The kinetochore is a multi-protein structure assembled on eukaryotic centromeres mediating chromosome attachment to spindle microtubules. Here we identified the kinetochore proteins Nuf2 and Ndc80 in the apicomplexan parasite Toxoplasma gondii. Localization revealed that kinetochores remain clustered throughout the cell cycle and colocalize with clustered centromeres at the centrocone, a structure containing the spindle pole embedded in the nuclear envelope. Pharmacological disruption of microtubules resulted in partial loss of some kinetochore and centromere clustering, indicating microtubules are necessary but not strictly required for kinetochore clustering. Generation of a TgNuf2 conditional knock-down strain revealed it is essential for chromosome segregation, but dispensable for centromere clustering. The centromeres actually remained associated with the centrocone suggesting microtubule binding is not required for their interaction with the spindle pole. The most striking observation upon TgNuf2 depletion was that the centrosome behaved normally, but that it lost its association with the centrocone. This suggests that microtubules are essential to maintain contact between the centrosome and chromosomes, and this interaction is critical for the partitioning of the nuclei into the two daughter parasites. Finally, genetic complementation experiments with mutated TgNuf2 constructs highlighted an apicomplexan-specific motif with a putative role in nuclear localization.
The protozoan parasite Toxoplasma gondii is a member of the Apicomplexa. This phylum comprises approximately 5000 species many of which infect humans and/or animals. Opportunistic infection of Toxoplasma in people with a compromised or immature immune response can result in severe disease and pathology due to recurring lytic cycles by the tachyzoite stage of the parasite, whereas infection during pregnancy can result in a variety of birth defects (Montoya and Liesenfeld, 2004). To complete the lytic cycle, the parasite employs a unique mode of cell division known as endodyogeny, unfolding by the formation of two daughter cells within the boundaries of a mature mother parasite, which is consumed at the end of the process (Sheffield and Melton, 1968; Striepen et al., 2007). This exclusive process is marked by an unusual relationship between mitosis and cytokinesis, as the formation of daughter buds begins during S phase before the onset of mitosis (Radke et al., 2001; Hu et al., 2002; Striepen et al., 2007; Gubbels et al., 2008; Nishi et al., 2008). T. gondii tachyzoites are haploid and contain 14 chromosomes totalling 65 Mb of DNA sequence. Mitosis is closed, meaning that segregation of duplicated chromosomes (which do not or only limitedly condense) occurs within the confines of the nuclear envelope (Senaud, 1967). The spindle pole is embedded in the nuclear envelope in a structure known as the centrocone, which contains the protein MORN1 (Gubbels et al., 2006) and interacts with the cytoplasmic centrosome (Striepen et al., 2007). In Toxoplasma duplication of the centrosome during S phase serves to prompt assembly for the daughter cytoskeletons (Striepen et al., 2000; Hu et al., 2002; Hartmann et al., 2006; Anderson-White et al., 2012; Chen and Gubbels, 2013).
The centrocone is retained throughout the cell cycle suggesting that the chromosomes are persistently anchored to the spindle pole (Gubbels et al., 2006). This model was recently confirmed in a study using the histone H3 variant CENP-A (CenH3) where it was shown that throughout the G1 phase of the cell cycle CENP-A is visible as single nuclear spot, associated with the spindle pole, whereas two spots became visible upon mitosis (Brooks et al., 2011). The single or double nuclear spots indicate that the 14 centromeres remain clustered throughout the cell cycle and are at any point during the cell cycle, including G1, associated with the spindle pole. The reason for this persistent clustering is not entirely clear, but it would avoid the need to capture the chromosomes before mitosis as seen in mammalian mitosis, whereas it could also be a mechanism to ascertain the maintenance of a full set of chromosomes throughout development.
As in all eukaryotes, a microtubular spindle is formed during mitosis to facilitate the segregation of spindle poles, culminating in equal division of the duplicated chromosomes (Striepen et al., 2007; Gubbels et al., 2008). Chromosomes attach to the spindle microtubules via the kinetochore, an extensive protein complex that assembles on the centromere (Ciferri et al., 2007). At the primary sequence level there is much divergence between the kinetochore proteins across species, but in nearly all eukaryotes the kinetochore is functionally conserved (Akiyoshi and Gull, 2013). The Ndc80 complex is the kinetochore component mediating attachment to the spindle microtubules (Wigge and Kilmartin, 2001; Deluca et al., 2002; 2006; McCleland et al., 2004). The tetrameric Ndc80 complex (Ndc80 is known as Hec1 in humans) is composed of two heterodimeric subcomplexes, Nuf2-Ndc80 and Spc24-Spc25, which interact with each other through coiled-coil regions (Ciferri et al., 2005; Wei et al., 2005). The globular region of the Nuf2-Ndc80 heterodimer binds to microtubules through calponin homology (CH) domains present in the N-terminals of both subunits (Cheeseman et al., 2006; Ciferri et al., 2007; Alushin et al., 2010; Sundin et al., 2011).
Since the Toxoplasma spindle pole is embedded in the nuclear envelope, the persistent association of the centromeres with the spindle pole could suggest that kinetochore-mediated attachment to microtubules may be dispensable in tachyzoites. It is not unusual for chromosomal regions to associate with the nuclear envelope (e.g. heterochromatin in Schizosaccharomyces pombe (King et al., 2008) and telomeres in Plasmodium falciparum (Figueiredo et al., 2002) as well as clustered centromeres and telomeres in humans (Solovei et al., 2004) thus it can be imagined that the clustered centromeres of Toxoplasma associate with the nuclear envelope as well. To test this hypothesis we identified and subsequently characterized the Toxoplasma kinetochore proteins Nuf2 and Ndc80 by generating fluorescent protein fusions and specific antisera. These reagents demonstrated that the kinetochores are retained throughout the cell cycle and remain clustered. Pharmacological disruption of microtubules resulted in stray clusters indicating microtubules are necessary, but not strictly required for clustering of kinetochores. Moreover, through the generation of a conditional knock-down line we were able to show that TgNuf2 is essential for chromosome segregation. Surprisingly, TgNuf2 was also necessary for the interaction between the cytoplasmic centrosome and the centrocone. However, clustering of the centromeres and their association with the centrocone is retained in the absence of TgNuf2, which suggests that a distinct mechanism independent of chromosome-microtubule attachment underlies this phenomenon.
