Progesterone-induced Warburg Effect is Regulated by Cell-type-specic Interaction of Progesterone Receptor Membrane Component 1 and Hexokinases

Background: Progesterone receptor membrane component 1 (PGRMC1) is a non-canonical progesterone (P4) binding protein. PGRMC1 is elevated in a variety of cancers and its phosphorylation state associated with hormone responsiveness in breast cancer. Metabolic reprogramming is a key factor for tumor growth during malignancies. Recently, we reported that the P4-inducedWarburg effectinHEK293cells is associated with altered post-translational modications (PTMs) of PGRMC1, including phosphorylation, SUMOylation, and ubiquitination, which were linked to rapid proteasomal degradation of the protein. The previous study also identied hexokinase (HK) as a potential novel interacting partner of PGRMC1. HKs catalyze the rst essential step of glucose metabolism and directly couple glycolysis to mitochondrial respiration. Therefore, in the present study, P4’s effects on glycolysis and PTMs of PGRMC1 as well as its interaction with HKs were compared between HEK293 and HepG2 cells to unravel the signaling pathways that mediate cell-type-specic metabolic reprogramming. Methods: P4-induced glucose metabolism in wild-type and PGRMC1-decient cells wasassessed using the Seahorse ux analyzer, while PTMs of PGRMC1/HKs and protein-protein interaction were studied using immunoprecipitation, isoelectric focussing, phosphomimetics, and mass spectrometry.Translocation of HKs to different subcellular organelles were studied using subcellular fractionation, and the cell-type-specic effect of PGRMC1-deciency on endoplasmic reticulum (ER) and mitochondria; ultrastructure were examined by electron microscopy. Results:P4 treatment caused a rapid increase in glycolysis in HEK293 cells, whereas it decreased glycolysis in HepG2 cells. In addition, PGRMC1 was not degraded in HepG2 cells which is in contrast to HEK293 cells where rapid proteasomal degradation of PGRMC1 Effect and (AB, EF, IJ, & MN): Immunoblots showing an abundance of HK1 and HK2 in the ER/mitochondrial fractions (F1 fractions in HepG2 cells) and nuclear fractions from the vehicle and P4-treated HEK293 and HepG2 cells.Veh: vehicle; Tx: Treated with 10 (cid:0) M P4 for 30 mins. (CD, GH, KL, & OP): Scatter plots showing the relative abundance of HK1 and HK2 in the ER/mitochondrial fractions (F1 fractions in HepG2 cells) and nuclear fractions from the vehicle and P4-treated HEK293 and HepG2 cells. The oriole-stained gel images in the bottom panels of B, F, J&N represent total protein loadings. N=9 replicates from 3 independent experiments. P-value by t-test (unpaired).

isoforms are frequently referred to as mitochondrial HKs, as they differ from other isoforms by their unique capacity to directly interact with the voltage-dependent anion channel 1 (VDAC1) on the cytosolic side of the outer mitochondrial membrane (OMM) [44]. VDAC1 is a mitochondrial metabolite transporter that allows diffusion of small hydrophilic molecules including ADP/ATP [45]. Several experimental approaches led to the conclusion that mitochondrial HKs directly couple the rst committed step of glucose metabolism to oxidative phosphorylation (OXPHOS) [43,44]. During coupled respiration, the mitochondrial hexokinases selectively use intra-mitochondrial ATP as substrate, even if cytosolic ATP is available, and the ADP that is generated is channeled back into the mitochondria through VDAC1 to support OXPHOS [43,44]. Such a feedback system ensures an overall rate of glucosemetabolism commensurate with cellular energy demandsand helps to avoid an excessive production of lactate due to the Warburg effect. Therefore, it is easy to envision that any alteration in the mitochondrial HKs may contribute to a metabolic shift which may provide a selective advantage to the cells, such as, for example, the highly glycolytic phenotype observed in the cancer cells. Taking these ndings together, it was hypothesized that P4-induced PGRMC1 signaling-mediated rapid increase in glycolysis is cell-typespeci c and linked to the interaction of PGRMC1with mitochondrial HKs.
Human Protein Atlas (HPA)-based tissue expression pro les have shown that PGRMC1 is highly expressed in the liver [46]. Therefore, wild-type and PGRMC1-depleted HEK293 and hepatoma-derived HepG2 [47] cells were used to perform a comparative evaluation of P4-induced glucose metabolism as well as study the interaction of PGRMC1 with HKs. Furthermore, P4-induced PTMs and proteasomal degradation of PGRMC1, and the intracellular distribution of both PGRMC1 and HKs, were compared between the two cell types to examine the underlying mechanism of P4-induced metabolic reprogramming.
Isolation and enrichment of endoplasmic reticulum (ER)/mitochondrial fraction Mitochondria were isolated by a method previously described [3,48]. Brie y, 80-90% con uent cells were washed with PBS and harvested in mitochondrial stabilization buffer (MSB-1) containing 70 mM sucrose, 210 mM mannitol, 5 mM HEPES pH 7.2, and 1 mM EGTA. The cells were disrupted with a Te on Dounce homogenizer and the homogenate was centrifuged at 800g for 10 min at 4°C. Following centrifugation, the supernatant was decanted through 2 layers of cheesecloth into a separate tube and centrifuged at 8000g for 10 min at 4°C. After removal of the supernatant, the pellet was resuspended in MSB-1 and the centrifugation was repeated. The nal pellet was resuspended in lysis buffer and used for immunoblotting.

Western blotting and immune-detection and quanti cation
Relative quanti cation of proteins was done by SDS-PAGE-based separation of total proteins followed by nitrocellulose membrane transfer and immunoblotting-based detection using HRP-conjugated secondary antibodies. The detailed methodology for the relative quanti cation of the proteins-of-interest in the immunoblots has been previously described in detail [48]. Table 1 lists the primary antibodies used in this study. The cell lysates were prepared in 1× RIPA lysis and extraction buffer (ThermoFisher Scienti c, Cat No: 89900) supplemented with 1× Halt protease and phosphatase inhibitor cocktail (ThermoFisher Scienti c, Cat No: 78441). The protein lysates were denatured in Laemmli buffer containing 2% SDS, 10% glycerol, 0.002% bromophenol blue, and 0.75 M Tris-HCl pH 6.8 supplemented with 100 mM DTT. The immunoblots were quanti ed using ImageJ (version 1.48) Software [49].
