Progesterone receptor membrane component 1 inhibits tumor necrosis factor alpha induction of gene expression in neural cells

Progesterone membrane receptor component 1 (Pgrmc1) is a cytochrome b5-related protein with wide-ranging functions studied most extensively in non-neural tissues. We previously demonstrated that Pgrmc1 is widely distributed in the brain with highest expression in the limbic system. To determine Pgrmc1 functions in cells of these regions, we compared transcriptomes of control siRNA-treated and Pgrmc1 siRNA-treated N42 hypothalamic cells using whole genome microarrays. Our bioinformatics analyses suggested that Pgrmc1 plays a role in immune functions and likely regulates proinflammatory cytokine signaling. In follow-up studies, we showed that one of these cytokines, TNFα, increased expression of rtp4, ifit3 and gbp4, genes found on microarrays to be among the most highly upregulated by Pgrmc1 depletion. Moreover, either Pgrmc1 depletion or treatment with the Pgrmc1 antagonist, AG-205, increased both basal and TNFα-induced expression of these genes in N42 cells. TNFα had no effect on levels of Rtp4, Ifit3 or Gbp4 mRNAs in mHippoE-18 hippocampal control cells, but Pgrmc1 knock-down dramatically increased basal and TNFα-stimulated expression of these genes. P4 had no effect on gbp4, ifit3 or rtp4 expression or on the ability of Pgrmc1 to inhibit TNFα induction of these genes. However, a majority of the top upstream regulators of Pgrmc1 target genes were related to synthesis or activity of steroids, including P4, that exert neuroprotective effects. In addition, one of the identified Pgrmc1 targets was Nr4a1, an orphan receptor important for the synthesis of most steroidogenic molecules. Our findings indicate that Pgrmc1 may exert neuroprotective effects by suppressing TNFα-induced neuroinflammation and by regulating neurosteroid synthesis.

Introduction Progesterone receptor membrane component 1 (Pgrmc1) is an ancient and somewhat enigmatic molecule with a diverse range of functions and multiple intracellular locations. Structurally, it is a 28-kDa protein with an N-terminal extracellular region, a single transmembrane domain and a cytoplasmic region. The cytoplasmic region contains a cytochrome b5-like heme-binding domain that allows interaction with a number of steroidogenic and drugmetabolizing cytochrome P450 enzymes [1][2][3][4]. As suggested by its name, Pgrmc1 also contains a high-affinity progesterone (P 4 ) binding site [5][6][7][8][9], most likely within the transmembrane domain and the initial segment of the C terminus [10] and near the heme-binding site [7]. Finally, the Pgrmc1 molecule contains sites that may allow interaction with SH2-and SH3-domain signaling proteins [11].
In view of these diverse molecular structural features, it is not surprising that Pgrmc1 has been implicated in such wide-ranging functions as steroid synthesis [4], heme sensing and Ssynthesis [12], regulation of fatty acid 2 hydroxylase [13], stabilization of tyrosine kinase receptors in cell membranes [14,15], suppression of p53 and Wnt/β-catenin pathways [16], inhibition of ovarian granulosa cell apoptosis [17] and promotion of breast cancer cell survival and tumor growth [18]. Most studies delineating Pgrmc1 functions have been performed in non-neural tissues and cell lines, but this molecule is also widely distributed throughout the brain [19][20][21][22]. Unfortunately, information about Pgrmc1 in neural cells is rather sparse and no unifying concept regarding its neural functions or signaling pathways has emerged.
To better understand the molecular and cellular functions of Pgrmc1 in neural cells, we compared transcriptomes of hypothalamic N42 cells with and without Pgrmc1 knockdown. We independently verified our findings using QPCR and used several bioinformatics tools to identify pathways and neural processes likely regulated by Pgrmc1. Results of these analyses suggest that Pgrmc1 blocks expression of genes downstream of proinflammatory cytokines in a P 4 -independent manner.
We also used QPCR to examine effects of Pgrmc1 knockdown on Pgrmc1 mRNA levels in N42 cells. We isolated RNA using the RNeasy Mini Kit (Qiagen) and reverse-transcribed using the QuantiTect Reverse Transcription Kit (Qiagen). QPCR was performed in a Stratagene Mx3000P thermocycler (Agilent Technologies; Wilmington, DE) programmed as follows: 95 o C, 10 min; 40 cycles of 95 o C for 15 sec; 60 o C for 60 sec; and 72 o C for 60 sec. We used the QuantiTect SYBR Green Kit (Qiagen) for QPCR following the manufacturer's protocol.
We determined primary efficiency over a range of cDNA concentrations and included samples with no cDNA as negative controls. Primer specificity was validated by observing a single fluorescence peak in each QPCR reaction and also using 2% agarose gel electrophoresis to verify that single products were obtained following the reaction. Further verification entailed a melting curve analyses in which samples were heated to 95 o C and fluorescence measurements recording at incremental increases of 0.5 o C for 80 cycles.
Data for each sample and controls were obtained using MxPro QPCR analysis software (Agilent Technologies). We used the ΔΔCt method to compare treatment and control samples [25], converting data to percent control in each pairwise comparison and calculating means for treatment groups (Pgrmc1 siRNA vs scramble control siRNA). Means of treatment pairs were compared using student t-tests.

