A comprehensive data mining study shows that most nuclear receptors act as newly proposed homeostasis-associated molecular pattern receptors

An experimental data mining strategy (Fig. 1) was used to analyze the expression profiles of mRNA transcripts of NR genes in 21 different human and 17 mouse tissues including the heart and vasculature. We utilized an experimentally verified mRNA expression in the expressed sequence tag (EST) databases of the National Institutes of Health (NIH)/National Center of Biotechnology Information (NCBI) UniGene (http://​www.​ncbi.​nlm.​nih.​gov/​sites/​entrez?​db=​unigene) to determine the transcription profile of nuclear receptors in tissues of interest. Transcripts per million of genes of interest were normalized to that of housekeeping gene β-actin in each given tissue to calculate the arbitrary units of gene expression. A confidence interval of the expression variation of housekeeping genes was generated by calculating the mean plus two times that of the standard deviation of the arbitrary units of three randomly selected housekeeping genes (PRS27A, GADPH, and ARHGDIA in human; Ldha, Nono, and Rpl32 in mouse) normalized by β-actin in the given tissues. If the expression variation of a given gene in the tissues was larger than the upper limit of the confidence interval, the high expression levels of genes in the tissues were considered statistically significant. Gene transcripts where the expression level was lower than one per million were technically considered as no expression.

Microarray datasets were collected from the Array Express of European Bioinformatics Institute, which stores data from high-throughput functional genomics experiments (https://​www.​ebi.​ac.​uk/​arrayexpress). These data include the information of the expression of nuclear receptors through experiments submitted directly to Array Express or imported from the NCBI Gene Expression Omnibus database. We used data from the following databases: (1) Metabolic disease: (a) adipose tissue and liver in high fat diet-induced obese mouse model versus normal diet controls, (b) aortic arch segment of the atherogenic apolipoprotein E gene knockout (apolipoprotein E (ApoE−/−)) mice versus wild-type mouse aorta controls, (c) pancreatic islets and white fat of leptin receptor mutant db/db type II diabetic mice versus control mice, (d) oxidized low-density lipoprotein (Ox-LDL)-stimulated mouse endothelial cells versus control endothelial cells, and (e) high-concentration homocysteine (Hcy)-treated human aortic smooth muscle cells (HASMCs) versus low-concentration homocysteine (Hcy)-treated vascular smooth muscle cells (VSMCs); (2) CD4+Foxp3+ regulatory T cell (Treg) polarization/differentiation—we examined the expression changes of the nuclear receptors in Tregs versus effector T cells in mice, as well as in vitro, with cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) ligation; (3) mRNA expression of NR changes due to the stimulus with pro-/anti-inflammation conditions; and (4) we screened the datasets among energy metabolic nuclear receptors of tricarboxylic acid (TCA) cycle and respiratory chain. The modulation of nuclear receptor expression in cancers were determined by analyzing the Cancer Genome Atlas database.

Genome-wide association studies (GWASs) continue to be a widely used approach to detect genetic association with a phenotype of interest in well-defined populations. Various anthropometric measures serve as surrogates for obesity, with body mass index (BMI) (HGVPM 1111 and 564) and waist-hip ratio (HGVPM 1114) as the most frequently used markers in epidemiologic studies aimed at assessing obese disease risks. Anti-cyclic citrullinated peptide-positive rheumatoid arthritis status (HGVPM 38) and rheumatoid arthritis (HGVPM 235) are the most frequently used markers in epidemiologic studies aimed at rheumatoid arthritis risk. Fasting plasma glucose (HGVPM 825), homeostatic model assessment of β-cell function (HGVPM 827), fasting insulin (HGVPM 822), homeostatic model assessment of insulin resistance (HGVPM 826), glycated hemoglobin levels (HGVPM 1081), glycosylated hemoglobin (HGVPM 569), 2-h glucose challenge (HGVPM 769), type II diabetes status (HGVPM 4 and 5), early onset type II diabetes mellitus (HGVPM 74), proinsulin levels (HGVPM 1538), and a-glucose (HGVPM 3639) are the most frequently used markers in epidemiologic studies aimed at diabetes risks. Serum cholesterol (HGVPM 568), lipids (CH3) (HGVPM 3602), lipids (CH2) (HGVPM 3611), lipids (CH2CO) (HGVPM 3616), lipids (CH=CH*CH2CH2) (HGVPM 3640), and systolic blood pressure (HGVPM 563) are the most frequently used markers in epidemiologic studies aimed at vascular atherosclerosis risks. With the advent of large GWASs, we now have the ability to identify NRs associated with dangerous risks for specific disease.