Identification of kinetochore proteins in T. gondii
We identified kinetochore proteins Ndc80 and Nuf2 in the Toxoplasma genome (http://www.toxodb.org) through reciprocal blast searches with yeast and mammalian orthologues. TgNuf2 (TGME49_309380) encodes a protein of 608 amino acids (Fig. 1A) whereas its binding partner TgNdc80 (TGME49_260420) is comprised of 728 amino acids (Fig. 1B). Formation of the Ndc80-Nuf2 heterodimer is mediated by coiled-coil regions (Ciferri et al., 2005; Wei et al., 2005), which we readily identified in the C-terminal halves of TgNuf2 and TgNdc80 (Fig. 1). Moreover, we were able to identify CH domains, which mediate contact with microtubules, in both TgNuf2 and TgNdc80 (Fig. 1). Sequence alignments of the CH domains with Nuf2 and Ndc80 orthologues that we identified in other apicomplexan genomes and several model organisms revealed several conserved features. For example, several hydrophobic residues have been identified in the human Nuf2 and Ndc80 sequences that are critical for microtubule binding (Wei et al., 2007; Ciferri et al., 2008; Sundin et al., 2011). While most of these residues are not positionally conserved, analysis of the CH domain alignments identified several lysine residues in these domains, specifically three Lys residues that are positionally conserved in the TgNdc80 CH domain (Lys165, Lys207 and Lys217 in Fig. 1B) (Wei et al., 2007; Ciferri et al., 2008; Sundin et al., 2011). Interestingly, the apicomplexan CH domain in Nuf2 contains a strongly conserved stretch (IGNLR residues 121–125) specific to the Apicomplexa. The function of this conserved sequence is currently unknown (Fig. 1A; marked by asterisks).
TgNuf2 and TgNdc80 colocalize with the centromeres throughout the cell cycle
We first confirmed by RACE-PCR and DNA sequencing that the predicted coding sequences for TgNuf2 and TgNdc80 on ToxoDB were correct (results not shown). Subsequently, we set out to establish the subcellular localizations of TgNuf2 and TgNdc80 throughout the cell cycle using three parallel strategies. First, the coding sequences of both kinetochore proteins were cloned as C-terminal autofluorescent protein fusion constructs in Toxoplasma expression plasmids driven by the α-tubulin promoter. Additionally, we raised specific antisera in guinea pigs against the CH-domain containing N-termini of TgNuf2 and TgNdc80 expressed as recombinant proteins. Lastly, we generated transgenic strains in which we endogenously tagged the genomic locus of TgNuf2 and TgNdc80 at the C-terminus with a triple myc-epitope tag (gNuf2- and gNdc80-myc3) and a YFP tag (gNuf2- and gNdc80-YFP) (Huynh and Carruthers, 2009). Western blot analysis using a myc antibody revealed bands consistent with the predicted sizes for triple myc-tagged TgNuf2 (73 kDa) and TgNdc80 (86 kDa) (Fig. S1). Our reagents all produced consistent results and revealed that both proteins colocalize and are present as one or two spots in the nucleus (Fig. 2A).
To validate the putative localization of TgNuf2 and TgNdc80 to the kinetochore we used two independent markers of the centromeres. First we employed parasites stably expressing CENP-A with a triple haemagglutinin tag fused to the C-terminus (Brooks et al., 2011). These parasites were transfected with a TgNdc80-YFP encoding plasmid (Fig. 2B). This demonstrated that TgNdc80-YFP and CENP-A-HA3 colocalize as single or double nuclear spots (Fig. 2B). The second marker we employed was a recently identified chromodomain-containing protein, TgChromo1, which binds with high affinity to H3K9me3, a chromatin mark highlighting the centromere (Gissot et al., 2012). As expected, a specific antiserum against TgChromo1 colocalized with gNuf2-YFP (Fig. 2C). Taken together, these results firmly validated the localization assignment of both TgNuf2 and TgNdc80 to the Toxoplasma kinetochores assembled on the centromere clusters.
The presence of either one or two spots likely corresponds with differences in the cell cycle, as previously shown for CENP-A (Brooks et al., 2011) and indicates that, like the centromeres, the kinetochores are equally clustered at all points in tachyzoite development. To test this directly we co-stained parasites with several cell cycle markers. First, we used a centrin antibody to mark the centrosome, which duplicates right before the onset of S phase (Hartmann et al., 2006). As shown in the upper panels of Fig. 2D, TgNuf2 localizes in close proximity to the centrosome during interphase. In the lower panels of Fig. 2D the centrosomes have duplicated, marking the transition from G1 to S phase. At this point the TgNuf2 signal is visible as a single, yet sometimes slightly elongated structure flanked by the centrosomes, which likely represents the metaphase stage of mitosis. In addition to the centrosome marker, we used the centrocone marker MORN1. MORN1 is additionally present at the basal complex as well as the apical complex, although this latter localization is relatively weak and usually invisible. In addition, MORN1 localizes at an early stage to the new daughter buds and therefore serves as an additional marker for cytokinesis (Gubbels et al., 2006; Anderson-White et al., 2012). Significant, yet imperfect colocalization of TgNuf2 and MORN1 in the spindle pole during the metaphase to anaphase transition in mitosis (Fig. 2E, upper panel) and following the completion of mitosis at mid-budding (Fig. 2E, lower panel) demonstrates the association of the kinetochores with the spindle pole. Collectively, these results demonstrate that throughout the Toxoplasma cell cycle the clustered kinetochores remain at any point associated with the centromere on one side and the spindle pole on the other side.
