Hypercholesterolemia Increases Colorectal Cancer Incidence by Reducing Production of NKT and γδ T Cells from Hematopoietic Stem Cells

Obesity will soon surpass smoking as the most preventable cause of cancer (1, 2). Studies to determine which metabolic disorder or combination of disorders in obese people increases their cancer risk have been inconclusive (3). Hypercholesterolemia, a common metabolic disorder in obese people, has been shown to increase cancer risk and substantial epidemiologic evidence links hypercholesterolemia to an increased risk of colorectal cancer (4). It was originally proposed that hypercholesterolemia exerts a systemic, conditional influence that affects immunosurveillance against colorectal cancer. In support of this hypothesis, hypercholesterolemia has been shown to reduce the frequency and function of the cellular components of tumor immunosurveillance (5, 8). The mechanism by which hypercholesterolemia reduces the number and function of innate immune cells is unknown, nor is it clear if such an effect compromises immunosurveillance against colorectal cancer.

Emerging evidence suggests that, although hematopoietic stem cells (HSC) maintain an undifferentiated state, activating epigenetic marks and low-level expression of lineage-associated genes, a process known as lineage priming, keep HSCs responsive to physiologic and pathologic demands of immune cells (9–12). Recently, we have shown that hypercholesterolemia induces oxidant stress in HSCs that accelerates their aging and impairs their repopulation capacity (13). With these findings, we hypothesize that hypercholesterolemia-induced oxidant stress reduces HSC lineage priming toward innate immune cells and thereby impairs immunosurveillance against colorectal cancer.

Here, we show that hypercholesterolemia-induced oxidant stress downregulates Ten Eleven Translocation 1 (Tet1) in HSCs, resulting in increased DNA hypermethylation and histone modifications in the genes critical to natural killer T cell (NKT) and γδ T-cell differentiation. These effects reduced the number and function of terminally differentiated NKT and γδT cells in the thymus, the colon submucosa and at the early stages of tumorigenesis and thereby impaired immunosurveillance against colorectal neoplasia.

All mice were purchased from Jackson Laboratories. Care of mice was in accordance with NIH guidelines. ApoE^−/− and wild-type (WT) mice were fed standard mouse chow (5.4 g fat/100 g diet, 0% cholesterol). HCD mice were fed a diet with 10 g fat/100 g diet, 11.25 g cholesterol/100 g diet (Research Diets). NAC was given for 8 weeks (150 mg/kg/day via drinking water).

293T (CRL-3216) cell lines were obtained from the ATCC repository. Cell line characterization by ATCC is conducted by short tandem repeat (STR) analysis. OP9-DL1 cells were kindly provided by Dr. Juan Carlos Zúñiga-Pflücker (University of Toronto, Toronto, Canada). Upon receiving the cell lines, frozen stocks were prepared within one to five passages and new stocks were thawed frequently to keep the original condition. The cell lines were passaged for less than 3 months after receipt or resuscitation. Cells were authenticated by morphology, phenotype, and growth, routinely screened for mycoplasma.

The colorectal neoplasia experiments were performed as described in previous publications (14)_. Three-month-old mice were subcutaneously injected with a solution of AOM at a dose rate of 15 mg/kg body weight, once weekly for 3 successive weeks. Two percent of DSS was given in the drinking water over 5 days in the last week. Mice were sacrificed 10 weeks after the last injection of AOM. Tumor counts and histopathologic staging of tumors were performed by a cancer pathologist in a blind fashion.

Cells were stained with mAbs conjugated to various fluoroprobes. These antibodies included: cKit (2B8), Sca-1 (E13-161.7), CD4 (L3T4), CD8 (53-6.72), CD90.1, CD25, CD44, TCRβ, NK1.1, γδTCR, CD45.1, CD45.2. The lineage cocktail consisted of CD4, CD8, B220 (RA3-6B2), TER-119, Mac-1 (MI/70), and Gr-1 (RB6-8C5). All antibodies were purchased from BD Bioscience. CD1d-GalCer tetramer was obtained from the NIH Tetramer facility. FACS analysis was carried out on a FACS Diva or MoFlow. HSCs were isolated from the bone marrow and defined as cKit⁺ sca-1⁺ CD90.1^lo/−Lin⁻ (KTLS).