Cultured cells were grown in 24 well assay plates overnight. Cells were then washed thrice in pre-warmed serum free XF assay medium (non-buffered DMEM) supplemented with 1 mM sodium pyruvate (pH 7.4).
In the glycolysis stress test, an appropriate volume of glucose-free pre-warmed XF assay media was added into the individual wells of the XF24 analyzer plate and incubated in a non-CO 2 incubator for 1 hour. Pre-warmed glucose (10 mM), oligomycin (1 µM) and 2-deoxyglucose (50 mM) were added sequentially into the injector ports A, B and C of the sensor cartridge, respectively. The concentrations mentioned are nal concentrations. The data were analyzed using WAVE software (Agilent Technologies, version 2.6.0).

Site-directed mutagenesis and plasmid transfection
The N-terminal Flag-PGRMC1 ORF cloned in the pcDNA3.1+/C vector was obtained from GenScript (Catalog: OHu21154). The Flag-PGRMC1 plasmid was used to create S57A, S57D and S57E mutants using a Q5 site-directed mutagenesis kit (New England Biolabs Inc. Catalog number: E0554S) and the primer sequences listed in Table-1. The plasmids were sequenced using T7 and Sp6 sequencing primers to verify the desired mutations. The phosphomimic mutant plasmids were transfected using Lipofectamine 3000 reagent (Thermo Fisher, catalog number:L3000001).
Immunoprecipitation of Flag-PGRMC1 and mass spectrometry The N terminal Flag tagged PGRMC1 plasmid was transfected into HEK293 and HepG2 cells and the cells were serum starved for 24 hours. The cells were lysed using 1× phosphate buffered saline (PBS) supplemented with 1% IGEPAL-CA630 (Sigma, Cat. No: I8896) and 1.5% n-Dodecyl β-D-maltoside (Sigma, Cat. No: D4641). The Flag-PGRMC1 was then pulled from total cellular lysates using Bs3-conjugated (bissulfosuccinimidyl-suberate, ThermoFisher, Cat. No-A39266) as well as unconjugated Dynabeads-protein G-anti-Flag antibody. Dynabeads-protein G (ThermoFisher, Cat. No: 10004D) was used as a negative control in the immunoprecipitation. The immunoprecipitated Flag-PGRMC1 complex was denatured, reduced, alkylated and precipitated. The precipitated proteins were dissolved in 6 M urea, with volume adjusted using 20 mM Tris buffer (pH 8.0), and digested overnight at 37°C on a shaking platform using a Trypsin/Lys-C protease mixture (Promega, Cat. No: V5071). The digested peptides were desalted using Transmission electron microscopy (TEM) ForTEM, cells were grown on Matrigel coated 2 µm nitrocellulose membranes [50] and kept in medium without serum for 24 hours. Alternatively, the ER/mitochondrial fractions were precipitated as a pellet onto a nitrocellulose membrane. The cells/pellets were then xed with 2% glutaraldehyde in Sorenson's buffer (pH 7.4) at 4°C for 2 hour. Following xation, the membranes were dehydrated and nally embedded in epoxy resin and polymerized at 60°C overnight. Sections were cut on a Leica EM UC7 ultramicrotome; at rst, semi-thin sections (0.5 µm) were stained with toluidine blue and examined under light microscopy to identify the area of interest and con rm the tissue orientation. Ultrathin sections (70 nm) were then transferred to copper grids (Ted Pella Inc), and stained with uranyl acetate and lead citrate, and examined on a Phillips CM 100 Compustage (FEI) transmission electron microscope and digital micrographs were captured with an AMT CCD camera (Deben).

Statistical analysis
Statistical analysis was performed using Prism version 7.00 (GraphPad Software). The means of more than 2 groups were compared using one-way ANOVA (randomized) [51,52]. followed by Dunnett's posthoc multiple comparison test to compare the means of multiple experimental groups against a control group mean [51]. Comparisons between two groups were performed using Student's t test (unpaired). Differences were considered signi cant with P <0.05.

Results
Primary adult human tissue-speci c expression of PGRMC1.
Tissue-speci c expression of PGRMC1 was studied by reverse transcription PCR (RT-PCR) using primary adult human adipose, artery, bone marrow, cerebellum, cortex, intestine, liver, smooth muscle, and skin tissue derived RNAs. A set of primers was designed to amplify the full-length PGRMC1 open reading frame (Table-1, Fig 1A). Long-range RT-PCR ampli cation using cDNAs equivalent to 100 ng total RNA revealed the presence of an expected band corresponding to the full-length PGRMC1 amplicon in the adipose, artery, cortex, intestine, liver, skeletal muscle and skin tissues ( Fig 1B). Based on this expression pro le, the choice of hepatoma-derived HepG2 [47] cells as a model system to study P4-PGRMC1 signaling-induced glucose metabolism was justi ed [3].
Seventy kDa PGRMC1 is abundant in the ER and mitochondrial fractions in HepG2 cells.