MouseMine (www.​informatics.​jax.​org) is a new data warehouse for accessing mouse data from Mouse Genome Informatics (MGI). The main source of MouseMine data is MGI, which includes a wealth of information about the structure and function of the mouse genome, developmental gene expression patterns, phenotypic effects caused by mutations, and annotations of human disease models. The “Human-mouse: disease connection” tool (www.​informatics.​jax.​org/​humanDisease.​shtml) supports uploading a list of nuclear receptor gene IDs/symbols and getting back certain information about those nuclear receptors, such as those associated human diseases and abnormal mouse phenotypes reported in adipose, cardiovascular, metabolic, and endocrine systems.

The concentrations of S-adenosyl methionine (SAM) and S-adenosyl homocysteine (SAH) were measured in six tissues (heart, liver, lung, kidney, spleen, and brain) in C57BL/6J (n = 4) mice from 13.4 to 18 weeks of age. Mouse tissues were collected and homogenized in 0.4 mol/L perchloric acid (PCA) solution. The homogenized tissues were centrifuged for 10 min at 2000 rpm. Supernatant was collected and stored at −80 °C. SAM and SAH levels were analyzed by liquid chromatography-electrospray ionization-tandem mass spectrometry (LC-ESI-MS/MS; Institute of Metabolic Disease, Baylor Research Institute, Dallas, TX). The unit of SAH level in tissues is nanomole per gram [24].

As summarized in Table 1, the NR superfamily includes 48 NRs classified into seven families, such as class I-thyroid hormone receptor-like family (19 members), class II-retinoid X receptor-like family (12 members), class III-estrogen receptor-like family (9 members), class IV-nerve growth factor IB-like family (3 members), class V-steroidogenic factor-like receptor family (2 members), class VI-germ cell nuclear receptor-like family (1 member), and class O-miscellaneous family (2 members). In addition, we summarized seven common features of the NR superfamily in Table 2. One of the most striking features of NRs is that in addition to transduce steroid, thyroid, retinoid, and other hormone signals, NRs can also serve as metabolic sensors and xenobiotic sensors for high-affinity ligands and low-affinity molecular patterns [25]. Several reports showed that NRs not only bind to specific ligands but also recognize structural patterns (Table 3), which raises a possibility for NRs to recognize many endogenous metabolites that can act as HAMPs that are yet to be identified/characterized [5].

