A functional analysis of TOEFAZ1 uncovers protein domains essential for cytokinesis in Trypanosoma brucei

The protozoan parasite Trypanosoma brucei is the causative agent of African sleeping sickness in humans and nagana in cattle, both of which are sources of substantial hardship in sub-Saharan Africa (Fèvre et al., 2008). T. brucei is an obligate extracellular pathogen that has evolved a unique cell morphology as a means to survive within its mammalian and insect hosts (Gull, 1999). Trypanosomes have a long, tapered cell body shaped by a subpellicular microtubule array that underlies the cell surface (Fig. 1A) (Anderson and Ellis, 1965; Vickerman, 1962). A single flagellum nucleates near the posterior end of the cell and is attached along the cell body, extending towards the cell anterior (Langousis and Hill, 2014). The attachment and positioning of the flagellum are essential for proper motility and evasion of the host immune response (Engstler et al., 2007; Heddergott et al., 2012). Along with replicating and partitioning cellular organelles, cell division in T. brucei must faithfully transmit this complex cellular morphology to ensure the viability of daughter cells (Farr and Gull, 2012; Wheeler et al., 2013). Unlike many other organisms, trypanosomes maintain their high degree of polarity throughout cell division, which puts unique constraints on this process.

The cytoskeletal components associated with the flagellum play an important role in maintaining cell morphology (Fig. 1A). A single basal body nucleates the flagellum from the base of the cell surface invagination known as the flagellar pocket (FP), which is the only compartment in trypanosomes that is competent for endo- or exo-cytosis. The top of the FP is tightly apposed to the flagellar membrane by a series of cytoskeletal structures including the flagellar pocket collar (FPC), the hook complex and the centrin arm (which together were previously known as the ‘bilobe structure’) (Bonhivers et al., 2008; He et al., 2005; Morriswood, 2015; Morriswood and Schmidt, 2015). Once the flagellum emerges from the FP it is attached to the cell surface by the flagellum attachment zone (FAZ), which is made up of a series of junctional complexes similar to desmosomes, and a set of four unique microtubules that are thought to have opposing polarity to the microtubules present in the subpellicular array (Sunter and Gull, 2016; Robinson et al., 1995). These microtubules, known as the microtubule quartet (MtQ), nucleate near the basal body, wrap around the side of the FP, traverse the hook complex and FPC, and then extend beside the FAZ filament as part of the subpellicular array (Lacomble et al., 2009).

Cytokinesis, the late cell cycle event that segregates duplicated organelles and cytoplasm to produce daughter cells, proceeds via a unique mechanism in T. brucei (Fig. 1B). In most eukaryotes, the motor protein nonmuscle myosin II and actin filaments arranged in an actomyosin ring provide the force necessary to separate daughter cells (Fenix et al., 2016; Glotzer, 2005; Laplante et al., 2016). Trypanosomes appear to lack a nonmuscle myosin II homolog and actin is not essential in the insect-resident (procyclic) form of the parasite, suggesting that the protein does not participate in a core conserved function such as cell division (García-Salcedo et al., 2004). Instead, the trypanosome cleavage furrow initiates from the anterior tip of the new FAZ and progresses towards the cell posterior along an indentation in the cell membrane known as the cleavage furrow fold (Fig. 1B, inset). The posterior end of the cell contains plus-end microtubule binding proteins such as XMAP215 and is a major microtubule nucleation site for the subpellicular array (Wheeler et al., 2013). A new nucleation site forms during division to delimit the subpellicular microtubule arrays of the two daughter cells (Robinson et al., 1995; Sherwin et al., 1987). The assembly of the nascent cell posterior for the cell containing the old flagellum and FAZ is a separate event from cleavage furrow ingression, although the two events happen in concert to complete cytokinesis.

While cytokinesis in trypanosomes has been well described morphologically, little is known about the specific proteins that are involved. Owing to a lack of conserved cytokinetic components, this process is likely to be mechanistically unique. Removing essential FAZ components or perturbing cytoskeletal structures that are necessary for new FAZ assembly causes cleavage furrow misplacement or an outright block in cytokinesis (He et al., 2005; Sunter et al., 2015; Vaughan et al., 2008). Depleting proteins involved in flagellar assembly, such as intraflagellar transport (IFT) components, yields shorter daughter cells, most likely due to misplaced furrow ingression caused by a shortened new FAZ (Kohl et al., 2003). Recently, the polo-like kinase homolog in trypanosomes, named T. brucei (Tb)PLK, was shown to localize to the anterior tip of the new FAZ during cell division; inhibiting TbPLK activity early in the cell cycle blocks cytokinesis, suggesting that the kinase may have some mechanistic role in this process (Hammarton et al., 2007; Ikeda and de Graffenried, 2012; Li et al., 2010; Lozano-Núñez et al., 2013; Yu et al., 2012).

We recently conducted a screen that employed phosphoproteomics and proximity-dependent biotin identification (BioID) to identify novel TbPLK interactors and substrates (McAllaster et al., 2015). One of our candidate proteins, which we named TOEFAZ1 (for ‘Tip of the Extending FAZ protein 1’; Tb927.11.15800), colocalizes with TbPLK at the tip of the new FAZ during cell division (McAllaster et al., 2015; Zhou et al., 2016a). Depletion of TOEFAZ1 blocks recruitment of TbPLK to the tip of the new FAZ and causes substantial defects in cytokinesis, initially producing many anucleate and multinucleate cells followed by a block in cell division. TOEFAZ1 is the first protein identified in T. brucei that appears to have a specific role in cytokinesis, as its depletion does not cause additional defects in new FAZ or cytoskeletal assembly. Subsequent work by others chose an alternative name for the protein (Cytokinetic Initiation Factor 1; CIF1) and proposed that in its absence an alternative ‘back-up’ cytokinetic mechanism could produce a furrow that ingresses from posterior to anterior to generate viable daughter cells (Zhou et al., 2016a).

TOEFAZ1 is conserved among kinetoplastids but lacks clear homologs in other organisms such as yeast and mammals. The protein contains three distinct domains: an N-terminal α-helical domain, a central intrinsically disordered protein (IDP) domain and a C-terminal zinc finger (ZnF) domain (Mezulis et al., 2015; Xu et al., 2014). We sought to determine how the protein functions by dissecting the contribution of each domain through both overexpression and RNAi complementation strategies. We show that the N-terminal domain may function as a protein interaction domain that interacts with other FAZ proteins to localize TOEFAZ1. The IDP domain controls the timing of protein expression during the cell cycle and TbPLK retention at the tip of the new FAZ, but is not essential. The C-terminal ZnF domain is a homo-oligomerization domain that is necessary but not sufficient for TOEFAZ1 recruitment to the tip of the new FAZ. We further show that cells lacking TOEFAZ1 rapidly develop aberrant DNA states that are not consistent with further productive cell divisions, calling into question the viability of alternative cytokinetic pathways.

