Testing Novel Pyrimidinyl Rexinoids: A New Paradigm for Evaluating Rexinoids for Cancer Prevention

Rexinoids are selective ligands for retinoid X receptors (RXR), which regulate the expression of numerous genes (1). Even though rexinoids were initially developed for metabolic disorders like diabetes (2), their important roles in proliferation, differentiation, and apoptosis are highly relevant to cancer (3, 4), making RXRs an attractive cancer target. The rexinoid bexarotene, is FDA approved for the treatment of cutaneous T-cell lymphoma. This drug has been tested in several clinical trials for lung cancer (5–9), and the combination of bexarotene (bex) and chemotherapeutic agents significantly improved median survival in a subset of patients with advanced non–small cell lung cancer (NSCLC; refs. 8, 10). Rexinoids are also effective for prevention in a variety of preclinical cancer models, including breast cancer (MMTV-neu mouse model, nitrosomethylurea NMU rat model), lung cancer (vinyl carbamate induced A/J mouse model, and a genetically engineered mouse model of lung cancer), pancreatic cancer (KPC mouse model), and other cancers (3, 11–18). Either as single agents or in combination with other drugs, rexinoids significantly delayed tumor development and reduced tumor burden in prevention protocols.

In spite of their profound effects on many cellular pathways with direct relevance for the pathogenesis of human diseases (16), including cancer, diabetes, atherosclerosis, and Alzheimer's disease, safe and effective rexinoids have not yet entered routine clinical practice, for either prevention or treatment of cancer or any other chronic diseases. Bexarotene has limited efficacy as a single agent. Moreover, triglyceride levels were elevated (5) and the thyroid axis was suppressed in patients treated with this drug (19). More potent and selective RXRs rexinoids have been developed, many with promising in vitro activity (20–24). LG100268 (LG268) is one of the most potent and selective of these newer rexinoids, with a 1,000-fold increase in selectivity for RXR binding compared with RAR binding (20), but elevated triglyceride levels and patent issues have prevented clinical development of this promising drug (3, 16). Although the known side effects of rexinoids can be tolerated for treatment of cancer, they do not meet the elevated degree of safety required for long-term prevention protocols in humans.

Therefore, there is still a great need to develop new rexinoids with greater potency and less toxicity. Here, we present data from both in vitro and in vivo studies of new rexinoids, the pyrimidine (Py) derivatives of Bex and LG268. These new rexinoids were tested for their ability to prevent lung cancer and reduce side effects in a widely used A/J mouse model. When challenged with vinyl carbamate, these mice develop Kras mutations and adenocarcinomas, so the model is clinically relevant for evaluating new drugs for lung cancer (25). As there is now abundant evidence that immunomodulatory and inflammatory effects play an essential role in carcinogenesis (26), we explored the immunomodulatory actions of rexinoids during lung carcinogenesis and their effects on macrophage polarization. We also screened 10 new rexinoids for their ability to suppress induction of the proinflammatory enzyme iNOS, an assay that correlates with suppression of carcinogenesis in this lung cancer model. In addition, their activation of RXR, as well as SREBP (sterol regulatory element-binding protein), a transcription factor involved in triglyceride synthesis, was used in initial screens to evaluate potency and toxicity, respectively. Upon validation, this set of in vitro assays will be used to guide the synthesis of additional compounds to generate drugs that are more effective and/or less toxic than existing rexinoids.

Rexinoids were synthesized at Arizona State University (Glendale, AZ) as described previously (21, 22). For in vitro assays, drugs were dissolved in DMSO to make 10 mmol/L stock concentrations and then diluted in the appropriate cell culture media (described below) to generate the final working concentrations listed in each figure or figure legend. Controls containing equivalent concentrations of DMSO were included in all experiments. For in vivo testing, drugs were incorporated in diet (40 mg/kg diet and 80 mg/kg diet) as described previously (14). In brief, rexinoids were dissolved in 50 mL vehicle [1 part ethanol: 3 parts Neobee Oil, (Thermo Fisher Scientific), a highly purified coconut oil triglyceride used for formulation of drugs given to humans] per kg of diet and then mixed into powdered AIN-93M diet (BioServ) for 20 minutes using a commercial (KitchenAid) food mixer to assure homogeneity of drug in diet. Rexinoids are soluble in this vehicle, and providing drugs in diet yields better bioavailability and steady-state drug levels than giving suspensions of drug as a bolus by gavage. The diets were stored in the cold room at 4°C for up to 4 weeks; we have confirmed the stability of multiple rexinoids in diet for this length of time by liquid chromatography–mass spectrometry.

