Abstract
Cdk1 has been found to phosphorylate the majority of its substrates in disordered regions. These phosphorylation sites do not appear to require precise positioning for their function. The mitotic kinesin-5 Cin8 was shown to be phosphoregulated at three Cdk1 sites in disordered loops within its catalytic motor domain. Here, we examined the flexibility of Cin8 phosphoregulation by analyzing the phenotypes of synthetic Cdk1-sites that were systematically generated by single amino-acid substitutions, starting from a phosphodeficient variant of Cin8. Out of 29 synthetic Cdk1 sites that we created, eight were non-functional; 19 were neutral, similar to the phosphodeficient variant; and two gave rise to phosphorylation-dependent spindle phenotypes. Of these two, one site resulted in novel phosphoregulation, and only one site, immediately adjacent to a native Cdk1 site, produced phosphoregulation similar to wild-type. This study shows that, while the gain of a single phosphorylation site can confer regulation and modulate the dynamics of the spindle, to achieve optimal regulation of a mitotic kinesin-5 motor protein, phosphoregulation has to be site-specific and precise.
Introduction
Mitotic cell division is an essential and highly regulated process by which genomic information in the form of duplicated chromosomes is faithfully transmitted from mother to daughter cells. This process is mediated by the mitotic spindle, a microtubule (MT)-based structure that undergoes a distinct set of dynamic morphological changes with precise temporal and spatial regulation. One of the major coordinators of the mitotic events is the conserved cyclin-dependent protein kinase 1 (Cdk1). The mitotic functions of Cdk1 are activated by association with the B-type cyclins (Clb(1-6) in yeast). The levels of the six cyclins oscillate (Morgan & Roberts, 2002), thus helping to provide stage-specific functions for Cdk1 (Nasmyth, 1996). Cdk1-dependent phosphorylation of mitotic substrates is also affected by the activities of protein phosphatases such as the PP2Acdc55 (Queralt, Lehane et al., 2006) and Cdc14 (D’Amours & Amon, 2004, Stegmeier & Amon, 2004), which dephosphorylate Cdk1 target sites. Thus, mitotic events are coordinated by the balance between the activities of Cdk1 and opposing phosphatases.
Previous work identified a large class of “flexible” Cdk1 substrates that were enriched for phosphorylation sites throughout evolution, but where the precise site position was not strictly conserved (Holt, Tuch et al., 2009). These phosphorylation sites tended to occur in disordered regions, including loops and termini of proteins. It was suggested that these phosphates might interact with modular phospho-binding motifs or disrupt a protein-protein interface, forms of regulation that are relatively tolerant of changes in site position (Holt et al., 2009). A well-characterized example of a Cdk1 target with a flexible phosphorylation cluster is the Sic1 protein, a Cdk1 inhibitor in Saccharomyces cerevisiae (Mendenhall, Jones et al., 1987, Schwob, Bohm et al., 1994). For proper phosphoregulation of Sic1, a minimal set of priming phosphorylation sites and phospho-degrons is required, while distances between the sites in the cluster can be flexible (Koivomagi, Ord et al., 2013, Koivomagi, Valk et al., 2011a). A second, smaller set of substrates was found that maintained a precise position throughout long periods of evolution. These sites were often found in ancient metabolic enzymes, suggesting that they may have evolved early and adopted a highly efficient form of phosphoregulation that relies on exact conformation changes (Holt et al., 2009). To date, there is very little experimental data directly testing the flexibility of phosphoregulation.
To address this issue, we experimentally examined the flexibility of phosphoregulation by Cdk1 using the S. cerevisiae mitotic kinesin-5 Cin8 as a model-protein. These conserved bipolar kinesin-5 motors perform essential roles in mitotic spindle dynamics of eukaryotic cells by crosslinking and sliding apart antiparallel microtubules of the spindle (reviewed in (Ferenz, Gable et al., 2010, Goulet & Moores, 2013, Hildebrandt & Hoyt, 2000, Kashina, Rogers et al., 1997, Singh, Pandey et al., 2018, Valentine, Fordyce et al., 2006, Valentine & Gilbert, 2007, Waitzman & Rice, 2014)). Their function had been shown to be regulated by Cdk1 in various organisms (Avunie-Masala, Movshovich et al., 2011, Blangy, Arnaud et al., 1997, Blangy, Lane et al., 1995, Cahu, Olichon et al., 2008, Chee & Haase, 2010, Crasta, Huang P Fau - Morgan et al., 2006, Gardner, Bouck et al., 2008, Goldstein, Siegler et al., 2017, Shapira & Gheber, 2016, Sharp, McDonald et al., 1999). The budding yeast S. cerevisiae encodes two kinesin-5 homologues, Cin8 and Kip1, that partially overlap in spindle assembly and maintenance of the bipolar spindle structure (Hoyt, He et al., 1992, Roof, Meluh et al., 1992, Saunders & Hoyt, 1992), in focusing kinetochore clusters (Gardner et al., 2008, Tytell & Sorger, 2006, Wargacki, Tay et al., 2010), in anaphase B spindle elongation (Saunders, Koshland et al., 1995, Straight, Sedat et al., 1998), and in stabilizing and organizing the middle-spindle midzone, an overlapping array of antiparallel MTs (Fridman, Gerson-Gurwitz et al., 2009, Fridman, Gerson-Gurwitz et al., 2013, Gerson-Gurwitz, Movshovich et al., 2009, Ibarlucea-Benitez, Ferro et al., 2018, Movshovich, Fridman et al., 2008).
