Uncovering therapeutic targets for macrophage-mediated T cell suppression and PD-L1 therapy sensitization

Summary

Tumor-associated macrophages (TAMs) and other myelomonocytic cells are implicated in regulating responsiveness to immunotherapies, including immune checkpoint inhibitors (ICIs) targeting the PD-1/PD-L1 axis. We have developed an ex vivo high-throughput approach to discover modulators of macrophage-mediated T cell suppression, which can improve clinical outcomes of ICIs. We screened 1,430 Food and Drug Administration (FDA)-approved small-molecule drugs using a co-culture assay employing bone-marrow-derived macrophages (BMDMs) and splenic-derived T cells. This identified 57 compounds that disrupted macrophage-mediated T cell suppression. Seven compounds exerted prominent synergistic T cell expansion activity when combined with αPD-L1. These include four COX1/2 inhibitors and two myeloid cell signaling inhibitors. We demonstrate that the use of cyclooxygenase (COX)1/2 inhibitors in combination with αPD-L1 decreases tumor growth kinetics and enhances overall survival in triple-negative breast cancer (TNBC) tumor models in a CD8⁺ T cell-dependent manner. Altogether, we present a rationalized approach for identifying compounds that synergize with ICI to potentially enhance therapeutic outcomes for patients with solid tumors.

Graphical abstract

Highlights

• •HTS screening of macrophage-targeting agents to improve anti-PD-L1 response in cancer
• •Identification of synergistic chemotherapeutic agents to enhance T cell activity
• •COX1/2 and PD-L1 inhibition reduce TNBC tumor growth in a CD8⁺ T cell-dependent manner

Kumar et al. demonstrate a high-throughput screening approach to discover macrophage-targeting drugs for improving response to PD-L1-targeted therapy for cancer. This approach identified COX1/2 inhibitors, which significantly improved the anti-tumor response of the PD-L1 blocking antibody in TNBC tumor models in a CD8⁺ T cell-dependent manner.

Introduction

Burgeoning research toward understanding mechanisms of response and resistance to immune checkpoint inhibitors (ICIs) has highlighted the impact of myeloid-dependent inflammation on clinical responses to ICIs.¹ The density of some myeloid lineage cell types correlates with poor clinical outcomes in several solid tumor types,^2,3 and preclinical data indicate that myelomonocytic cells suppress T cell recruitment,⁴ proliferation,^5,6 and/or functionality^7,8; thus, identifying targets that relieve macrophage-mediated T cell suppression is of urgent medical need. Preclinical studies with murine models of human cancer have reported synergy when tumor-bearing mice are treated with anti-programmed cell death 1 (αPD-1) or anti-programmed death-ligand 1 (αPD-L1) blocking monoclonal antibodies (mAbs) combined with inhibitors of myeloid-based pathways such as CSF1/CSF1R (colony-stimulating factor 1 and its receptor),⁹ phosphoinositide 3-kinase gamma (PI3K-γ),^10,11,12 and others (for comprehensive reviews, see Pittet et al., Cassetta and Pollard, DeNardo and Ruffell, Mantovani et al., and de Visser and Joyce^{2,3,13,14,15,16}). Based on these preclinical findings, clinical trials are now evaluating similar combinations in patients with advanced cancers.^2,15,17

Reversal of intratumoral T cell suppression is an important therapeutic strategy that could improve outcomes for patients with cancer by increasing response rates of ICIs.¹ Current approaches to identify myeloid-based vulnerabilities that relieve T cell suppression are largely dependent on deep understanding and characterization of in vivo murine models, human tumor explants, and, more recently, heterotypic organoid cultures.¹⁶ These approaches are slow and resource-intensive; alternative approaches relying on in silico¹⁸ and cell-based screening^19,20 have higher throughput but have thus far not yielded widespread reproducible in vivo synergy.

Based on our prior in vivo preclinical murine studies manipulating macrophage presence or functional status in vivo,^{7,8,21,22,23,24,25,26,27,28,29} and reported synergy between macrophage antagonists and ICIs,^9,12 we postulated that drugs targeting T cell-suppressive molecules emanating from macrophages would be synergistic with PD-1/PD-L1 axis inhibitors and increase CD8⁺ T cell activity. Based on this, we designed an ex vivo high-throughput screen (HTS) as a drug discovery platform for evaluating small molecules that block macrophage-mediated T cell suppression in a co-culture system. We screened 1,430 Food and Drug Administration (FDA)-approved therapeutic agents and identified 57 candidate compounds, seven of which exerted synergistic T cell expansion activity when combined with αPD-L1, and one with αPD-1. As proof-of-principle for the ex vivo assay approach and its power to identify synergistic drug combinations, we evaluated synergy of the top hit, with αPD-L1 antibodies in two orthotopic syngeneic mouse models of triple-negative breast cancer (TNBC), and demonstrated in vivo synergy, indicating feasibility and utility of our HTS drug discovery approach.

Results

A high-throughput screening assay monitoring macrophage-dependent T cell suppression ex vivo

Macrophages suppress various aspects of CD8⁺ T cell functionality is now well described (for a comprehensive review, see the study by DeNardo and Ruffell¹⁴). This inhibition can be modeled ex vivo by culturing bone-marrow-derived macrophages (BMDMs), differentiated under CSF1 supplemented culture conditions for 5 days, followed by co-culture with spleen-derived enriched CD4⁺ and CD8⁺ T cells (Figures 1A and S1A–S1C). After 5 days of co-culture, T cell proliferation was suppressed in the presence of BMDMs as compared to mixed co-culture of CD4⁺ and CD8⁺ T cells (Figure 1A). BMDM-mediated T cell suppression was blocked by inclusion of 1400W, an iNOS inhibitor (Figure 1B) indicating BMDM-dependent inhibition.⁶ Since in vivo, tumor-associated macrophages (TAMs) exhibit a wide range of transcriptional states, we utilized a generalized maturation approach based on CSF1, thus allowing for the theoretical presence of multiple macrophage states³⁰ in the in vivo assay.