RAW264.7 macrophage–like cells were purchased from the ATCC and cultured in DMEM with 10% FBS and 1% penicillin/streptomycin. HepG2 cells were purchased from the ATCC and cultured in RPMI1640 media with 10% FBS and 1% penicillin/streptomycin. Media and supplements were purchased from Corning Cellgro (Mediatech). Bone marrow–derived macrophages (BMDM) were isolated from the femurs of adult C57BL/6 mice as described previously (27). Femurs were dissected out of the mice, and all remaining tissues removed before using scissors to cut the proximal and distal ends. The bone marrow was flushed out of the femur with 5 mL of RPMI using a 25G needle and syringe. All samples from a mouse were combined and resuspended in RPMI + 10% FBS and 1% penicillin/streptomycin.

RAW264.7 cells were plated in 96-well plates (20,000 cells/well) and after overnight attachment, cells were treated with various concentrations of rexinoids (0–1,000 nmol/L) and then stimulated with 1–2 ng/mL LPS (dissolved in saline) for 24 hours. NO production was measured using the Griess reaction as described previously (14).

RAW264.7 cells were treated with 300 nmol/L rexinoid and stimulated with 1 ng/mL LPS for 24 hours. Bone marrow–derived macrophages (BMDM) were cultured in RPMI1640 media supplemented with 10% FBS and 20 ng/mL M-CSF (BioLegend) for 7 days to induce an M2 phenotype (27). Then BMDMs were treated with 100 nmol/L rexinoids and stimulated with 100 ng/mL LPS for 24 hours. Total RNA was isolated with TRIzol (Invitrogen). Two micrograms of RNA was used to synthesize cDNA using the SuperScript III Reverse Transcriptase Kit (Invitrogen). Validated IL6 (PPM03015A), CCL9 (PPM02957F), IL1β (PPM03109F), TNFα (PPM03113G), IL10 (PPM03017C), and β-actin (PPM02945B) primers were purchased from Qiagen. iQ SYBR Green Supermix (Bio-Rad Laboratories) and the ABI 7500 FAST Real-Time PCR system were used to detect gene expression [95°C for 10 minutes followed by 40 cycles of 95°C for 15 seconds (melt) and 60°C (extend/anneal) for 1 minute]. The recommended protocol supplied by the manufacturer of each kit was followed for all experiments. Relative expression was calculated by the ΔΔC_t method (14).

All animal studies were approved by the Institutional Animal Care and Use Committee (IACUC) at Michigan State University. Eight-week-old female A/J mice (Jackson Laboratories, average weight = 20 g) were injected intraperitoneally (16 mg/kg body weight) with vinyl carbamate (Toronto Research Chemicals). A 1.6 mg/mL stock solution of vinyl carbamate (dissolved in isotonic saline) was prepared, and approximately 200 μL injected into each mouse. One week after injection, the mice were randomized into the control group–fed AIN-93G diet (BioServ) mixed with vehicle (1 part ethanol: 3 parts Neobee oil; 50 mL vehicle/kg diet) or fed rexinoids dissolved in the same vehicle and mixed into the diet (14). After 16 weeks on diet, mice were euthanized and lungs were inflated with PBS. The entire left lobe of the lungs was fixed in neutral buffered formalin (NBF) for histopathology. Right lungs were used immediately for flow cytometry (the superior and middle lobes) or were flash frozen (the other two lobes: inferior and post-caval). The tumor number, size, and histopathology were assessed on two separate sections of the left lung by two independent investigators. Sections were step sectioned (200-μm apart starting at the medial hilar section) and the slides stained with hematoxylin and eosin. The samples were coded with random numbers and then randomized before being read, thus blinding the investigators to the treatment group. The histopathology classifications were based on published criteria established by a lung pathologist (28).