Affymetrix microarray analysis
For these studies, we chose to use microarray technology rather than RNA-Seq because the former is better suited for detecting genes with relatively low expression levels [26]. RNA from N42 cell siRNA experiments described above was used for Mouse Genome 430 2.0 Gene Chip assays (Affymetrix; Palo Alto, CA) performed by the Keck Microarray Institute at Yale University. RNA quality was assessed using an Agilent 2100 Bioanalyzer and RNA 6000 Nano Lab-Chips (Agilent Technologies).
For each treatment group (cells transfected with Pgrmc1 siRNA or with control scramble siRNA), we used three pools of RNA extracted from three replicate studies of transfected N42 cells. The quality of total RNA in each sample was assessed using a Nanodrop Spectrophotometer (Thermo Scientific; Wilmington, DE) and a 2100 Agilent Bioanalyzer (Agilent Technologies). Samples were accepted for analysis if the 260/280 ratio was at least 1.8 and the RNA Integrity Number was greater than 8.0. Preparation of labeled cRNA for hybridization onto Affymetrix GeneChips followed the recommended Affymetrix protocol.
The Yale Center for the NIH Neuroscience Microarray Consortium carried out microarray analysis. Double-stranded cDNA was synthesized from 1 to 5 μg of total RNA using a Superscript Choice System (Life Technologies; Carlsbad, CA) with an HPLC-purified oligo (dT) primer containing a T7 RNA polymerase promoter sequence at the 5'-end (Proligo LLC; Boulder, CO). We synthesized the second cDNA strand using E. Coli DNA polymerase I, RNase H and DNA ligase, and then generated labeled cRNA using a GeneChip IVT labeling kit (Affymetrix) following the manufacturer's instructions. The labeling procedure incorporated biotinylated synthetic analogs by using a pseudouridine reagent and MEGAscript T7 RNA Polymerase (Life Technologies). Biotin-labeled cRNA was purified using GeneChip cleanup module (Affymetrix). The cRNA was incubated at 94 o C for 35 min in fragmentation buffer and the resulting 35-to 200-base fragments were hybridized to the arrays for 16 h at 45 o C.
After hybridization, we washed arrays using an Affymetrix fluidics station and stained them with streptavidin-phycoerythrin (10 μg/ml; Life Technologies). Arrays were inspected for hybridization artifacts and then scanned with an Affymetrix GeneChip Scanner 3000. Images were analyzed using Affymetrix Microarray Suite 5.0, scaling to a target average intensity. Quality controls included sense strand probes and three housekeeping genes (gapdh, hexokinase and β-actin). We also evaluated spiked controls, background values, scanner noise (Q value) and scaling factors. The CEL files containing levels of probe intensities were analyzed with Partek Genomics Suite, version 6.15 (Partek; St Louis, MO) with the Robust Multi-array Average method of normalization. P values were corrected for multiple hypothesis testing using the Benjamini-Hochberg method to control for false discovery. Finally, we performed hierarchical clustering analyses of genes regulated by at least 1.2-fold with p<0.05.

