Estrogen Receptor α Signaling Exacerbates Immune-Mediated Nephropathies through Alteration of Metabolic Activity

Systemic lupus erythematosus (SLE) is an autoimmune disease that affects 1 in 2000 individuals in the United States and ∼5.5 per 100,000 individuals worldwide (1). Up to 50% of SLE patients develop some degree of renal involvement, with up to 20% progressing to end-stage renal disease, depending on racial background (2). Lupus nephritis (LN) is characterized by anti-nuclear Ab production, immune complex deposition, and immune-mediated kidney damage. Development of LN remains a sign of poor prognosis and is a significant cause of morbidity and mortality (3).

Because 9 of 10 SLE patients are women, the role of the sex hormones estrogens in this disease is of key interest. Although men and women develop LN, the disease often occurs earlier and is more severe in men (4). Estrogens signal through two receptors: estrogen receptor α (ERα) and estrogen receptor β (5). In contrast to estrogen receptor β, ERα is found in female reproductive organs, yet is robustly expressed in kidney, liver, heart, and lungs in males and females, as well as on most immune cells (6); however, the kidney is considered the most estrogenic nonreproductive organ.

Earlier reports have demonstrated that the absence of ERα in lupus-prone mouse strains ameliorates renal damage (7). However, the diminished renal disease was associated with decreased autoantibody production, confounding the determination of a direct effect of ERα on the kidney. In this article, we show that ERα does not affect the humoral immune response but exacerbates immune-induced renal damage by altering PPAR signaling and reducing retinol metabolism in the glomeruli, resulting in increased inflammation and fibrosis. Furthermore, by sequencing of blood samples from lupus patients, we found similar decreased expression of genes associated with lipid and retinol metabolisms in lupus patients compared with healthy controls, confirming the relevance of these pathways in human disease. These results reveal the significance of metabolic pathways in renal disease for mice and humans, as well as demonstrate the profound impacts of hormonal environments on disease pathogenesis.

C57BL/6 (B6) and B6.ERα-knockout (KO) mice were bred and maintained according to the guidelines of the Lewis Katz School of Medicine at Temple University’s University Laboratory Animal Resource Office, an Association for Assessment and Accreditation of Laboratory Care International–accredited facility. Experimental procedures were conducted according to the Institutional Animal Care and Use Committee guidelines. B6.ERα-KO mice were originally obtained from Dr. G. Gilkeson (Medical University of South Carolina). B6.ERα-KO mice were bred as heterozygotes, and the genotyping was performed by standard PCR, as described elsewhere (The Jackson Laboratory).

Patients were randomly selected from the Temple Lupus Cohort, which is an Institutional Review Board–approved prospective cohort of lupus patients followed in the Lupus Clinic at Temple University Hospital. After informed consent was given, blood was obtained from patients who fulfilled at least four of the Systemic Lupus International Collaborating Clinics Criteria (8). Blood was processed as described in the Microarray section. Renal disease activity was defined by the SLE Disease Activity Index (9) and was determined on the day that blood was obtained. We enrolled 21 female SLE patients, and 21 healthy age-, sex-, race-matched controls in the study. Of the SLE patients, 7 were Hispanic, 12 were African American, and 2 were white.

Chronic graft-versus-host disease (cGVHD) is an inducible model of lupus disease that is characterized by the development of typical lupus autoantibodies, such as anti-dsDNA and anti-chromatin. cGVHD was induced in the mice as described previously (10). Briefly, 8-wk-old mice were injected i.p. with 10⁸ splenocytes from B6.Bm12 mice. Mice were bled, and serum was collected weekly via the tail vein for up to 6 wk postinjection.

Nephritis was induced using nephrotoxic serum (NTS), as shown previously (11). Briefly, 8-wk-old mice were injected i.p. with NTS (8 μl/g body weight). To assess the degree of kidney disease, blood urea nitrogen (BUN) levels were measured using Azostix (Siemens) during each tail bleed before injection, as well as every 2–4 d following the injection, for the duration of the experiment.

Eight-week-old wild-type (WT) and B6.ERα-KO mice were irradiated (550 cGy) twice, 3 h apart. Twenty-four hours later, irradiated mice were injected i.v. with 100 million bone marrow cells harvested from donor WT and B6.ERα-KO mice. After injection, irradiated mice were treated with 5 mg/kg meloxicam for 2 wk and with antibiotics for 4 wk. Blood was collected via the tail vein and genotyped to check chimerism at 6 wk postinjection.

