Cyclophilin D Extramitochondrial Signaling Controls Cell Cycle Progression and Chemokine-directed Cell Motility*

Background: Signals originated from mitochondria can affect other intracellular compartments. Results: Decreased levels in mitochondrial cyclophilin D promote cell proliferation and cell motility via chemokine/STAT3 signaling. Conclusion: Mitochondria regulate nuclear gene expression. Significance: Interorganelle signaling affects cell proliferation and cell motility. Mitochondria control bioenergetics and cell fate decisions, but how they influence nuclear gene expression is understood poorly. Here, we show that deletion or reduction in the levels of cyclophilin D (CypD, also called Ppif), a mitochondrial matrix peptidyl prolyl isomerase and apoptosis regulator, results in increased cell proliferation and enhanced cell migration and invasion. These responses are associated with extensive transcriptional changes, modulation of a chemokine/chemokine receptor gene signature, and activation of the pleiotropic inflammatory mediator, STAT3. In the absence of CypD, active STAT3 enhances cell proliferation via accelerated entry into S-phase and stimulates autocrine/paracrine cell motility through Cxcl12-Cxcr4-directed chemotaxis. Therefore, CypD directs mitochondria-to-nuclei inflammatory gene expression in normal and tumor cells. This pathway may contribute to malignant traits under conditions of CypD modulation.

Cyclophilin D (CypD, Ppif) 2 is a mitochondrial matrix protein with cis-trans peptidyl prolyl isomerase activity (1) that participates in the regulation of organelle permeability transition and the initiation of apoptosis (2). Considerable effort has been devoted to elucidate a role of CypD in various aspects of mitochondrial cell death (3), map its molecular arrangement in an organelle permeability transition pore (4), and validate its therapeutic value as a target for improved cytoprotection (5) or, conversely, induction of apoptosis (6). Dynamically regulated by posttranslational modifications, including nitrosylation (7), deacetylation (8), and chaperone-directed folding (9), CypD has been implicated in additional mitochondrial functions, for instance stabilization of the glycolytic enzyme hexokinase II (10) at the outer membrane (11) and quality control of damaged organelles via modulation of autophagy (12). However, the possibility that CypD may connect to extramitochondrial signaling mechanism(s) and, in particular, changes in nuclear gene expression has not been investigated.
The concept of interorganelle signaling has been modeled on the cellular response to proteotoxic stress (13). Accordingly, defects in the protein folding environment in the endoplasmic reticulum initiates a complex gene expression response in the nucleus (14) aimed at restoring homeostasis while also dampening mitochondrial cell death pathways (15). There is also evidence that signals emanating from a mitochondrial unfolded protein response (16) may affect cellular homeostasis, participating in metabolic reprogramming, especially in tumor cells (9), and inducing secondary endoplasmic reticulum-dependent responses (17). Although some of these mitochondria 3 nuclei "retrograde" signaling mechanisms (18) are evolutionarily conserved and may affect broad homeostatic pathways, including aging, metabolism, and adaptive responses (19), their effector molecules in mitochondria have not been delineated clearly. In this study, we explored a potential role of CypD in mitochondria-to-nuclei interorganelle signaling.

EXPERIMENTAL PROCEDURES
Cell Lines-Human glioblastoma LN229 cells, monocytic THP-1 cells, breast adenocarcinoma MCF-7 cells, non-transformed NIH3T3 fibroblasts, and pancreatic adenocarcinoma MiaPACA cells were obtained from the ATCC and maintained in culture as recommended by the supplier. LN229 cells stably transfected with control, non-targeting shRNA or shRNA directed to CypD have been described (20). Clones of stable transfectants were maintained in the presence of 800 g/ml of G418 sulfate (Sigma). Tumor cell types were maintained in high-glucose DMEM (except for THP-1 cells, which were propagated in RPMI 1640) with glutamine and supplemented with 10% FBS and 1% penicillin/streptomycin. WT or CypD Ϫ/Ϫ mouse embryonic fibroblasts (MEFs) were generously provided by Dr. Nika Danial, Harvard Medical School, and described previously (1).
