Hypovitaminosis A Drives the Progression of Tubulointerstitial Lupus Nephritis through Potentiating Predisease Cellular Autoreactivity

Systemic lupus erythematosus (SLE or lupus) is a multisystem autoimmune disorder with no definite cure. This might be in part due to disease heterogeneity among SLE patients (1) where persistent inflammation in multiple organs as a consequence of SLE pathogenesis could lead to collateral tissue damage of these organs (2). During the last decades, the management plans for lupus treatments are based on nonselective immunosuppressants (3). Unfortunately, the effects of these therapies are limited especially for end-stage renal disease (3, 4). In addition, long-term use of immunosuppressants aggravates the susceptibility of lupus patients to infections and predisposes them to a higher risk of infection-related mortality (3). Indeed, treatment-related complications could even overweigh those from disease manifestations (5). With the worldwide increase in the incidence of SLE and limited long-term prognosis (6), it is imperative to define and understand the immunopathological mechanisms of cofactors (e.g., diet and environment) that can potentially accelerate disease initiation and/or aggravate lupus flares.

It is widely accepted that environmental factors such as dietary nutrients have potent implications on health and disease (7–9). Micronutrients, especially vitamin A (VA; or retinol), have been shown to modulate various immunological processes (10, 11). The most active metabolite of VA, all-trans retinoic acid, could shape the pathogenesis of different autoimmune disorders (12), including SLE (13, 14). We and others have shown the protective effects of retinoid treatment on inflammation of the kidney, or lupus nephritis (13–15), which is one of the most common and life-threatening manifestations of SLE (16, 17).

In contrast, VA deficiency (VAD) or hypovitaminosis A could aggravate inflammatory responses and provoke a higher risk of irreversible collateral tissue damage (18). Interestingly, hypovitaminosis A has been reported in SLE patients (19) with an incidence double that in healthy individuals (19). Cases of lower serum levels of β-carotene (the inactive form of VA or provitamin A) and retinol have been reported preceding the diagnosis with SLE (20). Similarly, in murine lupus models, we have found reduced serum and liver retinol levels with disease progression (13, 14). However, our understanding of whether VAD is a contributing factor for severe SLE and how it affects the initiation and/or the progression of lupus nephritis is limited. Two VAD studies in lupus-prone mice have so far been reported (21, 22), both with VAD initiated after disease onset and neither showing an impact of VAD on lupus nephritis. Whether VAD initiated before disease onset would affect lupus nephritis is unknown. In this study, we established VAD as a driving factor for the initiation of lupus nephritis in genetically predisposed conditions. Utilizing the MRL/lpr murine lupus-prone model, we delineated the changes in disease immunopathology with predisease hypovitaminosis A initiated either prenatally or postweaning. Our findings established the importance of monitoring VA levels in SLE patients and urge for supplementation for SLE patients with hypovitaminosis A, especially during pregnancy and nursing.

Genetically-prone lupus MRL/lpr mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and maintained in a specific pathogen-free animal facility at Virginia Maryland College of Veterinary Medicine, Virginia Tech. All procedures were carried out according to an approved Institutional Animal Care and Use Committee protocol (no. 19-258). To establish VAD as a driving factor for the progression of lupus disease, we generated VAD breeders through dietary modulation. Due to the strong female bias in SLE (23–25), only female mice were analyzed. Thus, to generate 10 female mice per group, we set up four breeder pairs for each condition (experimental setups below), and timed mating was performed. Five-week-old MRL/lpr mice were used for breeding. Because VA is essential for embryogenesis (26), diet modulation started on the seventh gestational day (GD). For VAD breeders, the hormone-free NIH-31 modified 6% mouse/rat diet was replaced by an irradiated VAD diet (Envigo-TD.86143, which contained no retinol). For control mice, an irradiated vitamin A–sufficient (VAS) diet (Envigo-TD.91280, which contained 20,000 IU of VA per kilogram of diet, a level within the range found in standard rodent diets) was used for replacement also at the seventh GD. The nutritional formulations of VAD and VAS diets are shown in Supplemental Table I. Both diet formulations provided 3.8 kcal/g with similar amounts of other micronutrients and macronutrients. Notably, all vitamin levels were similar between both diets. In particular, choline, vitamin D, and vitamin E were added back as line items in TD.86143, which initially lacked these nutrients in addition to lacking VA. Males were removed from the breeding cages on day 14 of the breeding setup to ensure that no pregnancy occurs while lactating dams are on the special diets. The special diets continued after the pups were born, and some pups were sacrificed at weaning (Supplemental Fig. 1A, gestational VAS [G-VAS] versus gestational VAD [G-VAD]). Pups generated through G-VAD continued on the VAD diet postweaning (see Fig. 1A, VAD-D group) or were switched to the VAS diet (VAD-S group), whereas some G-VAS pups were switched to the VAD diet upon weaning (VAS-D group). Other G-VAS weanlings were kept on the VAS diet (VAS-S group). All other factors related to housing, handling, and light cycle (12-h light/12-h dark) were consistent. Food and water were provided ad libitum.

