Long-term platinum-based drug accumulation in cancer-associated fibroblasts promotes colorectal cancer progression and resistance to therapy

Abstract

A substantial proportion of cancer patients do not benefit from platinum-based chemotherapy (CT) due to the emergence of drug resistance. Here, we apply elemental imaging to the mapping of CT biodistribution after therapy in residual colorectal cancer and achieve a comprehensive analysis of the genetic program induced by oxaliplatin-based CT in the tumor microenvironment. We show that oxaliplatin is largely retained by cancer-associated fibroblasts (CAFs) long time after the treatment ceased. We determine that CT accumulation in CAFs intensifies TGF-beta activity, leading to the production of multiple factors enhancing cancer aggressiveness. We establish periostin as a stromal marker of chemotherapeutic activity intrinsically upregulated in consensus molecular subtype 4 (CMS4) tumors and highly expressed before and/or after treatment in patients unresponsive to therapy. Collectively, our study underscores the ability of CT-retaining CAFs to support cancer progression and resistance to treatment.

Fig. 1. Platinum accumulates within fibroblasts resilient to treatment.

a Tumor analysis in MTO-injected C57BL/6J mice treated with oxaliplatin. Left panel: percentage of tumor growth upon treatment. Right panels: percentage of CD31, CD45 and α-SMA intratumoral positivity. Control (blue; n = 5) and treated conditions (red; n = 8). Values are mean ± sd. p values are indicated. Drawing modified from Tauriello et al.. b Biological activity of oxaliplatin against HT29-M6, PDO and CCD-18Co. Values are mean ± sd. EC₅₀ are indicated. Representative of n = 3 biologically independent experiments. c 12-days follow-up of HT29-M6 and CCD-18Co cells survival upon oxaliplatin treatment. n = 3 biologically independent experiments. Values are mean ± sd. p value is indicated. d Oxaliplatin uptake in HT29-M6 (CRC) after 4 h treatment and in CCD-18Co (Fib) after 4 h treatment and 90 days after treatment. n = 6 biologically independent experiments. Mean quantities ± sd and p value are indicated. e Oxaliplatin biodistribution representative of eight independent patient tumors resected after CT. Time post CT is indicated. Left panels: platinum (Pt) uptake map. Right panels: corresponding hematoxylin and eosin staining, Scale bars: 100 μm. Drawing modified from Rivas, Linares et al.. f Mean stromal (red) and cancer (blue) Pt uptake in samples from eight CRC patients. p value is indicated. PDO patient-derived tumor organoids, C cancer, S stroma, CT chemotherapy, cps count per second. Two-sided, unpaired (a, c, d) and paired (f) t-test p values (p) are indicated. Source data are provided as a Source Data file.

Fig. 2. Platinum-stimulated fibroblasts promote CRC progression.

a Kaplan–Meier plot displays tumor initiation overtime in nude mice injected subcutaneously with 15,000 HT29-M6 cells alone (gray; n = 24), co-injected with 50,000 CCD-18Co non-treated (blue; n = 50) or pre-treated with oxaliplatin (red; n = 44). **p = 0.0015, ***p = 0.0002. Drawing modified from Calon et al.. b Quantitative analysis of oxaliplatin-treated HT29-M6 cells cultured with conditioned media (CM) from CCD-18Co non-treated or pre-treated with oxaliplatin. n = 3 biologically independent experiments. Values are mean ± sd. p value is indicated. c Kaplan–Meier curve displays DFS for GSE17536 patients (n = 177) presenting low (blue; n = 112) or high expression levels of aFib-RS (red; n = 65). HR, CI, p value are indicated. d Kaplan–Meier curve displays DFS for GSE39582 patients (n = 519) presenting low (blue; n = 332) or high expression levels of aFib-RS (red; n = 187). HR, CI, p value are indicated. e aFib-RS levels in n = 1029 CRC patients classified by CMS subtypes (CMS1 n = 175; CMS2 n = 445; CMS3 n = 147, CMS4 n = 262). Central mark indicates the median, box extends from the 25th to 75th percentiles, whiskers represent the maximum and minimum data point. p values are indicated. f Kaplan–Meier curve displays DFS for GSE39582_CMS4 patients (n = 121) presenting low (blue; n = 81) or high expression levels of aFib-RS (red; n = 40). HR, CI, p value are indicated. g GSEA of aFib-RS in the GSE72970 subset of tumor samples collected before treatment comparing patients responding to CT (n = 20) and unresponsive (n = 12) patients. h GSEA of aFib-RS in the GSE14333 subset of tumor samples collected before treatment comparing relapsing (n = 13) to non-relapsing (n = 38) patients after CT. ES enrichment score, NES normalized enrichment score, FDR false discovery rate. TFS tumor-free survival, CRC colorectal cell line (HT29-M6). Fib fibroblasts (CCD-18Co), DFS disease-free survival, HR hazard ratio, CI confidence interval, Ox oxaliplatin, GSEA gene set enrichment analysis, RLU relative luminescence unit. Two-sided, unpaired t-test p values (p) are indicated for b, e. Log-rank (Mantel–Cox test) p values (p) are indicated for a, c, d, f. GSEA nominal p value (p) and FDR-adjusted p value are indicated for g, h. Source data are provided as a Source Data file.

