Abstract
Cancer cells accommodate multiple genetic and epigenetic alterations that initially activate intrinsic (cell-autonomous) and extrinsic (immune-mediated) oncosuppressive mechanisms. Only once these barriers to oncogenesis have been overcome can malignant growth proceed unrestrained. Tetraploidization can contribute to oncogenesis because hyperploid cells are genomically unstable. We report that hyperploid cancer cells become immunogenic because of a constitutive endoplasmic reticulum stress response resulting in the aberrant cell surface exposure of calreticulin. Hyperploid, calreticulin-exposing cancer cells readily proliferated in immunodeficient mice and conserved their increased DNA content. In contrast, hyperploid cells injected into immunocompetent mice generated tumors only after a delay, and such tumors exhibited reduced DNA content, endoplasmic reticulum stress, and calreticulin exposure. Our results unveil an immunosurveillance system that imposes immunoselection against hyperploidy in carcinogen- and oncogene-induced cancers.
“Oncogenic stress,” generally resulting from the activation of oncogenes or inactivation of oncosuppressor genes, can ignite an array of intrinsic programs for tumor suppression, some of which are linked to the DNA damage response (1). For instance, oncogenic stress and DNA damage can induce senescence or apoptotic cell death (1, 2), two cell-intrinsic mechanisms for the suppression of cancer cells, or stimulate the expression of ligands for NKG2D receptors (3) or CD95/FAS (4), which can increase the susceptibility of tumor cells to killing by immune effectors (3–5), elicit an immune response against tumor antigens (6, 7), or both. One prominent mechanism of tumorigenesis involves an initial event of tetraploidization. This can result from illicit cell-to-cell fusion among somatic cells (8), endomitosis (during which cells with duplicated chromosomes fail to divide) (9, 10), or endoreplication (during which cells enter two consecutive rounds of DNA replication that are not separated by mitosis) (11). Tetraploidy has been observed in the early stages of esophageal, colorectal, mammary, and cervical cancers (12). Tetraploidy may act as a stochastic generator of genomic instability because tetraploid cells can gain or lose chromosomes progressively during aberrant rounds of bipolar mitosis or undergo multipolar mitosis (10, 12–15). This oncogenic polyploidization-depolyploidization cascade is normally avoided by cell-autonomous, oncosuppressive control mechanisms (9, 12, 15–19). We report the unexpected finding that the chromosomal content is also controlled indirectly, by an immunosurveillance mechanism that ensures the elimination of hyperploid cells.
Cytotoxic anticancer agents only induce immunogenic cancer cell death if they provoke the translocation of calreticulin (CRT) to the plasma membrane surface as a consequence of endoplasmic reticulum (ER) stress (6, 20). CRT facilitates the phagocytosis of stressed and dying cells by macrophages as well as by antigen-presenting dendritic cells (21–23). Among 480 compounds that perturb signal transduction pathways, we identified 12 agents that caused a CRT–green fluorescent protein (GFP) fusion protein transduced into human osteosarcoma U2OS cells to accumulate in granules and to move to the cellular periphery (24). These agents included calyculin A, a protein phosphatase-1 inhibitor that stimulates CRT exposure (6), as well as four agents that can induce polyploidization, namely one cytokinesis inhibitor and three inhibitors of microtubule function (Fig. 1A and fig. S1, A to D). The redistribution of endogenous CRT to the surface of the plasma membrane was confirmed by three distinct methods (figs. S1, E to J, and S2). CRT exposure was induced by multiple hyperploidy-promoting conditions in immortalized nontransformed or malignant murine and human cell lines, as well as in Tp53−/− primary mouse mammary gland cells, intestinal epithelial cells, and colon crypt organoids, all of which are susceptible to pharmacologically induced or spontaneous polyploidization (9) (figs. S3 and S4, A to D). Human colon carcinoma DLD-1+7 cells, which are trisomic owing to the introduction of one additional copy of chromosome 7 (25), did not display a