The Tet1 specific and control shRNA plasmids were purchased from Santa Cruz Biotechnology. The plasmid with Tet1 catalytic domain (pTYF-U6-shCONT-EF1-Puro-2A-CD1) was a gift from Dr. Yi Zhang (Boston Children's Hospital, Boston, MA). The envelope and helper plasmids were purchased from ABM. The lentiviral particles were prepared according to the kit instructions. Fresh isolated KSL cells were transduced with lentivirus for 24 hours and then selected with puromycin (2 μg/mL; Santa Cruz Biotechnology) for 72 hours.

Native LDL (nLDL) was purchased from Sigma. OxLDL was prepared by incubating nLDL with 10 μmol/L CuSO4 at 37°C for 24 hours. The material was dialyzed against a sterile solution (150 mmol/L NaCl, 1 mmol/L EDTA, 100 μg/mL polymyxin B, pH 7.4) and then sterilized by filtration. The extent of LDL oxidation was estimated by agarose gel electrophoresis and by measuring the amount of thiobarbituric acid reactive substances generated (Supplementary Fig S1A and S1B).

The coculture was performed as described (15, 16). KTLS cells were seeded at 4 × 10³ cells/well into 12-well tissue culture plates containing a confluent monolayer of OP9-DL1 cells. All cultures were performed in the presence of 5 ng/mL IL2, 10 ng/mL GM-CSF (Stem cell Technology), 5 ng/mL IL7, and 5 ng/mL mFLT3 (PeproTech).

We used a standard protocol to detect NKT and γδT cells in colon and tumor tissues. The antibodies were purchased from BD Biosciences. For indirect IHC, we used rabbit-specific IgG conjugated with FITC or PE (Chemicon) as a secondary antibody. Fluorescent images were obtained using a confocal laser scanning microscope (Carl Zeiss LSM 510 system; Carl Zeiss).

Chromatin immunoprecipitation (ChIP) was performed following a standard protocol. A detailed description can be found in Supplementary Methods.

We reverse transcribed cDNAs from total RNA isolated from each cell fraction using RNAqueous-Micro Kit (Ambion). Transcription to cDNA was performed using SuperScript III (Invitrogen). All PCRs were carried out in triplicate using an Eppendorf Mastercycler (Eppendorf).

Genomic DNA was extracted with a standard protocol. PCRs were performed using a converted DNA by 2xHiFi Hotstart Uracil+Ready Mix RCR Kit (Kapa Biosystems). Please see Supplementary Methods for detailed description.

Total RNA was isolated using the mirVana miRNA Isolation Kit according to manufacturer's instructions (Applied Biosystems). The predicted target genes of differentially expressed miRNAs were obtained using the following tools: TargetScan v6.2 and miRDB. Please see Supplementary Methods for detailed description.

All data were shown as means ± SD. Statistical analyses were carried out with either GraphPad Prism (GraphPad Software) or SPSS v19 (IBM) software. Statistical significance was evaluated by using a one- or two-way ANOVA or an unpaired t test. Significance was established for P values of at least <0.05.

We first induced colorectal neoplasia with azoxymethane (AOM) in two common mouse models of hypercholesterolemia, the ApoE^−/− mouse, and the C57BL/6 mouse fed a high cholesterol diet (HCD). The average tumor number was almost two-fold higher in hypercholesterolemic mice than in WT mice (Fig. 1A), indicating that hypercholesterolemia increases the incidence of colorectal neoplasia. The tumors at late stages of tumorigenesis, including adenoma+++ and carcinomas, were not found in WT control mice, but constituted more than 10% of the tumors found in hypercholesterolemic mice. Meanwhile the, tumors at the early stages of tumorigenesis, including hyperplasia and adenoma+, were dramatically reduced in hypercholesterolemic mice (Fig. 1B), indicating that hypercholesterolemia significantly increases the incidence and the pathological severity of colorectal neoplasia.

