Type I IFNs promote cancer cell stemness by triggering the epigenetic regulator KDM1B

IFN-I administration drives enrichment and de novo induction of CSCs

To investigate the impact of the IFN-I → IFNAR axis on the appearance of cancer cells with a stem-like phenotype (hereafter referred to as CSCs), we selected a panel of cancer cell lines of distinct origin (epithelial or mesenchymal) and species (human or mouse) and treated them for 72 h with 6 × 103 U ml–1 IFN-I before assessing, by flow cytometry, the levels of prominin 1 (Prom1, best known as CD133), CD24 and CD44 surface markers, whose expression, alone and in combination, has been associated with putative CSCs. In this setting, we observed that IFN-I favors the enrichment of rare CD133+CD24+CD44+ putative CSCs (IFN–CSCs) in all analyzed murine cancer cell lines. Specifically, we identified two main populations of IFN–CSCs in MCA205 sarcoma cells: the CD133+CD24+CD44+low (CD44L, ~7 times higher compared with the untreated condition, (CTR)) and the CD133+CD24+CD44+high (CD44H, ~9 times higher compared with the CTR) CSC subsets (Fig. 1a). Putative IFN–CSCs were also detected in AT3 breast carcinoma, namely the CD133+CD44+CD24+low (CD24L, ~3.5 times higher compared with the CTR) and CD133+CD44+CD24+high (CD24H, ~2.6 times higher compared with the CTR) CSC subsets, but we focused on the former, the widely recognized CSC subpopulation in breast carcinoma19 (Fig. 1a). Similarly, we found (1) CD133+CD44+CD24+ in CT26 colon carcinoma cell line and (2) CD133+CD44+CD24+low and CD133+CD44+CD24+high in B16.F10 melanoma cell line (Extended Data Fig. 1a). These results are in line with the intra- and intertumoral heterogeneity often ascribed to CSCs20. To assess whether this phenomenon was exclusive of the murine cancer model, we treated human osteosarcoma (U2OS), breast carcinoma (MCF7, HMLER) and mammary epithelial (MCF10A) cell lines with recombinant human IFN-α2a and then analyzed the expression of standard human CSC markers. We detected IFN–CSC subpopulations in U2OS (CD133+CD44+ and CD44v6+CD24+) and MCF7 (CD44+CD24−low and CD44v6+CD24−low) but not in the nontumorigenic MCF10A and in the highly CSC-enriched HMLER (CD44+CD24−low) (Extended Data Fig. 1b).

Fig. 1: Emergence of CSCs following IFN-I treatment.figure 1

a, Multiparametric flow cytometry analysis of the illustrated CSC surface markers in MCA205 and AT3 cells treated with mock (CTR) or IFN-I (6 × 103 U ml–1, 72 h). Representative biparametric plots and histograms showing CD133+CD24+CD44+ percentages (mean ± s.e.m. with individual data point, n = 3 and n = 4 independent experiments) are shown. For more details on gating strategies, see Supplementary Fig. 1. b, Flow cytometry analyses of CD44L and CD44H percentages (top) and qRT–PCR analyses of the reported TF (bottom) in FACS-isolated CD133− and CD133+ MCA205 cells treated as in a. Mean ± s.e.m. with individual data point, n = 3 independent experiments. qRT–PCR data are reported as mean fold change (FC) ± s.e.m. over CTR after Ppia intrasample normalization, n = 3 and n = 2 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001; for exact P values, see Supplementary Table 1. c, SP (Hoechst 33342− within propidium iodide, PI−) in MCA205 and AT3 cells left untreated (black), treated with VRP (100 μM, light green), IFN-I (blue) or VRP + IFN-I (dark green). Mean ± s.e.m. with individual data point, n = 9 and n = 6 independent experiments. d, TF expression levels in IFN-I-treated MCA205 cells. Data are reported as in b, n = 3 and n = 4 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, see Supplementary Table 1 for exact P values. e, Clonogenicity of MCA205 and AT3 cells plated in soft-agar upon treatment as in a. The number (mean ± s.e.m. and individual data point) of biologically independent samples collected over three independent experiments is shown. f, Ex vivo flow cytometry of CD44L and CD44H cells within the CD45 negative (CD45−) fraction of MCA205 tumors from C57Bl/6J mice either treated with one single dose (1 × 105 U) or repeated doses (2 × 104 U) of IFN-I. Mean ± s.e.m. and individual data points for 10 mice per group from two experimental replicates. a,b,d Unpaired two-sided Student’s t-test and unpaired two-sided Student’s t-test with Welch’s correction compared with CTR. c,f, Brown–Forsythe test with Dunnet’s correction and ordinary one-way ANOVA test followed by Bonferroni’s correction. e, Unpaired two-sided Student’s t-test with Welch’s correction and two-tailed Mann–Whitney test compared with CTR.