Interrogation of microarray data
We used multiple bioinformatics strategies to identify P 4 -independent pathways likely regulated by Pgrmc1 in neural cells. Integrating data generated with tools that use different algorithms and databases allowed us to develop a more comprehensive mechanistic model.
Identification of the most highly regulated genes. We first identified genes regulated by at least 1.2-fold and with a significance of p<0.05 in each of the two comparison groups (S2 Dataset). We then used GeneCards (www.genecards.org) and NIH Gene (www.ncbi.nlm.nih. gov/gene) to determine general functions of these most highly regulated genes.
Pathway analysis. We compared the transcriptomes of Pgrmc1 siRNA-and scramble siRNA-treated N42 cells with Gene Set Enrichment Analysis (GSEA; http://www.broad institute.org/gsea/index.jsp), software that allows detection of sets of affected genes overrepresented in specific pathways without consideration of the fold change of genes induced by treatment. We used the Hallmark Gene Set from the Molecular Signatures Database of genes known to be involved in specific biological or biochemical relationships or shown to be coexpressed or co-regulated. The Hallmark Gene Sets include 50 sets of genes that represent well characterized biological processes (http://www.broadinstitute.org/gsea/msigdb/collections. jsp).
We also compared transcriptomes of Pgrmc1 siRNA-treated and control cells using Ariadne Pathway Studio (www.elsevier.com/solutions/pathway-studio-biological-research). The Pathway Studio program is based on a manually curated database of biological relationships derived from published articles beyond the public domain [27].
Ingenuity upstream regulator analysis. We used the Ingenuity Pathway Analysis (Qiagen; https://www.qiagenbioinformatics.com/blog/discovery/publication-roundup-ingenuitypathway-analysis-3 to identify upstream transcriptional regulatory cascades linked to the gene changes observed in response to Pgrmc1 knockdown in the absence or presence of P 4 . For this analysis, we selected the genes that changed by at least 1.2-fold and at a significance level of p<0.01. Database for Annotation, Visualization and Integrated Discovery (DAVID) analysis. We used DAVID Bioinformatics Resources 6.8 program [28,29] to functionally categorize genes into clusters and determine the fold-enrichment of genes in these clusters. DAVID uses different databases than GSEA and simplifies the process of determining the likely biological significance of microarray data by integrating gene identifier and functional category terms from multiple bioinformatics databases. DAVID uses different databases than GSEA, including Kyoto Encyclopedia of Genes and Genomes, Clusters of Orthologous Groups, Gene Ontology Consortium and the PANTHER Classification System. DAVID converted the data to DAVID IDs and removed redundancies so that genes detected with multiple probes were not overrepresented in the analysis. We used Functional Annotation Clustering analysis with default databases and stringency settings except that we increased the Kappa Similarity Overlap to 3, Similarity Threshold to 0.7 and the Final Group Membership to 3 in order to minimize duplication of genes in different clusters.

QPCR verification of microarray findings
We used QPCR to verify changes in gene expression of targets identified in our microarray analysis. For QPCR studies, we transfected N42 cells with scrambled control siRNA or Pgrmc1 siRNA as described for microarray studies. We then isolated mRNA, performed reverse transcription, measured mRNA levels and verified Pgrmc1 knockdown using QPCR and western blots as described above. Primer pairs for gene targets (Table 1) were based on sequences from Table 1. Primers used for QPCR validation of targets identified on microarrays to be significantly regulated by Pgrmc1 knockdown. . Primary efficiency and specificity were verified as described above. Data for each sample and controls were obtained using MxPro QPCR analysis software (Agilent Technologies). We used the ΔΔCt method to compare treatment and control samples [25], converting data to percent control in each pairwise comparison and calculating means for treatment groups (Pgrmc1 siRNA vs scramble control siRNA). Means of treatment pairs were compared using student t-tests.

