A MORN1-associated HAD phosphatase in the basal complex is essential for Toxoplasma gondii daughter budding

Identification of HAD2a

To identify proteins directly interacting with MORN1 we mined a tachyzoite cDNA library by Y2H screening. Because bait comprising full-length MORN1 resulted in auto-activation, we used two partial constructs, a N-terminal construct (MORN1 aa 1–144) and a C-terminal construct (MORN1 aa 196–363) (Fig. S1), not causing auto-activation. A total of ~120 000 yeast colonies were screened for Y2H interactions across both constructs, which resulted in 44 yeast clones corresponding with 33 unique Toxoplasma genes (Fig. S1, Table S1). Among these was MORN1 itself (Y2H clone ‘MORN1 N-terminal interacting candidate’ (MNIC-13) as well as Hsp20 (clones MNIC-6 and ‘MORN1 C-terminal interacting candidate’ (MCIC-29), the latter is known to localize to the IMC and the basal end of the parasite (de Miguel et al., 2008). We focused our attention on genes that encode hypothetical proteins and shortlisted nine because of their gene expression profile (peaking in S/M phase of parasite replication; Behnke et al., 2010) and our ability to obtain sub-cellular localization information (Table S1). Yeast clone MNIC-18 became our lead candidate as its product showed strong interaction with MORN1 across three different Y2H interaction reporter assays (Fig. 1A). Sequence analysis identified a haloacid dehalogenase (HAD) phosphotransferase domain (Fig. 1B and C). Based on these data, our phenotypic observations described in the succeeding texts and the recent characterization of a HAD phosphatase in P. falciparum, PfHAD1, (Guggisberg et al., 2014) with a putative ortholog in Toxoplasma, this gene was named HAD2a. A BLAST search identified three additional paralogs, which we subsequently named HAD2b-d (HAD2b: TGME49_246110; HAD2c: TGME49_271960; HAD2d: TGME49_223870). HAD2a-d share a conserved motif of ~200 amino acids (Fig. 1B) flanked on both the N- and C-terminus with unique regions. The conserved region is shared among the Alveolates and is characterized by a DxDx(T/V) motif (Fig. 1C, red; Fig. S2) representative of the HAD superfamily of phosphotransferases (Burroughs et al., 2006; Allen and Dunaway-Mariano, 2009; Seifried et al., 2013). The majority of these evolutionarily wide-spread enzymes are phosphatases that use Asp as a nucleophile (the first residue in the DxDx(T/V) motif) in a Mg²⁺-dependent fashion for catalytic activity (Allen and Dunaway-Mariano, 2009). Consistent with this, the three Asp residues known to associate with the Mg²⁺ ion are conserved in this group of Toxoplasma proteins (Fig. 1C, red). Moreover, a modelling search of crystal structures deposited on SwissProt provided the best model for HAD2a and HAD2b using the crystal structure of a HAD family hydrolase from Pseudomonas aeruginosa (Fig. S3). Furthermore, the key residues for the four conserved motifs defining the HAD hydrolases according to Burroughs et al. (Burroughs et al., 2006) are conserved in this family of Toxoplasma HADs (Fig. 1C, yellow/red). The unmodeled loop between motifs I and II is known as the cap domain (Fig. 1C), which endows phosphatase specificity and impacts the ability to recognize either micromolecule or macromolecule as substrates (Burroughs et al., 2006; Seifried et al., 2013). Between the four members of this HAD family we noticed a considerable divergence in the cap domain suggesting different substrate specificities. Collectively, these data imply that this group of proteins is a family of phosphatases.

