MYC regulates the antitumor immune response through CD47 and PD-L1

See allHide authors and affiliations

Science  08 Apr 2016:
Vol. 352, Issue 6282, pp. 227-231
DOI: 10.1126/science.aac9935

This article has a correction. Please see:

Oncogene control of antitumor immunity

Recent clinical success of cancer immunotherapy has intensified interest in how tumors normally evade the immune response. Whether and how oncogenes contribute to this process are not well understood. In a study of mice, Casey et al. found that the MYC oncogene, which is aberrantly activated in many human cancers, up-regulates the expression of genes encoding proteins that dampen the antitumor response. These include two proteins that are often overexpressed on tumor cells and that serve as immune checkpoints. One of them (PDL1) sends to the immune system a “don't find me” signal, and the other (CD47) sends a “don't eat me” signal. Thus, therapies aimed at suppressing MYC may help promote an immune response against tumors.

Science, this issue p. 227


The MYC oncogene codes for a transcription factor that is overexpressed in many human cancers. Here we show that MYC regulates the expression of two immune checkpoint proteins on the tumor cell surface: the innate immune regulator CD47 (cluster of differentiation 47) and the adaptive immune checkpoint PD-L1 (programmed death–ligand 1). Suppression of MYC in mouse tumors and human tumor cells caused a reduction in the levels of CD47 and PD-L1 messenger RNA and protein. MYC was found to bind directly to the promoters of the Cd47 and Pd-l1 genes. MYC inactivation in mouse tumors down-regulated CD47 and PD-L1 expression and enhanced the antitumor immune response. In contrast, when MYC was inactivated in tumors with enforced expression of CD47 or PD-L1, the immune response was suppressed, and tumors continued to grow. Thus, MYC appears to initiate and maintain tumorigenesis, in part, through the modulation of immune regulatory molecules.

The transcription factor MYC regulates the expression of a multitude of gene products involved in cell proliferation, growth, self-renewal, differentiation, and apoptosis (14). The MYC gene is genetically activated and overexpressed in many human cancers (14), and this overexpression has been causally linked to tumorigenesis (5, 6). Studies involving inducible transgenic mouse models have shown that growth of Myc-induced tumors is dependent on continuous expression of MYC (14, 710). For example, in the tetracycline-off mouse model (where Myc expression can be turned off by the addition of tetracycline or doxycycline), tumors grow only when Myc is “on.” When Myc is turned “off,” tumors regress.

In mouse models, MYC inactivation results in tumor regression through the induction of proliferative arrest and apoptosis (13, 7, 8, 1012). We have demonstrated that complete tumor clearance that occurs after the inactivation of oncogenes, including Myc, requires the recruitment of CD4+ T cells and the secretion of thrombospondin-1 (13, 14). Hence, a host-dependent immune response is required for sustained tumor regression. However, the mechanism by which oncogene inactivation elicits this immune response is unknown.

The host immune system generally serves as a barrier against tumor formation (15). Activation of the immune response can contribute to tumor regression (13, 16, 17) through both adaptive and innate immune effectors (1820). Programmed death–ligand 1 (or PD-L1, also known as CD274 and B7-H1) sends a critical “don’t find me” signal to the adaptive immune system (2123), whereas CD47 (cluster of differentiation 47) sends a critical “don’t eat me” signal to the innate immune system and acts as a regulator of the adaptive immune response (24, 25) (fig. S1A). These and similar molecules are often overexpressed on human tumors (22, 25). Therapeutic suppression of PD-L1 and other immune checkpoint molecules elicits an immune response against tumors. Recently, this strategy has been applied in clinical practice, with very encouraging results (2629).

