A ‘molecular guillotine’ reveals the interphase function of Kinesin-5

ABSTRACT Motor proteins are important for transport and force generation in a variety of cellular processes and in morphogenesis. Here, we describe a general strategy for conditional motor mutants by inserting a protease cleavage site into the ‘neck’ between the head domain and the stalk of the motor protein, making the protein susceptible to proteolytic cleavage at the neck by the corresponding protease. To demonstrate the feasibility of this approach, we inserted the cleavage site of the tobacco etch virus (TEV) protease into the neck of the tetrameric motor Kinesin-5. Application of TEV protease led to a specific depletion and functional loss of Kinesin-5 in Drosophila embryos. With our approach, we revealed that Kinesin-5 stabilizes the microtubule network during interphase in syncytial embryos. The ‘molecular guillotine’ can potentially be applied to many motor proteins because Kinesins and myosins have conserved structures with accessible neck regions. This article has an associated First Person interview with the first author of the paper.


INTRODUCTION
Cytoskeletal motor proteins, including myosins, dyneins and kinesins, convert the chemical energy of ATP hydrolysis into mechanical work. Motor proteins are widely involved in multiple fundamental cellular processes such as intracellular transport, cell division, cell shape change and migration (Schliwa and Woehlke, 2003). The structure of motor proteins is conserved. They contain a motor domain, referred to as the head, which catalyzes ATP and binds microtubules or F-actin. The catalytic cycle links ATP hydrolysis to a conformational change of the protein that translates into unidirectional movement of the motor protein on the filament. A second part of the protein, the stalk, links the head to the cargo binding site, contains coil-coiled structures for oligomerization or associates with other subunits. Head and stalk are parts of the same polypeptide, which is functionally relevant as a tight link of head and stalk is essential for transmission of mechanical force (Kent and Lele, 2017).
Genetic analysis of the physiological function of motor proteins has been hampered by their essential function for the cell or organism. For example, Kinesin-5 serves indispensable functions during mitosis (Heck et al., 1993), making an analysis of its function in interphase or in terminally differentiated cells difficult. Conditional mutations, such as temperature-sensitive alleles, can overcome these limitations of genetic analysis (Enos and Morris, 1990). Gene knockdown by RNAi approaches relies on protein turnover, leading to insensitivity of stable proteins. Pharmacological approaches with small-molecule inhibitors or specific antibodies provide an alternative and have been applied for motor protein inhibition (Peterson and Mitchison, 2002;Sharp et al., 2000Sharp et al., , 1999. However, chemical approaches cannot be generalized, and need to be developed case by case.
Kinesin-5 proteins are members of kinesin superfamily, consisting of a motor domain and a coiled-coil rod with the central bipolar assembly (BASS) domain. Kinesin-5 can crosslink anti-parallel aligned microtubules by forming bipolar homo tetramers (van den Wildenberg et al., 2008). The motor activity enables filament sliding, e.g. during formation and elongation of the mitotic spindle (Kapitein et al., 2005;Shimamoto et al., 2015;Waitzman and Rice, 2014). In Drosophila syncytial embryos, Kinesin-5 (also known as Klp61F) is enriched at mitotic spindles (Barton et al., 1995) and is essential for spindle formation and chromosome segregation. Injection of antibodies specific for Kinesin-5 into embryos leads to a collapse of newly formed spindles and the formation of mono-asters (Sharp et al., 2000(Sharp et al., , 1999. Making proteins susceptible to proteolytic cleavage represents a generally applicable strategy for generation of conditional alleles (Harder et al., 2008;Oliveira et al., 2010;Pauli et al., 2008). Here, we apply this concept to motor proteins by inserting a proteolytic site between the head and stalk region (the 'neck'). We designated this strategy a 'molecular guillotine' (Fig. 1A). We chose the wellcharacterized Kinesin-5 proteins in order to demonstrate the feasibility of this approach. As a protease, we employ tobacco etch virus (TEV), which is highly specific. No match with the TEV recognition motif within the Drosophila proteome has been identified, and flies expressing TEV are viable and fertile (Harder et al., 2008).