Figure 1.

Quantification of CD8⁺ T cells by multispectral flow cytometry in an ex vivo co-culture assay

(A and B) Quantification of cell content in an ex vivo co-culture assay using either flow cytometry-based (A) or a luminescence-based surrogate cell viability where ATP released from cells allow luciferase to generate luminescence (B). The one-way ANOVA (A) or Mann-Whitney t test (B) was performed for statistical significance. Results in A and B are from two repeat experiments. Data are represented as mean ± SEM.

(C) A schematic workflow of a high-throughput drug screening strategy. Briefly, BMDMs were prepared 5 days prior to setting up the HTS assay. CD4⁺ T cells and CD8⁺ T cells from spleen were enriched a day prior to setting up the screen. The BMDM, CD4⁺ T cells, and CD8⁺ T cells were co-cultured in a 396-well assay plate, which was precoated with CD3 antibody. The co-cultures were treated with drug library after 24 h of incubation. Co-cultures were further incubated with drugs from the library for another 5 days. The HTS assay was terminated after 5-day incubation by adding the cell titer glow (CTG) assay reagent followed by data collection and analysis.

(D) Classification of the FDA-approved drug library based on the drug targets or drug’s in vivo activity and the percentage of the number of drugs represented in each class when compared to the total no. of drugs in the library.

(E and F) HTS dot plot showing an increase in luminescence signal generation from each of the 1,430 drugs when compared to the untreated culture condition in the co-culture or BMDM-only condition. Relative change in luminescence signal in the assay is plotted for each drug as compared to the untreated control.

(G) Graph showing HTS data as the percentage of each pharmacological class of drugs in the library represented in the total 57 “hits” identified in the drug screen, as well as the “hits” also further analyzed as the percentage of the total number of drugs comprising each class.

(H) The “inflammation modulators” among 57 identified drugs are further categorized by their activity in vivo.

Utilizing dynamic change in luminescence signal in the co-culture of T cells due to BMDM addition or supplementation of 1400W (Figures 1B and S1F), we next converted the co-culture assay into a high-throughput screening (HTS) approach to enable the analysis of 1,430 small-molecule drugs at seven concentrations in three replicates (Figure 1C; Tables S1 and S2). Of the 1,430 drugs that impact a diversity of cellular pathways, ∼11% were known inflammatory modulators (Figure 1D; Table S1).

In the HTS assay, to rule out any signal derived from BMDM expansion, we compared results from BMDM-T cell co-culture to BMDM-only conditions (Figures 1E, 1F, and S1G). This strategy enabled the identification of 57 compounds (out of 1,430) that reduced BMDM-T cell suppression based on increased luminescence intensity in co-culture reflecting T cell expansion (Figures 1E and S1G; Table S2). Approximately 50% of the 57 compounds identified using this strategy represented known inflammatory modulators, including several cyclooxygenase (COX)1/2 inhibitors (Figures 1G and 1H; Table 1). Based on these findings, we concluded that the co-culture and HTS approach contained sufficient specificity to identify mediators of macrophage-mediated T cell suppression.

Table 1.

