Wild-type BALB/c and C57BL/6N mice, aged 8 to 16 weeks, were obtained from Japan SLC. All mice were used in accordance with protocols approved by the respective institutional review board (IRB). They were housed under specific pathogen-free (SPF) conditions, with ad-libitum water and food supply to maintain their health and well-being.

CT26 and LLC cell lines were cultured in RPMI-1640 medium (catalog # 189-02025, FUJIFILM Wako Pure Chemical Co., Ltd. Wako) with 10% (v/v) FBS (Lot # 2166446, Gibco) and 1% (v/v) Penicillin/Streptomycin mixed solution (catalog # 26253-84, Nacalai Tesque).

CT26 (1×10⁶) or LLC (5×10⁵) cells were injected intradermally (i.d.) into the right flank of BALB/c or C57BL/6N mice respectively on day 0. Mice were treated with Paquinimod (catalog # SML2883-25MG, Sigma Aldrich) dissolved in DMSO (catalog # 031-24051, FUJIFILM Wako Pure Chemical Co., Ltd. Wako) and diluted to 1mg/ml in PBS (catalog #14249-24, Nacalai Tesque) by intraperitonal (i.p.) injection daily, from day 0 to day 11. Control mice were injected with solvent solution (DMSO diluted in PBS). For CPI treatment, mice were injected i.p. with anti-PD-L1 monoclonal antibody (clone 1-111A, produced in-house, 50 µg/mouse) every 3 days, starting day 6 after tumor inoculation. Treatment with recombinant mouse S100A9 protein, CF (R&D, catalog # 2065-S9) (2 µg/mouse in 20 µl in PBS) was administered intratumorally on days 7 and 11. Control mice received an intratumoral injection of an equal volume (20 µL) of PBS (Nacalai Tesque). Tumor volumes were measured every other day using electronic calipers, and the tumor volume was calculated using the formula: tumor volume = π × (length × breadth × height)/6. Mice were sacrificed either on day 12 for early analysis of tumor-infiltrating immune cells in tumor, or on day 19-21 for growth curve analysis, total tumor and spleen weight and late analysis of immune cells in spleen and tumor.

Mice were euthanized and tumors and spleens harvested into tubes containing RPMI-1640 medium. Tumor samples processed by first mincing into 1- to 2-mm pieces using scissors and then digestion using collagenase type II (catalog # CLS2, Worthington Biochemical Corporation) (100 units/mL) and DNase I (catalog # 11284932001, Roche) (20 µg/mL) on a gentleMACS Dissociator (catalog # 130-093-235, Miltenyi Biotec). The digested tumor tissue was passed through a 100 μm filter to generate a single cell suspension, washed and resuspended in RPMI-1640 with 10% FBS. Spleens were homogenized by mashing through a 100 µm filter and subsequently treated with ammonium-chloride-potassium (ACK) (catalog # A1049201, Thermofisher Gibco™) lysing buffer for 5 minutes to remove red blood cells. Spleen cell suspensions were then washed and resuspended in RPMI with 10% FBS.

Single cell suspensions obtained from harvested organs or primary cell cultures were seeded on v bottom plates, washed in PBS and stained for 10 min at room temperature using the Zombie Aqua™ Fixable Viability Kit (catalog # 423102, BioLegend) to evaluate cell viability. Subsequently, cells were washed with a buffer solution containing PBS, 2% FBS, and 2 mM EDTA (catalog #06894-85, Nacalai Tesque), and treated with Trustain FcX (BioLegend, catalog #101320) for 10 minutes at 4°C, to minimize non-specific binding by blocking Fc receptors. Finally, cells were stained with a master mix of fluorescent antibodies for 30 minutes at 4°C, washed and resuspended in flow buffer. FMO controls were used to define positive and negative events. Samples were analyzed on the SONY ID7000 spectral analyzer and data processing was performed using FlowJo software v10.10 (BD Biosciences). Quantitative analysis of the flow cytometry data is detailed in the “Statistical analysis” section below.