Mice were anesthetized using isoflurane. Twenty-one-day slow-release 0.5 mg 17β-estradiol pellets (Innovative Research of America) were used as the source of estrogen. Pellets were inserted s.c. between the shoulders using a 10-gauge trocar.

Serum anti-dsDNA and anti-chromatin Abs were detected by ELISA, as previously described (12).

Eight-week-old mice were injected i.m. with 100 μg of OVA in a 1:1 ratio in CFA. One week later, mice received a second i.m. injection of OVA in a 1:1 ratio in IFA. Blood was collected via the tail vein weekly for 4 wk.

ELISA plates were coated with 30 μg/ml OVA in BBS. Following blocking (3% BSA in BBT), serum samples were diluted 1:500 in BBT and incubated overnight at 4°C. The secondary Ab used was AP-conjugated goat anti-mouse IgG (1:5000; Fcγ specific; Jackson ImmunoResearch). The plates were developed using 1 mg/ml pNPP (Sigma). A standard curve was generated using OVA in 2-fold dilutions from 1:500 to 1:256,000.

Ten-micrometer frozen kidney sections were fixed in ice-cold acetone, stained with Mayer’s hematoxylin (Sigma), and washed with sodium bicarbonate for bluing. Sections were then stained with eosin (Sigma) and dehydrated, and coverslips were mounted. H&E sections were scored for glomerular damage and interstitial inflammation, as described (13).

Fibrotic tissue was detected in kidney sections using a Trichrome Stain kit (Abcam), following the manufacturer’s instructions. Briefly, frozen tissue sections were immersed in Bouin’s Fluid for 60 min. After 10 min of cooling, slides were rinsed in tap water and stained with working Weigert’s iron hematoxylin solution for 10 min. Slides were rinsed in running tap water and then stained with Biebrich scarlet-acid fuchsin solution for 10 min, followed by a 15-min differentiation in phosphomolybdic/phosphotungstic acid solution. Next, Aniline blue was applied for 10 min, followed by acetic acid solution for 5 min. Slides were dehydrated in 95 and 100% alcohol, and coverslips were mounted.

Lipids were detected in the kidney using a specific Abcam kit and following the manufacturer’s instructions. Briefly, frozen tissue sections were fixed in 10% formalin and then incubated in propylene glycol. Slides were stained with Oil Red O solution and then differentiated in 85% propylene glycol. Slides were cover stained with hematoxylin and rinsed, and coverslips were mounted using VectaMount AQ Mounting Medium (Vector Laboratories).

Ten-micrometer frozen kidney sections were fixed in acetone and then blocked (5% goat serum/2% BSA in PBS) for 45 min. Sections were incubated with primary Ab for 1 h at room temperature. The Abs used were FITC-conjugated goat anti-mouse IgG (Fcγ specific) (Jackson ImmunoResearch) and FITC-conjugated goat anti-mouse C3 F(ab′)₂ fragment (MP Biomedicals). Coverslips were mounted using anti-fade mounting medium (Vector Laboratories). IgG and C3 deposition was measured using ImageJ software (National Institutes of Health, Bethesda, MD).

Frozen 10-μm kidney sections were cut using a cryostat and mounted on glass slides. Before staining, the sections were fixed in 4% paraformaldehyde for 15 min at room temperature, washed, and incubated with blocking buffer for 60 min at room temperature. Primary Abs were added for 2 h at 37°C, and after washing, secondary Abs were added for 1.5 h at room temperature. After three washes, coverslips were applied using VECTASHIELD Antifade Mounting Medium with DAPI (Vector Laboratories) to detect nuclei. Primary Abs used were polyclonal guinea pig anti-synaptopodin (1:200; Antibodies-online.com, ABIN1742349) and polyclonal rabbit anti–smooth muscle actin (SMA) (1:200; ab5694; Abcam). Secondary Abs were FITC-goat anti-rabbit IgG (H+L) (1:200; Thermo Fisher) and Rhodamine Red-X goat anti-guinea pig IgG (H+L) (1:200; 106295003; Jackson ImmunoResearch).