To determine whether hypercholesterolemia-induced oxidant stress causes an HSC-autonomous defect that increases the incidence of colorectal neoplasia by impairing HSC lineage specification, we induced colorectal neoplasia in a chimeric model whereby hematopoiesis was reconstituted in lethally irradiated WT recipient mice (CD45.1) with HSCs from either ApoE^−/− or WT mice. In the recipient WT mice, the thymus was repopulated with cells that were derived from the donor HSCs (CD45.2; Supplementary Fig. S1C). As expected, the serum cholesterol and white blood cell counts of the recipient mice were normal and comparable to those of WT mice (Supplementary Fig. S1D and S1E). Despite normal serum cholesterol levels, the average tumor number and their histopathologic severity was significantly greater in recipients that received HSCs from ApoE^−/− mice than in those that received HSCs from WT mice (Fig. 1C and D). Thus, both chimeric models in which WT mice received HSCs from ApoE^−/− mice recapitulated the increased tumor incidence and greater histopathologic severity seen in hypercholesterolemic mice. These results show that the increased incidence of colorectal neoplasia in hypercholesterolemic mice is due to a HSC-autonomous mechanism.

Upon TCR activation, NKT and γδT cells rapidly secrete a variety of cytokines that are critical for the antitumor functions of cytotoxic T cells. In addition to the cytokines produced by antigen-presenting cells with which NKT and γδT cells interact, these cytokines recruit and stimulate the antitumor functions of cytotoxic T cells, boosting innate and adaptive antitumor responses. Activated NKT and γδT cells both have strong cytotoxic effector activity (17–19). In this context, NKT and γδT cells function as major participants in tumor immunosurveillance.

The populations of γδT cells and NKT cells in the thymus were significantly lower in hypercholesterolemic mice than in WT mice (Fig. 2A and B; Supplementary Fig. S2A and S2B). Intrathymic NKT-cell development in ApoE^−/− mice was identical to WT at phases 1 (CD44⁻NK1.1⁻) and 2 (CD44⁺NK1.1⁻; Supplementary Fig. S2C) as well as the CD4 subsets of NKT cells (Supplementary Fig. S2D). T-cell developmental intermediates were similar in all groups (Supplementary Fig. S2E and S2F). Except for a slight decrease in B cells and a slight increase in NK cells, FACS analysis did not show any significant change in CD3e⁺, CD4⁺, and CD8⁺ T-cell populations in peripheral blood of hypercholesterolemic mice (Supplementary Fig. S2G). In the thymus of WT recipient mice reconstituted with HSCs from ApoE^−/− mice, we observed a nearly identical decrease in differentiation toward NKT and γδ T cells as that seen in ApoE^−/− mice (Fig. 2C and D). These results indicate that hypercholesterolemia specifically induces a cell-autonomous reduction in lineage specification of HSCs toward NKT and γδ T cells.

Consistent with the reduction of NKT and γδ T-cell maturation in the thymus, we also found a robust reduction of NKT cells (up to 6-fold) and γδ T cells (3-fold) in the colon submucosa of ApoE^−/− mice (Fig. 2E and F). Meanwhile, the other major cellular components of cancer immunosurveillance, including NK cells, CD11b⁻ dendritic cells, CD11b⁺ dendritic cells, CD11c⁻ macrophages, and CD11c⁺ macrophages in the colon of hypercholesterolemic mice did not show significant changes (Supplementary Fig. S2H).

To confirm if this substantial reduction of NKT and γδT cells impairs tumor immunosurveillance, we determined the incidence of AOM-induced colorectal neoplasia in Tcrd^−/− mice, which lack γδT cells, and CD1d^−/− mice, which lack NKT cells. Both mice strains showed significantly higher incidence and greater histopathologic severity of colorectal neoplasia than their control strains (Fig. 2G and H). In addition, we also found significantly fewer NKT and γδT cells infiltrated into the early but not later stages of tumor progression from hypercholesterolemic mice than in those from WT mice (Fig. 2I and J). Together, these results indicate that hypercholesterolemia reduces the differentiation of HSCs toward NKT and γδT cells, resulting in impaired tumor immunosurveillance against colorectal neoplasia.