Cells were stained with the following antibodies: Anti-FoxP3 (catalog #11-5773-82, clone PJK-169, Invitrogen), Anti-I-A/I-E(catalog #107606, clone M5/114.152, BioLegend), Anti-CD11b (catalog #101226, clone M1/70, BioLegend), Anti-CD45 (catalog #564279, clone 30-F11, BioLegend), Anti-CD86 (catalog #564199, clone 6L1, BioLegend), Anti-CD8a (catalog #50-1886-U100, clone 2.43, Tonbo Biosciences), Anti-F4/80 (catalog #123110, clone BM8, BioLegend), Anti-CD80 (catalog #104708, clone 16-10A1, BioLegend), Anti-CD4 (catalog #100536, clone RM4-5, BioLegend), Anti-Arg-1 (catalog #17-3697-82, clone A1exF5, Invitrogen), Anti-Ly6G (catalog #127622, clone 1A8 BioLegend), Anti-Ly6C (catalog #560593, clone AL-21, BD Pharmingen), Anti-CD274 (catalog #124336, clone 10F.962, BioLegend), Anti-CD279 (catalog #135218, clone 29F.1A12, BioLegend).

Ly6G⁺ cells from the spleens were isolated from spleen single-cell suspensions by employing the Anti-Ly-6G MicroBeads UltraPure kit (catalog #130-120-337, Miltenyi Biotec) following the manufacturer’s recommended protocols. Spleens from naive, control, and Paquinimod (Sigma-Aldrich) treated mice were harvested and pooled based on treatment groups. The spleens were mechanically dissociated by pressing them through a 70 µm cell strainer using a syringe plunger to obtain a single-cell suspension. The cell suspension was centrifuged at 400×g for 5 minutes, and the pellet was resuspended in ACK (Ammonium-Chloride-Potassium) lysis buffer (Gibco) to lyse red blood cells. After another centrifugation at 400×g for 5 minutes, the cells were resuspended in a buffer containing phosphate-buffered saline (PBS) (Nacalai Tesque), 0.5% bovine serum albumin (BSA) (catalog # A8531, Sigma-Aldrich), and 2 mM EDTA (Nacalai Tesque), The cells were labeled with Anti-Ly-6G MicroBeads UltraPure (Miltenyi Biotec) by adding 10 µL of MicroBeads per 10⁷ cells and incubating at 4°C for 10 minutes. After washing to remove unbound MicroBeads, the cell suspension was applied to a MACS^® LS Column (catalog #130-042-401, Miltenyi Biotec) placed in the magnetic field of a MACS Separator (catalog # 130-042-303, Miltenyi Biotec). The labeled Ly6G⁺ cells were retained in the column, while unlabeled cells passed through. The column was removed from the magnetic field, and the retained Ly6G⁺ cells were eluted.

Effector Ly6G⁺ cells were generated from single-cell suspensions of spleens from both treated and naïve mice using Ly6G⁺ beads sorting, as described above. A portion of naïve splenocytes from non-tumor bearing mice were kept unsorted for use as target cells. Effector Ly6G⁺ cells of interest were seeded in triplicates in varying numbers in U-bottom 96-well plates (catalog # 353077, Corning Incorporated) in RPMI-1640 medium (Wako) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS) (Gibco) and 1% (v/v) penicillin-streptomycin mixed solution (Nacalai Tesque). Target cells, consisting of total splenocytes harvested from naïve mice, were stained with the CellTrace-CSFE Cell Proliferation kit (catalog # C34554, Invitrogen) and added at 200.000 cells/well density to the effector cells, yielding different ratios of effector to target cells. T cells in the splenocyte target population were stimulated by adding anti-CD3 and anti-CD28 monoclonal antibody-coated Dynabeads (catalog # 11453D, Thermo Fisher Scientific, Gibco). The co-cultures were incubated for 4 days at 37°C in 5% CO₂. Subsequently, the cells were harvested, stained for CD11b (catalog # 101226, clone M1/70, BioLegend), CD4 (catalog # 100536, clone RM4-5, BioLegend) and CD8 Anti-CD8a (catalog # 50-1886-U100, clone 2.43, Tonbo Biosciences) and analyzed using the SONY ID7000 spectral cell analyzer. Proliferation was determined by measuring the decrease in CellTrace-CFSE fluorescence intensity in CD4⁺ and CD8⁺ T cells. Splenocytes stimulated with anti-CD3 and anti-CD28 monoclonal antibody-coated Dynabeads alone was used as the control and were considered 100% proliferated.