Source data

We then isolated MCA205 CD133+ and CD133− (that is, non-CSC) cell fractions by fluorescence-activated cell sorting (FACS) and exposed them to IFN-I. By flow cytometry, we found that IFN-I treatment led to a significant increase in the CD44H and CD44L cell fraction and in the levels of the pluripotency transcription factor (TF) SRY (sex determining region Y)-box 2 (SOX2) in both the CD133+ and CD133− subsets (Fig. 1b). In parallel, by quantitative PCR with reverse transcription (qRT–PCR) analyses of common stem-related TFs and CSC markers, we found that exogenous IFN-I significantly upregulates Kruppel-like factor 4 (Klf4), POU domain, class 5, transcription factor 1 (Pou5f1, best known as Oct3/4), Sox2 and nestin (Nes) in FACS-isolated CD133− cells and Nanog homeobox (Nanog) in FACS-isolated CD133− and CD133+ cells (Fig. 1b). These results suggest that IFN-I-mediated CSC enrichment depends on the co-occurrence of positive selection of rare, pre-existing CSCs and de novo generation of CSCs.

Phenotypic and transcriptional profiles of IFN–CSCs revealed that IFN-I-treated epithelial cancer cells (AT3 and B16.F10) acquired a typical stem-like elongated morphology (Extended Data Fig. 1c). Moreover, IFN-I promoted the emergence of the side population (SP, a bona fide CSC feature) accompanied by a significant increase in cell death (Fig. 1c and Extended Data Fig. 1d). As expected, SP was significantly reduced by cotreatment with verapamil (VRP), the blocker of ATP-binding cassette transporters. Accordingly, IFN-I exposure induced significant upregulation of Klf4, Oct3/4, Sox2, Nanog, hes family bHLH transcription factor 1 (Hes1) and Nes (Fig. 1d and Extended Data Fig. 1e), and endowed MCA205 and AT3 cancer cells with increased sphere-forming ability (Fig. 1e). Moreover, when serially replated in standard CSC culture conditions, only spheres pre-exposed to IFN-I retained a CSC-related phenotypical and transcriptional profile (Extended Data Fig. 1f).

Notably, the local treatment of MCA205-derived tumors in syngeneic immunocompetent C57Bl/6J mice with one single dose of 105 U IFN-I promoted a significant accumulation of CD44H CSCs, while treatment with repeated doses of 2 × 104 U IFN-I did not enrich for CSCs (Fig. 1f). Moreover, at odds with one single 6 × 103 U ml–1 IFN-I administration (Fig. 1a), repeated treatment with lower doses IFN-I (3 × 103 U ml–1 and 103 U ml–1) did not induce CSC accumulation in MCA205 and AT3 cell lines (Extended Data Fig. 1g).

Collectively, these data demonstrate that depending on the dose and time of administration, IFN-I may favor the appearance of putative CSCs in multiple murine and human cancer cell lines.