We previously showed that hypercholesterolemia induces an oxLDL-dependent increase of oxidant stress in HSCs that accelerated their ageing and impaired their repopulation capacity, both of which were reversed by the antioxidant N-acetylcysteine (NAC; ref. 13). Interestingly, NAC administration rescued the otherwise impaired differentiation toward NKT and γδ T cells in hypercholesterolemic mice (Supplementary Fig. S3A and S3B). NAC also significantly decreased the average tumor number in ApoE^−/− and HCD mice. Although NAC reduced the histopathologic severity of tumors in HCD mice, the reduction in the histopathologic severity of tumors in ApoE^−/− mice did not reach statistical significance (Supplementary Fig. S3C and S3D). Finally, NAC increased significantly the infiltration of NKT and γδ T cells in early stages of tumor development in both ApoE^−/− and HCD mice (Supplementary Fig. S3E and S3F). Regression analysis between the degree of HSC oxidant stress and the number of tumors per mouse revealed a remarkable linear correlation between these variables (R² = 0.87; Supplementary Fig. S3G). Together, these findings show that hypercholesterolemia-induced HSC oxidant stress directly mediates the reduction of HSC differentiation toward NKT and γδ T cells and the consequent increase in tumor number and histopathologic severity.

The cell autonomous defect in HSC differentiation caused by hypercholesterolemia-induced oxidant stress raised the possibility that oxidant stress disrupted an epigenetic regulatory pathway necessary for proper HSC differentiation to NKT and γδT cells. In support of the possibility, we observed an oxidant stress-dependent reduction in the expression of Tet1 in HSCs from hypercholesterolemic mice (Fig. 3A and B).

Within the Tet family, Tet2 has been shown to have a critical role in regulating the self-renewal, proliferation, and differentiation of HSCs (20–23), whereas the role of Tet1 in HSC differentiation is as yet unknown. To determine if this reduction of Tet1 expression is directly responsible for the decrease in NKT and γδT cells and the impairment of tumor immunosurveillance in hypercholesterolemic mice, we measured the percentage and total number of NKT and γδT cells in the thymus of WT and Tet1^−/− mice. Consistent with our findings in hypercholesterolemic mice, the percentage and total number of NKT and γδT cells in the thymus was significantly lower in Tet1^−/− mice than WT mice (Fig. 3C and D). The percentage and number of NKT and γδT cells was also decreased in the colon submucosa of Tet1^−/− mice (Fig. 3E and F). Meanwhile, Tet1^−/− mice did not show significant changes in T-cell intermediate populations in the thymus (Supplementary Fig. S4A and S4B). With the exception of a slight increase in NK cells and a slight decrease in B cells, we did not find significant changes in CD3e⁺, CD4⁺, and CD8⁺ cells in peripheral blood of Tet1^−/− mice (Supplementary Fig. S4C). The other major cellular components of cancer immunosurveillance, including NK cells, CD11b⁻ dendritic cells, CD11b⁺ dendritic cells, CD11c⁻ macrophages, and CD11c⁺ macrophages did not show significant changes in the colon of Tet1^−/− mice (Supplementary Fig. S4D). In an in vitro HSC differentiation assay, knockdown of Tet1 expression in HSCs from either WT or ApoE^−/− mice greatly reduced their differentiation toward NKT and γδT cells (Supplementary Fig. S5A–S5C).

The overexpression of Tet1 in HSCs from WT or ApoE^−/− mice caused a seven-fold increase in WT and almost 20-fold increase in ApoE^−/− mice in HSC differentiation toward NKT cells both in vitro and in vivo. The overexpression of Tet1 in HSCs also caused 10-fold increase in WT and 20-fold increase in ApoE^−/− mice in HSC differentiation toward γδT cells (Fig. 3G–I; Supplementary Fig. S5A, S5D, and S5E). The overexpression of Tet1 in HSCs did not affect the daughter T-cell intermediate populations in thymus as well as B cells, NK cells, CD3e⁺, CD4⁺, and CD8⁺ cells in peripheral blood of recipient mice (Supplementary Fig. S5F–S5H). These results further support the novel and specific role of Tet1 in HSCs lineage specification toward NKT and γδT cells.