HAD2a and HAD2b display phosphatase activity

As shown in Fig. 1C, sequence analysis and modelling identified a HAD domain conserved across this family of proteins. To provide direct evidence for the predicted phosphate hydrolase activity we expressed and purified recombinant 6xHis-tagged HAD2a and HAD2b proteins and performed phosphatase activity assays. We also generated constructs wherein the first nucleophilic Asp residue in the conserved DxDx(T/V) motif is mutated into an Ala residue (HAD2a^D262A and HAD2b^D171A), a mutation known to abrogate phosphatase activity (Collet et al., 1999; Li et al., 2003). We assessed the purity of the proteins by SDS-PAGE and Coomassie Brilliant Blue staining and demonstrated that the purified HAD2a and HAD2b proteins, besides some contaminants in the case of HAD2a, display strong protein bands around the predicted 64 and 42 kDa protein sizes respectively (Fig. S4). Expression and purification of HAD2a^WT and HAD2b^WT as well as their respective active site mutants resulted in comparable purity within the pairs (Fig. S4). We then used para-Nitrophenyl Phosphate (p-NPP) to test for enzyme activity of the recombinant proteins (Takai and Mieskes, 1991). p-NPP is a non-specific substrate for alkaline and acid phosphatases with a demonstrated ability in detecting Ser/Thr as well as Tyr protein phosphatase activity (Zhang et al., 1992; Zhuo et al., 1993). Using this assay, we readily detected phosphatase activity for HAD2a and HAD2b, which was abolished in the active site mutants (Fig. 2). To test which co-factors are prerequisites for phosphatase activity we utilized the known cations Mg²⁺ and Mn²⁺ (Zhuo et al., 1993) and detected a vital dependence on Mg²⁺ for catalytic activity of both HAD2a and HAD2b (Fig. 2). Taken together, HAD2a and HAD2b have demonstrable Mg²⁺-dependent phosphatase activity consistent with their HAD domain.

HAD2a dynamically associates with the mostly basal cytoskeleton

The strong Y2H interaction between HAD2a and MORN1 together with their closely matched mRNA expression profiles suggested that these proteins might localize to the same subcellular compartments in vivo. To test this we first determined the dynamics of HAD2a and HAD2b localization by fluorescence microscopy. We generated specific polyclonal HAD2a and HAD2b antisera raised against the full-length 6xHis-tagged proteins to study their localization. The antisera are specific as in Toxoplasma lysates the HAD2a and HadD2b antisera recognize bands of 64 and ~40 kDa, corresponding with predicted molecular weights of 64.5 and 41.7 kDa respectively (Fig. S5A and B). We assessed the dynamics of HAD2a throughout daughter cell budding using IMC3 as a development marker for both mother and daughter cortical cytoskeletons. In nearly all parasites we observe HAD2a at the basal complex of the mother (Fig. 3A, asterisks). In early daughter buds we observed HAD2a predominantly in the forming basal and apical complexes (Fig. 3A, early budding, arrows). During mid-budding, the signal disperses slightly over the whole nascent daughter cytoskeleton though remains associated with the apical and basal complex (Fig. 3A, mid budding, arrow). In late buds when the basal complex is constricting, HAD2a transitions into a nearly exclusive association with the basal complex of the daughter buds (Fig. 3A, late budding, arrows). To determine whether HAD2a co-localizes with MORN1 we co-stained parasites expressing Myc2-MORN1. Indeed, we observe co-localization of both proteins at the basal complex of mother (asterisk) and daughter (arrow) cells (Fig. 3B). To assess whether the localization of HAD2a to the basal complex is dependent on MORN1 we stained for HAD2a in an inducible MORN1 knockdown cell line (Lorestani et al., 2010). Upon the loss of MORN1, HAD2a still localizes to the cortex of the daughter buds; however, we can no longer detect HAD2a signal at the basal complex (Fig. 3C, +ATc, arrows). This indicates that HAD2a localization to the daughter buds is independent of MORN1 but that HAD2a association with the basal complex requires MORN1. Taken together, HAD2a dynamically associates with the daughter cytoskeleton in early buds but transitions to the basal complex when it starts to contract and is retained in the basal complex in the majority of mature parasites.