To ensure VA depletion, we confirmed VA (retinol) status using HPLC quantification following our previously published procedures (13). Initial serum and liver retinol/VA levels at weaning by 3 wk of age and endpoint levels around 15 wk of age were determined. Four to five pups (per group) generated from VAD and VAS dams were spared and euthanized for this purpose. Blood was collected for serum separation. Liver samples were harvested and snap-frozen in liquid nitrogen. Aliquots of 60 μl of serum sample and 0.05 g of liver samples were stored at −80°C until analysis. Dietary modulation successfully diminished the retinol levels by weaning (Supplemental Fig. 1B) and completely depleted its levels from both the circulation and hepatic reservoir by 15 wk of age (Supplemental Fig. 1C).

Total splenocytes and bone marrow cells were isolated and RBC exclusion was performed following our previously published procedures (13, 27). For renal leukocyte isolation, our previously reported procedures were conducted with minor modifications (28). Briefly, kidneys were finely minced into grain size 1- to 2-mm³ pieces and digested in 5 ml of digestion buffer containing 1 mg/ml collagenase and 0.2 mg/ml DNAse I (Sigma‐Aldrich, St. Louis, MO) in RPMI 1640 medium containing 10 mM HEPES, and incubated for 1 h with continuous gentle stirring at 37°C. Then, 10 ml of 1× ice-cold PBS containing 10 mM EDTA was quickly added and incubated for 10 min on ice. Cell suspensions were then vortexed several times and filtered through a 70-µm strainer, washed with 10 ml of HBSS-full (HBSS without Ca²⁺ and Mg²⁺ [Life Technologies] containing 5 mM EDTA, 0.1% BSA, and 10 mM HEPES). Suspensions were then centrifuged at 2000 rpm for 10 min at room temperature. The cell pellets were then resuspended in 5 ml of 30% stock isotonic Percoll (SIP; 100% SIP was prepared as 1 part vol of 10× PBS and 9 parts vol of Percoll [Fisher Scientific]). Suspensions were then carefully loaded on a 37% (5 ml)–70% (5 ml) SIP gradient. Tubes were centrifuged for 30 min at 1000 × g at room temperature with a deacceleration of 0, and enriched leucocytes were collected from the layer between 37 and 70% SIP. Splenocytes, bone marrow cells, and renal leukocytes were then processed for flow cytometry analysis as detailed below. PBMCs (29) were isolated as previously reported. PBMC pellets were snap-frozen and stored at −80°C until processed for RNA extraction and reverse transcription–quantitative PCR (RT-qPCR) as detailed below.