IL17 is a critical cytokine in both innate and adaptive immunity. CCR6 regulates the migration and recruitment of T cells during inflammatory and immunological responses (24–27). γδT cells from ApoE^−/− mice also showed significant decreases in the production of IL17 (Supplementary Fig. S5I and S5J). Interestingly, γδT cells derived from Tet1-overexpressing HSCs also displayed greater expression of CCR6 and IL17 than WT mice (Supplementary Fig. S5K and S5L). These results also indicate that Tet1 expression in HSCs is a pivotal determinant not only of the lineage specification of HSCs toward NKT and γδT cells but also of the function of terminally differentiated NKT and γδT cells.

We next sought to determine in vivo whether the overexpression of Tet1 in HSCs from hypercholesterolemic mice could restore their normal lineage specification toward NKT and γδT cells and thereby immunosurveillance against colorectal neoplasia. We reconstituted hematopoiesis of WT recipient mice with WT HSCs, Tet1-overexpressing WT HSCs or Tet1-overexpressing ApoE^−/− HSCs. When we tried to reconstitute WT and ApoE^−/− mice with HSCs that overexpress Tet1, all the mice died. We assumed these deaths were secondary to failed reconstitution of the bone marrow of the irradiated mice. To address this problem, the transplantation of Tet1-overexpressing WT HSCs was supported with normal, nontransduced WT HSCs and similarly the transplantation of Tet1 overexpressing ApoE^−/− HSCs was supported with ApoE^−/− HSCs, both at the ratio of 3:1 (Fig. 4A). Under these conditions, all mice survived and overexpression of Tet1 in ApoE^−/− HSCs restored the number of NKT and γδT cells in the thymus of recipient WT mice to that of recipient WT mice reconstituted with WT HSCs (Fig. 4B and C). Overexpression of Tet1 in ApoE^−/− HSCs restored the number of submucosal NKT and γδT cells in recipient WT mice (Fig. 4D and E). Most significantly, overexpression of Tet1 in ApoE^−/− HSCs reduced the average tumor number of colorectal neoplasia in recipient irradiated WT to a level similar to that of overexpression of Tet1 in WT mice (Fig. 4F and G). Moreover, overexpression of Tet1 in both ApoE^−/− and WT HSCs had a profound effect on the histopathologic severity of the tumors. Indeed, lethally irradiated WT recipient mice reconstituted with either Tet1-overexpressing WT HSCs or Tet1-overexpressing ApoE^−/− HSCs eliminated the progression of any tumors to the carcinoma stage in both groups (Fig. 4G). These results indicate that restoration of Tet1 expression in ApoE^−/−HSCs rescues both their reduced lineage specification toward NKT and γδT cell populations as well as their effective immunosurveillance against colorectal neoplasia. The increase in NKT and γδT cell populations in the mucosa and submucosa and the consequent reduction in the histopathologic severity of the AOM-induced tumors in WT mice transplanted with Tet1 overexpressing HSCs was an unexpected finding that might have potential immunotherapeutic implications.

We performed miRNA microarray analysis in HSCs isolated from WT and hypercholesterolemic ApoE^−/- mice (Supplementary Fig. S6A). MiR101c, which is predicted to directly target Tet1, showed a significant upregulation in HSCs from ApoE^−/− mice. This increased level of miR101c was further validated by RT-PCR (Fig. 5A; Supplementary Fig. S6B). The administration of NAC effectively reduced the overexpression of miR101c in HSCs from hypercholesterolemic mice (Fig. 5A; Supplementary Fig. S6B). Transfection of miR101c mimics in HSCs from WT mice significantly repressed Tet1 expression (Fig. 5B and C), whereas transfection of miR101c inhibitors in HSCs from ApoE^−/− mice significantly increased Tet1 expression (Fig. 5D and E). miR101c significantly repressed luciferase activity in cells transfected with the constructs containing the predicted miR101c binding sites in the Tet1 3′-UTR regions. In contrast, miR101c failed to alter luciferase activity when we mutated the Tet1-binding sites, thereby confirming that Tet1 is the direct binding target of miR101c (Fig. 5F). These findings represent the first known effects of miR101c on gene expression in vivo.