HAD2b associates with the apical cytoskeleton

We determined that HAD2b also strongly interacts with MORN1 in Y2H experiments (Fig. 1A). We also assessed whether HAD2a and HAD2b interacted with each other, but we did not observe any interaction between these two proteins (Fig. 1A). By IFA HAD2b localizes to the apical end of the cortical cytoskeleton in both mother (arrowhead) and daughter (arrow) parasites (Fig. 3D). During endodyogeny HAD2b is present in the early daughter buds (Fig. 3D, early budding, arrows). In emerging parasites we also recognized a strong HAD2b signal at the base of daughter buds where the mother's IMC is disassembling (Fig. 3D, late budding, arrowhead). Such a signal is suggestive of association with the conoid as the conoid is not disassembled till very late in division and is observed as a signal migrating in an apical to basal direction with the retracting mother cytoskeleton (Morrissette and Sibley, 2002). To further refine to which apical cytoskeleton structure HAD2b localizes we co-stained with markers for IMC sub-compartment protein 1 (ISP1), which localizes to the apical cap alveolus of the IMC (Beck et al., 2010), and RNG1, which localizes to the apical polar ring serving as the sub-pellicular microtubule organizing centre at the base of the conoid (Tran et al., 2010) (Fig. 3E). We observed that HAD2b sits apical of the ISP1 signal in mature parasites. On the other hand, the HAD2b signal partially overlaps with the RNG1 signal, but RNG1 extends further apically. Together, these data indicate that HAD2b is not part of the conoid but is associated with the very end of the IMC where it connects to the apical polar ring. Notably, MORN1 also has a minor presence at this position (Gubbels et al., 2006; Hu et al., 2006), suggesting that this localization could be dependent on MORN1. In conclusion, HAD2b associates with MORN1 by Y2H and localizes to the apical end of the IMC at the interface with the conoid.

HAD2a is an essential gene in Toxoplasma

Because our primary interest was to identify new proteins and mechanisms of the contractile-related machinery in T. gondii, we focused our efforts on HAD2a because of its co-localization with MORN1 during endodyogeny. We engineered a conditional knockout cell line for HAD2a using conditional diCre-LoxP technology (Andenmatten et al., 2012). In this LoxP-HAD2a line the had2a ORF is replaced with a LoxP flanked, Ty-tagged had2a allele that can be excised upon activation of diCre-mediated recombination by rapamycin treatment. We exchanged the single YFP ORF against a double YFP in the transfection plasmid to enhance detection of YFP when expressed under the endogenous had2a promoter following diCre recombination (Fig. 4A). A clonal LoxP-HAD2a line was established, and successful gene-swap and genomic integration of the transfected plasmid were verified by diagnostic PCR (Fig. 4B). We verified expression of the Ty-tagged HAD2a allele by western blot using Ty and the specific HAD2a antisera, which confirmed a protein of ~64 kDa in the LoxP-HAD2a as well as in a diCre control cell line (Fig. S5B). To induce recombination we incubated LoxP-HAD2a parasites overnight with 50 nM rapamycin and detected an average of ~10% recombination efficiency. Loss of Ty-HAD2a signal and expression of YFP-YFP in had2aKO parasites were confirmed by IFA experiments using anti-Ty and anti-GFP antisera (Fig. 4C, S5C). Loss of HAD2a did not result in a change of HAD2b expression and localization (Fig. S5D), indicating that there is no phenotypic compensation between these two proteins. To obtain sufficient YFP-positive had2aKO mutants for subsequent phenotypic characterization we applied fluorescence-activated cell-sorting (FACS) resulting in highly enriched YFP-positive parasite samples with only a minor (<10%) YFP-negative parasite population. Excision of the had2a ORF was verified by diagnostic PCR (Fig. 4D), and sorted parasites were used for standard plaque-assays taking the YFP-negative samples as controls for parasite fitness. Plaque assays revealed a severe proliferation defect for YFP-positive parasites, shown by 85% less plaques formed in comparison to control parasites (Fig. 4E). Residual plaques in YFP-positive samples could be assigned to residual YFP-negative parasites as evidenced by fluorescence microscopy performed four days after FACS sorting (Fig. 4F). Taken together, our results demonstrate an essential function for HAD2a in the lytic replication cycle of T. gondii.