Examination of Pgrmc1 antagonist and P 4 effects on expression of selected genes
To determine whether administration of the Pgrmc1 antagonist, AG-205, altered expression of genes found to be regulated by Pgrmc1, cells were seeded at a density of 50,000 cells/well in 24-well plates and allowed to grow for 48 h in 5% CO 2 at 37 o C. They were treated with 10 μM AG-205 (Sigma; optimal dosage determined in preliminary studies) or vehicle (cell culture grade DMSO) for 24 h. To determine whether these genes were also regulated by P 4 , we treated N42 cells with either P 4 (10 or 100 nM) or vehicle for 8 h. RNA was then isolated, reverse transcribed and the cDNA was then used as a template in QPCR studies as described above.

Examination of Pgrmc1 effects on TNFα induction of genes
Effects of Pgrmc1 knockdown or AG-205 on TNFα-induced expression of genes. We tested whether TNFα affects ifit3, gbp4 and rtp4 gene expression and whether Pgrmc1 depletion or blockade changes the responses to TNFα. We chose these targets because they were among the genes most highly regulated by Pgrmc1 knockdown, and they are known targets of cytokines [30][31][32][33]. To deplete cells of Pgrmc1, we transfected N42 cells with 10 nM Pgrmc1 siRNA or negative control siRNA, and 40 h later treated cells with 10 ng/ml (0.59 nM) TNFα for 8 h as described above. In a separate study, we treated N42 cells with 10 μM AG-205 as described above, incubating the cells for 24 h. RNA was isolated, reverse-transcribed and cDNA used for measurement of Ifit3, Gbp4 and Rtp4 mRNAs.
Effects of Pgrmc1 overexpression on TNFα-induced expression of genes. To determine the effects of overexpression of Pgrmc1 on TNFα-dependent gene expression, we constructed a Pgrmc1 expression vector with puromycin resistance using PCR. We amplified the mouse Pgrmc1 coding sequence with primers that also contained regions homologous to the PstI restriction enzyme cut site region on the 5' ends. The forward primer was: AAC CGG ATC CTC TAG AGT CGA TGG CTG CCG AGG ATG TGG TGG CG and the reverse primer was: CCC CAA GCT TGC ATG CCT GCT CAT TCA TTC TTC CGA GCT GTC. The pBApo-CMV Pur plasmid (Clontech) was digested with PstI (Promega) and ligated to the Pgrmc1 PCR product using the NEBuilder HiFi DNA Assembly Master Mix (New England Biolabs; Ipswich, MA). We then transformed competent E. coli with the resulting Pgrmc1 expression vectors and isolated the plasmids using a Qiagen Plasmid Midi Kit. We confirmed correct insertion of the full coding sequence of Pgrmc1 by sequencing. To generate the stable cell line, N42 cells were plated at a density of 200,000 cells/well in 6-well plates. The following day, cells were transfected with Pgrmc1 expression vector or empty vector using Genejuice transfection reagent (EMD Millipore) following manufacturer's directions. After 48 h, cells were trypsinized and split 1:4 in DMEM containing 3 μg/ml puromycin hydrochloride (Cayman Chemical; Ann Arbor, MI). Puromycin-containing media was changed every three days to allow for selection of stably-expressing cells. We verified overexpression of Pgrmc1 with western blots using the antibody and procedures described above in studies verifying Pgrmc1 downregulation by siRNA.
The resulting N42 cells stably expressing empty or Pgrmc1-containing vector were plated in 24-well plates at a density of 25,000 cells/well and grown in DMEM containing 1 μg/ml puromycin. Forty-eight h later, the cells were treated with 10 ng/ml mouse TNFα (Invitrogen) or vehicle (sterile 0.5% BSA in PBS) for 8 h. After lysing the cells in TRIzol (Invitrogen), RNA was extracted, reverse-transcribed and used in QPCR analyses as described above.
For these studies, means were compared among groups using Two-Way ANOVA with Pgrmc1 and TNFα treatments as main effects. Tukey's multiple comparison test was used for post hoc analyses.
Effects of P 4 on TNFα-induced expression of target genes. Considering that P 4 is a ligand of Pgrmc1, we tested whether TNFα-induction of ifit3, gbp4 and/or rtp4 expression was affected by P 4 as it is by Pgrmc1 manipulations . We cultured N42 cells as described above, then treated with either P 4 (10 or 100 nM) or vehicle for 1 h before treating with TNFα or vehicle for 8 h. RNA was then isolated using TRIzol (Invitrogen) using the manufacturer's protocol. RNA was reverse transcribed with MMLV-RT (Promega) in the presence of RNAsin (Promega) to prevent RNA degradation. The resulting cDNA was then used as a template in QPCR studies using the appropriate gene-specific primers (Table 1). Data were analyzed as described above.
Effects of Pgrmc1 on cytokine-induced expression of target genes in a hippocampal cell line. To determine whether the effects of Pgrmc1 on expression of gene targets of pro-inflammatory cytokines is similar in hippocampal cell lines, we used mHippoE-18 cells (CELLutions Biosystems, Inc.). Cells were grown, transfected with Pgrmc1 or negative control siRNA and treated with TNFα as described above. RNA was extracted, reversed-transcribed and cDNA was used to determine effects of Pgrmc1 on Ifit3, Gbp4 and Rtp4 mRNA levels. Data were analyzed using Two-Way ANOVA with Pgrmc1 and TNFα as main effects followed by Tukey's multiple comparison test.  Table 2). Eleven of these genes have been linked to immune/inflammatory response functions, five to Jak/Stat signaling and three to steroid hormone synthesis or signaling (NIH Gene; www.ncbi.nlm.nih. gov/gene and GeneCards; www.genecards.org).

