Transcriptional regulation of cyclophilin D by BMP/Smad signaling and its role in osteogenic differentiation

Cyclophilin D (CypD) promotes opening of the mitochondrial permeability transition pore (MPTP) which plays a key role in both cell physiology and pathology. It is, therefore, beneficial for cells to tightly regulate CypD and MPTP but little is known about such regulation. We have reported before that CypD is downregulated and MPTP deactivated during differentiation in various tissues. Herein, we identify BMP/Smad signaling, a major driver of differentiation, as a transcriptional regulator of the CypD gene, Ppif. Using osteogenic induction of mesenchymal lineage cells as a BMP/Smad activation-dependent differentiation model, we show that CypD is in fact transcriptionally repressed during this process. The importance of such CypD downregulation is evidenced by the negative effect of CypD ‘rescue’ via gain-of-function on osteogenesis both in vitro and in a mouse model. In sum, we characterized BMP/Smad signaling as a regulator of CypD expression and elucidated the role of CypD downregulation during cell differentiation.


INTRODUCTION
The mitochondrial permeability transition pore (MPTP) is a non-selective high-conductance channel within the inner mitochondrial membrane (IMM). MPTP opening leads to the increased permeability of the IMM and entry of solutes up to 1.5kDa of molecular mass, i.e. Mitochondrial Permeability Transition (Bernardi, Rasola, Forte, & Lippe, 2015). Although the MPTP molecular identity remains debatable, the ATP synthase converges as a potential locus. The consequences of MPTP opening span from physiological events such as regulation of synthasome assembly, oxidative phosphorylation (OxPhos), membrane potential (Δψm), ROS-induced ROS release, Ca +2 homeostasis, and epigenetic regulation (Bernardi &  Encoded by the nuclear gene PPIF, CypD is a chaperone protein -peptidyl-prolyl cis-trans isomerase F (PPIase) -thought to be involved in protein folding. However, very few reports demonstrate such an activity of CypD (Porter & Beutner, 2018). Regarded as the master regulator of MPTP, CypD is imported into mitochondria where it can actively bind to the mitochondrial ATP synthase, decreasing the threshold for pore opening and increasing opening probability. The exact molecular mechanism by which CypD regulates the MPTP and mitochondrial function remains unclear. A range of post-translation modifications, such as acetylation and phosphorylation, is known to exert regulatory properties on CypD activity, and consequently on MPTP opening shown that CypD also functions as a scaffold protein, being able to cluster various structural and signaling molecules together to modify mitochondrial physiology and bioenergetic response Porter & Beutner, 2018). Although several aspects affecting CypD physiologic regulation and its interplay with mitochondrial activity have been elucidated, how the CypD gene is transcriptionally regulated has yet to be described.
Cellular metabolism and mitochondrial function play an important role in cell differentiation, which in turn is a crucial process in organogenesis, tissue homeostasis and damage repair. CypD expression and activity are cell-specific (Laker et al., 2016) and can be regulated during differentiation. As cells become more active and increase their oxygen consumption, higher ROS levels are produced partially driven by an increase in the electron flow within the respiratory chain, and Δψm (Barja, 1999;Drose & Brandt, 2012;Suski et al., 2018;Turrens, 1997). Higher ROS levels due to OxPhos activation can lead to deleterious effects including opening of the MPTP and consequently blunting of mitochondrial function (Carraro & Bernardi, 2016). To inhibit MPTP opening, it is beneficial for differentiating cells moving towards a higher OxPhos usage, to downregulate CypD expression and/or activity to be better protected against oxidative stress and support OxPhos. Such an inhibition due to downregulation of CypD has been reported during neuronal and cardiomyocyte differentiation (Eliseev et  Upon BMP receptor activation and trans-phosphorylation, receptor regulated Smads (R-Smads) such as Smad1/5/8 are recruited and phosphorylated. Phospho-Smad1/5/8 binds to Smad4 forming a heterodimeric complex which is subsequently translocated into the nucleus to activate or repress the expression of its target genes. Osteogenesis and osteoblast (OB) differentiation are controlled at least in part via BMP/Smad canonical signaling. During osteogenic differentiation, activation of the transcription factor Runx2 is mediated by the BMP/Smad signaling pathway initiating a cascade of downstream osteogenic marker genes upregulation (Bianco & Robey, 2015