Kinesin motors undergo diverse phosphoregulation, including by Cdk1 (reviewed in (Gibbs, Greensmith et al., 2015, Goulet & Moores, 2013, Waitzman & Rice, 2014)). In particular, the mitotic functions of both S. cerevisiae kinesin-5 Cin8 and Kip1 were shown to be regulated by Cdk1 (Avunie-Masala et al., 2011, Chee & Haase, 2010, Goldstein et al., 2017, Shapira & Gheber, 2016). For Cin8, such phosphoregulation occurs primarily at three Cdk1 sites, which are located in disordered loops 8 and 14 within the catalytic motor domain (Avunie-Masala et al., 2011, Goldstein et al., 2017, Shapira, Goldstein et al., 2016). We therefore asked how easy it is to recapitulate or create new phosphoregulation by generating novel Cdk1 sites within Cin8, adjacent to and distant from the native sites. To answer this question, we examined the phenotypes of 29 new possible Cdk1-sites that were systematically generated by a single amino-acid substitution starting from a phosphodeficient allele of Cin8. By combining a comprehensive genetic, cell biological and biochemical characterization of these mutants we found that the three native sites in the catalytic domain of Cin8 can each alone support partial regulation of Cin8. We also found that out of 29 novel synthetic Cdk1 sites that we created, eight (28%) resulted in non-functional Cin8, 19 (65%) resulted in a neutral spindle phenotype similar to the phosphodeficient variant, and two (~7%) gave rise to phosphorylation-dependent spindle localization phenotypes. Of these two, one site, one amino-acid proximal to a native Cdk1 site, closely recapitulated the original phosphoregulation without destabilizing Cin8 or perturbing its function. This study shows that while the gain of a single phosphorylation site can confer complex regulation and thereby modulate the dynamics of the mitotic spindle, to achieve optimal regulation of a mitotic kinesin-5 motor protein, phosphoregulation has to be highly constrained.
Results
Strategy for generation of synthetic Cdk1 phosphorylation sites in Cin8
Our strategy to examine the flexibility of phosphoregulation of Cin8 by Cdk1 was to introduce novel consensus Cdk1 sites in the Cin8 ORF in a systematic manner. As a basis for mutagenesis, we used a phosphodeficient variant of Cin8 (Cin8-5A) carrying mutations of the phosphoacceptor serines and a threonine to alanine within the five native Cdk1 sites of Cin8, at positions S277, T285, S493, S763, and S1010. New Cdk1 sites were systematically created by single amino-acid replacement in most locations such that replacement resulted in a novel partial or full Cdk1 consensus site (Fig. 1A and B). We comprehensively created mutations around every proline as this amino acid is the major specificity determinant for Cdk1. For every proline, a serine was introduced at position -1 to generate a Cdk1 phosphorylation consensus sequence (S-P or S-P-x-K/R). If lysine or arginine were already present at the +3 position relative to the new serine, this mutation immediately generated a full site. Otherwise, a partial site was created (also referred to as a minimal consensus site). In these cases, a second synthetic allele was additionally generated with a lysine at position +3 to the serine, to produce the full Cdk1 phosphorylation consensus [S-P-x-K] (Fig. 1B class 1). In addition, we created Cdk1 sites by substitution of a proline only in cases where full Cdk1 phosphorylation consensus sites were created by this single change (i.e., [S-x-x-K] to [S-P-x-K]; Fig. 1B class 2). We limited the proline substitutions in this fashion because proline substitutions are often poorly tolerated due to the structural constraints of this amino acid. To avoid any bias, we generated the new sites based on the sequence only, without excluding structured and conserved regions. Using this approach, we both reintroduced three native Cdk1 sites at the motor domain of Cin8, termed Cin8 natural-phosphorylation variants (Cin8nat), and additionally generated 29 novel sites termed Cin8 synthetic-phosphorylation variants (Cin8syn). For variants that exhibited potential phosphoregulation phenotypes (see below), we created control mutations in which the serine within the consensus sequence was changed to alanine, which lacks a phosphoacceptor hydroxyl group (Fig. 1B) (see Materials and Methods).
Single Cdk1 sites at the native positions confer regulation
Re-generation of the native Cdk1 sites by the above described approach allowed us to examine the functionality of these sites as a sole source of Cdk1 phosphoregulation and compare their phenotypes to the wt Cin8 and the phosphodeficient Cin8-5A. We previously demonstrated that each of the native Cdk1 phosphorylation sites in the motor domain of Cin8, S277, T285, and S493, are independently phosphorylated by Cdk1 in vitro and that mutations to alanine of each of these sites cause different phenotypes during anaphase (Goldstein et al., 2017). Here, the native sites were examined as partial (suffixed by “p”) and full (suffixed by “f”) Cdk1 consensus sequences since they were generated as variants of class 1 (Fig. 1B). It is worth noting that although the native Cdk1 phosphorylation site at position 285 is composed of a threonine followed by a proline, here we introduced a serine to result in S-P consensus (see Materials and Methods). The native variants were first examined by general viability test, then by two dynamic spindle-localization features: translocation to the midzone at mid-anaphase and detachment from the spindle at late anaphase. Finally, the functionality of the variants was evaluated by analysis of anaphase spindle elongation dynamics (Goldstein et al., 2017).
In order to determine if modifications to the native sites substantially alter the functionality of Cin8, we conducted a yeast viability test in which the Cin8nat variants were expressed from a centromeric (CEN) plasmid as a sole source of kinesin-5 function in cells deleted for the chromosomal copies of CIN8 and KIP1. Cin8 and Kip1 have overlapping roles. However, at least one of them is required for proper spindle formation and yeast viability (Avunie-Masala et al., 2011, Duselder, Fridman et al., 2015, Gheber, Kuo et al., 1999, Hoyt et al., 1992, Roof et al., 1992, Saunders & Hoyt, 1992). We found that cells expressing native partial and full Cin8 phosphorylation variants (Cin8nat) and their controls, in which the serine at the Cdk1 site was mutated to alanine, are viable at all temperatures (Fig. 2A). Therefore, these single phosphorylation site alleles are sufficient for the essential function of Cin8.