Finally, the cell cycle co-ordinated expression of TgNuf2 was examined using the TgNuf2 antibody in conjunction with the aforementioned centrin antibody (Fig. 2F). The signal of the kinetochore, which is closely matched to that of the centrosome, reduces in intensity during the G1 phase of the cell cycle. Since we collected z-stacks and deconvoluted these images, it is possible we did not acquire a full z-series of images for all parasites in the figure. Therefore, we assured ourselves that we collected a full z-series for all parasites within panel F, which is provided in Supplementary Fig. S2. These data show a full series was obtained and therefore the reduced signal intensity of the centrosome and kinetochores in G1 parasites is valid. A potentially alternative artefact is that the epitopes recognized by the centrin and TgNuf2 antisera are masked in a cell cycle-dependent fashion. Although we cannot directly test this, we deem this scenario unlikely as both these antisera are polyclonal and masking is unlikely to span extensive stretches of sequence.
Microtubules contribute, but are not strictly required for kinetochore clustering
The above results indicate a direct relationship between centromere clustering and spindle pole anchoring; however, the experiments do not directly test that kinetochores are actually associated with microtubules at all points in the cell cycle. We reasoned this model can be tested either by depolymerizing the spindle microtubules or by depleting the kinetochore proteins responsible for mediating microtubule attachment (e.g. Nuf2 or Ndc80). For the first approach we used oryzalin, a microtubule depolymerizing herbicide, which disrupts the subpellicular Toxoplasma microtubules at a concentration of 0.5 μM and the spindle microtubules at a concentration of 2.5 μM (Stokkermans et al., 1996; Morrissette and Sibley, 2002; Ma et al., 2010). Hence, to determine the status of kinetochore clustering in the absence of microtubules we treated intracellularly developing parasites with 2.5 μM oryzalin (Morrissette and Sibley, 2002) (Fig. 3A). If spindle microtubules were required for spindle pole anchoring we would expect dissociation from the spindle pole and we would likely loose the clustering of the 14 chromosomes. By staining treated parasites with the TgNuf2 antiserum we observe multiple kinetochore clusters of varying intensity (Fig. 3B). We do not see a gross disruption of kinetochore clustering, but the weaker spots potentially represent kinetochores released from the spindle pole of a kinetochore cluster, indicating that the interaction of these connections is dependent on the presence of microtubules. To assess whether the weaker TgNuf2 spots were indeed stray kinetochores we determined whether these spots colocalize with the centromeres using a CENP-A monoclonal antibody (Francia et al., 2012). As shown in Fig. 3B, we observe weak TgNuf2 spots colocalizing with the centromere marker, indicating these are likely dissociated chromosomes. Careful observation of the two signals, however, shows that the TgNuf2 signal connects several of the weaker centromere spots, suggesting that multiple chromosomes are dissociated, which are still connected with each other through their kinetochores (Fig. 3B). We also observed poor colocalization, or at least vast variations in intensity between TgNuf2 and CENP-A signals (Fig. 3B). Quantification of dissociated centromeres and kinetochores revealed these were not rare events, with frequencies for both events present in around 20% of the parasites (Fig. 3C). In addition, the dispersed DAPI signal in these parasites illustrates extensive nuclear fragmentation in these nuclei. These data would suggest that we have centromeres without kinetochores, as well as kinetochores without centromeres. Therefore, disruption of microtubules by oryzalin appears to disrupt the connection between centromeres and kinetochores.
To directly monitor the association of kinetochores with the centrocones we stained parasites stably expressing myc2-MORN1 with the TgNuf2 antiserum following treatment with oryzalin (Fig. 3D). Strikingly, we observe that every TgNuf2 signal is associated with a centrocone signal following treatment with 2.5 μM oryzalin. To confirm the accuracy of this observation we also performed a co-stain of MORN1 and CENP-A (Fig. 3E). Additionally, co-staining with TgNuf2 and centrin revealed mitosis does not occur in oryzalin treated parasites and duplicated centrosomes fail to separate (Fig. 3F). It is of note we observed kinetochore clusters not associated with centrosomes at this point, indicating centrosome–centrocone association is microtubule dependent. Finally, staining oryzalin treated myc2-MORN1-expressing parasites with centrin antibody revealed multiple centrosomes in close association with a single centrocone (Fig. 3G), suggesting centrosome separation is microtubule dependent.
Generation of a conditional TgNuf2 knock-down parasite strain
Although the pharmacological disruption of microtubules results in dramatic changes in centromere and kinetochore organization concerns have been raised in other systems, notably budding and fission yeast, that pharmacological disruption of spindle microtubules is always incomplete (Sawin and Snaith, 2004; Castagnetti et al., 2010; Richmond et al., 2013). Hence, to bypass this potential artefact and to provide an independent approach to test if microtubule-attachment of chromosomes via the kinetochore is dispensable throughout the cell cycle of Toxoplasma, we employed reverse genetics. We generated a conditional TgNuf2 knock-down parasite strain (TgNuf2-cKD) by directly replacing the endogenous promoter with the tetracycline regulatable (TetO7sag4) promoter through double homologous recombination (Fig. 4A) (Meissner et al., 2002; Sheiner et al., 2011). Following transfection of the promoter swap plasmid, parasites showing pyrimethamine resistance were screened by PCR to confirm correct integration of the TetO7sag4 promoter and absence of the TgNuf2 promoter. We selected a single clone exhibiting successful promoter replacement (Fig. 4B) and determined its viability by plaque assays (Fig. 4C). No plaque formation was observed, even after prolonged culture of the conditional knock-down parasites in the presence of anhydrous tetracycline (ATc) for up to 20 days to detect potentially slow growing mutants (Fig. 4C). These results show that that TgNuf2 is essential for completion of the tachyzoite cell division cycle. Next, we determined the kinetics of protein depletion upon the addition of ATc using immunofluorescence assays (Fig. 4D). Conditional knock-down parasites grown in the presence or absence of ATc for 3, 6 or 9 h were stained with anti-TgNuf2 antiserum. Expression of TgNuf2 persisted for 6 h following addition of ATc but was no longer detectable 9 h post incubation (Fig. 4D). Since Nuf2 forms a heterodimer with Ndc80, we also determined whether TgNuf2 is required for the retention of its binding partner at the kinetochore. TgNdc80 shows similar kinetics as TgNuf2, as expression is no longer detectable following exposure to ATc for 9 h (Fig. 4E). Taken together, our data suggest that one division cycle (6–7 h) is needed to deplete TgNuf2 upon knock-down of the conditional allele, and that TgNuf2 knock-down results in simultaneous disappearance of its binding partner TgNdc80 from the kinetochore.