To explore whether and how MYC regulates the antitumor response, we examined its effect on the expression of CD47 and PD-L1 in the Tet-off transgenic mouse model of MYC-induced T cell acute lymphoblastic leukemia (MYC T-ALL). When MYC was turned on, both CD47 and PD-L1 were expressed. However, in vitro or in vivo Myc inactivation resulted in a rapid down-regulation of CD47 and PD-L1—both at the mRNA level, as detected by quantitative real-time polymerase chain reaction (qPCR), and at the protein level, as detected by flow cytometry (Fig. 1, A and B) and immunofluorescence (fig. S1B). Expression of other immune-related surface receptors was not affected by MYC inactivation (Fig. 1A). Consistent with these observations, suppression of MYC expression in the human T-ALL cell lines CCRF-CEM and Jurkat, either by treatment with a MYC-targeting short hairpin RNA (shRNA) (fig. S2A) or with the BET (bromodomain and extraterminal) inhibitor JQ1 (30), reduced the expression of CD47 and PD-L1 (Fig. 1C). Treatment of MYC T-ALL cells with the chemotherapeutic drugs prednisone, cytoxan, cisplatin, or vincristine, resulted in tumor cell death. However, CD47 and PD-L1 were either unaffected or showed increased expression (fig. S3A), and there was no effect on the expression of CD3, CD8, CD25, or CD69 (fig. S3, B to E).

Fig. 1 MYC regulates the expression of CD47 and PD-L1 in murine and human leukemia and lymphomas.

(A) Flow cytometry median fluorescence intensity (MFI) was used to determine the relative cell surface expression of CD47 (blue), PD-L1 (green), and other immune proteins after MYC inactivation in MYC T-ALL 4188 cells in vitro (n = 3 replicates). (B) Tumors were harvested from primary MYC-driven lymphomas 0 to 4 days after MYC inactivation. mRNA and protein levels were quantified by qPCR and flow cytometry MFI (n = 3 tumors per condition). Representative flow cytometry histograms are shown to the right. (C) CD47 (blue) and PD-L1 (green) protein levels in Jurkat and CCRF-CEM cells were quantified by flow cytometry MFI after MYC inhibition by conditional shRNA knockdown or 10 μM JQ1 treatment (n = 3 biological replicates). *P < 0.05; **P < 0.01; ***P < 0.001. Error bars indicate SEM.

We next investigated the effect of MYC inactivation on CD47 and PD-L1 in mouse and human solid tumors. In a Tet-off transgenic mouse model of hepatocellular carcinoma (HCC) (3), inhibition of MYC expression resulted in decreased levels of CD47 and PD-L1 protein (fig. S4, A and B) and mRNA (fig. S4B); expression of the two proteins was not affected by cisplatin treatment (fig. S4A). In the human HCC cell line HEPG2, shRNA knockdown of MYC caused a reduction in the levels of both CD47 and PD-L1 mRNA (fig. S4C). We also investigated the relationship between MYC expression and CD47 and PD-L1 expression in the human melanoma cell line SKMEL28 (Fig. 2A) and the human non–small cell lung cancer (NSCLC) cell line H1299 (Fig. 2B), as these cells represent tumor types that are often treated with immune checkpoint inhibitors in the clinic (31). We found that MYC shRNA knockdown and MYC functional suppression by JQ1 reduced the expression of CD47 and PD-L1 mRNA and protein as measured by qPCR and flow cytometry, respectively.

Fig. 2 MYC regulates CD47 and PD-L1 expression in human and mouse tumors and binds to the promoters of the corresponding genes.

(A and B) mRNA and protein levels of MYC (gray), CD47 (blue), and PD-L1 (green) in human melanoma SKMEL28 and human NSCLC H1299 cells as determined by qPCR and flow cytometry MFI, respectively, 48 hours after MYC inactivation in vitro. MYC was inactivated by 10 μM JQ1 treatment or MYC shRNA knockdown (n = 3 biological and 3 technical replicates for qPCR, n = 3 biological replicates for flow cytometry). *P < 0.05; **P < 0.01; ***P < 0.001. Error bars indicate SEM. (C) ChIP-seq analysis of MYC binding to the promoter sequence of the genes encoding CD47 and PD-L1 in mouse MYC T-ALL cells. Immunoglobulin G (IgG) was used as a negative control. ChIP-seq traces were generated from GSE44672 (34). Exons are represented as vertical bars, the untranslated region is represented by a black line, and arrows indicate the direction of transcription.