had2aKO parasites display a cytokinesis defect

Having established an essential role for HAD2a we dissected the phenotypic consequences of had2a loss. We compared intracellular division of had2aKO parasites to LoxP-HAD2a controls using IMC3 antiserum 24 h after host cell invasion and observed a severe division defect in the had2aKO population (Fig. 5A). Defects first appeared in intracellular vacuoles harbouring ~4 parasites. IMC3 staining revealed enlarged had2aKO parasites with two separated nuclei in the absence of daughter buds. Furthermore, we observed vacuoles in which had2aKO parasites passed several rounds of budding eventually leading to a distorted shape, lacking the rosette appearance of LoxP-HAD2a control parasites (Fig. 5A). When quantified 24 and 48 h after invasion, we found 57% and 82% respectively of the counted had2aKO vacuoles to display division defects, likely produced by incomplete separation of dividing tachyzoites (two nuclei per parasite in absence of large daughter buds and/or distorted shape in intracellular vacuoles) (Fig. 5B).

The sequence in Fig. 5A suggests an escalating phenotype rather than a clean arrest. To best understand the role of HAD2a the most information lies in the first step wherein the phenotype manifests, which differentiates subsequent secondary defects. Because the earliest changes in morphology and budding are reminiscent of the previously described phenotype for parasites with a loss of MORN1, where the basal complex fails to assemble and results in severely impaired daughter cell segregation (Heaslip et al., 2010; Lorestani et al., 2010), we focused our phenotype dissection on this step in had2aKO phenotype development. To validate the conjoined appearance of the had2aKO parasites we used SAG1 antiserum to visualize the parasite's plasma membrane and RNG1 antiserum to highlight the parasite's apical end (Fig. 6A). We found conjoined parasites carrying two apical ends as highlighted by RNG1 staining (Fig. 6A, arrows) but lacking a basal end, a characteristic previously reported for the double-headed and multi-headed parasites upon loss of MORN1 (Heaslip et al., 2010; Lorestani et al., 2010). We further confirmed the formation of double-headed parasites with the apical marker ISP1 (Fig. 6A, lower panel) and validated the continuous connection between the two ‘heads’ by differential interference contrast (Fig. 6A and Fig. S5C and D). To verify that the observed had2aKO conjoined parasites completed budding we stained with anti-IMC1 and anti-GAP45 antibodies. GAP45 is associated with the IMC only when division is concluded, allowing us to distinguish between mother and daughter cytoskeletons (Gaskins et al., 2004). In LoxP-HAD2a control parasites a closure of the basal end that separated one cell from the other was always observed (Fig. 6B, asterisks), contrary to had2aKO conjoined parasites that lacked a tapered closure and displayed two completely segregated nuclei in one continuous cell (Fig. 6B). The absence of large daughter buds indicated that conjoined parasites had completed budding and the presence of small daughter buds implied that the next round of budding had initiated even though parasites failed to conclude cytokinesis of the previous cycle (Fig. 6B, arrows). We also noticed a gap in had2aKO parasites that was neither stained by IMC1 nor GAP45 (Fig. 6B, arrow head), indicating that there are two separate cytoskeletons and that the parasite failed to close the cytoskeletons after the mother cell's cytoskeleton disassembled.

By counting the number of parasites per vacuole 24 h after host cell invasion we found aberrant parasite numbers almost exclusively in had2aKO vacuoles compared with LoxP-HAD2a vacuoles that displayed regular endodyogeny and geometric increase in numbers (Fig. 6C). Together, these data strongly suggest that had2a knockout impairs the conclusion of cytokinesis in Toxoplasma endodyogeny.