List of 57 “hits” and their classification based on the biological activity

S. No.	Compound	Class	Molecular weight	CAS number	Formula
1	Diclofenac sodium	Inflammatory modulators	318.13	15307-79-6	C14H10Cl2NNaO2
2	Ibuprofen	Inflammatory modulators	206.28	15687-27-1	C13H18O2
3	Indomethacin	Inflammatory modulators	357.79	53-86-1	C19H16ClNO4
4	Meloxicam	Inflammatory modulators	351.4	71125-38-7	C14H13N3O4S2
5	Tolfenamic acid	Inflammatory modulators	261.7	13710-19-5	C14H12ClNO2
6	Lornoxicam	Inflammatory modulators	371.82	70374-39-9	C13H10ClN3O4S2
7	Flunixin meglumin	Inflammatory modulators	491.46	42461-84-7	C21H28F3N3O7
8	Ketorolac	Inflammatory modulators	376.4	74103-07-4	C15H13NO3
9	Rofecoxib	Inflammatory modulators	314.36	162011-90-7	C17H14O4S
10	Tenoxicam	Inflammatory modulators	337.37832	59804-37-4	C13H11N3O4S2
11	Aspirin	Inflammatory modulators	180.16	50-78-2	C9H8O4
12	Parecoxib	Inflammatory modulators	370.42	198470-84-7	C19H18N2O4S
13	Etoricoxib	Inflammatory modulators	358.84	202409-33-4	C18H15ClN2O2S
14	Trilostane	Inflammatory modulators	329.43	13647-35-3	C20H27NO3
15	Vitamin D2	Inflammatory modulators	396.65	50-14-6	C28H44O
16	Niflumic acid	Inflammatory modulators	282.22	4394-00-7	C13H9F3N2O2
17	Budesonide	Inflammatory modulators	430.53	51333-22-3	C25H34O6
18	Flumethasone	Inflammatory modulators	410.45	2135-17-3	C22H28F2O5
19	Fluorometholone acetate	Inflammatory modulators	418.5	3801-06-7	C24H31FO5
20	Entinostat (MS-275)	Inflammatory modulators	376.41	209783-80-2	C21H20N4O3
21	Dexamethasone (DHAP)	Inflammatory modulators	392.46	50-02-2	C22H29FO5
22	Halcinonide	Inflammatory modulators	454.96	3093-35-4	C24H32ClFO5
23	Ruxolitinib (INCB018424)	Inflammatory modulators	306.37	941678-49-5	C17H18N6
24	Tofacitinib (CP-690550, Tasocitinib)	Inflammatory modulators	312.37	477600-75-2	C16H20N6O
25	Difluprednate	Inflammatory modulators	508.55	23674-86-4	C27H34F2O7
26	Halobetasol propionate	Inflammatory modulators	484.96	66852-54-8	C25H31ClF2O5
27	Azatadine dimaleate	Inflammatory modulators	522.55	3978-86-7	C28H30N2O8
28	Fingolimod (FTY720) HCl	Inflammatory modulators	343.9	162359-56-0	C19H34ClNO2
29	Tianeptine sodium	Neuromodulators	458.93	30123-17-2	C21H24ClN2NaO4S
30	Vecuronium bromide	Neuromodulators	637.73	50700-72-6	C34H57BrN2O4
31	Diphemanil methylsulfate	Neuromodulators	389.51	62-97-5	C21H27NO4S
32	Donepezil HCl	Neuromodulators	416	120011-70-3	C24H30ClNO3
33	Dexmedetomidine	Neuromodulators	200.28	113775-47-6	C13H16N2
34	Terazosin HCl	Neuromodulators	423.89	63074-08-8	C19H26ClN5O4
35	Otenabant (CP-945598) HCl	Neuromodulators	546.88	686347-12-6	C25H26Cl3N7O
36	Trifluoperazine 2HCl	Neuromodulators	480.42	440-17-5	C21H26Cl2F3N3S
37	Reserpine	Neuromodulators	608.68	50-55-5	C33H40N2O9
38	Emricasan	Tumor signaling modulators	569.5	254750-02-2	C26H27F4N3O7
39	Afatinib (BIBW2992)	Tumor signaling modulators	485.94	439081-18-2	C24H25ClFN5O3
40	Dacomitinib (PF299804, PF299)	Tumor signaling modulators	469.94	1110813-31-4	C24H25ClFN5O2
41	Selumetinib (AZD6244)	Tumor signaling modulators	457.68	606143-52-6	C17H15BrClFN4O3
42	LY2228820	Tumor signaling modulators	612.74	862507-23-1	C26H37FN6O6S2
43	Pictilisib (GDC-0941)	Tumor signaling modulators	513.64	957054-30-7	C23H27N7O3S2
44	Nifuroxazide	Tumor signaling modulators	275.22	965-52-6	C12H9N3O5
45	Napabucasin	Tumor signaling modulators	240.21	83280-65-3	C14H8O4
46	Evacetrapib (LY2484595)	Metabolism modulators	638.65	1186486-62-3	C31H36F6N6O2
47	Fluvastatin sodium	Metabolism modulators	433.45	93957-55-2	C24H25FNNaO4
48	Lovastatin	Metabolism modulators	404.54	75330-75-5	C24H36O5
49	Azithromycin dihydrate	Anti-microbial	785.02	117772-70-0	C38H76N2O14
50	Furaltadone HCl	Anti-microbial	360.75	3759-92-0	C13H17ClN4O6
51	Rifapentine	Anti-microbial	877.03	61379-65-5	C47H64N4O12
52	Zidovudine	Genome modulators	267.24	30516-87-1	C10H13N5O4
53	Amlodipine	Ion transporter	408.88	88150-42-9	C20H25ClN2O5
54	Macitentan	Other	588.27	441798-33-0	C19H20Br2N6O4S
55	Nelfinavir mesylate	Other	663.89	159989-65-8	C33H49N3O7S2
56	Dimesna	Other	326.34	16208-51-8	C4H8Na2O6S4
57	Sildenafil mesylate	Other	474.58 (free base)	1308285-21-3	C23H34N6O7S2

Identification of compounds synergizing with PD-L1 blockade

To identify compounds within the initial 57 that might synergize with PD-1/PD-L1 blockade and further expand T cells in culture, we modified the HTS co-culture assay to include either αPD-L1 or αPD-1 blocking mAbs (Figure 2A) and evaluated synergistic activity reflected by increased signal output in co-culture, but not in the BMDM-only culture (Figures S2A and S2B). An increasing signal was observed in CD4⁺ CD8⁺ T cell-only culture by a few drug treatments (Figure S2C; Table S4). Since we aimed to identify the drugs that reversed the macrophage-dependent T cell suppression, we utilized the macrophage and T cell co-culture screen data for hit identification. Out of the 57 compounds, eight evidenced an increased signal ≥1.3-fold by this approach (Figure 2A), six were synergistic with αPD-L1, and one with αPD-1 (Figures 2B and 2C; Table S3). Of these six compounds showing synergy with αPD-L1, six were known inflammatory modulators, of which four were COX1/2 inhibitors, and two were myeloid cell signaling inhibitors. The signal from the top hit, indomethacin (a COX1/2 inhibitor), was enhanced ∼3.5-fold when in the presence of αPD-L1 (combination index [CI] < 1, synergizing) (Figures 2B, 2D, and S2E). The only drug with a marginal effect on αPD-1 was emricasan, a pan-caspase inhibitor (CI < 1, synergizing) (Figures 2B, 2D, and S2L). Together, these results indicate that using the HTS approach with BMDM-T cell co-culture, drugs targeting myelomonocytic cells can be identified that possess synergistic activity with ICIs. Based on these observations and the prominent effect seen in combination with αPD-L1, we proceeded further with in vivo studies to validate the approach using the top hit indomethacin.

Figure 2.

Synergistic screen with αPD-L1 identifies COX1/2 inhibitors as combination therapy partners

(A) Schematic workflow of HTS for identifying synergistic “hits” with PD1/PD-L1 blocking inhibitors. Briefly, BMDMs were prepared 5 days prior to setting up the HTS assay. The CD4⁺ T cells and CD8⁺ T cells from spleen were enriched a day prior to setting up the screen. The BMDM, CD4⁺ T cells, and CD8⁺ T cells were co-cultured in a 396-well assay plate, which was precoated with CD3 antibody. The co-cultures were treated with drug library and anti PD-1 or anti PD-L1 after 24 h incubation. The co-cultures were further incubated with drugs and PD1/PD-L1 blocking inhibitors for another 5 days. The HTS assay was terminated after 5-day incubation by adding the CTG assay reagent followed by data collection and analysis.