Nuf2 is required for proper mitotic segregation of chromosomes
Based on the function of the Ndc80 complex in yeast and humans, we hypothesized that parasites deficient in TgNuf2 would likely exhibit missegregation of chromosomes. Indeed we already observed nuclear missegregation defects in Fig. 4D and E. To specifically monitor the nuclear material we stably transfected the TgNuf2-cKD strain with a histone 2B YFP fusion protein encoding plasmid (Hu et al., 2004) and cultured the parasites in the presence of ATc for 12 h. Besides those shown in Fig. 4D and E, this revealed an array of mutant phenotypes (Fig. S3). Some parasites demonstrated partial accumulation of two nuclei in a single parasite and complete and partial loss of nuclei (Fig. 4D and E, Fig. S3). The most striking phenotype is the complete accumulation of a single irregularly shaped nucleus, indicating chromosome missegregation or incomplete karyokinesis (Fig. 5A). We quantified the percentage of parasite vacuoles exhibiting the mutant phenotype of nuclear loss following addition of ATc. As shown in Fig. 5B, the phenotype is detected 9 h post induction with ATc and after 18 h 80% of the vacuoles contain a parasite from which the nucleus is missing. Together with the observation that it takes 9 h to deplete the protein, this observation indicates that the nuclear missegregation and fragmentation phenotypes are the result of nuclear partitioning gone awry during the cell division process.
In parasites without a nucleus we always observed a small amount of DAPI stained DNA (marked by an arrowhead in Fig. 5A). Since this DNA is devoid of H2B staining we reasoned it to be the apicoplast, whose genome in wild-type parasites is usually visible as a small spot of DAPI stained nuclear material apical of the nucleus. Moreover, it has been shown that plastid partitioning is mediated by association with the centrosome (Striepen et al., 2000) and therefore would be independent of the kinetochore. As shown in Fig. 5C, a marker for the plastid confirms this prediction (in Fig. 5E we observe that the centrosome and plastid DNA are co-segregating as well).
We next wanted to establish the localization of the centromere in the conditional TgNuf2 knock-down, as this is the site on the chromosome for assembly of the kinetochore. Specifically, we wanted to determine if we still observe centromere clusters and whether these clusters localize to fragmented nuclei. TgNuf2 depleted parasites stained with the CENP-A centromere marker display retention of clustered centromeres within the nuclei (Fig. 5D), indicating that depletion of TgNuf2 does not affect the clustering of the centromeres.
To establish whether the interaction between the cytoplasmic centrosome and spindle pole embedded in the nuclear envelope is maintained in the absence of TgNuf2 we expressed a myc2-MORN1 encoding construct in TgNuf2 depleted parasites to highlight the centrocone and simultaneously co-stained these parasites with anti-centrin to mark the centrosome. In wild-type parasites the spindle pole is always adjacent to the centrosome; however, depletion of TgNuf2 abolishes this interaction, as visible by the extended distance between these structures (Fig. 5E and F). As predicted by the intact apicoplast division, parasites exhibiting chromosome missegregation continue to duplicate and accurately segregate their centrosomes. The lower panel in Fig. 5E shows four parasites, each with a centrosome and associated plastid DNA; however, only two parasites have a nucleus with a single centrocone. In addition, clustered centromeres remain persistently associated with the centrocone in the absence of TgNuf2 (Fig. 5G). It has been shown that MORN1 localizes to both the cytoplasmic and the nuclear side of the nuclear membrane in the centrocone (Gubbels et al., 2006). Therefore, it is possible that MORN1 mediates the anchoring of the kinetochore to the centrocone. To test this model we studied kinetochore and centrosome localization in parasites wherein MORN1 was depleted (Lorestani et al., 2010). As shown in Fig. S4B, depletion of MORN1 does neither affect the clustering of kinetochores, nor affect the localization of the centrosomes, nor affect the progression of mitosis (Fig. S4B). Hence, MORN1 is not the protein anchoring the kinetochore to the centrocone.