Microarray analyses identified several Pgrmc1-regulated functions
An unbiased GSEA Hallmark study of the 19301 individual genes (S1 Dataset) identified 14 pathways and processes with NOM p and FDR q values of p<0.05 (Table 3). Nearly half of the pathways and processes were related to Jak/STAT, TNFα, cytokines (interleukins and interferons) and inflammatory regulators (Table 3; shown in bold text). Similarly, results of our Ariadne pathway analysis identified five pathways significantly (p<0.05) regulated by Pgrmc1 knockdown and four involved STAT signaling (Table 4). Table 5 shows results of our IPA study to identify upstream regulators of the same genes that are altered by Pgrmc1 knockdown. Targets were considered significant if the p value of the overlap between dataset genes and genes regulated by a putative upstream regulator was <0.05. A majority of the targets were steroids or molecules related to steroid action or synthesis. In addition, more than half have been linked previously to Pgrmc1 or P 4 , increasing confidence in our data and analyses. DAVID 6.8 Functional Clustering analysis identified 2112 DAVID IDs and three significantly (p<0.05) enriched Annotation Clusters: Guanylate Binding (11.8-fold enrichment), MAPK Activity (1.8-fold enrichment) and SMAD Protein Signal Transduction (1.4-fold enrichment).

QPCR studies verified microarray findings
We verified a significant (p<0.0001) decrease in Pgrmc1 mRNA levels after treatment of N42 cells with Pgrmc1 siRNA (scramble control = 100.0 ± 4.32; Pgrmc1 siRNA treated = 49.5 ± 4.25). Fig 3 shows results of QPCR studies using RNA from N42 cells treated with Pgrmc1 siRNA. Fig 4 shows similar studies verifying that Pgrmc1 knockdown also upregulates three proinflammatory cytokine gene targets, gbp4, ifit3, and rtp4, that were also among the genes identified as top targets on the microarray. Ten of 11 genes significantly regulated by Pgrmc1 in microarray analysis were verified by independent QPCR validation studies.

Pgrmc1 inhibits TNFα-induced upregulation of genes in N42 hypothalamic cells independent of P 4
Treatment of N42 cells with TNFα did not alter Pgrmc1 mRNA levels; however, Pgrmc1 knockdown significantly increased both basal and TNFα-induced expression of gbp4, ifit3 and  rtp4 (Fig 6). Treatment of cells with the Pgrmc1 antagonist, AG-205, increased basal expression of ifit3 and rtp4 and significantly enhanced the ability of TNFα to induce expression of gbp4, ifit3 and rtp4 (Fig 7). Conversely, Pgrmc1 overexpression in stably transfected N42 cells (Fig 8) completely blocked the ability of TNFα to induce expression of ifit3, gbp4 and rtp4 genes (Fig 9). P 4 had no effect on the ability of TNFα to increase the expression of these genes (Fig 10).