constitutive ER stress response or CRT exposure yet did expose CRT in response to hyperploidization with cytochalasin D (fig. S4, E and F). Hence, a major, nonphysiological increase in chromosome content stimulates CRT exposure. Similarly to mitoxantrone-induced CRT externalization (20, 26), the microtubular poison-induced exposure of CRT at the cell surface was accompanied by all hallmarks of the ER stress response, namely the PKR (protein kinase, RNA-activated)–like ER kinase (PERK)–associated phosphorylation of eukaryotic initiation factor 2 α (eIF2α), the up-regulation of X-box binding protein 1 (XBP1), and the perinuclear translocation of activating transcription factor 6 (ATF6) (Fig. 1B and fig. S5). Moreover, the depletion of PERK, caspase-8, or ERp57 (an ER chaperone that is required for CRT exposure); the knockout of Bax and Bak or Casp8; or a nonphosphorylatable mutant of eIF2α (S51A) all reduced CRT transport to the surface (fig. S6), thus revealing a CRT exposure pathway that was similarly induced by mitoxantrone and microtubular poisons. Mouse colon cancer CT26 cells succumbing in response to the microtubular inhibitor nocodazole protected immunocompetent BALB/c mice against rechallenge with live CT26 cells. The immunogenicity of CT26 cells treated with microtubule inhibitors was strongly reduced by depletion of CRT or ERp57, yet could be restored by the absorption of recombinant CRT to the cells (Fig. 1C and fig. S7).
Functional linkage of hyperploidization, CRT exposure, and immunosurveillance. (A and B) CT26 cells were treated for 48 hours with cytochalasin D (CytD), nocodazole (Noco), epothilone B (EpoB), or taxotere (TXT) and then either subjected to the cytofluorometric detection of CRT exposure (A) or analyzed by immunoblotting for eIF2α phosphorylation (B). R.U. indicates relative units. (C) BALB/c mice were injected subcutaneously with wild-type (WT) CT26 cells that had been treated or not with 100 nM Noco for 48 hours or with CT26 cells stably expressing a CRT-specific short hairpin RNA (shRNA) that had been preincubated or not with recombinant CRT (recCRT) and treated with 100 nM Noco for 48 hours. In all cases, mice were injected 1 week later with live WT CT26 cells, and tumor incidence was monitored. These experiments were done three times (15 to 20 mice per group). (D) Parental (P) and hyperploid (H) CT26 cells were stained to measure CRT exposure on the cell surface. (E) Constitutive eIF2α phosphorylation in H CT26 clones (two representative clones of at least n = 5), as determined by immunoblotting. Clonogenicity of hyperploid mouse embryonic fibroblasts (MEFs) (F) or CT26 cells (G) upon disruption of the CRT exposure pathway. The cloning efficiency of nocodazole-treated, FACS-purified WT cells or cells transfected with a scrambled control with 8n DNA content was considered as 100%. In vitro data were compared with one-tailed Student’s t tests, tumor incidence with the log rank test. *P < 0.05; **P < 0.01; n.s., nonsignificant, as compared to untreated P CT26 cells [(A), (B), (D), and (E)], untreated (F), or shCo-transfected H (G) CT26 cells, or mice challenged with phosphate-buffered saline (PBS) only (C). (H to M) Evidence for an immune response to hyperploid cancer cells. (H) P or H CT26 cell clones were injected into Rag γ or BALB/c mice. (I) Tumor growth of P or H CT26 cells treated with a shRNA (shCRT) to CRT was monitored in Rag γ or BALB/c mice. (J) Growth of P CT26 cells in naïve mice and in tumor-free BALB/c mice previously (2 months before) inoculated with H CT26 cells [as in (H)]. (K) H CT26 cells were inoculated into BALB/c mice that were depleted of CD4+ or CD8+ T lymphocytes by injection of suitable antibodies. (L and M) Contribution of IFN-γ and Ifnar1 to the immunosurveillance of hyperploid cells. H LLC (L) and MCA205 (M) cells were inoculated into wild-type [(L) and (M)], Ifng−/− (L), or Ifnar1−/− (M) C57Bl/6 mice. Tumor growth curves [(H) to (M), top graphs] were analyzed with one-tailed Student’s t test, whereas tumor incidence [(C) and (H) to (M), bottom graphs, illustrated with Kaplan-Meier curves] was compared by log rank test. Error bars indicate SEM. *P < 0.05, **P < 0.01, as compared with P cells (H), PBS-challenged mice [(J) and (K)], or immunocompetent C57Bl/6 mice (M).
The surface exposure of CRT and ERp57 was increased in CT26 clones derived from cells transiently exposed to nocodazole that contained close to twice the DNA content of parental cells (which we refer to as “hyperploid” cells), although the surface expression of most other membrane proteins was unaltered (Fig. 1D and fig. S8). This hyperploidy-associated increase in CRT exposure was also observed in mouse Lewis lung carcinoma (LLC) and fibrosarcoma MCA205 cells, as well as in human cancer cell lines (fig. S9). As compared to their parental counterparts, hyperploid clones exhibited constitutive PERK and eIF2α phosphorylation (Fig. 1E). Interruption of the CRT exposure pathway reduced the clonogenic potential of hyperploid cells (Fig. 1, F and G), suggesting a functional link between the ER stress–associated CRT exposure pathway and the fitness of hyperploid cells.
Because hyperploidization is linked to CRT exposure, we wondered whether cancer cells with increased DNA content might be subjected to immunosurveillance. We injected parental or hyperploid clones derived from CT26 [class I major histocompatibility complex (MHC) haplotype H2d], MCA205 (H2b), or LLC (H2b) cells subcutaneously into syngeneic immunocompetent wild-type (WT) or immunodeficient Rag2−/− γc−/− (Rag γ) mice. When inoculated into immunocompetent (but not immunodeficient) mice, hyperploid CT26 cells formed tumors less frequently than did their parental precursors, and such cancers grew less efficiently (Fig. 1H). Similarly, tumors arising from hyperploid LLC or MCA205 cells generally failed to grow or grew more slowly in C57Bl/6 mice, unless the immune system was compromised by the Rag γ phenotype (fig. S10, A and B). This difference in the growth of cancer cells on immunodeficient versus immunocompetent mice was lost upon depletion of CRT from hyperploid cells (Fig. 1I). Mice that did not develop tumors within 1 month after injection of hyperploid CRT-exposing cancer cells (40 to 50% of all mice) also failed to develop cancers after another inoculation with parental CT26 cells (Fig. 1J) yet allowed for the growth of unrelated cancer cells (fig. S11). The incidence and proliferation rate of tumors generated from hyperploid CT26 cells increased upon the depletion of CD4+ or CD8+ T lymphocytes from BALB/c mice (Fig. 1K) or when interferon γ (IFN-γ) or the type I interferon receptor 1 (Ifnar1) were inactivated (Fig. 1, L and M). Hyperploid cells were more efficient at priming T lymphocytes against tumor antigens than were their parental counterparts, as indicated by experiments in which tumor cells were injected into the footpad, followed by recovery of T lymphocytes from the draining lymph node, their restimulation with tumor antigens, and quantification of Ifnγ production. CRT depletion from hyperploid tumor cells reduced T lymphocyte priming in this system (fig. S10C).