Flash-frozen murine kidneys were halved using a sterile scalpel, and the cortex was separated from the medulla. The cortex was passed through a series of sieves (250, 180, and 106 μm) as described previously (14). Finally, the glomeruli were collected in a 40-μm cell strainer. The collected glomerular samples were composed of ∼70% glomeruli and 30% interstitium.

RNA was extracted from the isolated murine glomeruli and quantified on a NanoDrop ND-100 and Qubit 2.0 spectrophotometer, followed by RNA quality assessment on an Agilent 2200 TapeStation (Agilent Technologies, Palo Alto, CA). Fragmented biotin-labeled cDNA (from 10 ng of RNA) was synthesized using an Affymetrix WT Pico Kit. An Affymetrix GeneChip Mouse Transcriptome Array 1.0 (MTA 1.0) was hybridized with 5 μg of fragmented and biotin-labeled cDNA in 200 μl of hybridization mixture. Target denaturation was performed at 99°C for 5 min and then at 45°C for 5 min, followed by hybridization at 60 rpm for 16–18 h at 45°C. Arrays were washed and stained using a GeneChip Fluidics Station 450 and a GeneChip Hybridization, Wash, and Stain Kit (both from Affymetrix). Chips were scanned on an Affymetrix GeneChip Scanner 3000, using Command Console Software. Quality control of the experiment was performed by Expression Console Software v1.4.1. Quality assessment and Affymetrix GeneChip array were conducted by the Jefferson Cancer Genomics Laboratory at the Kimmel Cancer Center in Philadelphia, PA. Raw data files for the microarray were submitted to the National Center for Biotechnology Information Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE98626) under accession number GSE98626.

Chip data files were generated by Signal Space Transformation-Robust Multi-Chip Analysis normalization from Affymetrix cel files using Expression Console Software. The experimental group (ERα-KO females) was compared with the control group using Transcriptome Analysis Console (TAC) Software v3.1.0.5. Lists of differentially expressed genes (>2-fold, p < 0.05) were generated by TAC Software. Principal component analysis was completed using MeV software (15).

RNA was isolated from healthy and lupus patient whole blood using a PAXgene system (QIAGEN). RNA was depleted of rRNA, and strand-specific libraries were generated and run on an Affymetrix HT HG-U133+ array. RNA sequencing data (100-bp paired-end reads) were aligned to the human genome hg19 using TopHat (16), taking into account reads coming from splicing junctions (parameters were set to default). Counts were obtained using featureCounts (17). Library normalization was performed, and fragments per kilobase million for each RefSeq gene was calculated using the edgeR package in R (18).

RNA from glomerular isolates was extracted using a Direct-zol RNA MiniPrep Kit (Zymo Research). Total RNA (500 ng) was reverse transcribed using a High-Capacity Reverse Transcription Kit (Applied Biosystems). Tgfβ1 (Mm01178820_m1), Fabp1 (Mm00444340_m1), Cyp4a14 (Mm00484135_m1), β-actin (Mm02619581_g1), Gapdh (Mm99999915_g1), lactotransferrin (Ltf; Mm00434787_m1), and progesterone receptor (Pgr; Mm00435628_m1) expression was detected using TaqMan gene expression assays (Thermo Fisher) and an Applied Biosystems Step2 plus thermal cycler. Relative gene expression of Tgfβ1, Fabp1, and Cyp4a14 was normalized to β-actin and calculated by the ΔΔCt method. Relative gene expression of Ltf and Pgr was normalized to Gapdh and calculated by the ΔΔCt method.

OVA, cGVHD, and NTS-induced nephritis (NTN) experiments were analyzed using GraphPad Prism. For parametric data, unpaired and paired two-sample t tests were used to analyze differences among groups. Nonparametric data comparisons were analyzed by the Mann–Whitney U test. Mouse microarray gene expression was analyzed using TAC Software (Affymetrix). Human RNA sequencing data were analyzed using MeV software. Statistical significance was determined by a p value < 0.05.

ERα-KO mice have a truncated form of ERα that lacks the estrogen ligand–binding domain. This functional mutation prevents the receptor from responding to estrogen ligands but still allows the receptor to carry out estrogen-independent functions (19). To verify the lack of estrogen response in these mice, WT and ERα-KO female mice received a slow-release estrogen pellet implant that raised the estrogen to pregnancy levels. Ten days postimplantation, WT and ERα-KO mice exhibited increased estrogen levels in the serum (Supplemental Fig. 1A). In addition, WT mice exhibited increased expression of ERα-responsive genes, Pgr and Ltf, in response to estrogen in spleen, kidney, and uterus, whereas ERα-KO mice did not respond (Supplemental Fig. 1B) (20). To confirm the estrogen-independent functions of ERα-KO mice, we studied the responses of WT and ERα-KO splenocytes to epidermal growth factor (EGF). EGF binds to EGFR on the surface of the cells, leading to phosphorylation and activation of ERα independently from estrogen ligands (21). Splenocytes from WT and ERα-KO mice were incubated with 1 ng of EGF. EGF treatment increased Ltf expression in WT and ERα-KO splenocytes, whereas cells treated with the estrogen control showed similar results as in Fig. 1B, demonstrating that the mice still maintain estrogen-independent function (Supplemental Fig. 1C).

A defining characteristic of lupus is an aberrant humoral immune response (22). Because of the limited information concerning the state of the immune system in ERα-KO mice, we wanted to determine the ability of these mice to mount a normal immune response to a foreign Ag. Eight-week-old WT and ERα-KO female mice received OVA immunizations and were monitored for 4 wk. By 4 wk, WT and ERα-KO mice had generated similar levels of anti-OVA Abs (Fig. 1A), demonstrating that the ERα-KO humoral immune system can mount a typical immune response.

Supported by a large body of evidence that sex hormones differentially contribute to the development of lupus (23), we investigated whether ERα plays a role in the generation of autoantibodies. To this end, we used the cGVHD lupus model. This murine model of lupus is dependent on allogeneic T cell help and tests the activation of endogenous autoreactive B cells and their production of autoantibodies directed toward nuclear components (10, 24), a hallmark of lupus disease and major players in the initiation of LN (25, 26). We injected Bm12 splenocytes, which express ERα, into WT and ERα-KO female mice. Six weeks after cGVHD induction, WT and ERα-KO mice had developed similar levels of anti-chromatin and anti-dsDNA autoantibodies (Fig. 1B). These results indicate that ERα does not play a significant role in the humoral autoimmune response during the cGVHD lupus model.

Previous studies by the R. Caricchio laboratory have shown that male mice pretreated with 17β-estradiol develop less severe renal disease during NTN (27). This model uses serum containing anti-glomerular Abs, triggering a type II hypersensitivity reaction: complement-dependent autoimmune response (11). We hypothesized that estrogens contribute to the progression of renal damage in the NTN model, and we investigated the role of ERα, the primary estrogen receptor expressed in the kidney (28). We injected WT and ERα-KO female mice with NTS. Two weeks postinjection, ERα-KO mice developed significantly less severe renal disease compared with WT mice, as measured by BUN levels (Fig. 2A). WT mice progressed to end-stage renal failure, whereas ERα-KO mice maintained mild disease. H&E staining and pathology scoring of the kidneys also supported the BUN results that the presence of ERα is required for the full severity of NTN (Fig. 2B, 2C). In addition, ERα-KO mice had reduced trichrome staining within their glomeruli, indicating less fibrosis, compared with WT mice (Fig. 2D). It is well known that NTN in mice has a first phase (heterologous), during which the sheep IgG present in the NTS induce kidney damage, and a second phase (autologous), during which the immune system of the mouse produces Abs against the NTS that continue to damage the kidney (29). We found that, upon NTS injection, both strains of mice exhibited deposition of mouse IgG and C3 (Fig. 3). Ordinarily, mice do not exhibit IgG and C3 deposition in the kidneys; therefore, the presence of these deposits demonstrates that the absence of ERα did not impair their ability to generate Abs targeting the NTS or activate complement. Taken together, these results demonstrate that the absence of ERα protects females during NTN.

We generated ERα-KO chimeric mice by irradiating WT and ERα-KO mice and then adoptively transferring bone marrow cells from WT or ERα donor mice into the irradiated host mice. Seven weeks after the adoptive transfer, we induced nephritis in the mice using NTS. Two weeks after the NTS injection, ERα-KO mice that received ERα-KO bone marrow had less severe nephritis than controls (Fig. 4A, 4B), consistent with the previous results (Fig. 2). However, the absence of ERα in the immune cell or renal cell compartment resulted in a severity of renal disease similar to WT controls (Fig. 4C). All chimeric mice displayed IgG and C3 deposition within their glomeruli following NTS injection (Fig. 4D, 4E). From these results, we conclude that the expression of ERα on immune cells or renal cells is sufficient to contribute to the progression of renal disease during NTN and that complete absence of ERα is required to confer protection.