Single-cell suspensions were blocked with Fc receptor blocker CD16/32 (eBioscience) on ice for 10 min and then stained with fluorochrome-conjugated Abs for 15 min in the dark following our previously reported procedures (13, 27). All flow cytometry analysis was done on the live-cell population through Zombie Aqua (BioLegend) staining-based exclusion of dead cells. For the quantification of conventional dendritic cells (cDCs), plasmacytoid DCs (pDCs), and neutrophils, the following anti-mouse primary Abs were used: CD11c-allophycocyanin, CD11b-PerCP-Cy5.5, B220–Pacific Blue, Siglec-H–FITC, Ly6C-allophycocyanin-Cy7, and Ly6G-PE (BioLegend). For T cells, anti-mouse CD3-allophycocyanin, CD4-PerCP-Cy5.5, CD8-PE, CD44-FITC, and CD62L-allophycocyanin-Cy7 (BioLegend) were used. For B cells, anti-mouse CD19–Pacific Blue, CD27-PE, CD138-allophycocyanin, and CD44-PerCP-Cy5.5 (BioLegend) were used. Analysis was conducted using a BD FACSAria Fusion flow cytometer (BD Biosciences). Flow cytometry data were analyzed with FlowJo software.

Total RNA extraction was performed as we previously reported (13, 27). Snap-frozen preweighed kidney tissues at the indicated time points were homogenized in QIAzol lysis reagent (Qiagen) using a TissueLyser II homogenizer (Qiagen). Snap-frozen PBMC pellets were directly processed without homogenization. Total RNA was isolated using a RNeasy Plus universal kit (Qiagen) that also achieved the elimination of genomic DNA. RT was carried out using iScript RT supermix (Bio-Rad). qPCR was performed utilizing the Fast SYBR Green master mix and the ABI 7500 Fast real time-PCR system (Applied Biosystems). Relative quantities of transcripts were calculated using the 2^−ΔΔCt method after being normalized to the 18S rRNA housekeeping gene. Primer sequences are available upon request.

At euthanasia, whole blood samples were collected and allowed to coagulate, then serum samples were obtained and aliquots were stored at −80°C. For both anti-dsDNA IgG and total IgG Abs, ELISAs were performed. Anti-dsDNA IgG levels in diluted serum samples were detected following our previously reported procedures (13). Plates were read at 450 nm using a SpectraMax plate reader (Molecular Devices, Sunnyvale, CA).

Total IgG levels were determined and procedures were performed as previously reported (30) with some modifications. Briefly, ELISA plates were coated with 100 µl per well of 5 µg/ml goat anti-mouse capture IgG Ab (SouthernBiotech) in PBS. After overnight incubation at 4°C, plates were washed twice using 300 µl/well wash buffer–PBS containing 0.05% Tween 20 (PBST). Wells were then blocked by adding 250 µl per well of blocking solution (1% BSA in wash buffer) and incubated for 1 h at room temperature, followed by washing the plates three times and adding 100 µl per well of diluted mouse IgG standard (SouthernBiotech) and serum samples. After a 1-h incubation at 37°C, plates were washed followed by adding 100 µl per well of HRP-labeled goat anti-mouse kappa (SouthernBiotech) and incubated for another 1 h. Plates were then washed and 100 µl per well of tetramethylbenzidine substrate was added and incubated at 37°C for 10 min. Reactions were stopped by adding 100 µl of 2 N sulfuric acid to each well, and the reading the plates was performed as mentioned above.

The assessment of antinuclear Abs (ANAs) in the mouse serum was carried out at euthanasia using a HEp-2 cells ANA kit following the manufacturer’s procedures. Briefly, 20 μl of diluted serum samples, as well as positive and negative controls, were carefully loaded into each of the slide wells. Then, slides were placed in a moist chamber for 30 min at room temperature. After incubation, slides were rinsed thoroughly with PBS and incubated in PBS wash for one 10-min wash, followed by dipping in distilled water to remove the PBS residues. Fluid was drained and wells were immediately flooded with the fluorescent reagent (provided with the kit) and incubated for 30 min at room temperature followed by further washing. Finally, two to three drops of buffered glycerol were added to each slide and gently covered by coverslip for immediate imaging. Images were captured with a Zeiss LSM 880 confocal microscope. Image processing and quantification of the mean fluorescence intensity were performed with ZEN 2.1 Lite software. Analysis of the ANA pattern expression was performed in a blinded fashion using ImageJ software. We defined each pattern according to the demonstration from the ANA testing kit. Four different patterns were characterized as follows: peripheral, homogeneous, cytoplasmic spider web, and fine speckled. Each pattern was given an arbitrary number and counted utilizing the particle tracking function of ImageJ software.