Fig. 3. Platinum absorption increases TGF-beta activity in fibroblasts.

a GSEA of SenRS (left panel) and SASP-S (right panel) comparing oxaliplatin treated (n = 2) and non-treated (n = 2) CCD-18Co in culture. b Relative expression levels of CDKN1A and CDKN2A in CCD-18Co treated with oxaliplatin. n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. c β-galactosidase positive CCD-18Co quantification after oxaliplatin treatment. n = 4 biologically independent experiments. Values are mean ± sd. p value is indicated. d GSEA of SASP-S comparing relapsing (n = 13) to non-relapsing (n = 38) patients after CT (GSE14333; left panel) and comparing patients responding to CT (n = 20) to unresponsive (n = 12) patients (GSE72970; right panel). e Relative expression levels of TGFB1 in paired tumor samples from ten CRC patients before and after platinum-based CT. p value is indicated. f Relative expression levels of TGFB1 in CCD-18Co, CAF1 and CAF2 treated with oxaliplatin. n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. g Correlation between aFib-RS and Fib-TBRS or TGFB1 in n = 177 CRC tumors from GSE17536. Correlation values (R) and Spearman p values are indicated. h GSEA of Fib-TBRS comparing oxaliplatin treated (n = 2) and non-treated (n = 2) fibroblasts. i GSEA of Fib-TBRS comparing relapsing (n = 13) to non-relapsing (n = 38) patients after CT (GSE14333; left panel) and comparing patients responding to CT (n = 20) to unresponsive (n = 12) patients (GSE72970; right panel) GSEA gene set enrichment analysis, ES enrichment score, NES normalized enrichment score, FDR false discovery rate, CT chemotherapy, Ox oxaliplatin, Ct control. Two-sided, unpaired (b, c, f) and paired (e) t-test p values (p) are indicated. GSEA nominal p value (p) and FDR-adjusted p value are indicated for a, d, h, i. Source data are provided as a Source Data file.

Fig. 4. TGF-beta pathway autocrine activation in platinum-stimulated fibroblasts upregulates IL11 secretion.

a Correlation between IL11 and aFib-RS in n = 177 CRC tumors from GSE17536. Correlation (R) and Spearman p value are indicated. b Relative expression levels of IL11 in CCD-18Co treated with increasing concentration of oxaliplatin. n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. c Relative expression levels of IL11 in CAF1 and CAF2 treated with oxaliplatin. n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. d Relative expression levels of IL11 in CCD-18Co 6, 12, 90 and 180 days after oxaliplatin retrieval. n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. e Relative expression levels of IL11 in CCD-18Co treated with oxaliplatin w/o TGF-beta pathway inhibitor (TBRi). n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. f Kaplan–Meier curve displays DFS for GSE39582 patients presenting low (blue; n = 136) or high (red; n = 383) expression levels of IL11. HR, CI and p value are indicated. g IL11 levels in n = 1029 CRC patients classified by CMS subtypes (CMS1 n = 175; CMS2 n = 445; CMS3 n = 147, CMS4 n = 262). Central mark indicates the median, box extends from the 25th to 75th percentiles, whiskers represent the maximum and minimum data point. p values are indicated. h Kaplan–Meier curve displays DFS for GSE39582_CMS4 patients presenting low (blue; n = 26) or high (red; n = 95) expression levels of IL11. HR, CI and p value are indicated. i GSEA of CRC-IL11RS comparing relapsing (n = 13) to non-relapsing (n = 38) patients after CT in GSE14333. j Kaplan–Meier plot displays tumor initiation overtime in nude mice injected subcutaneously with 30,000 HT29-M6 control (blue; n = 10) or IL11-secreting cells (red; n = 10). p value is indicated. k Kaplan–Meier plot displays tumor initiation overtime in nude mice injected subcutaneously with 30,000 HT29-M6 control cells (dashed red; n = 12), TGF-beta secreting cells (red; n = 12), shIL11RA cells (dashed blue; n = 8) or with shIL11RA/TGF-beta-secreting cells (blue; n = 8). p value is indicated. CRC: HT29-M6; TFS tumor-free survival, DFS disease-free survival, HR hazard ratio, CI confidence interval, Ox oxaliplatin, Ct control, CT chemotherapy. Two-sided, unpaired t-test p values (p) are indicated for b–e, g. Log-rank (Mantel–Cox test) p values (p) are indicated for f, h, j, k. GSEA nominal p value (p) and FDR-adjusted p value are indicated for i. Source data are provided as a Source Data file.