TOEFAZ1 localizes to the tip of the extending FAZ and subsequently to the ingressing cleavage furrow. To further characterize its localization, we studied the distribution of the protein compared to the signal from two monoclonal antibodies: L3B2, which detects the FAZ filament protein FAZ1, and 1B41, which recognizes a subset of β-tubulin that is present in the FAZ, most likely in the MtQ (Gallo et al., 1988; Kohl et al., 1999; Vaughan et al., 2008). We generated cells where TOEFAZ1 was endogenously tagged at its N-terminus with three copies of the HA epitope and stained them with 1B41, anti-FAZ1 and anti-HA. Fig. 2A shows a panel of cells at different cell cycle stages. Trypanosomes maintain a mitochondrial DNA aggregate known as the kinetoplast as well as a nucleus; tracking the distribution of these DNA-containing organelles allows the cell cycle state to be easily distinguished. During the cell cycle, a cell that contains one nucleus and one kinetoplast (1N1K) first duplicates its kinetoplast (1N2K), and then its nucleus (2N2K), before cytokinesis produces two daughter cells with a single copy of each organelle (Fig. 1B). Early in the cell cycle, the 1B41 and FAZ1 labeling patterns are in close proximity and similar in length and cells lack TOEFAZ1 labeling (Fig. 2Ai). TOEFAZ1 appears just as the new FAZ can be resolved, which is most easily seen by the bright anterior 1B41 labeling on the new FAZ tip (Fig. 2Aii, asterisk). As the new FAZ continues to elongate, the 1B41 signal extends slightly beyond the most anterior labeling visible for FAZ1 and comes in close proximity to the old FAZ. In contrast, there is a distinct gap in the FAZ1 labeling between the new and old FAZ (Fig. 2Aiii,iv). The TOEFAZ1 signal accumulates on the bright puncta at the anterior end of the 1B41 signal on the new FAZ, which matches the gap between FAZ1 signals (Fig. 2Aiii,iv). TOEFAZ1 is present at the tip of the new FAZ until cleavage furrow ingression, when the protein localizes along the furrow (Fig. 2Av) (McAllaster et al., 2015; Zhou et al., 2016a).

We compared the localization of TOEFAZ1 and TbPLK in relation to the FAZ marker 1B41 to mark the migration pattern of the two proteins. At the early stages of the cell cycle, cells do not express TbPLK or TOEFAZ1 (Fig. 2Bi). TbPLK initially appears on the basal body, hook complex and centrin arm, all of which are proximal to the flagellar pocket (Fig. 2Bii,iii, arrowhead). TOEFAZ1 first localizes to the anterior end of the TbPLK signal, which is most likely the centrin arm and/or hook complex (Fig. 2Biii, empty arrowhead). This also coincides with the bright puncta of 1B41 signal. As the new FAZ extends, TbPLK and TOEFAZ1 are present on the tip of the extending structure, with TbPLK also on the flagella connector (FC) that links the tip of the new flagellum to the old flagellum (Fig. 2Biv, asterisk) (McAllaster et al., 2015; Moreira-Leite et al., 2001; Zhou et al., 2016a). As the cell cycle progresses, the TbPLK signal at the tip of the FAZ begins to decline, while the TOEFAZ1 signal remains constant (Fig. 2Bv). TbPLK is absent from the tip of the new FAZ once cleavage furrow ingression initiates (Fig. 2Bvi).

TOEFAZ1 has three distinct domains: an N-terminal α-helical domain (amino acids 1–319), a central IDP domain (aa 320-649), and a C-terminal ZnF domain containing two zinc finger motifs (amino acids 650–790) (Fig. 3A). As a first test of protein function, we expressed each domain individually using the doxycycline-inducible vector pLEW100 (Fig. 3A) (Wirtz et al., 1999). Each domain was tagged at its N-terminus with three copies of the Ty1 epitope tag, and induction of domain expression was assessed by western blotting with anti-Ty1 antibody (Fig. 3B). None of the expression constructs caused growth defects, showing that they lacked dominant-negative properties (Fig. S1). Full-length TOEFAZ1 migrates at a substantially higher molecular mass than the expected 93 kDa, most likely due to posttranslational modifications (Fig. 3B) (McAllaster et al., 2015; Zhou et al., 2016a). Similarly, all three domains appear at a higher molecular mass than expected (Fig. 3B). This is likely due to phosphorylation; nine potential phosphorylation sites have been detected in the N-terminal domain and at least 21 in the IDP domain (Urbaniak et al., 2012, 2013).

The localization of each domain was assessed in immunofluorescence microscopy experiments. The N-terminal α-helical domain was expressed at all cell cycle stages and primarily had a diffuse cytoplasmic localization, although some of the protein formed ring-like structures that were frequently in close proximity to the kinetoplast (Fig. 3C, asterisks). A portion of the protein also localized to the full length of the FAZ (Fig. 3C, arrowheads). The ring structures and FAZ labeling were susceptible to detergent extraction, suggesting that the N-terminal domain is not tightly associated with the cytoskeleton. The IDP domain localized exclusively to the cytoplasm, but was not expressed in the majority of 1N1K cells early in the cell cycle. Expression of the IDP was evident in cells where the kinetoplast had begun to divide and in 1N2K and 2N2K cells (Fig. 3D), which corresponds to the cell cycle window when full-length TOEFAZ1 is normally expressed (Fig. 2A). Considering that the IDP domain is being overexpressed from a vector that lacks the native 5′ and 3′ UTRs for TOEFAZ1, it is unlikely that translational control could be responsible for the observed cell cycle-restricted expression. The C-terminal ZnF domain localized primarily to the tip of the new FAZ (Fig. 3E), but the domain was also present along the length of the old and new FAZ, suggesting that it was not able to maintain as tight a localization with the FAZ tip structure as does the full-length protein (Fig. 3E, asterisk). The protein localized to the cytoplasm in cells that were not dividing.

