Parkin ubiquitination of Kindlin-2 enables mitochondria-associated metastasis suppression

Mitochondria are signaling organelles implicated in cancer, but the mechanisms are elusive. Here, we show that Parkin, an E3 ubiquitination (Ub) ligase altered in Parkinson’s disease, forms a complex with the regulator of cell motility, Kindlin-2 (K2), at mitochondria of tumor cells. In turn, Parkin ubiquitinates Lys581 and Lys582 using Lys48 linkages, resulting in proteasomal degradation of K2 and shortened half-life from ∼5 h to ∼1.5 h. Loss of K2 inhibits focal adhesion turnover and β1 integrin activation, impairs membrane lamellipodia size and frequency, and inhibits mitochondrial dynamics, altogether suppressing tumor cell–extracellular matrix interactions, migration, and invasion. Conversely, Parkin does not affect tumor cell proliferation, cell cycle transitions, or apoptosis. Expression of a Parkin Ub-resistant K2 Lys581Ala/Lys582Ala double mutant is sufficient to restore membrane lamellipodia dynamics, correct mitochondrial fusion/fission, and preserve single-cell migration and invasion. In a 3D model of mammary gland developmental morphogenesis, impaired K2 Ub drives multiple oncogenic traits of EMT, increased cell proliferation, reduced apoptosis, and disrupted basal–apical polarity. Therefore, deregulated K2 is a potent oncogene, and its Ub by Parkin enables mitochondria-associated metastasis suppression.

Mitochondria are signaling organelles implicated in cancer, but the mechanisms are elusive. Here, we show that Parkin, an E3 ubiquitination (Ub) ligase altered in Parkinson's disease, forms a complex with the regulator of cell motility, Kindlin-2 (K2), at mitochondria of tumor cells. In turn, Parkin ubiquitinates Lys581 and Lys582 using Lys48 linkages, resulting in proteasomal degradation of K2 and shortened half-life from 5 h to 1.5 h. Loss of K2 inhibits focal adhesion turnover and β1 integrin activation, impairs membrane lamellipodia size and frequency, and inhibits mitochondrial dynamics, altogether suppressing tumor cell-extracellular matrix interactions, migration, and invasion. Conversely, Parkin does not affect tumor cell proliferation, cell cycle transitions, or apoptosis. Expression of a Parkin Ub-resistant K2 Lys581Ala/Lys582Ala double mutant is sufficient to restore membrane lamellipodia dynamics, correct mitochondrial fusion/fission, and preserve single-cell migration and invasion. In a 3D model of mammary gland developmental morphogenesis, impaired K2 Ub drives multiple oncogenic traits of EMT, increased cell proliferation, reduced apoptosis, and disrupted basal-apical polarity. Therefore, deregulated K2 is a potent oncogene, and its Ub by Parkin enables mitochondria-associated metastasis suppression.
Mitochondria are invariably exploited in cancer (1,2). This involves adaptive changes in bioenergetics important for tumor growth (3,4) but also a plethora of other functions in redox balance (5); control of cell death (6); modulation of mitochondrial size, shape, and subcellular distribution, i.e., mitochondrial dynamics (7,8); and interorganelle cross talk (9) that enables key disease traits, including heightened cell motility and invasion (10).
Aside from the roles in tumor promotion, there is now emerging evidence that mitochondria may also participate in the opposite function of tumor suppression, specifically targeting cancer cell movements (11). Accordingly, ablation of disparate regulators of mitochondrial dynamics (12), bioenergetics (13) or mitophagy (14) enhances tumor cell invasion and metastasis (7). Molecules that selectively inhibit tumor cell movements without affecting other cancer traits have been described (15), but whether such metastasis suppressors involve mitochondria is unknown, and mechanistic link(s) between mitochondria and the signaling machinery of cell motility (16) have not been clearly delineated.
A potential regulator of mitochondria-associated tumor suppression is Parkin, an E3 ubiquitination (Ub) ligase and effector of mitophagy (17) altered in early-onset Parkinson's disease (18) and lost in virtually all human cancers (19). How Parkin antagonizes tumorigenesis, and whether this involves mitophagy (20), has remained unclear and variously linked to metabolic rewiring (21), regulation of mitosis (22), changes in iron homeostasis (23) and/or inhibition of serine metabolism (24). However, a recent Ub screen carried out in the absence of harsh and potentially nonphysiologic mitochondriadepolarizing conditions (25) suggested a different scenario and identified Parkin as a potent inhibitor of mitochondriafueled tumor cell movements (26). This pathway required Parkin E3 ligase activity but was independent of mitophagy (26).
Here, we identified K2 as a target of Parkin Ub, enabling a novel pathway of mitochondria-associated metastasis suppression.