Although the molecular mechanism underlying the differentiation of NKT and γδ T cells is still incompletely characterized, a group of genes that have been shown to mediate the differentiation and maturation of NKT and γδ T cells have been identified (28, 29). To determine the mechanism by which hypercholesterolemia-induced downregulation of Tet1 impairs differentiation of HSCs toward NKT and γδ T cells, we examined the expression and epigenetic regulation of genes necessary for NKT and γδ T-cell specification (Supplementary Table S1). Only five genes, Fyn, Sox13, IL15R, ITK, and SH2D1a, had lower expression in ApoE^−/− HSCs than in WT HSCs (Fig. 6A). Moreover, the expression of these genes increased when Tet1 was overexpressed in HSCs from WT and hypercholesterolemic mice (Fig. 6A; Supplementary Fig. S7A), thereby indicating a repression of the genes essential for NKT and γδ T differentiation in HSCs from hypercholesterolemic mice.

Because Tet-dependent DNA demethylation typically increases the transcription of target genes (23, 30), we next sought to characterize the changes in DNA methylation at the regulatory regions of the five genes whose expression was reduced in ApoE^−/− mice. Pyrosequencing analysis showed that Fyn, Sox13, IL15R, ITK, and SH2D1a were more hypermethylated in ApoE^−/− HSCs than in WT HSCs (Fig. 6B). In contrast, Tet1 overexpression in HSCs from WT and hypercholesterolemic mice significantly decreased the methylation and correspondingly increased the expression of these genes in both WT and ApoE^−/− HSCs (Fig. 6B and Supplementary Fig. S7B). Interestingly, several genes (ETV5, EGR2, SLAMF1, ZBTB16, and RELb) whose expression did not change significantly in ApoE^−/− HSCs were also increased after overexpression of Tet1 in HSCs of both WT and ApoE^−/− mice (Supplementary Fig. S7A), suggesting that Tet1 overexpression to levels higher than those found in WT HSCs can increase the expression of genes required for NKT and γδ T specification. Consistent with this possibility, methylation of ETV5, EGR2, and NFKB1c was significantly higher in cells derived from ApoE^−/− HSCs than those from WT HSCs, and this hypermethylation was reduced to levels at or below WT upon overexpression of Tet1 (Supplementary Fig. S7B). Taken together, these results show that Tet1 directly activates genes required for NKT and γδ T specification, and this activation is impaired upon Tet1 downregulation in hypercholesterolemic HSCs.

Tet proteins also participate in the regulation of histone modifications via distinct pathways. The O-linked N-acetylglucosamine (O-GlaNAc) transferase, OGT is an evolutionarily conserved enzyme that catalyzes O-linked protein glycosylation. Tet proteins were identified as stable partners of OGT in the nucleus (31–33). The interaction of Tet2 and Tet3 with OGT leads to the GlcNAcylation of Host Cell Factor 1 and contributes to the integrity of the H3K4 methyltransferase SET1/COMPASS complex, revealing that Tet proteins increase the level of H3K4me3, a modification that functions in transcriptional activation (34). Although an early observation showed that the interaction between Tet1 and OGT was limited to embryonic stem cells, our immunoprecipitation studies show that OGT also interacts with Tet1 in HSCs (Supplementary Fig. S8A). In accordance with the decrease in Tet1 expression, the Tet1–OGT interaction was significantly reduced in HSCs isolated from hypercholesterolemic mice. This overexpression of Tet1 significantly increased the interaction of Tet1 and OGT, but did not influence the expression or interaction of Tet3 and OGT in the cells (Supplementary Fig. S8A and S8B). Consistent with these findings, overexpression of Tet1 caused an increase of H3K4me3 methylation near the promoters of all genes investigated, except RELb and NFKB1. The results suggest that, by interacting with OGT, Tet1 also increases histone H3K4me3 levels and maintains active chromatin structure near many of the genes critical for the differentiation of HSCs toward NKT and γδT cells (Supplementary Fig. S8C). Consequently, Tet1 promotes the expression of genes driving NKT and γδT specification by multiple mechanisms.