Urinary albumin-to-creatinine ratios were calculated based on levels determined with a mouse albumin ELISA kit (Bethyl Laboratories) and a creatinine colorimetric assay kit (Cayman Chemical). Kidneys were harvested and fixed in 10% formalin. Processing of fixed tissues including paraffin embedding, sectioning, and staining with H&E was performed at the Histopathology Laboratory at the Virginia-Maryland College of Veterinary Medicine. Histopathological evaluation and scoring were performed in a blinded fashion by a certified veterinary pathologist. Slides were assessed using an Olympus BX43 microscope based on a global semiquantitative assessment as previously reported (31). Tubulointerstitial inflammatory lesions were graded on a scale of 0–3 for each of the following categories: peritubular mononuclear infiltrates, tubular damage, interstitial fibrosis, vasculitis, perivascular lymphoproliferative mononuclear cell infiltrate, development of squamous metaplasia, and tubulointerstitial neutrophilic nephritis. Glomerular lesions of at least 50 glomeruli from each kidney per mouse were evaluated and graded on a scale of 0–3 as we previously reported (13).

Gene expression data were processed in R as previously described (32) using Bioconductor packages (GEOquery, affy, affycoretools, limma, and simpleaffy). Unnormalized data were inspected for visual artifacts or poor RNA hybridization using quality control plots (32). Datasets were annotated using custom Brain Array chip definition files, which are maintained by the University of Michigan Molecular and Behavioral Neuroscience Institute (Ann Arbor, MI) and exclude probes with known nonspecific bindings (33). Probes missing gene annotation data were discarded. Raw data from the Affymetrix platform were background corrected and normalized using the guanine cytosine robust multiarray average algorithm. Gene expression data were analyzed with gene set variation analysis (GSVA), a nonparametric, unsupervised method that uses a Kolmogorov–Smirnoff-like rank statistic to estimate the variation in enrichment of predefined gene sets among dataset samples (34). Genes with interquartile range of expression equal to 0 were removed for GSVA analysis. The classical retinoic acid pathway gene set was derived from 26 upregulated genes categorized to be specific direct targets of the classical retinoic acid pathway (category 3) by Balmer and Blomhoff (35).

One-way ANOVA followed by a Tukey’s posttest was performed. For the comparison involving two groups only, a Student t test was employed. Data are shown as mean ± SEM. Significant differences were shown as follows: *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001. For survival analysis, a log-rank (Mantel–Cox and Gehan–Breslow–Wilcoxon) test was applied. Pearson correlation analysis was performed to determine correlations of the immunological changes. All analyses were performed with GraphPad Prism.

Hypovitaminosis A is reported in SLE patients (19), and we have observed a reduced circulatory and hepatic reservoir of retinol/VA in different murine models of lupus with disease progression (13, 14). Therefore, we sought to answer the question of whether VAD could be a driving factor for SLE initiation and/or progression. We first compared the transcript levels of proteins responsible for retinol transfer and cellular uptake including retinol-binding protein 4 (Rbp4) and stimulated by retinoic acid 6 (Stra6), respectively (36). We found reduced levels of these transcripts in the MRL/lpr mice compared to the congenic MRL control (Fig. 1), which suggest disrupted transfer and/or cellular bioavailability of retinol/VA in lupus mice.