(B and C) HTS dot plot showing an increase in luminescence signal generation in the co-cultures due to combination of the ICI and one of the “hits” (ICI+hit) when compared to the “hit”-only treated co-culture where ICI being either the αPD-1 or αPD-L1.

(D) Histogram showing the change in luminescence signal in the co-culture by identified eight “hits exhibiting synergy with ICI where data are represented as increase in luminescence signal due to treatment of the co-culture by the “hit” or with “hit”+ICI as compared to the untreated co-culture. Data are represented as mean ± SD.

Synergistic in vivo anti-tumor activity by indomethacin and αPD-L1 combination

Based on the synergy identified between indomethacin and αPD-L1 in the ex vivo HTS co-culture assay, we hypothesized that in vivo, combing these two drugs would yield a significant reduction in primary tumor growth kinetics, mediated by a CD8⁺ T cell-dependent mechanism. To address this, we employed two syngeneic, orthotopic TNBC models, e.g., 4T1 and EMT6, where tumor development is associated with the recruitment of diverse leukocyte lineages into the tumor microenvironment (TME) (Figures 3A, S3A, and S3B), containing significant myeloid cell infiltration (∼50% of all CD45⁺ leukocytes) with a high abundance of GR-1⁻ and Ly6G⁻-expressing granulocytic cells, as well as F4/80⁺ macrophages (Figure 3A). In both tumor models, F4/80⁺ macrophages expressed the highest levels of COX2 (Figure 3B); PD-L1 expression was significant on several myeloid subsets with highest levels on granulocytes and lower levels on T and natural killer (NK) subsets (Figures 3B, S4A, and S4B). On the other hand, PD-1 expression was highest on both CD4⁺ and CD8⁺ T cells and detectable on NK cells, as well as Ly6G-expressing granulocytes and TAMs (EMT6 model only) (Figure 3B). Tumor growth kinetics were variably impacted by αPD-L1 monotherapy, most significantly in EMT6 but without impact in 4T1 (Figures 3C and 3D). Indomethacin, on the other hand, exhibited significant single-agent efficacy in both 4T1 and EMT6 (Figures 3C and 3D). As hypothesized based on HTS results, the combination of indomethacin plus αPD-L1 exhibited significant reductions in tumor growth kinetics as compared to single agents (Figures 3C, 3D and S4A and S4B). Moreover, in both tumor models, combination therapy increased overall survival (OS) as compared to single agents (Figures 3E and 3F). To further validate our findings, we tested ibuprofen, another COX1/2 inhibitor identified in the HTS screen using just the EMT6 tumor mode with ibuprofen administered twice a day. Ibuprofen demonstrated a similar therapeutic benefit as indomethacin in reducing tumor growth kinetics, an effect that was further enhanced when combined with αPD-L1 (Figures S5A–S5C).

Figure 3.

In vivo translation of HTS results

(A) Immune complexity analysis of orthotopic mammary tumor models EMT6 and 4T1 analyzed by multispectral flow cytometry and major leukocyte lineages quantitated as the percentage of total CD45⁺ leukocytes infiltrating tumors.

(B) Qualitative analysis of target expression of COX2, PD-L1, and PD-1 in TME-derived leukocyte lineages.

(C and D) Growth kinetics of EMT6 (C) and 4T1 (D) tumors receiving αPD-L1 intraperitoneally, COX1/2i orally, or their combination. Data shown reflect two independent experiments conducted for each tumor model.

(E and F) Impact of drug combinations on overall survival was calculated using Kaplan-Meier method for overall survival in EMT6 (E) and 4T1 (F) tumor-bearing animals. Two-way ANOVA (C, D) or Wilcoxon rank-sum test (E, F) was used for testing statistical significance of the differences between experimental groups in each cohort. Data are represented as mean ± SD.

Indomethacin and αPD-L1 combination therapy increased CD8⁺ T cell functionality in tumors

We next investigated the mechanism of indomethacin and αPD-L1 synergy in vivo by evaluating whether blockade of COX2 and PD-L1 increased CD8⁺ T cell presence or functionality as predicted from the HTS. We first analyzed complexity of total leukocytes in both 4T1 and EMT6 tumors treated with either single or combination therapy, using multiplexed spectral flow cytometry (Figures S3A, S3B, and S6A–S6C). While neither indomethacin monotherapy nor the combination with αPD-L1 increased total CD8⁺ T cell frequency in 4T1 or EMT6 tumors, proportions of both the granzyme B (GzmB)-expressing and T cell factor 1 (TCF1)-expressing CD8⁺ T cells were significantly increased by combination therapy in both tumor models (Figures 4A and 4C). TCF1 is a transcription factor identified as a marker of stem-like memory T cells (T_scm).³¹ NK cells were also significantly increased in EMT6 tumors by indomethacin but not further increased when combined with αPD-L1 (Figures S6F and S6G). Prostaglandin receptors are expressed on NK cells, and prostaglandin E2 (PGE2) can attenuate NK cell-dependent immunity in pre-clinical tumor models.³²

Figure 4.

Combination blockade of αPD-L1 and COX1/2 in vivo requires CD8⁺ T cells

(A and C) Multispectral flow cytometry analysis of CD8⁺ T cell abundance as the percentage of total CD45⁺ leukocytes or CD8⁺ T cell phenotype marked by GzmB or TCF1 expression and quantified as the percentage of the total CD8⁺ T cells in EMT6 (A) or 4T1 (C).

(B and D) CD86 expression (geometric mean fluorescence intensity, gMFI) determined by multispectral flow cytometry analysis in F4/80⁺ macrophages infiltrating EMT6 (B) and 4T1 (D) tumors.

(E) MHCII expression (mean fluorescence intensity, gMFI) in F4/80⁺ macrophages infiltrating EMT6 (left) and 4T1 (right) tumors.

