Macrophage Migration Inhibitory Factor Promotes Tumor Growth and Metastasis by Inducing Myeloid-Derived Suppressor Cells in the Tumor Microenvironment

The inflammatory tumor microenvironment contributes to tumor growth and progression through a variety of mechanisms (1). One cellular component of the tumor microenvironment, myeloid-derived suppressor cells (MDSCs), strongly enhances tumor growth and metastasis in a murine breast cancer model (2). Circulating MDSCs have been demonstrated in the blood of patients with a variety of tumors, including head and neck (3), lung (3), breast (3), gastric (4), pancreatic (4), and renal cell carcinomas (5). MDSCs use a variety of mechanisms to suppress T cell and NK cell functions, thereby inhibiting anti-tumor immunity (6, 7).

In mice, MDSCs are defined as GR1⁺CD11b⁺ cells, but this population can be further subdivided into two distinct subsets. The granulocytic subset is characterized by high Ly6G expression and low to intermediate Ly6C expression, whereas the monocytic subset is characterized by high Ly6C expression and low Ly6G expression (6, 8, 9). These two subtypes of MDSCs also have differing mechanisms of T cell suppression. Both subsets express arginase, which contributes to T cell suppression through consumption of arginine. In addition, the granulocytic subset of MDSCs suppresses T cells by producing reactive oxygen species and the monocytic subset by producing NO (8). The granulocytic population of MDSCs is usually more increased in tumor models (6), however the monocytic subset is thought to be the more suppressive, as this population is able to suppress both Ag-specific and nonspecific T cell proliferation (9), likely through nitrosylation of either the TCR (10) or MHC complex (11). Additionally, the monocytic subset of MDSCs has been demonstrated to differentiate into macrophages in vitro (6), possibly linking MDSCs and tumor-supporting macrophages.

The macrophage migration inhibitory factor (MIF) was identified in the 1960s as a soluble protein factor secreted by T cells with the ability to inhibit the random migration of macrophages (12, 13). MIF is an inflammatory cytokine that plays a critical role in numerous models of inflammatory disease (14, 15). In addition, MIF is overexpressed in many solid tumors (16–21). In many cases, the degree of MIF overexpression has been correlated with tumor progression and/or metastatic potential, with examples in prostate (22), hepatocellular (23), gastric (24), and lung cancers (25).

We identified MIF as a target for covalent modification and inhibition by a class of natural product compounds with well-established anti-cancer activities (26). MIF has an enzymatic keto-enol tautomerase activity, with the N-terminal proline as its catalytic center (27), and we demonstrated that these compounds covalently modify MIF on this proline and inhibit MIF tautomerase activity (26). No physiological tautomerase substrate has been identified, but the importance of the tautomerase to the biological function of MIF is supported by observations of other groups (28, 29) that have also characterized small-molecule inhibitors of MIF that inhibit both this enzymatic activity and the biological activities of MIF.

Given the observation that MIF overexpression correlates with tumor growth and progression, several groups have explored the potential contribution of MIF to growth and transformation of tumor cells. As a result, current models proposed to explain the tumorigenic activities of MIF have focused on functional interactions of MIF with cell signaling pathways, tumor suppressors, and oncogenes (30–32). However, as an inflammatory cytokine, we hypothesized that MIF would contribute to the interaction between the tumor and the host immune response. In this study, we provide evidence for a mechanism of MIF action in promoting tumor growth and metastasis that is unrelated to effects within the tumor cells. We demonstrate that the tumor-promoting effects of MIF require interaction with a fully competent immune system, and we characterize MIF-dependent modulation of monocytic MDSCs within the tumor microenvironment.

4T1 and CT26 cells (American Type Culture Collection) were cultured at 37°C in 5% CO₂ in RPMI 1640, supplemented with 10% FBS and 1% penicillin/streptomycin. Luciferase-expressing 4T1 lines (Caliper Life Sciences) were cultured with heat-inactivated FBS as recommended. Anti-MIF was from Invitrogen and anti-tubulin from Sigma. l-Dopachrome methyl ester and periodate were from Sigma, and sulforaphane was from LKT Labs.