IFN-I during immunogenic chemotherapy triggers cancer stemness

As IFN-I plays a role during ICD11, we asked whether immunogenic chemotherapy could enrich for CSCs. We took advantage of a library of prevalidated MCA205-derived clones deficient for cardinal elements of the IFN-I pathway, including: (1) Ifnar1, (2) stimulator of interferon response cGAMP interactor 1 (Sting1, best known as Sting), (3) toll-like receptor 3 (Tlr3), (4) toll-like receptor adapter molecule 1 (Ticam1, best known as Trif), (5) interferon induced with helicase C domain 1 (Ifih1, best known as Mda5) and (6) mitochondrial antiviral-signaling protein (Mavs, also known as Ips-1) (Fig. 2a)11. We exposed these clones to the ICD inducer OXP (‘donor’ dying cells), then cocultured donor dying cells with untreated clones of the same genotype (‘receiving’ viable cells) for 24 h, and, finally, analyzed receiving cells at phenotypic and transcriptional levels (Extended Data Fig. 2a). Wild-type (Wt) clones responding to OXP displayed a significant increase in the two CD44H and CD44L CSC subpopulations (ICD–CSCs, Fig. 2b). On the contrary, the vast majority of clones deficient in the IFN-I pathway presented a certain degree of impairment of ICD–CSC enrichment (Fig. 2b), indicating dependence on IFN-I signaling. This effect was not paralleled by differential cell death induction, as all clones displayed similar sensitivity to OXP (Extended Data Fig. 2b). The comparison within each genotype revealed a significant ICD–CSC enrichment in OXP-treated versus untreated conditions in all but Ifnar−/− clones, suggesting a compensation between nucleic acid-sensing pathways (Fig. 2b). Accordingly, both IFN-I and OXP treatment induced the accumulation of CSC-related transcripts in Wt clones and, to a lesser and heterogeneous extent, in Sting1−/−, Tlr3−/−, Ticam1−/−, Ifih1−/− and Mavs−/− clones, but failed to do so in Ifnar−/− clones (Fig. 2c). Moreover, the abrogation of the AIM2 and RIG-I signaling significantly reduced, but did not completely abrogate ICD–CSC enrichment in Wt and Ifih1−/− clones (Extended Data Fig. 2c and Supplementary Fig. 2). Finally, DOX-mediated ICD induction favored a complete transcriptional rewiring toward pluripotency, enhancing the expression of the entire panel of TFs analyzed, while the non-ICD drug cisplatin (CDDP), which induces very low levels of IFN-I (ref. 11), affected the expression of only few TFs (Fig. 2d).

Fig. 2: CSC promotion during immunogenic chemotherapy.figure 2

a, Major intracellular pathways upstream of IFN-I and inflammation. b, Multiparametric flow cytometry analysis of CSC surface markers in MCA205 derived clones with the indicated genotypes left untreated (CTR) or treated with OXP (300 μM, 24 h). The histograms represent the percentage (mean ± s.e.m. and individual data points, n = 3 independent experiments) of CD44H and CD44L cells. c,d, Quantification by qRT–PCR of the expression levels of the illustrated reprogramming factors in MCA205 clones left untreated or exposed to OXP (3, 30, 300 μM, 24 h) or IFN-I (6 × 103 U ml–1) (c) and in MCA205 and AT3 cells left untreated or administered with DOX (0.25, 2.5, 25 μM), OXP (3, 30, 300 μM) or CDDP (1.5, 15, 150 μM) (d). Data are reported as mean FC over untreated condition after intrasample normalization to the expression levels of Ppia, n = 2, for c, and n = 3, for d. *P < 0.05, **P < 0.01, ***P < 0.001, see Supplementary Table 1 for exact P values. e,f, MCA205 tumors grown in C57Bl/6J mice treated intratumorally as illustrated. Ex vivo flow cytometric analysis of the percentage of CD44L and CD44H cells in the CD45 negative (CD45−) fraction are reported in e, while tumor growth curves (mean tumor surface ± s.e.m.) and the percentage of tumor-free mice are shown in f. In e, data are presented as mean ± s.e.m. along with individual data points for 15 and 12 mice from two experimental replicates; the results for CSC enrichment upon one single dose of 1 × 105 U of IFN-I or repeated doses of 2 × 104 U of IFN-I of this experiment are reported in Fig. 1f. In f, data are presented as mean ± s.e.m. along with individual data points for 6 and 8 mice from two experimental replicates. b, Unpaired two-sided Student’s t-test with Welch’s correction compared with CTR cells with each clone. d,e, Ordinary one-way ANOVA test followed by Bonferroni’s correction compared with CTR cells (d) and PBS-treated and DOX-treated mice (e). f, Ordinary two-way ANOVA test and log-rank (Mantel–Cox) test.

Source data

We then exploited DOX red fluorescence, observing two distinct cell subsets (DOX+low and DOX+high) in DOX-treated MCA205 cells differing for the capability to extrude DOX and Hoechst 33342 (Extended Data Fig. 2d). Notably, following drug withdrawal, only DOX+low cells survived and resisted rechallenge with distinct ICD inducers (Extended Data Fig. 2e), indicating multidrug tolerance/resistance21. To explore the in vivo appearance of ICD–CSCs, we evaluated the effect of DOX and CDDP on syngeneic immunocompetent mice bearing MCA205 tumor grafts, analyzing tumor growth control as well as CSC markers 15 days after (the first) treatment, that is, when starting to escape growth control11. We found a twofold increase of CD44H and NANOG+ cells upon DOX, but not CDDP administration (Fig. 2e and Extended Data Fig. 2f). Also, when used as an adjunctive to DOX treatment, repeated doses of 2 × 104 U IFN-I prevented ICD–CSC accumulation, favoring tumor control and animal survival (Fig. 2f).

