenase/dispase (Roche), collagenase type I/IV, and DNAase I (Sigma-Aldrich) for 90 minutes at 37°C and passed through a 40-μm membrane. For experiments measuring survival rate, all mice were injected intravenously with B16 cells (1 × 10⁵ cells/mouse) and day 6 they were randomly sorted into four groups (n = 10) and given different mAb every 2 days until day 16; thereafter, schedule was changed to longer intervals; every 3 days until day 28; and every week until day 60.

All mice were anesthetized, blood drawn from heart, and serum prepared using BD Microtainer. After sacrificing mice, single-cell suspensions [(1–5) × 10⁵ cells/reaction] from lung were prepared and fluorescently stained with marker Ab (1–10 μg/mL) to each leukocyte subpopulations, including CD45, CD4, CD8 T cells and their SD4⁺ or PD1⁺ subset, and CD11b⁺Gr1⁺ MDSC and their DC-HIL⁺ or PDL1⁺ subsets. After fluorescently labeling with secondary Ab (1 μg/mL), stained cells were analyzed by FACS Verse (BD Biosciences).

On day 5 postintravenous injection of B16 cells (2 × 10⁵ cells/mouse) into DC-HIL^−/− mice, mice were randomly sorted into four groups (n = 5) and given intravenous injection of PBS or CD11b⁺Gr1⁺ MDSC (1 × 10⁶ cells/mouse) purified from spleen cells of WT or DC-HIL^−/− mice bearing subcutaneous B16 melanoma (2 weeks after subcutaneous implantation; ref. 26). CD11b⁺Gr1⁺ fraction from WT mice contained 40% to 50% DC-HIL⁺ cells, and that from KO mice had no such cells. On day 6, all mice were treated with anti-PDL1 or control IgG every 2 days.

Previously, we established LL2 cell line transfected with a vector driving expression of sDC-HIL-V5 under the control of Tet-off system (27). All mice were given 2 mg/mL doxycycline (Dox)-water from day 3 to day 6. Day 0, all mice were injected intravenously with Tet-Off-sDC-HIL-V5-LL2 cells (2 × 10⁵ cells/mouse) and day 6 randomly sorted into 2 groups: one group was continuously supplied with Dox (2 μg/mouse intraperitoneal injection every 5 days) and the other discontinued (PBS injection). Since day 7, mice were treated with intraperitoneal injection of anti-PDL1 or control IgG every 2 days (n = 5). On day 17 (2 days after the last injection), lung was scored for metastasis indices. Growth of TetOff-sDC-HIL-LL2 cells in lung was measured by clonogenic assays: Total lung cells were seeded in a DMEM-complete media with 600 μg/mL G418 (Sigma-Aldrich). After 10 days in culture, colonies were fixed, stained with crystal violet, and counted. Tumor growth was quantified by colony-forming units (CFU) in total lung cells (lung-CFU).

To assess gp100-specific T-cell response in mAb-treated mice, spleen cells were prepared and seeded onto ELISPOT wells at varying cell densities with gp100 peptide (5 μg/mL) for 2 days. The IFNγ-producing cells were counted using ELISPOT assay (eBioscience). For Ag-nonspecific response, spleen cells were cultured for 3 days with anti-CD3/CD28 mAb (each 1 μg/mL) and Brefeldin A (Invitrogen), followed by surface staining of CD4 or CD8 and intracellular staining of IFNγ.

This PCR was performed with primers for DC-HIL, ADAM10, or GAPDH (13, 28). mRNA expression in each sample was expressed as the expression level relative to GAPDH gene, which was quantitated using the comparative Ct method and the formula 2-ΔΔCT.

Mann–Whitney U test or Student t test was used for comparison. Survival rate and association were evaluated by log rank (Mantel–Cox test) and Pearson's correlation coefficient, respectively. Slope of sDC-HIL levels was analyzed using a mixed-effects model. A P ≤ 0.05 was considered significant.