Identification of a putative apicomplexan-specific Nuf2 nuclear localization signal
Depletion of the kinetochore component TgNuf2 allows for complementation studies to identify residues or domains critical for Nuf2 function in Toxoplasma. Comparable studies on human Nuf2 identified the CH domain to be critical for progress of mitosis (Sundin et al., 2011). It was shown that stable microtubule-microtubule interactions were established, but there was a block in progression towards anaphase. Three lysine residues in the CH domain (Lys33, Lys41 and Lys115) were shown to be critical for this phenotype. However, structural studies showed that these are not directly involved in microtubule binding, since the Nuf2 domain does not directly interact with microtubules (Alushin et al., 2010). The current model is that the CH domain is involved in either stabilizing and/or positioning the Ndc80 complex on the microtubules (Sundin et al., 2011). Sequence alignments revealed the critical lysine residues identified in humans are not positionally conserved in TgNuf2, but there are lysine residues which likely function similarly (Fig. 1A). Our first experiment towards validating this model for Toxoplasma was the generation of a TgNuf2 CH-domain deletion mutant fused to a triple myc tag (ΔCH domain-myc3). We first validated whether expression levels and protein integrity of the mutant constructs was comparable to a wild-type complementation construct (Fig. 6B). As shown in Fig. 6C, the ΔCH domain-myc3 construct was unable to complement the TgNuf2-cKD mutant, whereas a control wild-type allele successfully restored function. Subsequently, we analysed the subcellular localization of the ΔCH domain-myc3 protein and determined that it was cytoplasmic (Fig. 6D). This is in stark contrast to results reported for the comparable human Nuf2 deletion construct, which correctly localized to the kinetochore (Sundin et al., 2011). This can be attributed to the disassembly of the nuclear envelope during mitosis in humans, which is not the case for Toxoplasma. Our observations therefore suggest that the TgNuf2 domain contains a nuclear localization signal (NLS). Hence, we subjected the TgNuf2 sequence to several NLS finder searches, which all returned negative results. We reasoned that if there is an NLS in the CH domain, it may be conserved across the Apicomplexa. As previously mentioned, the IGNLR region (aa 121–125) is conserved across the phylum, therefore to address the function of this region we generated a ΔIGNLR mutant and a construct wherein these residues are replaced by alanine residues (IGNLR:AAAAA). Growth restorations through genetic complementation with either construct were unsuccessful and both exhibited cytoplasmic localization (Fig. 6D). As such, these results suggest that this conserved stretch of sequence acts as an NLS, or at least is part of an NLS.
All living organisms need to accurately endow their offspring with nuclear material (Errington et al., 2005; Yanagida, 2005; Drechsler and McAinsh, 2012). The common denominator in eukaryotes is the assembly of a microtubular spindle to which the chromosomes attach, whereas there is variation regarding the organization of the spindle microtubules and modes of spindle pole separation (Drechsler and McAinsh, 2012; Akiyoshi and Gull, 2013). The kinetochore protein complex mediates attachment of chromosomes to microtubules, which is a structure conserved across the vast majority of eukaryotes, including the Apicomplexa (Akiyoshi and Gull, 2013). Since the primary structure of kinetochore proteins is quite variable between species, identification of kinetochore components in annotated genomes is challenging. However, we identified convincing candidates for Ndc80 and Nuf2 within the Toxoplasma genome, containing several conserved features (Fig. 1). Attempts to identify other kinetochore components by genome mining were unsuccessful (data not shown).
Subcellular localization studies of TgNuf2 and TgNdc80 revealed that both colocalize with clustered centromeres in close association with the centrocone (Fig. 2B and C). Moreover, kinetochore clustering is retained throughout the cell cycle (Fig. 2D and E), consistent with the recently shown persistent clustering of centromeres and centrocone association (Brooks et al., 2011). Clustering of centromeric chromosomal regions has been demonstrated in yeast (Funabiki et al., 1993; Janke et al., 2001; Richmond et al., 2013) and human cells (Solovei et al., 2004), although the biological significance remains poorly understood. In budding yeast, clustering has been proposed as a mechanism by which kinetochores are efficiently captured by spindle microtubules. In contrast to most other eukaryotes, only a single microtubule binds each kinetochore in budding yeast (Winey et al., 1995; Westermann et al., 2007). Capture of a single detached kinetochore by a single microtubule would result in the transport of all detached kinetochores towards the spindle pole, significantly increasing the chances of kinetochore-microtubule attachment (Richmond et al., 2013). Observations by Swedlow et al. indicate there are a limited number of microtubules in the Toxoplasma nucleus, as only a maximum of 11 microtubules were detected for 14 chromosomes (Swedlow et al., 2002). We propose persistent clustering and anchoring of chromosomes to spindle microtubules via kinetochore functions as means to ensure genome integrity during cell division. This model is particularly applicable for the coccidian parasite Sarcocystis neurona, which divides by endopolygeny. This model is marked by DNA replication in the absence of nuclear division, resulting in the formation of a polyploid nucleus intermediate and ultimately the production of 64 haploid daughter cells (Vaishnava et al., 2005; Striepen et al., 2007; Brooks et al., 2011). Retention of intranuclear spindle poles throughout the cell cycle suggests persistent attachment of chromosomes via kinetochores to spindle microtubules likely ensures genome integrity during increasing ploidy throughout Sarcocystis cell division and ensures each daughter emerges with a complete set of chromosomes. The more distantly related P. falciparum divides by schizogony, generating schizont intermediates wherein up to thousands of nuclei share a single cytoplasm, none of which appear to contain more than 2N (Arnot et al., 2011; Gerald et al., 2011). Therefore, in schizogony it is not necessary to organize the nucleus as strictly as in Sarcocystis. Consistent with this model, it has recently been demonstrated that clustering of centromeres in P. falciparum only occurs prior to the onset of mitosis (Hoeijmakers et al., 2012). Overall, these data suggest that persistent kinetochore clustering within the Apicomplexa is an innovation in the Sarcocystidae (encompassing Toxoplasma and Sarcocystis).
Interestingly, treatment with the microtubule depolymerizing agent oryzalin fails to completely disrupt clustering, indicating clustering of kinetochores in Toxoplasma is not strictly dependent on spindle microtubules (Fig. 3; see Fig. 7 for schematics of scenarios). Although stray kinetochores were seen in the absence of stable kinetochore-microtubule attachments, 14 nuclear spots corresponding to kinetochores of individual chromosomes were never detected (Fig. 3). Strikingly, we observed a high incidence of stray kinetochores without centromeres as well as centromeres without kinetochores following oryzalin treatment, suggesting loss of microtubules disrupts the assembly of kinetochores on centromeres (Fig. 3B and C). It is possible this unprecedented observation may be a secondary effect of the oryzalin treatment.
Microtubule-independent clustering of kinetochores is also observed in Saccharomyces cerevisiae, as nocodazole treatment results in the disconnection of some kinetochores from the spindle pole, but does not abolish clustering (Richmond et al., 2013). Nocodazole treatment failed to disrupt all kinetochore-microtubule attachments, as indicated by the localization of kinetochore clusters to the spindle pole (Sawin and Snaith, 2004; Castagnetti et al., 2010; Richmond et al., 2013). Moreover, a specific protein, Csi1, was recently identified in S. pombe that mediates centromere clustering and their anchoring to the nuclear envelope (Hou et al., 2012; 2013). It is not unlikely that a protein with similar function is present in Toxoplasma, which would likely be the result of a parallel evolution as we were unable to identify a Csi1 orthologue in the Toxoplasma genome (protein X in Fig. 7).
Upon knock-down of TgNuf2 we observed nuclear fragmentation (Fig. 5A, Fig. S3); however, we never observed complete loss of clustering (i.e. 14 or 28 independent centromeres) (Fig. 5D). Similar phenotypes have been observed in Nuf2 and Ndc80 temperature-sensitive (ts) mutants generated in yeast (Jin et al., 2000; Janke et al., 2001; Anderson et al., 2009). In addition, various other proteins have been associated with kinetochore clustering, including kinesin motor proteins Cin8, Kip1 and Kip3 (Gardner et al., 2008; Wargacki et al., 2010) but only moderate impacts on the disruption of kinetochore clusters were observed in these mutants. More recently it was revealed that disruption of spindle microtubules in conjunction with Slk19 depletion in budding yeast results in a more dramatic loss of kinetochore clustering, hinting at a specific role for Slk19 in clustering (Richmond et al., 2013). A putative functional orthologue termed CENP-F/mitosin has been identified in mammals (Holt et al., 2005). No direct orthologue of Slk19 or CENP-F could be identified in the Toxoplasma genome; however, based on these data and our results we predict that a protein with a similar kinetochore-cohesive function is likely present in Toxoplasma (protein Y in Fig. 7).
While depletion of TgNuf2 fails to disrupt kinetochore clustering, segregation of chromosomes is drastically affected. Missegregation of chromosomes resulted in the partial and complete loss of the nucleus as well as the accumulation of two nuclei in a single parasite (Fig. 5A, Fig. S3). As shown in Fig. 4D, TgNuf2-cKD parasites progress through one cell cycle before TgNuf2 and TgNdc80 are no longer detectable as a phenotype begins to manifest. The inability to detect TgNdc80 concurrently with TgNuf2 suggests they form a stable dimer and the association is required for the formation of a functional kinetochore. The stability of the Ndc80 complex has been experimentally validated in humans (Hori et al., 2003) using fluorescence recovery after photobleaching (FRAP). This analysis revealed that Nuf2 and Hec1 are stable components of the kinetochore, with recovery times of > 30 min. Hence, to detect a phenotype in TgNuf2-cKD parasites, completion of a cell cycle is required to exchange sufficient Nuf2 from the Toxoplasma kinetochores. In many cases, however, mitosis appears to progress quite normally as two nuclei of comparable size are observed (e.g. Fig. 4D and E). The loss of centrosome–centrocone association may therefore be independent of mitosis. As previously stated, due to kinetochore clustering it is conceivable that mitosis completes normally if only one or a few kinetochores are attached to spindle microtubules, therefore, the loss of the centrosome–centrocone connection appears to be secondary, as the centrocone stays associated with the nucleus (Fig. 5E). It has been reported that the centrosome migrates to the basal end of the nucleus to duplicate and segregate (Hartmann et al., 2006). Our TgNuf2-cKD data appear to be consistent with this dissociation/association of the centrosome with the centrocone. Our data suggest that the centrosome is anchored to the centrocone by spindle microtubules, which in turn are kept in place by association with the kinetochore inside the nuclear envelope (Fig. 7C). Extrapolating on this model, together with the observation that centromeres associate with the centrocone in absence of microtubules or the Ndc80 complex (Fig. 5G), this would predict spindle microtubules are not needed in G1 or interphase. Similar observations have been made in fission yeast where centromeres clustering during interphase is independent of spindle microtubules (Hou et al., 2012; 2013). Hence, the absence of microtubules would permit centrosome migration to the basal side of the nucleus for its duplication. We have recently acquired preliminary data supporting the absence of microtubules from the spindle pole in G1 (C.-T. Chen and M.-J. Gubbels, unpublished). However, the comparison with yeast stops at the centrosome/centrocone association since both fission and baker's yeast have spindle pole bodies (SPB), which is a functional merger of the spindle pole and centrosome. Interestingly, the S. pombe SPBs duplicate in the cytoplasm but are embedded in the nuclear envelope during mitosis, whereas the S. cerevisiae SPBs are at all times embedded in the nuclear envelope (Drechsler and McAinsh, 2012). Hence the S. pombe SPBs behave to some extend as the Toxoplasma centrosomes.
Additionally, we observed that knock-down of TgNuf2 results in fragmentation and total loss of nuclei (Fig. S3). These residual nuclear masses resulting from chromosome missegregation are strikingly similar to the phenotype observed in the nuclear actin-related protein 4a (ARP4a) temperature-sensitive (ts) mutant, which abolishes nuclear localization (Suvorova et al., 2012). The exact role of ARP4a in Toxoplasma chromosome segregation has not been resolved. By functional orthology with yeast ARP4 a role in kinetochore assembly is anticipated (Ogiwara et al., 2007), which fits with the observed similarity of the phenotype with the loss of TgNuf2.
The functional dissection of TgNuf2 by complementation studies using deletion mutants led to the surprising observation that the CH domain is required for nuclear localization (Fig. 6). The short IGNLR region (aa 121–125) conserved across the Apicomplexa proved to be required for nuclear localization. However, this region is not a typical NLS, but since there is a charged Arg residue typically found in the NLS region and there are several two conserved positively charged Arg/Lys residues nearby, it is possible the extended region acts as an NLS, and deletion of a single Arg residue abolishes its function. While our results strongly suggest this sequence acts as an NLS, currently we cannot exclude the possibility that mislocalization is result of misfolding of the mutant proteins. Alternatively, this motif may mediate binding with a protein that facilitates entry of TgNuf2 into the nucleus through its own NLS. Our observations hint at a potential mechanism towards how mitotic components gain access to the nucleus. Nuclear import pathways have not been extensively studied in Toxoplasma, although it is known they are RAN-based as in higher eukaryotes (Frankel and Knoll, 2009). Previous work identified a single importin5α, which has been shown to recognize the NLS in GCN5A (Bhatti and Sullivan, 2005) and ARP4a (Suvorova et al., 2012). In future work we will address how kinetochore components access the nucleus.
RH strain parasites and transgenic derivatives were maintained in human foreskin fibroblasts (HFF) as previously described (Roos et al., 1994). Stable parasites expressing transgenes were selected under 1 μM pyrimethamine and 20 μM chloramphenicol. The TgNuf2 conditional knock-down was induced at a concentration of 1.0 μg ml−1 anhydrous tetracycline (ATc). The RHΔKu80ΔHX (Huynh and Carruthers, 2009) (kindly provided by Vern Carruthers, University of Michigan), TATiΔKu80 (Sheiner et al., 2011) and CENP-A-HA3 (Brooks et al., 2011) (Boris Striepen, University of Georgia) parasite strains were used in this study.
To validate the gene models on ToxoDB and to identify the 5′ and 3′ end of TgNuf2 and TgNdc80 we used the Gene Racer kit (Invitrogen) using RH strain RNA following the manufacturer's instructions. Primer sequences are available in the Supplementary Table S1.
All primer sequences are listed in the Supplementary Table S1. The plasmid ptub-Ndc80-CherryRFP/sagCAT was generated by amplifying the Ndc80 CDS using primers Ndc80-Bc-N-F1 and Ndc80-AvrII-R. The insert, digested with BclI/AvrII, was cloned into the ptub-mCherry-RFP2/sagCAT (Chtanova et al., 2008) plasmid using BglII/AvrII restriction enzymes. The plasmid ptub-Nuf2-YFP/sagCAT was generated by amplifying the Nuf2 CDS using primers Nuf2-Bam-F and Nuf2v2-Avr-R. The insert, digested with BamHI/AvrII, was cloned into the ptub-YFP2(MCS)/sagCAT plasmid using BglII/AvrII restriction enzymes. The plasmid pndc80-Ndc80-YFP/sagCAT was generated by amplifying the promoter (1.5 kb upstream from the start codon) using primers pndc80-PmeI-F and pndc80-BglII-R. The insert, digested with PmeI/BglII was cloned into the ptub-YFP2(MCS)/sagCAT (Gubbels et al., 2003) plasmid using PmeI/BglII restriction enzymes. The Ndc80 CDS was then cloned into the intermediate plasmid using the abovementioned primers.
The endogenous tagging constructs for TgNuf2 and TgNdc80 were generated as previously described (Huynh and Carruthers, 2009). Briefly, pNuf2-YFP-LIC/DHFR was cloned by amplifying 2 kb upstream of the stop codon using RH genomic DNA as template with primers Nuf2-LIC-F and Nuf2-LIC-R. pNdc80-myc3-LIC/DHFR was generated by amplifying 2.9 kb upstream of the stop codon using the primers Ndc80-LIC-F and Ndc80-LIC-R. pYFP-LIC/DHFR and pmyc3-LIC/DHFR (kindly provided by Vern Carruthers, University of Michigan and Michael White, University of South Florida respectively) were digested with PacI. Prior to transfection, pNuf2-YFP-LIC/DHFR and pNdc80-myc3-LIC/DHFR were digested within the homologous region using the restriction enzymes NarI and NsiI respectively.
Three-way Gateway cloning (Invitrogen) was employed to generate the TgNuf2 conditional knock-down vector. The entry clone plasmid pDONR221-5′Nuf2R1/R4 was generated by amplifying approximately 1 kb upstream of the promoter region (1.5 kb upstream of start codon) using RH genomic DNA as a template with primers 5′Nuf2-F-B1 and 5′Nuf2-F-B2 and cloned into the pDONR221 P1/P4 plasmid (Rosowski et al., 2011) (kindly provided by Jeroen Saeij, MIT). The entry clone plasmid pDONR221-DHFR-T7S4 R4/R3 was generated by amplifying the DHFR selectable marker and Tet7Sag4 region from p2NdeI_DHFR_T7S4_myc3 (kindly provided by Lilach Sheiner and Boris Striepen) with primers DHFR-T7S4-F-B4r and DHFR-T7S4-R-B3r and cloned into the pDONR221 P4r/P3r plasmid (Jeroen Saeij). The entry clone plasmid pDONR221-Nuf2 R3/R2 was generated by amplifying 1.1 kb beginning at the start codon using RH genomic DNA as a template with primers Nuf2-F-B3 and Nuf2-R-B2 and cloned into the pDONR221 P3/P2 plasmid (Jeroen Saeij). The conditional knockout vector pTgKO2-T7S4pnuf2 was generated by combining the pTgKO2 (Jeroen Saeij), pDONR221-5′Nuf2R1/R4, pDONR221-DHFR-T7S4 R4/R3 and pDONR221-Nuf2 R3/R2 plasmid in an LR reaction. The plasmid linearized prior to digestion using restriction enzyme PciI.
The full-length complementation vector, ptub-Nuf2-myc3/sagCAT was generated by amplifying the Nuf2 CDS using primers Nuf2-Bam-F and Nuf2v2-Avr-R. The insert, digested with BamHI/AvrII, was cloned into ptub-CactinCD1-3myc3/sagCAT plasmid using BglII/AvrII restriction enzymes. The deletion and mutant complementation constructs were cloned in the same manner. ptub-Nuf2ΔCH domain-myc3/sagCAT was generated by fusion PCR amplification (Szewczyk et al., 2006) using primers Nuf2-Bam-F, Nuf2Δ34-187-R, Nuf2Δ34-187-F, Nuf2-R-AvrII. ptub-Nuf2ΔIGNLR-myc3/sagCAT was generated by fusion PCR amplification using primers Nuf2-Bam-F, Nuf2-ΔIGNLR-R, Nuf2-ΔIGNLR-F, Nuf2-R-AvrII. QuikChange Site-directed mutagenesis was used to generate ptub-Nuf2-IGNLR:AAAAA-myc3/sagCAT with primers Nuf2-IGNLR:A-F and Nuf2-IGNLR:A-R.
Immunofluorescence assays were performed as previously described (Gubbels et al., 2006). Parasite strains of choice were inoculated in six-well plates containing coverslips with HFF cells and fixed and permeabilized with either 100% methanol or 4% paraformaldehyde and blocked in 1% BSA w/v in 1× PBS. The following primary antibodies were used: rabbit α-MORN1 1:2000 (Gubbels et al., 2006), rat α-IMC3 1:2000–1:3000 (Anderson-White et al., 2011), rabbit α-human centrin 1:1000 (kindly provided by Iain Cheeseman, Whitehead Institute), mouse α-TgChromo1 1:500 (Gissot et al., 2012) (kindly provided by Mathieu Gissot, Institute Pasteur de Lille) mouse α-CENP-A 1:20 (Francia et al., 2012) (kindly provided by Boris Striepen, University of Georgia), rabbit α-β-tubulin 1:500 (Morrissette and Sibley, 2002) (kindly provided by Naomi Morrissette, University of California, Irvine), rat α-HA 1:3000 (Roche), mouse α-myc 1:50 (Santa Cruz) rabbit α-GFP 1:500 (Abgent), guinea pig α-Nuf2 1:2000 and guinea pig α-Ndc80 1:2000. Alexa fluorophores A488 and A594 1:200 (Invitrogen) conjugated to α-rat, α-rabbit, α-guinea pig and α-mouse secondary antibodies were used. Streptavidin conjugated to Alexa 594 1:500 was used to highlight the plastid (Jelenska et al., 2001). Nuclear material was co-stained with 4′,6-diamidino-2-phenylindole (DAPI). A Zeiss Axiovert 200 M wide-field fluorescence microscope equipped with a α-Plan-Fluar 100×/1.3 NA and 100×/1.45 NA oil objectives and a Hamamatsu C4742-95 CCD camera was used to collect images, which were deconvolved and adjusted for phase-contrast using Volocity software (Improvision/Perkin Elmer).
Six-well plates confluent with HFF cells were inoculated with 50 freshly lysed Nuf2-cKD, TATiΔKu80, or complemented Nuf2-cKD parasites in Ed1 medium and incubated for 7 or 20 days in the presence or absence of 1.0 μg ml−1 ATc. The monolayer was then fixed with 100% ethanol for 10 min and stained with crystal violet to visualize the plaques (Roos et al., 1994).
Generation of antisera
To generate His6 N-terminal fusion proteins, the 5′ end prior to the coiled-coil domain of Nuf2 (1–160 aa) and Ndc80 (1–243 aa) were PCR-amplified using cDNA as template and cloned into the pAVA0421 plasmid by LIC (Alexandrov et al., 2004). Fusion proteins were expressed in BL21 Escherichia coli using 30 μM IPTG at 30°C for 4 h and purified under denaturing conditions over Ni-NTA Agarose (Invitrogen). Polyclonal antisera were generated by guinea pig immunizations (Covance, Denver, PA). Nuf2 antisera were affinity-purified against Nuf2 (1–160 aa) recombinant protein using published methods (Gubbels et al., 2006).
Six-well plates confluent with HFF cells were inoculated with freshly lysed parasites in Ed1 medium and incubated for 2 h. Intracellular parasites were treated with 2.5 μM oryzalin (Supelco) for 24 h and then methanol-fixed and analysed by immunofluorescence.
Western blots were performed according to previously published methods (Gubbels et al., 2006). Essentially, a 4–12% NuPAGE Bis-Tris gel (Invitrogen) was loaded with 60 × 106 parasites. Proteins were transferred to a PDVF membrane (Bio-Rad) and blocked with a solution of 5% milk and 1% BSA. Blots were probed with mouse α-c-myc-HRP antibody 1:2000 (Santa-Cruz Biotech) or mouse anti-α-tubulin 12G10 (University of Iowa Hybridoma Bank). Following a 1 h incubation the membranes were washed three times with 1× PBST for 10 min and once with 1× PBS for 5 min. Signals were detected using chemiluminescent HRP substrate (Millipore) treatment and subsequently exposed to X-ray film.
Sequences were accessed from ToxoDB.org and NCBI (see Fig. 1 for accession numbers). Nuf2 and Ndc80 domains were predicted using the program smart (Schultz et al., 1998; Letunic et al., 2012). Multiple sequence alignments were generated using the program muscle (Edgar, 2004) and analysed using Jalview (Waterhouse et al., 2009). Coiled-coil domains were predicted using the COILS server (Lupas et al., 1991) using a window width of 28 and a minimum score of 0.9.
We thank Mathieu Gissot, Boris Striepen, Jeroen Saeij, Michael White, Vern Carruthers, Naomi Morrissette and Iain Cheeseman for kindly sharing reagents. This work was supported by National Institutes of Health Grant AI081924 and a March of Dimes Basil O'Connor Starter Scholar Research Award (5-FY09-08) to M.-J.G.