The same two lobes (superior and middle) of the right lung were harvested from A/J mice for flow cytometry. Freshly harvested lung tissue was chopped and incubated in digestion media (DMEM with 10% FBS and 1% penicillin/streptomycin) containing collagenase (300 U/mL, Sigma), dispase (1 U/mL, Worthington), and DNAse (2 U/mL, Calbiochem) for 30 minutes at 37°C. Cells were then passed through a 40-μm cell strainer (BD Falcon). Lysing solution (eBioscience) was used to eliminate red blood cells. Single-cell suspensions were stained with 5 μg/mL anti-mouse Fc block (eBioscience) and two optimized panels of validated antibodies for 30 minutes at 4°C. Panel 1: CD45-VioGreen (Miltenyi Biotec, clone 30F11, 3 μg/mL), Gr-1-PE (Miltenyi Biotec, clone RB6-8C5, 3 μg/mL), CD11b-FITC (Miltenyi Biotec, clone M1/70, 3 μg/mL). Panel 2: CD45-VioGreen (Miltenyi Biotec, clone 30F11, 3 μg/mL), CD4-FITC (Miltenyi Biotec, clone GK1.5, 3 μg/mL), CD3-PE (BioLegend, clone 145-2C11, 2 μg/mL), CD8-PerCP/Cy5.5 (BioLegend, clone 53-6.7, 2 μg/mL). Flow cytometry was performed using a LSR II flow cytometer with DIVA 6.2 software (BD Biosciences) and three laser sources (488, 633, and 407 nm). Data analysis was done using FlowJo x.10.0.7r2 software (Tree Star).

The entire lungs were inflated with PBS and then the entire left lobe of each lung was separated, inflated with PBS, and fixed in 10% NBF for 48 hours. Lungs were then step-sectioned for histopathology and IHC analysis. Citrate buffer (Vector Laboratories, catalog no: H3300) was used for antigen retrieval. Slides immersed in the buffer were microwaved to the boiling temperature and then kept around 90°C by microwaving another 3–5 seconds every 15–20 seconds for a total of 20 minutes. Endogenous peroxidase activity was quenched in 3% hydrogen peroxide for 10 minutes. Sections were immunostained with CD45 (1:100, BioScience) and a biotinylated anti-rat secondary antibody (Vector Laboratories), or F4/80 (1:50, Invitrogen) antibodies and a biotinylated anti-rat secondary antibody (Vector). Signal was amplified by Vectastain ABC (Vector Laboratories) and detected using a DAB Kit (Cell Signaling Technology). Sections were counterstained with hematoxylin (Vector Laboratories). Samples were coded, randomized, and blinded as described above. Negative controls without primary antibody were also done to assure the quality and specificity of the primary antibodies (Supplementary Fig. S1).

HepG2 cells treated with rexinoids were lysed in RIPA buffer (1 mol/L Tris-Cl, 5 mol/L NaCl, pH 7.4, 0.5 mol/L EDTA, 25 mmol/L deoxycholic acid, 1% Triton-X, 0.1% SDS) with protease inhibitors (1 mmol/L phenylmethylsulfonylfluoride, 2 μg/mL aprotinin, and 5 μg/mL leupeptin). The BCA assay (Sigma-Aldrich) was used to determine protein concentrations. Proteins (20 μg/well) were separated by 10% SDS-PAGE gels and transferred to nitrocellulose membranes. SREBP-1c (Active Motif, 1:1,000) and vinculin (Cell Signaling Technology, 1:4,000) primary antibodies were used to detect the corresponding proteins. Secondary antibodies (anti-rabbit or anti-mouse antibodies conjugated to horseradish peroxidase) were purchased from Cell Signaling Technology. Signal was detected using the ECL Western blotting substrate (14). Images shown are representative of three independent experiments. Protein expression levels were quantified by ImageJ.

Sections of livers were homogenized in 5% NP-40/ddH₂O (50 mg tissue/1 mL solution), and total triglyceride levels in liver and plasma were evaluated using a Triglyceride Quantification Assay Kit from Abcam. Cholesterol levels in plasma were measured using a Cholesterol Quantification Kit from Sigma-Aldrich. The recommended protocol supplied by the manufacturer of each kit was followed.