(F) Change in the presence of MHCII^HI F4/80⁺ TAMs as a function of therapy (as indicated), revealed by multispectral flow cytometry in EMT6 (left) and 4T1 (right) tumors. One-way ANOVA was used for calculating the statistical significance of the differences in A–D and F.

(G–I) Growth kinetics of EMT6 tumors receiving αPD-L1 (intraperitoneally) and COX1/2i (orally) as a combination therapy with or without αCSF1 (G), αCD4 (H), or αCD8 (I) (intraperitoneally). Two-way ANOVA was used for calculating the statistical significance of the differences in G–I. Data are represented as mean ± SD.

With regards to myeloid lineages, no significant change in the frequency of overall TAMs or Gr1⁺ immature myeloid cells (iMCs) was observed with therapies; however, indomethacin mono- and combination therapy in EMT6 tumors led to a significant increased presence of Ly6⁺ monocytes (Figures S6D and S6E). Instead, examination of biomarkers indicative of cell state revealed increased expression of the co-stimulatory molecule CD86 on F4/80⁺ macrophages (Figures 4B and 4D), as well as increased expression of major histocompatibility complex class II (MHCII) on TAMs correlating with increased frequency of mature MHCII⁺ macrophages (Figures 4E and 4F) by both mono- and combination therapy.

To determine if therapeutic efficacies were CD8⁺ T cell dependent, we depleted CD8⁺ T cells in 4T1 and EMT6 tumors prior to combination therapy (Figure 4G). Whereas CD8⁺ T cell depletion did not impact tumor growth kinetics in untreated mice, the benefits of combination therapy were subverted in both EMT6 and 4T1 tumors (Figures 4G, S7A, S7B, and S7G–S7J).

We have previously reported that agents targeting the CSF1/CSF1R pathway in myelomonocytic cells reduce TAM presence in mammary carcinomas but have no impact on mammary tumor growth kinetics as monotherapy; however, when given in combination with a cytotoxic therapy, either chemo- or radiotherapy, they significantly reduce primary tumor growth kinetics and metastases by CD8⁺ T cell-dependent mechanisms.^7,8,22 Moreover, T_H2-CD4⁺ T cells directly regulate the immunosuppressive and pro-metastatic activities of TAMs by their abundant expression of interleukin-4 (IL-4)^23,33; depletion of CD4⁺ T cells or neutralization of IL-4 blocks the synergy of αCSF1 mAb plus chemotherapy combinations, whereas mere depletion of CD4⁺ T cells does not impact primary tumor growth as we have previously reported.^23,33 Based on these data, and since our in vivo analyses did not include cytotoxic agents, we anticipated that blockade of the CSF1/CSF1R pathway or depletion of CD4⁺ T cells would fail to impact combined COX1/2 inhibition/αPD-L1 responses. To no great surprise, and as anticipated, neither blockade of the CSF1/CSF1-R pathway and reducing TAM presence nor CD4⁺ T cell depletion impacted the efficacy of combination therapy on tumor growth kinetics (Figures 4G, 4H, and S7A–S7F).

Discussion

With the goal of identifying rational therapeutic approaches to improve outcomes for patients whose tumors are ICI-non-responsive and/or myeloid dense, we developed an ex vivo HTS assay leveraging three critical criteria: (1) direct ability to screen drugs in an ex vivo co-culture setting with either mono- or combinations of drugs to directly evaluate the impact on macrophage-dependent T cell suppression; (2) potential to identify macrophage regulatory mechanisms suppressing T cell functionality; and (3) high fidelity of the signal-to-noise ratio in the HTS assay associated with T cell rather than BMDM expansion. Our HTS screen, which identifies compounds blocking macrophage-mediated T cell suppression ex vivo, permits the identification of compounds that are synergistic with ICI in vivo, which rapidly accelerates the discovery process while also increasing the probability of success in vivo of combination therapy with PD-1/PD-L1 inhibitors. These attributes enabled narrowing of candidates for combination therapy with αPD-L1, namely indomethacin, the top hit identified from the combination screen.

Indomethacin is a non-steroidal anti-inflammatory drug (NSAID) that inhibits COX1 and COX2 enzymes involved in lipid metabolism and thereby limits the release of paracrine mediators (prostaglandins) in extracellular tissue microenvironments. Similar to TAMs, COX1/2 inhibitors are involved in tumorigenesis and cancer progression where significant literature has predicted their use as preventive strategies.^27,34 More recently, the expression of prostaglandin receptors has been documented on PD-1⁺ dysfunctional CD8⁺ T cells in preclinical models of viral immunity as well as tumor development.³⁵ In these studies, PGE2 reduced the presence of CD8⁺ T cells, TCF1-expressing T_SCM expansion and activation, and blunted cytotoxic immune response.³⁶ Our in vivo results herein demonstrate that the combined targeting of COX1/2 and PD-L1 increases the frequency of TCF1⁺ T_SCM coincident with a durable anti-tumor response (Figures 4A and 4C). Interestingly, when tested in ad hoc analyses of ICI-treated patients with cancer, those who used NSAIDS reported significantly improved ICI response.³⁷ Our study not only identified COX2 biology as an important regulator of PD-L1 therapy response but also identified key COX2-targeting drugs that can be used for this purpose.

Our proof-of-concept in vivo tumor studies revealed that a highly inflamed tumor microenvironment harbors COX2-expressing myelomonocytic cells (Figure 3B). Indeed, COX2 expression is tightly regulated by the activation of interferon and Toll-like receptor signaling in macrophages, and its activity is further dependent upon nitric oxide production: nitric oxide synthase is another key enzyme associated with classical activation of macrophages.^38,39 Moreover, COX2 expression in solid tumors is associated with PD-L1 expression, which together correlates with OS. Stimuli known to increase PD-L1 expression in myelomonocytic cells also increase COX2 expression in vitro.^40,41 Interestingly, in the Keynote-355 and Impassion130 trials, the best response was reported in patients with a high percentage of PD-L1 expression in the TME.^42,43 Therefore, in the wake of our results from the present study and previously published analyses,⁴⁴ we predict that patients with macrophage-dense solid tumors, and patients with TNBC in particular, who exhibit high PD-L1 and COX2 expression, will benefit by the use of combinations of COX2 inhibitors, perhaps indomethacin, and PD-1/PD-L1-inhibiting ICIs. Our data indicate that COX1/2i can increase cytotoxic T cell response by inducing immunostimulatory changes in myelomonocytic cells. However, the impact of combination therapy on other cells in the TME cannot be ruled out. Altogether, this study demonstrates an approach where rationalized screening of drugs targeting ICI resistance pathways can identify combination approaches and advance translation to the clinic.

Limitations of the study

This study utilized BMDMs and spleen-derived T cells in the macrophage-T cell co-culture assay. It did not compare these results to TAMs and tumor-infiltrating T cells isolated from frank tumors. We tested “hits” and the combination therapy using anti-PD-L1 in two TNBC tumor models and did not evaluate other tumor types where macrophages are known mediators of either disease progression or T cell suppression—these should be evaluated to expand on implications from the results to identify other indications for potential efficacy.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
BUV395 Rat Anti-Mouse CD45	BD biosciences	Cat#564279; RRID: AB_2651134
BUV563 Rat Anti-Mouse CD44	BD biosciences	Cat#741227; RRID: AB_2870781
BUV661 Rat Anti-Mouse CD19	BD biosciences	Cat#612971; RRID: AB_2870243
BUV737 Rat Anti-Mouse CD3 Molecular Complex	BD biosciences	Cat#612803; RRID: AB_2870130
BUV805 Rat Anti-Mouse CD8a	BD biosciences	Cat#612898; RRID: AB_2870186
FOXP3 Monoclonal Antibody (FJK-16s), eFluor™ 506	Invitrogen	Cat#69577382; RRID: AB_2637367
MHC Class II (I-A/I-E) Monoclonal Antibody (M5/114.15.2), eFluor™ 450	Invitrogen	Cat#48532182; RRID: AB_1272204
BV480 Mouse Anti-Ki-67	BD biosciences	Cat#566109; RRID: AB_2739511
Brilliant Violet 510™ anti-mouse/human CD11b Antibody	BioLegend	Cat#101245; RRID: AB_2561390
Brilliant Violet 570™ anti-mouse Ly-6C Antibody	BioLegend	Cat#128029; RRID: AB_10896061
Brilliant Violet 605™ anti-mouse CD11c Antibody	BioLegend	Cat#117333; RRID: AB_11204262
Brilliant Violet 650™ anti-mouse CD86 Antibody	BioLegend	Cat#105035; RRID: AB_11126147
Brilliant Violet 711™ anti-mouse CD274 (B7-H1, PD-L1) Antibody	BioLegend	Cat#124319; RRID: AB_2563619
BV750 Rat Anti-Mouse CD103	BD biosciences	Cat#747478; RRID: AB_2872154
Brilliant Violet 785™ anti-mouse F4/80 Antibody	BioLegend	Cat#123141; RRID: AB_2563667
PerCP/Cyanine5.5 anti-mouse CD80 Antibody	BioLegend	Cat#104722; RRID: AB_2291392
EOMES Monoclonal Antibody (Dan11mag), PerCP-eFluor™ 710	Invitrogen	Cat#46487582; RRID: AB_10597455
TOX Antibody, anti-human/mouse, PE	Miltenyi Biotec	Cat#130120716; RRID: AB_2801780
PE/Dazzle™ 594 anti-mouse CD335 (NKp46) Antibody	BioLegend	Cat#137630; RRID: AB_2616666
Granzyme B Monoclonal Antibody (NGZB), PE-Cyanine5.5	Invitrogen	Cat#35889882; RRID: AB_2848329
PE/Cyanine7 anti-mouse CD39 Antibody	BioLegend	Cat#143806; RRID: AB_2563394
Purified Rat Anti-Mouse CD16/CD32 (Mouse BD Fc Block™)	BD biosciences	Cat#553142; RRID: AB_394657
CD279 (PD-1) Monoclonal Antibody (J43), APC	Invitrogen	Cat#17998582; RRID: AB_11149358
CD197 (CCR7) Monoclonal Antibody (4B12), Alexa Fluor™ 700	Invitrogen	Cat#56197182; RRID: AB_657687
APC/Cyanine7 anti-mouse Ly-6G Antibody	BioLegend	Cat#127624; RRID: AB_10640819
APC/Fire™ 810 anti-mouse Ly-6G/Ly-6C (Gr-1) Antibody	BioLegend	Cat#108470; RRID: N/A
COX2 Monoclonal Antibody	Invitrogen	Cat#358200; RRID: AB_2533224
TCF1/TCF7 (C63D9) Rabbit mAb (Alexa Fluor® 647 Conjugate)	Cell Signaling Technology	Cat#6709; RRID: AB_2797631
InVivoMAb Anti-mouse PDL1 (B7-H1)	BioXCell	Cat#BE0101; RRID: AB_10949073
InVivoMAb rat IgGb isotype control	BioXCell	Cat#BE0090; RRID: AB_1107780
InVivoPlus anti-mouse CD8α	BioXCell	Cat#BP0061; RRID: N/A
InVivoPlus anti-mouse CD4	BioXCell	Cat# BE0003-1; RRID: AB_1107636
InVivoPlus anti-mouse CSF1	BioXCell	Cat# BE0204; RRID: AB_10950309
Ultra-LEAF™ Purified anti-mouse CD3ε Antibody	BioLegend	Cat#100359; RRID: AB_2616673
Ultra-LEAF™ Purified anti-mouse CD28 Antibody	BioLegend	Cat#102115; RRID: AB_11150408
Chemicals, peptides and recombinant proteins
Recombinant Murine M-CSF	PeproTech	Cat#315-02
Indomethacin	Thermo Scientific Chemicals	Cat#AAJ6325506
Ibuprofen	Sigma-Aldrich	Cat# I7905
1400W 2HCL	Selleckchem	Cat#S8337
Captisol	CYDEX pharmaceuticals	Cat#RC-0C7-100
RPMI-1640	Corning	Cat#10-040-CM
DMEM High Glucose	Corning	Cat#10-017-CM
FBS	Corning	Cat#35-015-CV
Antibiotic Antimycotic solution	Corning	Cat#30-004-CI
Fixation/Permeabilization Diluent	Invitrogen	Cat#00522356
Fixation/Permeabilization Concentrate	Invitrogen	Cat#00512343
Perm/Wash Buffer	BD biosciences	Cat#554723
Collagenase A	Roche	Cat#11088793001
LIVE/DEAD™ Fixable Blue Dead Cell Stain Kit, for UV excitation	Invitrogen	Cat#L34962
384 well plate	Greiner Bio-One	Cat#781080
96 well plate	Greiner Bio-One	Cat#650185
15 cm cell culture plates	Fisher Scientific	Cat#FB012925
10 cm cell culture plates	Fisher Scientific	Cat#FB012924
Critical commercial assays
Cell titer Glo cell viability assay	Promega	Cat#G7570
Cell titer Glo 3D cell viability assay	Promega	Cat#G9683
EasySep Mouse CD8+ T cell isolation kit	STEMCELL Technologies	Cat#19853A
EasySep Mouse CD4+ T cell isolation kit	STEMCELL Technologies	Cat#19852A
FDA approved compound library	Enzo Life Sciences Inc.	Cat#BML-2843
Experimental models: Cell lines
4T1	ATCC	Cat#CRL-2539: RRID:CVCL_0125
EMT6	ATCC	Cat#CRL-2755; RRID:CVCL_1923
Experimental models: Organisms/strains
C57BL/6J	Jackson Laboratory	Strain#000664; RRID:IMSR_JAX:000664
Balb/cJ	Jackson Laboratory	Strain#000651; RRID:IMSR_JAX:000651
Software and algorithms
GraphPad Prism (version 9.2)	GraphPad	RRID:SCR_002798
FlowJo	FlowJo	RRID:SCR_008520
Biorender	Biorender	RRID:SCR_018361