Female 6- to 12-wk-old BALB/c mice and 6- to 8-wk-old SCID/bg mice were purchased from Charles River Laboratories. All animal experiments were performed with the approval of the University of Virginia Animal Care and Use Committee.

For MIF-depleted cell line, the sequences 5′-gatccATGCCtATGTTCATCGTgATTCAAGAGATcACGATGAACATaGGCATTTTTTTACGCGTg-3′ and 5′-aattcACGCGTAAAAAAATGCCtATGTTCATCGTgATCTCTTGAATcACGATGAACATaGGCATg-3′ were annealed and ligated into pSiren-RetroQ plasmid (Clontech) and used to make retroviruses. Parental 4T1 cells were infected with either the targeting hairpin or an empty vector and selected with 10 μg/ml puromycin. In some experiments, a line expressing a nontargeting RNA directed against the human MIF sequence was used in place of the empty vector control. For preparation of MIF-depleted CT26, parental CT26 cells were infected with targeting or nontargeting hairpin and selected with 10 μg/ml puromycin.

For the reconstituted cell lines, the coding region for wild-type (WT) human MIF or P2G MIF, in which the codon for the proline at position 2 was mutagenized to glycine [both resistant to the short hairpin RNA (shRNA) targeting the murine MIF sequences], was inserted into PQCXI-neo vector and used to produce retroviruses. MIF-depleted 4T1 MIF KD cells were infected with empty vector, mutant P2G MIF, or WT MIF expressing viruses and selected with 500 μg/ml neomycin.

Tautomerase activity was measured as previously described (33). Briefly, 4 mM dopachrome and 8 mM sodium periodate were mixed together in a 1:1 ratio and incubated for 5 min. This substrate was diluted 1:9 in tautomerase buffer (25 mM Na⁺PO₄, pH 6, 1 mM EDTA) and added to normalized cell lysates. Absorbance was monitored over time at 475 nm in a 96-well plate reader (Bio-Rad). The velocity is the rate of decolorization of the colorimetric substrate (measured as a decrease in absorbance at 475 nm) per unit time.

WT and MIF-depleted 4T1 or CT26 cells (5 × 10⁴) were plated in triplicate in 6-well plates and incubated at 37°C in 5% CO₂ for a total of 4 d. Cells were counted each day using a hemocytometer. For colony formation, 4 × 10⁴ 4T1 cells were suspended in media containing soft agar (Lonza) and incubated at 37°C for 2 wk. Photos were taken of three independent fields, and the number of colonies in each field was enumerated.

4T1 cells (1 × 10⁴) were injected into the mammary fat pad of 6- to 12-wk-old female BALB/c mice. CT26 cells (2 × 10⁴) were injected s.c. above the flank of 6- to 12-wk-old female BALB/c mice. Tumor volumes were estimated using the formula V = 0.4 × L × W² from two perpendicular measurements made with electronic calipers. In experiments with the reconstituted cell lines, tumors were removed at 18 d postimplantation. The mice were euthanized 8 d later for enumeration of metastasis.

Mice were treated with 200 μg sulforaphane (SFN) or saline vehicle by i.p. injection, starting the day of tumor injection and continuing daily until tumor harvest.

Tumors were dissected from the mammary fat pad and digested with 10,000 U collagenase I (Worthington Biochemical) for 60 min at 37°C. Cell suspensions were strained through a 70-μm membrane and then stained for flow cytometry.

Isolated lungs were digested with 5 mg collagenase IV (Worthington Biochemical) and 30 U elastase (Calbiochem) for 75 min at 4°C. Strained cell suspensions were plated in IMDM plus 10% FBS, 1% penicillin/streptomycin, containing 60 μM 6-thioguanine. Plates were incubated for at least 10 d at 37°C, fixed with methanol, and stained with crystal violet for enumeration of colonies.