BMP/Smad signaling is a transcriptional repressor of Ppif gene
To determine potential mechanisms involved in CypD regulation on a transcriptional level, we performed an in silico analysis of the CypD gene, Ppif, promoter for transcription factor (TF) binding sites using PROMO online platform (Messeguer, Escudero et al. 2002). We found multiple BMP-dependent Smad-binding elements (SBE) in the 1.1kb 5' upstream region of both human PPIF and mouse Ppif genes ( Figure 1A). BMP/Smad signaling pathway is an important driver of differentiation in various lineages including osteogenic lineage. Therefore, to test if BMP/Smad signaling regulates CypD gene transcription, we transfected mouse osteogenic bone marrowor calvaria-derived cell lines ST2 and MC3T3-e1, respectively, with a pCMV-Smad1 vector. We observed that CypD mRNA was downregulated after Smad1 over-expression, whereas osteogenic marker Runx2, a readout of BMP/Smad activity, was upregulated ( Figure 1B-C). The efficiency of Smad1 transfection was confirmed by real time RT-qPCR.
To confirm interaction of BMP-dependent Smad1 with the Ppif promoter, we performed a chromatin immunoprecipitation (ChIP) DNA-binding assay using nuclear fractions from ST2 cells supplemented with BMP2 or vehicle-control for 24 hours. PCR analysis of the reversed cross-linked protein-DNA immunoprecipitate complexes was done ( Figure 2A) using primers to amplify the distal region within the Ppif promoter containing multiple SBEs ( Figure 2B). Smad1-Ppif interaction was found to be present at low levels in the vehicle-treated control cells. This finding is consistent with the fact that intrinsic levels of BMP activity are present even in undifferentiated osteogenic cells. Importantly, BMP2 treatment significantly induced Smad1-Ppif interaction.
To characterize the functionality of the SBEs found within the Ppif promoter, we subcloned the 1.1kb mouse Ppif promoter into the pGL4.10 luciferase reporter construct ( Figure 3A). We then co-transfected ST2 cells with the above promoter-reporter construct and either pCMV-Smad1 or pCMV-EV and analyzed the luciferase signal. Figure 3C shows that Smad1 significantly decreased the luciferase signal from the 1.1kb Ppif-luc promoterreporter. Furthermore, a promoter bashing approach was used and five other various length promoter constructs were generated (Supplementary Figure S1A): -0.37kb to -0.1kb, or -0.62 to -0.45kb, or -1.1 kb to -0.95kb, or -0.62kb to -0.1kb, or -1.1kb to -0.45kb deletion mutants that correspond to a cluster of the two most proximal (P), three middle (M), five most distal (D), five middle + proximal (M+P), and eight distal + middle (D+M) SBEs, respectively. We co-transfected ST2 cells with the above promoter-reporter constructs and either pCMV-Smad1 or pCMV-EV and analyzed the luciferase signal for all our constructs. Middle and Distal regions did not show any luciferase signal difference when compared to EV-transfected cells, whereas Proximal, M+P, and D+M showed activation of Ppif activity (Supplementary Figure S1B). Thus, only the full-length 1.1kb Ppif-luc reporter showed Ppif activity repression upon Smad1 transfection ( Figure 3C). Knowing that the full-length Ppif promoter is the actual region controlling the gene activity and that Smads work by forming oligomeric structures binding various regions of the promoter (Massague, Seoane, & Wotton, 2005; Wang et al., 2014), we considered the data using the 1.1kb Ppif-luc reporter as the most relevant. Of note, in all these experiments, activation of Smad1 signaling following transfection was confirmed by the increase in the activity of 12xSBE, a BMP/Smad signaling luciferase reporter ( Figure 3B). In sum, our data indicate that BMP/Smad signaling is a transcriptional repressor of CypD.
To further delineate the role of BMP/Smad signaling in Ppif repression we used the inhibitory Smad7 which rescued Ppif promoter activity in Smad1-transfected cells ( Figure 3C). Smad7 competitively inhibits the interaction of R-Smads to the cytoplasmic domain of their respective cell receptors, therefore preventing R-Smad recruitment and phosphorylation, and consequently nuclear translocation. Additionally, cells transfected with the 1.1kb Ppif-luc reporter were either treated with BMP2 with or without the BMP inhibitor Noggin, or induced in osteogenic media and assayed for luciferase signal. BMP2 and osteogenic media downregulated the luciferase signal, whereas Noggin rescued Ppif activity ( Figure 3D). These results were also confirmed in MC3T3e1 cells ( Figure 3E-F). Taken together, these data support our hypothesis that BMP/Smad signaling transcriptionally represses Ppif promoter activity and consequently downregulates CypD expression during osteogenic differentiation.