Immunoblotting using an anti-PGRMC1 mouse monoclonal antibody (Table 1) revealed the presence of p25, p70, and p90-110 PGRMC1 in HepG2 cells grown overnight in serum-supplemented as well as serum-deprived media ( Fig 1C). In HepG2 cells, p70 PGRMC1 is the predominant form as previously observed in HEK293 cells [3]. A non-detergent-based cell fractionation method previously used to separate cytosol, enriched endoplasmic reticulum (ER)/mitochondria, and nuclear fractions in HEK293 cells [3] and EA.hy926 endothelial cells [48] was used to fractionate/enrich subcellular organelles in HepG2 cells ( Fig  S1A). Both immunoblotting, as well as transmission electron microscopy (TEM), were used to analyze HEK293 and HepG2 cell structures and the nature and enrichment of subcellular fractions. HepG2 cells were grown on the nitrocellulose membrane to study the cellular ultrastructure by TEM (Fig 1E). TEM images revealed the presence of elongated tubular rough endoplasmic reticulum (RER) in the HepG2 cells ( Fig 1F). The majority of the mitochondria in HepG2 cells were found in contact with RER and arranged in an orderly manner sandwiched between the tubules of RER (Fig 1F, yellow arrow). Following initial homogenization, the resultant subcellular fraction was designated as fraction-1 (F1), which was enriched in RER and mitochondrial structures ( Fig 1GH). The nuclear fraction was found with perinuclear ER and mitochondrial structures (Fig S1A-C, red dotted areas in BC). Therefore, the nuclear fraction was treated with a mild non-ionic detergent (0.5% NP40) to dissociate the attached perinuclear organelles (Fig S1A). The detergent treatment resulted in a smooth rounded nuclear fraction which was devoid of any residual ER/mitochondrial structures (Fig 1I). Based on this procedure and ndings, the mild non-denaturing detergent-extracted perinuclear ER/mitochondrial fraction was designated as fraction 2 (F2). Next, a series of subcellular organelle-speci c protein markers were used to validate the fractionation process and to identify PGRMC1 intracellular distribution (Fig 1J-O). α-Tubulin (TUB1A), calnexin (CANX) [53], VDAC1 [54], and histone-H3 (H3) were used as a cytosolic, ER, mitochondria, and nucleus-speci c markers, respectively (Fig 1J-N). The Oriole-stained gel indicated considerable differences between the total protein pro les of different subcellular fractions ( Fig 1O) which provided additional support for the e ciency of the enrichment process. The p70 PGRMC1 was found relatively abundant in the ER/mitochondrial fraction compared to the nucleus ( Fig 1J). The p25 PGRMC1 was not detectable in the subcellular fractions (within detection limit). The p90-100 PGRMC1 was observed as faint bands following longer exposure of the immunoblots and, therefore, it was not shown. Overall, p70 is the dominant form of PGRMC1 in HepG2 cells as previously observed in HEK293 cells [3]. Expression of an Nterminal Flag-tagged PGRMC1 in HepG2 cells indicated localization of Flag-PGRMC1 in the ER (Fig S2A, yellow arrows). Flag-PGRMC1 appeared as a p25, p50, and >p75 proteins (Fig S2B-C).
P4-induced Warburg effect is cell-type-speci c.
P4-induced cellular glucose metabolism was interrogated by simultaneous measurement of oxygen consumption rate (OCR) and extracellular acidi cation rate (ECAR) using the Seahorse ux analyzer as described previously [3]. ECAR approximates glycolysis and OCR is an important metric for mitochondrial function [3,55]. The glycolysis stress test [56] was performed to study the P4-induced Warburg effect. In the glycolysis stress test, the glucose-induced response is reported as the rate of glycolysis under basal conditions (Fig 2A, grey-colored area). ECAR, prior to glucose injection, is referred to as non-glycolytic acidi cation which is caused by processes in the cell other than glycolysis ( Fig. 2A, yellow-colored area). The injection of oligomycin, an ATP synthase inhibitor, inhibits mitochondrial ATP production, and shifts energy production to the glycolysis. A subsequent increase in ECAR reveals maximum glycolytic capacity in the cell ( Fig. 2A, cyan area). Finally, injection of 2-deoxyglucose (2-DG), a glucose analog that inhibits glycolysis through competitive binding to HKs, the rst enzyme in the glycolytic pathway, resulted in a decrease in ECAR and con rmed that ECAR produced in the experiment is due to glycolysis ( Fig. 2A). As reported previously [3], treatment of HEK293 cells with 10 µM P4 caused a signi cant increase in basal glycolysis and a concomitant decrease in respiration within 30-40 min of the treatment compared to vehicle-treated cells (Fig. 2AB). In contrast, the P4-induced glycolysis as well as respiration was found to be signi cantly decreased in HepG2 cells ( Fig 2AB). Previously, the dose-dependent analysis revealed that 10 µM P4 elicited a maximum glycolytic response in HEK293 cells within 30 min of treatment [3]. Therefore, we performed a dose-dependent analysis in HepG2 cells which revealed that 10 µM P4 is also the most effective concentration to elicit a signi cant alteration in glycolytic response ( Fig S2D). Thus, the use of 10 µM P4 to compare the P4-induced Warburg effect between HEK293 and HepG2 cells was justi ed. Interestingly, oligomycin treatment increased glycolysis (glycolytic reserve volume) in vehicletreated HEK293 cells but not in HepG2 cells (Fig 2A). In contrast, P4-treatment caused a signi cant increase in the glycolytic reserve volume in HEK293 cells, whereas the effect was marginal in HepG2 cells ( Fig 2C). Overall, these results indicate that the P4-induced Warburg effect is cell-type-speci c.
In contrast to the cell-type-speci c differential effect of P4 on glycolysis, cellular respiration rate (OCR) was found signi cantly decreased in both cell types within 30 min after P4 treatment (Fig 2D-E). Thus, it is evident that P4-mediated regulation of glycolysis and respiration may be independent events.
P4-induced proteasomal degradation of PGRMC1 is cell type-speci c.
Previously, it was demonstrated that 10 µM P4 caused rapid (within 30 min of treatment) proteasomal degradation of p70 PGRMC1 in HEK293 cells, which was inhibited with a proteasome inhibitor [3]. Based on this, it was concluded that P4-induced PGRMC1 degradation and the Warburg effect are physiologically connected events. Therefore, it was expected that PGRMC1 would not be degraded in HepG2 cells as it has exhibited the opposite effect in P4-induced glycolysis compared to HEK293 cells (Fig 2A-C). As anticipated, immunoblotting revealed that p70 PGRMC1 is not degraded in HepG2 cells within 30 min to 1.5 hours after P4 treatment compared to the HEK293 cells ( Fig 2FG). Immunoblottingbased quanti cation revealed a signi cant decrease in p70 PGRMC1 levels in HEK293 cells (Fig 2H) whereas, it remained unaltered in HepG2 cells during the course of P4-treatment ( Fig 2I). Thus, it is likely that the P4-induced Warburg effect and proteasomal degradation of PGRMC1 are interconnected and cell-type speci c.
PGRMC1 half-life is cell-type-speci c.