Western Blotting-Protein lysates were prepared from various cell types in extraction buffer containing 150 mM NaCl, 50 mM Tris (pH 8.0), and 1% Triton X-100 in the presence of protease and phosphatases inhibitors. 40 g of protein was separated on SDS-polyacrylamide gels, transferred to PVDF membranes, and incubated with primary antibodies of various specificities for 16 h at 4°C. After washes and incubation with secondary antibodies, immunoreactive bands were detected by chemiluminescence.
Fluorescence-activated Cell Sorting-Various cell types were fixed in glacial 70% ethanol for 24 h followed by incubation with propidium iodide (2.5 g/ml) in the presence of RNAse A for an additional 16 h at 4°C. Ten thousand events were acquired on a Calibur flow cytometer followed by analysis using Cell Quest Pro software (BD Biosciences). For cell cycle synchronization experiments, cells were arrested at the G 1 phase of the cell cycle by double thymidine block and processed for analysis of DNA content at increasing time intervals (2-24 h) after release.
For determination of reactive oxygen species (ROS), cells were harvested, washed, and incubated in PBS (pH 7.4), in the presence of a carboxyl derivative of fluorescein (CM-H 2 DCFDA, Invitrogen) at a final concentration of 10 M. After 30-min incubation, cells were centrifuged, washed, and suspended in PBS (pH 7.4) for flow cytometry analysis at excitation and emission wavelengths set at 488 nm and 575 nm, respectively.
Cell migration and Cell Invasion Assays-All migration experiments were carried out using uncoated transwell chambers (Becton Dickinson, catalog no. 353097) using NIH3T3conditioned medium in the lower compartment as a chemoattractant. In the case of poorly invasive MCF-7 cells, transwell membranes were coated with collagen (15 g/ml) before analysis of cell motility. All invasion experiments were carried out using Matrigel-coated transwell chambers (Becton Dickinson, catalog no. 354483) with NIH3T3-conditioned medium in the lower compartment as a chemoattractant.
Gene Expression Profiling-Total RNA was extracted using the TRI reagent (Ambion) method and treated with DNAse I (Fermentas). For microarray analysis, RNA quality was determined using the bioanalyzer (Agilent). Only samples with RNA Integrity numbers Ͼ 7.5 were used for further studies. Three WT (CypD ϩ/ϩ ) and three CypD knockout (CypD Ϫ/Ϫ ) samples were used. Equal amounts (400 ng) of total RNA were amplified under the different conditions and hybridized to the HumanWG-6 v2 whole genome bead arrays as recommended by Illumina.
Transfections-Various cell types were seeded at 5 ϫ 10 4 /ml and transfected with control non-targeting siRNA or siRNA pool directed to the individual target gene(s) using Oligofectamine (25 pmol/ml). Plasmid DNA was transfected using Lipofectamine (0.5 g/ml) in Opti-Mem, which was replaced with complete culture medium after 5-h incubation.
STAT3 Luciferase Reporter Assay-Cells were transfected with 0.25 g of STAT3 reporter plasmid in the presence of 0.025 g of Renilla plasmid as normalizer, using X-tremeGENE transfection reagent (2 ml). After 24 h, cells were lysed and analyzed for normalized fluorescence expression in a luminometer.
Chemokine Release-Cells (5 ϫ 10 4 cells/ml) were serumstarved for 24 h, and medium was collected, centrifuged, and incubated onto membranes (catalog no. AAM-CYT-3-4, Raybiotech, Norcross, GA) spotted with antibodies against 62 different mouse cytokines. After washes, membranes were incubated with a mix of biotin-conjugated antibodies to the same cytokines, and immunoreactive spots were detected using HRP-conjugated streptavidin and chemiluminescence. Dots were quantified using ImageJ and normalized against positive controls.
Microarray Data Analysis-Illumina BeadStudio v.3.0 software was used to export expression levels and detection p values for each probe of each sample. Arrays were quartilenormalized between each other and filtered to remove noninformative probes (with detection p Ͼ 0.05). Differentially expressed genes between WT and CypD Ϫ/Ϫ groups were identified using two-tail unpaired Student's t test. The false discovery rate (FDR) was calculated according to the Storey procedure (22), and only genes with FDR Ͻ 5% were called significant. Heat maps for a list of genes were generated using two-way hierarchical clustering with a normalized Euclidean distance to cluster samples and a Spearman correlation distance to cluster the genes. Heat map color intensities were proportional to a value calculated as a ratio between the gene expression in a single sample and the geometric mean expression of the gene across all samples. A pathway enrichment analysis was carried out with Ingenuity Pathways Analysis software using Ingenuity Core Analysis (IPA 8.0, Ingenuity Systems) with Benjamini-Hochberg correction for multiple testing and using p Ͻ 0.05 as a significance threshold. Enrichments of gene ontology terms and INTERPRO annotations in a gene list were done with DAVID (23) software. Results were filtered to satisfy criteria of FDR Ͻ 5% and fold enrichment Ͼ 4.
Statistical Analysis-Data were analyzed using GraphPad Prism 4.03 for Windows. Unless otherwise stated, data analyses were carried out using unpaired Student's t test. p Ͻ 0.05 was considered statistically significant.

CypD Regulation of Cell Cycle
Progression-We began this study by examining a potential role of CypD in extramitochondrial signaling in normal or tumor cells representative of different genetic makeups. We found that glioblastoma LN229 cells with stable shRNA knockdown of CypD (20) exhibited increased cell proliferation compared with control cells transfected with non-targeting shRNA (Fig. 1A). Deletion of CypD (CypD Ϫ/Ϫ ) in MEFs (1) also resulted in accelerated cell proliferation compared with WT MEFs (Fig. 1B). To test the specificity of this response, we next transfected CypD in CypD Ϫ/Ϫ MEFs and looked at changes in cell proliferation. Consistent with the data above, CypD Ϫ/Ϫ MEFs showed increased cell proliferation compared with WT cultures (Fig. 1C). Conversely, re-expression of CypD in CypD Ϫ/Ϫ cells reversed this effect and restored proliferation rates quantitatively comparable with WT MEFs (Fig. 1C). In contrast, transfection of a CypD mutant cDNA defective in peptidyl prolyl isomerase activity did not modulate proliferation of CypD Ϫ/Ϫ MEFs (Fig. 1D).
When analyzed in cell cycle-synchronized cultures, siRNA knockdown of CypD in LN229 (Fig. 1E) or breast adenocarcinoma MCF-7 cells (data not shown) resulted in faster and more sustained transition into S-phase, with accumulation of a larger fraction of cells with G 2 /M DNA content, compared with control transfectants (Fig. 1F). CypD Ϫ/Ϫ MEFs exhibited a similar phenotype with appearance of a larger S-phase population (data not shown) and further expansion of the G 2 /M fraction (Fig. 1G).
CypD Regulation of Cell Motility and Invasion-Stable shRNA depletion of CypD in LN229 cells ( Fig. 2A) or acute knockdown of CypD in pancreatic adenocarcinoma MiaPACA cells (Fig. 2B) produced a second phenotype of increased cell migration and cell invasion compared with control transfectants (Fig. 2, A and B). Similar results were obtained in a more physiologic three-dimensional model of organotypic spheroids embedded in a collagen matrix where CypD knockdown considerably increased cell invasion compared with control transfectants (Fig. 2C). Consistent with modulation of cell motility, CypD Ϫ/Ϫ MEFs exhibited remodeling of the actin cytoskeleton with increased expression of cellular protrusions and lamellipodia formation compared with a rounded morphology of WT MEFs (Fig. 2D). In reconstitution experiments, transfection of a CypD cDNA reversed the increased cell migration and invasion of CypD Ϫ/Ϫ MEFs, whereas a control vector had no effect (Fig. 2E).