Decreased CypD expression and MPTP activity during osteogenic differentiation
BMP/Smad signaling is a major driver of differentiation. It is particularly important for osteogenic lineage.
During osteogenic differentiation, cell energy metabolism shifts towards OxPhos as was shown by us and others  Figure   S2A), potentially leading to oxidative stress and higher probability of the MPTP opening, it is beneficial for actively respiring cells such as OBs to decrease MPTP activity, for example by downregulating CypD. We, therefore measured CypD expression and MPTP activity in several osteogenic cell types. Figure 4 shows that primary  Figure S3). We then measured MPTP activity using calcein-cobalt assay and flow cytometry ( Figure 4F). In this assay, cells are incubated with calcein AM in the presence of CoCl 2 which quenches cytosolic but not mitochondrial calcein unless mitochondria are permeabilized due to MPTP opening. The increase in the calcein signal indicates higher resistance to MPTP opening and decreased pore activity (Petronilli et al., 1999). The assay showed that calcein signal increased at day 14 of OB differentiation when compared to day 0 indicating lower MPTP activity ( Figure 4F). No changes were detected in the total mitochondrial mass during OB differentiation as labeled by Nonyl Acridine Orange ( Figure 4G), therefore confirming that calcein signal increase is in fact caused by a lower pore activity and not by an increased mitochondrial compartment. This is consistent with our previous report showing that mitochondrial mass and mtDNA do not increase during OB differentiation (Shares et al., 2018;Shum, White, Mills, et al., 2016). Altogether, these data indicate that CypD expression is downregulated and MPTP activity is decreased during osteogenic differentiation.

Smad1
Osteogenic differentiation is a complex process that involves the coordination and crosstalk of several signaling pathways. BMP/Smad signaling is a potent driver of osteogenic differentiation in BMSCs. Once the BMP response is activated, other signaling molecules can influence Smad activity and osteogenic differentiation.
For instance, after Smad1-Smad4 complex nuclear translocation, its dephosphorylation is accompanied by the dissociation of the complex and export to the cytoplasm. Smad4 then is believed to interact with β-catenin and  Figure 5B) and decreased Ppif promoter activity when co-transfected with the full-length 1.1kb Ppif-luc reporter ( Figure 5C). Altogether, our data strongly indicate that BMP/Smad signaling exerts inhibitory effect on the Ppif gene as a direct transcriptional repressor and not as an indirect result of osteogenic differentiation ( Figure 5D).

Downregulation of CypD is important for osteogenic differentiation
We previously reported that  Figure 7G). Although bones with AAV-DIO-delivered CypD re-expression showed decreased cortical thickness and torsional rigidity when compared to their respective contralateral PBS-injected tibiae ( Figure 7F and 7H), we further analyzed our data using data normalization. In both Creand Cre + mice, AAV-DIO injected tibiae were normalized to the contralateral PBS-injected limb to account for unspecific differences in bone phenotype between animals, differences caused by CypD deletion among experimental and control mice, and unforeseen effects of virus infection. Normalized data showed a significant decrease in cortical thickness in the virus-infected Cre + mice ( Figure 7I) and a decrease in tibial bone volume fraction in these mice which did not reach statistical significance ( Figure 7J). These changes were sufficient to decrease tibial biomechanical properties and increase bone fragility in virus-injected bones from Cre + mice when compared to To our knowledge, this is the first evidence for TF-mediated regulation of PPIF transcription. Although our main focus was on osteogenic cells, our findings can be potentially extrapolated and studied in other cell types.