Altogether, these results demonstrate that IFN-I production upon ICD can promote CSC enrichment, both in vitro and in vivo, pointing to this effect as an adaptive response deployed by cancer cells to escape therapy control.

Nucleic acid transfer transduces stem signaling between cancer cells

To dissect the molecular mechanisms underlying ICD–CSC enrichment, we cocultured OXP-treated donor MCA205 cells with untreated receiving MCA205 cells alone or in combination with benzonase (BNZase), which degrades all nucleic acids, or RNase A, RNase H or DNase, which selectively degrade single-strand RNAs, double-strand RNAs or DNA. We observed differential effects in the two CD44H and CD44L ICD–CSC subsets, with BNZase preventing the enrichment of both CSC populations, while RNase A, RNase H and DNase significantly affecting only CD44L cells (Fig. 3a). Accordingly, BNZase halved the proportion of ICD–CSCs in receiving AT3 and CT26 cells (Extended Data Fig. 3a). The observation that only the depletion of all nucleic acids nullifies ICD–CSC enrichment, again suggests that this phenomenon depends on intact IFN-I signaling.

Fig. 3: Cell-to-cell horizontal transfer of nucleic acids and dedifferentiating factors during immunogenic chemotherapy.figure 3

a, Multiparametric flow cytometry analysis of CSC surface markers in receiving viable MCA205 cells upon coculturing with donor MCA205 cells left untreated or previously treated with OXP (300 µM, 24 h) alone or in combination with the indicated nucleases. Columns represent the percentage of CD44H and CD44L cells, expressed as mean ± s.e.m. and individual data points. Number of biologically independent experiments are reported. b, Fluorescence microscopy (left) or flow cytometry (right) analysis of the internalization (at 37 °C and 4 °C) of donor cell-derived, PKH26-stained EVs by receiving MCA205 cells. Scale bar, 100 μm. One representative experiment out of two is shown. c, Multiparametric flow cytometry analysis of CSC surface markers in receiving MCA205 cells cocultured with donor MCA205 cell-derived EVs in the presence of cyto D (0.5 μM). Data are expressed as mean ± s.e.m. and individual data points; number of biologically independent experiments is reported. d,e, Assessment of the expression levels of the indicated reprogramming factors by qRT–PCR in receiving MCA205 cells stimulated with donor MCA205 cell-derived EVs alone or in the presence of cyto D, as before (d) and inside EVs (e). Data are reported as mean FC ± s.e.m. over control conditions, n = 2 and n = 3, for d, n = 2, n = 3, n = 4, n = 6, n = 7, n = 9 and n = 10 for e, independent experiments, after intrasample normalization to Ppia expression levels. *P < 0.05, **P < 0.01, ***P < 0.001, see Supplementary Table 1 for exact P values. See also Extended Data Fig. 3. a,c,d, Ordinary one-way ANOVA test followed by Bonferroni’s correction. e, Unpaired two-sided Student’s t-test.

Source data

We next investigated the involvement of extracellular vesicles (EVs) in ICD–CSC enrichment. EVs isolated from donor MCA205 cells and stained with the nontoxic fluorescent membrane dye PKH26 were added to receiving MCA205 cells (Extended Data Fig. 3b). EV uptake in receiving cells, confirmed by fluorescence microscopy and flow cytometry (Fig. 3b), induced a considerable increase in CD44H and CD44L cells and in the expression of most TFs, which was impaired by cotreatment with the actin inhibitor cytochalasin D (cyto D) (Fig. 3c,d). Intriguingly, EVs from OXP-treated cancer cells carried messenger RNAs (mRNAs) for TFs (Myc, Oct3/4, Sox2, Nanog, Hes1, Nes), invasion molecules (Twist-related protein 1 (Twist1, also known as bHLHa38)), ICs (programmed cell death 1 ligand 2 (Pdcd1lg2, also known as Pdl2), lectin, galactose binding, soluble 9 (Lgals9, best known as galectin-9)) and Ifnb1 (Fig. 3e), suggesting their contribution to cancer cell dedifferentiation and aggressiveness upon ICD.