Ly6C^high cells from bone marrow cells were obtained from the femur and tibia of naïve C57BL/6 mice by flushing the bones with PBS containing 2% FBS and 2 mM EDTA. After lysing erythrocytes, the remaining cells were washed and resuspended in PBS with 2% FBS and 2 mM EDTA for fluorescence-activated cell sorting (FACS). Bone marrow cells were stained with Zombie Aqua (catalog # 423102, BioLegend) for identification of live cells, followed by blocking Fc receptors with anti-CD16/32 (catalog #101320, BioLegend) antibodies. The following antibodies were used for subsequent staining: Anti-CD11b APC-Cy7 (catalog # 101226, clone M1/70, BioLegend), Anti-Ly6G APC (catalog # 127614 clone 1A8, BioLegend), Anti-Ly6C PE-Cy7 (catalog #560593, clone AL-21, BD Pharmingen), and Anti-CD45 FITC (catalog # 103108, clone 30-F11, BioLegend). The Ly6C^high target cell population to be used in the migration assay, identified as Zombie^–CD45⁺CD11b⁺Ly6G^–Ly6C^high cells, was sorted using a BD FACSMelody^™ sorter (BD Biosciences). The purity of the sorted cells was confirmed, and cells were resuspended in RPMI-1640 containing 10% FBS (Gibco) and 1% pen/strep (Nacalai Tesque) for use in the Boyden chamber migration assay. Gating strategy for sorting and results from the purity assessment is shown in Supplementary Figures 8 and 9 .

The Boyden chamber assay was performed using upper and lower compartments separated by a porous membrane with an 8 μm pore size (catalog # 3422, Corning Incorporated). FACS- sorted Ly6C^high bone marrow cells were seeded in upper chamber (insert). To test the effect of the inhibitor, cells in the lower chamber were treated with or without various doses of the inhibitor under these conditions: RPMI-1640 (Wako) medium alone (negative control), RPMI-1640 (Wako) medium with the chemoattractant complement C5a (catalog # 2150-C5-025/CF, R&D) (1 µg/mL, positive control), RPMI-1640 medium with recombinant mouse S100A9 protein (R&D), CT26 tumor cell line supernatant, CT26 supernatant with recombinant mouse S100A9 protein (R&D). The CT26 supernatant was collected from CT26 cells cultured to confluence in T75 flasks. After reaching confluence, the supernatant was harvested, centrifuged at 400 g for 5 minutes to remove residual cells, and filtered through a 28 mm 0.45 µm SFCA membrane syringe filter (catalog # 431220, Corning Incorporated) to ensure purity. The inhibitor was added to the lower chamber at various doses, to assess its effect on migration under these conditions.

After 24 hours of incubation at 37°C in 5% CO2, migrated cells in the lower chamber were stained with NucBlue™ Live ReadyProbes™ Reagent-Hoechst 33342, (catalog # R37605, Invitrogen) as per the manufacturer’s instructions. Briefly, cells were stained with 1 drop/2 mL NucBlue™ Live ReadyProbes™ Reagent-Hoechst 33342 and incubated for 15 minutes at 37°C and 5% CO2. Images of the Transwell lower chamber bottom were taken with a 10x objective using a Keyence BZ-X800 All-in-One Fluorescence microscope (Keyence Corporation). The number of migrated cells stained with NucBlue™ Live ReadyProbes™ Reagent-Hoechst 33342 was counted using Keyence BZ-800 image analysis software Hybrid Cells count module and Macro cell count options (Keyence Corporation).