To determine whether tissues have functional differences in sensing metabolic stressors and xenobiotic stressors via NRs, we hypothesized that various tissues express differential levels and certain types of NRs under physiological conditions. To examine this hypothesis, the expression of 48 NR genes in 21 human tissues and 17 mouse tissues were examined (fewer mouse tissues were examined due to unavailability of gene expression data for four types of mouse tissues, i.e., nerve, trachea, stomach, and vascular tissues in the NIH UniGene database) (Additional file 1: Figure S1). The results showed that some human tissues such as muscle (17), trachea (14), and nerve (10) express a large variety of NRs at high expression levels (Tables 4 and 5). This data suggests that the gene expression, differentiation, and function of these tissues may largely be regulated by NRs under normal physiological levels. Comparatively, eyes (7), adrenal gland (6), kidney (5), and adipose tissue (5) express more variety of NRs than the heart, liver, and pancreas (Table 4). Similarly, when comparing the human NR expression profile to that of the mouse, human tissues express much more types of NRs at high expression levels than mice. For example, although human and mouse muscles contain more variety of NRs at high levels relative to other tissues studied, human muscle expresses 17 NRs whereas mouse muscle expresses only 7 NRs. Among the 17 human muscle-expressed NRs, the higher expression of THRB, RORA, ESR1, ESRRA, NR3C2, and NR4A3 in human muscle is not seen in mouse muscle (Tables 4 and 5). Therefore, this indicates that these receptors were evolutionally gained, and addition of these NRs in humans may be responsible for the development of new muscle functions in response to environmental changes/nutritional changes that humans face. Furthermore, nearly half of the tissues examined (including the heart, liver, pancreas, brain, and lymph node) did not contain a large variety of NRs at high expression levels. These results suggested that the gene expression, differentiation, and function of these tissues may be largely dependent on those expressed NRs rather than the non-expressed NRs. Similarly, the human skin, spleen, stomach, vascular, blood, and lung tissue had minimal varieties of nuclear receptors in physiological conditions, since less than 4 out of 48 nuclear receptors are highly expressed (Additional file 2: Figure S2).

Based on the distribution pattern of highly expressed NRs among the tissues, we classified NRs into following four groups: very highly distributed, highly distributed, moderately distributed, and scarcely distributed (Table 6). In order to determine whether very highly distributed and highly distributed groups of NRs have any functional differences from that of moderately distributed and scarcely distributed group of NRs, we analyzed the potential signaling pathways with the Ingenuity Pathway Analyzer for these two major groups of NRs. The results in Table 6 show that among the top 10 pathways examined for each group, the two major NR groups share four signaling pathways such as FXR/retinoid X receptor (RXR) activation, hepatic cholestasis, aryl hydrocarbon receptor signaling, and RAR activation. The very highly distributed and highly distributed group of NRs have six specific top pathways including peroxisome proliferator-activated receptor (PPAR) signaling, glucocorticoid receptor signaling, melatonin signaling, estrogen receptor signaling, adipogenesis pathway, and PPARα/RARα activation. In contrast, the moderately distributed and scarcely distributed groups of NRs have another six specific top pathways including Oct4 stem cell pluripotency, pregnane X receptor (PXR)/RXR activation, LPS/IL-1-mediated inhibition of RXR function, retinoic acid-mediated apoptosis signaling, 25-dihydroxyvitamin D3 (vitamin D3) receptor (VDR)/RXR activation, and liver X receptor (LXR)/RXR activation. Of note, the NRs that have vitamin A, vitamin D, and retinoids as ligands are all included in the scarcely distributed group. Therefore, these data suggest that the tissue expression levels and distribution pattern of NRs can be used as an indicator of functional differences in tissues.

A previous paper reported mouse nuclear receptor tissue expression profile using nucleic acid-binding-based RT-PCR technique [26, 27]. However, NR superfamily expression using more accurate DNA sequencing-based technique has not been profiled in human tissues. Comparing with that reported for mouse NR expression by Bookout et al. [26], our results on highly expressed NRs have the following features (Table 7): (1) our expression sequence tag (EST)-based data were more precise; (2) our data included 21 human tissues, but the previous report only examined mouse tissues; and (3) our data implicated that there are significant differences between human and mouse NR expressions, which had never been investigated before. Our data shows that humans have more NRs expressed in the central nerve system (CNS, 19 versus 11), metabolic system (40 versus 13), and cardiovascular system (19 versus 8). Therefore, our data of NR tissue expression profiles have provided valuable insight over potential NR functions in human tissues.