To image lung metastasis in experiments with the luciferase-expressing cell lines, mice were injected i.p. with 100 μl/10 g body weight of a 15 mg/ml solution of luciferin 5 min prior to euthanasia. Lungs were excised and immersed in a 300 μg/ml solution of luciferin in PBS for imaging on the Xenogen IVIS System (Caliper Life Sciences).

For flow cytometry analysis, cells were stained with Abs for CD45 PerCP (BD Biosciences), F4/80 allophycocyanin–eFluor780 (eBioscience), Ly6G FITC (BioLegend), Ly6C allophycocyanin (BioLegend), and CD11b Pacific blue (Invitrogen), along with Live/Dead Fixable Dead Cell Stain (Invitrogen). Live cells were gated first using Live/Dead Fixable Dead Cells Stain (negative population), then forward and side scatter, followed by CD45⁺ and, finally CD11b⁺. A representative series of gated plots is shown in Supplemental Fig. 2A. The cells were analyzed with the Beckman Coulter CyAN ADP LX 9 Color Flow Cytometer.

CD8⁺ T cells were purified from a naive BALB/c mouse using CD8–PE Ab (BioLegend) and the EasySep PE kit (Stem Cell Technologies). MDSCs were purified from tumors by first depleting T cells with CD90.2 magnetic beads (Miltenyi) and subsequently purifying GR1⁺ cells with GR1–PE Ab (BioLegend) and the EasySep PE kit. T cells (5 × 10⁴ per well) were stimulated with anti-CD3 (Cedarlane) and anti-CD28 (eBioscience) at dilutions of 1:1000 each. MDSCs from either an MIF-containing or MIF-deficient tumor were added in 1:1 or 4:1 ratios. After a 72-h incubation, the culture was pulsed with [³H]thymidine and harvested 18 h later.

4T1 WT or knockdown cells (3 × 10⁵) were incubated in RPMI 1640 supplemented with 10% heat-inactivated FBS and 1% penicillin/streptomycin for 5 d to generate tumor cell conditioned media. For the differentiation assay, spleens from tumor-bearing mice were harvested and incubated with anti-CD11b beads (Miltenyi) and purified using the Automacs system. Purified CD11b⁺ cells (6 × 10⁶) were incubated in RPMI 1640 supplemented with 10% heat-inactivated FBS and 1% penicillin/streptomycin along with the conditioned media with and without SFN for 5 d before flow staining as described above. All cultures were supplemented with 10 ng/ml GM-CSF (R&D Systems) to ensure viability.

Mice were injected i.p. with 500 μg of either rat IgG (Invitrogen) or GR1 Ab (clone RB6-8C5) on days 11, 14, and 17 relative to tumor implantation.

All data are graphed as mean ± SEM. Groups were compared by unpaired two-tailed Student t test for Figs. 1 and 2 and by ANOVA for Figs. 3–7.

To address the role of MIF in tumor growth and progression and to test the hypothesis that MIF functions through impact on the immune tumor microenvironment, we used the spontaneously metastatic, syngeneic, and orthotopic 4T1 model of breast cancer (34). This model mimics stage IV breast cancer, with an organ preference hierarchy for spontaneous metastasis similar to human breast cancer (35). We generated an MIF inhibitory shRNA knockdown (KD) cell line and a vector control line from the parental 4T1 cell line. Effective MIF depletion was demonstrated by immunoblotting (Fig. 1A) and assay of tautomerase activity, measured as the rate of decolorization of a colorimetric substrate per unit time (Fig. 1B). Although MIF depletion has been reported to decrease the proliferative capacity of some cell lines (31) and increased expression of MIF has been shown to increase colony formation (30), we observed no difference in proliferation between the MIF-containing (vector) and MIF-depleted (KD) 4T1 cell lines (Fig. 1C). We repeated this experiment using a third control cell line expressing a nontargeting shRNA and again detected no differences in proliferation (data not shown). Because the serum used in the 4T1 cultures likely contains some bovine MIF, we repeated the proliferation assay in low serum conditions, and still the MIF-containing and MIF-depleted lines exhibited similar proliferation rates (data not shown). We compared the ability of the MIF-containing and MIF-depleted cell lines to grow in soft agar as a measure of anchorage-independent growth and observed no difference in either the number or size of the colonies (Fig. 1D). Together, these results demonstrate that MIF depletion does not impact in vitro growth properties of 4T1 cells.