New rexinoids were screened using the iNOS suppression assay described above in addition to RXR (EC₅₀) and SREBP activation. For RXR activation screening, full dose–response curves were generated with ligand concentrations ranging from 1 × 10⁻⁹ to 0.3 × 10⁻⁵ mol/L in transfected HCT-116 cells (male Homo sapiens colorectal carcinoma epithelium) using an RXR mammalian two-hybrid system. Although not lung or liver cancer cells, these cells can be transfected with a variety of plasmids and are thus useful for screening. HCT-116 cells were plated overnight at 80,000 cells/well in 24-well plates and maintained in DMEM/high glucose (Hyclone) enhanced with 10% FBS (Invitrogen), 1 mmol/L sodium pyruvate (Invitrogen), 100 μg/mL streptomycin, and 100 U/mL penicillin. The cells were cotransfected using a human RXR-binding domain (BD) vector, a human RXR activation domain (AD) vector, a luciferase reporter gene containing BD-binding sites, and Renilla control plasmid. Transfection was achieved via 2 μL/well of Express-IN transfection reagent (Thermo Fisher Scientific), which was allowed to incubate for 24 hours with the cells. The cells were subsequently treated with ethanol vehicle (0.1%) or analogues (1.0, 2.5, 5.0, 7.5, 10, 25, 50, 75, 100, 250, 500 nmol/L, 1, 2, 3 μmol/L) and incubated for 24 hours. The amount of rexinoid activity at each concentration was measured by luciferase output utilizing a dual-luciferase reporter assay system according to the manufacturer's protocol (Promega) in a Sirus luminometer (Berthold Detection System). Three independent assays were conducted with triplicate samples for each treatment group. The EC₅₀ values were derived from dose–response curves of ligand concentration versus normalized luciferase activity.

For SREBP-based screening, HCT-116 cells were maintained as above, followed by cotransfection of 250 ng of the pBP1c(6500)-Luc reporter gene, which contains an LXRE in the context of about 6,500 base pairs of flanking DNA from the mouse SREBP-1c natural promoter (29) along with 50 ng of CMX-hLXRα, 50 ng of pSG5-hRXRα and 20 ng of the Renilla control plasmid. The transfection was initiated with 2 μL/well of Express-IN transfection reagent (Thermo Fisher Scientific) used for liposome-mediated DNA delivery for 18 hours. The cells were then incubated for 24 hours posttransfection with ethanol vehicle, 10⁻⁷ mol/L T0901317 (LXR ligand), or 10⁻⁷ mol/L bexarotene or analogues. After a 24-hour incubation period, the amount of SREBP promoter activity was measured by luciferase output utilizing a dual-luciferase reporter assay system according to the manufacturer's protocol (Promega) in a Sirius luminometer (Berthold Detection System). Data were expressed as the percentage of T0901317-induced SREPB activation. Three independent assays were conducted with triplicate samples for each treatment group.

The in vitro experiments were performed in triplicate samples for each concentration of drug, and independent experiments were repeated at least three times. Results were expressed as the mean ± SD or mean ± SEM as indicated in each specific figure or table. For the in vitro and in vivo experiments, results were analyzed using a two-tailed t test when only two groups were compared; when more than two groups were compared, data were analyzed using one-way ANOVA followed by a Tukey test for multiple comparisons (to control the type I error) if the data fit a normal distribution; the Kruskal–Wallis one-way ANOVA on ranks was used followed by the Dunn test for multiple comparisons if the data did not fit a normal distribution (SigmaStat 3.5). For the histopathology, McNemar's Z test was used to compare proportions. P < 0.05 was considered statistically significant. A Benjamini–Hochberg procedure with false discovery rate of 25% was performed for the in vivo study (including tumor number, tumor size, tumor burden, liver weight and cholesterol levels in the plasma); all the P values with this procedure were smaller than the original calculated P value.

In an effort to develop more potent and selective rexinoids, pyrimidine derivatives of LG268 (PyLG268) and bexarotene (PyBex) were synthesized (Fig. 1A). These two derivatives were tested in vivo because they were similar in potency to Bex but gave higher plasma concentrations at C_max than bexarotene and produced differential gene expression in Sprague Dawley rats (30). In the current experiments, A/J mice were injected intraperitoneally with vinyl carbamate. As confirmed in our model, vinyl carbamate induced a mutation in codon 61 (Supplementary Fig. S2A) of the Kras gene as well as additional mutations across all of the chromosomes (Supplementary Fig. S2B; Supplementary Table S1).

One week after initiation, the mice were fed control diet or rexinoids in diet (40–80 mg/kg diet) for 16 weeks. Tumor number, size, and burden were then evaluated, and Table 1 summarizes these results. Despite a trend toward lower tumor numbers in the rexinoid-treated mice (Fig. 1B; Table 1), only high dose LG268 (80 mg/kg diet) significantly (P < 0.05) reduced the average number of tumors, by 61%. All of the high dose (80 mg/kg diet) rexinoid groups significantly (P < 0.05) reduced the average size of the lung tumors (Table 1) by 49%–69% (range 0.07 ± 0.01 mm³ to 0.12 ± 0.02 mm³ vs. controls 0.24 ± 0.09 mm³).