K2 is a novel substrate of Parkin Ub
To identify physiologic targets of Parkin in cancer, a Ub screen was carried out in the absence of mitochondriadepolarizing conditions (25) using stable isotope labeling by amino acids in cell culture (SILAC) (26). A top hit in the screen was K2, which exhibited a two-fold increase in SILAC labeling in the presence of Parkin, compared with vector (Fig. 1A). The tandem mass spectroscopy spectrum of the triply charged m/z 646.3459 ion identified the K2 peptide FQGGKKEELIGIAYNR, with both K581 and K582 as candidate Ub sites for Parkin (Fig. 1A). In a structural model of the -COOH terminus FERM (F3) domain of K2, both K581 and K582 are surface-exposed (Fig. 1B), upstream of Q(614)W and other functionally relevant sites in K2 (reviewed in the study by  (27)). By sequence alignment, K582 was conserved in all 3 K-family proteins, whereas K581 was conserved in K1 and K2 and substituted to R in K3 (Fig. 1B, inset). On the other hand, only K2 was identified as a potential Parkin Ub substrate in our SILAC screen (26). Next, we asked if Parkin interacted with K2 in cancer. Immune complexes of Myc-Parkin expressed in PC3 cells contained GFP-K2 (Fig. S1A). Reciprocally, K2 immunoprecipitates were associated with Parkin (Fig. S1B). IgG immune complexes were unreactive (Fig. S1, A and B). Consistent with these data, Parkin and K2 exhibited extensive colocalization at mitochondria, by confocal fluorescence microscopy (Figs. 1C and S1C). To test whether the two candidate Ub sites were important for this interaction, we next generated a K2 K581A/ K582A double mutant (DM) that abolishes Parkin-predicted Ub sites (Fig. S1D). In these experiments, both K2-WT and K2-DM were comparably associated with Parkin at mitochondria, by confocal microscopy ( Fig. 1C; Fig. S1, C and E).

Parkin Ub regulates K2 stability
K2, but not vector, precipitated from Parkin-expressing PC3 cells reacted with an antibody to total Ub (Fig. 1D), as well as K48 Ub, whereas K63 Ub was unreactive (Fig. 1E). To further characterize this pathway, we next generated clones of PC3 cells with conditional expression of Parkin (TET-ON system) induced by treatment with doxycycline (Doxy) (Fig. S1F). Immune complexes of K2-DM precipitated from Doxyinduced Parkin-expressing cells showed reduced reactivity for total Ub and K48 Ub, compared with K2-WT ( As an independent approach, we expressed Parkin E3 ligase-defective mutant S65A or C431S in PC3 cells (Fig. 1G). These mutants did not reduce the levels of endogenous K2 or mitofusin-2, a known target of Parkin Ub, whereas both molecules were lost in the presence of wildtype (WT) Parkin (Fig. 1G) Fig. S2E).

Regulation of plasma membrane dynamics by Parkin-K2 Ub
K2 regulates plasma membrane dynamics of cell-ECM interactions (29) and cell motility (28) and the role of Parkin Ub in this process was next investigated. Consistent with previous observations (26), Parkin expression inhibited focal adhesion (FA) turnover, which is essential for cell motility (16), reducing the fraction of new and decayed FA while expanding the population of stable FA (Fig. 2, A and B). This was associated with impaired membrane lamellipodia dynamics (Fig. 2, C and D), another requirement of cell movements (31). In these experiments, Parkin expression decreased the lamella size and speed of lamella retraction while prolonging lamella persistence at the plasma membrane (Figs. 2E and S3A). The frequency of lamella formation was also inhibited in these settings, by single-cell Stroboscopic Analysis of Cell Dynamics (SACED) quantification (Figs. 2F and S3B). A third upstream requirement of tumor cell movements is heightened mitochondrial dynamics (32). Reconstitution of tumor cells with Parkin inhibited this process (26), reducing the rate of both organelle fusion and fission events (Fig. S4, A and B). Based on these collective results, we asked if K2 Ub contributed to Parkin suppression of plasma membrane and mitochondrial dynamics in the control of cell movements. Accordingly, reconstitution of Parkin-expressing PC3 cells with K2-DM, but not K2-WT, restored the kinetics of FA turnover (Fig. 2, A and B), maintained membrane lamellipodia dynamics (Fig. 2, C-F; Fig. S3, A and B), and rescued mitochondrial fusion and fission events (Fig. S4, A and  B), when compared to controls. Reconstitution with Ubresistant single K2 K581A or K582A mutant also rescued mitochondrial dynamics in the presence of Parkin (Fig. S4B).

Parkin-K2 regulation of tumor cell-ECM interaction and cell motility
Consistent with inhibition of membrane dynamics, Parkin expression inhibited tumor cell adhesion (Fig. S5A) and spreading (Fig. S5, B and C) onto the ECM substrate, fibronectin. This was accompanied by impaired activation of β1 integrin (Fig. S5D), a process regulated by K2 (29), whereas total β1 integrin levels were unchanged (Fig. S5E). In addition, Parkin expression inhibited single-cell motility (26), reducing the speed of tumor cell movements and the total distance traveled by individual cells (Fig. 3, A and B). Tumor cell invasion across Matrigel-coated inserts (Fig. 3, C and D), as well as 3D invasion of tumor spheroids embedded in collagen (Fig. 3, E and F), was also inhibited in these settings. Conversely, reconstitution of Parkin-expressing cells with K2-DM or K2 mutants K581A or K582A restored tumor cell adhesion (Fig. S5A) and spreading (Fig. S5, B and C), rescued single-cell movements (Fig. 3, A and B), and enabled tumor cell invasion across Matrigel inserts (Fig. 3, C and D), as well as in 3D tumor spheroids (Fig. 3, E and F). Reconstitution with K2-WT had limited to no effect (Fig. S5, A-C; Fig. 4, E and F).
As an independent approach, expression of E3 ligasedefective Parkin S65A or C431S mutant (Fig. S6A) failed to reduce tumor cell adhesion to fibronectin or fibrinogen (Fig. S6B) and did not affect single-cell motility (26), compared with WT Parkin. Similarly, treatment with the proteasome inhibitor, MG132 prevented Parkin degradation of K2 in cycloheximide block-and-release experiments (Fig. S6C) and preserved the speed of tumor cell movements and total distance traveled by individual cells, when compared to controls (Fig. S6, D and E).
Finally, we asked if Parkin inhibition of tumor-ECM interactions and cell motility was specific. In these experiments, Parkin expression in PC3 cells had no effect on cell proliferation (Fig. S7A), DNA content (Fig. S7B), or distribution of cell cycle phases (Fig. S7C). Similarly, annexin V labeling, which quantifies apoptosis, was unchanged in control or Parkinexpressing cells (Fig. S7D).