To determine if MIF promotes tumor growth and/or metastasis, we implanted the MIF-containing or MIF KD 4T1 cells into the mammary fat pad of BALB/c mice. Both sets of mice developed tumors that were detectable and of similar size at approximately day 13. However, from 15 d after tumor implantation until harvest, the tumors from MIF-depleted KD 4T1 cells grew more slowly than those expressing MIF (Fig. 2A, left). Moreover, we detected fewer metastases in the lungs of mice bearing 4T1 MIF KD tumors than in the lungs of mice bearing MIF-containing tumors (Fig. 2A, right). This decrease in metastatic burden was not solely due to the difference in size of the primary tumors, as evidenced by comparing the metastasis as a function of tumor mass at the time of harvest. In this analysis, comparing tumors of similar mass, the MIF KD tumor–bearing mice still harbored far fewer metastases (data not shown).

To ensure that our observations were not unique to the 4T1 cell line, we generated a MIF-depleted CT26 murine colon carcinoma cell line, along with a control line expression a nontargeting shRNA. Effective MIF depletion was confirmed by immunoblot (Supplemental Fig. 1A). Similar to the results observed with 4T1 cells, depletion of MIF did not alter the in vitro proliferation of CT26 cells (Supplemental Fig. 1B). However, MIF depletion significantly impaired the growth of s.c. CT26 tumors in syngeneic BALB/c mice, despite the tumors appearing at similar sizes at early time points (Supplemental Fig. 1C). This suggests that similar to the 4T1 breast tumor model, MIF promotes CT26 tumor growth without influencing the in vitro growth properties of the cells.

Because MIF suppression did not alter cell growth in vitro but caused reduced tumor growth and metastasis in an intact animal, we hypothesized that the difference in tumor growth and metastasis observed in the BALB/c mice was due to an influence of MIF on the immune system. To test this, we implanted MIF-containing and MIF-depleted 4T1 cells expressing luciferase in the mammary fat pad of SCID/bg mice that lack T cells, B cells, and NK cells but have an intact myeloid compartment. In contrast to the results seen in the immunologically intact BALB/c mice, the resulting tumors grew at similar rates in the SCID/bg mice over the duration of the experiment (Fig. 2B, left). Moreover, ex vivo imaging of the metastatic burden in the lungs at the time of harvest demonstrated that the MIF-containing and MIF-depleted tumors were similarly metastatic in immunodeficient SCID/bg animals (Fig. 2B, right). These results suggest that the contribution of MIF to tumor growth and metastasis is due to influence over some aspect of the immune system.

To confirm that the effects observed in the MIF-depleted cells were due to loss of MIF, we reconstituted the KD cells with MIF using an expression plasmid encoding an shRNA resistant version of the MIF gene. In parallel, to test the importance of the tautomerase activity of MIF in promoting tumor growth and metastasis, we reconstituted the MIF KD cell line with a similar plasmid encoding a tautomerase-deficient MIF protein containing a proline to glycine substitution at residue 2 (P2G). Expression of the MIF protein in the two reconstituted cell lines was equivalent and similar to the parental cells (Fig. 3A). Tautomerase activity was restored in the WT-reconstituted line but remained absent in the P2G line, as the P2G mutant abolishes the tautomerase activity (Fig. 3B). Incipient tumors resulting from either of these cell lines were detected in the mammary glands of injected mice at the same time, but tumors derived from cells reconstituted with WT MIF grew at a faster rate (Fig. 3C) compared with those from the MIF-depleted line mock-reconstituted with an empty vector. Importantly, reconstitution with the tautomerase-defective P2G MIF did not restore the rapid tumor growth seen in WT MIF-expressing tumors. This suggests that the tautomerase activity of MIF is required for the tumor growth–promoting activity of MIF. Furthermore, the tautomerase activity was also important for promotion of metastasis by MIF, as reconstitution with WT MIF restored the ability of tumors to produce pulmonary metastases, whereas the P2G mutant MIF did not (Fig. 3D).