Next, we examined the localization of Cin8 variants, tagged with 3GFP, during anaphase B spindle elongation (Fig. 2B). Consistent with previous results (Avunie-Masala et al., 2011, Goldstein et al., 2017), we observed that at the early stages of anaphase Cin8 is distributed along the spindle and later detaches from the spindle and becomes diffusively localized around the spindle pole bodies (SPBs), likely in the divided nuclei (Fig. 2B). In order to quantitatively compare the localization patterns of Cin8 variants during anaphase B, we performed line-scan intensity analysis of Cin8 distribution along the spindle at different anaphase spindle lengths, as previously described (Goldstein et al., 2017). We found that the phosphodeficient mutant Cin8-5A is mainly concentrated at the SPBs (Fig. 2C, black arrows), while wt Cin8 exhibited a lower concentration at the SPBs and higher levels along the spindle (Fig. 2C gray arrow). This result indicates that Cin8-5A is unable to translocate from the SPBs to the spindle midzone and that translocation is regulated via phosphorylation of Cin8 in at least one Cdk1 site. We previously obtained similar results with the Cin8-3A variant, carrying phosphodeficient mutations to alanine at the three Cdk1 sites in the catalytic domain of Cin8 (Goldstein et al., 2017). This indicates that, consistent with a previous report (Avunie-Masala et al., 2011), phosphoregulation of Cin8 is dependent mainly on phosphorylation of Cdk1 sites in its catalytic domain. Quantitative analysis of spindle localization also revealed that in short spindles of 3-4 μm, the partial and full Cin8nat277 variants and the full Cin8nat493f but not the partial Cin8nat285p had recruitment of Cin8 throughout the spindle length (Fig. 2C 3- 4 μm), similarly to wt Cin8. In contrast, the full Cin8nat285f exhibited a sharp peak of Cin8 at the middle of the spindle (Fig. 2C 3-4 μm). Overall, our data indicate that the S277 and S493 sites present phenotypes that are similar to the wt Cin8, while S285 site is similar to the Cin8-5A variant, consistent with our previous report (Goldstein et al., 2017).
To characterize the detachment of the native Cin8 phosphorylation variants from the spindle at late anaphase (Fig. 2B, yellow arrows), we performed quantitative analysis of Cin8 fluorescence signal perpendicular to the spindle (Fig. 2D) as previously described (Goldstein et al., 2017). We found that the detachment of wt Cin8 from the spindle in late stages of anaphase B was abolished by the phosphodeficient mutant Cin8-5A (Figs. 2E and EV3), indicating that phosphorylation by Cdk1 is required for this detachment (Avunie-Masala et al., 2011, Goldstein et al., 2017). We found that partial and full native variants at position S277 exhibit high degree of detachment, most similar to that of the wt Cin8; partial and full variants at position S285 being no different from the phosphodeficient CIn8-5A, while the variant at position S493 has an intermediate effect (Figs. 2E and EV3). This is consistent with previous reports that phosphorylation at the S277 site is the major regulator of Cin8 distribution along the spindle and its detachment from the spindle.
Categorization of the synthetic Cdk1 sites
Results presented so far demonstrate that single Cdk1 sites in the motor domain of Cin8 produce phenotypes markedly different from those of the phosphodeficient Cin8-5A (Fig. 2), indicating that phosphorylation at a single Cdk1 site can, at least partially, regulate the function of Cin8. Thus, we next examined the phenotypes of the synthetic phospho-variants of Cin8 (Cin8syn). Since phosphoregulation among kinesin-5 motors produces varied effects including effects on motor activity (Gibbs et al., 2015, Goldstein et al., 2017, Morfini, Schmidt et al., 2016, Ritter, Kreis et al., 2015, Waitzman & Rice, 2014, Wojcik, Buckley et al., 2013), predicting possible outcomes as a result of introducing novel sites for Cdk1 phosphorylation is not trivial. Yet we postulated hypothetical outcomes: we expect some mutations to result in loss of function due to perturbation of Cin8 structure. Other mutants may result in no structural perturbation and no observable phosphoregulation effect; these would be neutral mutations. And finally, there may be mutants that result in regulation of Cin8 function, either by recapitulating the original phosphoregulation or resulting in novel regulation. To distinguish between these possibilities, the 3GFP-tagged Cin8syn variants were first screened for their localization to anaphase spindles by live-cell imaging (Fig. 3) and for their ability to support the viability of cells deleted for the function of Cin8 and Kip1 as a sole source for kinesin-5 function (Fig. 3). Out of the 29 Cin8syn mutants that were generated, eight (28%) were found to carry deleterious mutations and resulted in cell death. These mutants exhibited mostly diffusive detachment of Cin8 from the spindle that was not abolished by the alanine control variants and thus were not due to phosphorylation (Fig. 3). Of these eight variants, five carried proline insertion mutations (Fig. 1B Class 2), likely disrupting the secondary structure of Cin8. The largest class of mutants were neutral. We found 19 mutants (~65%) that had no discernible effect on localization, i.e., Cin8-5A-like phenotype (Fig. 3) (see Materials and Methods). These sites could be neutral for several reasons. They may not be accessible for Cdk1 phosphorylation, phosphatases may keep steady-state phosphorylation levels low, or it could be that phosphorylation at these sites does not influence Cin8 function. Finally, two novel Cdk1 positions (7%) resulted in a phosphorylation-dependent spindle-localization phenotype, i.e., control variants abolished or corrected part of the phenotypes induced by the novel Cdk1 sites S148 and S276 (Figs. 3, 4, EV1, EV2, and EV3). The last were further investigated by quantitative dynamic localization analysis (Figs. 4 and EV3), in vitro kinase assays, and spindle elongation measurements (Figs. 5, 6, and EV6).
Two synthetic Cdk1 sites confer differential phosphoregulation
The novel sites at positions S148 and S276 exhibited differential effects on the dynamic detachment of Cin8 from the spindle (Figs. 3, 4A, B, EV1, EV2 and EV3). The partial Cin8syn148p variant showed no detachment, similar to the control variant Cin8syn148pc. However, the full Cin8syn148f variant exhibited high levels of detachment from the spindle at anaphase, comparable to wt Cin8. This novel regulation was lost in the non-phosphorylatable alanine control allele Cin8syn148fc (Figs. 4A, B, EV1, and EV2), indicating that Cin8 detachment from the spindle of the full Cin8syn148f is dependent on phosphorylation at this site. The full Cin8syn276f variant also exhibited detachment from the spindle but to a lower extent. This detachment was also abolished by the control Cin8syn276fc mutation (Figs. 4A, B, EV1, and EV2), indicating that this detachment is also phosphorylation dependent.
Cin8syn276f, which contains a full Cdk1 phosphorylation consensus sequence in position S276 and is of Class 2 (Fig. 1B), was viable at all temperatures, similar to the wt Cin8. In contrast, the mutant bearing a full Cdk1 consensus sequence at position S148 (Cin8syn148f) was temperature sensitive at temperatures higher than 33°C (Figs. 3 and 4C). The control variant Cin8syn148fc partially rescued the phenotype and grew at 33°C (Fig. 3 and 4C), indicating that gain of phosphate at these positions interferes with Cin8 function. All S148 variants were stable at 26°C and 33°C, similarly to the native and the synthetic S276 variants (Fig. EV4), suggesting that mutation at this site affects mainly protein function and not its stability.
Quantitative examination of anaphase spindle distribution (Goldstein et al., 2017) revealed that synthetic variants also regulated the translocation of Cin8 from the SPBs to the midzone (Fig. 4D), as indicated by reduced intensity at the SPBs and elevated intensity at the midzone compared to the phosphodeficient Cin8-5A. At short and intermediate anaphase spindles, the full Cin8syn148f variant exhibited higher distribution to the midzone and lower distribution to the SPBs compared to the partial Cin8syn148p counterpart (Fig. 4D 3-4 μm and 4-5 μm red arrows). These results indicate that the synthetic Cdk1 sites regulate the localization of Cin8 on anaphase spindles and that this regulation is more pronounced in full sites compared to partial Cdk1 consensus sequences. Finally, the translocation from the SPBs to the spindle of the full Cin8syn276f variant was less pronounced compared to the full variants at the S148 site. Taken together, our data indicate that the two novel sites produce different effects, with S276 mainly affecting detachment from the spindle (Fig. 4A, B, EV1, EV2, and EV3), and the S148 site affecting both Cin8 translocation to the midzone at mid anaphase and detachment from the spindle at late anaphase (Figs. 4, EV1, EV2, and EV3).
To examine whether the newly created Cdk1 sites are accessible for phosphorylation by Clb2/Cdk1, we performed in vitro phosphorylation assays using a purified Cin8 motor domain as a substrate, as previously described (Avunie-Masala et al., 2011, Goldstein et al., 2017). Our data indicate that all native variants, S277, S285, and S493, undergo phosphorylation in vitro. At the S277 and S285 sites, phosphorylation of the full Cdk1 phosphorylation consensuses is ~20-fold higher than of their partial counterparts (Fig. 5 right), consistent with previous reports (Chang, Begum et al., 2007, Koivomagi, Valk et al., 2011b, Loog & Morgan, 2005). These higher phosphorylation levels do not correlate with the relatively small differences, if any, in Cin8 localization patterns, along and perpendicular to the spindle, between the partial and full phosphorylation consensus variants of these positions (Fig. 2). In contrast, variants with Cdk1 consensus at position S148 exhibited a low degree of phosphorylation compared to wt Cin8 and a less than ~2-fold increase in phosphorylation levels between the partial and the full Cdk1 phosphorylation consensus sites. This result indicates that position S148 is less accessible for Cdk1/Clb2 phosphorylation; however, partial and full consensus sites at this position result in completely different phenotypes. We also included the Cin8syn168f variant in this assay, which exhibits S/T independent diffusive Cin8 localization and is also non-viable at 26°C (Fig. 3). This mutant exhibited no phosphorylation in vitro (Fig. 5 left), indicating that phenotypes exhibited by this mutant are a result of a deleterious mutation that disrupts Cin8 structure and prevents Cin8 from performing its essential roles. In addition, the Cin8syn268f mutant was included in the assay, exhibiting a neutral Cin8-5A-like phenotype, and is a few positions upstream to the native Cdk1 phosphorylation at position S277 in the disordered loop 8. This mutant underwent substantial phosphorylation in vitro, ~3-fold higher than wt Cin8 (Fig. 5 right); however, no phosphoregulation of Cin8 localization was evident with this variant (Fig. 3), indicating that extensive phosphorylation at this position caused no phosphoregulation. And finally, we also included Cin8syn99 partial and full variants, which were temperature sensitive (Fig. 3). These mutants also exhibited high levels of in vitro phosphorylation with ~20-fold increase between partial and full Cdk1 sites (Fig. 5, left), but no detachment from the spindle (Fig. 3). Taken together, these results indicate that the precise position of the phosphorylation site, rather than the degree of phosphorylation, is the most important determinant of Cin8 phosphoregulation.