Turnover and half-life of a protein are dependent on the synthesis and degradation rate in the cell [57]. Therefore, P4-induced cell-type-speci c differential proteasomal degradation of PGRMC1 may be associated with a potential difference in the half-life of PGRMC1 in HEK293 and HepG2 cells. One of the methods to determine the turnover rate of a protein is to use cycloheximide to block protein synthesis and the natural decay of the protein over time is measured by immunoblotting [58]. We treated HEK293 and HepG2 cells with 50µg/ml cycloheximide and examined the PGRMC1 level at different time points ( Fig   S3). Immunoblotting revealed that half-life of p70 PGRMC1 in HEK293 cells is 2 hours (Fig S3AB) whereas, it is 5-6 hours in HepG2 cells ( Fig S3CD). Thus, the cell-type-speci c difference in PGRMC1 halflife is a factor underlying the P4-induced bioenergetic differences.
PGRMC1 deletion in HepG2 cells failed to reverse the P4-induced decrease in glycolysis.
Multiple independently selected PGRMC1 knockout (PGRMC -/-) HepG2 cell clones were generated using CRISPR/Cas9-mediated genetic ablation of PGRMC1 (Fig 3A-H). Both exon 2 and exon 3 of PGRMC1 were targeted for the homology-driven insertion of a oxed reporter/selection cassette ( Fig 3A) which was subsequently deleted by transient Cre-recombinase expression in HepG2 cells. Immunoblotting revealed the loss of PGRMC1 protein in the PGRMC -/-HepG2 cell clones ( Fig 3B). Interestingly, the level of ERspeci c CANX protein was found to be signi cantly reduced in PGRMC -/-HepG2 cell clones compared to wild-type ( Fig 3CD). Also, the total protein pro les of the PGRMC -/-HepG2 cell clones were visually different from that of wild-type cells indicating a global difference in the abundance of different species of proteins (Fig 3E). The loss of PGRMC1 in knockout clones was further validated by long-range RT-PCR which revealed the absence of the PGRMC1 ORF in the selected clones, thus indicating complete deletion of the gene (Fig 3F). HepG2 cells express a paralogous PGRMC2 gene (Fig 3G). Therefore, we used multiplex RT-PCR with primers speci c to PGRMC1 & PGRMC2 to simultaneously co-amplify both genes in a single reaction. Multiplex RT-PCR revealed the selective loss of PGRMC1 expression in the PGRMC1 -/-HepG2 cell clones, whereas, PGRMC2 expression remained unaltered, thus showing that this targeted deletion approach selectively affected PGRMC1 (Fig 3H). In subsequent experiments, four independently selected PGRMC1 -/-HepG2 cell clones were used to eliminate clonal variation. Previously, in HEK293 cells, genetic ablation of PGRMC1 inhibited the P4-induced Warburg effect [3], therefore, it was expected that loss of PGRMC1 in HepG2 cells would inhibit P4-induced inhibition of glycolysis. Interestingly, the loss of PGRMC1 failed to inhibit the P4-induced decrease in glycolysis observed in wild type HepG2 cells ( Fig  3IJ). Further, PGRMC1 loss showed no effect on the P4-mediated decrease in cellular respiration ( Fig  3KL). Overall, these results indicate that HEK293 and HepG2 cells differed in terms of reversal of the P4 induced effect on glycolysis under PGRMC1 loss-of-function conditions. PGRMC1 regulates ER distribution and ER-mitochondrial contacts in HepG2 cells.
The ER and mitochondria are structurally and functionally linked thorough contact points de ned as mitochondria-associated ER membranes (MAM), which play a key role in cellular metabolism by controlling mitochondrial dynamics and function [59,60]. Therefore, the ultrastructure of MAM was studied in HepG2 and compared with HEK293 cells to understand the cell type-speci c differences in P4induced glucose metabolism. The ultrastructure of HEK293 cells was studied previously [3]. The TEMbased study using HepG2 cells revealed the presence of abundant elongated tubular RER in the wild-type cells (Fig 1F, S4A, yellow arrows). In contrast, fragmented RER was observed in PGRMC1 -/-HepG2 cells (Fig S4B, yellow arrows). Furthermore, the orderly distribution of mitochondria appearing as sandwiched between tubules of RER and the frequent contacts between these organelles (MAM) was lost in PGRMC1 -/-HepG2 cells compared to the wild-type (Fig S4CD, green arrows).
In HEK293 cells, the ER and MAM structures are not as abundant as compared to the HepG2 cells (Fig 1,  S4, and S5). The ultrastructure of HEK293 cell was previously reported [3], however, it is presented in this study to compare with the HepG2 cells. Also, loss of PGRMC1 in HEK293 cells caused no visibly discernible dramatic effect on ER and mitochondrial distribution or MAM structures compared to the HepG2 cells (Fig S4 and S5). In addition, TEM analysis of the enriched ER/mitochondrial fractions from both cell types exhibited relatively less frequent MAMs in HEK293 compared to the HepG2 cells (Fig S6, yellow arrows). Overall, HepG2 and HEK293 cells are dramatically different in terms of organization of the ER and MAM structures, and loss of PGRMC1 severely altered structural organization of the ER and MAM in HepG2 cells compared to the HEK293 cells.
Post-translational modi cations of PGRMC1 are cell type-speci c.