Because the presence of ERα in renal and immune cell compartments can affect NTN progression, we decided to investigate the kidney at the transcript level to determine how the absence of ERα conferred protection. Microarray data analysis of gene expression in the murine glomeruli revealed that, over the course of NTN, WT and ERα-KO mice undergo extensive changes in gene expression (Fig. 5A), and the identity of the changing genes is vastly different between the two strains (Fig. 5B). It is interesting to note that untreated WT and ERα-KO kidneys show a differential expression of 591 genes in homeostatic conditions. Eighteen hours after the induction of NTN, they differ by only 15 genes, indicating that the first reaction to the NTN insult is very similar in the two strains and eliminates the homeostatic differences. WT kidneys upregulate twice as many genes as ERα-KO kidneys, suggesting that ERα-KO kidneys have many proinflammatory genes constitutively activated. Six days after NTN, WT and ERα-KO kidneys changed their expression of hundreds of genes (286 and 336 genes, respectively) compared with 18 h and differed from each other by 457 genes, suggesting that the major divergence during NTN has occurred.

To elucidate the mechanisms responsible for the drastic differences in gene expression and pathology between WT and ERα-KO mice, we used TAC Software to determine which molecular pathways were associated with the gene changes during disease progression. Pathway analysis revealed that, at 6 d, WT mice upregulate pathways associated with cellular infiltration, fibroblast activation, and fibrosis, whereas ERα-KO mice use DNA damage repair and cellular proliferation pathways. Furthermore, from 6 d to 2 wk, WT mice downregulate many metabolic pathways, which can explain the severe cell death within the glomerulus (Fig. 2), whereas ERα-KO mice downregulate proinflammatory pathways, suggesting a response aimed at disease resolution. We used Cytoscape software (30) to generate a pathway map that summarized the overall changes and similarities occurring between WT and ERα-KO mice over the course of NTN (Fig. 6). Each node in Fig. 6 represents a molecular pathway, and the node size is proportional to the number of genes that are changing within that given pathway. The map demonstrates that the changes in the molecular pathway mimic the pathology seen in mice, where initially (t₀–t₁), WT and ERα-KO mice primarily undergo changes for different pathways, although some are similar. At 6 d, the two strains are undergoing changes in many of the same pathways, which is analogous to the identical increases in BUN seen in Fig. 2. Finally, the pathways that change for each strain become drastically different during the transition from t₂ to t₃, which is identical to the strong divergence in disease severity between WT and ERα-KO mice from 6 d to 2 wk (Fig. 2).

The gene expression analysis at 6 d post-NTN showed large differences between WT and ERα-KO kidneys (457 genes) (Fig. 7) at the time point when severity diverges between the two groups. Most of the differentially expressed genes were associated with lipid metabolism pathways. The two pathways with the greatest number of genes were PPAR signaling and retinol metabolism. Many of the PPAR-associated genes were upregulated in WT mice compared with ERα-KO mice (Fig. 8A). We validated these microarray results for two genes in the PPARα pathway, fabp1 (upstream of PPARα) and cyp4a14 (downstream of PPARα) measuring the expression of these two genes by quantitative RT-PCR (Fig. 8B). Next, we used STRING protein–protein interaction networks (31) to determine whether lipid metabolism pathways were linked to ERα. We entered a list of genes into the STRING program. In this case, the genes were Esr1 (ERα) and three genes that were significantly upregulated in WT mice at 6 d and were also associated with lipid PPAR signaling: ppara, fabp1, and acsl1. The program provided a network based on literature searches and gene databases showing the relationship, if any, between Esr1 and the three genes of interest. The pathway generated by STRING shows that the genes related to lipid metabolism (ppara, fabp1, and acsl1) are directly downstream of Esr1 (Fig. 9A). Increases in PPAR signaling have been shown to promote anti-inflammatory effects; therefore, we interpreted the higher expression of PPAR-related genes in WT mice as an attempt to control the progression of renal inflammation.