Altogether, these data indicate that ICD–CSC enrichment occurs through paracrine processes involving free and EV-mediated transfer of nucleic acids and stem-related mRNAs.

Behavioral and immunogenic features of IFN–CSCs and ICD–CSCs

We then analyzed FACS-isolated CD44H and CD44L ICD–CSCs separately, and analyzed hallmark CSC features, including chemorefractoriness, tumorigenic/metastatic potential and capability to escape immune control. We observed that CD44H and CD44L MCA205 cells exhibit a distinct sensitivity to ICD inducers, with only CD44H cells showing higher therapeutic resistance than parental (PAR) cells, both in vitro (Extended Data Fig. 4a) and in vivo, in immunocompetent mice (Fig. 4a). In vivo studies also revealed higher tumorigenicity and less immunogenicity of CD44H ICD–CSCs compared with CD44L ICD–CSCs. Although both subpopulations were able to generate tumors in immunocompromised NOD SCID γ (NSG) mice, only CD44H ICD–CSCs developed neoplasms at the lowest doses (Fig. 4b). Along with this, CD44H (but not CD44L) ICD–CSCs were able to overcome immunosurveillance, developing tumors at high incidence in immunocompetent hosts when injected at the highest number (Fig. 4b). Several findings confirmed the unique low immunogenicity of CD44H cells. First, DOX-treated PAR cells were able to vaccinate 85% of mice against PAR and CD44L ICD–CSCs, but only 30% of mice challenged with CD44H ICD–CSCs (Fig. 4c). Second, while only 15% of immunocompetent mice rejecting CD44H ICD–CSCs were vaccinated against viable PAR cells, CD44L ICD–CSCs and PAR cells conferred a higher long-term protection against this rechallenge (Extended Data Fig. 4b). Finally, when injected intravenously into immunocompetent mice, CD44H (but not CD44L) ICD–CSCs developed lung metastases (Fig. 4d). In this experiment, CD44L ICD–CSCs reacquired metastatic potential in immunocompetent mice depleted of CD4 and CD8 T cells and, even more, in immunodeficient NSG mice (Fig. 4d and Extended Data Fig. 4c), thus confirming their immune control. Of note, while a considerable fraction of CD44H ICD–CSCs divided asymmetrically (a common CSC feature), the vast majority of CD44L ICD–CSCs underwent symmetric division (Fig. 4e,f). Altogether, these results indicated that CD44H but not CD44L can be considered bona fide CSCs.

Fig. 4: Functional characterization of CSCs induced during immunogenic chemotherapy.figure 4

a, Tumor growth of PAR and CD44H MCA205 cells in C57Bl/6J mice either PBS- or DOX (2.9 mg kg–1)-treated. Growth curves show the mean tumor surface ± s.e.m. in one representative experiment out of two. Number of biologically independent mice and P values for DOX-treated CD44H versus DOX-treated PAR cells (purple) and DOX versus PBS treatments in PAR cells (black) are shown. See Supplementary Table 1 for exact P values, and Extended Data Fig. 4a. b, In vivo evaluation of the tumorigenicity of PAR, CD44H and CD44L MCA205 cells in C57Bl/6J (Wt) or NSG mice at the indicated dose. The percentage of tumor-free mice out of 12 and 15 mice per group from two experimental replicates is shown. Tumor-free mice from this experiment were rechallenged as reported in Extended Data Fig. 4b. c, In vivo evaluation of the vaccination potential of MCA205 cells. CTR or PAR MCA205 cells treated with 25 µM DOX (vaccination/VAX condition) were inoculated in the flank of C57Bl/6J mice. Seven days later animals were challenged with 1 × 105 PAR, CD44H or CD44L MCA205 in the other flank. The percentage of tumor-free mice out of six biologically independent mice per group in CTR and VAX conditions is shown. d, In vivo evaluation of the metastatic potential of parental or ICD–CSC MCA205 injected in the tail vein of C57Bl/6J mice, NSG mice or C57Bl/6J depleted of CD4 and CD8 cells. Representative macroscopic observation and quantification (mean ± s.e.m. and individual data points, n = 6 biologically independent mice per group) of the number of lung metastases 15 days post injection are reported. See also Extended Data Fig. 4c. e,f, Immunofluorescence analysis of cell divisions in FACS-isolated CD44H and CD44L MCA205 cells upon NUMB staining (e) and videomicroscopy analysis of cell divisions in FACS-isolated CD44H upon PKH26 staining (f, scale bar, 20 µm). In e, the percentage of asymmetric divisions upon image analysis quantification of the fluorescent signal in the two daughter cells is reported (n = 100, pool of three independent experiments, scale bar, 5 µm). a, Ordinary two-way repeated measures (RM) ANOVA test followed by Bonferroni’s correction. b,c, log-rank (Mantel–Cox) test. d, Ordinary one-way ANOVA test followed by Bonferroni’s correction.