The average tumor burden on lung sections was markedly reduced by PyBex and PyLG268 by 59% and 53% (40 mg/kg diet doses) compared with the controls (from 0.47 ± 0.19 mm³ in the control group to 0.19 ± 0.05 mm³ in the group treated with PyBex and 0.22 ± 0.05 mm³ in the PyLG268-treated group). In contrast, average tumor burden was 0.35 ± 0.08 mm³ in the group treated with the low dose (40 mg/kg diet) of bexarotene, a reduction of only 26%. Again, LG268 was the best rexinoid and reduced average tumor burden by 88%, to 0.06 ± 0.02 mm³ versus 0.47 ± 0.19 mm³ in the control group.

Although the effects on tumor number and size were not changed with the low doses of rexinoids, the severity of the lesions was reduced. Notably, the percentage of high-grade tumors (HH, both histologic and nuclear characteristics such as tumefactive fused trabecular architecture, distinct nucleoli, and conspicuous mitoses, as shown Fig. 1C) were significantly (P < 0.05) higher (Fig. 1D) in the mice fed bexarotene (63%, 40 mg/kg diet) than in the control group (38%) or in the group treated with PyBex (34%, 40 mg/kg diet). While the increase in more advanced tumors (HH) was no longer found in the group fed the higher dose (80 mg/kg diet) of bexarotene (40%), the percentage of advanced tumors was still higher than any of the other groups (30% PyBex, 28% LG268, and 24% PyLG268) (Table 1).

Because of the known lipidemic properties of rexinoids, we measured lipid levels in the A/J mice. All the mice were weighed weekly, and the liver was weighed at the end of the study. The rexinoids were well-tolerated and did not cause weight loss, as there were no significant changes in weight across any of the groups throughout the study (Supplementary Fig. S3). As fatty livers and hepatomegaly are commonly seen in mice treated for prolonged periods with LG268, we normalized liver weights to body weight and compared groups (Fig. 1E). All four rexinoids significantly (P < 0.05) increased the average liver weight (5.5%–9.4%) compared with control mice (4.8%). However, the increase in liver weight observed with PyLG268 (6.2%) was significantly (P < 0.05) reduced compared with LG268 (8.6%). There were no significant differences in liver weights between mice fed bexarotene and PyBex, but liver weights in these groups were significantly (P < 0.05) lower than LG268.

In addition, triglyceride levels in the liver and plasma and cholesterol levels in the plasma were measured (14). PyBex (40 mg/kg diet) reduced triglyceride levels (7.17 ± 0.92 μg/kg tissue) in the liver compared with bexarotene (40 mg/kg diet, 8.2 ± 0.93 μg/kg tissue) and plasma (3.6 ± 0.29 mg/mL vs. 4.8 ± 0.24 mg/mL, respectively, P < 0.05). In the control group, the triglyceride level was 4.75 ± 0.95 μg/kg tissue in the liver and 4.90 ± 0.65 mg/mL in the plasma. PyLG268 significantly reduced cholesterol levels (Fig. 1F) in the plasma compared with LG268 (1.6 ± 0.28 μg/μL vs. 2.5 ± 0.68 μg/μL, respectively; P < 0.05).

Most rexinoids, at least as single agents, have limited activity for inhibiting proliferation or inducing apoptosis of epithelial cancer cells in vitro. Although we and others have shown significant in vitro reduction of inflammatory cytokines and pathways by the rexinoids, the effects of rexinoids on immune cells in vivo in cancer studies have not been extensively investigated. Thus, we examined the modulation of immune cells by the two best rexinoids here. We harvested the superior and middle lobes of the right lung to characterize immune cell populations by flow cytometry. We evaluated levels of total immune cell populations (CD45⁺), macrophages (CD45⁺, CD11b⁺, Gr-1⁻), myeloid-derived suppressor cells (MDSCs, CD45⁺, CD11b⁺, Gr-1⁺), CD4 Th cells (CD45⁺, CD3⁺, CD4⁺), and CD8 cytotoxic T cells (CD45⁺, CD3⁺, CD8⁺), which are all very relevant to the pathogenesis and prognosis of lung cancer. MDSCs and tumor-associated macrophages are key players in immunosuppression and high infiltration of these cells is associated with poor prognosis (31, 32). CD8 T cells are essential for killing tumor cells and for responding to immunotherapies like anti-PD-1/PD-L1 (33). There was no increase of CD8 T cells with either LG268 or PyLG268 treatment. However, PyLG268, but not LG268, significantly increased total immune cell (CD45⁺) infiltration into the lung (Fig. 2A). PyLG268 also significantly decreased the tumor-associated macrophages and MDSCs in the lung (Fig. 2B and C, respectively). Changes in macrophage and MDSC populations were not significantly different in the LG268 groups, except for a decrease in macrophages in the high-dose group (Fig. 2B). The changes in the CD45⁺ populations in lung parenchyma and tumor-associated macrophages near tumors were confirmed by IHC (Fig. 2D).