CypD Deletion Results in Transcriptional Modulation of a Chemokine Gene Expression Signature-To begin elucidating how changes in CypD expression may affect cell behavior, we next looked at a potential differential mitochondrial production of ROS in control or CypD-depleted cells. In these experiments, ROS production was indistinguishable in control cul-tures or LN229 cells stably silenced for CypD or CypD Ϫ/Ϫ MEFs (Fig. 3A). Also, increasing concentrations of the ROS scavenger N-acetyl cysteine did not significantly affect proliferation of CypD Ϫ/Ϫ MEFs or CypD-silenced LN229 cells compared with control cultures (Fig. 3B).
On the basis of these results, we next asked whether CypD Ϫ/Ϫ MEFs secreted active chemokine(s) in their conditioned medium, potentially supporting autocrine/paracrine cell motility. Accordingly, exposure of MCF-7 or monocyte THP-1 cells to CypD Ϫ/Ϫ conditioned media resulted in increased cell migration compared with control cultures incubated with NIH3T3 or WT MEF-conditioned media (Fig. 4D). The enhancing effect on cell motility was concentration-dependent for increasing doses of CypD Ϫ/Ϫ -conditioned media (Fig. 4E). To identify which secreted chemokine was potentially responsible for the enhanced migratory phenotype induced by CypD loss, array profiling studies were carried out next. In these experiments, supernatants from CypD Ϫ/Ϫ MEFs exhibited a dramatic increase in Cxcl12 compared with WT cells (Fig. 4F). In contrast, IL-6 levels were increased less prominently, and other chemokines on the array did not change significantly between the two cell types (Fig. 4F). Consistent with these data, supernatants from CypD Ϫ/Ϫ MEFs contained high levels of Cxcl12 compared with WT cultures (Fig. 4G). Functionally, a neutralizing antibody to Cxcl12 inhibited cell migration mediated by CypD Ϫ/Ϫ -conditioned media (Fig. 4H), whereas a nonbinding IgG had no effect (not shown). Conversely, a neutralizing antibody to IL-6 did not significantly affect tumor cell migration or invasion in the presence of CypD Ϫ/Ϫ -conditioned medium (Fig. 4I).
Signaling Requirements of CypD-directed Cellular Responses-Analysis of transcriptional changes in CypD Ϫ/Ϫ MEFs suggested a potential involvement of STAT3 (supplemental Fig.  S1) (24). Consistent with these predictions, CypD knockdown in LN229 cells or deletion of CypD in MEFs resulted in increased STAT3 luciferase promoter activity compared with control cultures (Fig. 5A). In addition, CypD Ϫ/Ϫ MEFs exhibited increased mRNA expression of a panel of STAT3 target genes, compared with WT MEFs, by quantitative PCR amplification (Fig. 5B). Biochemically, this was associated with increased phosphorylation of STAT3 on Tyr-705, a marker of transcriptional activity (Fig. 5C). In reconstitution experiments, transfection of a CypD cDNA attenuated STAT3 phosphorylation in CypD Ϫ/Ϫ MEFs, whereas a control vector had no effect (Fig. 5C). Confirming the specificity of this response, CypD targeting in different cell types did not affect STAT3 phosphorylation on Ser-727 (Fig. 5D), an activation marker of mitochondria-localized STAT3 (25,26).

STAT3 Regulation of CypD-directed Cell Motility and Cell Proliferation-Pharmacologic inhibition of the STAT3 SH2
domain with a small molecule inhibitor, Stattic (21), reversed the increased migration and invasion of CypD-silenced LN229 cells compared with control cultures (Fig. 6A). In contrast, Stattic had no effect on tumor cell migration or invasion in the absence of CypD knockdown (Fig. 6A). Similarly, Stattic reversed the increased migration and invasion of CypD Ϫ/Ϫ MEFs but was without effect in WT MEFs (Fig. 6B). The more pronounced effect of STAT3 inhibition in CypD Ϫ/Ϫ MEFs (Fig.  6B) compared with LN229 shRNA transfectants (A) may reflect the incomplete knockdown of CypD in this cell type. As an independent approach, we next silenced STAT3 by siRNA and examined changes in cell motility in the presence or absence of simultaneous CypD knockdown. STAT3 silencing inhibited the migration and invasion of LN229 cells (Fig. 6C), supporting a role of this pathway in cell motility (24). Consistent with the data above, CypD knockdown increased both the migration and invasion of LN229 cells, and these responses were reversed by simultaneous STAT3 knockdown (Fig. 6C). To independently validate a role of CypD in this response, we next carried out additional reconstitution studies by transfecting a control vector or CypD cDNA in CypD Ϫ/Ϫ MEFs. In these experiments, reconstitution of CypD Ϫ/Ϫ MEFs with CypD, but not control cDNA, reduced cell proliferation (Fig. 6D, top panel), migration (center panel), and invasion (bottom panel) to the levels of WT CypD MEFs.
We next asked whether STAT3 signaling was also involved in the increased cell proliferation induced by CypD targeting. Analysis of cell cycle-synchronized cultures revealed that STAT3 was constitutively hyperphosphorylated on Tyr-705 in CypD Ϫ/Ϫ MEFs throughout cell cycle progression (Fig. 7A). In contrast, synchronized WT MEFs had largely undetectable levels of Tyr-705-phosphorylated STAT3 at all cell cycle phases (Fig. 7A). LN229 cells silenced for CypD also exhibited constitutive hyperphosphorylation of STAT3 on Tyr-705 at all cell cycle transitions (Fig. 7B). In contrast, phosphorylation of p53 on Ser-15 was unchanged in control or CypD-silenced LN229 cells (Fig. 7B). Stattic treatment prevented the accumulation of LN229 cells with S-phase DNA content in response to CypD knockdown, whereas control transfectants treated with Stattic progressed through the cell cycle and transitioned into S-phase  ( Fig. 7C). Consistent with these data, Stattic inhibited the proliferation of CypD-silenced LN229 cells (Fig. 7D) as well as CypD Ϫ/Ϫ MEFs (Fig. 7E), compared with control cultures.

DISCUSSION
In this study, we have shown that deletion or decreased expression of mitochondrial CypD activates interorganelle signaling in normal or tumor cell types, leading to transcriptional changes in gene expression, modulation of a chemokine/ chemokine receptor signature, and activation of the pleiotropic inflammatory mediator, STAT3. In turn, this resulted in accelerated cell cycle progression through S-phase and enhanced chemokine-dependent autocrine/paracrine cell migration and invasion.
CypD is the only component of a mitochondrial permeability transition pore (2) required for certain forms of cell death, for instance, oxidative stress (1,3,27). Although the role of CypD in this process has not been completely elucidated (11), there is evidence that inhibition of CypD protects from apoptosis and may have therapeutic potential in vivo (28), limiting the extent of tissue damage during cerebral (1), vascular (27), or pancreatic ␤ cell (29) injury. Although additional functions of CypD in autophagy (12) and regulation of glycolysis (11) have been proposed, a role of this molecule in nuclear control of gene expression, chemokine responses, and STAT3-dependent signaling (this study) has not been proposed previously.