To better understand the function of each domain within TOEFAZ1, we devised an RNAi complementation strategy that allows us to generate cells expressing only truncated forms of the protein that lack individual domains (Fig. 4A). We assembled an RNAi hairpin against the native TOEFAZ1 DNA sequence (nt 578–1090) in the doxycycline-inducible vector pTrypSon by using Gibson assembly (McAllaster et al., 2016). The plasmid was integrated into cells carrying the necessary machinery for doxycycline induction (the cell line 29-13), and then one allele of TOEFAZ1 was endogenously tagged with a triple-Ty1 tag at its N-terminus. This cell line, carrying the TOEFAZ1 RNAi hairpin and a natively coded Ty1-tagged wild-type (WT) TOEFAZ1 allele, was the base cell line used for all further modifications. We then recoded a segment of the TOEFAZ1 DNA sequence (nt 4–1096) so that it used different codons from the native gene but still encoded the same amino acids, making it insensitive to the RNAi hairpin that targets the native coding sequence (Horn, 2008). Thus, induction of RNAi should selectively deplete the mRNA produced from the native allele. The recoded sequence was included in an endogenous tagging construct that also introduces a triple-HA tag onto the N-terminus of TOEFAZ1. Because the two versions of the gene are differentially epitope tagged, we can readily distinguish the native from recoded alleles.

As a proof of principle, we introduced full-length TOEFAZ1 with the recoded segment tagged with HA into our base cell line. Induction of RNAi by addition of doxycycline effectively depleted the native copy of the protein, as assessed by anti-Ty1 immunofluorescence and western blotting, but the protein produced from the recoded HA-tagged allele localized effectively to the tip of the new FAZ and resisted depletion (Fig. S2A,B). Cells expressing only the recoded allele grew at the same rate as cells carrying a native copy of the gene, which shows that the recoded allele is fully functional and that our RNAi hairpin has no off-target effects (Fig. S2C).

With the efficacy of the RNAi complementation strategy established, we introduced RNAi-insensitive mutants of TOEFAZ1 into our base cell line to observe the effect of removing individual domains of the protein. We confirmed the correct integration of each of these domain mutant cell lines by western blotting, loci PCR and sequencing of the modified locus (Fig. S3). Our first mutant, termed ΔN-terminus, removed the N-terminal α-helical domain from TOEFAZ1. Immunofluorescence with anti-Ty1 and anti-HA antibodies showed that ΔN-terminus TOEFAZ1 mutant colocalized with WT TOEFAZ1 (Fig. 4B). Depletion of the WT TOEFAZ1 by RNAi caused the rapid loss of Ty1 signal as determined by western blotting (Fig. S3A) and immunofluorescence (Fig. 4B) analyses, while the HA signal, which shows the mutant RNAi-insensitive protein, remained detectable in western blots. In the absence of full-length TOEFAZ1, the ΔN-terminus mutant was no longer able to localize to the tip of the new FAZ, suggesting that the mutant relies on the presence of the full-length protein for localization (Fig. 4B). After 1 day of WT protein depletion, cells expressing only the ΔN-terminus TOEFAZ1 display significant cell division defects, including slow growth and an increase in the number of 2N2K cells (Fig. 4C,E). After 2 days, DNA counting of cells expressing only the ΔN-terminus TOEFAZ1 showed increasing cytokinetic delay, with 1N1K cells making up only 24% of the population and an elevated number of 1N2K and 2N2K cells (21% and 20%, respectively; Fig. 4D,E). Aberrant cells containing no nuclei (0N1K, 0N2K) or multinucleate cells also appeared. At 4 days after depletion of the WT protein, 1N1K cells comprise only 14% of the population, while the anucleate and multinucleate cells make up over 62% of the population. We also noted that nearly half of the 2N2K cells in the experimental condition had new flagella tips that were detached after 2 days of RNAi induction, which suggested that the FC had resolved before cleavage furrow ingression, as we previously saw in the case of full-length TOEFAZ1 depletion (Fig. 4F) (McAllaster et al., 2015). This indicates that cleavage furrow ingression is likely delayed in cells expressing only the ΔN-terminus construct, although it is not formally known if FC resolution is independent of cleavage furrow ingression (Varga et al., 2017).

We next generated a cell line that contained a recoded mutant TOEFAZ1 allele that lacked the IDP domain. Depletion of the WT allele proceeded with similar kinetics to the cell line containing the ΔN-terminus TOEFAZ1 construct (Fig. S3B). The mutant protein was able to localize to the tip of the new FAZ in the presence or absence of the full-length TOEFAZ1 (Fig. 5A), although the ΔIDP protein had an additional labeling pattern in some cells just to the posterior side of the extending FAZ (Fig. 5A, asterisk). Surprisingly, the IDP domain of TOEFAZ1 does not appear to be essential for cell viability, as cells expressing only the ΔIDP mutant had just a 10% deficit in the doubling rate (Fig. 5B). After 3 days of RNAi induction, ΔIDP mutant cells had slightly lower levels of 1N1K cells compared to controls (62% versus 80% 1N1Ks, respectively) and a concomitant increase in 2N2K cells; there was also a small but noticeable population of zoid cells, which are cells that contain no nuclei and one kinetoplast (Fig. 5C). These results argue that while the ΔIDP TOEFAZ1 mutant can support near-WT levels of growth, there is still a small delay or defect in cytokinesis. This cytokinetic defect is further evidenced by the premature release of the FC in almost 40% of the ΔIDP 2N2K cell population (Fig. 5D). We also noticed additional TOEFAZ1 labeling in the posterior of 1N1K cells, which most likely corresponds to the point at which cleavage furrow ingression terminates prior to nascent posterior end formation (Fig. 5E, arrowhead).

Finally, we generated cells lacking the TOEFAZ1 C-terminal ZnF domain using our complemented RNAi strategy. Induction of RNAi with doxycycline led to a rapid depletion of the full-length native coding protein, while the HA-tagged recoded protein persisted (Fig. S3C). This mutant was unable to localize to the tip of the new FAZ in the presence or absence of the full-length protein (Fig. 6A). Cells expressing only the ΔZnF version of TOEFAZ1 began to show a rapid decline in the rate of cell division, with minimal growth observed after 4 days of RNAi depletion of the full-length protein (Fig. 6B). Assessment of cell cycle stage and morphology showed a similar pattern to what was observed for cells lacking the N-terminal α-helical domain. There was an initial increase in 2N2K cells, followed by the appearance of anucleate and multinucleate cells, which make up 63% of the population after 4 days of TOEFAZ1 RNAi induction (Fig. 6C). Over 40% of 2N2K cells had prematurely detached new flagella tips, suggesting a delay or defect in cleavage furrow ingression (Fig. 6D). This result suggests that the ZnF domain is absolutely required for TOEFAZ1 localization to the new FAZ tip and for its proper function.

