The kinesin-8 member Kif19 alters microtubule dynamics, suppresses cell adhesion, and promotes cancer cell invasion

Metastasis is one of the deadliest aspects of cancer. Initial Metastatic spread is dependent on the detachment and dissemination of cells from a parent tumor, and invasion into the surrounding tissue. In this study, we characterize the kinesin-8 member Kif19 as a promoter of cancer cell invasion that suppresses cell-cell adherens junctions and cell-matrix focal adhesions. Initial analysis of publicly available cancer patient data sets demonstrated that Kif19 expression correlates with worse overall survival probability in several cancers and that Kif19 expression is increased in metastases of colorectal and breast carcinoma compared to the primary tumor. Depletion of Kif19 from two human cancer cell lines (DMS53 and MDA-MB-231) did not alter viability, but decreased the cells’ ability to invade a Matrigel matrix by half and impaired the invasion of spheroids into a primary cell monolayer. Ectopically expressed Kif19 localized to, and partially depolymerized, microtubules in the cell periphery. However, Kif19 depletion increased microtubule dynamicity and sensitivity to pharmacological depolymerization without altering total microtubule polymer levels. These data indicate that Kif19 can both depolymerize and stabilize microtubules. Given this activity, we then studied Kif19’s effect on focal adhesions and adherens junctions, which are both regulated by microtubule dynamics. Kif19 knockdown increased the proportion of cell surface area covered by Vinculin focal adhesions. Further, Kif19 depletion increased whole cell E-cadherin expression and the accumulation of E-cadherin at cell-cell adherens junctions. Conversely, ectopic overexpression of full-length Kif19 led to proportionally smaller focal adhesions and impaired E-cadherin accumulation at cell-cell junctions. Our current hypothesis is that aberrant Kif19 expression in cancer alters focal adhesion dynamics and suppresses E-cadherin expression, which enhance cell invasiveness. Further, we propose that these changes in cell adhesion are due to modification of peripheral microtubule dynamics by Kif19, potentially through disruption of local rho GTPase activity.


47
Introduction 48 Metastasis is one of the primary mechanisms by which cancer causes mortality. More

237
For plasmid transfection, the Lipofectamine 3000 kit (Thermo Fisher Scientific) was used 238 with a modified protocol. Cells were plated on a 24-well plate at a density of 40k per well 24 239 hours prior to transfection. Each well was transfected using 50ul Opti-MEM (Thermo Fisher at 37°C prior to fixation. Cells were fixed following the protocol described above using solutions 372 at RT.

373
For the tubulin tyrosination assays, DMS53 or MDA-MB-231 cells stained for α-tubulin 374 and tyrosinated-tubulin were imaged and pre-processed as described above. For a given cell 375 ROI, the total fluorescent intensity of each fluorescent channel was measured and corrected for 376 background to generate the corrected total cell fluorescence (CTCF). The α-tubulin CTCF was 377 used to represent the total microtubule polymer levels per cell. The tyrosinated-tubulin CTCF 378 was divided by the α-tubulin CTCF to generate a ratio describing what proportion of the MT 379 array was tyrosinated. This same process was completed for the tubulin detyrosination assays, 380 except cells were stained for α-tubulin and detyrosinated-tubulin.

381
To examine a total microtubule polymer levels following nocodazole treatment or ectopic

382
Kif19 expression, cells were stained for α-tubulin ± Myc (for Kif19 construct overexpression 383 studies). Cells were imaged and preprocessed as described above, and the α-tubulin CTCF was 384 calculated to represent the total microtubule polymer levels per cell. the visibility of vinculin adhesions, we ran "enhance contrast" with saturated pixels at 0.35%, 405 followed by "subtract background" with a rolling ball radius of 50 pixels, then "despeckle," and 406 lastly "8-bit." To create masks from these processed images, the FIJI "auto local threshold" 407 function was run with the Phansalkar method and a radius at 20. To generate a list of ROIs for 408 each focal adhesion in the mask, the FIJI "analyze particles" function was run with a size range 409 of 0.04-20um 2 (for DMS53) or 0.12-40um 2 (for MDA-MB-231) and circularity range of 0.00-1.00.

410
Subsequently, this list of ROIs was applied over the original Vinculin z-projection and the ROIs 411 were manually edited to ensure they accurately represented the foal adhesions on the image.