Based on the variety of NRs expressed in tissues, we classified tissues examined into three categories (Fig. 2), high variety (expressed NRs n ≥ 10; n = number of different types of highly expressed NRs), moderate variety (expressed NRs 5 ≤ n < 10), and low variety (expressed NRs n ≤ 4) in a new nuclear receptor pyramid model shown in Fig. 2 in humans. Similarly, we classified mouse NR pyramid model as high variety (expressed NRs n ≥ 7; n = numbers of the highly expressed NRs), moderate variety (expressed NRs 3 ≤ n < 7), and low variety (expressed NRs n < 3) (Fig. 2). These results suggested that the super high variety and moderate variety of NRs are found in tissues such as the muscle, trachea, and nerves in humans and in the muscle and skin in mice. Therefore, it can be concluded that these tissues may use NR pathways the most to regulate gene expression in response to developmental, physiological, and environmental stimulation. However, a high variety expression of NRs in the trachea has not been extensively reported [28]. It has been reported that NRs regulate skeletal muscle mitochondrial function [29] and the nervous system [30]. In addition, those tissues that have low variety of NRs may need fewer variety of NR pathways to regulate genes in response to developmental, physiological, and environmental stimuli; thus, they may also have other redundant pathways to carry out similar functions to that of NRs.

Genome-wide association studies (GWASs) have investigated potential genetic factors that explain inter-individual variations in response to NR ligand stimulations in various pathologies [50]. Given that susceptibility to complex human metabolic diseases is likely a result of genes operating as part of functional modules rather than individual effects, association analysis methods hold promise in discovering additional associations from existing GWAS data [51]. Previous GWAS studies have been reported for NRs in some diseases such as liver injury [50], osteoporosis, sarcopenia, and obesity. However, it is unclear whether the GWAS data on NRs are associated with globally increased genetic risks for metabolic diseases and autoimmune disease, such as rheumatoid arthritis, obesity, diabetes, and vascular atherosclerosis in human populations.

To address this issue, we examined the GWAS database (http://​www.​gwascentral.​org/​) for all the NRs. As shown in Tables 8 and 9, 45 out of 48 NRs with sequence changes or mutations were associated with rheumatoid arthritis, obesity, diabetes, and vascular disease and atherosclerosis. In addition, two NRs such as PPARA and NR3C2 variations were associated with certain lipid metabolite traits (Table 8). Despite the fact that AR exerts pro-inflammatory effects like PPARD and RXRA, it was much less associated with development of obesity and diabetes unlike PPARD and RXRA (Table 8). Finally, NR2F2 variations were not associated with the diseases examined except in one diabetes study.

These results suggest that NRs may be very important factors in determining the susceptibility and progression of metabolic disorders including obesity, diabetes, and atherosclerosis. Also, our GWAS analysis suggests that NRs may play an important role in the progression of autoimmune disorders such as rheumatoid arthritis. Most interestingly, the NR mutations associated with various metabolic disorders and autoimmune diseases are different. This observation can be supported by multiple publications that had demonstrated NRs play an important role in immune cells. Especially, the PPARs are highly expressed in human CD4+ T cells [52], and the role of PPAR agonists in the treatment of autoimmune disorders had been extensively discussed [53, 54]. It was shown that activation of T cells was dramatically decreased in the presence of PPARA and PPARG agonists and suppressed pro-inflammatory cytokine secretion [52]. In addition to PPARs, other NRs such as ROR-γt were found to regulate differentiation of CD4+ T helper 17 (Th17) subset [55]. However, RAR/RXR dimerization exerts contrasting effects to that of ROR-γt by enhancing Foxp3 transcription factor positive inducible T-regulatory cells (Tregs) while inhibiting Th17 differentiation [56]. Therefore, it is evident that the cross talk between NRs play a critical role in the immunity and development of autoimmune disorders.