In this context, active STAT3 is a recognized effector of multiple inflammatory and homeostatic responses (30). Activated by cytokines, IL-6, IL-10, and IL-23, growth factor receptors, and non-receptor tyrosine kinases (Abl, Src, Syk), STAT3-dependent transcription of target genes maintains Th17 cells (31), modulates B and T regulatory populations (32), and contributes to antigen presentation by dendritic cells (33). STAT3 signaling is also exploited in cancer (34), influencing disparate mechanisms of cell differentiation, proliferation, apoptosis, angiogenesis, and cell motility/invasion (24,35). Here, loss of mitochon-drial CypD resulted in increased production of IL-6 (35), and, to a much greater extent, Cxcl12 (36), two well known inducers of STAT3. In turn, this was associated with canonical markers of STAT3 activation, including phosphorylation on Tyr-705, enhanced promoter activity, increased mRNA expression of STAT3 target genes, and STAT3-dependent increased cell proliferation and cell motility. Pharmacologic blockade of JAK2, but not Src (35), reversed STAT3 responses, consistent with a role of cytokine signaling in STAT3 induction after CypD loss.
Intriguingly, an oncogenic pool of STAT3 has been identified in mitochondria (37) and implicated in regulation of oxidative phosphorylation (38), potentially modulated by the SIRT1 deacetylase (25). Although the biochemical requirements of this pathway are still controversial (39), there is evidence that mitochondrial STAT3 may improve organelle performance and antagonize apoptosis during acute vascular injury (40), consistent with a non-transcriptional function of STAT3 in cellular responses (41). At variance with this paradigm, STAT3 induced by CypD depletion did not significantly accumulate in mitochondria, and its phosphorylation on Ser-727, a marker of the mitochondrial pool of the molecule (25,26), was not affected in CypD-targeted cells.
The mechanistic links that connect changes in mitochondrial CypD levels to STAT3 activation remain to be fully elucidated. There is prior evidence that changes in mitochondrial integrity or function can induce retrograde modulation of nuclear gene expression (18). Potential mediators of these interorganelle responses may include modulation of Ca 2ϩ homeostasis (42), production of ROS (43), proteotoxic stress (44), or, more recently, metabolic unbalance with secondary activation of an endoplasmic reticulum-dependent unfolded protein response (9). At variance with previous models of retrograde signaling, CypD Ϫ/Ϫ cells did not produce excess ROS, which could influence gene expression, and their enhanced proliferative kinetics were unaffected by a ROS scavenger. Conversely, other soluble mediators of mitochondria retrograde signaling can be postulated, in particular changes in Ca 2ϩ homeostasis, which, reminiscent of CypD depletion (3,27), may influence nuclear gene expression (45). Consistent with this model, CypD Ϫ/Ϫ cells exhibited increased NFB-dependent transcription 3 . This is a common hallmark of mitochondria 3 nuclei retrograde signaling (18), including in response to Ca 2ϩ perturbations (45), and pivotal for the activation of many of the cytokines found up-regulated by CypD deficiency (24).
The pathophysiological context of these observations is highlighted by the critical role of STAT3 and chemokine signaling in tumors and their microenvironment. In particular, the ligand-receptor pair Cxcl12-Cxcr4, which in our studies mediated autocrine/paracrine cell motility after depletion of CypD, has been implicated in the interplay between tumor cells and stroma, promoting stem cell mobilization (46), cell invasion, recruitment of bone marrow-derived myeloid cells, angiogenesis (47), and further activation of STAT3 (36). In this scenario, constitutively decreased levels of CypD in malignancies of the cervix and pancreas, as predicted by bioinformatics analysis 3 , may contribute to tumor progression via inhibition of oxidative cell death (1,27) and activation of chemokine-and STAT3-dependent inflammatory signaling (this study). As an alternative scenario, therapeutic inhibition of CypD, either as a strategy to enhance cardioprotection during ischemia-reperfusion injury (28) or as a consequence of cyclosporine A-mediated immunosuppression (48), may enhance local inflammation and cooperate with loss of immune surveillance to favor malignant transformation in organ transplant patients (49).
In summary, we have identified CypD (1) as a pivotal effector of a novel form of mitochondrial retrograde signaling (18) that promotes STAT3 activation and chemokine/chemokine receptor-dependent cell motility. These studies reinforce the concept of interorganelle signaling as a regulatory network poised to affect broad cellular homeostatic responses (9) and identify an unexpected role of CypD (1) in extramitochondrial control of inflammatory gene expression.