had2aKO parasites fail to constrict the basal complex

The formation of conjoined parasites and imperfect basal end closure upon had2a loss suggests that HAD2a might act on the basal complex and/or MORN1 itself. To first verify that had2aKO parasites possess the scaffold of the basal complex we tested if MORN1 localization is impaired in had2aKO parasites. Hereto we stably expressed mCherry-tagged MORN1 in LoxP-HAD2a parasites. As expected, we localized mCherry-MORN1 to the spindle pole and the mature basal complex in LoxP-HAD2a parasites (Fig. 7A, basal complex marked with an asterisk). In had2aKO parasites we also localized mCherry-MORN1 to the spindle pole and detected the MORN1- labelled basal complex at either the parasites' basal end (Fig. 7A, arrow second panel) or at the mid point of cells where conjoined parasites were fused (Fig. 7A, arrow third panel). We rarely observed basal complexes in segregating parasites for the uninduced parent line (LoxP-HAD2a). This indicates that loss of HAD2a does not interfere with the association of the basal complex scaffolding protein MORN1 with the basal extremity of the IMC. Furthermore, we noticed an unusually large diameter of the basal complex for both the midpoint basal complexes and at the basal end of had2aKO parasites as compared with LoxP-HAD2a controls. Measurement of the basal complex diameter in LoxP-HAD2a controls showed an average of 0.94 µm, whereas had2aKO parasites had a much wider average basal complex diameter of 1.47 µm, with 83% of the analysed had2aKO parasites displaying a basal complex width greater than 1 µm (compared with only 29% of the LoxP-HAD2a controls) (Fig. 7B). Thus, these data indicate that some had2aKO parasites are still able to complete cytokinesis but with incomplete basal complex contraction. Mechanistically, MORN1 recruitment does not appear to be hampered in had2aKO parasites but basal complex contraction is impaired.

The Toxoplasma basal complex has been resolved into three sub-compartments. Starting from the apical end, a ring harbouring MORN1, IMC9 and IMC13 is followed by a structure harbouring IMC5 and IMC8, which in turn is bordered at the very basal end by Centrin2 (Anderson-White et al., 2011). We demonstrated the presence of MORN1 and asked next whether the other structures are formed. As a representative of the middle compartment we endogenously tagged IMC8 with a triple Myc tag in LoxP-HAD2a parasites and determined its localization after induction of the had2aKO (Fig. 7C). We observed the previously reported basal localization in uninduced parasites (Fig. 7C, asterisks), whereas loss of HAD2a resulted in a clustered IMC8 Myc signal at the mid-point (Fig. 7C, arrowhead). These data indicate that this basal complex component is still recruited to the right place, but does not assemble into a proper ring.

To test for Centrin2 localization in had2aKO parasites, we stably transfected the LoxP-HAD2a cell line with a Myc2-Centrin2 expression plasmid and induced the had2aKO. We confirmed Centrin2 localization at the apical end, the apical annuli, the centrosome and the basal complex in the uninduced parasites (Fig. 7D) (Hu et al., 2006; Hu, 2008). In the conjoined phenotype we detected Centrin2 at the apical end and the apical annuli at both ends of double-headed had2aKO parasites (Fig. 7D, arrows). However, we were unable to detect Centrin2 at the midpoint of conjoined parasites where we expected to find the basal complex (Fig. 7D, arrowhead). However, the typically weak signal of Centrin2 at the basal complex (Hu et al., 2006; Hu, 2008) might be below the detection limit and therefore this result in not conclusive.

In summary, the absence of HAD2a does not affect recruitment of MORN1 to the basal complex but results in defective formation of the IMC8 sub-compartment. Moreover, the absence of HAD2a results in an overall wider basal complex diameter. Thus, HAD2a operates downstream of MORN1 recruitment to the basal complex in properly assembling basal complex sub-structures and HAD2a absence results in defective basal complex contraction and a cytokinetic defect (Fig. 7E).