The ability of the TOEFAZ1 mutant lacking the IDP domain to support cell growth raised the possibility that the N-terminal domain and the C-terminal ZnF domain could function together in trans, either by forming a complex or by performing their functions independently. To test this possibility, we altered our RNAi complementation strategy to allow for individual expression of the two TOEFAZ1 domains (Fig. S4A). We tagged one of the native TOEFAZ1 alleles at its N-terminus with the GFP variant mClover3 so that we could use the HA and Ty1 epitope tags for the individual domains (Bajar et al., 2016). We modified the second TOEFAZ1 allele so that it contained a recoded version of the N-terminal domain tagged with HA and the ZnF domain tagged with Ty1, each with its own start and stop codons. The two domains were separated by an intergenic sequence from the tubulin locus, which allows the two segments to be transcribed from a single promoter but to be processed into two unique mRNAs (Imboden et al., 1987). The three variants of the protein could be detected individually through western blotting with anti-HA antibody, anti-Ty1 antibody, and antibodies against GFP. Depletion of the full-length mClover3-tagged TOEFAZ1 by RNAi proceeded similarly to depletion of the Ty1-tagged protein in our conventional RNAi complementation approach, leaving cells expressing only the separate N-terminal and ZnF domains (Fig. S4B). In control cells, the Ty1-tagged ZnF domain and full-length mClover3-TOEFAZ1 proteins were able to localize to the new FAZ tip, but no signal was evident for the HA-tagged N-terminal domain (Fig. S4C), although we could observe protein expression by anti-HA antibody western blotting. Depletion of the full-length TOEFAZ1 caused the mislocalization of the ZnF domain (Fig. S4C). The loss of WT TOEFAZ1 caused a rapid slowdown in growth, with kinetics similar to the ΔN-terminus and ΔZnF mutants in previous experiments (Fig. S4D). This result suggests that these domains cannot operate in trans and need to be physically linked to function.

The cell line expressing the individual domains of TOEFAZ1 provided an opportunity to test the oligomerization characteristics of the ZnF domain, as the Ty1-tagged ZnF domain localizes to the new FAZ tip only when full-length TOEFAZ1 is present (Fig. S4C). We added the eight-atom crosslinker disuccinimidyl suberate (DSS) or a vehicle control (DMSO) to lysates and determined the oligomerization state of the Ty1-tagged ZnF domain (Fig. S5A). In lysates treated with vehicle control, the ZnF domain migrates at its expected size (19 kDa). In the presence of DSS, we observed higher order species that migrated at the expected molecular mass of a ZnF tetramer and a second higher-order species. We also tested whether the ZnF domain was able to oligomerize with full-length TOEFAZ1 (Fig. S5B). We treated live cells with DSS or DMSO, and then subjected them to lysis and immunoprecipitation with GFP-Trap resin to capture full-length mClover3-tagged TOEFAZ1. In the presence of DSS, we were able to capture the Ty1-tagged ZnF domain with GFP-Trap resin, showing that the domain can interact with full-length TOEFAZ1. Blotting the immunoprecipitated fractions with anti-GFP also showed that the full-length TOEFAZ1 formed an oligomer in the presence of DSS.

To address the differences in localization observed when the domains are overexpressed, we generated cells expressing the individual N-terminal and C-terminal domains of TOEFAZ1 at endogenous levels. We integrated HA-tagged RNAi-resistant versions of the N-terminal α-helical domain and the C-terminal ZnF domain into the base cell line for our RNAi complementation approach (Fig. S6). After confirming correct integration, we checked the localization pattern of the individual domains (Fig. S6A,B). The N-terminal α-helical domain on its own was not able to localize to the tip of the new FAZ either in the presence or absence of the full-length protein (Fig. S6A), as was the case with the ΔZnF mutant and trans cell line (Fig. 6A; Fig. S4C), suggesting that the FAZ labeling and ring structures we observed previously arise only when the protein is expressed at high levels (Fig. 3C). The ZnF on its own was able to localize to the new FAZ tip as long as full-length TOEFAZ1 was present in cells (Fig. S6B), as is the case with the ΔN-terminal mutant and the trans cell line (Fig. 4B; Fig. S4C). When the protein was localized, it showed a high degree of colocalization with native full-length TOEFAZ1, which suggests that the broader expression pattern along the length of the FAZ observed previously (Fig. 3E) is also due to overexpression.

We next assessed the localization of TbPLK in cells expressing only the domain deletion mutants of TOEFAZ1. We classified TbPLK localization into four categories based upon its localization pattern throughout the cell cycle (Fig. 2B): no TbPLK expression (none); localization to the basal body, centrin arm, and hook complex (pocket); FAZ exclusively or FAZ and FC (FAZ/FC); or exclusively to the FC (FC). We first looked at cells expressing only the ΔIDP TOEFAZ1 mutant because it maintained localization to the FAZ tip and these cells showed only a slight growth defect, which made it possible that TbPLK localization was not altered. In control conditions where the full-length TOEFAZ1 was present, TbPLK localized to the pocket region early in the cell cycle, then was present on the tip of the new FAZ and FC as those structures migrate during later stages in cell division (Fig. 7A,B). In cells expressing only the ΔIDP mutant after 5 days of RNAi induction, TbPLK recruitment to the pocket region was the same as in control cells. However, recruitment of the kinase to the tip of the new FAZ was blocked, similar to what is observed when full-length TOEFAZ1 is depleted from cells (McAllaster et al., 2015; Zhou et al., 2016a). This block is remarkable because the ΔIDP mutant cells grow at nearly the same rate as control cells, which strongly argues that TbPLK localization to the tip of the new FAZ at later cell cycle stages is dispensable for growth and more specifically for cytokinesis.

We also determined TbPLK localization in cells expressing only the ΔN-terminus or ΔZnF mutants of TOEFAZ1 (Fig. S7). Because both of these mutants were unable to localize to the new FAZ tip on their own, it was likely that TbPLK recruitment would be blocked, as we saw previously with full-length TOEFAZ1 depletion (McAllaster et al., 2015). As expected, exclusive expression of either mutant led to a near-total block of TbPLK recruitment to the new FAZ tip, while the kinase was still able to localize to the pocket region early in the cell cycle.