Synthetic Cdk1 sites control spindle dynamics
Finally, we examined the control of spindle dynamics by Cin8 phospho-variants (Figs. 6 and EV6). Spindle length was measured as a function of time, and rate was determined for the first (fast) and second (slow) phases of spindle elongation, as previously described (Avunie-Masala et al., 2011, Gerson-Gurwitz et al., 2009, Goldstein et al., 2017, Movshovich et al., 2008, Straight et al., 1998). For simplicity, Fig. 6 exhibits comparison of spindle elongation during the two phases between the full Cin8 variants, while Fig. EV6 presents the full comparison including the partial and control counterparts of these sites. Consistent with our previous report (Goldstein et al., 2017), we found no difference in the rate of elongation between the phosphodeficient Cin8-5A and wt Cin8 during the first phase (Fig. 6 left and EV6). In the slow phase, the rate of elongation of the phosphodeficient Cin8-5A variant was slightly faster than that of wt Cin8; however, this difference was not significant according to post-ANOVA comparison with Tukey procedure (Fig. 6 right). We found that the full variants at the native sites have no effect on spindle elongation during either elongation phase compared to wt Cin8 and Cin8-5A. On the other hand, our data indicate that the synthetic variants at positions S148 and S276 did alter anaphase spindle dynamics. The synthetic variant at position S148 significantly reduced the rate of elongation during the early fast elongation compared to the full native variants Cin8nat277 and Cin8nat285 (Fig. 6 left), while the full synthetic variant at position S276 significantly reduced the elongation rate during the slower second phase compared to the full native sites Cin8nat277 and Cin8nat285 (Fig. 6 right), with no effect on the fast phase. This suggests that whereas single native sites bearing full Cdk1 phosphorylation sites have little to no effect on spindle elongation rate, the synthetic sites are affecting Cin8 function during spindle elongation. Interestingly, although the synthetic site S276 is in very high proximity to the native site S277, their regulation of spindle elongation is very different (Fig. 6 right). Examination of rates of elongation of the partial and control variant also demonstrate a complex picture consistent with the notion that the synthetic variants differ from the native sites (Fig. EV6). For example, the variants bearing full Cdk1 phosphorylation consensus at native sites, Cin8nat277f and Cin8nat285f, exhibit faster elongation rates during the first phase compared to their partial counterparts, with control mutations having similar elevated rates (Fig. EV6 top). In contrast, the rate of the full variant Cin8syn148f is not faster than the partial variants and the control variants are significantly faster compared to the full variants (Fig. EV6 top). These results indicate that the synthetic phospho-variants affect spindle elongation in a unique manner, different from that of the native variants, likely affecting spindle localization and/or motor activity by mechanisms different from those of wt Cin8.
Discussion
In this study we examined the plasticity of phosphoregulation of the Cin8 mitotic kinesin-5 motor by Cdk1 in S. cerevisiae cells. We employed an extensive synthetic mutation experiment and characterized the functionality of Cin8 harboring both native and synthetic sites by four in vivo assays: viability test, translocation to the midzone in mid anaphase, release from the spindle in late anaphase, and rate of anaphase spindle elongation. We revealed the complexity of the effect of phosphorylation consensus at native Cdk1 sites and found two novel sites that conferred phosphoregulation, which did not, however, fully recapitulate the original phosphoregulation.
Possible mechanisms of phosphoregulation of Cin8 by native and synthetic Cdk1 sites
Most Cdk1 substrates contain multiple phosphorylation sites (Holt et al., 2009, Moses, Heriche et al., 2007). To initiate this study, we first investigated to what extent single sites were able to confer regulation. We found that each of the three native Cdk1 sites in the catalytic domain of Cin8 were functional alone. The three phosphorylation sites together could give more robust regulation, or perhaps fine-tune the activity of Cin8. Indeed, not all sites are exactly equal and, consistent with our previous work (Goldstein et al., 2017), there is evidence of cooperative regulation of Cin8. A number of kinesin-related proteins were shown to be regulated by phosphorylation at residues outside the catalytic domain (Andrews, Ovechkina et al., 2004, Bishop, Han et al., 2005, Blangy et al., 1995, Cahu et al., 2008, Drummond & Hagan, 1998, Fu, Ward et al., 2009, Giet, Uzbekov et al., 1999, Matsuoka, Ballif et al., 2007, Morfini, Szebenyi et al., 2002, Olsen, Vermeulen et al., 2010, Pigino, Morfini et al., 2009, Rapley, Nicolas et al., 2008, Sawin & Mitchison, 1995, Sharp et al., 1999, Smith, Hegarat et al., 2011, Wojcik et al., 2013, Zhang, Shao et al., 2011, Zhang, Ems-McClung et al., 2008). However, some kinesin motors, Cin8 included, were shown to be regulated by phosphorylation in their catalytic domain. In such cases, phosphorylation affected the motile properties of kinesin and their interaction with MTs (Avunie-Masala et al., 2011, Gerson-Gurwitz, Thiede et al., 2011, Hara & Kimura, 2009, Hizlan, Mishima et al., 2006, Mennella, Tan et al., 2009, Padzik, Deshpande et al., 2016, Shapira & Gheber, 2016), indicating that phosphorylation directly regulates the kinesin catalytic cycle. Thus, it is possible that the synthetic Cdk1 sites in the catalytic domain of Cin8 that confer phosphoregulation, directly affect its catalytic activity.
The native S493 site is located in loop 14 and is conserved among the kinesin-5 homologs (Fig. 7A). This site is located in the vicinity of the Cin8 ATP-binding pocket (Fig. 8), raising the possibility that phosphorylation may affect ATPase activity. The two other native sites, S277 and T285, are located in loop 8 of Cin8, which contains an unusually large insertion not present in the paralogous motor Kip1 and metazoan kinesin-5 motors. In the MT-bound state, loop 8 is located near the MT lattice (Cao, Wang et al., 2014, Goulet & Moores, 2013, Ogawa, Saijo et al., 2017, Wang, Cantos-Fernandes et al., 2017). Loop 8 was shown to regulate the directionality of Cin8 (Gerson-Gurwitz et al., 2011, Shapira & Gheber, 2016) and the non-canonical binding of Cin8 to MTs (Bell, Cha et al., 2017). One of the synthetic phospho-variants, S276, is located in loop 8, one amino-acid apart from the native S277 site (Fig. 7B). The phenotypes of this site resemble the phenotypes of the wt Cin8 and of the native Cin8nat277 variant (Figs. 2, 3, 4, and EV5), although they differ in in spindle elongation dynamics (Fig. 6), indicating that the mechanisms of regulation of the S276 and S277 sites are similar, but not identical. Two additional synthetic sites, Cin8syn286 and Cin8syn300 were introduced into loop 8, and one of these was efficiently phosphorylated in vitro, but these sites conferred no regulation (Fig. 3). It is possible that phosphorylation within loop 8, at the S277nat and S276syn positions, changes the conformation of loop 8, thus altering its interaction with the MT lattice and affecting the function of Cin8 and this conformational change requires a precise coordination of the phosphate.
Sequence alignment reveals that the second functional synthetic site, Cin8syn148, is also found in some kinesin-5 homologs (Fig. 7C), suggesting that this is an important site for kinesin-5 regulation that has been employed elsewhere in evolution. This site is located in α- helix 1 (Subbiah & Harrison, 1989) in the vicinity of loop 5 (Fig. 8) within the kinesin motor domain (Harrington, Naber et al., 2011, Maliga, Xing et al., 2006, Waitzman, Larson et al., 2011). Although the function of loop 5 is unknown, its proximity to the conserved nucleotide-binding P-loop element and the fact that several kinesin-5 inhibitors bind to this loop (Brier, Lemaire et al., 2004, Lad, Luo et al., 2008) indicate that loop 5 is important for kinesin-5 function. Thus, phosphorylation of the S148 site may affect the function of kinesin-5 motors by affecting the conformation of loop 5.
Correlation between in vitro phosphorylation and in vivo phosphoregulation
Results presented here indicate that the two synthetic variants at positions S148 and S276 affected spindle elongation rates in a phosphorylation-dependent manner (Fig. EV6). Changing from a partial to full consensus Cdk1 sequence decreased the elongation rate of the S148 variant, and introducing a non-phosphorylatable control mutation fully rescued this effect (Fig. EV6). This result suggests that synthetic phosphoregulation sites can modulate the dynamics of the spindle to various degrees in vivo. Thus, a single synthetic phosphoregulatory site can tune intracellular functions.
Previous results have indicated that phosphorylation of the three native sites in the catalytic domain of Cin8 is required for its detachment from the spindle (Avunie-Masala et al., 2011). However, results presented here indicate that there is only partial correlation between in vitro phosphorylation and in vivo spindle detachment of the different phospho-variants. For example, the synthetic variants at positions S99 and S268 exhibited a high level of phosphorylation in vitro that was considerably higher than phosphorylation of wt Cin8 or single native variants (Fig. 5, right). However, in spite of this high level of phosphorylation the synthetic variants don’t detach from the spindle at late anaphase, but rather exhibit a Cin8-5A-like phenotype (Fig. 3). In contrast, the full variant at position S148 exhibits a very low level of phosphorylation in vitro (Fig. 5 left), and yet exhibits high levels of detachment from the spindle (Figs. 3, 4B, EV1, EV2, and EV3). These results clearly indicate that although phosphorylation at a single amino-acid can induce detachment from the spindle, as with the native variants, a high level of phosphorylation in itself is insufficient to induce spindle detachment. Thus, the precise location of phosphorylation plays a critical role in phosphoregulation of Cin8.
Cell cycle networks are flexible, Cin8 regulation is constrained
To date there has been very little experimental data systematically testing the flexibility of phosphoregulation. Cell cycle networks have been extensively rewired over long evolutionary time-scales. For example, while the general cell-cycle network topologies are conserved from yeast to humans, many of the molecular components have been replaced (Cross, Buchler et al., 2011). Thus, it is clear that there is considerable flexibility in cell cycle regulation. This general feature is likely related to the fact that >90% of Cdk1-dependent phosphorylation events occur in disordered regions of proteins (Holt et al., 2009). Nevertheless, there is clearly constraint on phosphorylation sites (Nguyen Ba & Moses, 2010). As a number of kinesin-related proteins have been shown to be regulated by phosphorylation at different regions of the molecule we chose to investigate the flexibility of phosphoregulation using a mitotic kinesin-5 Cin8 as a model.
We found that two out of 29 Cin8syn mutations resulted in phosphorylation-dependent detachments. Of these two, one (position S276) was viable at all temperatures. Furthermore, although the phenotypes of the Cin8syn276 variant were similar to those of the native Cin8nat277 sites, the two alleles differed in their distribution along the spindle (Fig. EV5, left), timing of their detachment from the spindle (Fig. EV5, right), and in rate of anaphase spindle elongation (Fig. 6). Thus, although some of the aspects of phosphoregulation at the synthetic sites were similar to those of the native sites, none of the new synthetic sites precisely recapitulated the phenotypes of the native sites.
In summary, the data presented here indicate that phosphoregulation of the kineisn-5 Cin8 by Cdk1 is site-specific and constrained. A synthetic Cdk1 site only one amino-acid distant from a native site within a disordered loop could not fully recapitulate the native regulation. Kinesin-5 motors function by binding and moving along MTs via a precise catalytic cycle, with directionality, motor activity, and spindle binding all subject to phosphoregulation. Thus, regulation must be highly constrained to maintain optimal control of mitotic kinesin motors such that they can efficiently perform their role in mitotic spindle dynamics.