The dependence of MIF tumor-promoting activity on an intact immune system suggests that MIF does not directly impact an intrinsic property of the tumor cells but rather regulates an aspect of the tumor growth–promoting immune microenvironment. To identify MIF-dependent alterations in the immune cell infiltrates within the tumors, we examined the phenotype of cells prepared from the MIF-containing and MIF-depleted tumors using a panel of markers to detect MDSCs. Specifically, we examined the two major MDSC populations, defined as Ly6C^int/lowLy6G^high granulocytic MDSCs and Ly6C^highLy6G^low monocytic MDSCs (6). The CD11b⁺ myeloid cells in the WT MIF-reconstituted tumors contained a significantly larger proportion of monocytic MDSCs compared with the P2G- and vector-reconstituted tumors (Fig. 4A). No differences were observed in this subpopulation in either the blood or spleen of these animals (Supplemental Fig. 2B). Because MIF is named for its effects on macrophage migration, we also examined F4/80⁺ macrophage populations in the 4T1 tumors. Over the same three experiments in which we detected significant differences in the monocytic MDSC population, there was no difference in the GR1^−/low, F4/80⁺ macrophage population (Supplemental Fig. 2C). Therefore, MIF expression in the tumor cells acts specifically to increase the abundance of the monocytic population of MDSCs within the tumor. A reduction in monocytic MDSC abundance was also observed in the MIF-deficient CT26 tumors (Supplemental Fig. 1D). These results suggest that MIF may promote tumor growth and metastasis by enhancing the immune-suppressive properties of the tumor microenvironment through influence over MDSC populations.

The differences in MDSC abundance were not simply the result of differences in tumor size, as MDSC populations characterized in WT MIF-expressing tumors in SCID/bg mice also displayed a greater fraction of monocytic MDSCs than those from MIF-depleted or P2G-reconstituted tumors (Fig. 4B), despite all of the tumors being of similar size. Significantly, although the MIF-dependent differences in monocytic MDSC populations were observed in the immunocompromised mice, tumors that express MIF do not display accelerated tumor growth and metastasis in these mice (Fig. 2B). Therefore, although MIF is still able to influence the abundance of monocytic MDSCs in the SCID/bg hosts, in the absence of an anti-tumor immune response mediated by T cells and NK cells, there is no impact of MIF on tumor growth and metastasis.

Parallel to this genetic demonstration of the importance of the tautomerase activity of MIF in regulation of MDSC populations, we used SFN, a natural product inhibitor of MIF (26), to suppress pharmacologically the MIF tautomerase. Treatment of 4T1 tumor–bearing mice with SFN reduced the prevalence of monocytic MDSCs in tumors, similar to the effects of MIF depletion from tumors (Fig. 4C). Together with the failure of the tautomerase-deficient P2G mutant to rescue MDSC recruitment, this suggests that MIF controls the prevalence of monocytic MDSCs through a mechanism that requires its tautomerase activity.

Finally, to confirm that the MIF status of the tumor and the resultant shift in MDSC populations resulted in functional changes within the tumor microenvironment, we isolated GR1⁺ cells from MIF-containing and MIF-depleted tumors and evaluated their ability to suppress CD8⁺ T cell proliferation in vitro. The purified GR1⁺ cells isolated from MIF-expressing tumors suppressed T cell proliferation significantly more than those from the MIF-depleted tumors (Fig. 5) when comparing an equivalent number of cells. In fact, at a 1:1 ratio of MDSCs to T cells, the MDSCs isolated from the MIF-depleted tumor failed to suppress proliferation significantly, whereas the cells from the MIF-containing tumor were effective. These results demonstrate that the MIF-expressing tumors have a more highly suppressive MDSC infiltrate and therefore likely benefit from a more immunosuppressive microenvironment.