In addition to the GWAS analysis, we further performed the cause-effect analysis using the mouse genome informatics (MGI) database (www.​mousemine.​org) that contains a comprehensive compilation of genomic and phenotypic data from NR transgenic and gene knockout mouse models. In Table 10, 26 NR deficiencies lead to four groups of abnormalities, including (1) hormone insufficiency/insensitivity/resistance, (2) cancers, (3) autoimmune/immunodeficiency diseases, and (4) metabolic cardiovascular diseases. The NRs in the hormone insufficiency/insensitivity/resistance group include NRs THRA, THRB, RARB, NR1H3, VDR, NR2E3, NR2F1, NR2F2, ESR1, ESRRB, NR3C1, NR3C2, PGR, AR, NR4A2, NR4A3, NR5A1, and NR0B1. The NRs in the cancer group include RARA (acute promyelocytic leukemia), NR1H4 (hepatocellular carcinoma), ESR1 (breast cancer), and AR (prostate cancer). The NRs in the autoimmune/immunodeficiency disease group include PPARG (systemic lupus erythematosus), RORC (immunodeficiency), and RXRA (systemic lupus erythematosus). The NRs in the metabolic cardiovascular diseases group include RARA (cardiomyopathy), PPARD (diabetes mellitus), PPARG (diabetes mellitus, lipodystrophy, obesity, pulmonary hypertension), HNF4A (diabetes mellitus), NR2F2 (congenital heart defects), ESR1 (myocardial infarction), NR3C2 (hypertension), AR (obesity), and NR0B2 (obesity). Taken together, the results of the GWAS association studies and MGI casual analysis suggest that NRs play significant roles in maintaining the homeostasis and suppressing various hormonal diseases, cancers, autoimmune diseases/immunodeficiency, and metabolic cardiovascular diseases.

To further determine whether many NRs serve as homeostasis-associated molecular pattern receptors (HAMPRs) by inhibiting inflammation, we conducted an extensive literature survey to find out experimentally validated data to prove our hypothesis. As shown in Table 11, the level of 10 hormone ligands of NRs was changed with inflammatory diseases. The ligands of class-I thyroid hormone receptor-like group including vitamin A, fatty acids, and prostaglandins levels were reduced in the presence of inflammatory disorders, suggesting that they have the potential to exert anti-inflammatory effects. In addition, retinoids and estrogen inhibited inflammatory intestinal disease and atherosclerosis respectively. Moreover, testosterone suppressed Crohn’s disease.

Finally, we also searched for the evidence in the literature where gene knockout and activation approaches of NRs were used to determine the pathological phenotypes. Twelve out of 15 NRs including NR1A1, NR1C3, NR1D1, NR1H3, NR1H2, NR1H4, NR2F2, NR3A1, NR3B2, NR4A1, NR4A3, and NR0B2 have anti-inflammatory roles as shown in Table 12. The three NRs NR1C2, NR2B1, and NR3C4 did not show any anti-inflammatory properties. Taken together, these results suggested that most human and mouse NRs have anti-inflammatory functions in various tissues and cell types.

In order to determine the overall roles of NRs in modulating the pathogenesis of human autoimmune diseases, metabolic diseases, and cancer, we examined the expression changes of 48 NRs in eight human diseases using the microarray datasets (https://​www.​ncbi.​nlm.​nih.​gov/​gds/​) deposited by other investigators in the NIH-GEO dataset database. The microarray datasets we analyzed were conducted on various pathological settings including autoimmune disease rheumatoid arthritis, and five metabolic diseases such as familial hypercholesterolemia, type 2 diabetes, type 1 diabetes, obesity, hyperhomocysteinemia, and also hypertension. We analyzed The Cancer Genome Atlas (TCGA) database to determine NR expression changes in human cancers.

As shown in Table 13 (A), three NRs were upregulated but nine NRs were downregulated in the synovial tissue of patients with rheumatoid arthritis. Similarly, in Table 13 (B), 7 NRs were upregulated and 11 NRs were downregulated in T cells from patients with familial hypercholesterolemia. Also, we analyzed the monocytes isolated from patients with familial hypercholesterolemia, peripheral blood from patients with metabolic syndrome, arterial tissue from patients with type 2 diabetes, peripheral blood mononuclear cells from patients with type 1 diabetes, adipose stem cells and omental adipose tissue from morbidly obese patients, aortic smooth muscle cells from patients with hyperhomocysteinemia, and carotid artery atheromatous plaques from patients with hypertension. The results showed that NRs have the tendency to be downregulated during metabolic disorders and autoimmune disorders rather than being upregulated. However, this trend was not observed in morbidly obese patients where equal numbers of NRs were upregulated and downregulated (Table 13 (B)). To further consolidate the finding, we analyzed the NR expression changes in the presence of proatherogenic stimulus oxidized low-density lipoprotein (Ox-LDL) in human aortic endothelial cells (HAECs). This analysis also showed that NRs tend to be downregulated than upregulated with prolonged Ox-LDL treatment (Table 13 (C)). Taken together, these results suggested that NRs have the tendency to be downregulated than upregulated during human autoimmune rheumatoid arthritis and metabolic diseases, and this tendency of NRs was more obvious in autoimmune arthritis than in metabolic diseases.