Materials and Methods
Generation of synthetic Cdk1 phospho-variants and their categorization
The synthetic Cdk1 sites of Cin8 were generated on the basis of the phosphodeficient variant of Cin8 in which serines or threonines in all five native Cdk1 sites (S277, T285, S493, S763, and S1010) were mutated to alanine. Sites for novel synthetic Cdk1 phospho-variants were chosen in a systematic manner, with each original proline in the sequence of Cin8 being targeted as a potential Cdk1 phosphorylation site. Out of 20 prolines in the sequence of full length Cin8, five are occupied by the native sites and the remaining fifteen prolines were tested for potential Cdk1 phosphoregulation sites. In these cases, to create a Cdk1 site, a serine was added replacing an amino-acid at position -1 to the proline, resulting in a partial ([S/T]-P) or full ([S/T]-P-x-[K/R]) Cdk1 phosphorylation site (Fig. 1B class 1). In one case, at position 465, a threonine was added at position -1 to an existing proline. In the case where a partial Cdk1 phosphorylation site is created by this mutagenesis, a sequential mutation introduced a lysine at position +3 to the serine to result in a full Cdk1 phosphorylation consensus ([S/T]-P-x-[K/R]) (Fig. 1B class 1). In addition, if an [S/T]-x-x-[K/R] sequence is present in a non-coiled coil region of Cin8, a proline is introduced at position +1 to the serine/threonine, resulting in a full Cdk1 phosphorylation consensus [S/T]-P-x-[K/R] (Fig. 1B class 2). The addition of serine at position -1 to the proline and the addition of lysine at position +3 to the serine were preferred over threonine and arginine, respectively, since it results in a stronger phosphorylation consensus (Chang et al., 2007, Koivomagi et al., 2011b, Loog & Morgan, 2005). By this strategy, three of the five native sites, at positions 277, 285, and 493, were converted to [S-P] sites although position 285 originally contained a threonine followed by a proline [T-P]. The two native sites outside of the motor domain, at positions 736 and 1010, were not sampled since they were previously shown to have no phosphoregulation properties (Avunie-Masala et al., 2011). Two sites of class 2, at positions 712 and 271, were not sampled due to technical reasons. Overall, 29 new Cdk1 sites were generated. The novel sites were examined by live cell imaging for possible anaphase spindle-detachment phenotypes (Avunie-Masala et al., 2011, Goldstein et al., 2017). Mutants that were viable at 26°C and exhibited no detachment of Cin8 from the spindle were categorized as Cin8-5A-like mutants, to indicate they are not phosphoregulated similarly to wt Cin8 (colored orange in Fig. 3). Typically, Cin8-5A-like mutants were viable at all temperatures with five exceptions: (a) Cin8syn39, which exhibits slight temperature sensitivity at 35°C; (b) Cin8syn99 partial and full variants, which exhibit temperature sensitivity, which is partially rescued by the control counterparts; (c) Cin8syn103 and Cin8syn99 are exhibiting reduced viability at 33°C and 35°C; (d) Cin8syn700, which exhibits reduced viability of the full variant but not of the partial variant; and (e) Cin8syn881, which exhibits temperature sensitivity at 33°C and 35°C. Since these variants did not detach from the spindles, they were classified as Cin8-5A-like variants and not examined further by quantitative analyses.
Mutants that exhibited detachment from the spindle were targeted as possible Cdk1 phosphoregulation sites and additional control mutants were generated at these positions to determine if the observed phenotypes were phosphorylation-dependent. These control mutants contained a phosphodeficient alanine replacing the serine or threonine within the Cdk1 consensus sequence. In addition, the variants were examined by yeast viability test to assess their ability to function in cells as a sole source of kinesin-5 activity (see below). Eight of the variants that exhibited detachment from the spindles at anaphase were not viable at 26°C as full Cdk1 sites (colored in black in Fig. 3). One site in this category, at position 134, was non-viable as a partial Cdk1 phosphorylation consensus of its control variant (Cin8syn134pc) partially rescues the phenotype. Since the full variant at this position (Cin8syn134f) is non-viable, this site was not examined further by quantitative analysis. Another exception is at position 194, in which the partial variant Cin8syn194p was viable at 26°C and 33°C but temperature sensitive at 35°C, with no detachments from the spindle at 26°C. However, its full counterpart was non-viable at 26°C, which was not rescued by the control mutant and therefore was categorized as a non-viable variant.
Variants that exhibited phosphorylation-dependent spindle localization phenotypes, at positions S148 and S276, were further investigated by quantitative analysis of their distribution along the spindle and detachments.
Yeast viability test
Viability test of cells expressing the Cin8nat and Cin8syn variants as a sole source of kinesin-5 function was performed as previously described (Avunie-Masala et al., 2011, Duselder et al., 2015, Goldstein et al., 2017). Yeast strains used for this assay were deleted for their chromosomal copies of CIN8 and KIP1 and contained an endogenic recessive cycloheximide resistance gene (cin8Δkip1Δcyhr). These cells were supplemented with a plasmid (pMA1208) encoding for wt Cin8 and a wt cycloheximide sensitivity gene. After transformation with a plasmid encoding the Cin8 variant of interest, the initial pMA1208 plasmid was shuffled out by growth on YPD medium containing 7.5 μg/mL cycloheximide for 3 to 4 days, at 26°C and elevated temperatures, in a serial dilution (1:1, 1:10, 1:102, 1:103, 1:104) starting with 0.2 O.Dλ=600nm (approximately 2.55×106 cells/ml).
In vitro phosphorylation assay
In vitro phosphorylation assays were performed as previously described (Avunie-Masala et al., 2011, Goldstein et al., 2017). In brief, bacterially expressed Cin8 (590)-TEV-EGFP-6His variants were purified using standard nickel affinity chromatography, eluted with 300mM imidazole that was subsequently removed using Zeba Spin Desalting Columns 40K (Thermo Scientific). For a phosphorylation assay, equal concentrations of Cin8 variants were mixed with TAP-purified Clb2-Cdk1-Cks1 complex in kinase assay mixture [50mM HEPES, pH 7.4, 150mM NaCl, 5mM MgCl2, 8% glycerol, 0.2 mg/ml BSA, 500nM Cks1, and 500μM ATP (with added γ-32P-ATP (PerkinElmer)]. Reactions were stopped after 10 and 20 min with SDS-PAGE sample buffer and proteins were separated by SDS-PAGE. Gels were stained with Coomassie Brilliant Blue (CBB) R-250 (Sigma) and incorporation of 32P into the proteins was visualized by autoradiography.