In addition to the immunomodulatory effects of LG268 and PyLG268 on the percentage of tumor-associated macrophages (TAM) in the lung, we further investigated the effects of these two rexinoids on macrophage polarization in vitro. Macrophages with M1 and M2 phenotypes have opposite effects on lung cancer progression (34). M1 (immunostimulatory) macrophages enhance antigen presentation and induce cell death of tumor cells, while M2 (immunosuppressive) macrophages promote tumor growth and invasion. Skewing the polarization of macrophages from an M2 to an M1 phenotype has been a promising strategy for cancer therapy. Here, we isolated primary bone marrow–derived macrophages (BMDM) from C57/BL6 wild-type mice and skewed them to an M2 phenotype by treating with M-CSF for 7 days. After differentiation, we treated the BMDMs with LG268 or PyLG268 and stimulated with LPS for 24 hours. Both LG268 and PyLG268 significantly (P < 0.05 vs. control) increased the production of M1 cytokines (TNFα and IL1β, Fig. 2E) and decreased the production of an M2 cytokine (IL10, Fig. 2E). Furthermore, PyLG268 was significantly (P < 0.05) more potent than LG268 in skewing the polarization of BMDMs. These experiments suggest that rexinoids, especially PyLG268, can reprogram the differentiation of macrophages from an M2 to an M1 phenotype.

Inflammation plays critical roles in tumor development (35). We have previously shown that the rexinoids LG100268, LG101506, and NRX194204 modulate the production of inflammatory cytokines and pathways induced by LPS in RAW264.7 macrophage-like cells (14, 36). To characterize the anti-inflammatory effects of new pyrimidine rexinoids, we first performed the in vitro iNOS suppression assay (11). RAW264.7 cells were treated with 10 nmol/L or 100 nmol/L rexinoids, challenged with 2 ng/mL LPS for 24 hours, and NO was measured in the media. LG268 remained the most potent rexinoid for inhibiting the induction of NO and reducing lung carcinogenesis. PyLG268 is comparable in potency to LG268, and is significantly (P < 0.05) more potent than bexarotene. PyBex is significantly (P < 0.05) more potent for inhibiting iNOS than bexarotene, the parent compound (Fig. 3A). Notably, a correlation between in vitro efficacy in the iNOS assay and in vivo efficacy in the lung cancer model is observed with other published rexinoids. When the same concentration of five known rexinoids was directly compared in a separate iNOS assay, the rank order was the same for iNOS suppression and reduction in tumor burden in the vinyl carbamate–induced lung cancer model (14, 28). LG268 was the most effective rexinoid in both assays (100% ± 4% iNOS suppression at 10 nmol/L; 69% reduction in tumor burden), followed in order of potency by NRX194204 (98% ± 6% iNOS suppression; 64% reduction in tumor burden), PyLG268 (48% ± 10% iNOS suppression; 53% reduction in tumor burden), LG101506 (45% ± 8% iNOS suppression; 50% reduction in tumor burden), and bexarotene (31% ± 12% iNOS suppression; 26% reduction in tumor burden).

Next, the expression of a series of proinflammatory cytokines was compared by qPCR analysis. Total RNA was isolated from RAW264.7 cells treated with 300 nmol/L rexinoid and stimulated with 1 ng/mL LPS. CCL9 is a known downstream target of RXRs and plays an important role in leukocyte recruitment (37); rexinoids suppressed CCL9 production (Fig. 3B) in a similar manner as in the NO assay. LG268, PyLG268, and PyBex were significantly (P < 0.05) better for reducing CCL9 expression than bexarotene. IL6 and IL1β are not direct targets of RXR, but all four rexinoids almost completely abolished production of mRNA for these cytokines (Fig. 3C and D).