We chose NSCLC because of the high prevalence and high expression of DC-HIL by NSCLC cells and circulating MDSC (29). We asked whether patients with NSCLC produce sDC-HIL in the blood. Recruited patients (n = 76 and the demographics are summarized in Supplementary Table S1) were measured for tumor size before ICI treatment (6, 24) and evaluated for response to treatment at weeks 6 to 12 (25). Plasma samples were collected at week 0. Blood sDC-HIL was detected at pretreatment (week 0, ≤25 ng/mL) and the levels did not correlate with baseline tumor size (Fig. 1A). We then examined correlation of baseline sDC-HIL levels and tumor response (Fig. 1B), characterizing cases as having progressive disease (PD, n = 33), stable disease (SD, n = 38), or partial response (PR, n = 5) at the first evaluation. Responders (SD and PR) displayed blood sDC-HIL levels (3.3 ng/mL ± 1.2) that are not different from healthy donors (2.9 ± 0.5, P = 0.9). In contrast, nonresponders (PD) had significantly higher sDC-HIL levels (P < 0.0001 by Mann–Whitney U test) at the baseline. The extent of variation in sDC-HIL levels is similar to that of serum vascular endothelial growth factor, a marker associated with poor outcomes to chemotherapy (30). Among nonresponders, the percent change in sDC-HIL levels within the first 6 weeks after treatment correlated significantly with percent change in tumor size (Fig. 1C, P < 0.00001).

We also examined fluctuations in sDC-HIL during treatment (Fig. 2). In 28 cases for which sDC-HIL levels were available for more than three time points (weeks 0, 2, and 6), we analyzed individual kinetics. Among nonresponders, 14 of 15 cases showed increasing or persistently elevated levels (>3.3 ng/mL, the median of SD/PR groups), whereas one case had low levels at all the time (<3.3 ng/mL). For responders, 12 of 13 cases had decreasing or persistently low levels, with 1 PR patient 94 who showed increasing levels. We then focused on patients (n = 13) plotted in the zone of 4 to 6 ng/mL sDC-HIL overlapping between SD/PR and PD groups (Fig. 1B). Partial response patients 10 and 98 expressed 5.1 ng/mL sDC-HIL at baseline, which was higher than 3.3 ng/mL. This measure gradually fell at later time points, during which time the treatment response was characterized as PR. Patient 70 (representative of 6 patients) had a baseline measure of 4.3 ng/mL that rose to 11 of 13 ng/mL at weeks 2 and 6; this case's treatment response was judged as PD. In the case of patient 117 (representative of 5 patients), the baseline value of 5.8 ng/mL went down to 3.3 and 2.3 ng/mL and this treatment response was SD. We tested for significant differences in slope among the three groups using a mixed-effects model. The estimated slopes and their 95% confidence intervals (CIs) are as follows: PD: slope = 0.468, 95% CI: 0.273–0.662. PR: slope = −0.145, 95% CI: −0.579 to 0.289. SD: slope = −0.203, 95% CI: −0.441 to 0.035. There was significant difference in slope between SD and PD (P = 0.001). There was no significant change in sDC-HIL levels of healthy donors in 2 weeks (Supplementary Table S2). These cases illustrate fluctuations in blood sDC-HIL levels during the treatment, with considerable variation in magnitude and direction of the changes.

We probed for which cell types as likely producers of sDC-HIL in patients with NSCLC (Fig. 3). To examine DC-HIL expression on tumor cells, we used three NSCLC lines established from cancer biopsies. Among these lines, only H1957 NSCLC line expressed DC-HIL on the cell surface; it was greater than SK-MEL-28 melanoma cell line (Fig. 3A) but lower levels of DC-HIL mRNA (Fig. 3B). In contrast, all cells expressed PDL1 on the surface constitutively. Despite high levels of surface expression, H1957 line did not produce sDC-HIL in the culture supernatant even when the cell expressed ADAM10 mRNA. For leukocytes, PBMC from patients with NSCLC (n = 3) were fluorescently stained with anti-HLA-DR and anti-CD14 Ab and sorted into four fractions; Fr. 1 (HLA-DR^no/lo CD14⁺ cells); Fr. 2 (HLA-DR^med/hi CD14⁺); Fr. 3 (HLA-DR^med/hi CD14^neg); and Fr. 4 (HLA-DR^no/lo CD14^neg). Each fraction was assayed for DC-HIL and PDL1 expression (Fig. 3C). Among these fractions, HLA-DR^no/lo MDSC and HLA-DR^med/hi CD14⁺ cells expressed highest levels of DC-HIL (ΔMFI: 91 and 62, respectively), with no expression in the other fractions. We then examined sDC-HIL production by these leukocytes. Because monocytic MDSC are a minuscule fraction, PBMC were sorted into CD14⁺ monocytes, CD15⁺ granulocytes, and the other (CD14^neg CD15^neg, which contain B and T cells). DC-HIL mRNA was highest in CD14 monocytes, with no expression by CD15 granular cells (which contain granulocytic MDSC) nor by other leukocytes (Fig. 3D). DC-HIL⁺ cells also expressed ADAM10 mRNA and secreted sDC-HIL into culture (54–113 pg/mL). Thus, CD14 monocytes, but not tumor cells nor granulocytes, may be the primary source of sDC-HIL.