Oncogenic properties of deregulated K2
To test how a Parkin-K2 Ub axis may affect mechanisms of tumorigenesis, we next expressed K2-DM in normal, nontransformed breast epithelial MCF10A cells, which contain endogenous Parkin (26). In these experiments, non-Parkin ubiquitinatable K2 increased focal adhesion kinase phosphorylation (Tyr397); upregulated EMT markers, vimentin, and SNAIL (Fig. S8A); and heightened cell migration, compared with control (Fig. S8B). Expression of K2-WT had a more limited effect (Fig. S8, A and B). In a 3D model of mammary gland developmental morphogenesis, expression of K2-DM, but not K2-WT, increased the surface area of 3D acini (Fig. 4,  A and B) and reduced acini circularity (Fig. 4C), two markers of deregulated epithelial morphogenesis. This was accompanied by increased cell proliferation, reduced apoptosis, and upregulation of EMT vimentin (Fig. 4, D and E). In addition, K2-DM disrupted apical-basal polarity in 3D acini, resulting in deregulated expression of apical (GM130) and basal (hDlg) polarity markers and aberrant filling of the acini lumen, by phalloidin staining (Fig. S8C).
In summary, we have identified Parkin Ub-dependent degradation of K2 (29) as a novel mechanism of mitochondriaassociated metastasis suppression, shutting off multiple upstream requirements of cell-ECM interaction and cell motility   16). This is consistent with key roles of K2 in FA turnover (28), integrin activation (33) and signaling requirements of cell movements (28), exploited for tumor progression (34). Mechanistically, there is precedent for K2 regulation by Ub (35) and participation in mitochondrial functions (36), in line with the formation of a Parkin-K2 complex at mitochondria described here. On the other hand, a role of K581 and K582 or K48 Ub linkages in K2 stability has not been proposed (35). The Parkin-K2 axis uncovered here selectively inhibited tumor cell motility and invasion, without affecting other tumor traits of cell proliferation, cell cycle, or apoptosis. This selectivity is a hallmark of metastasis suppressors (15), identifying Parkin as a novel, mitochondria-associated antagonist of metastasis. In this context, the nearly universal loss of Parkin in cancer (19) is expected to impair K2 Ub and proteasomal degradation. This enables potent oncogenic properties of unbridled K2 in normal epithelia, disrupting tissue morphogenesis and heightening tumor cell motility and invasion to promote metastasis. . ***p < 0.0001. E, PC3 cells with Doxy-induced Parkin expression were reconstituted with K2-WT or K2-DM, seeded in 3D spheroids in a collagen-containing matrix, and cell invasion was visualized by light microscopy (top) and INSIDIA software image reconstruction (bottom) for quantification of core area (green) and invasive area (red). The scale bar represents 200 μm. F, the conditions are as in (E), and the invasion area of 3D spheroids was quantified (min to max; n = 7-9). For all panels, numbers correspond to p values by one-way ANOVA with Tukey's multiple comparisons test. DAPI, 4 0 ,6-diamidino-2-phenylindole; Doxy, doxycycline.

Plasma membrane dynamics
FA turnover was quantified by time-lapse videomicroscopy (12). Plasma membrane lamellipodia dynamics was imaged using a 40× objective on a Nikon TE300 inverted time-lapse microscope, and lamella size, persistence, and frequency were quantified by SACED (38).

Fluorescence microscopy
Mitochondrial dynamics was quantified using CellLight Mitochondria-RFP, BacMam 2.0-transduced cells using a Leica TCS SP8 X inverted laser scanning confocal microscope with a 63X 1.40NA oil objective (26). For each cell, the volume of mitochondria over time was analyzed in four different areas (10 mitochondria per area) in 3D images. Variations in mitochondrial volume were evaluated by fold-change over time: a fold-change >1.3 denoted a fusion event, and a foldchange <0.7 denoted a fission event. The averaged fission and fusion events in four different areas were used for each cell analyzed. Parkin-K2 colocalization at mitochondria was examined with an antibody to TOM20 by confocal fluorescence microscopy.

Single-cell motility
Experiments were carried out using 2D chemotaxis chambers (Ibidi) and a gradient setup with NIH3T3-conditioned medium (26). Time-lapse videomicroscopy was performed over 10 h with 10-min intervals. At least 30 cells were tracked using the Manual Tracking plugin for ImageJ, and the tracking data from independent time-lapse experiments were pooled and exported into the Chemotaxis and Migration Tool v2.0 (Ibidi) to calculate the speed of cell movements and the total distance traveled by individual cells.

Cell adhesion, spreading, and invasion
Cells (4 × 10 4 , 50% confluency) were starved for 1 h in a reduced serum medium, plated onto a fibronectin-coated (10 μg/ml) 96-well plate, fixed in 4% paraformaldehyde, and stained with crystal violet. Cell adhesion was quantified by absorbance at 570 nm. For analysis of cell spreading, cells seeded onto an Ibidi 8-well μ-Slide were imaged with a Nikon TE inverted microscope after 24 h, and the cell surface was quantified by ImageJ. Tumor cell invasion across Matrigelcoated 8-μm PET Transwell chambers (Corning) was quantified in the presence of an NIH3T3-conditioned medium as a chemoattractant (26). The 3D organotypic tumor spheroids embedded in a collagen matrix were quantified for maximum invasion distance and invasion area (38).

Statistical analysis
Data are expressed as mean ± SD or mean ± SEM of results from a minimum of three independent experiments. Unpaired, two-tailed Student t test was used for two-group analyses. For multiple-group comparisons, analysis of variance was used. Statistical analyses were performed using a GraphPad software package (Prism 9.0) for Windows. A p value of <0.05 was considered statistically significant.

Data availability
All data are contained within the article. Requests for reagents should be addressed to DC Altieri (daltieri@wistar.org).
Supporting information-This article contains supporting information.