143b-OS cells results indicate that Smad-mediated CypD downregulation/Ppif repression is not exclusive nor
indirect to osteogenic differentiation. As previously discussed, BMP is ubiquitously expressed and described to induce cellular differentiation and maturation in various tissues. Accordingly, some cells induced by BMP/Smad signaling, reprogram their metabolic profile to a higher energetic state during differentiation and are shown to decrease CypD mRNA expression and/or activity, such as cardiomyocytes and neuronal cells. We previously reported that BMP2 induction stimulate mitochondrial OxPhos during OB differentiation and that such activation Biomechanical torsion testing: immediately following μCT scanning, tibiae were subjected to biomechanical testing. The ends of the tibiae were cemented (Bosworth Company) in aluminum tube holders and tested using an EnduraTec TestBench™ system (Bose Corporation, Eden Prairie, MN). The tibiae were tested in torsion until failure at a rate of 1°/sec. The torque data were plotted against rotational deformation to determine maximum torque and torsional rigidity. Data from virus-injected tibiae were normalized to the contralateral PBS-injected (intra-mouse control) tibiae data.
Histology: After biomechanical testing, tibiae bones were NBF-fixed and processed for histology via decalcification in Webb-Jee 14% EDTA solution for one week followed by paraffin embedding. Sections were cut to 5 μm in three levels of each sample, and then stained with either TRAP or immunofluorescence (IF).
Histomorphometry: TRAP or IF-stained slides were scanned in an Olympus VSL20 whole slide imager at 40x magnification and evaluated with VisioPharm automated histomorphometry software. TRAP-stained slides were analyzed to measure the TRAP positive area relative to total bone area in the tibiae shaft. IF-stained slides probed for osteocalcin were analyzed to measure the fluorescence signal intensity relative to total bone area of tibiae shaft. Three different levels were counted per mouse and averaged.
Chromatin immunoprecipitation assay: Chromatin immunoprecipitation assay was performed using SimpleChIP ® Enzymatic Chromatin IP Kit -Magnetic Beads (Cell Signaling Technology, Inc., Danvers, MA, USA) according to the manufacturer's instructions. Briefly, cells were fixed and DNA cross-linked with formaldehyde and homogenized. Nuclei were pelleted and chromatin was digested enzymatically, sonicated and immunoprecipitated with either anti-Smad1 antibody (RRID:AB_628261), or negative control immunoglobulin G, or positive control Histone H3 antibody using protein G magnetic beads. Chromatin was eluted from immunoprecipitate complexes and cross-links reversed with NaCl and proteinase K. After purification, DNA was amplified by a PCR reaction using primers to amplify the distal SBE region (-1107 to -947) within the Ppif promoter.
Cloning of Ppif promoter into luciferase reporter and reporter activity assay: Mouse Ppif promoter fragments containing the -1.1kb to -0.1kb, or -0.37kb to -0.1kb, or -0.62 to -0.45kb, or -1.1 kb to -0.95kb, or -0.62kb to -0.1kb, or -1.1kb to -0.45kb region were PCR amplified from purified C3H/HeJ mouse DNA. Primers were designed to introduce CTCGAG XhoI 5′ and AAGCTT HindIII 3′ flanking sequences. The fragments were then subcloned into the XhoI 3'/HindIII 5' site of the promoterless pGL4.10 luciferase reporter vector. The correct insert orientation of the resulting promoter reporters was verified by sequencing. To evaluate promoter activities, the constructed Ppif-Luc reporters were transfected into ST2, MC3T3e1, or 143b-OS cells at 0.8μg per well in twelve-well plates. The promotorless renilla luciferase vector pRL (Promega, Madison, WI, USA) was cotransfected at 50ng per well as a reference. Smad1 activity was further activated with either 0.8μg/mL pCMV-Smad1 co-transfection, 50ng/mL BMP2, or osteogenic media induction. The role of BMP/Smad was delineated by co-transfecting inhibitory Smad7 using 0.8 μg/mL pCMV-Smad7 vector, or using BMP inhibitor, Noggin (R&D systems 1967-NG-025/CF) at 0.1 µg/mL. Firefly and renilla luciferase activities were measured using an Optocomp 1 luminometer and a Dual Luciferase Reporter Assay System (Promega) according to the manufacturer's protocol. The firefly luciferase signal was normalized to renilla luciferase signal and expressed as relative luminescence units fold change to pCMV empty vector control.
Statistics: A power analysis on normalized biomechanical data was performed since it showed the highest variance. It was determined that some quantitative outcomes would require 6 mice per group. We set the significance level at 5% (α=0.05) and Type II error (β) to ≤20%. For statistical analysis, we compared the difference of two simple groups independently, therefore an unpaired t-test was used when the frequency distribution of the differences between the two groups fitted a normal distribution. When left and right tibiae from the same animal was compared, we used a paired t-test. Although we analyzed independent variables and therefore independent hypothesis in the multi-group graphs (CypD f/f : Cre+ or Cre-x AAV-DIO or PBS), we performed Ordinary one-way ANOVA using Dunnett's multiple comparisons test with a single pooled variance to further validate our statistical findings significance. Since no differences in significance were found, we maintained our t-test results.       to account for differences in bone phenotype between animals. Cre + mice showed decreased bone volumetric parameters and mechanical properties when compared to Cremice; I) Cortical thickness (Ct Th); J) Bone over total volume (BV/TV); K) Torsional rigidity; L) Yield torque; M) Maximum torque. Plots show the actual data points from six independent mice per group, calculated means and P value determined by an unpaired t-test. Paired ttest was used when left and right tibia from the same mouse were compared. Specimens' genotype guide: