Tretinoin improves the anti-cancer response to cyclophosphamide, in a model-selective manner

Abstract

Background: Chemotherapy is included in treatment regimens for many solid cancers, but when administered as a single agent it is rarely curative. The addition of immune checkpoint therapy to standard chemotherapy regimens has improved response rates and increased survival in some cancers. However, most patients do not respond to treatment and immune checkpoint therapy can cause severe side effects. Therefore, there is a need for alternative immunomodulatory drugs that enhance chemotherapy.

Methods: We used gene expression data from cyclophosphamide (CY) responders and non-responders to identify existing clinically approved drugs that could phenocopy a chemosensitive tumor microenvironment (TME), and tested combination treatments in multiple murine cancer models.

Results: The vitamin A derivative tretinoin was the top predicted upstream regulator of response to CY. Tretinoin pre-treatment induced an inflammatory, interferon-associated TME, with increased infiltration of CD8 + T cells, sensitizing the tumor to subsequent chemotherapy. However, while combination treatment significantly improved survival and cure rate in a CD4⁺ and CD8⁺ T cell dependent manner in AB1-HA murine mesothelioma, this effect was model-selective, and could not be replicated using other cell lines.

Conclusions: Despite the promising data in one model, the inability to validate the efficacy of combination treatment in multiple cancer models deprioritizes tretinoin/cyclophosphamide combination therapy for clinical translation.

Keywords: Combination treatment; Cyclophosphamide; Interferon; Sensitisation; Tretinoin.

Fig. 1

Tretinoin sensitises AB1-HA murine mesothelioma to cyclophosphamide chemotherapy. (A). The generation of the cyclophosphamide (CY) responder signature. Tumors from complete responders and progressors were sequenced and the differentially expressed genes upregulated in complete responders were made into a ‘CY responder gene signature’ (n = 178 genes). (B) Drugs predicted to induce a CY-sensitive tumor microenvironment, as determined by Ingenuity upstream regulatory analysis. (C) Experimental design for tretinoin and CY combination treatment. Mice were inoculated with AB1-HA s.c. and treated with tretinoin i.p. starting 3 days prior to cyclophosphamide and continued for 9 total doses at 10 mg/kg. CY was dosed at 200 mg/kg on day 12. (D) Survival of mice inoculated with AB1-HA and treated with CY, tretinoin or the combination (n = 10) (E) Heatmap of complete responders (CR) and progressors (PR) to CY treatment clustered using a ‘tretinoin induced genes’ set. (F) Gene set enrichment analysis of tretinoin associated gene sets in complete CY responders or progressors. Positive NES indicated a gene set is enriched in complete responders. Significance was determined using the Mantel-Cox log-rank test. *p < 0.05 **p < 0.01

Fig. 2

Tretinoin does not increase the sensitivity of AB1-HA to cyclophosphamide or its metabolites in vitro. (A) Viability of AB1-HA cells treated with increasing concentrations of tretinoin. (B-D). Viability of AB1-HA treated with increasing concentrations of acrolein (B), mafosfamide (C) or cyclophosphamide (D) with or without 50 µM of tretinoin. Viability was determined by the MTT assay. Data points are the mean ± SD of three independent experiments

Fig. 3

The efficacy of combination tretinoin–CY treatment is both CD8⁺ and CD4⁺ T cell dependent. (A-B) Survival of mice inoculated with AB1-HA and treated with 200 mg/kg CY with or without tretinoin. Mice were depleted of CD8 (B) or CD4 (C) T cells using 100 µg αCD4 or αCD8 i.p., 3days prior to inoculation and depletion was maintained throughout the experiment. Mice were dosed daily with 10 mg/kg tretinoin, beginning 3 days before CY for nine total doses. (C) Enrichment analysis of CD4 tretinoin-associated gene set in complete responders or progressors to CY. Positive NES indicates a gene set is enriched in CR. Significance for survival analysis was determined using the Mantel-Cox log-rank test. *p < 0.05 **p < 0.01. n = 5 for tretinoin, αCD8 + CY and αCD8 + CY + tretinoin, n = 10 for all other groups

Fig. 4

Tretinoin does not increase the efficacy of cyclophosphamide in alternative cell lines (A–E). Tumor bearing mice were treated with tretinoin 3 days before CY, for a total of nine doses at 10 mg/kg. CY was dosed at 200 mg/kg. Survival of mice inoculated with (A) CT26 colon carcinoma (n = 10, n = 5 for tretinoin group), (B) WEHI164 sarcoma (n = 5), (C) 4T1 breast cancer (n = 5), (D) AE17 murine mesothelioma (n = 5) or € Lewis lung carcinoma (n = 5, n = 10 for PBS). Significance was determined using the Mantel-Cox log-rank test. “ns” = not significant

Fig. 5

The expression of retinoic acid receptors and inflammatory genes in AB1-HA, AE17,CT26 and WEHI164 from bulk RNA sequencing data. Tumors from mice inoculated with AB1HA, AE17,CT26 or WEHI164 were harvested, and RNA was extracted and sequenced. (A–C) Expression of the three retinoic acid receptors. (D–F) Expression of the three retinoic X receptors. Counts were normalised to adjust for sequencing depth. (G) Heatmap of inflammatory genes. Counts were normalized and z scaled. (H) Inflammatory score of each tumor type, as calculated by taking the average expression of inflammation associated genes in each sample. (H) Immune score for each tumor as determined by ESTIMATE. Data are shown as mean ± SD and significance was determined using Mann-Whitney U tests corrected for multiple comparisons *p < 0.05, **p < 0.01 ***p < 0.001. n = 26 for AB1-HA, n = 8 for AE17, n = 10 for CT26 and n = 15 for WEHI164.