Fig. 5. POSTN is marker of platinum-induced TGF-beta activity in CAFs.

a Volcano plot displays genes upregulated and downregulated in fibroblasts treated with oxaliplatin (n = 2) compared to untreated ones (n = 2). X axis is the linear fold change and Y axis the −log10 p value of the ANOVA p values between conditions. POSTN probes are indicated. b Relative expression levels of POSTN in CCD-18Co treated with oxaliplatin w/o TGF-beta pathway inhibitor (TBRi). n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. c Relative expression levels of POSTN in CAF1 and CAF2 treated with oxaliplatin. n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. d Left panel: POSTN protein expression levels in tumors from MTO-injected mice treated with oxaliplatin (n = 8) compared to untreated control (n = 5). Values are mean ± sd. p value is indicated. Middle and right panels: representative pictures of POSTN stained tumor sections. Scale bar: 50 μm. e Correlation between POSTN and aFib-RS (left panel), Fib-TBRS (center panel) or TGFB1 (right panel) in n = 177 CRC tumors from GSE17536. Correlation values (R) and Spearman p values are indicated. f POSTN protein expression intensity in tumor samples obtained from untreated patients and cultured ex vivo with oxaliplatin in presence or absence of TGF-beta pathway inhibitor (TBRi). n = 4 biologically independent experiments. Values are mean ± sd. p values are indicated. g Representative POSTN immunostaining in explants from f. Scale bars: 100 μm. IU intensity unit, C cancer, S stroma, Ct control, Ox oxaliplatin. Two-sided, unpaired t-test p values (p) are indicated for b–d, f. Source data are provided as a Source Data file.

Fig. 6. POSTN is a stromal marker of resistance to chemotherapy.

a POSTN protein expression intensity in IHC-CRC cohort (CMS1 n = 22; CMS2 n = 31; CMS3 n = 15, CMS4 n = 41). Values are mean ± sem. p values are indicated. b Percentage of stroma in IHC-CRC cohort (CMS1 n = 22; CMS2 n = 31; CMS3 n = 15, CMS4 n = 41). Values are mean ± sem. p values are indicated. c Kaplan–Meier curve displays DFS of CRC patients in IHC-CRC cohort presenting low (blue; n = 62) or high (red; n = 45) protein expression of POSTN. HR, [CI], p value are indicated. d Kaplan–Meier curve displays DFS of CRC patients in IHC-CRC cohort presenting low (blue; n = 47) or high (red; n = 60) stromal content. HR, [CI], p value are indicated. e Kaplan–Meier curve displays DFS of CRC patients in IHC-CRC cohort presenting low (blue; n = 28) or high (red; n = 19) protein expression of POSTN in tumors with high stromal content. HR, [CI], p value are indicated. f Kaplan–Meier curve displays DFS of CRC patients in IHC-CRC cohort presenting low (blue; n = 34) or high (red; n = 26) protein expression of POSTN in tumors with low stromal content. HR, [CI], p value are indicated. g Left gray panel: protein expression levels of POSTN in all (n = 28) paired tumor samples collected before (blue) and after (red) CT. Right panel: protein expression levels of POSTN in tumor samples categorized by patients responsive (R; n = 17) and unresponsive (NR; n = 11) to therapy. Values are mean ± sem. p values are indicated. h POSTN detection by IHC in paired tumor samples from a NR patient collected before (left panel) and after CT (right panel), representative of n = 11 patients. Scale bars: 100 μm. IU intensity unit, C cancer, S stroma, CMS consensus molecular subtype, CT chemotherapy, HR hazard ratio, CI confidence interval, NR unresponsive to CT, R responsive to CT, DFS disease-free survival. Two-sided, unpaired (a, b) and paired (g) t-test p values (p) are indicated. Log-rank (Mantel–Cox test) p values (p) are indicated for c–f. Source data are provided as a Source Data file.

Fig. 7. Platinum-induced expression of POSTN isoform 4 in the tumor stroma enhances resistance to treatment.

a Relative expression levels of POSTN isoforms in CCD-18Co treated with oxaliplatin. n = 4 and n = 3 biologically independent experiments for POSTNi4 and for POSTNi1-3, 5-8 respectively. Values are mean ± sd. p value is indicated. b Relative expression levels of POSTNi4 in CAFs treated with oxaliplatin. n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. c Relative expression levels of POSTNi4 in CCD-18Co treated with oxaliplatin w/o TGF-beta pathway inhibitor (TBRi). n = 3 biologically independent experiments. Values are mean ± sd. p values are indicated. d Relative expression levels of POSTNi4 in CCD-18Co 6, 12, 90, 180 days after oxaliplatin retrieval. n = 3 biologically independent experiments. Values are mean ± sd. p value is indicated. e POSTN protein levels in genetically engineered PDOs with upregulated expression of POSTNi4 compared to control cells. Bottom panel shows β-Actin protein levels as normalization control. Representative of n = 3 biologically independent experiments. f Growth kinetics upon oxaliplatin treatment of tumors derived from subcutaneous injection into NSG mice of PDOs-Ct or POSTNi4-secreting PDOs. Left panel: PDOs-Ct injected mice tumors treated (red; n = 10) or non-treated tumors (blue; n = 9). Right panel: POSTNi4-PDOs injected mice tumors treated with oxaliplatin (red; n = 9) or non-treated tumors (blue; n = 7). Values are mean ± sem. p values are indicated. PDO patient-derived tumor organoids, Fib CCD-18Co, Ct control, Ox oxaliplatin. Two-sided, unpaired t-test p values (p) are indicated for a–d, f. Source data are provided as a Source Data file.