To study mechanisms of how DC-HIL attenuates beneficial effect of ICI therapy, we searched mouse tumor lines for the model that exhibits ICI resistance. To measure antitumor activity of anti-PDL1 mAb, we employed experimental metastasis that is the simplest and ideal model because tumor cells are accurately quantified and the target of therapies is often the endpoint of metastasis (31).

We first examined the effects of ICI on pre-established lung mets of LL2 lung cancer. Syngeneic mice were infused intravenously with GFP-transfected LL2 lung cancer cells (that do not express DC-HIL); 6 days later, all mice were treated with anti-DC-HIL, anti-PDL1 mAb alone, or in combination (200 μg/mouse) injected intraperitoneally (six injections total at intervals of 2, 3 or 4 days; Fig. 4A). Mice treated with control IgG yielded 1,526 ± 439 GFP⁺ LL2 cells per 10⁶ lung cells. This index was reduced by anti-PDL1 mAb to 74 ± 37 cells and by anti-DC-HIL to 32 ± 13 cells (P = 0.08 vs. anti-PDL1; Fig. 4B and C). Combined treatment did even better (9 ± 4 cells, P = 0.01 vs. anti-DC-HIL). Monotherapy and combined treatment similarly increased CD4 and CD8 T cells in tumor-infiltrating lymphocytes (TIL; including SD4⁺, PD1⁺, and IFNγ⁺ subsets; Supplementary Table S3). Interestingly, the same treatment led to reductions in % CD4 (but not % CD8) Treg and % MDSC in TIL. These MDSC expressed DC-HIL but not PDL1 (Supplementary Fig. S1A). We also examined RM9 prostate cancer and noted similar results (Fig. 4D and E): both mAb individually inhibited growth of RM9 lung mets, and combined treatment achieved better results (P = 0.01 to anti-DC-HIL), including reduced CD4⁺ Treg and elevated CD8⁺ IFNγ response (Supplementary Table S3). Even more dramatic benefits from monotherapy and combined treatment were noted for MC38 colon cancer (P = 0.001 between combination vs. anti-DC-HIL alone; Fig. 4F and G), with enhanced T-cell activation phenotype and decreased MDSC in tumor microenvironment (Supplementary Table S3). Thus, all three-tumor models are highly sensitive to both anti-PDL1 and anti-DC-HIL treatment.

We then examined effects of anti-PDL1 mAb on pre-established B16 melanoma metastasis in lung. Similarly, C57BL/6 mice were infused intravenously with B16 melanoma cells; 6 days later, all mice were treated with anti-DC-HIL or control IgG (200 μg/mouse) injected intraperitoneally (six injections total at different intervals of 2, 3, or 4 days). On day 18, we quantified the following metastasis indices: lung weight, number of foci, and melanin content per lung (Fig. 5A), while also measuring blood sDC-HIL. Anti-DC-HIL mAb reduced all metrics in a dose-dependent manner, with the greatest effect noted for the most frequent treatment protocol every 2 days. We then compared the efficacy of anti-DC-HIL with anti-PDL1 mAb: B16 cells express the targets of both treatments (18). Anti-DC-HIL mAb markedly inhibited growth of B16 lung mets and increased the IFNγ response in draining lymph nodes (P < 0.01, Fig. 5B), whereas anti-PDL1 had a marginal effect. Surprisingly, combined treatment with both mAbs negated the benefit from anti-DC-HIL mAb (Fig. 5B). Antitumor activity of anti-DC-HIL mAb is most likely due to its ability to block interaction of DC-HIL with its ligand because the mAb cannot transduce the DC-HIL-specific intracellular signals (14) and because it blocks binding of DC-HIL to activated T cells (Supplementary Fig. S2).