Statistical analysis was performed using Prism v9.3 (Graphpad). Treatment groups were compared using the two-sided unpaired t test (for two groups), or one-way or two-way ANOVA (for data with one or two independent variables, respectively) with Tukey’s multiple comparisons test (for more than two groups). No technical replicates were used to derive statistics in this study. All statistical analyses have been performed using 3 or more biological replicates. For all experiments, the alpha level was set to 0.05. All p values are displayed, with *, ** and *** indicating p < 0.05, p < 0.01 and p < 0.001, respectively.

During initial investigations, we searched for correlations between the responsiveness to anti-PD-L1 responsiveness and immune cell infiltration in multiple tumor models. The mice injected with CT26 colon carcinoma cell line showed typically two patterns of tumor growth: a small number of mice (2 out of 10 mice) responded well to anti-PD-L1 treatment and rejected the tumors, while the majority of mice showed only a partial response to the treatment, with tumors progressing, although at reduced speed ( Figure 1A ). We found that the tumor burden tightly correlated with splenomegaly ( Figure 1B ). The main cellular subset associated with splenomegaly was that of Ly6G⁺ granulocyte-like cells, which were significantly increased in tumor-bearing mice compared with the naïve mice, or mice responding to anti-PD-L1 treatment and rejecting the tumors ( Figure 1C ). Other immune cell subsets evaluated, like CD8⁺ T cells, CD4⁺ T cells or macrophages were not altered (data not shown, gating strategy shown in Supplementary Figure 1 ). The expanded splenic Ly6G⁺ cells from tumor-bearing mice showed typical functional phenotype of MDSCs, as they effectively inhibited the proliferation of T cells ex vivo ( Figure 1D ). Such immunosuppressive phenotype was not observed with Ly6G⁺ cells from naïve, non-tumor bearing mice ( Figure 1D , suppression assay outline shown in Supplementary Figure 2 ). In anti-PD-L1 non-responding mice, both the frequency and the suppressive function of Ly6G⁺ cells were similar to that of untreated tumor-bearing mice ( Figure 1C ). In contrast, anti-PD-L1 responding mice, lacked both the expansion of Ly6G⁺ cells as well as the suppressive function of these cells, which were similar to those in naïve mice ( Figures 1C, D ).

The frequencies of immune cells infiltrating the CT26 tumors were quite high, with CD45⁺ cells making up about 40% of live cells, with further significant increase to 60% in mice responsive to anti-PD-L1 therapy ( Figure 1E ). A large proportion of tumor-infiltrating immune cells were myeloid cells, mainly F4/80⁺ tumor associated macrophages (TAMs) and monocytic (Ly6C^high), with smaller populations of CD4⁺ T cells, CD8⁺ T cells and granulocytic (Ly6G⁺) myeloid cells ( Figure 1F , gating strategy shown in Supplementary Figure 3 ). Anti-PD-L1-responsive mice showed a significantly higher frequency of CD8⁺ T cells and Ly6C^high myeloid cells infiltrating the tumor, indicating that these cells contribute to a successful anti-tumor immune response, in addition to the absence of systemic Ly6G⁺ suppressive cells. Thus, the CT26 appears to be a good model to test drugs aiming of inhibiting the generation and function of suppressive myeloid cells and improve efficacy of checkpoint inhibitor (CPI) therapies.

Previous studies, utilizing genetic knockdown, monoclonal antibodies and pharmaceutical inhibitors, have identified the protein S100A9 or the protein complex S100A8/A9 as playing a major role in the generation of MDSCs as well as suppressive TAMs (8, 9, 29). S100A9 either as homodimers or heterodimers with S100A8, were found binding and signaling through TLR4, in addition to RAGE and other receptors (2, 30). The interaction between S100A9 and TLR4, and possibly also with the receptor RAGE, can be blocked using the inhibitor Paquinimod, a quinoline-3-carboxamide derivative. Due to its previous testing in a clinical setting, safe clinical profile and blocking of both human and mouse S100A9 signaling through TLR4, we decided to investigate this drug. Specifically, we tested whether Paquinimod could target the detrimental differentiation of myeloid cells into MDSCs and/or reduce the systemic accumulation of these suppressive myeloid cells, leading to the suppression of anti-tumor immunity and CPI resistance.