The undesirable toxicities of rexinoids are a consequence of activation of heterodimers of RXR with other nuclear receptors, such as LXR, RAR, and PPARγ. Activation of the LXR–RXR complex is usually associated with elevated plasma triglycerides due to the upregulation of SREBP-1c, a master regulator of cellular lipid metabolism and homeostasis (38). Here, we evaluated the predictive value of SREBP elevation as a marker of in vivo toxicity in our lung cancer model. HepG2 cells were treated with the four rexinoids, and expression of SREBP mRNA and protein levels were detected. LG268 induced the highest levels of SREBP mRNA (Fig. 4A) and protein (Fig. 4B). Consistent with the in vivo results, PyLG268 significantly (P < 0.05) reduced the expression of SREBP mRNA compared with LG268, even though the SREBP expression was still elevated compared with the control (Fig. 4A). Bexarotene and PyBex also increased SREBP mRNA expression compared with the control, but neither was as promising as PyLG268. In addition to mRNA and protein expression, the effects of rexinoids on transcriptional activity of SREBP were measured. A luciferase reporter gene containing the LXRE sequence from the mouse SREBP-1c promoter was transfected into HCT-116 cells. T0901317 (T0, a LXR ligand) was used as a positive control and results were normalized to induction with T0. All four rexinoids induced luciferase activity, but again PyLG268 was less lipogenic than LG268 (Fig. 4C).

With the in vivo and in vitro data supporting the value of the iNOS and SREBP assays, we then screened a series of 10 new rexinoids in three in vitro assays. Results are summarized in Table 2. Several of the new rexinoids inhibited NO secretion in the iNOS assay comparably to LG268, previously our most potent rexinoid, and some of them (compounds 6, 8, 12, 13) were even more potent than LG268. RXR activation was better than bexarotene (EC₅₀ = 55 nmol/L) for almost all of the new compounds. Induction of SREBP, a transcription factor that induces enzymes of triglyceride synthesis, was used as a biomarker of toxicity. SREBP was induced with T0901317, a LXR ligand and known SREPB activator, and data expressed as the percentage of T0901317-induced SREPB activation. SREBP activation was lower with compounds 12, 11, 7, and 10 compared with LG268 and bexarotene. On the basis of all of the in vitro screening assays listed in Table 2, compounds 7, 11, and 12 appear to be the most promising new rexinoids, with potentially lower toxicity and higher potency. To test our predictions, these compounds will be screened in our lung carcinogenesis assay. Once validated, these screening assays will be used to guide chemical synthesis and decision making for future in vivo testing.

In this study, two newly synthesized pyrimidine derivatives of bexarotene (PyBex) and LG268 (PyLG28) were evaluated in a preclinical lung cancer model. Both PyBex and PyLG268 were more potent than bexarotene for inhibiting lung carcinogenesis, particularly in reducing tumor burden. The reduction in the severity of the lesions also suggests that rexinoids can prevent tumor progression. In addition, in terms of toxicity, a better toxicity profile was observed with pyrimidine derivatives compared with their parent compound, with reduced hepatomegaly and lipids levels (triglycerides and cholesterol). This study supplied direct evidence showing the feasibility of using synthetic chemistry to produce superior rexinoids than bexarotene, the FDA-approved rexinoid.

The lung cancer model we used in this study is highly clinically relevant. A/J mice are widely used to evaluate chemopreventive agents. Although these mice can develop spontaneous adenomas in the lung, injection of a variety of carcinogens including urethane, vinyl carbamate, benzo(a)pyrene, and 4-(methylnitrosamino)-1-(3-pyridyl)-1-butanone (NNK) accelerates carcinogenesis (39–41). Vinyl carbamate induces Kras mutations and lung adenocarcinomas in contrast to the adenomas induced by urethane and NNK (42); there are currently no effective treatments for tumors driven by Kras mutations. Adenocarcinomas are the most common type of lung cancer and Kras mutations are found in 30% of lung adenocarcinomas (43), most frequently in smokers and Caucasian patients (44), making the A/J mouse model particularly relevant (25). In addition, treatment with rexinoids was started post-initiation, in contrast to evaluation of many chemopreventive agents, which are administered before initiation with a carcinogen.