We also examined effects of mAb treatment on survival rate of mice with B16 lung metastases (Fig. 5C). Tumor-bearing mice were treated as before, but this time the mAb treatments were sustained through 60 days. All control mice died by day 19 and all anti-PDL1–treated mice died by day 20. In contrast, 60% of anti-DC-HIL–treated mice remained alive, surviving through day 35 (P = 0.0004), with the exception of 1 mouse surviving by day 60. Similar results were noted in a repeat experiment. The longest surviving DC-HIL–treated mouse with melanoma appeared to develop memory T-cell response to B16 melanoma (Supplementary Fig. S3). Thus, B16 melanoma offers the great opportunity to study the role of DC-HIL molecules in suppressing response to ICI treatment.

We next examined the immune status of lung mets during the course of ICI therapy (over 14 days). Reproducibly, anti-PDL1 treatment decreased by 40% of blood sDC-HIL levels on day 14, whereas anti-DC-HIL did by 80% (Supplementary Fig. S4A). CD45⁺ TIL were isolated from B16 lung mets and analyzed for frequency of immune cell subsets (Supplementary Fig. S4B). In control and treated mice, TIL were increased on day 7, followed by a reduction on day 14. Control mice showed the progressive increase in CD4 and CD8 T cells within TIL. Anti-DC-HIL mAb led to a greater increase over control in percentage of CD4 and CD8 T cells and percentage of SD4⁺ and PD1⁺ subsets and in their IFNγ responses. Anti-PDL1 mAb also produced rises in T cells and IFNγ responses but to a lesser degree compared with anti-DC-HIL mAb treatment. Combined treatment had no added value over monotherapy. With respect to MDSC, in control mice percent CD11b⁺Gr1⁺ MDSC in TIL was unchanged through 14 days, but there was a marked increase in the number of MDSC per gram of lung (Supplementary Fig. S4). Interestingly, DC-HIL⁺MDSC subset expanded rapidly from 0.2% on day 0 to 45% on day 14, with hardly any MDSC expressing PDL1. Anti-DC-HIL mAb prevented MDSC expansion greater than did anti-PDL1 mAb, and combined treatment negated these effects. Thus, the two mAb appeared to compete with each other in their impact on the tumor microenvironment.

Among tumor lines we used, B16 melanoma is unique in expressing DC-HIL (ref. 27; Supplementary Fig. S5). We thus examined whether tumor-DC-HIL expression accounts for resistance to anti-PDL1 therapy. We generated DC-HIL-knocked out (DC-HIL^−/−) B16 cells that do not express DC-HIL but retain high expression of PDL1. This engineered cell line had extensive in vitro growth ability, but with significantly slower in vivo tumor growth ability than parental B16 cells (18). Using this cell line, we examined influence of tumor-DC-HIL to anti-PDL1 response (Fig. 6A). DC-HIL^−/− B16 cells showed as a poor response to anti-PDL1 as did the parental cells, indicating that tumor-DC-HIL has no influence to this resistance. We next addressed whether host DC-HIL affects anti-PDL1 response (Fig. 6B). DC-HIL^−/− mice bearing B16 lung metastases were treated with anti-PDL1 or control IgG. Anti-PDL1 treatment led to a remarkable decrease in all metastasis indices in DC-HIL^−/− mice, including lung weight (0.23–0.18 g), melanin content (0.53–0.03 mg), and number of foci (127–14), with a 10-fold increase in IFNγ response. Although there was no difference in CD45⁺ TIL number between anti-PDL1 and control treatments, anti-PDL1 mAb increased total CD4 and CD8 T cells and their SD4⁺ and PD1⁺ subsets, while decreasing total MDSC in the lung (Fig. 6C; Supplementary Fig. S6).

Because DC-HIL is also expressed by nonhematopoietic cells, we dissected influence of hematopoietic cell- versus nonhematopoietic cell-derived DC-HIL to anti-PDL1 response. We generated BM-chimeric mice, in which hematopoietic cells are originated from transplanted BM cells of DC-HIL^−/− mice. Deletion of DC-HIL gene from hematopoietic cells (while nonhematopoietic DC-HIL expression remains) failed to reverse the high response to anti-PDL1 (Fig. 6D), indicating that DC-HIL on nonhematopoietic cells is dispensable for B16 melanoma resistance to anti-PDL1 treatment.

Because melanoma-bearing hosts express DC-HIL as the receptor on MDSC and the soluble form (sDC-HIL), we addressed which molecular form is more important in regulating tumor responsiveness to anti-PDL1 mAb. For the impact of DC-HIL receptor, we examined whether adoptive transfer of DC-HIL⁺ MDSC into DC-HIL^−/− mice bearing B16 lung mets reverses anti-PDL1 tumor response (Fig. 6E; Supplementary Fig. S7A). As control, DC-HIL^neg MDSC were also transferred into the host. Reproducibly, anti-PDL1 inhibited growth of B16 lung mets almost completely in DC-HIL^−/− mice. This antitumor activity was diminished markedly by infusion of DC-HIL⁺ MDSC, whereas injection of DC-HIL^neg MDSC had no effect. For sDC-HIL, we used LL2 cells (which do not express DC-HIL) transfected with Tet-Off-controlled sDC-HIL gene (Fig. 6F). Immediately after intravenous injection of the LL2 transfectant, mice were given Dox for 6 days and then sorted into two groups; one was continued with Dox and the other was discontinued. These groups were also treated with anti-PDL1 or control IgG and scored for metastasis indices on day 17. sDC-HIL remained very low through day 6 but quickly increased to 8.4 to 9.6 ng/mL after Dox discontinuation, and low sDC-HIL levels (1–1.4 ng/mL) were maintained in Dox-continued group that exhibited very high response to anti-PDL1 mAb compared with IgG-treated mice (Fig. 6G). In contrast, doxycycline-discontinued group (sDC-HIL expressed) had markedly reduced antitumor activity. This reduced response was associated with increased total MDSC or DC-HIL⁺ subset of lung metastases (Supplementary Fig. S7B). Furthermore, adoptive transfer of DC-HIL⁺ (but not DC-HIL^neg) MDSC into DC-HIL-KO mice bearing LL2 tumor markedly reversed the benefit of anti-PDL1 treatment (Supplementary Fig. S8). Thus, both DC-HIL⁺ MDSC and sDC-HIL converted tumor cells from an anti-PDL1-responsive to resistant tumor.

Primary resistance to ICI therapy has been attributed to tumor-cell deficiency in PDL1 expression or in IFNγ signaling (and thus insensitivity to IFNγ killing) and/or host defects in mounting T-cell responses (32, 33). T-cell suppressors also play a pivotal role particularly in attenuating the tumor Ag-specific T cells restored by intervening the PD1 pathway (34, 35). We underscore the importance of this suppressor mechanism by finding the significant correlation of high blood sDC-HIL levels and poor response to ICI immunotherapy for patients with advanced NSCLC. Furthermore, our animal studies have shown the possible interaction of host-derived T-cell-inhibitory DC-HIL molecules with the restored T cells by anti-PDL1 mAb in B16 melanoma and LL2 lung cancer.

In tumor-bearing hosts, DC-HIL is expressed as a transmembrane protein on myeloid cells and as a soluble form (sDC-HIL) secreted into peripheral blood. Although both forms share common structures, they are not identical in function. We have previously identified MDSC as the most abundant myeloid-cell sector expressing DC-HIL receptor in B16 melanoma-bearing mice (26). Lung metastases of B16 melanoma (and other tumors) recruited DC-HIL⁺MDSC subsets such that these cells comprised up to 45% of lung-TIL. Surprisingly, PDL1⁺MDSC subset numbered very few among lung-TIL, despite the fact that it comprised approximately 30% among total MDSC in subcutaneous B16 melanoma and in spleen of tumor-bearing mice (26). We speculate that lung-infiltrating DC-HIL⁺MDSC, which are unaffected by anti-PDL1 mAb and suppress the function of T cells revitalized by anti-PDL1 mAb.

Unlike the cellular DC-HIL receptor, binding of sDC-HIL to the ligand SD4 on T cells is incapable of transducing its inhibitory signals, suggesting that sDC-HIL interferes with binding of DC-HIL receptor to the ligand (36, 37). In disagreement with this functional property, soluble PDL1 (sPDL1) was reported to transmit its inhibitory signal through activated CD4⁺ T cells (38). Recently, we showed sDC-HIL to also bind to the surface of endothelial cells (EC) of blood vessels with higher avidity than to T cells (28). Intriguingly, sDC-HIL bound to the surface of EC prevents selectively transendothelial migration of T cells into cancer tissues (17, 28). These findings led us to hypothesize that blood sDC-HIL hampers recruitment of T cells into tumors, resulting in reduced tumor-CD8 density, a requisite for successful anti-PDL1 treatment (39).

Combining anti-DC-HIL with anti-PDL1 mAb produced synergistic outcomes for treatment of all tumors except B16 melanoma, in which the latter mAb neutralized the effect of anti-DC-HIL mAb. Because combination treatment led to elevated expression of sDC-HIL in the host, we presume that treatment of B16 melanoma cells with anti-PDL1 mAb upregulated sDC-HIL expression that saturated the Ag-binding capacity of anti-DC-HIL mAb.

Given high costs, potential toxicities, availability of alternative therapies, and limited predictive value of existing biomarkers, ICI therapy can benefit from the identification of new markers that more reliably discriminate between response and resistance. Tumor-PDL1 expression has been used to predict ICI response for many cancers, but the lack of a standard method to measure this marker has made it inconsistent at best (40). Furthermore, PDL1-negative tumors may still benefit from ICI, making clinicians reluctant to base clinical decisions on the result. sPDL1 and sPD1 are generated by alternative splicing and secreted by cancer cells into blood, but their prognostic ability suffers from contradictory findings: one study of adenocarcinomas correlated high blood levels with better prognosis (41), whereas other studies of diffuse large B-cell lymphoma, hepatocellular carcinoma, and lung cancer showed a converse outcome (42, 43). Tumor-infiltrating lymphocytes are a pre-existing barometer of the T-cell response within tumors; however, its value is rendered null in most cases because the cancer is already bereft of T cells. Genetic profiling of cancers (tumor mutation burden, tumor cell gene expression, and microsatellite instability) provides good biologic characterization but its predictive ability has yet to be established (44). These foregoing markers reflect properties intrinsic to the cancer. In contrast, TIL and peripheral blood markers like DC-HIL on MDSC and sDC-HIL echo the immune response (45). In addition, blood elements have the advantage of being readily assayed and in a repeated fashion during the course of treatment.

Gpnmb/DC-HIL has been studied as a treatment target for melanoma, triple-negative breast cancer (TNBC), and NSCLC. Gpnmb mRNA level within tumors correlated with poor prognosis for melanoma (46) and increased risk for relapse of TNBC (47). The NSCLC cell lines express Gpnmb and secrete the soluble form, which can promote tumor growth in immunocompromised mice (48). However, we failed to detect sDC-HIL in the culture supernatant of NSCLC cells expressing DC-HIL on the cell surface. Our study showed CD14 monocytes (including MDSC) to be the primary secretors of sDC-HIL, which is a predictive marker for response of patients with NSCLC to ICI.

In sum, DC-HIL offers the promise of useful predictive blood markers (receptor on MDSC and the soluble moiety sDC-HIL) that embody a real-time measure of the cancer's interaction with the immune system. Because our animal studies suggest that the significance of elevated DC-HIL levels as a predictive marker is not limited to treatment of lung cancer with ICI, our insights may be applicable to other malignancies and their immunotherapy. Nevertheless, adequately designed large prospective studies are warranted to establish blood sDC-HIL as a predictive biomarker for immunotherapy.

No potential conflicts of interest were disclosed.

Conception and design: D.E. Gerber, K. Ariizumi

Development of methodology: J.-S. Chung, M. Kobayashi

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.-S. Chung, V. Ramani, M. Kobayashi, V. Popat

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.-S. Chung, S. Zhang

Writing, review, and/or revision of the manuscript: J.-S. Chung, V. Popat, S. Zhang, P.D. Cruz, D.E. Gerber, K. Ariizumi

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Fattah

Study supervision: K. Ariizumi

We are grateful to Irene Dougherty for technical assistance and Rolf Brekken for NSCLC cell lines. This study was primarily supported by VA Merit Award (1I01BX004069) and Department of Defense Lung Cancer Research Program (W81XWH-18-1-0312) and also partially supported by American Cancer Society/Melanoma Research Alliance (132330-MRAT-18-114-01-LIB/597700), The University of Texas SPORE in Lung Cancer, and the Harold C. Simmons Comprehensive Cancer Center's Biomarker Research Core (1P30-CA142543–03).

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.