412
The size measurements for each individual focal adhesion ROI was saved. To calculate focal

420
Differences between treatments were analyzed using a student's t-test to compare two groups.

421
Outliers were removed from independent experiment data sets using the ROUT  expression correlates with decreased overall survival probability in several cancers (Fig S1), so  (Fig S2A,B). Following Kif19 knockdown, we 440 observed no change in the viability of MDA-MB-231 or DMS53 cells (Fig 1A). indicated that, in colorectal and breast carcinoma, Kif19 expression was significantly higher in 450 samples from metastases than from primary tumors (Fig S3A,B). in a more physiologically relevant system, we seeded siRNA-treated DMS53 spheroids onto a 455 monolayer of primary human umbilical vein endothelial cells (HUVECs) (Fig 1D). Spheroids Kif19 siRNA, respectively, p=0.017) (Fig 1E). From these data, we propose that the pro-  (Fig 2A).

491
We then examined localization of Kif19 FL along MTs using IF imaging. We observed a gradient 492 of Kif19 concentration along MTs, with accumulation at MT plus-ends in the cell periphery (Fig   493  2B). This is characteristic of kinesin-8 members and is often reliant on the kinesin's CTT as well 494 as the motor domain (52-54). As expected, neither a Kif19 construct lacking the CTT (Kif19 1-575 ) 495 nor a construct lacking the kinesin motor domain (Kif19 331-998 ) concentrated along microtubules, suggesting that Kif19 behaves similar to other kinesin-8s (Fig 3). This accumulation at MT ends 497 is often due to the presence of a secondary MT-binding site within the kinesin-8 CTT, which 498 results in highly processive movement of kinesin-8s along microtubules (50, 53, 81). We 499 observed that Kif19 331-998 chiefly localized to the nucleus, but was present on microtubules when 500 highly overexpressed, suggesting that a secondary binding site may be present (Fig 3C).

520
Kif19's motor domain was previously reported to rapidly depolymerize purified GMPCPP-stabilized MTs, and we hypothesized that overexpression of Kif19 constructs in cells would 522 similarly depolymerize MTs (64, 65). However, DMS53 and U2OS cells expressing Kif19 FL or 523 Kif19 1-575 often contained a complete MT array and only rarely exhibited apparent MT 524 depolymerization (Fig 2 and Fig 4A,B). We then quantified how Kif19 plasmids (described in   (Fig 5A). However, this result is not outside the realm of possibility for Kif19. In addition to  (Fig 5B,C). Similarly, Kif19 knockdown in MDA-MB-231 cells resulted in a 9.3% 559 increase in tyrosinated-tubulin (p<0.0001) and a corresponding 6.8% decrease in detyrosinated-560 tubulin (p=0.0021) (Fig 5B,C). This implies that Kif19 is targeting dynamic (i.e. tyrosinated) MTs, 561 either by selective depolymerization or stabilization. nocodazole treatments (7.2% change, p=0.12) (Fig 5D) 19.5% increase in FA density (total FA area / total cell area) (p=0.013) (Fig 6B). This appears to 603 be primarily mediated by an increase in the average FA size per cell (12.3% increase, 604 p=0.0015), with only a slight, non-significant increase in the number of adhesions per cell (4.5% 605 increase, p=0.40) (Fig 6E). Similarly, Kif19 depletion in MDA-MB-231 cells led to a 13.5% 606 increase in FA density (p=0.0012), though individual FA size and number could not be 607 accurately calculated due to extensive overlapping between adjacent adhesions (Fig 6C and 608 Fig S4A). Overexpression of myc-tagged Kif19 FL in DMS53 cells led to a 16.3% decrease in FA 609 density (p=0.0060), supporting our knockdown data (Fig 6D). However, this change in density 610 was meditated by decreases in both FA number (6.3% decrease, p=0.44) and average FA size 611 (5.9% decrease, p=0.11), with neither variable significant on its own (Fig 6F).  corrected for "effective" area rather than whole cell area (see Fig S4). Data  and are excluded from sites of cell-cell contacts (Fig 6A). We qualitatively observed that Kif19 640 knockdown in DMS53 cells increased cell-cell contact area and thus decreased the area over 641 which FAs were observed (Fig S4B). To correct for this, we calculated an "effective" cell area 642 consisting of a 2um band along the cell perimeter, excluding areas of cell-cell contact (Fig S4C).

643
The increase in FA density following Kif19 knockdown was over 2-fold greater within this 644 effective area compared to the whole cell area (49.9% increase in effective FA density 645 compared to control, p<0.0001) (Fig 6G). A similar effect was observed with overexpression of 646 Kif19 FL plasmids in DMS53 (26.7% decrease in effective FA density compared to control, 647 p<0.0001) (Fig 6I). However, there was no difference after correcting for effective area in MDA-

648
MB-231 cells after Kif19 knockdown (8.9% increase in effective FA density compared to control, 649 p=0.083) (Fig 6H). Kif19 regulates E-cadherin concentration at cell-cell contacts 672 We observed a qualitative increase in cell-cell contacts within areas of moderate 673 confluency in Kif19-depleted DMS53 cells relative to controls (Fig S4B).