We have demonstrated immunomodulatory effects of rexinoids that are highly relevant in lung carcinogenesis, as increasing evidence supports the importance of the immune system in cancer progression (26). The potent anti-inflammatory effects of rexinoids also support the important roles of these drugs on immune cells in the tumor microenvironment. PyLG268 decreased macrophage and myeloid-derived suppressor cells, which is correlated with better clinical outcomes (32, 45). Although the effects on macrophages were observed at both doses, the effect on myeloid-derived suppressor cells was only observed in the 40 mg/kg diet group, suggesting a concentration-dependent effect or a different mechanism in some immune populations. Future studies will address these questions. PyLG268 is also more potent than LG268 in skewing the polarization of macrophages from an M2 to an M1 phenotype. Future studies will examine the functional consequences of these changes in immune cells in lung carcinogenesis as well as examine newly synthesized rexinoids. The incorporation of immunomodulatory and anti-inflammatory assays into the paradigm for evaluating new chemopreventive agents should help to further the practical development of new drugs for preventing cancer.

More importantly, we are not only reporting two new rexinoids, we also are developing a valuable in vitro screening system to guide our future rexinoid program for synthesizing superior rexinoids. On the basis of our years of experience on rexinoid development, a correlation between in vitro anti-inflammatory efficacy and in vivo chemopreventive efficacy was postulated and tested here. Combined with the activity of RXR activation in vitro, inhibition of iNOS may help us better predict the efficacy in vivo. Moreover, we also evaluated SREBP-1c expression and activity in vitro to determine whether it could serve as a biomarker for predicting liver toxicity in vivo. As shown in Table 2, we have established these in vitro assays in a relatively high throughput way to rank the efficacy and toxicity of new rexinoids. Selection of the best compounds across our three in vitro screening assays will be tested in vivo in the lung cancer model for both efficacy (reduced tumor burden) and safety (reduced liver weight, triglycerides, and cholesterol levels). If validated with additional compounds, this screening system provides us a more effective and efficient way to develop better rexinoids.

More and more studies support the potential clinical use of rexinoids for both cancer prevention and treatment, and RXR remains an attractive therapeutic target. We and other groups are making great efforts to push better rexinoids into the clinic. In addition to new efforts in synthetic chemistry, other strategies can be pursued. One way is changing the delivery system. Zhang and colleagues reported that delivery of aerosolized bexarotene inhibits lung tumorigenesis without increasing plasma triglyceride and cholesterol levels (18), but this approach would be limited to lung cancer and other diseases of the lung. Another strategy is to utilize drug combinations to enhance efficacy and reduce side effects by using lower doses of complementary drugs. Previous work in many laboratories has shown notable effects with the combination of a rexinoid and other drugs (3, 16). Lowering the dosing or using intermittent dosing (46, 47) can help balance between efficacy and toxicity. One more possibility is to design new rexinoids with appropriate interactions with coactivators and corepressors, as has been successful for synthesis of new SERMs and other ligands targeting the nuclear receptor superfamily. With these strategies and better understanding of the mechanisms of rexinoids in cancer, there is great potential that the eventual goal of a clinically useful rexinoid will be achieved. Evaluation of the significant anti-inflammatory and immunomodulatory effects of rexinoids offers a new paradigm for such development.

No potential conflicts of interest were disclosed.

Conception and design: D. Zhang, P.W. Jurutka, M.B. Sporn, C.E. Wagner, K.T. Liby

Development of methodology: D. Zhang, K.T. Liby

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. Zhang, A.S. Leal, S. Carapellucci, P.H. Shahani, J.S. Bhogal, S. Ibrahim, S. Raban, P.W. Jurutka, P.A. Marshall

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Zhang, A.S. Leal, S. Ibrahim, P.W. Jurutka, P.A. Marshall, K.T. Liby

Writing, review, and/or revision of the manuscript: D. Zhang, P.W. Jurutka, P.A. Marshall, M.B. Sporn, C.E. Wagner, K.T. Liby

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Ibrahim, M.B. Sporn

Study supervision: K.T. Liby

Other (synthesis of novel, tested compounds): C.E. Wagner

We thank Kayla Zydeck and Nicole Chaaban for their assistance with the lung carcinogenesis studies. The flow cytometry analysis was done in the Flow Cytometry Core at Michigan State University, and we thank Dr. King, Director of the MSU core, for his assistance and members of Dr. Rockwell's lab at MSU for their help with sample preparation for flow cytometry. Martine Cao completed preliminary studies with PyLG268, and Jonathan Rennhack assisted with the analysis of the whole exome sequencing. This work was supported by Michigan State University (Startup funds, to K.T. Liby) and the Audrey and Theodor Geisel School of Medicine at Dartmouth (a research grant for developing new rexinoids, to K.T. Liby). Patent applications covering the new compounds described in this work have been applied for on behalf of Arizona Board of Regents (C.E. Wagner, P.W. Jurutka, P.A. Marshall).

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.