To test whether these findings in mouse models of hypercholesteremia were applicable to human HSCs, human HSCs were exposed to oxLDL, the primary source of oxidant stress in the HSCs of hypercholesterolemic mice (13) and their differentiation capacities toward NKT and γδ T cells were examined. We observed an oxLDL concentration–dependent decrease in the differentiation of human HSCs toward NKT and γδ T cells (Supplementary Fig. S9A and S9B). Congruent with our mouse studies, the treatment with oxLDL inhibited Tet1 expression in human HSCs in a dose dependent manner (Supplementary Fig. S9C). Thus, these results in human HSCs are parallel to those in hypercholesterolemic mice, suggesting that these findings may be generalizable to humans.

Hypercholesterolemia has been shown to increase the risk of all-cause mortality, including an increased risk of death from cancer (12, 24). Here, we show that hypercholesterolemia increases the incidence and pathologic severity of experimental colorectal neoplasia by inducing an oxidant stress dependent downregulation of Tet1 in HSCs that impairs their differentiation toward NKT and γδT cells. This reduction of Tet1 expression downregulates genes critical to the differentiation of HSCs toward NKT and γδ T cells in large part by loss of activating histone H3K4me3 modifications and gain of repressive DNA methylation marks near the promoters of the key differentiation genes. These Tet1-induced effects on HSC differentiation reduce the number of terminally differentiated NKT and γδ T cells in the circulation, the submucosa of the gut, and finally within the early stages of tumor development. These findings reveal a novel mechanism by which innate immunity can be modulated by a metabolic abnormality at the level of HSC rather than at terminally differentiated immune cells. Thus, this pathologic change in gene expression was “preprogrammed” in HSCs and carried through bone marrow progenitor cells, thymic intermediate cells, and terminally differentiated NKT and γδ T cells.

Although genome-wide ChIP studies established the central role of epigenetic modifications in HSC fate decision (9–12), little is known of how the epigenetic regulators of lineage priming are linked to a given lineage specification. In this study, we show that the inhibition of Tet1 in HSCs imposed repressive modification of the genes that controls their differentiation toward γδT cells and NKT cells. This overexpression of Tet1 led to active modification of the genes and greatly increased the differentiation of HSCs toward NKT and γδT cells both in vivo and in vitro. These finding indicate that Tet1 is a master epigenetic regulator of HSC differentiation toward NKT and γδT cells.

Emerging evidence suggests that Tet1 functions as a critical tumor suppressor in multiple human cancers, including colorectal cancer (35). The analysis of tumor methylomes showed that Tet1, as a methylated target, is frequently methylated and downregulated in cell lines and primary tumors of multiple carcinomas and lymphomas, including gastric and colorectal carcinomas (36–38). The expression of the Tet1 catalytic domain is able to inhibit the CpG methylation of tumor suppressor gene promoters and reactivate their expression (37). The downregulation of Tet1 leads to repression of the DKK gene and constitutive activation of the WNT pathway, resulting in the initiation of tumorigenesis in colon. In addition, the reexpression of Tet1 in colon cancer cells inhibits their proliferation and the growth of tumor xenografts even at late stages (38). These studies provide convincing evidence that Tet1 is a critical regulator in preventing the malformation and transformation of colon cells. We found that Tet1, by supporting the differentiation from HSCs toward NKT and γδ T cells, also functions as a critical regulator of innate immunity against colorectal neoplasia. Our findings provide new evidence for understanding how tumors escape immunosurveillance in the context of hypercholesterolemia.