In the previous study [3], PGRMC1 was found to be phosphorylated (P), acetylated (Ac), SUMOylated, and ubiquitinated in HEK293 cells. MS analysis identi ed P-S57, P-Y107, P-Y180, and Ac-K136 residues in PGRMC1. It was also found that P4-treatment caused a signi cant reduction in P-PGRMC(Ser) in HEK293 cells. Therefore, we studied the PTMs of PGRMC1 in HepG2 cells and compared them with HEK293 cells. Flag-PGRMC1 was transiently-expressed in HepG2 cells and immunoprecipitated. Immunoblotting revealed the presence of p25, p63, p100, and p>245 kDa PGRMC1 in the anti-Flag immunoprecipitate ( Fig  4A-B). Immunoblotting of the Flag-PGRMC1 immunoprecipitate using anti-P-Ser, anti-SUMO2/3/4, and anti-ubiquitin antibodies revealed that Flag-PGRMC1 protein in HepG2 cells is phosphorylated, SUMOylated, and ubiquitinated (Fig 4C-E). Next, we performed MS analysis of the Flag-PGRMC1 and identi ed P-T178, P-Y180, and P-S181 residues of PGRMC1 in HepG2 cells (Fig 4F- Table   1). The PTMs (phosphorylation and acetylation) of PGRMC1 differed in HEK293 and HepG2 cells as summarized in Figure 4F. The difference in phosphorylation is highlighted in the crystal structure of PGRMC1 in Figure 5GH. The P-S47, P-Y107, and Ac-K136 residues of PGRMC1, previously identi ed in HEK293 cells, were not detected in HepG2 cells, however, corresponding non-phosphorylated or nonacetylated peptide fragments were detected in the MS analysis (Fig 4I). The detection of P-Ser positive bands in the Flag-PGRMC1 immunoprecipitate validated this MS-based detection of P-PGRMC1(S181) in HepG2 cells, while the ubiquitination and SUMOylation of PGRMC1 provided an explanation for the higher molecular mass forms of the protein. Single ubiquitination or SUMOylation adds approximately 10 kDa or 8.6 kDa molecular mass to a protein, respectively; therefore, it is possible that either multiple residues are involved or there may be poly-ubiquitination/SUMOylation [3].
P4-induced post-translational modi cation of PGRMC1 is cell type-speci c.
In the previous study, it was found that P4 treatment of HEK293 cells caused a signi cant alteration in PGRMC1 PTMs (phosphorylation, SUMOylation, and ubiquitination) [3]. Therefore, the Flag-PGRMC1 immunoprecipitate from P4 and vehicle (ethanol)-treated HepG2 cells were subjected to immunoblotting using anti-P-Ser, anti-SUMO2/3/4, and anti-ubiquitin antibodies (Fig 5A-F). A faint p25, a p60, and a strong p100 PGRMC1 bands were detected in both P4 and vehicle (ethanol)-treated immunoprecipitate (Fig 5A-C). The p100 PGRMC1 reacted positively for anti-P-Ser (Fig 5D), anti-SUMO2/3/4 (Fig 5E), and anti-ubiquitin (Fig 5F) antibodies. In contrast to HEK293 cells [3], Image J-based analysis of the immunoblots revealed no signi cant difference in the PTMs of PGRMC1 in P4 treated and vehicle-treated HepG2 cells. It is possible that the pan anti-P-Ser antibody may not detect all phosphorylated epitopes of PGRMC1 in HepG2 cells. Therefore, total proteins from the P4 and vehicle-treated Flag-PGRMC1 overexpressed HepG2 (Fig 5G-I) and HEK293 (Fig 5J-L) cells were subjected to isoelectric focusing (IEF) followed by SDS-PAGE and immunoblotting. IEF/SDS-PAGE analysis revealed the presence of pH 8-9 (blue rectangle), pH 6-7 (green rectangle), and pH 5-6 (red rectangles) fractions of PGRMC1 containing p60, p75 and p>75 proteins, each corresponding to a speci c isoelectric point (pI) and molecular mass representing different PTMs in HepG2 cells (Fig 5GI). A comparison between the charged fractions of PGRMC1 revealed that P4 treatment caused the disappearance of p75 (pI 8-9) fraction (green arrows) and shifted p60 (pI 5-6) fraction (red/white arrows) towards more negative pI/pH (Fig 5G-I). IEF/SDS-PAGE using the P4 and vehicle-treated HEK293 cell-derived total proteins revealed the presence of pH 9-10 (blue rectangle), pH 8-9 (green rectangle), and pH 3-4 (red rectangles) fractions of PGRMC1 containing p60 and p75 proteins (Fig 5JK). A comparison revealed that P4 treatment caused the disappearance of a fraction of p60 (pI 8-9) PGRMC1 (black arrows) and reduced the abundance of the p60 (pI 3-4) fraction (Fig 5J-L blue & white arrows) compared to the vehicle-treated cells. Thus, it is evident that the basal (vehicle-treated), as well as P4-treated, PTMs of PGRMC1 are cell-type-speci c which may account for the differential effect of P4 on glucose metabolism in the different cells.
Cell-type-speci c interaction of PGRMC1 and HKs.
To validate potential cell-type-speci c interaction of PGRMC1 and HKs, and to study the effect of P4 on such interaction, we subjected a Flag-PGRMC1 immunoprecipitate from the vehicle and P4-treated HEK293 and HepG2 cells to immunoblotting using anti-HK1 and -2 antibodies. Immunoblotting revealed the presence of a p25, p60, and p100 PGRMC1 in ag-PGRMC1 transiently expressed HEK293 cells ( Fig  6AB). The absence of IgG light and heavy chain speci c bands in the Oriole/Coomassie-stained gels ( Fig   6A) was due to the chemical (BS3) crosslinking of the primary antibodies with the protein G beads. Immunoblotting of the Flag-PGRMC1 immunoprecipitate from the vehicle and P4-treated HEK293 and HepG2 cells revealed the presence of a p100 HK1 in both cell-types (Fig 6C-F). Increased exposure of the immunoblots revealed a p135 HK1 associated with both vehicle and P4 treated HEK293 cells (Fig 6D). In contrast to HEK293 cells, multiple p≥135 HK1 proteins were found in HepG2 cells to associate with P4treated PGRMC1, whereas, only p100 HK1 associated with PGRMC1 in the vehicle-treated cells (Fig 6F). Interestingly, a p100 HK2 was found associated with PGRMC1 in HEK293 cells but not in HepG2 cells ( Fig   6GH). Overall, these results indicate a cell type-speci c differential interaction between PGRMC1 and HKs which may be responsible for the observed difference in the P4-induced metabolic reprogramming.
HK2 is differentially modi ed in HEK293 and HepG2 cells.