Activation of retinol metabolism results in anti-inflammatory and profibrotic effects. In this study, we found that the genes associated with this pathway are downregulated in WT mice compared with ERα-KO mice, suggesting a role for retinol metabolism in kidney damage during NTN (Fig. 8A). Agonists of retinoic acid receptors have been shown to decrease Tgfb1 expression, fibrosis, mesangial cell proliferation, and glomerular damage (32–35). Quantitative RT-PCR validation supported the increase in Tgfb1 expression at 6 d in WT mice (Fig. 8C). In addition, WT mice had increased mesangial cell proliferation compared with ERα-KO females during NTN (Fig. 8D). We generated a STRING protein interaction diagram to investigate potential interactions between Esr1 and retinol receptors (Fig. 9B) and found that Esr1 is directly upstream of retinoid X receptor α and might interact with it.

Retinol metabolism has been shown to promote cholesterol efflux from macrophages, resulting in decreased intracellular lipid accumulation and foam cell formation (36, 37). We hypothesized that the downregulation of retinol metabolism may increase intraglomerular lipids and foam cell formation in WT kidneys, but not in ERα-KO kidneys, contributing to the more severe kidney damage that is present in WT mice. Oil Red O staining of the kidneys at 14 d confirmed that WT kidneys have an accumulation of lipid that is absent in ERα-KO kidneys (Fig. 8E).

These results demonstrate that a decrease in retinol metabolism in WT females during NTN is associated with increased glomerular damage, mesangial proliferation, and lipid accumulation, which may be due to interactions between Esr1 and retinol receptors; however, further experimentation is needed to determine a direct causality.

To investigate whether the lipid metabolism was also altered in human SLE, we performed RNA sequencing on whole blood samples of 21 lupus patients and 21 healthy controls. Clinical characteristics of the lupus patients are listed in Table I. We analyzed gene expression differences between healthy and SLE samples using MeV software. The DAVID Bioinformatics Database provided the molecular pathways associated with the genes that were differentially expressed between healthy and SLE patients. As expected, many genes were involved in type I IFN responses. We also found that other significant genes were associated with PPAR signaling and retinol metabolism (Fig. 10, Table II). Similar to the findings in WT mice, most of the genes involved in PPAR signaling were upregulated in SLE patients, whereas the retinol metabolism genes were downregulated. In addition, we analyzed a publically available dataset of RNA sequencing generated from whole blood samples from the Michigan patient cohort, consisting of 99 SLE patients and 18 healthy controls (Gene Expression Omnibus accession number GSE72509). These data also displayed an upregulation of PPAR signaling and a downregulation of retinol signaling genes in SLE patients compared with healthy controls (Supplemental Fig. 2). These data provide novel evidence for a role for lipid metabolism in SLE, both in peripheral immune cells and in target tissues.

Earlier studies demonstrated that the presence of estrogens or the inhibition of estrogen receptors can reduce disease in lupus-prone mice (7, 23, 38); however, the mechanism by which ERα directly contributes to LN pathogenesis has not been investigated. We found that the absence of ERα confers protection from nephritis in female mice and propose two novel metabolic mechanisms for such effects.

Initially, our investigation focused on the impact of estrogen on autoantibody production, because their deposition within the kidney is the initial step in the development of nephritis (25, 26). The results from the cGVHD lupus model experiment show that the absence of ERα does not impair the ability of autoreactive B cells to produce autoantibodies in response to aberrant T cell help provided by allogeneic T cells. Together with the results that ERα-KO mice produce normal amounts of Abs in response to immunization, these data suggest that ERα does not affect development of the autoimmune humoral response.

ERα is highly expressed in the kidney (28); therefore, we hypothesized that the role of estrogens in LN may be involved in the direct renal damage downstream of autoantibody deposition. The NTN experiments demonstrated that ERα is involved in nephritis pathogenesis, and its absence resulted in protection in the mice, especially in females. Because of the higher levels of estrogen, it is predictable that a lack of ERα would have a greater impact in females than in males, in which we have previously shown are protected from Parp-1–induced renal damage (27). Type-I IFNs are elevated in the serum of lupus patients and are important drivers of proinflammatory responses. Studies have implicated a connection between ERα and type I IFNs in murine B cells, in which females exhibit an increased IFN signature that can activate a feedback loop consisting of upregulation of ERα-responsive genes and IFN-inducible genes (39). In addition, estrogens have been shown to mediate the expression of p200 IFN-inducible genes (40). In view of these data, ERα-KO mice may have reduced levels of IFNs and IFN-inducible genes, which could have contributed to the reduced renal disease; however, more experimentation is needed to test this hypothesis.