We decided to determine whether the IDP domain was absolutely necessary for interaction with TbPLK, as TbPLK no longer localizes to the new FAZ tip at later cell cycle stages in cells expressing only the ΔIDP protein, even though the ΔIDP protein is present at this location. Mapping this interaction was complicated by the fact that TOEFAZ1 forms oligomers (Fig. S5B), so it was essential to perform the experiment in the cells expressing only the ΔIDP mutant. We induced RNAi against the WT protein for 3 days in the ΔIDP cell line and then performed anti-TbPLK antibody immunoprecipitations. We blotted the eluted proteins from both immunoprecipitations with anti-Ty1 and anti-HA antibody to identify the WT full-length and ΔIDP TOEFAZ1, respectively (Fig. 7C). We were able to immunoprecipitate both full-length TOEFAZ1 and ΔIDP TOEFAZ1 under control conditions. Interestingly, in cells expressing only the ΔIDP TOEFAZ1, TbPLK precipitation was still able to capture ΔIDP TOEFAZ1. Immunofluorescence examination using antibodies to detect the FAZ, TbPLK and the HA-tagged ΔIDP protein in cells only expressing the ΔIDP mutant showed that TbPLK colocalized with the ΔIDP protein at the early stages of cell division, when TbPLK localizes to the pocket region just as the new FAZ is being formed (Fig. 7D). These experiments demonstrate that the IDP domain is not essential for TbPLK interaction, but is necessary to maintain the kinase on the new FAZ tip at later points in the cell cycle.

Several of the TOEFAZ1 domain mutant cell lines show significant decreases in growth and aberrant DNA content in the absence of full-length protein. While this is consistent with our previous work, it contrasts with other reports, where a diminished but constant growth rate is observed in the absence of TOEFAZ1 (Zhou et al., 2016a). It is possible that the isolated domains that remain in our experiments exert some additional dominant-negative effect that is lacking when TOEFAZ1 is completely removed, but we did not observe any detrimental phenotype from overexpression of individual domains (Fig. S1). To address this discrepancy, we tagged the second TOEFAZ1 allele in our base cell line with a triple-HA tag but did not alter the codon usage, retaining its sensitivity to our TOEFAZ1 RNAi hairpin. We referred to this cell line as our double-tag line and used it in conjunction with our previously constructed cell line containing an HA-tagged TOEFAZ1 allele resistant to RNAi (hereafter referred to as the add-back line) to probe the effect of complete protein depletion (Fig. 8A).

We induced TOEFAZ1 RNAi in the add-back and double-tag cell lines and monitored parasite growth, DNA state, and morphology over the course of 8 days. The add-back cell line grew at the same rate both in the presence or absence of TOEFAZ1 RNAi induction, as in our previous experiment, showing that our RNAi is specific (Fig. 8B). In the double-tag cell line, induction of TOEFAZ1 RNAi caused a rapid decrease in cell growth. Over the 8 days of the experiment the growth of TOEFAZ1-depleted cells continued to decrease, and from day 6 to day 8 the cells were not able to complete a single doubling. Western blotting of lysates from both the add-back and double-tag lines showed that the RNAi-insensitive allele was preserved in the add-back line while both copies of TOEFAZ1 were depleted in the double-tag line, as expected (Fig. S8).

We focused our DNA and morphologic analysis on the double-tag cell line because it was the only cell line to show growth defects (Fig. 8B). After 1 day of depletion, cells lacking TOEFAZ1 showed a substantially diminished number of 1N1K cells and began to accumulate at the 2N2K stage, suggesting a delay in cytokinesis (Fig. 8C). After 2 days, the population of 2N2K cells decreased and anucleate cells containing only kinetoplasts became prevalent. Multinucleate cells also began to appear. This trend continued on days 3 and 4, with anucleate cells now comprising more than half the population and cells with conventional DNA states making up fewer than 20% of the total cells. Many of these conventional cells may have contained normal DNA states (i.e. 1N1K, 1N2K, and 2N2K), but showed a wide range of abnormal morphologies that were difficult to categorize. Quantifying DNA state and morphology after day 4 was not possible due to the complexity of observed morphologies and large multinucleate cells.

In this work, we have dissected the function of the three modular domains of TOEFAZ1. The N-terminal domain has low affinity for the FAZ and cytoskeleton, the IDP domain is responsible for retention of TbPLK at the new FAZ tip, and the C-terminal ZnF domain is involved in TOEFAZ1 oligomerization. TOEFAZ1 most likely functions as a scaffold for organizing and assembling different protein complexes to perform their functions at different stages of cell division. The protein is unlikely to function as a direct initiator of cytokinesis, but rather it helps to recruit and localize other factors that are responsible for this process.

The N-terminal α-helical domain has low affinity for the cytoskeletal elements to which TOEFAZ1 targets during cell division (Fig. 3C). The N-terminal domain is likely to interact with other proteins on the tip of the new FAZ, although it cannot localize when expressed at endogenous levels (Fig. S6A). Among potential N-terminal binding partners is the protein CIF2, which was identified in a BioID screen with TOEFAZ1 and is essential for recruitment of the protein to the tip of the new FAZ (Zhou et al., 2016b). It is likely that TOEFAZ1 is part of a larger complex that is responsible for the timing and positioning of cleavage furrow ingression, and that the N-terminal domain is responsible for these interactions. The N-terminal domain has two predicted coiled-coil regions that could act as a stable scaffold for protein association (Zhou et al., 2016a). The weak interaction of the isolated N-terminal domain at endogenous levels may be due to its monomeric state; since TOEFAZ1 appears to be an oligomer, the presence of additional copies of the N-terminal domain in the same complex may boost binding affinity and specificity in the full-length protein. However, we cannot exclude the possibility that the N-terminal domain mislocalizes under high expression levels, although its localization pattern shares similarities to the WT protein.

The TOEFAZ1 IDP domain is important for protein degradation at the end of cell division and for maintaining interaction with TbPLK, although neither of these functions is essential. IDP domains are unstructured and flexible domains that can adopt specific conformations upon binding to another protein, allowing them to bind to different partners (Malaney et al., 2013; Wright and Dyson, 2015). IDPs have also been thought of as flexible coils or springs that provide the motion necessary for protein function or as separation between functional domains (Young et al., 2001). IDPs are common sites for posttranslational modifications, especially phosphorylation (Bah et al., 2015). There are several potential TbPLK phosphorylation sites on TOEFAZ1, although these sites appear to be clustered in the N-terminus (McAllaster et al., 2015). Depletion of full-length TOEFAZ1 blocks recruitment of TbPLK to the tip of the new FAZ without interfering with the localization of the kinase to the pocket region or FC. Thus, our immunoprecipitation data (Fig. 7C) and work by others (McAllaster et al., 2015; Zhou et al., 2016a) suggests that TbPLK recruits TOEFAZ1 to the tip of the FAZ, and TOEFAZ1 maintains this interaction. However, small-molecule inhibition of TbPLK in synchronized cells shows that kinase activity is not necessary late in the cell cycle (Li et al., 2010; Lozano-Núñez et al., 2013). We now formally show that the presence of TbPLK at the new FAZ tip is not essential for cytokinesis to proceed, as TOEFAZ1 lacking the IDP domain is able to support near WT levels of growth without maintaining TbPLK at this location. The essential functions of TbPLK appear to be restricted to early cell cycle events, such as duplication of the FPC, basal body rotation and nucleation of the new FAZ (Ikeda and de Graffenried, 2012; Lozano-Núñez et al., 2013). It is possible that TbPLK is loaded onto the new FAZ tip during its initial assembly and remains there as a consequence of this early event without having a discrete function.