High throughput mass spectrometry analysis documented multiple phosphorylation and ubiquitination sites at functionally relevant residues in HK1 and HK2 archived in the PhosphositePlus database ( Fig 6IJ) [55,61]. Differential PTMs of HK2 may be responsible for the observed loss of interaction with PGRMC1 in HepG2 compared to the HEK293 cells. Therefore, two-dimensional IEF/SDS-PAGE analysis was performed to study the charged PTM fractions of HK2 in HEK293 and HepG2 cells. HK2 appeared as several charged fractions within the pH~4-5 region in vehicle-treated HEK293 cells (Fig 6K, red rectangle); these fractions shifted to a more negative pH in the P4-treated cells (Fig 6K, white rectangle). In contrast, HK2 appeared as several charged fractions within the pH~4-5 (red rectangle) and pH~8-10 (blue rectangle) regions in the vehicle-treated HepG2 cells (Fig 6L), which shifted to the more negative pH 3-4 region in the P4-treated cells (Fig 6L, yellow rectangle). It is interesting to note the HK2 pH~8-10 fractions (blue rectangle) were not detected in the HEK293 cells (Fig 6K) compared to HepG2 cells (Fig 6L). Also, the pI/pH shift in the charged fractions of HK2 in the P4-treatment HEK293 and HepG2 cells was not identical in the two cell types (Fig 6KL, white rectangles in the overlaid images at the bottom panel). Thus, differential PTMs may be responsible for the loss of interaction between HK2 and PGRMC1 in HepG2 cells compared to HEK293 cells. Identifying these modi cations of HK2 is beyond the scope of this study.
Phosphomimics revealed that S57 phosphorylation is not responsible for cell type-speci c interaction of PGRMC1 with HK2.
We used phosphomimetics to determine if differential phosphorylation at S57 in HEK293 and HepG2 cells (Fig 4F-H) is responsible for the cell-type-speci c interaction of PGRMC1 with HK2 (Fig 5 & 6). The S57 residue overlaps a potential Src homology 3 (SH3) domain on PGRMC1 [62] and the SH3 domains have been identi ed as protein modules that recognize proline-rich sequences [63]. The existence of a multimeric protein complex involving HK2 and Src tyrosine kinase has been reported [64]. Therefore, sitedirected mutagenesis was used to generate phosphonull PGRMC1(S57A) and phosphomimic PGRMC1(S57D/E) mutants ( Fig S7A). DNA sequencing con rmed the desired mutations in the plasmids expressing Flag-tagged PGRMC1 (Fig S7B). Next, we transiently transfected Flag-PGRMC1 mutants in HEK293 cells and the anti-Flag immunoprecipitate was subjected to immunoblotting using HK-speci c antibodies. Immunoblotting revealed the expression of Flag-PGRMC1 mutants in HEK293 cells (Fig S7C), while Oriole-staining and immunoblotting revealed Flag-PGRMC1 mutants could be immunoprecipitated (Fig S7D-E). Immunoblotting revealed the presence of both HK1 and HK2 in the mutant Flag-PGRMC1(S57A/D/E) immunoprecipitates indicating S57 phosphorylation is not involved in the cell-typespeci c interaction of PGRMC1and HK2 (Fig S7F). However, plot pro ling revealed the relative abundance of an unknown p75 kDa interacting protein (Fig S7D, blue arrow) which is relatively more abundant in the PGRMC1 phosphonull mutant (S57A) immunoprecipitate compared to the wild-type (S57) or phosphomimic mutants (S57D/E). These observations suggest that the phosphorylation state of S47 may in uence the interaction of PGRMC1 with other proteins but not HKs (Fig S7G). Identi cation of the p75 interacting protein will be the subject of future studies.

Distribution of HKs in different subcellular compartments.
The subcellular distribution of HK1 and HK2 has been proposed to regulate metabolic preference by cells [43]. A dual role has been proposed for HK2 based on its subcellular location, which involves channeling glucose-6-phosphate (G6P) into glycogen and PPP when localized in the cytoplasm. In contrast, shuttling G6P to glycolysis and OXPHOS occurs when HK2 is bound to mitochondria [65].
P4 treatment causes cell type-speci c differential distribution of HKs in subcellular compartments.
Immunoblotting was used to compare the HK content in subcellular fractions from the vehicle and P4treated HEK293 cells. This analysis revealed that P4 treatment caused a signi cant reduction of both HK1 and HK2 in the ER/mitochondrial fraction compared to the vehicle-treated cells (Fig 8A-D). In contrast to HEK293 cells, the HK1 and HK2 levels were found signi cantly increased in the ER/mitochondrial fractions (F1) of P4-treated HepG2 cells compared to vehicle-treated cells (Fig 8E-H). In both HEK293 and HepG2 cells, P4-treatment signi cantly increased translocation of HK1 to the nucleus compared to the vehicle-treated cells (Fig 8IK & MO). In contrast to HK1, the HK2 level was signi cantly decreased in the HEK293 nuclear fraction (Fig 8JL), but increased in the HepG2 nuclear fraction compared to vehicle-treated cells (Fig 8LP). Overall, these results indicate a considerable difference in the effect of P4 on subcellular translocation of HK1 and HK2 proteins in HEK293 and HepG2 cells. This differential action may account for the metabolic differences induced by P4.

Discussion
This study provides evidence that the ability of P4 to modulate glycolysis is cell-type-speci c. P4 caused a rapid increase in glycolysis in HEK293 cells, whereas it decreased glycolysis in HepG2 cells. These two extremes of P4-induced glucose utilization were found associated with differential PGRMC1 PTMs, and proteasomal degradation of PGRMC1 and cell-type-speci c interaction with HKs. The identi cation of HKs as an interacting partner of PGRMC1 is novel. Also, PGRMC1-mediated regulation of ER and MAM structures in HepG2 cells is a novel nding that provides a direct link to potential metabolic regulatory mechanisms. Furthermore, P4 treatment of HEK293 and HepG2 cells caused a cell-type-speci c differential translocation of HK1/2 to the ER and mitochondrial compartments as well as the nucleus which may account for the metabolic responses by these two cell types to P4.