In additional experiments, we found that MYC shRNA knockdown (fig. S2B) or JQ1 treatment of four independent primary human T-ALL samples reduced both CD47 and PD-L1 cell surface expression (fig. S5). Cisplatin treatment increased CD47 and PD-L1 expression, whereas CD8 expression was unaffected by the treatments (fig. S5). Last, we examined publicly available gene expression data derived from human primary tumors. Notably, in human HCC, renal cell carcinoma, and colorectal carcinoma, MYC expression significantly correlated with the expression of both CD47 and PD-L1 (fig. S6). Thus, MYC regulates CD47 and PD-L1 expression in multiple human tumor types.

MYC can act as a general transcriptional amplifier (that is, it can generally increase expression of many genes rather than specific target genes), but dosage-dependent specific effects have been reported (3236). We analyzed chromatin immunoprecipitation sequencing (ChIP-seq) data from mouse MYC T-ALL cells (34) and the human B cell line P493-6 (37, 38) and found high levels of MYC bound to the promoter regions of the genes coding for CD47 and PD-L1 (Fig. 2C and figs. S7 and S8). In contrast, we observed that both MYC T-ALL (fig. S7) and P493-6 (fig. S8) cells with high levels of MYC had lower, often nonspecific, binding to the promoters of other cell surface immune molecules such as CD8a and CD25. Oncogenic levels of MYC bound the CD47 and PD-L1 gene promoters in human osteosarcoma U2OS cells, whereas low levels of MYC did not (fig. S9). In a nuclear run-on assay with P493-6 cells, MYC induced expression of the CD47 gene, along with other well-known target genes such as PDK1, CHEK1, CDK2, LDHA, and ODC1 (fig. S10, A and B). PD-L1 expression was too low to measure changes in this experiment. Thus, we conclude that MYC binds to the promoters and directly regulates the expression of the CD47 and PD-L1 genes. An alternative but not mutually exclusive possibility is that MYC suppression acutely affects CD47 and PD-L1 surface protein expression by reducing the half-lives of the two proteins. However, we did not observe the increased turnover of CD47 or PD-L1 proteins compared with other immune surface proteins in mouse MYC T-ALL cells when we inhibited protein synthesis by cycloheximide treatment (fig. S11).

We have shown previously that MYC inactivation in mouse tumor models results in recruitment of immune cells to the tumors (13). To investigate the role of CD47 and PD-L1 in this process, we engineered MYC T-ALL 4188 cells to constitutively express CD47 or PD-L1 (fig. S12A). In this overexpression system, CD47 and PD-L1 mRNA levels were unaffected by MYC inactivation (fig. S12B). The recruitment of luciferase-labeled CD4+ T cells (Fig. 3A), CD69+-activated T cells, and F4/80+ macrophages (Fig. 3B and fig. S13) after MYC inactivation was suppressed when CD47 and PD-L1 were constitutively expressed by the tumor cells. Expression of CD47 or PD-L1 prevented the sustained tumor regression that has been observed upon MYC inactivation (Fig. 4A), without affecting MYC expression (Fig. 4B). Enforced expression of CD47 or PD-L1 increased minimal residual disease (tumor cells remaining), resulting in tumor recurrence (Fig. 4, C and D). Conversely, shRNA knockdown of CD47 or PD-L1 prevented the growth of MYC T-ALL cells in vivo (fig. S14).

Fig. 3 Constitutive expression of CD47 and PD-L1 in mouse MYC T-ALL 4188 cells prevents recruitment of immune effectors after MYC inactivation.