Design of a 'molecular guillotine'
We inserted three copies of the TEV recognition motif at one of two positions, G394 or Q499, into the stalk region. G394 and Q499 are located within conserved coiled-coil regions next to the head domain ( Fig. 1B,C). In addition, we fused GFP to the C-terminus, which does not affect the function of Kinesin-5, as previously reported (Cheerambathur et al., 2008). These constructs were expressed as transgenes in levels comparable to the endogenous allele with a ubiquitin promoter, as assayed by western blot (Fig. 1D). As a result of the C-terminal GFP moiety, the constructs showed a slower mobility in SDS-PAGE compared with wild-type Kinesin-5. The TEV sites do not affect the functionality of Kinesin-5 as the construct with the insertion at G394 (Kin-5[G394tev]-GFP) complemented the lethality of the Klp61f 07012 mutation. For this, we recombined Kin-5[G394tev]-GFP with a Klp61f mutation. The resulting flies only expressed Kin-5[G394tev]-GFP, were viable and fertile and can be kept as a homozygous stock. In embryos from this line, Kinesin-5 was detected only at the molecular mass corresponding to transgenic Kin-5[G394tev]-GFP, which confirms the absence of endogenous Kinesin-5 (Fig. 1E).

Kinesin-5 cleavage in vivo
We expressed TEV protease in stripes in embryos under the control of the engrailed promoter. Control embryos with no TEV expression showed uniform Kin-5[G394tev]-GFP expression. In contrast, the GFP signal was strongly depleted in stripes upon expression of TEV ( Fig. 2A). Next, we turned to syncytial embryos, which are characterized by their rapid and synchronous nuclear division cycles and the associated remodeling of the cytoskeleton. During mitosis, microtubules and their motors are important for the formation and function of mitotic spindles, and for chromosome segregation, whereas they function in nuclear arrangement and stabilization of the nuclear array in interphase (Kanesaki et al., 2011;Winkler et al., 2015). Kinesin-5 localizes to the mitotic spindle and is involved in chromosome segregation during mitosis (Cheerambathur et al., 2008;Sharp et al., 2000Sharp et al., , 1999. We microinjected TEV protease into syncytial embryos and recorded GFP fluorescence. Following injection of TEV protease but not buffer, GFP fluorescence rapidly dropped (Fig. 2B). Correspondingly, the specific staining pattern, such as labeling of mitotic spindles or cytoplasmic asters was lost in TEV-injected embryos (Fig. 2D). Quantification of total GFP fluorescence provided an estimate for an approximate half life of about 30 min (Fig. 2C). Kinesin-5 was specifically cleaved, since the electrophoretic mobility of Kin-5[G394tev]-GFP was higher in TEV-injected compared with buffer-injected embryos (Fig. 2E). Kin-5[G394tev]-GFP embryos were lysed about 30 min after injection and the C-terminus of Kinesin-5 extracts analyzed by western blotting. The observed difference in electrophoretic mobility was consistent with proteolytic cleavage at the TEV sites at the neck and corresponding loss of the head domain. As we detected a single band, proteolytic cleavage was close to complete under our experimental conditions (Fig. 2E).

Cleavage of Kinesin-5 causes the loss of function in mitosis
Next, we analyzed the functional consequences of Kinesin-5 cleavage. To track the nuclear cycles and behavior of chromosomes, we co-injected fluorescently labeled histone-1 and TEV protease into Klp61f null embryos expressing the Kin-5[G394tev]-GFP transgene (ΔKin5, Kin-5-tev-GFP). Following TEV injection, we observed a failure of chromosome separation and monoastral spindles (Fig. 3A,B). These phenotypes were observed in individual spindles interspersed between normally appearing spindles. These phenotypes were consistent with the previously reported mitotic defects following injection of antibody against Kinesin-5 (Sharp et al., 1999). We quantified the percentage of failed spindles in the following embryos after TEV injection: (1) ΔKin-5, Kin-5-tev-GFP, (2) Kin-5-tev-GFP and (3) wild type (Fig. 3C). Whereas no defects were observed in wild-type embryos, about three-quarters of the spindles did not form or collapsed in embryos containing only the TEV-cleavable Kinesin-5 protein. The proportion of abnormal spindles was ∼10%, if both TEV-cleavable and wild-type Kinesin-5 was present. This indicates that the TEV-cleavable form of Kinesin-5 exerts a dominant-negative activity on the wild-type protein, consistent with Kinesin-5 acting as a homo-tetramer.

Interphase function of Kinesin-5 in syncytial embryos
Compared with the well-established mitotic function, much less is known about how Kinesin-5 is involved in microtubule organization and function in interphase. In interphases of syncytial embryos, nuclei and the associated centrosomes form a regular arrangement (Fig. 4A), and Kinesin-5-GFP is strongly enriched at the centrosomes and associated asters. In addition, dynamic extended structures between adjacent asters were detected (Fig. 4B). These transient signals may represent microtubules coated with Kinesin-5 and possibly antiparallel aligned microtubules. We noticed that in Kinesin-5-depleted embryos, two nuclei originating from different spindles fused together after successful separation from their own sister nuclei (Fig. 4C), which implies an interphase function of Kinesin-5 for keeping nuclei separate and at a distance. As hypothesized previously (Kanesaki et al., 2011;Winkler et al., 2015), Kinesin-5 may be involved in nuclear positioning and formation of the nuclear array in syncytial Drosophila embryos. We examined the position of the nuclei in interphase carefully (Fig. 4D,E) and quantified the number of neighboring cells for each nucleus (Fig. 4F). In control embryos, 50% of nuclei had 6 neighbors, whereas 33% and 13% had 5 and 7 neighbors, respectively. In contrast, the distribution was different in the Kinesin-5-depleted embryos, with significantly fewer nuclei with 6 neighbors and significantly more with fewer than 5, or more than 7 neighbors (Fig. 4F). As a measure of the regularity of the nuclear arrangement, we calculated the variation in distances to neighboring cells [standard deviation (σ) normalized to the distance average (µ)]. We detected a much higher variation in Kinesin-5depleted embryos compared with the control embryo (Fig. 4G). In summary, our data indicate that Kinesin-5 is involved in interactions between adjacent asters leading to a normal arrangement of the nuclei in interphase.  Table S1. Kinesin-5 bound to anti-parallel aligned microtubules may push adjacent asters away from each other and thus generate a repulsive force, which may lead to uniform internuclear distances. In this model, Kinesin-5 would promote movements of centrosome and their associated asters. Alternatively, Kinesin-5 may crosslink microtubules from adjacent asters and stabilize the syncytial microtubule network. In this alternative model Kinesin-5 would suppress movement of centrosomes and associated asters (Fig. 5A). To distinguish these two models, we recorded the dynamics of centrosomes in the scale of seconds and calculated second-scale fluctuations of the centrosomes defined as previously reported (Fig. 5B,C) (Winkler et al., 2015). We calculated fluctuations, D i (t j ), as the deviation from slow (minute-scale) drift movements averaged over time and multiple centrosomes. These fluctuations have the dimension of a diffusion constant and do not contain slow drift movements. Comparing buffer-and TEV-injected embryos, we found that the fluctuations were strongly increased (about threefold) in Kinesin-5-depleted embryos. The baseline for passive, non-ATP-dependent fluctuations is about five times below that of the wild-type (Winkler et al., 2015) (Fig. 5D). Owing to the increased mobility of centrosomes in Kinesin-5-depleted embryos, we conclude that Kinesin-5 functions in the stabilization of the array of microtubule asters by crosslinking, for example.