Next, we monitored the progression of lupus-like disease in mice deprived of VA through dietary modulation, where VA deprivation could further exacerbate the lack of VA availability. In MRL/lpr mice, whereas lupus nephritis is not apparent before 8 wk of age, the inflammation of lymphoid tissues suggesting initiation of lupus is evident at weaning (37). As pregnancy and lactation could alter VA metabolism and kinetics (38), and because VAD represents a concern during pregnancy (39), we sought to answer the question of whether VAD on the maternal/neonatal interface could further aggravate SLE under genetically prone conditions. We initiated VAD either during the gestation (VAD-D group) or after weaning (mice were derived from VAS dam and then maintained on a VAD diet after weaning; VAS-D group) (Fig. 2A). By the weaning age of 3 wk, pups delivered by VAD dams had significantly larger bodyweight (Fig. 2B). Importantly, the spleen-to-body weight ratios were already markedly higher at the weaning age (Fig. 2C), suggesting that VAD during the embryonic and early neonatal life could potentiate the initiation of SLE immunopathogenesis. Furthermore, all VAD mice also exhibited significantly higher mortality rates (Fig. 2D). Increased mortality was observed starting at 12.4 wk of age in VAD-D mice and 14.4 wk of age in VAS-D mice, compared with no mortality in the VAS-S group. It is noteworthy that switching gestational VAD mice into VA-sufficient diet after weaning (VAD-S group) completely reversed the adverse effect of prior VAD and returned the survival rate back to normal (Supplemental Fig. 1D). These results suggest that VAD initiated either during gestation or postweaning accelerated SLE-associated lymphadenopathy and increased mortality. We went on to focus our investigation on both VAD groups for mechanistic understanding of how predisease VAD contributes to deteriorated lupus.

Lupus nephritis affects up to 60% of SLE patients and represents one of the most serious disease outcomes of lupus (16, 17). Therefore, we further analyzed the histopathological changes of renal inflammation. Even though all VAD mice (both VAS-D and VAD-D groups) had significantly lower glomerular inflammatory scores compared to VAS-S mice (Supplemental Fig. 1E), VAD induced dramatic changes in the tubulointerstitial compartments. VAD mice displayed marked tubulointerstitial nephritis (Fig. 3A) with massive neutrophilic infiltration (Fig. 3B, upper panels) that was also characterized by significant tubular damage (Fig. 3C) and interstitial fibrosis (Fig. 3D). Importantly, VAD significantly compromised the integrity of the renal epithelium (Fig. 3B, lower panels), leading to squamous metaplasia in the renal pelvic urothelium (Fig. 3E). Although squamous metaplasia can occur following VAD in physiologically intact, nonlupus mice as previously reported (40), our findings suggest that VAD could aggravate lupus-associated renal failure that was concomitant with increased mortality in lupus-prone mice. This is supported by the finding that the urinary albumin-to-creatinine ratio was significantly increased in the VAD-D group at 10 wk of age (Fig. 3F). This further supports the potential detrimental role of VAD in aggravating the pre-existing autoimmune inflammation.

Note that we found a significant or nearly significant reduction of the spleen (Supplemental Fig. 2A) and mesenteric lymph node (Supplemental Fig. 2B) to bodyweight ratios at 15 wk of age in VAD-D mice. On the contrary, a trending increase in the kidney-to-bodyweight ratio was noticed in these mice (Supplemental Fig. 2C). This observed shift in lymphoid organ weight to kidney weight (Supplemental Fig. 2D) could indicate a potential infiltration of activated immune cells from lymphoid organs to the kidney in VAD-D mice, as depicted by the massive neutrophilic infiltration in this group of mice (Fig. 3B, upper panels). Future mechanistic investigation will explore the hypothesis that VAD increases the trafficking of activated immune cells from lymphoid organs to the kidney to aggravate lupus nephritis.