The IDP domain is heavily phosphorylated and its expression is regulated during the cell cycle, most likely via proteolysis (Fig. 3D). There are a series of potential degradation motifs in the IDP, including a D-box (amino acids 366–369), a phosphodependent degron (amino acids 198–204), and a PEST domain (amino acids 399–411) (Lindon and Pines, 2004; Al-Zain et al., 2015; Li et al., 2012). Similar motifs were recently shown to control the stability of TbPLK via a ubiquitin-mediated degradation pathway (Hu et al., 2017). TOEFAZ1 lacking the IDP domain appears to persist in cells that have completed cytokinesis in a location similar to where the cleavage furrow ends, which is likely where the protein is removed and degraded (Fig. 5E). While the IDP is not required for TOEFAZ1 function, the N-terminal and C-terminal domains are not able to work in trans, either by associating with one another or by performing their functions individually, so they must be present on the same polypeptide chain in order to function (Fig. S4). This strongly argues that the individual domains cannot interact with one another in cells and that the IDP domain acts as a physical linker to coordinate the function the N-terminal and ZnF domains of TOEFAZ1.

The ZnF domain in TOEFAZ1 plays an important role in assembling the protein into higher-order structures (Fig. S5). Zinc fingers are ubiquitous folds in eukaryotes and have expanded their early-described function as nucleic acid-binding domains to encompass many different types of interactions (Bates et al., 2008; Gamsjaeger et al., 2007). The full range of zinc fingers in trypanosomes has not been established, although the parasite contains many of the CCCH type, which are involved in binding and stabilizing mRNAs (Kramer et al., 2010). Some zinc fingers have been shown to interact exclusively with other proteins. For example, the C-terminal zinc finger domain in the transcription factor Ikaros has been shown to function as a highly specific dimerization domain that excludes closely related proteins (Laity et al., 2001; McCarty et al., 2003). We propose that the TOEFAZ1 ZnF domain plays a similar role by oligomerizing the protein. This higher-order structure now includes at least two copies of the N-terminal localization domain that can interact with other proteins to localize TOEFAZ1. The ZnF domain can recruit itself to the tip of the new FAZ, but only if full-length TOEFAZ1 is also present (Fig. S6B). This likely means that the ZnF domain does not contain localization information on its own but relies on the N-terminal domain to target it to the correct cellular location once the protein is oligomerized.

Our data suggest that TOEFAZ1 is not absolutely essential for cytokinetic furrow ingression but is required for its correct timing and positioning (Fig. 8C). The appearance of anucleate cells in the TOEFAZ1 domain mutants as well as in complete TOEFAZ1 depletion conditions shows that a cytokinetic event must occur in the absence of the protein. The lack of an increasing 2N1K population argues against zoid production from 2N2K cells, which has been seen in cells lacking TbCentrin4 (Shi et al., 2008). A possible source for the anucleate population are multinucleate cells that expel zoids due to incorrect furrow positioning. An additional possibility is premature cytokinesis in 1N2K cells. Cells that do not complete abscission, the last step of cytokinesis, have been shown to restart the cell cycle, although these cells have already undergone cleavage furrow ingression and are only connected by a small posterior segment (Wheeler et al., 2013). In parasites lacking TOEFAZ1, 2N2K cells delayed in cleavage furrow ingression began to re-express TbPLK even though they have not yet completed the previous round of cell division (Fig. S7) (McAllaster et al., 2015). This indicates that the cytoplasmic cell cycle stage in these cells has progressed beyond their corresponding cytokinetic state. If cells with an ‘advanced’ cytoplasmic state complete division after an initial delay in cleavage furrow ingression, the daughter cells would be primed to start division prematurely. We occasionally observed TOEFAZ1-depleted 1N2K cells with what appeared to be cleavage furrow folds and released FCs, suggesting that these cells were attempting to undergo cytokinesis to produce a 1N1K cell and a zoid. This constant cytoplasmic cycling combined with a delay in cleavage furrow ingression would lead to increasing asynchrony between the cell cycle events tasked with organelle duplication and those that partition them, which would result in cells with such aberrant DNA states that viable cells could no longer be produced. This mimics the decreasing viability of cell divisions that we observe in our cell counting experiments.

Others have shown that TOEFAZ1 depletion blocks cleavage furrow ingression in the anterior-posterior direction, and suggested that this triggers a novel cytokinetic pathway in which a furrow ingresses from the posterior end of the cell towards the anterior to produce viable cells (Zhou et al., 2016a). We occasionally saw cells in our TOEFAZ1-depleted lines that had multiple posteriors while lacking clear ingression from the anterior end, which could be classified as posterior ‘furrows’. However, it should be noted that nascent posterior end formation is an essential cell division event that creates a new posterior end for the cell containing the old flagellum and FAZ (Robinson et al., 1995; Wheeler et al., 2013). Taking these data together, it is more likely that the reported ‘back-up’ cytokinetic mechanism is the result of nascent posterior end formation and subsequent subpellicular microtubule remodeling, and is not the reverse ingression of a posterior cleavage furrow. As the cleavage furrow fold is still produced in TOEFAZ1-depleted cells (Zhou et al., 2016a), cells could pull apart along the fold at the point of nascent posterior end formation due to the force of flagellar beating. In essentially every published image of a ‘back-up’ cytokinetic event in trypanosomes, the cells have already begun to replicate their kinetoplasts and even their nuclei beyond the 2N2K state (Zhou et al., 2016a,b). This indicates that back-up cytokinesis is either an extremely slow process or is triggered after a prolonged delay, both of which would lead to asynchrony between cytoplasmic and cytokinetic states. While some organisms do employ alternative cytokinetic pathways to account for differences in growth conditions or defective proteins, these cells tend to lack a high degree of polarity and frequently use adherence to a surface as a mechanism to separate daughter cells (Choudhary et al., 2013; Gerisch and Weber, 2000; Rancati et al., 2008; Uyeda and Nagasaki, 2004). The inability of trypanosomes to sustain growth in the absence of TOEFAZ1 strongly argues that this method of cell division cannot be relied upon to provide viable progeny. The differences in our results compared to those of others that have depleted TOEFAZ1 could reflect a difference in RNAi targeting efficiency or rate of depletion, although we note that we have shown the specificity of our RNAi hairpin with an add-back control. Long-term live-cell imaging, which is still a challenge in T. brucei, will be required to determine the contribution of the back-up cytokinesis to T. brucei cell division.

