Syntaphilin Ubiquitination Regulates Mitochondrial Dynamics and Tumor Cell Movements

Although most tumors reprogram their metabolism towards glycolysis even when oxygen is present, that is the “Warburg effect” (1), there is now mounting evidence that mitochondria continue to play a fundamental role in cancer (2, 3). This is especially important in advanced disease, where mitochondrial function has been linked to tumor repopulation after oncogene ablation (4), drug resistance (5), and tumor progression in vivo (6). In addition, mitochondria have been implicated in tumor cell movements and metastasis (7) through modulation of oxidative metabolism (8), organelle biogenesis (8), buffering of oxidative stress (9), and cycles of organelle fusion and fission, that is dynamics (10).

In this context, recent evidence has demonstrated that in response to stress stimuli of the microenvironment, tumor cells reposition mitochondria from their constitutive perinuclear localization to the peripheral (or cortical) cytoskeleton (11). In turn, these cortical mitochondria function as a “spatiotemporal” energy source to fuel membrane lamellipodia dynamics, kinase signaling and actin cytoskeleton remodeling, promoting increased cellular chemotaxis and invasion (11–13).

There are intriguing similarities between these observations in cancer (11–13) and the process of mitochondrial trafficking in neurons, which also repositions mitochondria at sites of high energy demands (14). These suggest that tumors may hijack mitochondrial trafficking as a bioenergetics requirement of cell motility, especially under stress (11), facilitating metastatic dissemination in vivo. Accordingly, a genome-wide siRNA screen has recently identified syntaphilin (SNPH), a molecule known to inhibit mitochondrial trafficking in neurons (15), as a regulator of tumor cell movements (16). Although considered neuronal-specific, SNPH is, in fact, broadly expressed in cancer, suppresses mitochondrial trafficking to the cortical cytoskeleton, and restricts tumor cell movements and metastasis in vivo (16). In addition, analysis of genomic databases suggests that SNPH is progressively downregulated or lost during tumor progression, correlating with worse disease outcome (16). How this pathway is regulated, however, has remained elusive. In this study, we examined a role of SNPH ubiquitination in the control of mitochondrial trafficking and metastasis.

Prostate adenocarcinoma PC3 cells were obtained from the ATCC along with authentication, and maintained in culture according to the supplier's specifications. Yumm 1.7 cells were a gift from Dr. Marcus Bosenberg (Yale University, New Haven, CT). All cell lines were obtained within 5 years, stored in liquid nitrogen and maintained in culture for less than 30 passages. Mycoplasma detection is carried out in our teams by direct PCR amplification of culture supernatants with Bioo Scientific Mycoplasma Primer Sets (Catalog No. 375501) and Hot Start polymerase (QIAGEN Catalog No. 203203). Conditioned media used for cell motility assays was prepared from exponentially growing cultures of NIH3T3 cells in DMEM supplemented with 4.5 g/L d-glucose, sodium pyruvate, 10 mmol/L HEPES and 10% FBS for 48 hours. Nutrient starvation was carried out in the presence of 0.8% FBS for 16 hours.

An affinity-purified, custom-made rabbit polyclonal antibody against human SNPH (aa 207-221; NEO Group Inc.) has been described (17). Antibodies to β-tubulin, β-actin, and Flag were from Sigma. Antibodies to MTC02, Kif2a, and Clasp-1 were from Abcam. Antibodies to ubiquitin, VDAC, C terminus of Hsc70-interacting protein (CHIP), USP7, Ser616-phosphorylated Drp1, Drp1, and Rac1 were from Cell Signaling Technology. CHX was from Sigma. An antibody to ubiquitin K11 linkage was from EMD Millipore. Antibodies to ubiquitin K48 and K63 linkages were from Cell Signaling Technology. MitoTracker Green, Phalloidin Alexa Fluor 488, Mitotracker-Deep Red FM, and secondary antibodies for immunofluorescence were from Molecular Probes.

The following recombinant proteins were used: UBE1 (E1, UBPBio; Catalog No. #B1100), His-UBE2D1 (E2, UBPBio; Catalog No. C1400), His-UBE2D3 (E2, UBPBio; Catalog No. C1600), His-UBE2E3 (E2, UBPBio; Catalog No. C2000), ubiquitin (UBPBio; Catalog No. E1100). Recombinant CHIP (Stub1) was purchased from Boston Biochem. A SNPH mutant lacking the kinesin-binding domain (ΔKBD, Δ337-422 aa) has been characterized previously (16). Ubiquitination-defective SNPH mutants K111R, K153R, K175R, K295R, and K415R were generated using Stratagene QuikChange II XL Site-Directed Mutagenesis Kit (Agilent Technologies), and confirmed by DNA sequencing. Cells were transfected with 2 μg of pDNA plus 4 μL X-treme gene HP (Roche) for 24 hours in complete medium, washed and processed for individual experiments. Adenoviral vectors expressing LacZ, wild-type (WT) SNPH, or the various ubiquitination-defective SNPH mutants were produced using Gateway Technology (Thermo Fisher Scientific) with pDONR221 vector and recombination into the adenovirus expression vector pAd/CMV/V5-DEST. After digestion with PacI, the individual constructs were transfected in 293A cells for production of adenoviruses after 7 days.