(A) Quantification of CD4+ T cells in transplanted control (gray) or constitutive CD47- or PD-L1–expressing (colored) tumors before 2, 4, 11, or 21 days after MYC inactivation. Control, CD47-expressing, or PD-L1–expressing MYC T-ALL 4188 tumor cells were transplanted into FVB RAG1−/− mice 1 week after reconstitution with fLuc+ CD4+ T cells. Administration of Dox to inactivate MYC in established tumors occurred on day 0. (Left): Representative bioluminescence images of tumor-bearing RAG1−/− animals. MSCV, murine stem cell virus. (Right) Average bioluminescence signal of the T cells (n = 5 tumors per group). (B) Quantification of F4/80+ or CD69+ cells in transplanted control (gray) or constitutive CD47- or PD-L1–expressing (colored) tumors before or 4 days after MYC inactivation, by immunohistochemistry using markers for macrophages (F4/80) and activated T cells (CD69). Tumor cells were transplanted into wild-type (WT) FVB hosts. Administration of Dox to inactivate MYC in established tumors occurred on day 0. The y axis denotes the number of positively-staining cells per field. For representative images, see fig. S13. Data represent mean ± SEM derived from measurements of three independent tumors and three measurements per tumor. *P < 0.05; ***P < 0.001.

Fig. 4 Down-regulation of CD47 or PD-L1 is required for tumor regression, shutdown of angiogenesis, and induction of senescence upon MYC inactivation.

(A) Survival after MYC inactivation of syngeneic FVB/N mice that had been transplanted with either MSCV control (gray), CD47-expressing (blue), or PD-L1–expressing (green) fLuc+ MYC T-ALL cells. MYC was inactivated when tumors reached 1.5 cm3 (day 0) (n = 5 tumors for control, n = 10 tumors for CD47, and n = 5 tumors for PD-L1). (B) MYC expression before (day 0) or after MYC inactivation (day 4). (C) Bioluminescence imaging measurement of tumor burden before and after MYC inactivation in control (gray), CD47-expressing (blue), and PD-L1–expressing (green) tumors. Data for three representative animals per group are shown. (D) Minimal residual disease (remaining tumor cells) after MYC inactivation was measured by bioluminescence imaging. (E) Angiogenesis was measured 0 and 4 days after MYC inactivation in control, CD47-expressing, and PD-L1–expressing tumors growing in WT FVB hosts by immunofluorescence for CD31. For representative images, see fig. S15. (F) Control, CD47-expressing, and PD-L1–expressing tumors [as described in (E)] were analyzed by immunostaining for senescence-associated β-gal. The y axis denotes the number of positively-staining microvessels (E) or cells (F) per field. For representative images, see fig. S15B. Data represent mean ± SEM derived from measurements of three independent tumors and three measurements per tumor. *P < 0.05; **P < 0.01; ***P < 0.001.

Inactivation of MYC induces tumor regression through both cell-autonomous mechanisms, including proliferative arrest and induction of apoptosis, and host-dependent mechanisms, such as inhibition of tumor angiogenesis and induction of tumor cell senescence (14, 13). We investigated the effect of enforced expression of CD47 or PD-L1 on these mechanisms and found that CD47 or PD-L1 expression prevented the shutdown of angiogenesis after MYC inactivation, as measured by the presence of CD31+ microvessels (Fig. 4E and fig. S15A) and expression of Ang2 and Tie2 (fig. S15C). The induction of tumor cell senescence, as measured by β-galactosidase (Fig. 4F and fig. S15B) and p15Ink4b and p19ARF levels (fig. S15D), was also affected, but we did not observe any effect on apoptosis or proliferation, as evaluated by annexin V and 7-AAD (fig. S16A), cleaved caspase 3 (fig. S16, B and D), and phospho-histone H3 (fig. S16, C and E). Therefore, the down-regulation of CD47 and PD-L1 appears to be required for the induction of sustained tumor regression, the shutdown of angiogenesis, and senescence induction promoted by MYC inactivation.

We conclude that MYC regulation of CD47 and PD-L1 expression has a direct role in the initiation and maintenance of MYC-driven tumorigenesis (Fig. 4). The overexpression of MYC may be one general mechanism by which tumor cells up-regulate the expression of immune checkpoint regulators, thereby evading immune surveillance. MYC inactivation has been proposed to restore the immune response against tumors (fig. S17) (3941).

The suppression of MYC rapidly resulted in decreased mRNA and protein expression of CD47 and PD-L1, which suggests a transcriptional regulatory mechanism. MYC is a general transcriptional amplifier that can regulate gene expression through a multitude of mechanisms (3234). However, as noted above, MYC also exhibits transcriptional effects dependent on gene dosage (36, 42). The relatively high levels of MYC expression that are associated with rapid proliferation and tumorigenesis may induce CD47 and PD-L1 expression.

Because MYC regulates transcription of their genes, CD47 and PD-L1 may be expressed at higher levels in the steady state as compared with other membrane proteins in tumors. Notably, MYC activation of the CD47 and PD-L1 genes appears to require higher levels of MYC binding to the CD47 and PD-L1 promoters compared with genes involved in normal cell growth; they may, therefore, represent promoters that have been invaded by oncogenic MYC levels (33, 42). Thus, these genes may be particularly sensitive to MYC withdrawal.

Through the suppression of immune surveillance against tumor cells, MYC activation may influence cancer immunoediting. We propose that during tumor evolution, high levels of MYC expression result in increased expression of CD47 and PD-L1, suppressing both the innate and adaptive immune responses and favoring tumor growth (fig. S17). Upon MYC inactivation, loss of the “don’t find me” and “don’t eat me” signals allows for the destruction of residual tumor cells and, consequently, sustained tumor regression.

Although the effects of MYC on the expression of CD47 and PD-L1 were modest, the consequences on tumor regression were dramatic, consistent with reports that small influences on immune regulators can have marked effects (25). Both CD47 and PD-L1 may also contribute to the tumor microenvironment via their influence on T cell activation and angiogenesis (13, 14, 22, 4345). CD47 is the receptor for thrombospondin-1, which may regulate cellular programs including angiogenesis, self-renewal, and senescence (13, 14, 45). We speculate that therapies suppressing MYC expression and activity may restore an immune response against human cancers. Humans cancers that overexpress MYC may be especially vulnerable to an immune checkpoint blockade.

Supplementary Materials

Materials and Methods

Figs. S1 to S17

Table S1

References (4654)

References and Notes

Acknowledgments: We thank E. Shroff, A. Deutzmann, J. Braun, D. Fruman, I. Weissman, and P. Betancur for helpful advice; P. Chu for pathology; N. Lacayo and G. Dahl for deidentified clinical specimens; C. Dang for nuclear run-on data sets; and the Small Animal Imaging Facility at the Stanford Center for Innovation in In-Vivo Imaging (SCI3). This work was supported by NIH R01 grants CA 089305, CA 170378, CA 184384, U01 CA 188383, and U01 CA 114747 and a Cancer Research Institute Clinic and Laboratory Innovation Program grant (D.W.F). S.C.C. was supported by NIH fellowships 1F32CA177139 and 5T32AI07290. K.N.F. was supported by a grant from Alex’s Lemonade Stand Foundation. M.E. and S.W. were supported by Deutsche Forschungsgemeinschaft (DFG) grant Ei222/12-1 (to M.E.) and by the Deutsche Krebshilfe via the Comprehensive Cancer Center Mainfranken. I.G. was supported by DFG grant GU 1046/2-1. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. All data are stored at the Division of Oncology, Department of Medicine, Stanford University School of Medicine.

Stay Connected to Science

Navigate This Article