Activation of autoreactive B cells and generation of autoantibodies represent a nidus for the development of lupus nephritis. To determine the mechanisms through which VAD deteriorates lupus nephritis, we examined the activation of B cells. Because MRL/lpr is clinically quiescent up to 8 wk of age (37) but the immunogenicity preceding overt signs of lupus nephritis is established by 6 wk of age (14), we examined mice at 7 wk of age, using the same experimental setting, to assess the predisease condition. We found a significant increase in the percentage of splenic plasma cells gated as CD19⁻CD27⁻CD44⁺CD138⁺ (Fig. 4A) with VAD-D over both the control group (VAS-S) and postweaning VAD group (VAS-D). Consistently, these VAD-D mice had significant increases in the levels of anti-dsDNA IgG (Fig. 4B) and total IgG Abs (Supplemental Fig. 2E). The level of anti-dsDNA Abs was positively correlated with the frequency of splenic plasma cells (Fig. 4C), suggesting that VAD had accelerated the early activation of plasma cells leading to increased production of autoantibodies. Because anti-dsDNA autoantibodies could fluctuate with disease flares (41–43) and autoantibodies bound to other nuclear antigens could provide further clinically relevant and diagnostic details (44), we also examined the level of circulatory ANAs. Using the well-established HEp-2 cell immunofluorescent assay, we found increased titers of ANAs in VAD-D mice as indicated by higher fluorescence intensity of IgG-FITC binding to ANAs from diluted serum (Fig. 4D, Supplemental Fig. 2F). Notably, the ANA fluorescence intensity was positively correlated with the frequency of plasma cells (Fig. 4E). Interestingly, VAD-D mice showed ANA expression patterns that are linked to SLE (44, 45), including peripheral and fine speckled antinuclear patterns (Fig. 4F). Importantly, we did not find significant changes in the renal deposition of immune complexes (Supplemental Fig. 2G) as indicated by comparable immunohistochemical staining for IgG Abs (Supplemental Fig. 2H) and C3 complement protein (Supplemental Fig. 2I) at 7 wk of age, supporting the notion that autoantibodies could be generated before any overt symptoms of SLE (46). Taken together, these results indicate that VAD-D accelerated the early expansion of plasma cells and autoantibody production.

T cell alterations have been reported in SLE (47–50), and T cells could promote SLE pathogenesis through potentiating the activation of autoreactive B cells (51) and/or sustaining an inflammatory cytokine milieu (52, 53). Importantly, VA is known to be able to modulate T lineage commitment (54–56), T cell tissue homing capacities (57–59), and their fate (60). All-trans retinoic acid supplementation could mitigate autoimmune and inflammatory disorders through modulating T cell differentiation and inducing their suppressive functions (61–65). In contrast, VAD could induce the differentiation of proinflammatory Th1 cells (66, 67). Therefore, we examined the effect of VAD on T cell populations during the predisease stage. Even though we did not find a significant change in the percentage of CD3⁺ T lymphocytes (Supplemental Fig. 3A), VAD significantly modulated the expansion of splenic T lymphocytes. VAD-D had significantly fewer CD8⁺ T cells (Fig. 5A) that have protective roles in murine lupus (68, 69). Despite no alteration of CD4⁺ T cells (an autoreactive T cell population in SLE [70], Supplemental Fig. 3B), we found a significantly higher CD4/CD8 ratio with VAD-D (Fig. 5B) due to the decrease in CD8⁺ T cells. In parallel, VAD-D significantly increased the expansion of double-negative T cells capable of producing IL-17 to promote lupus (71) (Fig. 5C). Furthermore, VAD-D induced a significant reduction in naive T cells (Fig. 5D) and a prominent increase of effector memory T (T_EM) cells (Fig. 5E). Taken together, these findings indicate that VAD-D accelerated the expansion of splenic and effector T lymphocytes early in the predisease stage. Importantly, this might have contributed to plasma cell activation as evidenced by a strong positive correlation between T_EM (or an inverse correlation between naive T cells) and plasma cells (Supplemental Fig. 3C).

Infiltration of leukocytes such as DCs and monocytes into the kidney has been shown as a principal component of lupus nephritis (72). Our group has previously shown that cDCs accumulate in the renal compartments of MRL/lpr lupus-prone mice with disease progression that promote the activation of renal-infiltrating CD4⁺ T cells, suggesting T cell–dependent mechanisms by which cDCs deteriorate lupus nephritis (28). In this study, we show that VAD-D significantly increased the accumulation of CD11c⁺CD11b⁻B220⁺Siglec-H⁺ pDCs in renal-infiltrating leukocytes (Fig. 6A; gating strategy shown in Supplemental Fig. 3D). VAS-D, in contrast, led to a trending increase of renal pDCs. pDCs can control the initiation of lupus nephritis, where their early and transient depletion could hinder the initiation of renal inflammation associated with lupus (73). Because pDCs could accelerate the initiation of lupus nephritis through secretion of pleiotropic cytokines including IFN-α (74), we measured the transcript levels of pan-IFN-α. We detected a trending increase of its level in renal tissues of VAD-D mice compared to VAS-S controls (Fig. 6B). Interestingly, we also found nearly significant positive correlations between circulatory anti-dsDNA IgG Abs (Fig. 6C) or ANA levels (Fig. 6D) and the frequencies of renal pDCs, suggesting that increased levels of these circulatory autoantibodies could drive the accumulation of pDCs in the renal compartment to initiate local tissue damage. Additionally, we found a significant reduction of pDCs in the spleens (Supplemental Fig. 3E), resulting in a significant increase of renal to the splenic ratio of pDCs in VAD-D mice (Fig. 6E). Taken together, our findings suggest that VAD-D increases the migration of pDCs from the spleen to the kidney, which subsequently initiated renal failure in these mice.