Pgrmc1 inhibits cytokine upregulation of genes in mHippoE-18 hippocampal cells
As shown in Fig 11, Pgrmc1 knockdown also increased both basal and TNFα-induced expression of gbp4, ifit3 and rtp4 in hippocampal cells without altering pgrmc1 expression. However, in contrast to our findings in N42 cells, TNFα did not significantly increase expression of any of these genes in control cells that contained Pgrmc1.

Discussion
These microarray and bioinformatics findings provide evidence that Pgrmc111 acts independently of P 4 to regulate signaling important for neuroimmune functions. Among the genes most highly upregulated by Pgrmc1 depletion were gbp4, ifit3, and rtp4, genes induced by Table 3. Results of GSEA analysis (Hallmark Collection database) of pathways and biological processes that differed significantly between N42 cells transfected with scramble or Pgrmc1 siRNA. Pathways were considered significant if the normalized enrichment score was at least 1.5-fold and nominal P values and false discovery rate Q values were significant (p<0.05). Pathways involved in inflammatory processes are shown in bold text. proinflammatory cytokines such as interferons, interleukins and Tnfα [30][31][32][33]. Our follow-up studies verified that at least one of these cytokines, TNFα, upregulated expression of gbp4, ifit3, and rtp4 and expression was inhibited by Pgrmc1 in both hypothalamic and hippocampal cell lines. The effect of Pgrmc1 on the expression of these genes was independent of P 4 , but our analysis of upstream regulators suggests that Pgrmc1 may alter synthesis or signaling of steroids that alter neuroinflammatory signals [34]. Overall, our findings provide new insights into how Pgrmc1 may exert neuroprotective effects. Pathway analyses of the entire transcriptomes of Pgrmc1 siRNA-and scramble siRNAtreated N42 cells identified pro-inflammatory signaling pathways as regulatory targets of Pgrmc1. In addition, 60% of the most highly regulated genes in our data set were downstream of proinflammatory cytokines. It is unlikely that our findings are due primarily to activation of an innate immune response to siRNA for several reasons. First, gbp4, ifit3, and rtp4 were upregulated by a Pgrmc1 antagonist in cells not treated with siRNA. Moreover, basal levels of expression were significantly higher in cells treated with Pgrmc1 siRNA than in those treated with scramble control siRNA. Finally, cells engineered to constitutively overexpress Pgrmc1 had lower levels of Gbp4 and Ifit3 mRNA. Instead, our findings indicate that Pgrmc1 suppresses proinflammatory cytokine signaling in neural cells.