Live cell imaging
The S. cerevisiae strains and plasmids used in this study are described in SI Tables S1, S2, and S3. Live cell imaging was performed as previously described (Avunie-Masala et al., 2011, Gerson-Gurwitz et al., 2009, Movshovich et al., 2008). Images were acquired using a Zeiss Axiovert 200M-based microscope setup equipped with a cooled CCD Andor Neo sCMOS camera. Images of Z stacks of eleven planes were obtained in three channels with 0.5 μm separation. Time-lapse images were obtained using a Zeiss Axiovert 200M-based Nipkow spinning-disc confocal microscope (UltraView ESR, Perkin Elmer, UK) with an EMCCD camera. Z stacks of 32-36 slices with 0.2 μm separation were acquired at one-minute intervals for 70 minutes. Data analysis was performed using MetaMorph (MDS Analytical Technologies) and open source ImageJ software. Spindle elongation rates were determined as previously described (Avunie-Masala et al., 2011).
Fluorescence intensity distribution of Cin8 along the spindle
Line scan analysis along the spindle was employed to quantify the distribution of 3GFP-tagged Cin8 variants, as previously described (Fridman et al., 2013, Goldstein et al., 2017). The fluorescence intensity profile was determined along a line tracing the spindle from mother to bud. The background signal was calculated by averaging the intensity outside the nucleus and that value was subtracted from the fluorescence intensity measured at each point. Finally, intensity was interpolated and divided into 100 segments of equal length using Origin software (OriginLab). Normalization of the Cin8-3GFP fluorescent signals was performed by dividing the intensity at each point by the total Cin8-3GFP fluorescence intensity at each spindle (Figs. 2C and 4D). The average intensity was calculated for each length-point for 10-20 cells.
Cin8 detachment from the spindle determined from fluorescence intensity perpendicular to the spindle
Line scan analysis perpendicular to the spindle was performed to quantify the detachment of Cin8 variants from the spindle during anaphase as previously described (Goldstein et al., 2017). First, the Cin8-3GFP fluorescence profile was measured along a line of 40 pixels (5.12 μm) drawn perpendicular to the spindle. The center of this line was set 2 pixels towards the midzone from a point of highest intensity, usually near the SPB (Figs. 2D). Eleven outer pixels on each side were assigned as background (Fig. 2D, dashed dark grey). The intensity of 5 pixels in the center of the line was assigned as the intensity near the SPB or the spindle and was not considered in calculating Cin8 detachment (Fig. 2D, dashed light grey). The intensities of pixels 12-18 and 23-29 were considered as the intensity resulting from Cin8-3GFP being localized in the nucleus due to its detachment from the spindle (Fig. 2D, green). Following subtraction of the background, the intensity along the line perpendicular to the spindle was normalized to the highest intensity value, usually observed near the SPB. Then, the intensity of Cin8-3GFP detached from the spindle (pixels 12-18 and 23-29) was averaged for each cell. Finally, the average Cin8-GFP intensity perpendicular to the spindles of 6-7 μm, was calculated in all cells expressing the same variant (10-20 cells – one exception is Cin8syn148pc in which only four cells were analyzed). Cin8 detachment from the spindle was analyzed as a function of spindle length (Figs. 2E and 4B) and as a function of time, based on time-lapse images as in Fig. EV1 (Figs. EV3 and EV5 right).
Statistical analysis was done using post-ANOVA all-pairwise comparison performed with Tukey procedure using OriginLab software.
Multiple sequence alignments (MSAs)
MSAs presented in Fig. 8 were conducted according to a phylogenetic tree as in (Wojcik et al., 2013). Strains chosen (as presented from top to bottom in Fig. 8): Cin8 and Kip1 of Saccharomyces cerevisiae, Ashbya gossypii (Eremothecium gossypii), Candida guilliermondii (Meyerozyma guilliermondii), Scheffersomyces stipites, Lodderomyces elongisporus, Candida orthopsilosis, Schizosaccharomyces japonicus, Schizosaccharomyces pombe, Schizosaccharomyces cryophilus, Schizosaccharomyces octosporus, and Drosophila melanogaster. The MSA was calculated by MUSCLE algorithm via UGENE program. Color coded by percentage identity with a 30% threshold.
Stability assay at room and elevated temperatures
Yeast strains were grown in 26°C and 33.5°C for five hours before crude protein extraction with 0.1M NaOH. Cell lysate was diluted 1:1 with SDS-PAGE sample-buffer. The samples were separated by SDS-PAGE and blotted by Western blot to a PDVF membrane and developed with α-GFP HRP conjugated anti-body (Santa Cruz GFP Antibody (B-2): sc-9996).
Author contributions
A.G. planned and performed experiments, analyzed data, designed figures, wrote and revised the manuscript. D.G. performed experiments. E.V. performed experiments. M.L. supervised experiments and revised the manuscript. L.J.H. conceived the study, planned experiments, wrote and revised the manuscript. L.G. conceived the study, planned experiments, supervised experiments, wrote and revised the manuscript.
Conflict of interest
The authors declare that they have no conflict of interest.
Acknowledgements
We thank Kumar-Singh Sudhir, Pandey Himanshu, Siegler Nurit, and Popov Mary from Ben-Gurion University in the Negev, Beer-Sheva, Israel, for critical reading of this manuscript. This work was supported in part by the Israel Science Foundation (ISF) (grant 165/13) awarded to L.G.; The United States - Israel Binational Science Foundation grant (BSF-2015851), awarded to L.G.; the William Bowes Foundation and Vilcek Foundation grant awarded to L.J.H.; The ERC Consolidator Grant Nr 649124, Phosphoprocessors, awarded to M.L.; and Estonian Science Agency grant Nr. IUT2-21, awarded to M.L.