The mice were injected with Paquinimod intraperitoneally, from day 0 to day 12 after CT26 tumor inoculation ( Supplementary Figure 4 ). Contrary to our expectations, Paquinimod treatment resulted in increased CT26 tumor growth and also eradicated the responsiveness of CT26 to anti-PD-L1 ( Figure 2 ). There were individual variations within the treatment groups, but Paquinimod injection alone or in combination and anti-PD-L1, resulted in even larger variations (maximal and minimal) tumor volume within the respective groups ( Figure 2A ). In mice injected with Paquinimod, we observed a significant increase in tumor volume from day 15 ( Figure 2A ) and a significant doubling in average tumor weight at day 19 ( Figure 2B ). Thus, combining Paquinimod with CPI resulted in increased tumor growth in non-responding mice and abrogation of the overall reduction in tumor growth compared with control mice. The protumor effect of Paquinimod was reproduced in the LLC tumor model ( Supplementary Figure 5 ). Paquinimod injection at later timepoints showed no pro or anti-tumor effect, and neither affected the growth of CT26 tumor cells in vitro (data not shown). Thus, it seems that there is a critical time-window during early tumor growth, for a beneficial effect of S100A9 on anti-tumor control, which is blocked by Paquinimod. These findings suggest that extracellular S100A9 signaling restricts tumor growth indirectly and contribute to CPI response, through early immune-mediated control of tumor growth.

Following the results from the tumor challenge experiments, we were curious to know whether the induction of suppressive myeloid cells, i.e. spleen Ly6G⁺ myeloid population, was affected by Paquinimod treatment. Based on a previous study reporting the role of S100A8/A9 in sustaining the accumulation of MDSC in all lymphoid tissues, we expected that Paquinimod would result in a systemic reduction of the tumor-induced expansion of myeloid cells (8). We found, however, that Paquinimod injection alone did not impact the expansion of myeloid cells, including Ly6G⁺ cells in the spleen, and the splenomegaly was not reduced but rather exacerbated by Paquinimod ( Figures 3A, B ).

T cell proliferation assays using Ly6G⁺ cells isolated from spleens of control or Paquinimod-treated tumor bearing mice, as well as naïve, non-tumor bearing mice, revealed that Paquinimod did not change the function of spleen Ly6G⁺ cells, which suppressed T cells to the same degree as Ly6G⁺ from tumor-bearing control mice ( Figure 3C ). We conclude that Paquinimod treatment does not provide any relief from myeloid immune suppression and that the generation of both suppressive Ly6G⁺ and BM-MDSCs likely do not require extracellular S100A9 signaling through TLR4.

To explore as to why Paquinimod treatment caused increased tumor growth, we harvested tumor tissue on day 12, and analyzed the infiltrating immune cells by flow cytometry. It is worth noting that at this timepoint, the tumor size was relatively similar between control and treated groups. This is important, as we have observed variations in immune cell infiltration in untreated tumors that correlate with tumor size (data not shown). By analyzing at day 12, an early timepoint, we expect most variations in tumor-infiltrating cells to be caused by the treatment and not by the differences induced by variations in tumor mass and their potential to influence local and systemic immune system.

We found that the frequency of tumor-infiltrating immune cells among live cells was reduced by more than half in Paquinimod-treated groups, compared with non-treated or anti-PD-L1 only control groups ( Figure 4A ). This result indicates that the pro-tumor effect of Paquinimod treatment is caused by lower immune cell infiltration, causing a less effective anti-tumor immune response. The composition of immune infiltrating cells was also changed by Paquinimod treatment. The frequency of Ly6C^high myeloid cells, the most prevalent immune cell among live cells in tumor at this early timepoint, was strongly reduced in both Paquinimod and anti-PD-L1 and Paquinimod groups compared with untreated and PD-L1 groups, respectively ( Figure 4B ). Also, there were smaller, albeit significant reductions in the frequency of both CD4⁺ T cells and CD8⁺ T cells ( Figure 4C ).