Pathway or Process
One of the identified cytokines, TNFα, is particularly abundant in hypothalamic and hippocampal regions [35][36][37] wherein pgrmc1 expression is also highest [21]. TNFα regulates a wide range of physiological functions controlled by the hypothalamus and hippocampus including sleep [38], food intake [39,40], learning, memory and anxiety-like behaviors [41]. Moreover, dysregulation of these functions can occur when production of TNFα increases in response to infection or injury [40]. Little is known about Pgrmc1 regulation of these functions under normal conditions, but our findings suggest that Pgrmc1 may mitigate pathological effects of elevated TNFα levels by repressing expression of cytokine effector genes.
A neuropathology that involves TNFα-mediated neuroinflammation and Pgrmc1 signaling is Alzheimer's disease (AD), a progressive neurological disease that results in cognitive impairment and memory loss. There is some debate about the primary pathophysiology underlying AD. According to the amyloid cascade hypothesis [42], the pathology begins with dysregulation of amyloid beta (Aβ) production and clearance. Aβ deposition, in turn, is thought to disrupt synaptic functions and induce a cascade that eventually produces neuronal loss and deficits in neural transmission underlying impaired memory and cognition. Others Table 5. Top 10 upstream regulators of genes also regulated by Pgrmc1 in N42 hypothalamic cells. Regulators shown in boldface type are steroids, steroid receptors or molecules known to regulate steroid synthesis or activities. Those shown in italic boldface are known to be regulated by Pgrmc1 or P 4 . argue that, rather than amyloid deposition triggering the disease, it may be a result of neuroinflammation mediated by TNFα [43][44][45]. Previous work combined with findings reported herein show that Pgrmc1 can impact both Aβ deposition and TNFα signaling. The literature in this area is complicated because Pgrmc1 was thought for several years to be the same molecule as the sigma 2 receptor [46]. It is now clear that they are separate molecules encoded by  two different genes [47][48][49], but both alter Aβ deposition and TNFα signaling. Importantly, Pgrmc1 antibodies or Pgrmc1 siRNA knockdown, as well as antagonists to sigma 2 receptor, prevent Aβ deposition and improve synaptic functioning [50,51]. Thus, they are logical therapeutic targets for AD treatment. However, such treatments may be complicated because in the present studies we found that Pgrmc1 likely represses TNFα signaling and others showed that sigma 2 agonists inhibit tnfα gene expression [52]. Consequently, while Pgrmc1 and sigma 2 antagonists decrease Aβ binding to neurons, they may also block the protective effects of Pgrmc1 and sigma 2 agonists exerted through TNFα inhibition. Further studies will be required to determine how Pgrmc1 and sigma 2 receptor interactions might be manipulated to prevent both Aβ deposition and TNFα-dependent inflammation associated with AD. The downstream targets of TNFα are not entirely clear in neurons, but our work shows that TNFα upregulates ifit3, gbp4 and rtp4, genes among the most highly upregulated when Pgrmc1 was depleted. In addition, both Pgrmc1 depletion and treatment with the Pgrmc1 antagonist, AG-205, enhanced the ability of TNFα to activate expression of ifit3, gbp4 and rtp4. Unfortunately, these genes have not been studied extensively in neural cells, so it is difficult to predict how they may affect cellular functions during the inflammatory process. Ifit3 (also known as retinoic acid induced gene G protein) is a Jak-Stat regulated target of interferon and retinoic acid that blocks cell proliferation by upregulating p21 [53]. The other two genes upregulated by TNFα and suppressed by Pgrmc1, gbp4 and rtp4 each play a role in G protein signaling. Rtp4 chaperones G-protein coupled receptors to the cell surface [54,55] and Gbp4 hydrolyzes GTP to both GDP and GMP [56]. Rtp4 also increases trafficking of mu-delta heterodimeric opioid receptors to the cell surface, thereby changing the cellular responses to endogenous ligands [55]. Both mu and delta opioid agonists repress TNFα production [57,58], suggesting a possible feedback mechanism in the system. Further studies are necessary to determine what role these genes might play in neuroinflammation.

Upstream Regulator Direction of Change Symbol Molecule Type P-Value of Overlap
Although P 4 did not affect basal or TNFα induction of Pgrmc1 target genes, one interpretation of our bioinformatics data is that Pgrmc1 indirectly exerts neuroprotective effects by modulating neurosteroid synthesis or signaling. Interrogation of the entire Pgrmc1-dependent transcriptome without considering fold-change determined that over half of the top 20 upstream regulatory pathways identified involved steroid signaling or synthesis. It was beyond the scope of this paper to examine whether steroidogenesis occurs in N42 cells or if Pgrmc1 might alter the process, but previous work shows that Pgrmc1 directly binds to and activates cytochrome P450 enzymes critical for steroid synthesis [1,2,4,59]. In addition, we found that Pgrmc1 regulates expression of nr4a1 (also known as nur77, tr3, nak-1 and ngf1b), an immediate early gene required for the induction of a number genes encoding steroidogenic enzymes [60][61][62][63]. Thus, an important question to explore in future studies is whether Pgrmc1 alters synthesis of neurosteroids including progesterone, estrogen or steroid metabolites that are neuroprotective [63].
In summary, our findings suggest that Pgrmc1 may exert neuroprotective effects by blocking TNFα induction of downstream gene targets including gbp4, ifit3, and rtp4. In addition, Pgrmc1 may alter neuroinflammation by regulating neurosteroid synthesis, both directly through steroidogenic enzyme activation and indirectly through nr4a1 regulation of steroidogenic enzyme expression. Our findings provide important new information about actions of