The universal mechanism of intermediate filament transport

Intermediate filaments (IFs) are a major component of the cytoskeleton that regulates a wide range of physiological properties in eukaryotic cells. In motile cells, the IF network has to adapt to constant changes of cell shape and tension. In this study, we used two cell lines that express vimentin and keratins 8/18 to study the dynamic behavior of these IFs. We demonstrated that both IF types undergo extensive transport along microtubules. This was an unexpected result as keratin filament remodeling has been described to depend on actin dynamics. We established the role of kinesin-1 in vimentin and keratin IF transport by knocking out KIF5B, the ubiquitous isoform of kinesin-1. Futhermore, we demonstrated that unlike typical membrane cargoes, transport of both types of IFs does not involve kinesin light chains, but requires the presence of the same region of the kinesin-1 tail, suggesting a unified mechanism of IF transport.


ABSTRACT 16
Intermediate filaments (IFs) are a major component of the cytoskeleton that regulates a wide 17 range of physiological properties in eukaryotic cells. In motile cells, the IF network has to adapt 18 to constant changes of cell shape and tension. In this study, we used two cell lines that express 19 vimentin and keratins 8/18 to study the dynamic behavior of these IFs. We demonstrated that 20 both IF types undergo extensive transport along microtubules. This was an unexpected result as 21 keratin filament remodeling has been described to depend on actin dynamics. We established the 22 role of kinesin-1 in vimentin and keratin IF transport by knocking out KIF5B, the ubiquitous 23 isoform of kinesin-1. Futhermore, we demonstrated that unlike typical membrane cargoes, 24 transport of both types of IFs does not involve kinesin light chains, but requires the presence of 25 the same region of the kinesin-1 tail, suggesting a unified mechanism of IF transport. the major microtubule motor kinesin-1 has been typically suggested to be involved in IF 51 transport. Kinesin-1 in mammals is represented by three isoforms, KIF5A, KIF5B or KIF5C, with KIF5B being the most abundant ubiquitous version. Several reports suggested that various 53 types of cytoplasmic IFs might be potential cargo for kinesin-1. In axons for example, 54 neurofilaments are transported by KIF5A (14,15), while in muscle, KIF5B has been reported to 55 be essential for the delivery of desmin and nestin IFs to the growing tip of myotubes (16). 56 Recently, knock down of KIF5B has been shown to reduce anterograde transport of IFs in 57 migrating astrocytes (13). where another cycle of particle formation takes place (17-20). Although microtubule-dependent 66 motion of keratin particles has been observed previously (21, 22), the contribution of 67 microtubules and/or microtubule-based motors for keratin filament dynamics has been neglected 68 as the rapid transport of fully polymerized keratin filaments has never been reported. 69 In this work, we used a combination of photoconversion experiments and CRISPR/Cas9 genome 70 editing of KIF5B to compare the dynamics of keratin and vimentin IFs and the role of 71 microtubules and microtubule motors. Surprisingly, we found that the dynamic properties of both 72 classes of IFs include transport of long mature filaments along microtubules by kinesin-1 and the 73 same domain of the kinesin tail is involved in transport, strongly suggesting that all types of IFs 74 move along microtubules using an identical mechanism.

Vimentin intermediate filaments are transported along microtubules by kinesin-1. 77
Kinesin antibody injection and shRNA knock down experiments have suggested a role for 78 kinesin-1 in vimentin IF transport (9, 13). In humans, kinesin-1 heavy chain is encoded by three 79 genes; KIF5A, KIF5B and KIF5C. We used CRISPR/cas9 genome-editing to KO KIF5B, the 80 major gene coding for kinesin-1 in RPE cells. Several clones were amplified and the KO was 81 verified using western blot analysis with an antibody (CT) directed against a peptide in the tail 82 domain of kinesin heavy chain common to all three isoforms of kinesin-1. Two of the clones 83 were selected for further analysis. The specificity of the KO was further confirmed using a blot 84 with an antibody (HD) that recognizes the motor domains of multiple kinesins. This blot 85 demonstrated that only the band corresponding to kinesin-1 was absent from the lysates of the 86 KO cells ( Figure 1B). We checked the functional implications of KIF5B KO by analyzing the 87 distribution and motility of known kinesin cargos. As expected, KIF5B KO induced the 88 retraction of mitochondria from the cell periphery as described previously ((23), Figure S2 AB). 89 In contrast, the motility of lysosomes was not affected ( Figure S2C), as lysosome transport is 90 driven, not only by kinesin-1, but by multiples kinesins (Reviewed in (24)). 91 We performed vimentin immunostaining to determine how the absence of kinesin-1 impacts 92 vimentin filament distribution. In control cells, the vimentin filament network extended all the 93 way to the cell periphery as delineated by actin staining. In contrast, in the absence of KIF5B the 94 majority of the mature vimentin filaments retracted from the leading edge, with only a few short 95 IF and non-filamentous particles left behind ( Figure 1A). We quantified the results using the 96 procedure described in Figure S1, and confirmed the initial visual observation that the KO of 97 KIF5B correlated with the retraction of the vimentin network towards the nucleus (Figure, 1C).