ERα is expressed in nearly every immune cell and renal cell type (6, 28, 41). The results from the bone marrow chimera experiments reveal that estrogens act through ERα in the immune and the renal cell compartments to cause renal damage, suggesting a specific role for estrogen in the activation of proinflammatory immune cells and in the development of inflammation and cell death in renal cells.

The results of the microarray analysis conducted on glomerular RNA from nephritic mice supported the divergence in disease severity seen in the histology of WT and ERα-KO females, especially beginning at day 6. When comparing WT mice with ERα-KO mice at day 6, many of the differentially expressed genes were associated with lipid metabolism pathways downstream of PPARs. In fact, WT females maintain increased PPAR gene expression and decreased retinol gene expression compared with ERα-KO females. These data reveal that the presence of ERα induces intrinsic differences in the expression of metabolic genes. When WT mice experience an inflammatory condition, in the case of retinol metabolism, they are more prone to the downstream pathological outcomes of reduced gene expression.

Activation of lipid metabolism is associated with inflammation in several ways. Immune cells have been shown to shift their metabolic processes toward increased lipid metabolism to expand their Golgi bodies and increase protein synthesis (42). In this way, these cells can release more cytokines and exacerbate the proinflammatory responses, promoting damage. Pathologically, the increased inflammation related to altered lipid metabolism, especially decreased retinol metabolism, has been shown to exacerbate tissue fibrosis (43) and promote mesangial cell proliferation (44, 45), both of which we found to be increased in the WT female kidney during NTN. In addition, this metabolic shift has been shown to result in greater accumulation of lipids within renal cells during nephritis (46). These lipids then are phagocytosed by mesangial cells and infiltrating macrophages, resulting in foam cells and the release of MCP-1, which leads to greater macrophage recruitment to the site of inflammation (47, 48). Interestingly, an earlier report demonstrated, in a breast cancer model, that Esr1 is able to promote survival of cancer cells by binding of retinoic acid response elements in the genome and preventing retinoic acid receptor–mediated transcription of tumor-protective genes (49). We hypothesize that Esr1 may act in a similar fashion to block the anti-inflammatory effects of retinol metabolism during NTN.

Nephrotic syndrome has been shown to induce abnormalities in metabolic processes (50); however, other studies have demonstrated that administration of agonists for these pathways, especially retinoids, is able to protect during acute models of nephritis (41, 42). Therefore, we propose that these metabolic changes are not merely secondary to the nephrotic syndrome but are directly involved in progression of the disease. Our interpretation is supported by the time of occurrence of the changes in the expression of metabolic genes, at day 6 after NTN induction, which is 1 wk before the development of the nephrotic syndrome, as suggested by the levels of BUN.

To determine whether the results in mice supported the human pathogenesis, we investigated the gene expression of the lipid metabolism pathway in whole blood from lupus patients. Gene expression analysis revealed that whole blood from lupus patients follows a similar gene expression pattern for PPAR signaling and retinol metabolism genes as in the mouse glomeruli. These data support our hypothesis that altered lipid metabolism by peripheral immune cells and intrinsic renal cells is an important pathogenic mechanism of LN, as suggested by the bone marrow chimera experiments.

In conclusion, our results demonstrate that, during LN, estrogen signals through ERα on peripheral immune cells and resident renal cells, which causes a shift in metabolic pathways. This metabolic shift enables the cells to release proinflammatory cytokines, proliferate, and enhance the proinflammatory environment, ultimately leading to tissue damage and renal failure. Correlation with the results found from the patients shows that these results may be translated to human disease and provide further insight into the impact of hormonal environments in lupus disease pathogenesis.

We thank Dr. Philip Cohen, Dr. Stefania Gallucci, and Dr. Marc Monestier for critical reading of the manuscript and numerous useful suggestions and Dr. Brendan Hilliard and Austin Parrish for technical support.

The authors have no financial conflicts of interest.