Resource availability

Lead contact

Further information and any requests for resources and reagents should be directed to the lead contact, Sanjay V. Malhotra (malhotsa@ohsu.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •Data: The data reported in this paper will be shared by the lead contact upon request.
• •Code: This study did not result in any development of original code.
• •Any additional information required to reanalyze the data reported in this work paper is available from the lead contact upon request.

Experimental model and study participant details

Cell lines

4T1 (Cat. No. # CRL-2539) and EMT-6 (Cat. No. # CRL-2755) were purchased from ATCC, USA. 4T1 cells were cultured in RPMI-1640 growth medium supplemented with 10% FBS and 1% PSA at 5% CO₂ and 37°C. EMT-6 cells were cultured in Waymouth’s MB 752/1 medium supplemented with 2.0 mM L-glutamine, 15% FBS and 1% PSA at 5% CO₂ and 37°C.

Primary cell cultures

Bone marrow-derived macrophages (BMDMs) were prepared 5 days prior to establishing a macrophage-T cell co-culture assay, as previously described.⁴⁵ Briefly, bone marrow cells were isolated from the tibia and femur of hind legs of C57BL6 mice using a syringe. Cells were then cultured in DMEM (high glucose) with 10% FBS and 10 ng/mL CSF-1 for 5 days. Adherent cells were then used as BMDMs (Figure S1A). Splenic CD4⁺ or CD8⁺ T cells were isolated from C57BL6 mice by negative selection using antibody cocktails provided in a CD8⁺ T cell or CD4⁺ T cell selection kit (Stem cell inc.) (Figure S1A).

Mouse models

C57BL/6J (Strain#000664) Balb/cJ (Strain#000651) (7–8 weeks old) mice were purchased from the Jackson Laboratory and housed under standard laboratory conditions. Animal studies were approved by OHSU’s Institutional Animal Care and Use Committee (eIACUC number: IP00003247). 4T1 (1 x 10⁵) and EMT6 (1 x 10⁵) cells were suspended in 1x PBS and matrigel (1:1 ratio) and implanted in mammary fat pad in each mouse.

Method details

Ex vivo macrophage T cell co-culture assay

Either a 96 U-bottom ultra-low attachment microwell plate or a 384 microwell plate were used for co-culture assays, where plates were pre-coated with αCD3 antibody at a 5 μg/mL concentration at 4°C for at least 16 h prior to cell addition. CD4⁺ T cells, CD8⁺ T cells, and BMDMs were then co-cultured in a 1:1:1 ratio in these pre-coated microwell plates in T cell assay medium (RPMI-1640 with 10% FBS and β-mercaptaethanol) at 37°C for another 5 days followed by flow cytometry-based analyses or luminescence-based viability analysis using Cell-titre Glo (Promega). We used 5000 cells for the 384-well plate assay and 50000 for the 96-well plate assay. The iNOS inhibitor, 1400W, was used as a positive control for relieving macrophage-mediated T cell suppression in the HTS assay at 50 μM, as previously reported.⁶