PGRMC1 was found differentially phosphorylated in HEK293 and HepG2 cells under basal conditions ( Fig 4&5). P-PGRMC1 (T-178, Y180, and S181) residues were detected in HepG2 cells, which is in contrast to previously identi ed P-PGRMC1 (S57 & S181) residues in HEK293 cells [3]. The signi cance of PGRMC1 phosphorylation is not completely known. Differential phosphorylation (charged fraction) of PGRMC1 in estrogen receptor-positive (ER) and ER-negative breast cancers was reported byNeubauer et al. [66]. Breast tumors are sex hormone-sensitive and 60-70% of all breast cancers are either estrogen receptor (ER)-positive or P4-receptor (PR)-positive [67,68]. Clinically, hormone-receptor-positive cancers tend to grow more slowly and have favorable clinical outcomes than hormone-receptor-negative tumors [69]. The exact mechanism for such a difference in terms of disease outcome is not completely understood. The most frequently observed PGRMC1 phosphorylation sites listed in the PhosphositePlus database are S57 and S181 followed by Y113, Y180, and T178 [3,55,61]. The conserved S57 and S181 residues overlap predicted Src homology domain binding motifs and also match to a theoretically predicted site for casein kinase 2 (CK2) activity [62]. Previously, reduced levels of P-PGRMC1(Ser) were found associated with the P4-induced Warburg effect in HEK293 cells [3]. Interestingly, in HepG2 cells, P4induced reduction in glycolysis is not associated with a reduction in P-PGRMC1(Ser) levels. Also, the absence of P-PGRMC1(S57) in HepG2 cells under basal condition indicates a correlation with PGRMC1 phosphorylation state and P4-induced metabolic reprogramming. Neubauer et. al. showed that stable transfection of S57A and S180A phosphonull mutants in MCF7 breast cancer cells protected cells from peroxide-induced cell death [66]. In a recent study, progestin treatment of MCF7 and T47D breast cancer cell lines increased P-PGRMC1 (S181) levels which were subsequently reduced by treatment with a CK2 inhibitor [70]. Thus, a link between PGRMC1 phosphorylation involving S57 and S180 residues and breast cancer pathology has been established, however, the mechanistic role of these two sites in disease pathophysiology is still not well established. Another important aspect of PGRMC1 S57 and S180 phosphorylation state is that it may affect potential interacting protein binding to the putative SH domain. Based on that understanding, we used S57 phosphomimic (S57D/E) and phosphonull (S57A) mutants to study the interaction of PGRMC1 with HK2 ( Fig S7). Interestingly, phosphomimetics revealed that S57 phosphorylation status is not responsible for the cell type-speci c interaction of PGRMC1 and HK2, however, the Oriole-stained immunoprecipitated proteins revealed an unknown 75 kDa interacting protein that exhibited differential interaction with S57A compared to the native S57 and S57D/E PGRMC1 in HEK293 cells (Fig S7G).
PGRMC1 is a 25 kDa protein. Complex PTMs of PGRMC1 including SUMOylation and ubiquitination were found associated with the appearance of higher molecular mass forms (70-100 kDa) of the protein in the ER/mitochondria and nuclear fractions in HEK293 cells [3]. SUMO and ubiquitin have a molecular weight of ∼8-10 kDa [71,72], therefore, it was suggested that the appearance of high molecular weight PGRMC1 was due to polyubiquitination and/or SUMOylation [3]. Multiple ubiquitination cycles add polyubiquitin chain on a protein which acts as a signal for proteasomal degradation [71]. A previous study demonstrated that P4-induced glycolysis and rapid proteasomal degradation of p70 PGRMC1 in HEK293 cells is associated with altered levels of PGRMC1 SUMOylation and ubiquitination. Therefore, the absence of any alteration in the PGRMC1 SUMOylation and ubiquitination is in agreement with the resistance of PGRMC1 to P4-induced proteasomal degradation and reduced glycolytic activity in HepG2 cells. Furthermore, phosphorylation cross-talks with SUMOylation and ubiquitination to determine the fate of a protein [73]. Interestingly, the P-PGRMC1 (Ser) level in P4-treated HepG2 cells was unaltered, which is in contrast to the HEK293 cells where P4-induced reduction of P-PGRMC1 (Ser) level was demonstrated.
The difference in the PTMs of PGRMC1 under basal condition may also be responsible for the cell-type-speci c differential half-life of the protein (Fig S3). Thus, it is evident that phosphorylation and SUMOylation/ubiquitination of PGRMC1 are linked to proteasomal degradation in the context of P4induced cell type-speci c metabolic reprogramming, however, the exact mechanism needs to be elucidated in the future.
This study highlighted cell-type-speci c differential interaction of HK1/2 and PGRMC1 (Fig 7), as well as a differential distribution of HK1/2 in various subcellular organelles following P4-stimulation (Fig 8). Both processes may contribute to P4-induced cell type-speci c metabolic reprogramming. Mammalian hexokinases are allosteric enzymes catalyzing ATP-dependent phosphorylation of glucose to G6P [74,75]. Despite sequence similarities, the kinetic features, regulatory properties, intracellular distribution, and expression pattern of HKs differ signi cantly [44]. For example, the HK-mediated reaction product G6P inhibits both HK1 and HK2 at micromolar levels; however, inorganic phosphate (Pi) only relieves G6P inhibition of HK1 [75]. Another important fundamental feature that distinguishes these isoforms is the unique capacity of HK1 and HK2 to directly interact with mitochondria, both physically and functionally [44]. Wilson et al. proposed that mitochondria-bound HK1 couples cytosolic glycolysis to intramitochondrial OXPHOS, whereas HK2 serves a primarily anabolic function to provide G6P for glycogen or lipid synthesis via the PPP (Fig S6). Therefore, P4-induced translocation of HK1 and 2 away from the ER/mitochondria may suppress OXPHOS activity (respiration) but increase glycolysis in HEK293 cells. On the other hand, increased translocation of HK1 and HK2 to the ER/mitochondria in P4-treated HepG2 cells may lead to a suppression of both glycolysis and respiration. It has been suggested that the primary effect of P4 on carbohydrate metabolism is the promotion of glycogen storage in the liver [30]. P4 has been shown to induce glycogen synthesis in a dose-dependent manner in primary human endometrial cells which express a high level of PGRMC1 [76]. In this context, it is important to note that in our study the rate of glycolysis was measured in terms of ECAR (due to lactate secretion) which does not re ect the true status of glycolysis in the cell. For example, if G6P production due to increased glycolytic activity is channeled to glycogen synthesis or the PPP, then ECAR values will not change. As a result, the Seahorse ux analyzer will not be able to detect an increase in glycolysis. Therefore, increased channeling of G6P to glycogen synthesis or the PPP in response to P4 cannot be excluded in HepG2 cells. However, in a contrasting study, HK1/2 overexpression in CHO cells revealed that HK1 exclusively promotes glycolysis, whereas HK2 has a more complex role, promoting glycolysis when bound to the mitochondria and glycogen synthesis when located in the cytosol [77]. Thus, the exact mechanism of P4-induced cell typespeci c differential translocation of HKs and metabolic reprogramming cannot be deciphered in this study and more in-depth studies in the future will be necessary to shed light on this important aspect of metabolic regulation.