Source data

We thus focused on the CD44H ICD–CSCs subset. To gain insights into their immunogenicity, we analyzed the proliferation rate of isolated CD8+ H-2Kb/ovalbumin (OVA)-specific OT-1 T cells previously primed with dendritic cells (DCs) that had taken up apoptotic OVA-expressing CD44H (CD44H-OVA) ICD–CSCs or PAR cells, and then boosted with viable cells of the same type. In line with the immune privileged nature observed in vivo (Fig. 4a–d), CD44H-OVA ICD–CSCs induced a significantly lower expansion of OT-1 CD8 T cells than PAR counterparts (Fig. 5a) and resisted CD8-mediated killing (Fig. 5b). These data prompted us to hypothesize that CD44H ICD–CSCs could escape immune control by inducing CD8 T cell exhaustion. To pursue this hypothesis, we analyzed common IC ligands, finding an increase in the percentage of cells positive to PDL1, PDCD1LG2, CEA1 and LGALS9 in CD44H cells (Fig. 5c). Consistently, CD8+ T tumor-infiltrating lymphocytes isolated from MCA205-bearing mice 15 days after intratumoral injection of DOX (when CSC enrichment occurs), but not of CDDP, displayed a significant increase in the fraction of cells expressing the LGALS9 receptor IC Hepatitis A virus cellular receptor 2 (HAVCR2, best known as TIM3) (Fig. 5d). We extended the characterization of ICD–CSCs to AT3 cells (that is, the CD24L cell subset), confirming the increase in the percentage of cells displaying PDL1, PDCD1LG2 and LGALS9 (Fig. 5c).

Fig. 5: Phenotypic and functional profiling of IFN–CSC immunogenicity.figure 5

a, Flow cytometry analysis of proliferation rate of CFSE-stained CD8+ OT-1 T cells stimulated with PAR or CD44H OVA-expressing cells. The histograms represent the FC (mean ± s.e.m. and individual data points, n = 3 independent experiments) of nonproliferating CFSE+highCD8+ cells. b, Flow cytometry analysis of CD45− OVA-expressing PAR and CD44H cell resistance to CD8+ OT-1-mediated killing. The histograms represent the FC (mean ± s.e.m. and individual data points, n = 3 independent experiments) of dying PI+CD45− cells. c, Multiparametric flow cytometry analysis of the indicated IC molecules in MCA205 or AT3 cells. Data are presented as mean ± s.e.m. and individual data points, with number of biologically independent samples collected over three independent experiments reported. d, Flow cytometry analysis of TIM3 in CD8+ tumor-infiltrating lymphocytes from MCA205-derived tumor grafts 15 days post in vivo treatment with PBS, DOX (2.9 mg kg–1), or CDDP (2.5 mg kg–1). Data are presented as mean ± s.e.m. and independent data points for 15 mice per group from three experimental replicates. e, Quantification of released chemokines in supernatants from MCA205 and AT3 cells by Luminex Multiplex Assay. One representative experiment out of two is shown. fi, Time-lapse analysis of H-2Kb splenocyte migration towards PAR and CD24L AT3 cells in microfluidic devices. Plots in (f) represent individual splenocyte trajectories towards target cancer cells (black spots) upon time-lapse recording. Quantification of interaction times between individual splenocytes and PAR or CD24L ICD–CSCs are shown in g, see also Supplementary Videos 14. Pictures of splenocytes in competition microfluidic devices (scale bar, 100 μm) and quantification of splenocytes migrated towards PAR or CD24L ICD–CSCs are shown in h and i. Data are expressed as mean ± s.e.m. and individual data points; number of biologically independent samples collected over three (f,g) and two (h,i) independent experiments is reported. See also Extended Data Fig. 4. ac, Unpaired two-sided Student’s t-test and unpaired two-sided Student’s t-test followed by Welch’s correction. d,f,g, Two-tailed Mann–Whitney test compared with PBS (d) and CTR (f,g). i, Ordinary two-way RM ANOVA test followed by Bonferroni’s correction.

Source data

To further characterize ICD–CSC immunogenicity, we measured cytokine production through Luminex Multiplex Assay, observing a unique chemokine secretion pattern in CD44H MCA205 and CD24L AT3 ICD–CSCs compared with their respective PAR cells. This encompasses reduced levels of proinflammatory chemokines CCL2 and CCL5, which mediate inflammatory monocyte trafficking and DC-T cell interactions22, and enhanced capability to secrete CXCL1 and CXCL2 (the latter in CD24L AT3 cells), which promote chemoresistance and metastasis23 (Fig. 5e). Notably, CD24L AT3 cells also showed higher levels of the regulatory T cell chemoattractant CCL22 (ref. 24) than PAR AT3 cells. Accordingly, when CD24L ICD–CSCs or PAR AT3 cells were confronted with histocompatible splenocytes in ad hoc microfluidic devices25 and then analyzed by videomicroscopy for their in vitro capability to recruit immune cells, only PAR cells were able to attract and stably interact with splenocytes at as early as 24 h (Fig. 5f,g and Supplementary Videos 14). At odds, CD24L ICD–CSCs failed to do so and, instead, migrated towards splenocytes starting a transient and unproductive interaction only upon 48 h. Finally, when we confronted PAR and CD24L AT3 cells in a microfluidic ‘competition’ device26 (Extended Data Fig. 4d), immune cells selectively migrated towards PAR cells, moving away from CSCs (Fig. 5h,i).

Altogether, these results indicate the existence of a mechanism of adaptation of cancer cells to immunogenic chemotherapy that actively contributes to intratumor heterogeneity, as the collection of induced CSC subpopulations has differential therapeutic response, aggressiveness and immunogenicity.

Global chromatin remodeling downstream of IFN-I

To dissect the mechanisms underlying cancer cell reprogramming downstream of IFN-I, we mapped the chromatin landscape of PAR (P) and CD44H (H) MCA205 cells by the assay for transposase-accessible chromatin with high-throughput sequencing (ATAC–seq) (Fig. 6a–c). By analyzing ATAC–seq peaks, we conceived a closed-to-open (C → O) and an open-to-closed (O → C) logic, and stratified genes in four groups. The CPCH and OPOH groups comprise genes with peaks permanently closed (that is, putatively repressed) or open (that is, putatively expressed) in both samples, while the CPOH and OPCH groups comprise genes whose peaks are closed in PAR cells and open in CD44H IFN–CSCs and vice versa. In particular, we focused on the CPOH group containing genes putatively more expressed in CSCs. As expected, we found genes dictating the CSC phenotype and behavior, including, but not limited to, cancer stemness (Myc and Sox) and epithelial-to-mesenchymal transition (EMT) (Gata6 and Tfcp2). We also found genes involved in immune evasion, including the negative regulator of the antigen presentation machinery Gpr17 and the inhibitor of granzyme activity Serpin (Fig. 6a). Consistently, the OPCH group contains tumor suppressor genes (Cdh, Cdk2ap1, Dlg2, Ripk3 and Fbxw2) and genes involved in antigen presentation machinery (Tap1, Tap2 and Ctsl) and inflammation (Il24, Il27, Gsdmd and Uba7) (Fig. 6a). Integration with RNA-sequencing (RNA-seq) analyses confirmed an increased expression of genes involved in tumorigenesis, tumor progression, invasiveness (Csf1r, Trpm4, Itga5, Wee1, Baiap2, Ttll7 and Spire1) and immune escape (Gpr17), coupled with repression of genes involved in tumor suppression and immune recognition (Cdh1, Il12b, Tlr5, Cdk2ap1, Il34, Il16 and Ctsl) in CD44H IFN–CSCs (Extended Data Fig. 5a).

Fig. 6: IFN-I-driven chromatin remodeling.figure 6

ad, ATAC–seq (ac) and RNA-seq (d) analysis in PAR or CD44H MCA205 cells. Heatmap illustrating global open (O) or closed (C) genes and representative gene subgroups in PAR/P and CD44H/H are shown in a, representative Kdm1b loci within CPOH group in b, TF binding motifs enriched more than twofold in PAR (black) or CD44H (purple) cells (x-axis, TF motif enrichment log FC in target/nontarget cells; y-axis, significance enrichment level) in c, and GO network analysis of upregulated (red) and downregulated (blue) genes in CD44H cells (nodes, enriched GO terms, node size, false discovery rate-adjusted enrichment P value (q value)) in d. e, Multiparametric flow cytometry analysis showing CD44H cell percentages upon OXP or OXP + TCP. Mean ± s.e.m. and individual data points. Number of biologically independent samples collected over two independent experiments is reported. f, Schematic experimental protocol of in vivo KDM1B inhibition and multiparametric flow cytometry analysis of CD44H and CD8+TIM3+ percentages in tumors from mice upon DOX or DOX + TCP treatment. Mean ± s.e.m. and individual data points for 12 and 15 mice per group from three experimental replicates. g, In vivo MCA205 tumor growth control in mice treated as illustrated. Tumor growth curves (mean tumor surface ± s.e.m. for 15 and 16 mice per group from three experimental replicates) and tumor-free mice percentages are reported. h, Ex vivo multiparametric flow cytometry analysis of CD44H percentages in PAR and Kdm1b-overexpressing (Kdm1bOVER) MCA205-derived tumors. Mean ± s.e.m. and individual data points for 12 mice per group from two experimental replicates. ik, In vivo evaluation of Kdm1bOVER and Kdm1b-depleted (Kdm1bKD) MCA205 metastatic potential (i), DOX-based therapeutic response (j) and tumorigenicity (k) in C57Bl/6J (ik) and NSG (k) mice. Mean ± s.e.m. and individual data points for 6 mice per group from two experimental replicates (i, j), and for 12 and 6 mice per group from two experimental replicates (k). See also Extended Data Figs. 5 and 6. c, One-sided binomial test. e, Ordinary one-way ANOVA test with Bonferroni’s correction. f, Kruskal–Wallis test with Dunn’s multiple comparisons. g,k,j, Ordinary two-way RM ANOVA test with Bonferroni’s correction (g) and log-rank (Mantel–Cox) test (g,k). h, Two-tailed Mann–Whitney test compared with PAR. i, Unpaired two-sided Student’s t-test with Welch’s correction.

Source data

Next, we performed TF-binding motif enrichment with the HOMER motif software, revealing considerable differences between CSCs and PAR cells for accessible motifs, indicating extensive global chromatin remodeling in CSCs (Fig. 6c and Supplementary Fig. 3a). In particular we found enrichment of motifs for various TFs of the helix–turn–helix superfamily (that is, RFX, Rfx1, Rfx2, Rfx5 and X-box), the Homeobox basic helix–loop–helix (bHLH) member Pitx1:Ebox, the Rel homology domain family member NFkB-p65 and the zinc-finger family member ZBTB in CD44H cells. Conversely, the zinc-finger motifs CTCF, BORIS and NRSF, the transcriptional enhanced associate domain (TEA, TEAD) motifs (that is, TEAD and TEAD1-4), the Rel homology domain-basic leucine-zipper superfamily member NFAT-AP1, the ETS, RUNT, the interferon-sensitive response element and the CCAAT box-binding transcription factor motifs were more accessible in PAR cells. We finally reconstructed protein–protein interaction subnetworks and biological processes specifically modulated in CD44H IFN–CSCs using the clusterProfiler and enrichPlot R packages (Fig. 6d and Supplementary Fig. 3b). Gene ontology (GO) analysis showed that most of the upregulated genes in CD44H cells (red module) have significant functional connections with stemness maintenance, tissue remodeling, immune suppression, response to stress and enhanced chromatin accessibility.

Altogether, these results provide clues about a global chromatin remodeling and a modular reorganization of specific pathways downstream of IFN-I.

Epigenetic regulation of cancer stemness by KDM1B

Among the genes specific for the CSC fraction (CD44H cells), we identified multiple ISGs, including (but not limited to) Ifi27l2a, Ifi27l2b and the epigenetic regulator Kdm1b (Fig. 6a,b). We were particularly intrigued by Kdm1b given the crucial role of chromatin remodeling in cancer evolution, cellular plasticity and immune escape5,27,28,29.

We first performed ATAC–seq studies on MCA205 cells engineered to either

留言 (0)

沒有登入
gif