DISCUSSION
The function of Kinesin-5 in spindle formation and elongation during mitosis is well established (Heck et al., 1993;Sharp et al., 2000). Consistently, depletion of Kinesin-5 by our guillotine method induced defects in chromosome segregation in syncytial Drosophila embryos consistent with previous reports (Sharp et al., 1999). Aside from its role in mitotic spindle formation, our insights are limited into the functions of Kinesin-5 during interphase. Study of interphase is hampered by the difficulty that interphase functions are obscured by mitotic defects. In the case of Drosophila, embryos depleted of Kinesin-5 cannot be obtained, since Kinesin-5 has an essential function during oogenesis (Radford et al., 2017). By employing small-molecule inhibitors, Kinesin-5 was found to be involved in maintenance of neuron dendritic structure (Kahn et al., 2015) and intracellular transport from Golgi to the cell surface in cultured cells (Wakana et al., 2013). As another approach for conditional interference, we developed a method for decapitating motor proteins by proteolysis or the 'molecular guillotine'. This method is potentially suitable for many motor proteins. We demonstrate the feasibility of this approach with Kinesin-5 and TEV protease. We revealed an interphase function of Kinesin-5 in syncytial Drosophila embryos for the stabilization of the microtubule network and keeping adjacent nuclei at a distance.
In syncytial embryos, the microtubule asters originating from centrosomes can directly interact with neighboring asters, since they are not physically separated by plasma membranes. These interactions lead to the formation of an extended network covering the embryonic cortex. The phenotypic behavior of centrosomes and their associated nuclei reflect their intrinsic properties but also, as part of the network, include any influences from the neighbors. Adjacent microtubule asters potentially interact via crosslinkers such as Feo/Ase1p, bundling proteins or motors  Table S1.
with sliding activity, such as Kinesin-5. Here, we tested the hypothesis that Kinesin-5 generates repulsive forces between adjacent astral microtubules in interphase. We expected that a loss of force generation would lead to a reduced mobility of the network and its nodes, the centrosomes. Using the fluctuations of centrosomes as an indicator of network dynamics, we rejected our hypothesis, because we measured an increased mobility of the centrosomes, when Kinesin-5 was inactivated. These data suggest that the function of Kinesin-5 as a crosslinker is more dominant in vivo than its function for sliding of anti-parallel aligned microtubules and thus pushing apart adjacent microtubule asters. The in vivo function of Kinesin-5 is similar to that of kinesin-1, which is enriched at the cortex and in F-actin and actin caps (Winkler et al., 2015). Both may be involved in anchoring microtubule asters to the cortex and in this way counteract the fluctuation movements of centrosomes. Having identified a stabilizing function of Kinesin-5, the question remains about the origin of the forces driving centrosome fluctuations. Fluctuations are due to an active component, since ATP depletion leads to loss of fluctuations. The minus-end directed motor kinesin-14 may serve as a force generator (Sharp et al., 2000).
The 'molecular guillotine' is potentially a versatile method for conditional inactivation of motor proteins. TEV protease has been employed for inactivation of cohesin in yeast (Uhlmann et al., 2000) and fly (Pauli et al., 2008), as well as for inactivation of claudin in Drosophila (Harder et al., 2008). However, this approach had not been applied for inactivation of motor proteins to our knowledge. The approach of a 'molecular guillotine' as reported in this study can be applied widely to members of the motor protein families. Unlike using the small-molecule inhibitor (Engelke et al., 2016;Peterson and Mitchison, 2002), TEV protease can be specifically and conditionally expressed using the UAS-GAL4 system and its temperature-dependent variants in any genetically tractable cell type. This allows 'decapitation' of the selected motor protein in a tissue-and developmental stage-specific manner. The molecular guillotine offers an alternative approach to other previously reported conditional approaches based on protein degradation such as the auxin-inducible degradation system (Gray et al., 2001), the degron (Dohmen et al., 1994) or deGradFP systems (Caussinus et al., 2012), which rely on the ubiquitin-mediated protein degradation machinery. In the case of TEV cleavage, a single peptide bond is cleaved, whereas in the degron or deGradFP systems, the protein complex is subjected to complete proteolysis in the proteasome. Direct comparison of these systems will reveal the respective advantages and disadvantages with respect to kinetics and effectiveness of motor depletion. A second intrinsic difference to degradation-based systems is a potential dominant effect of the molecular guillotine. The single cut generates two parts, which may interact with uncut counterparts and thus induce a dominant effect. We detected evidence for such a dominant effect, when the phenotype caused by addition of TEV protease is also detected in heterozygous embryos containing a Kinesin-5 wild-type allele and a copy of the Kin-5-tev-GFP transgene.
In summary, the novel approach of a molecular guillotine enabled us to investigate a specific function of the motor protein Kinesin-5 in interphase. Potentially, the decapitation approach can be correspondingly applied to other kinesin motors, as well as myosins, as they have a related domain structure in common.

Cloning
A sequence coding for three recognition sites of TEV protease (PS ENLYFQG PR ENLYFQG GS ENLYFQG PR) was inserted behind the  Table S1. codons of G394 or Q499 of the Klp61f cDNA (SK-Klp61f, Drosophila Genomic Resource Center, Bloomington). The resulting coding sequence (Kinesin-tev) was cloned behind the ubiquitin promoter (Lee et al., 1988;Oda and Tsukita, 2001) and 5′ to eGFP into the multiple cloning site of a pUASt vector derivative lacking the UAS and hsp70TATA sites. The Kinesin-5-GFP fusion constructs were generated by cloning the Klp61f cDNA (from SK-klp61f ) with GFP inserted at the C-terminus into the transformation vector sGMCA (Kiehart et al., 2000), which contains the spq promoter for ubiquitous expression. Sequence information and details of the cloning procedure are available upon request.

Fluctuation analysis
Centrosome tracking and measurement of fluctuation were carried out as previously described (Winkler et al., 2015). The custom-made software code is available on request.