The loss of PGRMC1 altered ER and mitochondrial organization in HepG2 cells. The ER is a complex dynamic organelle, primarily involved in protein and lipid synthesis as well as secretion, calcium homeostasis, and interactions with other organelles [78]. The structural variations of ER in different celltypes are a re ection of functional heterogeneity [79]. Therefore, the differences in ER structure in HepG2 and HEK293 cells were anticipated and this difference may account for the P4-induced metabolic difference in these cell-types to some extent. The ER tubules in PGRMC1-de cient HepG2 cells were fragmented compared to the wild-type suggesting PGRMC1 has a role in the maintenance of ER structural integrity (Fig 4). The continuity of the ER lumen is vital for this function and is required for the e cient diffusion and equilibration of ions [80]. The integrity of the ER and MAM is also required for hormonal signaling and glucose homeostasis [81] and its alteration is associated with the pathogenesis of metabolic diseases [82]. The MAM interface acts as a nutrient-regulated hub that adapts mitochondrial metabolism to the nutritional state of the cell [83]. In nutrient-limited conditions, the ER-mitochondria interaction doubles in length in liver cells [84], whereas high glucose levels disrupt MAM integrity in the liver [59]. These ndings suggest that the liver could adapt to metabolic transition through a MAMdependent mechanism. In mammalian MAM, the ER and mitochondria are physically tethered to each other by the inositol 1,4,5-trisphosphate (IP3) receptor (IP 3 R/ITPR)-glucose-regulated protein 75 (Grp75/HSPA9)-VDAC trimeric complex (Fig S6) [85]. VDAC interacts with adenine nucleotide translocase (ANT), an inner membrane protein (Fig S6) [86]. The resultant VDAC-ANT-mitochondrial creatine kinase (mCK) complex regulates the exchange of ATP and ADP between the mitochondria and cytosol [87]. Both HK1 and HK2 directly interact with and promote the 'open' con guration of VDAC that permits anionic metabolite exchange across the outer mitochondrial membrane (OMM) [44,88]. Functional coupling between glycolysis and OXPHOS is mediated by ADP-ATP exchange through the VDAC-ANT complex ( Fig   S5) [44]. Therefore, the association of PGRMC1 with HKs, P4-induced alteration of the subcellular distribution of HKs and metabolic reprogramming, and abnormal MAM in PGRMC1-de cient cells all point to a MAM-based metabolic transition switch controlled by P4-PGRMc1 signaling.

Conclusions
Glucose is the preferred fuel for ATP production by most cells. The fate of pyruvate, a product of glycolysis, depends upon whether oxygen is present. For example, when oxygen levels are low, skeletal muscle cells rely on glycolysis to meet their intense energy requirement and generate lactic acid as a byproduct [89]. In the presence of oxygen, the pyruvate is converted into acetyl CoA and subsequently enters the mitochondria to generate 36 molecules of ATP through the tricarboxylic acid (TCA) cycle and oxidative phosphorylation (OXPHOS). Thus, it is conceivable that burning glucose through glycolysis, TCA cycle, and OXPHOS pathways is a pro table metabolic preference for a cell. However, in tumors and other proliferating or developing cells, the rate of glucose uptake dramatically increases and the metabolic preference shifts to the production of lactate, even in the presence of oxygen and fully functioning mitochondria [90]. This process, known as the Warburg effect or aerobic glycolysis [91], is favored by tumor cells because the interior cells in a tumor mass may experience uctuations in oxygen tension that in turn limit nutrient oxidation; therefore, shifting the metabolic preference to aerobic glycolysis gives an advantage for tumor cell survival [90]. Besides, increased glucose utilization generates high amounts of lactate, which creates an acidic environment and facilitates tumor invasion [90]. The metabolic preference for aerobic glycolysis over OXPHOS also depends on other factors, for example, hormones. Decoding the mechanism of metabolic reprogramming is vital to our understanding of the diseases associated with metabolic dysfunction. Thus, our discovery of a P4-PGRMC1 signaling mediated cell type-speci c metabolic regulation pathway is an important step towards unraveling potentially new mechanisms underlying the hormone-responsive tumor phenotype.     The PGRMC1 PTMs in HEK293 cells were identi ed in a previous study [3]. (G&H): Crystal structures showing PGRMC1 cytochrome b5-like domain bound to haem and the N and C terminal sequences in HEK293 (G) and HepG2 (H) cells. The cell-type-speci c differential post-translationally modi ed residues (bold and red/blue) of PGRMC1 are highlighted in the crystal structure. Adapted from PDB ID: 4X8Y [26].

List Of Abbreviations
The β-strands and the α-helices are highlighted in pink and blue colors, respectively. The Tyr113, Tyr107, Lys163, and Tyr164 are the haem-binding residues [26]. The P-S57 is localized in the N-terminal region and the P-T178, P-Y180 and S181 residues are in the C-terminal region. The N-terminal transmembrane motif is highlighted in bold. (I) Table showing representative MS/MS spectrums of the PGRMC1 posttranslationally modi ed peptides identi ed in HepG2 cells. The spectrum is written in the form "X.Y", where X is a unique identi er for a particular tandem mass spectrum in this data set and Y is an identi er for this particular sequence solution.Log(e): the base-10 log of the expectation that any particular peptide assignment was made at random (E-value).m+h: the calculated mass of the protonated parent ion for this sequence assignment.Sequence: the sequence of the assigned peptide sequence. n: the number of observations of the respective peptide sequence in the Global Proteome Machine Database (GPMDB) [92].
The modi ed residues in the peptides are highlighted in red.