Specifically, our data shows that NR1C1 (PPARα) is among the downregulated genes in familial hypercholesterolemia. NR1C1 is one of the primary modulators in fatty acid oxidation and apolipoprotein synthesis [23]. This receptor was also found abundantly in the vascular wall and in human macrophages and was shown to exert anti-inflammatory and anti-atherogenic effects [57]. Therefore, downregulation of this gene may contribute to hypercholesterolemia and also to progression of atherosclerotic events. PPARα agonists are widely used to correct hyperlipidemia and were shown to reduce mortality and morbidity due to cardiovascular events [58]. Furthermore, we observed that NR1C3 (PPARγ) is downregulated in patients with rheumatoid arthritis. Previously, PPARγ was reported to have a negative effect on oxidative stress, and therefore, it was suggested that concomitant use of PPARγ agonists with other treatments will give additional therapeutic benefits against rheumatoid arthritis [59].

NRs play an important role in the development and progression of cancers. For an example, the roles of androgen receptors in breast and prostate cancers are well documented [60‐62]. We analyzed the NR expression in 17 different types of cancers in TCGA database. Similar to the observation we elaborated above, our data revealed that the tendency of NRs to be downregulated is more than being upregulated. NR1H2 receptor was downregulated in as many as seven types of cancers, NR2B2 in six types, and NR1B1 and NR1A1 in five types of cancers (Table 14). However, specifically NR1A1 (RAR α) and also NR1B1 (RAR β) are associated with progression of estrogen-dependent breast cancers [63]. This is contrasting to our observation of the expression of these two receptors in other types of cancers. Nevertheless, activation of NR1H2 which falls in to liver X receptors was shown to inhibit proliferation of HT29 colorectal cancer cells [64]. Therefore, this suggests that NR1H2 can be a potential therapeutic target for the treatment of many types of cancers.

To determine the features of those human diseases-modulated NRs, we performed Venn analysis as we previously reported [15]. The Venn analysis/diagram is a very useful analytical tool as it helps to clearly visualize the NRs that are shared between the different diseases analyzed. In Fig. 6a, b, the results show that NR expression changes in human diseases are not shared. In four human diseases analyzed by the Venn analysis, 13 NRs were upregulated, 20 NRs were downregulated, and 15 NRs were not changed in their expression levels (Fig. 6c). Of note, 10 out of 13 upregulated NRs in human diseases were from the scarcely distributed group shown in Table 6, 11 out of 20 NRs downregulated in human diseases were from the very highly distributed and highly distributed groups in Table 6, and 13 out of 15 NRs whose expressions were not changed in human diseases were from the moderately distributed and scarcely distributed groups shown in Table 6. Notably, the tissue expression of 3 NRs out of 20 disease-mediated downregulated NRs including NR1C1, NR1H3, and NR1C3 were correlated with tissue hypomethylated index SAH levels (Fig. 5b), 4 out of 15 NRs whose expressions were not changed were correlated with hypomethylated index SAH levels, and none of the NRs in the disease-upregulated group were correlated with hypomethylated index SAH levels (Fig. 6c). These findings are in a good correlation with tissue hypomethylation function in promoting inflammation as we reported [32, 65], suggesting that hypomethylation-promoting hyperhomocysteinemia may facilitate inflammation via inhibiting the expression of those human disease-downregulated NRs and also keep the stable expression of those non-disease-changed NRs.

In order to determine functional significances of disease-modulated NR expression, we analyzed the top 10 signaling pathways with the disease-upregulated NRs, disease-downregulated NRs, and non-disease-modulated NRs using the Ingenuity Pathway Analyzer (https://​www.​qiagenbioinforma​tics.​com/​products/​ingenuity-pathway-analysis/​). As shown in Fig. 6d; in addition to the shared pathways among three groups of NRs, two pathways, Wnt/β-catenin signaling, and Nur77 signaling were specifically associated with the disease-upregulated NRs; two other pathways such as peroxisome proliferator-activated receptor (PPAR) signaling and thyroid hormone receptor (TR)/retinoid X receptor (RXR) activation were specifically associated with the downregulated NRs; and four pathways including LPS/IL-1 inhibition, pregnane X receptor (PXR)/RXR activation, liver X receptor (LXR)/RXR activation and 1α, and 25-dihydroxyvitamin D3 (vitamin D3) receptor (VDR)/RXR activation were specifically associated with the non-disease-modulated NRs. These results have provided novel insight on the potential functions of various NRs in modulating the pathogenesis of human autoimmune arthritis and metabolic diseases.

Since we found that upregulation and downregulation of certain NRs can be shared in several human diseases (Fig. 6a), we examined whether the NRs shared in human diseases can be used as biomarkers for the diseases and complications. To test this issue, we organized the analysis results in Fig. 6e. The results showed that upregulation of three NRs such as NR1H3, NR1I1, and NR3A1 can be used as biomarkers for rheumatoid arthritis and that upregulation of NR1I1 alone and downregulation of only two NRs NR1A1 and NR3C3 can be used as potential biomarkers for rheumatoid arthritis with familial hypercholesterolemia. In addition, upregulation of NR3C4 can be used as potential biomarker for type 2 diabetes whereas downregulation of two NRs including NR3C3 and NR4A3 can only be used as biomarkers for diabetes with rheumatoid arthritis as a complication. Moreover, upregulation of three NRs such as NR4A1, NR4A2, and NR4A3 can be used for the potential biomarkers for obesity while downregulation of NR1D1 can be used as a potential biomarker for obesity with rheumatoid arthritis as a complication. Finally, downregulation of NR1A1 alone can be used as the biomarker for obesity complicated with rheumatoid arthritis and familial hypercholesterolemia. The results suggest that the NRs shared in human diseases may be highly valuable in serving as potential biomarkers for detection of autoimmune arthritis, metabolic diseases, and their complications.

We then hypothesized that the expression of NRs is regulated by numerous inflammation-modulating pathways. To test this hypothesis, we examined the expression of NRs in various gene-deficient mouse models and cells with overexpression of genes of interests. First, two NRs such as Nr1h3 and Nr1i1 were found to be upregulated, and four other NRs were downregulated in the aortic arch in apolipoprotein E (ApoE)−/− mice fed with 24 weeks of high fat diet (Table 15). However, only one NR, Nr2f1, was found to be downregulated in the aortic arch of ApoE−/− mice fed with 8 weeks of high fat diet. These findings suggest that NR modulation in ApoE−/− in the aortic arch requires prolonged high fat diet feeding.

In another study, it was shown that expression of NRs were upregulated and nine NRs were downregulated in aortic macrophages of low-density lipoprotein receptor (LDL-R)-deficient mouse aortic macrophages relative to wild type (Table 15). In a separate study, one NR Nr1b2 was found to be upregulated but three NRs were downregulated in diabetic db/db glomerular endothelial cells (Table 15). These results suggest that once again in proatherogenic models and type 2 diabetes model, there is a less tendency for NRs to be upregulated than downregulated.

Second, we examined whether inflammatory cytokine signaling pathways can downregulate NR expression. In Table 16, in interferon-γ (IFN-γ)-stimulated endothelial cells, interleukin-1β (IL-1β)-stimulated endothelial cells, IL-1β-stimulated human peripheral mononuclear cells (PBMCs) and tumor necrosis factor-α (TNF-α)-stimulated PBMCs, the NR expressions were either upregulated and downregulated in similar numbers or less upregulated than downregulated.

Third, in Table 17, we examined whether anti-inflammatory cytokine pathways and inhibition of the pro-inflammatory transcription factor regulate NR expressions. We observed that NR expressions were modulated in hepatocellular carcinoma cells stimulated with transforming growth factor-β (TGF-β), palatal mesenchyme cells from TGF-β knockout (KO) mice, in conventional T cells stimulated with anti-CTLA-4 (cytotoxic T-lymphocyte-associated protein 4, also known as CD152, a T cell co-suppressor) antibody, in regulatory T cells (Tregs) stimulated with anti-CTLA-4 antibody and in hearts extracted from cardiac-specific transgenic PPARα mice. In a study with NF-kB inhibitor-treated cells, six NRs were upregulated and five NRs were downregulated. These results demonstrated that immune suppressor pathways CTLA-4, NF-kB inhibitor, and Treg suppress inflammation by significantly upregulating the expression of nine NRs including NR4A1 (5.6–8.3 folds), NR4A2 (6.7–15.8 folds), NR4A3 (4.6–8.1 folds), NR1B2 (6.7 folds), NR1D1 (3.4 folds), NR2A2 (3.2 folds), NR1H4 (2.7 folds), NR2C1 (5.4 folds), and NR3A2 (4.2 folds).

Fourth, in Table 18, we examined whether a key enzyme of tricarboxylic acid (TCA) cycle, isocitrate dehydrogenase (IDH), regulates the expression of nuclear receptors. The results showed that IDH mutation in isogenic epithelial cells results in significant upregulation of six NRs and downregulation of eight NRs. Fifth, in Table 19, we examined whether four key enzymes of the mitochondrial respiratory chain including nicotinamide adenine dinucleotide (quinone) (NADH) dehydrogenase (Nd2), succinate dehydrogenase (SDH), cytochrome c oxidase-4 (COX4), and mitochondrial respiratory chain complex IV regulate the expression of nuclear receptors. The results showed that the mutations of these enzymes result in significant changes in the expression of NR2F2 (Nd2 mutation induced sevenfold upregulation), NR5A2 (Nd2 mutation induced 27-fold downregulation), NR1A2 (COX4 mutation induced 46.5-fold upregulation), and NR2F6 (Cox4 mutation induced 3.3-fold downregulation). Taken together, these results suggest that the expressions of nuclear receptors are regulated by numerous inflammation-modulating pathways and mitochondrial energy metabolic enzymes.

In Fig. 4, we found that innate immune sensor PRRs such as NOD1, NOD2, NOD4, and IFI16 may either act as upstream regulators or downstream targets of NRs in tissues and that the NLR-mediated regulation on NR expression in tissues are evolutionally conserved and mainly act toward suppression of NRs during microbial infection-triggered inflammations. To consolidate this finding, we determined whether gene deficiencies of caspase-1 and other inflammasome components affect the expression of NRs.

As shown in Table 20, deficiency of caspase-1 in ApoE−/− mouse aorta, adipose tissue, deficiency of caspase-1 in associated speck-like protein containing a CARD (ASC)−/− background, deficiency of histone deacetylase and caspase-1 substrate sirtuin 1 (Sirt1) in Tregs, and deficiency of NLRP3 in adult and children PBMCs led to mostly upregulation of NRs instead of downregulation of NRs. For example, the deficiency of NLRP3 led to upregulation of 26 NRs (54%) but downregulation of 9–11 NRs (19–23%). These results suggest that caspase-1/NLRP3 inflammasome pathways play a critical role in regulating the expression of NRs. In addition, in Table 21, we also noticed that the inflammasome/caspase-1 deficiencies upregulated 29 NRs (60%), downregulated 10 NRs (21%), but did not change the expression of 9 NRs (19%).