Notably, the differences observed for autoantibodies, plasma cells, T cells, and pDCs were trending but not significant for VAS-D, which are consistent with the intermediate impact of postweaning VAD on mouse survival, lymphadenopathy, and lupus nephritis. This suggests more deleterious effects of VAD at the perinatal period.

To establish relevance of our mouse findings to human lupus nephritis, we employed GSVA to analyze expression of retinoic acid–specific genes at the individual patient level in human lupus nephritis samples. Raw data from microarray analysis of human kidneys were derived from GSE32591, a publicly available dataset, accessed within the Gene Expression Omnibus (GEO) repository (https://www.ncbi.nlm.nih.gov/geo/). GSE32591 comprises samples from control (n = 15) and lupus nephritis (n = 32) renal biopsies for both the glomerular and tubulointerstitial compartments of the kidney. Samples from lupus nephritis classes II (n = 8), III (n = 9), and IV (n = 13) were analyzed. Two patients with noninflammatory class V lupus nephritis were excluded. Results showed that the classical retinoic acid pathway signature, which represents specific genes that have been described as targets of retinoic acid signaling (35), was decreased in the tubulointerstitial region of class III/IV lupus nephritis as compared to controls, whereas expression in the glomerulus was unchanged (Fig. 7). This suggests that VAD specifically impact tubulointerstitial lupus nephritis in both human and mouse SLE.

VAD is reported in both SLE patients (19) and murine models of lupus (13, 14). VA metabolism and kinetics are altered during pregnancy and lactation, making VAD a public health concern for women of childbearing age (75) who are also more susceptible to aggressive lupus flares (76–78). However, whether VAD could affect the initiation and/or progression of SLE under genetically prone conditions was unclear. Specifically, whether the potentially detrimental effects of disrupted VA levels on the maternal/neonatal interface could drive more severe lupus flares in susceptible progenies was unknown. In this study, we utilized the genetically-prone classical murine lupus model MRL/lpr to unravel the effects of VAD on SLE immunopathogenesis and, in particular, SLE-associated renal inflammation or lupus nephritis. Feeding MRL/lpr mice with a VAD diet either starting in utero or postweaning resulted in progressive renal failure with marked neutrophilic tubulointerstitial nephritis. Although the glomerular pathology was markedly less in all VAD mice, tubulointerstitial nephritis was concomitant with accelerated mortalities. Indeed, it is the severity of tubulointerstitial nephritis, not glomerular injury, that is associated with a greater risk for renal failure in SLE patients (79). This suggests a potential detrimental effect of hypovitaminosis A on SLE prognosis. Interestingly, these determinantal outcomes were more prominent when the manipulation of VA levels was initiated during the prenatal stage. Mechanistically, we propose that VAD potentiated immunopathogenic leukocyte autoreactivity before apparent disease onset by increasing the infiltration of autoreactive cells into the kidney, thereby accelerating renal damage.