Gene knockdown experiments by siRNA were carried out as described (16). The following sequences were used: control, ON-TARGET plus nontargeting siRNA pool (Dharmacon D-001810), human SNPH siRNA (Dharmacon L-020417 or Santa Cruz sc-41369), Kif2a siRNA (Dharmacon L-004959-00), and Drp1 siRNA (Dharmacon L-012092-00). USP7 siRNA and CHIP siRNAs were synthesized by Dharmacon. The USP7 siRNA sequence was 5′-AAATTATTCCGCGGCAAA-3′ and CHIP siRNA sequence was 5′-CGAGCGCGCAGGAGCTCAA-3′. Tumor cells were transfected with the individual siRNA pools at 30 nmol/L in Lipofectamine RNAiMAX (Invitrogen) at a 1:1 ratio (vol siRNA 20 μM: vol Lipofectamine RNAiMAX). After 48 hours, transfected cells were confirmed for protein knockdown by Western blotting, and processed for subsequent experiments. For reconstitution experiments, PC3 cells were transiently silenced for endogenous SNPH expression using siRNA sequence 5′-GCUGCUGACAUUACCAUUAUUUU-3′ targeting the 3′-UTR of human SNPH mRNA. Cultures were confirmed for silencing of endogenous SNPH expression by Western blotting and transfected with WT SNPH or other SNPH mutant constructs followed by functional analyses. Alternatively, PC3 cells stably expressing shRNA targeting the 3′UTR of human SNPH: TRCN 0000147900 and TRCN 0000128545 were used and reconstituted with the various SNPH constructs (17). An empty pLKO-based lentivirus was used as control, and selection of stable clones was carried out in the presence of puromycin (2 μg/mL), as described (17).

Protein lysates were prepared in RIPA buffer (150 mmol/L NaCl, 1.0% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, 50 mmol/L Tris, pH 8.0) in the presence of EDTA-free Protease Inhibitor Cocktail (Roche) and Phosphatase Inhibitor Cocktail (Roche). Equal amounts of protein lysates were separated by SDS gel electrophoresis, transferred to PVDF membranes, and incubated with primary antibodies of various specificities. Protein bands were visualized by chemiluminescence. For ubiquitination analysis, cells were washed with 10 mmol/L N-ethylmaleimide containing PBS. The cell pellet was lysed by sonication in 50 μL of 2% SDS containing 50 mmol/L Tris-HCl (pH 7.5) and boiled at 95°C in a heat block for 10 minutes. Nine hundred and fifty microliters of 50 mmol/L Tris-HCl (pH 7.5) buffer was added to the boiled lysates followed by centrifugation. For immunoprecipitation experiments, aliquots of total cell lysates from control or PC3 transfectants were incubated with anti-Flag M2 affinity agarose gel (sigma) for 16 hours at 4°C. The immune complexes were separated by SDS gel electrophoresis, and analyzed by Western blotting. For protein half-life, PC3 cells transfected as indicated in individual experiments were incubated with 100 μg/mL of the protein synthesis inhibitor cycloheximide (CHX), released in complete medium, and aliquots of cell lysates collected at increasing time intervals after release (2–10 hours) were analyzed by Western blotting. For in vitro ubiquitination reactions, Flag-SNPH was incubated with E1 (100 nmol/L), E2 (1 μmol/L), E3 (1–2 μmol/L), and Ub (100 μmol/L) in buffer containing 50 mmol/L Tris, pH 8.0, 0.1% Triton X-100, 5 mmol/L MgCl₂, 2 mmol/L ATP, 1 mmol/L DTT for 1 hour at 37°C. The reaction was terminated with addition of 4× SDS sample buffer and protein bands separated by electrophoresis were visualized by Western blotting.

Immunoprecipitates of WT Flag-SNPH were separated on an SDS-gel for approximately 5 mm followed by fixing and staining with colloidal Coomassie. The entire region of the gel containing protein was excised and digested with trypsin. Tryptic peptides were analyzed by LC/MS-MS on a Q Exactive Plus mass spectrometer (Thermo Scientific) coupled with a Nano-ACQUITY UPLC system (Waters). Samples were injected onto a UPLC Symmetry trap column (180 μm i.d. × 2 cm packed with 5 μm C18 resin; Waters), and tryptic peptides were separated by RP-HPLC on a BEH C18 nanocapillary analytical column (75 μm i.d. × 25 cm, 1.7 μm particle size; Waters) using a 95-min gradient. Eluted peptides were analyzed in data-dependent mode, where the mass spectrometer obtained full MS scans from 400 to 2,000 m/z at 70,000 resolution. Full scans were followed by MS-MS scans at 17,500 resolution on the 20 most abundant ions. Peptide match was set as preferred, the exclude isotopes option and charge-state screening were enabled to reject singly and unassigned charged ions. MS-MS spectra were searched using MaxQuant 1.5.2.8 (18) against the UniProt human protein database (August 2015). MS-MS spectra were searched using full tryptic specificity with up to two missed cleavages, static carboxamidomethylation of Cys, and variable oxidation of Met and protein N-terminal acetylation. For identification of ubiquitinated sites, variable modification of diglycine ubiquitin remnant on Lys was also included. Consensus identification lists were generated with FDRs of 1% at protein and peptide levels. As cutoff, we focused on candidate SNPH-interacting proteins with at least two MS-MS counts in both positive immunoprecipitation experiments, and either showed at least a five-fold enrichment of both MS protein intensities and MS-MS counts or had no spectra identified in the control samples. Fold-changes were calculated versus control values floored to 1,000,000 for intensity (minimal detected) and one for MS-MS counts.

PC3 cells silenced for endogenous SNPH by siRNA and reconstituted with WT SNPH or ubiquitination-deficient SNPH mutants were analyzed for ATP generation (BioChain, Catalog No. Z5030041) or oxygen consumption rate (OCR; ENZO Lifesciences, Catalog. No. ENZ-51045-1), according to the manufacturer's specifications.

Mitochondrial fractions were prepared using a Mitochondria Isolation Kit (Fisher Scientific), as described (17).