The primary function of TOEFAZ1 is to correctly position cytokinetic components to direct cleavage furrow ingression along the correct plane and at the correct time. The fact that cell division events can occur along an incorrect plane in the cell highlights the malleability of the cytoskeleton, but also emphasizes the importance of the precise placement of the cleavage furrow. Future research is needed to identify the molecular components that drive furrow ingression, remodel the surrounding membrane, and modify the subpellicular microtubule corset. The function of TOEFAZ1 homologs in the related kinetoplastids Leishmania and T. cruzi, which have FAZ homologs but do not attach most of their flagella, also remains to be studied (Wheeler et al., 2016).

Wild-type procyclic T. brucei brucei 427 strain cells were used to perform experiments, as well as 427 cells that were modified to carry doxycycline-inducible machinery (29-13) (Wirtz et al., 1999). 427 cells were passaged in Beck's Medium (Hyclone, GE Healthcare, Logan, UT) supplemented with 10% fetal calf serum (Gemini Bioproducts, West Sacramento, CA), while 29-13 cells were passaged in Beck's Medium supplemented with 15% doxycycline-free fetal calf serum (Takara Bio USA, Mountain View, CA), 50 µg ml⁻¹ hygromycin (ThermoFisher Scientific, Waltham, MA) and 15 µg ml⁻¹ neomycin (Sigma-Aldrich, St Louis, MO). 427 and 29-13 medium also included 10 µg ml⁻¹ gentamycin (ThermoFisher Scientific) and 500 µg ml⁻¹ penicillin-streptomycin-glutamine (Hyclone). All cells were maintained at 27°C. Cells were counted using a particle counter (Z2 Coulter Counter, Beckman Coulter, Brea, CA).

Antibodies are from the following sources and used at the following dilutions: anti-Ty1 antibody (1:300) from Cynthia He (National University of Singapore, Singapore), 1B41 (1:1000) from Linda Kohl (Centre National de la Recherche Scientifique, Paris, France), and anti-FAZ1 antibody (1:150) from Keith Gull (Oxford University, Oxford, UK). The anti-HA antibody (1:500) was purchased from Sigma-Aldrich (clone 3F10) and the anti-α tubulin antibody (1:50,000) from ThermoFisher Scientific (clone B-5-1-2). The anti-TbPLK (1:150) and anti-TbCentrin2 (1:250) antibodies, as well as the anti-GFP polyclonal antibody (1:10,000), were previously described (de Graffenried et al., 2008; Ho et al., 2006; Ikeda and de Graffenried, 2012).

All constructs were created by PCR amplification of inserts using Q5 polymerase (NEB, Ipswich, MA) followed by either restriction-ligation or Gibson assembly into a sequencing vector (PCR4Blunt) or pLEW100 as previously described (McAllaster et al., 2016). Each DNA construct was validated by sequencing and transfected into cells using an electroporator (GenePulser xCell, Bio-Rad, Hercules, CA). Clonal cell lines were created by selection and limiting dilution.

The endogenous sequence corresponding to each domain of TOEFAZ1 (N-terminus, nt 1–957; IDP, nt 958–1947; ZnF, nt 1948–2376) was PCR amplified and inserted into a pLEW100 vector that contained a 5′ triple-Ty1 epitope tag. Overexpression pLEW100 constructs were linearized with NotI for transfection into the doxycycline-inducible 29-13 cell line and selected with 40 µg ml⁻¹ Zeocin (Invivogen, San Diego, CA). Isolated clones were induced and screened by western blotting and immunofluorescence microscopy.

Endogenous tagging of TOEFAZ1 with triple-Ty1, triple-HA, and mClover3 tags was carried out by targeting the tag to the TOEFAZ1 locus using 500 bp of the 5′ UTR and the first 500 bp of the TOEFAZ1 coding sequence. The endogenous tagging constructs were digested with PacI and NsiI for transfection into either 427 cells or 29-13 cells carrying the doxycycline-inducible TOEFAZ1 RNAi construct (McAllaster et al., 2015), and selected with either 20 µg ml⁻¹ blasticidin (Invivogen) for the triple-Ty1 and mClover3 constructs or 1 µg ml⁻¹ puromycin (Invivogen) for the triple-HA construct. For endogenous tagging in the 29-13 cell line, selection with blasticidin was used at a concentration of 10 µg ml⁻¹.

The recoded TOEFAZ1 sequence was generated from nt 4–1096 of the published TOEFAZ1 5′ coding sequence (927 accession Tb927.11.15800) (Aslett et al., 2009). Optimal T. brucei codons were used when possible (Horn, 2008). To create the RNAi-insensitive domain constructs, the recoded TOEFAZ1 sequence that overlapped with the N-terminal and IDP domains was amplified and annealed to the remaining endogenous TOEFAZ1 sequence for each domain construct. For the trans N-terminus-ZnF construct, the tubulin intergenic sequence was inserted between the recoded N-terminus and endogenous ZnF domain sequence.

Domain constructs were assembled into a sequencing vector containing a 5′ triple-HA tag, were digested as above and transfected into the doxycycline-inducible TOEFAZ1 RNAi cell line containing an endogenous triple-Ty1 or mClover3-tagged TOEFAZ1. Domain analysis constructs were targeted to the second endogenous TOEFAZ1 allele using 500 bp of the 5′ UTR and either 500 bp of 3′ coding sequence (WT recoded control, ΔIDP and ΔN-terminus constructs) or 3′ UTR (ΔZnF, ΔN-terminus-IDP, ΔIDP-ZnF, and trans N-terminus-ZnF constructs). Clonal cell lines were selected with 1 µg ml⁻¹ puromycin (while maintaining selection for the initial endogenous tag and the inducible TOEFAZ1 RNAi construct with 10 µg ml⁻¹ blasticidin and 10 µg ml⁻¹ Zeocin, respectively) and isolated clones were screened for proper recombination through endogenous locus PCR of genomic DNA, western blotting and immunofluorescence microscopy. Domain truncation cell lines were further validated by sequencing of the endogenous locus.