HTS assay for screening small molecule library

A surrogate T cell proliferation assay was developed utilizing the macrophage-T cell co-culture assay. Luminescence signal generated by ATP release from cellular contents of each well in a microwell plate for high-throughput screening (HTS) using 3D Cell-titre Glo (Promega); luminescence signal was then analyzed to calculate relative change from untreated co-culture wells. Three replicates were used in the HTS screen to identify “hits” defined as increased T cell proliferation, in a similar manner as the 1400w-treated conditions. A BMDM only plate was utilized to rule out any BMDM-dependent spike in luminescence signal. Data for cell viability was collected and % viability calculated compared to control. ‘Hits’ were identified using the following selection strategy: 1) no activity in coculture and only BMDM culture (rejected); 2) activity in both coculture and only BMDM culture (rejected); 3) Activity in coculture with no activity in BMDM only culture (accepted as ‘Hit’) (Figure S1F). Cells were plated on day 1 on CD3-precoated white 384 well plates using a BioTek Multiflo FX fluid dispenser and incubated overnight at 37°C. On day 2, cells were drugged with SCREEN-WELL FDA-approved drug library V2 (Enzo Life Sciences Inc. # BML-2843) using Sciclone ALH3000 liquid handler with 384-well mandrel and Z-8. The library was in quantitative high-throughput screening (qHTS) format with 8 half-log dilutions from 10 μM to 3.0 nM. Positive control (1400w at 50 μM) was included in each plate and Z score was calculated for each plate. Cumulative Z score for primary and secondary screening was −0.7331 ± 0.7703 (Mean ± SEM). During the secondary screening, cells were drugged using HP D300e digital dispenser using T8+ dispense cassettes. After five days of incubation, 10 μL Cell titer Glo was added to each well using a BioTek Multiflo FX fluid dispenser and incubated for 10 min at 37°C. Luminescence signals were measured using Twister II microplate handlers with an integrated barcode reader, and BioTek Synergy 4 microplate reader using Gen5 software. The combination index (CI) was calculated using the Bliss independence method. CI values less than 1.0, equal to 1.0, and greater than 1.0 indicate synergism, additivity, and antagonism, respectively.⁴⁶

Animal tumor models for in vivo studies

4T1 (Cat. No. # CRL-2539) and EMT-6 (Cat. No. # CRL-2755) were purchased from ATCC, USA. 4T1 cells were cultured in RPMI-1640 growth medium supplemented with 10% FBS and 1% PSA at 5% CO₂ and 37°C. EMT-6 cells were cultured in Waymouth’s MB 752/1 medium supplemented with 2.0 mM L-glutamine, 15% FBS and 1% PSA at 5% CO₂ and 37°C. Animal studies were approved by OHSU’s Institutional Animal Care and Use Committee (eIACUC number: IP00003247). To generate all in vivo data, wild-type female Balb/cJ mice were purchased from the Jackson Laboratory and maintained within the OHSU animal care barrier facility according to IACUC procedures. Two different syngeneic orthotopic tumor models were used in these studies, e.g., EMT6 and 4T1, where 1.0 x 10⁵ tumor cells were implanted in mammary fat pads and drug treatment began 7 days post tumor implantation. Indomethacin was formulated in 20% captisol and administered by oral gavage 5 days a week at a dose of 3.0 mg/kg of animal body weight. For αPD-L1 mAb studies, mice were injected i.p. with αPD-L1 (10F.9G2, BioXcell) or isotype IgG2b (LTF-2, BioXcell) control antibody at a dose of 250 μg every 4 days until endpoint. For CD8⁺ and CD4⁺ T cell depletion, mice were injected i.p. with αCD8 mAb (BP0061, BioXcell) or αCD4 mAb (BE0003-1, BioXcell), at a dose of 500 μg three days prior to treatment followed by 250 μg twice with αPD-L1 mAb treatment. Similarly, αCSF1 mAb (BE0204, BioXcell) was used for reducer macrophage presence. Tumor burden was analyzed using electronic calipers every 3^rd day post implantation by measuring longest and shortest tumor lengths. Tumor volumes were calculated using the formula tumor volume = 0.5 × L × W². Terminal tumor burden was measured as weight after euthanasia and tissue harvest. For survival studies, mice were weighed at the time of randomization into experimental groups and weighed twice a week while on therapy. Survival endpoints were defined in accordance with OHSU IACUC requirements. A survival endpoint was considered reached when any of the following conditions were met: 1) weight gain or weight loss of >15% of baseline body weight; 2) development of abdominal swelling; 3) respiratory distress defined as respiratory rate >50% of baseline; 4) diminished cage mobility. Upon reaching a surrogate survival endpoint (per IACUC guidelines), mice were sacrificed per OHSU IACUC protocols.

Tumor harvest and multiplexed spectral flow cytometry

To prepare single cell suspensions from murine orthotopic mammary tumors, tissues were minced manually and then digested at 37°C for 45 min in a solution of 2.0 mg/mL Collagenase A (Roche) and 50 units/mL DNase I (Roche). Cells were filtered through a 70 μm nylon filter and pelleted by centrifugation at 1,200 rpm × 5 min, followed by washing 1x with PBS. Nearly 10⁶ cells were then incubated on ice for 30 min in a solution consisting of a 1:10 dilution of Fc Receptor Binding Inhibitor (eBiosciences) and 1:500 Live/Dead Aqua stain (Invitrogen) diluted in PBS. Cells were then incubated with fluorescently-labeled monoclonal antibodies as previously described⁴⁷ in a solution of PBS, 5% FCS, 1.0 mM EDTA (FACS buffer). After 30 min incubation on ice, cells were washed 1 x with FACS buffer followed by intracellular staining. The stained cells were treated with permeabilization/fixation buffer (eBioscience) on ice for 10 min followed by 1x wash using permeabilization buffer (eBioscience). Cells were then incubated with fluorescently labeled monoclonal antibodies for intracellular staining in a solution of PBS, 5% FCS, 1.0 mM EDTA (FACS buffer) for 30 min on ice. Flow cytometry data was acquired on a Cytek Aurora (Cytek biosciences) spectral flow cytometer and data analyzed using FlowJo software v9.5.

Quantification and statistical analysis

Statistical analyses were performed using Prism 9. Specific tests included Mann-Whitney (unpaired, nonparametric two-tailed), unpaired t-test, one-way ANOVA, and Wilcoxon rank-sum as indicated in Figure Legends.

Published: August 23, 2024