The MIF-dependent increase in prevalence of monocytic MDSCs in the 4T1 tumors could result from either increased recruitment of these cells into the tumor or differentiation of the MDSCs toward the monocytic phenotype, perhaps within the tumor. Others have reported that factors such as GM-CSF can control differentiation between the two subsets of MDSCs (36). Using a similar approach (diagrammed in Supplemental Fig. 3), we evaluated the effect of MIF on MDSC differentiation in vitro by flow cytometry. After purification of CD11b⁺ MDSCs from the spleen of a tumor-bearing mouse, the majority of cells were of the less suppressive granulocytic phenotype, with the monocytic subset composing <10% of the total population (Fig. 6A). Tumor-conditioned media from MIF-containing 4T1 cells induced differentiation of CD11b⁺ myeloid cells into monocytic MDSCs (Fig. 6B). Conditioned medium from MIF-depleted (KD) 4T1 cells resulted in a significantly reduced proportion of the CD11b⁺ cells differentiating into monocytic MDSCs. Cell viability in the cultures at the end of the incubation ranged between 25 and 50% (Supplemental Fig. 3B). Notably, there was no difference in cell viability between the cultures treated with the conditioned media from the MIF-containing versus the MIF-depleted 4T1 cells, suggesting that the differences seen in monocytic MDSCSs cannot be solely attributed to differences in survival. These observations suggest that either MIF itself or some MIF-induced secreted factor is responsible for the accumulation of monocytic MDSCs in this assay.

Addition of the MIF inhibitor SFN to the supernatant from the MIF-expressing 4T1 cells significantly decreased the induction of monocytic MDSCs (Fig. 6B). In contrast, SFN did not alter the induction of monocytic MDSCs in the cultures containing MIF KD supernatants. These results suggest that the differing prevalence of MDSCs in the MIF-containing versus MIF-depleted 4T1 tumors results from MIF-induced differentiation of myeloid cells in the microenvironment into the monocytic MDSC population. Inhibition of this process by SFN suggests that it occurs through a mechanism dependent on MIF tautomerase activity.

To verify that the contribution of MIF to tumor growth and metastasis was due to the reduction of MDSC abundance in the tumor, we depleted MDSC populations from mice and determined the effect on tumor growth and metastasis. In a pilot experiment, we demonstrated that i.p. delivery of three doses of GR1 Ab on days 11, 14, and 17 after tumor implantation resulted in a near complete depletion of MDSCs from the tumor when assayed on day 17 (Supplemental Fig. 4B). Subsequently, mice bearing either MIF-containing or MIF-deficient tumors were treated with anti-GR1 Ab to deplete MDSCs using the same dosing regimen (Supplemental Fig. 4A) with depletion initiated at day 11 postimplantation, before the presence of a measurable tumor. The experiment was harvested at day 20. Depletion of MDSCs in mice bearing MIF WT tumors resulted in impaired tumor growth (Fig. 7A) and inhibition of metastasis to lung (Fig. 7B). In contrast, depletion of MDSCs in mice bearing MIF KD tumors had no further impact on tumor growth (Fig. 7A) compared with MIF KD tumors in mice administered an isotype control Ab. Because MIF KD tumors fail to metastasize to lung (Fig. 7B), it was not possible to determine if MDSC depletion had any further impact on pulmonary metastasis in the absence of MIF. From this, we conclude that the contribution of MIF to 4T1 tumor growth and metastasis can be attributed to the effects of MIF in controlling MDSCs in the tumor microenvironment. Taken together, our results lead us to propose the model shown in Fig. 8.