In general, NKT and γδ T cells are critical components of innate immunity against cancer and initiate the cascade of immune reactions to recognize and eliminate transformed cells in the early stage of immunosurveillance (1, 39, 40). Suppression of the T-cell responses to tumors has been observed frequently in cancer patients, specifically NKT and γδ T cells have been shown to be decreased or functionally hyporeactive in both cancer bearing mice and humans (41–43). In agreement with these studies, we also found that both NKT and γδ T cells of hypercholesterolemic mice were decreased in the colon submucosa and at early stages of tumorigenesis. The overexpression of Tet1 significantly increased the differentiation of HSCs toward NKT and γδ T cells both in vivo and in vitro. Despite reconstituting less than 10% of the T cells in the thymus of the transplanted recipient mice, γδ T cell and NKT cell numbers in the thymus and most importantly in the submucosa of the colon were normal. The clinical implications of these findings could be important.

The immunoscore is a more accurate predictor of tumor-free survival in patients who have colorectal neoplasia than is the classical Dukes Classification or the TMN score (44). Although, weak immune infiltrates have been postulated to be due to a defect in the host immune response, no etiology for this impaired host immune response has heretofore been identified (45). Thus, the identification of a new level of regulation of innate immunity by the epigenetic regulation of HSC differentiation may provide a context to understand why patients are found to have impaired Immunoscores in response to tumors. Creation of chimeric mice that received either ApoE^−/− or WT HSCs that overexpressed Tet1 had a substantial effect on the pathologic stage of the tumors where not only were all carcinoma tumors eliminated in the mice receiving ApoE^−/− transfected cells that overexpress Tet1 but so did the chimeric mice that received WT HSCs that overexpressed Tet1. Thus, a miR101c-Tet1 based HSC immunotherapy might not only be effective in patients found to have an impaired immunoscore but also in those who have relatively normal Immunoscores. However, much remains to be done to determine how these changes in HSC gene expression are carried through the complex process of their differentiation into terminally differentiated NKT and γδT cells.

Finally, the relationship between cardiovascular risk factors and cancer incidence is being actively investigated and sometimes referred to as CardioOncology (1). Cardiovascular disease and cancer share many similar risk factors. The relationship of hypercholesterolemia and colorectal cancer has been studied for decades. Some studies found an inverse relationship between serum hypercholesterolemia and colorectal cancer that raised questions about the relationship. However, it was learned subsequently that colonic adenocarcinoma cells can aggressively metabolize cholesterol and so studies that look for the relationship between these variables in patients with established cancers can lead to an erroneous conclusion. An additional issue is the findings that serum levels of 27-hydroxycholesterol (1), a metabolite of cholesterol, is associated with many cancers, especially breast cancer. We assume this is the result of a direct effect of 27-hydroxycholesterol on tumor cells. It will be important to see if it also affects lineage priming of HSCs.

No potential conflicts of interest were disclosed.

Conception and design: G. Tie, L.M. Messina

Development of methodology: G. Tie, J. Yan, J. Kang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Tie, J. Yan, L. Khair

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Tie, J. Yan, J.A. Messina, A. Deng, L.M. Messina

Writing, review, and/or revision of the manuscript: G. Tie, J. Yan, L. Khair, J.A. Messina, T. Fazzio, L.M. Messina

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Tie, J. Yan, L.M. Messina

Study supervision: L.M. Messina

Other (advice, consultation, and technical support): T. Fazzio

We thank Dr. Oliver Rando and Katelyn Sylvia (University of Massachusetts Medical School) for their great support. We thank Dr. Juan Carlos Zuniga-Pflucker (University of Toronto) for providing the OP9-DL1 cells. We thank Dr. Yi Zhang (Mass General Hospital, Boston, MA) for providing pTYF-U6-shCONT-EF1-Puro-2A-CD1. We thank Dr. Tin-Lap Lee (The Chinese University of Hong Kong) for providing the pmirGLO vector. We thank the NIH Tetramer Facility for the CD1d tetramer.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.