Cells (2 × 10⁴) growing on high optical-quality glass bottom 35-mm plates (MatTek Corporation) were treated with 100 nmol/L Mitotracker-Deep Red FM dye for 1 hour and imaged with a 63 × 1.40 NA oil objective on a Leica TCS SP8 X inverted laser scanning confocal microscope. Short duration time-lapse sequences were carried out on a Tokai Hit incubation chamber equilibrated to 37°C and 5% CO₂ bidirectional scanning at 8,000 Hz using a resonant scanner. Time lapse was performed for 1,000 seconds (10 seconds per frame). Individual 12-bit images were acquired using a white-light supercontinuum laser (2% at 645 nm) and HyD detectors at 2× digital zoom with a pixel size of 90 nm × 90 nm. A pinhole setting of 1 Airy Units provided a section thickness of 0.896 μm. Each time point was captured as a stack of approximately 11 overlapping sections with a step size of 0.5 μm. At least five single cells per condition were collected for analysis. Initial post-processing of the three-dimensional (3D) sequences was carried out with Leica LAS X software to create an iso-surface visualization. Time-lapse sequences were imported into ImageJ Fiji and individual mitochondria were manually tracked using the Manual Tracking plugin. Mitochondria (approximately 10 mitochondria per cell) were tracked along the stacks until a fusion event prevented continued tracking. The speed and distance for each time interval were used to calculate the mean speed and cumulative distance traveled by each individual mitochondria.

Mitochondrial activity was quantified using automated image analysis approaches. Mitochondria were segmented in each image frame and tracked throughout the image sequence. The segmentation and tracking used algorithms previously developed for measuring organelle dynamics (19), utilized from within the LEVER framework for live cell microscopy image analysis (20–22). Briefly, the segmentation uses an adaptive thresholding on the 3D image data, followed by a connected component analysis to identify individual mitochondria. A single segmentation parameter specifying the expected minimum object radius is required, here set at 0.5 μm for all of the image data processed. Tracking is done using the multitemporal association tracking (MAT) algorithm (19, 23) that uses a minimum-spanning tree approach to solve the data association problem in polynomial time across a time window here set at three frames into the future. No parameters beyond the minimum expected object radius are required for the tracking algorithm. Following segmentation and tracking, we identify fission and fusion events from the mitochondrial tracking results as follows. Tracks lasting a minimum number of frames (here set at three frames) that originate (or terminate) after the first frame of the video are considered to have originated from a fission (or fusion) event. For each movie, we normalize the number of fission and fusion events using the number of foreground voxels identified in the first frame of the image sequence. The result is a count of the fission and fusion events per frame per voxel for each movie.

Mitochondria/F-actin composite images were analyzed in ImageJ, as described (11). Briefly, the F-actin channel was used to manually label the cell boundary and a belt extending from the boundary towards the inside of the cell was marked as “cortical mask.” This cortical mask was subsequently applied to the mitochondrial channel to measure intensity at the cortical region, and normalized to total mitochondrial intensity per cell and cell area. For quantification of total mitochondrial mass, composite images were analyzed in ImageJ. The cell border was manually traced on the F-actin channel and this “cell mask” was subsequently applied to the mitochondria channel to measure the total mitochondria signal per cell. Maximum intensity was monitored to ensure no pixel saturation (e.g., max intensity <256 for 8-bit images). Mitochondrial mass was normalized to total cell area. A minimum of 30 cells was analyzed in each independent experiment to obtain mean values.

Cells were fixed in formalin/PBS (4% final concentration) for 15 minutes at 22°C, permeabilized in 0.1% Triton X-100/PBS for 5 minutes, washed, and incubated in 5% normal goat serum (NGS; Vector Labs) diluted in 0.3 M glycine/PBS for 60 minutes. Primary antibodies against Tom20 (diluted 1:300), β-tubulin (diluted 1:200), SNPH (diluted 1:500), or MTC02 (diluted 1:500) were added in 5% NGS/0.3 M glycine/PBS and incubated for 18 hours at 4°C. After three washes in PBS, secondary antibodies conjugated either to Alexa Fluor 488, TRITC, or Alexa Fluor 633 were diluted 1:500 in 5% NGS/0.3 M glycine/PBS and added to cells for 1 hour at 22°C. Where indicated, F-actin was stained with phalloidin Alexa Fluor 488 (1:200 dilution) for 30 minutes at 22°C. Slides were washed and mounted in DAPI-containing Prolong Gold mounting medium (Invitrogen). At least seven random fields were analyzed by fluorescence microscopy using a Nikon i80 microscope.

Experiments were carried out as described (11). Briefly, 2 × 10⁴ PC3 or Yumm 1.7 cells were seeded in four-well Ph+ Chambers (Ibidi) in complete medium and allowed to attach overnight. Videomicroscopy was performed over 10 hours, with a time-lapse interval of 10 minutes. Stacks were imported into ImageJ Fiji for analysis. Images were aligned according to subpixel intensity registration with the StackReg plugin for ImageJ Fiji (24). At least 25 to 30 cells were tracked using the Manual Tracking plugin for ImageJ Fiji, and the tracking data were exported into the Chemotaxis and Migration Tool v2.0 (Ibidi) for graphing and calculation of mean and standard deviation of speed, accumulated distance, and Euclidean distance of movement.

Experiments were carried out essentially as described (16) using Growth Factor Reduced Matrigel-coated 8 μm PET transwell chambers (Corning). Tumor cells were seeded in duplicates onto the coated transwell filters at a density of 1 × 10⁵ cells/well in medium containing 0.1% BSA and conditioned media from NIH3T3 was placed in the lower chamber as chemoattractant. Cells were allowed to invade for 16 to 24 hours, noninvading cells were scraped off the top side of the membranes and the invasive cells on the transwell insert were fixed in methanol. Membranes were mounted in medium containing DAPI (Vector Labs) and analyzed by fluorescence microscopy. Five random fields at ×10 magnification were collected for each membrane. Digital images were batch imported into ImageJ Fiji, thersholded and analyzed with the analyze particles function. Experiments were repeated three times.

Studies involving rodents were carried out in accordance with the Guide for the Care and Use of Laboratory Animals of the NIH. Protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of The Wistar Institute (Protocol No. #112625). For a syngeneic model of metastasis, Yumm 1.7 cells stably expressing mCherry (17) were transiently transfected with pCMV6 vector, or cDNA encoding WT SNPH or ubiquitination-defective SNPH mutant K111R or K153R mutant and selected with G418 at 400 μg/mL for 15 days. Stably transfected cells (3.5 × 10⁵) were injected into the flanks of syngeneic 8-week-old male C57BL6/NCr mice (NCI Inbred mice; Charles River Strain code no. #556). One week later, tumor cells disseminated to lungs were identified and quantitated based on expression of the mCherry transgene by IHC.

Lungs were fixed in neutral formalin (Fisher Scientific; SF93-4) for 36 hours, transferred to 70% ethanol for 3 days, and then paraffin-embedded. Five-μm sections were stained with a rabbit anti-mCherry polyclonal antibody (Novus; NBP2-25157) as follows. Slides were warmed at 50°C for 30 minutes; deparaffinized in xylene for 20 minutes, then xylene/ethanol 1:1 for 5 minutes; and rehydrated in alcohol series (100%, 95%, 90%, 70%, 50%, 30% ethanol and dH₂O, 5 minutes each). Antigen retrieval was done in citrate-based solution (Vector Laboratories; H-3300) at pH 6.0 in a pressure cooker for 5 minutes, followed by cooling to room temperature. Next, slides were washed once in PBS for 5 minutes and incubated in 3% hydrogen peroxide for 20 minutes. Slides were then washed three times with PBS (5 minutes each) and blocked in 10% normal goat serum/PBS for 1 hour at room temperature. The primary antibody was diluted 1:500 in 10% normal goat serum/PBS and incubated in a humidified chamber overnight. The next day, slides were washed three times with PBS for 5 minutes each, incubated with anti-rabbit HRP-labeled polymer (Dako; K4002) at 22°C for 30 minutes, and washed three times with PBS for 5 minutes each. Slides were developed with a DAB+ substrate chromogen system (Dako; K3467) for 30 minutes, rinsed in dH₂O, and stained with Mayer's hematoxylin solution (Sigma-Aldrich; MHS16) for 10 seconds. Slides were dehydrated in dH₂O and serial ethanol rinses of 5 minutes each, immersed in xylene for 15 minutes and mounted with Permount mounting medium (Fisher Scientific; SP15-100).

Nine lungs per group were examined for the presence of mCherry⁺ cells using a Nikon i80 upright microscope. An average of five to nine individual microscopy fields were quantified per lung condition and each mCherry⁺ cell was photographed at 40× magnification and manually counted using ImageJ software. For each animal, the average number of mCherry⁺ cells per lung was calculated and represented.

Data expressed as mean ± SD of multiple independent experiments or replicates of representative experiments out of a minimum of two or three independent determinations. Two-tailed Student t test or Wilcoxon rank sum test was used for two-group comparative analyses. For multiple-group comparisons, including quantification of disseminated tumor cells (DTC), ANOVA with Bonferroni posttest was used to compare the means between the groups and derived pairwise comparison's P values. All statistical analyses were performed using GraphPad software package (Prism 6.0) for Windows. A P value of <0.05 was considered as statistically significant.

We began this study by conducting a proteomics screen to identify SNPH-associated proteins in model prostate adenocarcinoma PC3 cells (16). Using predetermined mass spectrometry cutoff parameters (see Materials and Methods), we identified 217 putative interacting targets (Supplementary Fig. S1A), and the top 50 hits were further examined for pathway enrichment (Supplementary Fig. S1A; Supplementary Table S1). In this analysis, protein ubiquitination was identified as the most represented pathway in SNPH-associated molecules (P = 1 × 10⁻¹⁹; FDR, 0%) with a total of 34 proteins (Supplementary Table S1). Other SNPH-associated proteins clustered in RNA translation (P = 2 × 10⁻¹⁹; FDR, 0%) with 29 proteins, DNA damage response and G₂–M checkpoint (P = 5 × 10⁻⁷; FDR, 0%) with 13 proteins, cytoskeleton (P = 4 × 10⁻⁵; FDR, 0.07%) with 17 proteins, mitochondrial function (P = 0.0016; FDR, 1.58%) with 7 proteins, and endocytosis signaling (P = 7 × 10⁻⁵; FDR, 0.12%) with 6 proteins (Fig. 1A; Supplementary Table S1). As the most abundant class of SNPH-associated molecules, we next focused on protein ubiquitination.

Consistent with the data above, SNPH immune complex(es) precipitated from PC3 cells included heavily ubiquitinated proteins by Western blotting (Fig. 1B). Using ubiquitin-specific antibodies, most of SNPH ubiquitination under these conditions involved Lys63 linkages (Fig. 1C). Lys48 linkages were predominantly involved in the formation of higher molecular weight SNPH (poly)ubiqitination, whereas Lys27 linkages were not associated with SNPH (Fig. 1C). Based on these findings, we next used LC/MS-MS to identify potential ubiquitinated SNPH residues. Mass spectrometry analysis of tryptic digests of Flag-SNPH immunoprecipitated from PC3 cells revealed partial ubiquitination of K111, K175, K295, and K415 (Fig. 1D). An additional residue, K153 was reported as a site of SNPH ubiquitination in an independent global ubiquitome screen (25). Although modification of that residue was not detected in our analysis possibly because of sparse peptide coverage in the tryptic digests, we further pursued K153 as a putative SNPH ubiquitination site. In the predicted SNPH structure, K111 and K153 localize in the microtubule-binding domain (MBD), K175 and K295 in the interdomain region upstream of the LC8-binding domain and K415 in the kinesin-binding domain (KBD; Supplementary Fig. S1B). To begin validating these results, we first focused on SNPH ubiquitination sites in the MBD, K111 and K153 (Supplementary Fig. S1B). Accordingly, SNPH mutant K111R or K153R immunoprecipitated from PC3 cells showed significantly reduced ubiquitin labeling, compared with WT SNPH (Fig. 1E).

Next, we asked whether stress conditions typical of the tumor microenvironment affected SNPH ubiquitination. Exposure of PC3 cells to nutrient deprivation (0.8% FBS for 16 hours) increased SNPH ubiquitination (Supplementary Fig. S1C), compared with control cultures. Accordingly, nutrient-deprived PC3 cultures exhibited markers of cellular starvation, including increased phosphorylation of AMPK (Thr172), its downstream substrate acetyl-CoA carboxylase (ACC, Ser 79) and the autophagy regulator ULK1 (Ser555; Supplementary Fig. S1D). Conversely, exposure of PC3 cells to hypoxia (1% O₂) or oxidative stress (oxidant paraquat, PQ) reduced SNPH ubiquitination, compared with control (Supplementary Fig. S1E).

We next sought to identify the regulators of SNPH ubiquitination. Potential candidate SNPH-associated proteins in this pathway included the chaperone-regulated E3 ligase, CHIP, also known as STUB1, and the deubiquitinase, USP7 (Supplementary Fig. S1A; Supplementary Table S1). Accordingly, WT SNPH or SNPH mutant K111R or K153R readily associated with CHIP and USP7 in coimmunoprecipitation experiments from PC3 cells (Fig. 2A). Functionally, silencing CHIP by siRNA (Supplementary Fig. S2A) suppressed SNPH ubiquitination, compared with control transfectants (Fig. 2B). Reciprocally, siRNA silencing of USP7 (Supplementary Fig. S2A) or pharmacologic inhibition of USP7 with the small molecule antagonist P5091 increased ubiquitination of immunoprecipitated SNPH in vivo (Fig. 2C). We next reconstituted an in vitro SNPH ubiquitination reaction using recombinant proteins. In these experiments, the addition of recombinant CHIP plus UBE2D1 (UbcH5a, E2) and ubiquitin (E1) resulted in direct SNPH ubiquitination, in vitro (Fig. 2D). In contrast, UBE2E3 (UbcH9, E2) or UBE2D3 (UbcH5c, E2) was ineffective (Fig. 2D). To further validate a direct role of CHIP in SNPH ubiquitination, we next transfected serum-deprived PC3 cells with WT CHIP or a CHIP H260Q mutant, defective in E3 ligase function. Flag SNPH immunoprecipitated from these cells (Supplementary Fig. S2B) was heavily ubiquitinated in the presence of WT CHIP, whereas expression of CHIP H260Q mutant inhibited SNPH ubiquitination to the levels of nontransfected cells (Fig. 2E).

Next, we asked whether CHIP ubiquitination of SNPH affected protein stability. In CHX block experiments, there was no difference in the half-life of WT SNPH or the ubiquitination-defective SNPH mutant K111R and K153R (Supplementary Fig. S2C and S2D). Similarly, siRNA silencing of CHIP or USP7 (Supplementary Fig. S2A) did not affect SNPH half-life in CHX block experiments (Supplementary Fig. S2E and S2F). Conversely, siRNA silencing of CHIP increased the stability of RUNX1 used as control protein in CHX block experiments (Supplementary Fig. S2E and S2F). Finally, small molecule inhibitors of proteasomal (MG132) or lysosomal (Bafilomycin A1, BafA1) functions did not affect steady-state levels of endogenous SNPH or Flag-SNPH transfected in PC3 cells (Supplementary Fig. S2G). We conclude from these results that CHIP ubiquitination does not affect SNPH stability in tumor cells.

Our proteomics screen identified tubulin-binding proteins that co-associate with SNPH, including kinesin 13 family members of tubulin depolymerases, Kif2a and Kif2c, and the microtubule-stabilizing protein, Clasp1 (Supplementary Fig. S1A). Consistent with these results, WT or ubiquitination-defective SNPH mutant K111R or K153R SNPH immunoprecipitated from PC3 cells contained co-associated Kif2a and Clasp1 (Fig. 2A). In these experiments, SNPH mutant K111R or K153R showed considerably reduced co-association with tubulin, compared with WT SNPH (Fig. 2A). To test a potential role of SNPH ubiquitination in direct tubulin binding, we next depleted Kif2a by siRNA (Fig. 2F). In the absence of Kif2a, binding of SNPH mutant K111R to tubulin was considerably reduced, compared with WT SNPH (Fig. 2F). Similar results were obtained with a SNPH double mutant (DM) K111R/K153R, which also showed reduced co-association with tubulin after Kif2a silencing (Fig. 2F). As an independent approach, we next used a SNPH mutant lacking the KBD (ΔKBD), and therefore defective in Kif2a binding. In coimmunoprecipitation experiments, tubulin association with ΔKBD-SNPH was significantly reduced compared with WT SNPH (Fig. 2G), reinforcing a role of Kif2a in tubulin recognition in the SNPH complex. Under these conditions, ubiquitination-defective ΔKBD-SNPH–mutant K111R or K153R minimally co-associated with tubulin (Fig. 2G). Finally, we carried out similar experiments in nutrient-deprived cells, a condition that increases SNPH ubiquitination (Supplementary Fig. S1C). In these studies, nutrient deprivation increased tubulin binding to ΔKBD-SNPH in vivo, whereas immunoprecipitated ΔKBD-SNPH mutant K111R or K153R showed minimal association with tubulin (Fig. 2G).

To test the implications of SNPH ubiquitination, we next used Yumm 1.7 melanoma cells, which express low levels of endogenous SNPH (17). Expression of ubiquitination-defective SNPH mutant K111R or K153R was sufficient to increase the speed (Fig. 3A and B), and distance traveled by individual mitochondria in Yumm 1.7 cells (Fig. 3B), compared with WT SNPH. Next, we silenced endogenous SNPH in PC3 cells by siRNA and reconstituted the cultures with different SNPH constructs. Consistent with previous data (16), siRNA silencing of SNPH in PC3 cells was sufficient to drive the active repositioning of mitochondria from a perinuclear localization to infiltrate the cortical cytoskeleton (Fig. 3C). Reconstitution of these cells with WT SNPH suppressed organelle trafficking and maintained a perinuclear distribution of mitochondria (Fig. 3C). Conversely, expression of ubiquitination-defective SNPH mutant K111R or K153R restored mitochondrial infiltration to the cortical cytoskeleton (Fig. 3C). We next quantified mitochondrial trafficking in these conditions, and we found that expression of WT SNPH in siRNA-silenced cells blocked the cortical infiltration of mitochondria to levels of control transfectants (Fig. 3D and E). Conversely, expression of SNPH mutant K111R or K153R sustained mitochondrial trafficking to the cortical cytoskeleton, indistinguishably from SNPH-depleted cells (Fig. 3D and E). Similar results were obtained in Yumm 1.7 cells, where reconstitution of siRNA-silenced cultures with WT SNPH suppressed mitochondrial infiltration to the cortical cytoskeleton, whereas SNPH mutants K111R or K153R restored mitochondrial motility (Fig. 3F).

In agreement with previous results (12, 16, 26), siRNA silencing of endogenous SNPH increased 2D chemotaxis of PC3 cells, characterized by heightened speed of cell movements and longer distance traveled by individual cells (Fig. 4A and B). Expression of WT SNPH suppressed 2D chemotaxis, whereas reconstitution with SNPH mutant K111R or K153R restored the increased speed of cell movements and distance traveled by individual cells (Fig. 4A and B). Biochemically, this response was associated with increased phosphorylation of mTOR and Akt kinases (11), as well as GTPase Rac1 (Fig. 4C). In contrast, reconstitution of these cells with WT SNPH did not affect kinase or Rac1 activation (Fig. 4C). Similar results were obtained with Yumm 1.7 cells, where expression of ubiquitination-defective SNPH K111R or K153R promoted 2D chemotaxis (Supplementary Fig. S3A and B), and increased phosphorylation of mTOR and Akt (Fig. 4D). Finally, inhibition of SNPH ubiquitination by siRNA knockdown of CHIP increased tumor chemotaxis with greater speed and distance traveled by individual cells (Fig. 4E and F). Conversely, siRNA silencing of USP7 suppressed 2D tumor chemotaxis (Fig. 4E and F).

A pool of SNPH localized to the inner mitochondrial membrane affects bioenergetics in tumor cells (17), and the role of ubiquitination in this response was next investigated. Consistent with earlier findings (17), siRNA silencing of endogenous SNPH in PC3 cells reduced ATP production (Supplementary Fig. S3C) and oxygen consumption (Supplementary Fig. S3D), a marker of oxidative phosphorylation. Reconstitution of these cells with WT SNPH or SNPH mutant K111R or K153R restored ATP production and oxygen consumption (Supplementary Fig. S3C and S3D), demonstrating that CHIP ubiquitination does not affect bioenergetics. Based on these data, we next looked at a potential effect on tumor cell proliferation. Also consistent with earlier results, depletion of endogenous SNPH in tumor cells (Supplementary Fig. S3E) was associated with reduced cell proliferation (Supplementary Fig. S3F), potentially reflecting decreased ATP production (Supplementary Fig. S3C) and higher ROS generation. Conversely, reconstitution of SNPH-depleted cells with adenovirus encoding WT SNPH or SNPH mutant K111R or K153R (Supplementary Fig. S3E) comparably restored tumor cell proliferation (Supplementary Fig. S3F).

Consistent with increased chemotaxis (Fig. 4A and B), expression of ubiquitination-defective SNPH mutant K111R or K153R, alone or as a DM, promoted greater invasion of PC3 cells across Matrigel-coated inserts, compared with control transfectants (Fig. 5A and B). Instead, expression of WT SNPH suppressed tumor cell invasion (Fig. 5A and B). When analyzed in reconstitution experiments, knockdown of SNPH also increased tumor cell invasion, in a reaction suppressed by WT SNPH, but restored by K111R or K153R mutant (Fig. 5C). Comparable results were obtained by targeting the regulators of SNPH ubiquitination. In these experiments, siRNA silencing of CHIP increased tumor cell invasion, whereas USP7 knockdown was profoundly inhibitory (Supplementary Fig. S4A and S5D).

We next asked whether ubiquitination of the SNPH residues outside the MBD (Supplementary Fig. S2B) influenced tumor cell invasion. For these experiments, we transduced PC3 cells with adenovirus constructs encoding Flag-tagged SNPH mutant K175R, K295R, or K415R, plus K111R and K153R used as control (Supplementary Fig. S4B). Consistent with the results of the proteomics screen, Flag-tagged SNPH mutant K175R, K295R, or K415R precipitated from PC3 cells showed decreased ubiquitination, comparably to immunoprecipitated K111R or K153R (Supplementary Fig. S4C). Functionally, expression of ubiquitination-defective SNPH mutants, including K175R, K295R, or K415R increased 2D chemotaxis, resulting in greater speed of cell movements and longer distance traveled by individual cells (Supplementary Fig. S4D). Similar results were obtained in reconstitution experiments, where transfection of each ubiquitination-defective SNPH mutant restored tumor cell invasion in siRNA-silenced cells, similar to SNPH depletion (Supplementary Fig. S4E). Finally, we asked whether SNPH ubiquitination was important for metastasis in vivo. In these experiments, stable transfection of Yumm 1.7 cells with WT SNPH suppressed metastatic seeding to the lung in a syngeneic mouse model of disseminated melanoma, compared with control transfectants (Fig. 5E and F). Conversely, expression of ubiquitination-defective SNPH mutant K111R or K153R significantly increased the frequency of DTC to the lung, compared with control transfectants (Fig. 5E and F).

We next used a novel tracking method based on the MAT algorithm (19, 23) with a minimum-spanning tree approach (see Materials and Methods) to quantify the impact of SNPH ubiquitination on mitochondrial dynamics (Fig. 6A). In these experiments, we observed that transfection of WT SNPH in Yumm 1.7 cells profoundly suppressed cycles of mitochondrial fusion and fission, compared with control transfectants (Fig. 6B and C). Conversely, expression of ubiquitination-defective SNPH mutant K111R or K153R restored mitochondrial dynamics in these settings, comparably to control cultures (Fig. 6B and C). Furthermore, expression of SNPH mutant K111R or K153R resulted in recruitment of the fission regulator, dynamin-related protein-1 (Drp1) to mitochondria, as well as increased Drp1 phosphorylation on the activating residue, S616 (Fig. 6D). Conversely, WT SNPH did not induce mitochondrial recruitment of Drp1 or Drp1 phosphorylation (Fig. 6D). Total Drp1 levels and S616-phosphorylated Drp1 expression in total cell extracts were not affected (Fig. 6D). Similar results were obtained after siRNA silencing of SNPH, which induced recruitment of Drp1 to mitochondria, and its phosphorylation on S616 (Fig. 6E). siRNA silencing of CHIP also caused increased phosphorylation of Drp1 on S616 and reduced phosphorylation of Drp1 on the inhibitory site, S637, whereas USP7 silencing had the opposite effect, decreasing Drp1 phosphorylation on S616 and upregulating the levels of S637-phosphorylated Drp1 (Fig. 6F).

Next, we asked if modulation of mitochondrial dynamics by SNPH ubiquitination was important for organelle trafficking and tumor cell motility. siRNA silencing of Drp1 (Supplementary Fig. S5A) suppressed mitochondrial motility in Yumm 1.7 cells (Supplementary Fig. S5B), restricting the distance traveled by individual mitochondria and the speed of organelle movements (Supplmentary Fig. S5C). This was associated with inhibition of PC3 cell invasion across Matrigel-coated inserts (Fig. 7A and B). Consistent with a role of SNPH in this response, expression of GTPase-defective (GD) Drp1 mutant (Supplementary Fig. S5D) prevented the increase in PC3 cell invasion after SNPH knockdown (Fig. 7C). To test whether this pathway involved SNPH ubiquitination, we next quantified mitochondrial movements in Drp1-silenced Yumm 1.7 cells expressing WT SNPH or SNPH mutant K111R (Supplementary Fig. S5E). Drp1 knockdown suppressed the increased mitochondrial speed (Fig. 7D and E) and distance traveled by individual mitochondria (Supplementary Fig. S5F) in the presence of SNPH mutant K111R. Silencing of Drp1 further reduced residual mitochondrial movements in cells expressing WT SNPH, affecting both mitochondrial speed (Fig. 7E; Supplementary Fig. S5G) and distance traveled (Supplementary Fig. S5F). Similar results were obtained in experiments of 2D chemotaxis, where depletion of Drp1 suppressed the increased speed of migration and distance traveled by individual cells in the presence of SNPH mutant K111R (Fig. 7F and G). In contrast, Drp1 knockdown minimally affected residual cell motility in the presence of WT SNPH (Fig. 7F and G).

In this study, we have shown that SNPH (14) is ubiquitinated in tumor cells on multiple lysine residues by the E3 ligase, CHIP under conditions of microenvironment stress. Instead of proteasomal degradation, CHIP ubiquitination enables a direct association of SNPH with tubulin, which is required to inhibit mitochondrial movements (Supplementary Fig. S6). Interference with SNPH ubiquitination increases mitochondrial trafficking to the cortical cytoskeleton, restores mitochondrial dynamics, and fuels tumor cell movements with heightened chemotaxis, invasion, and metastasis in vivo. Mechanistically, SNPH ubiquitination regulates the recruitment of the fission regulator Drp1 to mitochondria, which is required to support organelle trafficking to the cortical cytoskeleton and tumor chemotaxis (Supplementary Fig. S6).

Although first described in neurons as a mechanism to concentrate an efficient bioenergetics source at sites with high energy demands (15), mitochondrial trafficking may extend beyond the central nervous system and may provide a localized energy source to support membrane dynamics and directional migration in disparate cell types (26, 27). A similar model has been proposed in cancer, where exploitation of “neuronal” regulators of mitochondrial motility, including SNPH (16), repositions mitochondria to the cortical cytoskeleton, fueling tumor cell invasion (11), and metastasis (16). Here, an unbiased proteomics screen uncovered regulators of protein ubiquitination as the most abundant class of SNPH-associated molecules in cancer, and identified CHIP, or STUB1 (28) as a SNPH E3 ubiquitin ligase in vivo. Although the role of posttranslational modifications in mitochondrial trafficking has not been widely studied (29), CHIP has attracted attention for a dual regulation of protein homeostasis in cooperation with Hsp chaperones, especially Hsp70 and Hsp90 (30), and ubiquitination-coupled proteasomal destruction of protein targets (31). This pathway may contribute to aging (32) and various mechanisms of tumor suppression through inhibition of cancer metabolism (33), androgen receptor transcription (34), and NF-κB signaling (35). The data here that CHIP ubiquitination enables the function of SNPH in restricting tumor cell movements is consistent with the proposed tumor suppressor function, and may provide a mechanistic basis for other data that CHIP ubiquitination inhibits metastasis in vivo (36), and further reduces tumor cell motility via proteasomal degradation of integrin-linked kinase (37).

Unexpectedly, we found that CHIP ubiquitination of SNPH does not couple to proteasomal destruction. Experimentally, there was no difference in the half-life of WT or ubiquitination-defective SNPH mutants, and proteasome inhibitors or CHIP silencing had no effect on SNPH stability. There is precedent for nondegradative pathways of CHIP ubiquitination. For instance, CHIP ubiquitination controls the subcellular shuttling of the transcriptional corepressor, Daxx independently of proteasomal degradation (38), and participates in epigenetics control of gene expression via increased stability of the sirtuin family member, SirT6 (39). Biochemically, these nondegradative responses have been ascribed to noncanonical ubiquitin modification(s), independently of Lys40 (40), and data presented here suggest that most SNPH ubiquitination involves Lys63 linkages, with higher molecular weight (poly)ubiquitination moieties implicating Lys48.

Instead of proteasomal degradation, CHIP ubiquitination of SNPH on K111 and K153 in the MBD enabled direct binding to tubulin, a requisite to suppress mitochondrial movements and organelle dynamics (16). This is reminiscent of other nondegradative ubiquitination signals that mediate protein–protein interactions, in particular during immune and inflammatory responses (41). Conversely, other molecules found in the SNPH complex(es) bind tubulin, including members of the kinesin-13 family of microtubule depolymerases, Kif2a and Kif2c (42), and, conversely, a microtubule-stabilizing protein, Clasp1 (43). Whether a similar recruitment of tubulin-binding proteins occurs in nontransformed cells, including neurons (14), is unknown. However, for the role of Kif2a and Clasp1 in microtubule dynamics (44), it is possible that their association with SNPH controls the assembly of a stable microtubule track, potentially required for efficient mitochondrial transfer (45). In addition, we found that other SNPH lysine residues outside the MBD (K175, K295, and K415) are ubiquitinated in vivo and are relevant for tumor cell motility. The role of these additional SNPH ubiquitination sites remains to be fully elucidated, but it could be speculated that their posttranslational modification(s) contributes to stabilize SNPH–microtubule interaction mediated by the MBD.

Mimicking the effect of SNPH depletion (16), interference with SNPH ubiquitination was sufficient to reposition mitochondria to the cortical cytoskeleton of tumor cells through faster speed of organelle movements and longer distance traveled. Consistent with a model of regional mitochondrial bioenergetics to sustain tumor cell movements, loss of SNPH ubiquitination was associated with Akt and mTOR phosphorylation, heightened chemotaxis and invasion, and formation of lung metastasis in vivo. Together, these data point to SNPH as a novel ubiquitination-regulated metastasis suppressor (16), and, accordingly, pharmacologic or genetic targeting of the USP7 deubiquitinase abrogated mitochondrial movements, restricted chemotaxis, and blocked tumor cell invasion. Although this is consistent with an oncogenic function of USP7 and actionable therapeutic target (46), a role of this pathway in tumor cell motility has not been previously demonstrated.

Mechanistically, CHIP ubiquitination of SNPH regulated mitochondrial dynamics, suppressing cycles of organelle fission and fusion in tumor cells (47). In contrast, interference with SNPH ubiquitination restored mitochondrial dynamics and promoted recruitment of S616-phosphorylated Drp1 to mitochondria (47). These data mirror earlier findings, where SNPH depletion resulted in heightened mitochondrial dynamics, with increased frequency of fusion and fission cycles in tumor cells (16). As a regulator of organelle size, shape, and mass (47), there is evidence that mitochondrial dynamics is exploited in cancer (48), and may contribute to tumor cell invasion (10). Whether this predominantly involves fusion or fission has been debated (48). The data here that active Drp1 is required for mitochondrial trafficking and tumor chemotaxis are consistent with other findings that NIK signaling repositions mitochondria to the cortical cytoskeleton in a Drp1-dependent manner (26), and that mitochondrial fission enables faster transfer of mitochondria along microtubules in tumor cells (16). How mitochondrial fission couples to SNPH levels (16) or SNPH ubiquitination (this study) has not been determined. One possibility is that this response involves cytoskeletal remodeling at the mitochondrial interface (49), or mTOR activation (50), processes potentially linked to SNPH function in cancer (16).

In summary, we identified a critical posttranslational modification, that is ubiquitination that enables a potent, metastasis suppressor function of SNPH. The results reinforce the general exploitation of mitochondrial dynamics (48), including organelle trafficking (11–13) for tumor cell motility, and anticipate its potential as an actionable therapeutic target in metastatic disease (7).

No potential conflicts of interest were disclosed.

Conception and design: J.H. Seo, M.C. Caino, D.C. Altieri

Development of methodology: J.H. Seo, M.C. Caino, E.T. Kim, A.R. Cohen, D.C. Altieri

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.C. Caino, H.-Y. Tang

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.H. Seo, E. Agarwal, K.G. Bryant, M.C. Caino, A.V. Kossenkov, H.-Y. Tang, L.R. Languino, A.R. Cohen, D.W. Speicher, D.C. Altieri

Writing, review, and/or revision of the manuscript: J.H. Seo, E. Agarwal, K.G. Bryant, M.C. Caino, A.V. Kossenkov, D.I. Gabrilovich, A.R. Cohen, D.W. Speicher, D.C. Altieri

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.H. Seo, E. Agarwal, K.G. Bryant, E.T. Kim, D.C. Altieri

Study supervision: D.C. Altieri

We thank James Hayden and Frederick Keeney for assistance with time lapse microscopy. This work was supported by the NIH grants P01 CA140043 (to D.C. Altieri, L.R. Languino, and D.W. Speicher), R35 CA220446 (to D.C. Altieri), and R50 CA221838 (to H.-Y. Tang). Support for Core Facilities utilized in this study was provided by Cancer Center Support Grant (CCSG) CA010815 to The Wistar Institute.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.