Histological examination of tumors arising from hyperploid CT26 clones in either immunodeficient Rag γ or immunocompetent BALB/c mice revealed a difference in the mean nuclear diameter, which was decreased in tumors from BALB/c mice (Fig. 2A and fig. S12A). Hence, immunoselection, the passage of cancer cells in immunocompetent (as opposed to immunodeficient) mice, led to a reduction in nuclear size. Accordingly, immunoselected tumors from BALB/c hosts contained fewer chromosomes (Fig. 2B and fig. S12B) than tumors that had grown in vivo without immunoselection (i.e., recovered from Rag γ mice). Immunoselection led to a reduction of overall DNA content (Fig. 2C and fig. S12C) without preferentially affecting a particular set of chromosomes, as determined by comparative genomic hybridization and multicolor fluorescent in situ hybridization (fig. S13 and tables S1 and S2). Similarly, originally hyperploid LLC and MCA205-derived cancers exhibited reduced DNA content upon immunoselection (fig. S14, A and B). The amount of externalized CRT (ecto-CRT) decreased in hyperploid CT26, MCA205, and LLC cancer cells grown in immunocompetent mice yet remained constant or even increased (together with that of ecto-ERp57) upon growth in immunodeficient mice (perhaps because CRT exposure correlates with the fitness of hyperploid cells or because extracellular CRT exerts trophic, pro-inflammatory, and angiogenic effects) (27). The suppression of CRT exposure correlated with a reduction in ploidy, as well as with decreased eIF2α phosphorylation. Hyperploid cells recovered from immunodeficient mice (which conserved their hyperploid status and even increased CRT externalization in vivo) exhibited reduced ploidy and lower amounts of ecto-CRT after growth in immunocompetent mice (Fig. 2, D to G, and fig. S14, C to E). These results confirm the strong correlation between CRT exposure and hyperploidy.
Reduced DNA content and diminished CRT exposure in initially hyperploid cancers grown on immunocompetent mice. (A) Reduction in the nuclear diameter of immunoselected hyperploid cancer cells, as determined by hematoxylin and eosin staining of CT26 tumors growing in Rag γ and BALB/c mice. (B) Chromosome numbers of hyperploid CT26 cells immunoselected (in BALB/c mice) or grown in vivo without immunoselection (in Rag γ mice). (C) DNA loss in immunoselected hyperploids. CT26 hyperploids were immunoselected or grown without immunoselection [as in (A) and (B)], and the DNA content of isolated cancer cells was determined by cytofluorometry. The mean ploidy of the G0/G1 peak was determined for multiple tumors. (D to G) Immune response effect on CRT exposure. Hyperploid clones were either cultured in vitro (1), grown in vivo without immunoselection (in Rag γ mice) (2), or sequentially transferred into immunocompetent mice for immunoselection (3 to 5) (D), recovered and then subjected to the determination of ploidy (E), CRT exposure (F), or the phosphorylation of PERK and eIF2α, normalized to controls (G). In (G), representative results and quantitative data (obtained by densitometry on multiple tumors) are shown. One-tailed Student’s t test was used for statistical comparisons. Error bars indicate SEM. *P < 0.05, **P < 0.01, as compared to tumors untreated Rag γ mice (A), parental (P) CT26 cells cultured in vitro [(B), (C), (E), (F)], or hyperploid CT26 cells cultured in vitro (G).
We also investigated whether immunoselection may influence ploidy in spontaneous or genetically or chemically induced cancers. Indeed, carcinogen- or Myc-driven tumors exhibited larger nuclei and higher levels of eIF2α phosphorylation in immunodeficient (Rag γ, Stat1−/−, or Dnam1−/−) (28) mice than in immunocompetent hosts (Fig. 3, A to D, and fig. S15, A to D). To investigate the translational relevance of these findings, we determined the ploidy status and ER stress response on breast cancer tissues from 60 patients (table S3) with locally advanced breast cancer. In this context, the efficacy of neo-adjuvant chemotherapy is strongly influenced by anticancer immune responses (29–31). Successful chemotherapy reduced the nuclear size of the rare residual carcinoma cells from responders, correlating with an increased ratio of tumor-infiltrating CD8+ over immunosuppressive FOXP3+ cells. In contrast, tumor cells from nonresponders (who manifested discernible breast cancers in spite of six cycles of chemotherapy) exhibited increased nuclear size and increased phosphorylation of eIF2α (Fig. 3E and fig. S15E). These results correlate a local immune response with reduced nuclear size (in responders) and the absence of a local immune response with increased nuclear size (in nonresponders).
Immunoselection against hyperploidy and eIF2α phosphorylation. (A to D) Nuclear diameter and eIF2α phosphorylation in methylcholanthrene-induced fibrosarcomas (A and B), Eμ-myc–driven B cell lymphomas (C), or medroxyprogesterone acetate–induced mammary carcinomas (D) from immunodeficient (Rag γ, Stat1−/− or Dnam1−/−) versus immunocompetent mice. (E) Immunoselection in human mammary adenocarcinomas treated with neo-adjuvant chemotherapy. Tumor specimens obtained by surgery from 18 responders and 42 nonresponders before (pre) and after (post) treatment were stained to determine nuclear diameter, ratio of the number of tumor-infiltrating CD8+ to FOXP3+ lymphocytes, and eIF2α phosphorylation. One-tailed Student’s t test was used for statistical comparisons. Error bars indicate SEM. *P < 0.05, **P < 0.01, compared to tumor cells with the same nuclear diameter obtained from immunocompetent mice (A to D) or pretreatment (E). #P < 0.05, ##P < 0.01, compared to cells from immunodeficient mice with nuclear diameter < 9 μm (A), or to responders (E). †P < 0.05, compared to cells with a nuclear diameter < 10 μm from nonresponders (E). A.U., arbitrary units.
Our results underscore the relation between immunoselection, hyperploidy, and ER stress, which can induce CRT exposure. To test whether CRT might trigger immunosurveillance, we transduced diploid CT26 cells with a construct encoding CRT fused at the N terminus to the murine immunoglobulin (Ig) κ-chain leader sequence (which directs the protein to the secretory pathway) and at the C terminus to the platelet-derived growth factor receptor (PDGFR) transmembrane domain (which anchors the protein to the plasma membrane). CT26 cells stably expressing this membrane-exposed CRT [mCRT, which did not affect other surface molecules, including class I MHC (Fig. 4A)] efficiently proliferated in immunodeficient Rag γ mice yet rarely formed tumors in immunocompetent mice (Fig. 4B), and this only occurred coupled to mCRT down-regulation (Fig. 4C), resulting in increased aggressiveness upon reinjection into BALB/c mice (Fig. 4D). Similarly, the reestablishment of high ecto-CRT levels (by mCRT transfection) in immunoselected (former) hyperploid cells limited the aggressiveness of tumor cells that otherwise would have rapidly proliferated upon reinoculation into immunocompetent mice (Fig. 4E). Thus, ER stress–elicited exposure of CRT at the cell surface, as mimicked by mCRT transfection, constitutes a trait of hyperploid cells that can elicit immunosurveillance and is counterselected during tumorigenesis in immunocompetent mice.
Ecto-CRT underlies the immunogenicity of hyperploid cancer cells. (A to D) Immunosurveillance and immunoselection mechanisms activated by membrane-bound CRT on CT26 cells. Parental (P) CT26 cells were transduced with a construct encoding mCRT or vector only and subjected to immunostaining for the quantification of externalized CRT or MHC class I molecules (A). Alternatively, mCRT-expressing P cells were injected into Rag γ or BALB/c mice, followed by monitoring of tumor growth and incidence (B). Ecto-CRT was measured by immunostaining on recovered cancer cells (C). (D) Tumor growth and incidence of mCRT-expressing P cells sequentially injected into Rag γ (1) or BALB/c (2 and 3) mice. (E) Restoration of immunosurveillance on immunoselected (IS) hyperploid (H) cells by mCRT. H clones were injected into BALB/c mice for immunoselection (1). Then, tumor cells were recovered, transduced with a mCRT-encoding construct (3) or the empty vector (2), analyzed for CRT exposure, and re-injected into BALB/c mice, followed by monitoring of tumor growth and incidence. Experiments were performed three times, yielding comparable results. One-tailed Student’s t tests were used for statistical comparisons of in vitro data. Tumor growth and incidence (the latter being illustrated with Kaplan-Meier curves) were compared by one-tailed Student’s t and log rank tests, respectively. Error bars indicate SEM. *P < 0.05; **P < 0.01; n.s., nonsignificant compared with P CT26 cells [(A) to (C)], mCRT-expressing cells (D), or H cells (E) growing for the first time in BALB/c mice. ##P < 0.01, compared with mCRT-expressing CT26 cells grown in vitro (C) or P CT26 cells (E). ††P < 0.01, compared with mCRT-expressing CT26 cells grown in Rag γ mice (C).
Our results indicate the existence of an immunosurveillance mechanism for the control of ploidy that involves CD4+ and CD8+ T lymphocytes, as well as both type I and type II interferons, in line with results from other models of immunosurveillance (32–34). In our model, increases in ploidy were consistently accompanied by ER stress, resulting in the translocation of CRT to the plasma membrane surface. Conversely, immunoselection led to a coordinated reduction in ploidy, ER stress, and CRT exposure, both in established or developing cancers. Transfection-enforced expression of mCRT was sufficient to elicit immunoediting in the absence of hyperploidy. Thus, beyond oncogenic stress that elicits a DNA damage response (and a subsequent increase in the expression of NKG2D ligands and CD95/FAS) (3), hyperploidy-associated ER stress and CRT exposure may contribute to cancer immunosurveillance. Future studies must address which additional factors, such as changes in the expression and presentation of tumor antigens, determine whether a cancer cell is susceptible to or escapes from immunosurveillance.
Supplementary Materials
www.sciencemag.org/cgi/content/full/337/6102/1678/DC1
Materials and Methods
Figs. S1 to S15
Tables S1 to S3
References and Notes
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- ↵
- Acknowledgments: We are grateful to M. L. Albert (Institut Pasteur, Paris) for Ifnar1−/− and Ifng−/− mice, M. Colonna (Washington University School of Medicine, St. Louis, MO) for Cd226−/− mice (Dnam-1−/− mice), S. B. Horwitz (Albert Einstein College of Medicine, New York) for epothilone B–resistant A549 cells, R. Prywes (Columbia University, New York) for the GFP-ATF6–encoding plasmid, J. Yuan (Massachusetts Institute of Technology, Boston) for the XBP1-DBD-Venus–encoding construct, T. Reid (NIH, Bethesda, MD) for DLD-1 and DLD1+7 cell lines, and D. Metivier for assistance with fluorescence-activated cell sorting (FACS) experiments. G.K. is supported by the Ligue Nationale contre le Cancer (LNC, Equipe labelisée), Agence Nationale pour la Recherche (ANR), European Commission (Active p53, Apo-Sys, ChemoRes, ApopTrain), Fondation pour la Recherche Médicale (FRM), Institut National du Cancer (INCa), Cancéropôle Ile-de-France, Fondation Bettencourt-Schueller, and the LabEx Immuno-Oncology. F.M. is grateful to the Fonds Zur Förderung der Wissenschaftlichen Forschung (FWF, grants LIPOTOX, P23490-B12, P24381-B20, and W 1226-B18). L.S. and M.M. are supported by FRM. I.V., I.M., M.T., and S.A. are supported by LNC. M.N.-S. is supported by a postdoctoral contract of Junta de Extremadura (Spain) and G.M. by European Molecular Biology Organization. M.J.S. was supported by National Health and Medical Research Council (NH and MRC) Australia and the Victorian Cancer Agency. C.J.C. was supported by Leukemia Foundation of Australia, Monash University. N.M. was supported by Cancer Research Institute. J.M.P. and V.S. are supported by the Austrian Academy of Sciences, an Advanced European Research Council grant, and the European Union INFLA-CARE network. L.S., I.V., I.M., M.T., C.P., M.M., S.A., O.K., M.N.-S., S.S., G.M., A.C., A.B., B.J., S.L., F.G., A.S., T.Y., S.R.-V., C.L., V.P.-C., M.T., A.V., F.B., A.A., F.C., G.M.F., M.P., A.F., N.L.M., M.L., C.J.C., V.S., G.P., V.L., J.M.P., C.L.-O., M.J.S., and M.C. performed experiments. L.S., C.R., D.C.-G., F.M., L.Z., M.C., and G.K. designed the study. L.S., L.Z., M.C., and G.K. analyzed results. L.S., L.G., and G.K. assembled the figures and wrote the paper. The authors declare no conflicts of interest.