Cultures for overexpression and RNAi time courses were seeded at a density of 10⁶ cells ml⁻¹ and induced with 1 µg ml⁻¹ doxycycline (ThermoFisher Scientific) or treated with 70% ethanol as a vehicle control. Cells were maintained as above and reseeded with fresh medium and doxycycline or ethanol every 48 h. Cells were counted every 24 h, and samples were collected for western blotting and immunofluorescence microscopy as needed. For domain overexpression cell lines, the generation plots are representative of two or three independent clones. For functional RNAi analyses, the generation plot represents the average count of three independent experiments, and the error bars are the standard deviation.

Cells were collected by centrifugation (2400 g for 5 min), washed in PBS, and centrifuged onto coverslips. Cells were fixed in −20°C methanol for 20 min, followed by air-drying or direct rehydration in PBS. Cells were washed in PBS and blocked for either 1 h room temperature (RT) or 4°C overnight in blocking buffer (3% bovine serum albumin or 5% goat serum diluted in PBS). Coverslips were incubated with primary antibodies diluted in blocking buffer for 1 h at RT, washed three times in PBS, and were re-blocked for 20 min at RT. Coverslips were then incubated in Alexa Fluor 647-, 568-, and 488-conjugated secondary antibodies (Life Technologies, ThermoFisher Scientific) as above, washed in PBS, and mounted onto glass slides in DAPI-Fluoromount G (Southern Biotech, Birmingham, AL). For DNA state and morphology quantification, cells were harvested as above, fixed in 4% paraformaldehyde in PBS for 20 min at RT, washed in PBS, and either mounted directly or stained for immunofluorescence as above. Coverslips were imaged with a Zeiss Axio Observer.Z1 equipped with an ORCA-Flash 4.0 CMOS camera using a Plan-Apochromat 100×/1.4 NA oil lens. The microscope was controlled by using the ZEN 2 PRO program. All immunofluorescence images (with the exception of those in Fig. 3C,D) are maximum projections of the z-stack as signals were on different focal planes. The DIC image was overlaid on the merge. When comparing the presence or absence of immunofluorescence signal between cells, the images were set to use the same look-up tables. All images were analyzed and quantified in ImageJ (National Institutes of Health, Bethesda, MD), and assembled for publication in Adobe Photoshop and Illustrator (CC 2017).

For electrophoresis, cells were collected by centrifugation, washed with PBS, and lysed in SDS-PAGE loading buffer. A total of 2.5×10⁶ cell equivalents of lysate/lane was separated by SDS-PAGE, transferred onto a nitrocellulose membrane, and blocked for 1 h RT in blocking buffer [5% (w/v) non-fat dried milk dissolved in Tris-buffered saline (TBS) containing 0.1% Tween-20]. Blots were incubated overnight at 4°C with primary antibodies diluted in blocking buffer. Blots were then washed in TBS containing 0.1% Tween-20 and incubated for 1 h RT with secondary antibodies conjugated to horseradish peroxidase (Jackson ImmunoResearch, West Grove, PA) diluted in blocking buffer. Blots were washed and then imaged using Clarity Western ECL Substrate (BioRad) and a BioRad Gel Doc XR+ system. When required, blots were stripped with Restore Western Blot Stripping Buffer (ThermoFisher Scientific), washed, re-blocked, and incubated with antibodies for detection and imaged as above. All western blot images are representative of two independent experiments.

2.7×10⁸ trans N-terminus-ZnF cells were harvested, washed in PBS, and resuspended in 700 µl of cold amine-free lysis buffer [10 mM NaPO₄, 150 mM NaCl, 0.5% glucose, 7.0% sucrose and 0.5% NP-40 with protease and phosphatase inhibitors (ThermoFisher Scientific)]. Cells were lysed using the gentleMACS Dissociator (Miltenyi Biotec, Auburn, CA), and lysates were clarified by centrifugation at 17,000 g for 20 min at 4°C. The supernatant was separated into two equal fractions to which either 5 mM DSS (ThermoFisher Scientific) or DMSO was added. Lysates were incubated for 30 min at RT, and crosslinking was quenched using 10 mM Tris-HCl pH 7.5. The crosslinked lysates were diluted in SDS-PAGE loading buffer, and samples were separated using SDS-PAGE and western blotted as above. The western blot shown is representative of three independent experiments.

3.5×10⁸ trans N-terminus-ZnF cells were washed twice with cold Hank's balanced salt solution (HBSS) and gently resuspended in 3.6 ml of HBSS. The cell suspension was separated into equal fractions to which DMSO or 5 mM DSS was added. Cells were incubated at RT and the crosslinking reaction was quenched as above. Cells were then pelleted and resuspended in cold RIPA buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 2 mM EDTA, 1.0% NP-40, 0.5% Na-deoxycholate, 0.1% SDS with protease and phosphatase inhibitors). Cells were lysed and the lysate was clarified as above. The supernatant was gently rotated with 30 µl of GFP-Trap MA bead slurry (Chromotek, Hauppauge, NY) for 1 h at 4°C. The beads were washed three times with cold RIPA buffer and bound proteins were eluted by incubating the beads at 99°C for 10 min in SDS-PAGE loading buffer. Samples were separated using SDS-PAGE and transferred to nitrocellulose for western blotting as above. The western blot shown is representative of three independent experiments.

The ΔIDP domain mutant cell line was treated with either 70% ethanol or 1 µg ml⁻¹ doxycycline to induce TOEFAZ1 RNAi for 3 days. 2.5×10⁸ cells from each condition were harvested, washed with PBS, and resuspended in 650 µl of cold lysis buffer (10 mM NaPO₄, 150 mM NaCl, 0.5% glucose, 7.0% sucrose, and 0.5% NP-40 with protease and phosphatase inhibitors). The cells were lysed and the lysate was clarified by centrifugation as above. 30 µl of rat anti-TbPLK antibody was added to each supernatant and then gently rotated for 2 h at 4°C. 50 µl of anti-rat IgG magnetic bead slurry (NEB) was added to each condition and rotated for 45 min RT. The beads were then washed three times in lysis buffer and bound proteins were eluted in SDS-PAGE loading buffer as described above. Samples were separated using SDS-PAGE and western blotted as above. The western blot shown is representative of two independent experiments.

For DNA, morphology and immunofluorescence quantification, 300 cells were counted for both control and RNAi conditions per time point for three independent experiments. All error bars represent the standard deviation of three independent experiments unless otherwise stated. Counts were analyzed and graphed in Excel (Microsoft).

We would like to thank Nicholas Hilton and Jenna Perry for comments on the manuscript, Rebecca Page for TOEFAZ1 domain analysis, Richard Bennett for the use of his microscope, and Dan Weinreich and Sam Obado for helpful discussions.