6
To confirm that this phenotype was not due to an off-target effect, we performed a rescue 99 experiment using a mouse version of KIF5B (mKif5b), which is insensitive to the gRNA used to 100 KO human KIF5B. When mKif5b-Emerald was expressed in RPE KIF5B KO cells, the vimentin 101 filament network distribution was fully restored, demonstrating that the retraction of the network 102 was indeed caused by the absence of kinesin-1 ( Figure 1A, third column). This result 103 corroborates other observations (9, 11, 13) suggesting that vimentin is a kinesin-1 cargo. 104 We have previously visualized vimentin filaments transported along microtubules in RPE cells 123 (8). To directly demonstrate that this transport is powered by kinesin-1, we used photoconversion 124 of mEos3.2-vimentin (8). The emission of mEos3.2 changes from green to red when exposed to 125 ultraviolet (UV) light at 400 nm. By restricting photoconversion to a circular area of about 10 126 μm in diameter, we produced fiduciary marks on filaments permitting us to monitor their 127 transport in regions of cells with high filament density. Photoconverted mEos3.2-vimentin was 128 filaments robustly moved away from the central photoconverted area ( Figure 1D, left panel and 130 Supplemental Video S1). When cells were treated with 10µM nocodazole for 3 hrs to 131 depolymerize microtubules, vimentin filament transport was completely inhibited ( Figure 1D, 132 middle panels). To determine if the microtubule-dependent transport is driven by kinesin-1, The role of the actin cytoskeleton in the cycle of keratin assembly/disassembly has been 141 described in great details (Reviewed in (25)). However, even though microtubule-dependent 142 motion of keratin particles has been observed previously (21, 22), the role of microtubules in the 143 transport of keratin filaments has never been reported. In RPE cells, keratin filaments co-exist 144 with vimentin filaments. This allows us to use our RPE KO cells for analysis of keratin transport. 145 Immunostaining of the keratin network using a pan-keratin antibody shows an intricate network 146 of keratin filaments that extend to the cell edge ( Figure 2A). Interestingly, the absence of 147 kinesin-1 causes the retraction of the keratin network from the cell periphery (Figure 2A   cells is not well established. Therefore, we decided to use an alternative approach, replacing the 222 wild-type KIF5B in RPE cells with a truncated version of the motor lacking the region that 223 recruits KLC to the kinesin-1 complex. To accomplish this, we deleted the heptad repeats 224 (residues 775-802) of mKif5b responsible for KLC binding, creating (mKif5b Δ775-802 -Emerald) 225 ( Figure 4C). Pull-down experiments and western blot analyses were performed to confirm that 226 mKIF5B Δ775-802 -Emerald did not interact with KLC. As described previously, KLC binding to 227 kinesin-1 is required for their stability (31). As a consequence, neither KLC1 nor KLC2 were detectable by western blot analysis in crude extract of KIF5B KO cells ( Figure 4D). We found 229 that the rescue of KIF5B KO by expression of the full-length mKif5b-Emerald prevented 230 degradation of KLC. In contrast, KLC 1 and 2 remained undetectable in lysates from cells 231 expressing mKif5b Δ775-802 -Emerald ( Figure 4D). In addition to probing crude extracts, the kinesin-1 232 complex was enriched by pull down using GFP-binder and the pellets were probed for the 233 presence of KLC1 and KLC2. These experiments showed that the full-length mKif5b-Emerald 234 bound KLC1 and KLC2 while no KLC could be found even after enrichment of mKif5b Δ775-802 -235 Emerald ( Figure 4E). 236 Immunostaining of vimentin and keratin IFs was employed to compare the efficiency of full-237 length mKif5b and mKif5b Δ775-802 in rescuing IF distribution. The images showed that removal of 238 KLC and the region of kinesin tail that interacts with KLC had no effect on the capacity of 239 mKif5b to rescue keratin or vimentin distribution ( Figure 4F). This observation was reflected in 240 the quantification of vimentin and keratin fluorescence intensity at the cell edge, confirming that 241 both mKif5b and mKif5b Δ775-802 constructs rescued IF distribution to the same extent ( Figure 4G). 242 These results demonstrate that KLCs are not involved in the kinesin-dependent transport of IFs. residues 803-892 of the KIF5B tail ( Figure 5A). It is worth noting, that the constructs that 278 rescued vimentin distribution were also able to rescue the distribution of keratin and vice versa. 279 polymerized filaments (8, 12, 34, 35). In this paper, we established that kinesin-1 is essential to 295 vimentin IFs transport along microtubules. In KIF5B KO cells, vimentin filaments are depleted 296 from the cell edge ( Figure 1A) and active transport of vimentin filaments from the cell center to 297 the cell periphery is no longer observed ( Figure 1C). Furthermore, we showed that even fully 298 polymerized keratin IFs undergo constant transport along microtubules powered by kinesin-1 299 (Figure 2-3). Finally, we established, that KLC was not involved in vimentin or keratin IF 300 transport by kinesin (Figure 4) and that the same region of the kinesin tail is required for both 301 keratin and vimentin transport ( Figure 5). 302

Contribution of microtubules to keratin network dynamics 304
The dynamics of the keratin filament network has been extensively studied (25). It has been 305 clearly demonstrated that actin dynamics are responsible for the retrograde transport of keratin 306 filament precursors formed at the focal adhesion and long filaments to the perinuclear region (18, 307 20, 22, 36). Experiments using fluorescence recovery after photobleaching (FRAP) have been 308 very powerful to decipher the role of subunit exchange during the cycle of assembly and 309 disassembly of the entire keratin network (19, 37). This model was recently recapitulated in vivo 310 in an elegant study using YFP-tagged keratin in murine embryos (38). 311

312
In our study, we used photoconversion as an alternative approach to follow a small subset of 313 individual filaments at the cell center, the site where the filament network is especially dense. By 314 using this technique, we were able to observe for the first time the anterograde transport of fully 315 polymerized keratin filaments. 316 317 Our data complement very well the published keratin dynamics, demonstrating the contribution 318 of microtubule-dependent transport from the cell center to the periphery. We believe that the 319 contribution of microtubule-dependent transport to keratin IF dynamics is dependent on the 320 physiological context. We looked at keratin dynamics in two different cell lines; RPE cells, that 321 are highly motile, and A549 carcinomas cells, which are stationary and tend to form cell-cell 322 contacts like typical epithelial cells. We observed keratin filament transport in both cell types, 323 but it was more most robust in RPE cells, raising the possibility that microtubule-dependent 324 transport of keratin filaments could be upregulated as the cells change shape. We speculate that 325 anterograde transport of keratin filaments in epithelial cells delivers filaments to newly formed 326 areas of cytoplasm in migrating epithelial cells. 327 328

Kinesin-1 as a universal transporter of intermediate filaments 329
In this paper, we established using KIF5B KO that kinesin-1 is the anterograde motor responsible 330 for the transport of vimentin and keratin IFs. A role for kinesin-1 has been suggested previously 331 using less specific or less efficient approaches, such as antibody injection that inhibits multiple 332 kinesins (9), or shRNA that causes incomplete knock down (13). In both cases, we could not supplemented with 1% BSA, 0.1% Triton-X100) as described previously (12). Keratin network spreading analyses were performed on confocal images of vimentin and keratin 500 co-staining, acquired with 100x objective lens as described above using the FIJI software version 501 2.0. The vimentin signal was used to manually trace the area of single cell using the polygon 502 selection tool of FIJI. High contrasted image of the non-specific staining was used to determine 503 to outline of the vimentin KO cells. The corresponding channel for the keratin signal was auto-504 threshold using the Li method to determine the area (in pixel) covered by keratin signal. The 505 percentage of keratin spreading was determined by dividing the keratin area by the total area of 506 the cell. Data are representative of at least two independent experiments and are shown as the 507 mean with SD (n>30). Statistical significance was determined using the non-parametric Mann-508 Whitney test with a confidence interval of 95%. This test analysis compares the distributions of 509 two unpaired groups. 510 511

Live cell imaging and photoconversion 512
For all live-cell experiments, cells were plated on glass coverslips ~16hr before imaging. Cells 513 were maintained at 37˚C in 5% CO2 during imaging using a Tokai-Hit stage-top incubator 514 (Tokai-Hit, Fujinomiya City, Japan) and Okolab gas mixer (Okolab, Naples, Italy). Photoconversion mEos3.2-keratin 8/18 from green to red was performed using illumination from 523 a Heliophor LED light source in the epifluorescence pathway filtered with a 400-nm filter and 524 confined by a diaphragm. Photoconversion time was 5 s and the zone was 10 μm in diameter, 525 which was positioned at the cell center. Time-lapse sequences were acquired at 15s intervals for 526 3 min using 561nm laser. Images were analyzed in Fiji, and assembled in Illustrator. 527 528 Imaging of mEos3.2-vimentin was performed using TIRF on a Nikon Eclipse U2000 inverted 529 microscope equipped with a Plan-Apo TIRF 100x 1.49 NA objective and a Hamamatsu CMOS 530 Elements AR 4.51.01 software (Nikon, Melville, NY, USA). The angle of a 561 nm laser was 532 manually adjusted until near total internal reflection was reached as judged by imaging of 533 photoconverted mEos3.2-vimentin expressing cells. For photoconvertion, cells were exposed for 534 3 sec to UV light from a mercury arc in the epifluorescent light path filtered though a 400nm 535 excitation filtered and spatially restricted by a pinhole in the field diaphragm position. Time-536 lapse sequences were acquired at 15 sec intervals for 3 min using the 561nm laser.