VA can modulate lymphocyte functions and their fate (54–56, 80). We found that before the onset of clinical manifestations of lupus, VAD induced the expansion and activation of different immunopathogenic cells associated with exacerbation of SLE, including autoreactive B lymphocytes. VAD has been shown to induce abnormal Ig isotypes such as elevated IgG2b levels (40), which could be detrimental in lupus (81). We also found that VAD led to the expansion of autoreactive plasma cells and potentiated their production of various ANAs with distinct patterns associated with SLE. This was evident when the mice were 7 wk of age, suggesting activation of B cells as a nidus for accelerated renal failure in VAD mice. Indeed, nephritogenic autoantibodies could form immune complexes with self-antigens that subsequently deposit in the kidney to initiate renal damage in SLE (2, 82, 83). Because T cells could contribute to the activation of autoreactive B lymphocytes (51) to initiate this circuit, we also examined how VAD affected splenic T lymphocyte phenotype and function. Interestingly, by 7 wk of age, VAD significantly induced a higher ratio of CD4⁺ to CD8⁺ T cell populations. This indicates that VAD might have potentiated CD4⁺ T cell–driven help to B cells that possibly accelerated their activation. Importantly, VAD-D potentiated the expansion of T_EM cells, indicating that depletion of VA might lead to persistence of B cell autoreactivity to self-antigens, where T_EM cells could constantly recognize the self-antigens leading to accelerated B cell reactivation (84). This notion is supported by the strong positive correlation between the frequencies of splenic T_EM and plasma cells. Our future studies will aim to mechanistically delineate the potential interactions between different T cell subsets and B cells and how this could contribute to exacerbated tubulointerstitial lupus nephritis under VAD conditions.

VAD was previously reported to enhance the systemic accumulation of innate immune cells such as myeloid cells by hindering their apoptosis (85). In the current work, we did not find changes in the frequencies of different myeloid cells including neutrophils (CD11b⁺Ly6C⁺Ly6G⁺), macrophages (CD11b⁺F4/80⁺), and CD11b⁺ and CD11b⁻ cDCs (CD11c⁺B220⁻Siglec-H⁻) in the bone marrow, splenic, and renal compartments with VAD (data not shown). However, we found a significant reduction of splenic pDCs (CD11c⁺CD11b⁻B220⁺Siglec-H⁺) and, in parallel, a significant increase of their frequency in the kidney of VAD mice. Additionally, higher transcript levels of their major cytokines (pan-IFN-α) were detected in the kidney of VAD mice. Our group has previously shown that expansion of pDCs by 6 wk of age is an early indication of the start of immunogenicity in MRL/lpr mice (14). Indeed, pDCs are capable of driving the initiation of lupus nephritis, as early and transient depletion of pDCs could hinder disease onset in murine lupus (73). In parallel, abnormalities in circulating pDCs have been reported in SLE patients (86). pDCs have been shown to relocate from the periphery to the kidney of SLE patients, and their renal accumulation correlates with lupus nephritis (87). Similarly, pDCs patrolling the renal tubulointerstitium have been suggested to be responsible for nephritis pathogenicity in SLE patients (88). Therefore, our findings suggest that VAD might target the relocation of pDCs from lymphoid tissues (represented by the spleen) to the kidney to initiate renal damage. Notably, pDCs could also activate autoreactive B cells (89) and T cells (86, 87, 90) to drive autoimmunity. Future studies will explore whether VAD intrinsically modulates plasma cell autoreactivity, or extrinsically, for example, through the activation of pDCs to drive plasma cell autoreactivity.

Collectively, our findings indicate that disrupted VA levels could be a driving factor for tubulointerstitial lupus nephritis. Under genetically predisposed conditions, where VA availability or signaling could be impaired, further VAD can accelerate early (predisease) immunopathogenesis leading to more progressive tubulointerstitial nephritis during established disease. Remarkably, consistent with our mouse data, analysis of gene expression in human kidney reveals that an impaired retinoic acid pathway is potentially implicated in tubulointerstitial lupus nephritis.

These results highlight the importance of maintaining an optimal level of retinol/VA in SLE. Future studies will elucidate the underlying mechanisms of how VAD modulates cellular functions to exacerbate tubulointerstitial lupus nephritis.

We thank Melissa Makris for the use of the Flow Cytometry Core Facility, and Kristi Decourcy for use of the Fralin Imaging Core Facility at Virginia Tech.

The authors have no financial conflicts of interest.