Our results demonstrate a link between the proinflammatory cytokine MIF and the prevalence of monocytic MDSCs in the tumor microenvironment, leading to promotion of tumor growth and metastasis. This differs from previous reports that attribute the tumor growth–promoting effects of MIF to activities within the tumor cells (30–32, 37). Our work reveals a new mechanism for MIF as an immune modulator in the tumor microenvironment.

Significantly, we demonstrate that the tautomerase activity of MIF is required for promotion of tumor growth and metastasis. Using mutation of the catalytic proline and/or inhibition of MIF tautomerase with the enzyme inhibitor SFN, we further demonstrate that this enzymatic activity is important for inducing monocytic MDSCs both in tumors and in our in vitro differentiation assay. The importance of the MIF tautomerase activity has been debated (28, 38, 39). However, other inhibitors of the MIF tautomerase also influence downstream biological outcomes (28, 29, 40). Our current findings lend support to the conclusion that, although a physiological target has yet to be identified, the tautomerase activity is important for functions of MIF that contribute to its role in tumor growth and metastasis.

In this study, we observed MIF-dependent differences in the prevalence of monocytic MDSCs in the tumor microenvironment. As previously described, the monocytic MDSC subset is thought to have a higher capacity to suppress host anti-tumor response (9), ultimately supporting increased tumor growth and metastasis. Therefore, our demonstration that MIF contributes to the prevalence of these cells within the tumor suggests a possible mechanism through which MIF promotes tumor progression by influence over the tumor microenvironment.

Given that human cancers demonstrate increased MIF expression (17–21) and that MDSCs have been observed in the blood of human cancer patients (3–5, 41, 42), MIF is potentially a valuable therapeutic target for efforts aimed at inhibiting the tumor-supportive host immune response. Specifically, inhibition of MIF could block the accumulation of immune-suppressive MDSCs in the tumor microenvironment and restore effective anti-tumor immunity. This strategy, either alone or in combination with immunotherapy approaches designed to boost anti-tumor immune responses, offers an alternative to therapies that directly target the tumor cells. These approaches might be particularly useful in cancers that do not have recognized molecular therapeutic targets but that exhibit a strong inflammatory response or for those that are resistant to currently available targeted therapies. Finally, as tumors can recur and/or reappear as metastatic disease even years after primary tumor resection, our observation that MIF is important for spontaneous pulmonary metastasis suggests that MIF inhibition could represent a key approach not only for preventing primary tumor outgrowth but also for prevention of metastatic disease.

Our work not only furthers our understanding of the role of MIF in tumor progression but also suggests a previously unappreciated mechanism of action for cancer chemopreventives. We characterized the isothiocyanate chemopreventives, including SFN, as effective, covalent, natural product inhibitors of MIF (26). Epidemiological evidence and experimental models have long supported the conclusion that isothiocyanates exhibit anti-cancer activities (43–46). However, much of the function of isothiocyanates was thought to be through the detoxification of carcinogens (43–46), preventing the very earliest steps in tumor formation. Alternative models have focused on the ability of isothiocyanates to decrease proliferation of the cells (47, 48) or induce apoptosis (49, 50). Our study provides a novel molecular mechanism to explain isothiocyanate chemoprevention through inhibition of MIF and the resulting influence over the tumor microenvironment through specific impact on the immunosuppressive immune infiltrate.

We acknowledge the assistance of Joanne Lannigan and the Flow Cytometry Core Facility as well as Marya Dunlap-Brown and the Molecular Assessment and Preclinical Studies Core. We thank Angela Rose for technical assistance and Claudia Rival for advice on flow staining. Finally, we thank Sarah Conine, Denise Joplin, Carrie Leonard, Amy Bouton, Michael Gutknecht, and Scott Vande Pol for many helpful discussions and Thomas Braciale, Timothy Bender, Amy Bouton, Dan Conrad, Sheinei Saleem, and Scott Vande Pol for critical reading of the manuscript.

The authors have no financial conflicts of interest.