Both TAMs and Ly6C^high express high levels of TLR4, one important receptor engaged by extracellular S100A9 (2). We investigated if these cell subsets were altered in their activational state, by means of functional marker expression. We found no significant differences in the expression of the immune suppressive markers, like arginase (Arg)1 and CD206 (data not shown). For the immune activating markers, like CD80 and MHC-II, we found an increase in CD80 expression for TAMs and MHC-II expression for Ly6C^high after Paquinimod treatment ( Figures 4D, E ), but this did not reach a statistic significance. We conclude that, the strongest effect of Paquinimod treatment in this experimental setup is considerable reduction of Ly6C^high monocytic cells, associated with an overall reduction in immune cell infiltration.

The findings above suggest that S100A9 signaling would be beneficial for the initial anti-cancer immune responses. We tested this hypothesis, by directly injecting recombinant S100A9 into the tumors. We chose to do intratumoral injection of the protein at an early therapeutic window, that is, at day 7 after tumor inoculation, when the tumor size is big enough to control the injection site, and then, at day 11, just before the rapid tumor cells expansion, which typically start at day 12.

We found that recombinant S100A9 injection significantly reduced CT26 tumor growth from day 17, compared with untreated controls ( Figure 5A ). When examining individual mice, there was large variation within the groups, with typically only 1 responder among 9-10 mice that completely rejected the tumor and with the remaining mice clearly showing a delayed tumor growth ( Figure 5A ). There were no significant differences in final tumor weight ( Figure 5B ) or spleen weight ( Figure 5C ) in the S100A9 injected mice compared to untreated mice. In contrast to the Paquinimod treatment, S100A9 injection did not cause any clear changes to tumor immune cell infiltration ( Supplementary Figure 7 ). Taken together, these findings confirm an early anti-tumor effect of S100A9 in this mouse tumor model.

S100A9 is known to induce chemotaxis of neutrophils and MDSCs (9, 31). In the context of COVID-19, the induction of neutrophils by S100A9 was shown to be effectively blocked by Paquinimod (28). In our experiments, early inhibition of S100A9 with Paquinimod caused decrease in overall immune cell infiltration to tumors in CT26 mouse tumor model. Ly6C^high monocytic cells in the TME were the population that displayed the strongest reduction, but Ly6G⁺ neutrophils were not affected ( Figure 4B ). To investigate whether Paquinimod affects Ly6C^high cell chemotaxis directly, similar to previous work on neutrophils, we performed Boyden chamber assay, testing the migration capacity of Ly6C^high monocytes sorted from mouse bone marrow in response to different stimuli with and without Paquinimod. Gating strategy for FACS sorting of Ly6Chigh cells from bone marrow is shown in Supplementary Figure 8 and results from purity assessment is shown in Supplementary Figure 9 . We used recombinant S100A9 protein and CT26 tumor cell supernatant (CT26 sup) as migratory stimuli.

We found that S100A9 induced statistically significant migration of Ly6C^high cells, which was abolished by Paquinimod treatment at either 10 or 30 µg/ml ( Figure 6A ). CT26 supernatant had an even stronger chemotactic effect, which was also significantly inhibited by simultaneous treatment with Paquinimod at 30 µg/ml ( Figure 6B ). Paquinimod treatment in vitro did not show any direct cytotoxic effect on Ly6Chigh cells specifically or total immune cells from bone marrow in general ( Supplementary Figures 10, 11 ). We conclude that Paquinimod inhibits the chemotactic effect of S100A9 on Ly6Chigh cells, as well as reduce the chemotactic effect of CT26 tumor cell supernatant. These in vitro results indicate that Paquinimod likely functions by counteracting the potent migratory signals released from the tumors for the recruitment of proinflammatory Ly6C^high monocytes, which would provide a mechanistic explanation for the pro-tumor effect of Paquinimod when administered at earlier stages of tumor development.