The Plasmodium HAD2 ortholog resides in the cytokinetic ring of budding merozoites

Using a standard BLASTP approach we identified a P. falciparum protein that although divergent at the N- and C-terminus, showed considerable identity to the HAD-like domain of TgHAD2a (Fig. S6). The identified gene product (PF3D7_1205200) is annotated as a putative HAD domain ookinete protein (PlasmodDB.org) and gene expression peaks during late schizogony of the Plasmodium asexual replication cycle, consistent with a putative role in cell division (Bozdech et al., 2003). It is furthermore recognized as part of the established Plasmodium invasion-related network, PlasmoINT (Hu et al., 2010), which contains both invasion and division-related proteins. To approach the subcellular localization of Plasmodium HAD2 (PfHAD2) we replaced the endogenous ORF with a GFP-tagged version in 3D7 P. falciparum parasites (Fig. 8A). Successful replacement of the 3′ region was verified by diagnostic PCR (Fig. 8B) and subcellular localization assessed by microscopy (Fig. 8C). PfHAD2 was first visible in schizont stage parasites (~10 nuclei) localizing to several ring-like structures at the periphery of the mother cell (Fig. 8C, early budding). With proceeding schizogony, the PfHAD2 rings expanded and engulfed the separated nuclei and finally moved to the basal end of the nascent merozoites (Fig. 8C, mid budding). Upon conclusion of schizogony, PfHAD2 rested basal of the nascent merozoite's nuclei (Fig. 8C, late budding). The dynamic of PfHAD2 was further confirmed by co-localizing the protein with a marker of the parasite's IMC, GAPM2 (Bullen et al., 2009) (Fig. 8D). PfHAD2 first appeared as a wider ring compared with GAPM2 in early schizogony and eventually guides the IMC during merozoite formation to rest at the basal end of the cell (Fig. 8D, early – late budding). Thus, although considerably different in the amino acid sequence, Plasmodium HAD2 is shown to participate in the budding process, implying a conserved function for HAD2 in apicomplexan cell division.

Toxoplasma strains

All Toxoplasma parasites used are RH strain derivatives and were grown in confluent human foreskin fibroblast (HFF) cells (Roos et al., 1994). Parasite transfections and selections using 1 μM pyrimethamine, 20 μM chloramphenicol, or a combination of 25 mg ml⁻¹ mycophenolic acid and 50 mg ml⁻¹ xanthine were performed as described previously (Roos et al., 1994).

Toxoplasma plasmid constructs

All primer sequences are provided in Supplementary Table S2. All exogenous YFP or Myc2-tagging plasmids were based on replacing functional modules flanked by the restriction enzymes between brackets in the following plasmid: (PmeI)ptub-(BglII)YFP(AvrII)YFP(EcoRV)/sagCAT or dhfrDHFR-TS (Anderson-White et al., 2011). Insertional tagging of genomic loci was performed by gene-swap in RH ku80::diCre parasites (Andenmatten et al., 2012). For the gene-swap approach, the p5RT70 promoter sequence was replaced by a DNA fragment encoding 2500 bp upstream of the annotated had2a start ATG via the KpnI and EcoRI restriction sites. The respective gene coding sequence was placed between the AvrII and PacI sites, a fragment of 2500 bp after the annotated stop codon was cloned in the gene-swap plasmid using the SacI sites.

To generate recombinant His₆ N-terminal fusion proteins, full-length HAD2a and HAD2b CDS sequence were PCR-amplified and cloned into the pAVA0421 plasmid by LIC (Alexandrov et al., 2004). The Asp to Ala point mutants were generated by site directed mutagenesis.

Endogenous Myc tagging of IMC8 was performed by cloning 1.8 kb of the genomic DNA encoding the 3′-end of the ORF into (PmeI)ptub-(AvrII)Myc3(−)3′-dhfr(NotI)/DHFR (unpublished) via PmeI/AvrII cloning sites. The resulting plasmid was linearized by NheI before transfection of LoxP-HAD2a parasites.

All YTH constructs were made using Gateway Technology (Invitrogen). In short, full-length MORN1 (primer pair AttB1-F-MORN1-ATG and AttB2-R-MORN1-fl), the N-terminal (AttB1-F-MORN1-ATG and AttB2-R-MORN1-M6, including the 6th MORN domain) or the C-terminal half (AttB1-F-MORN1-M13and AttB2-R-MORN1-fl, from the 13th MORN domain onward) of MORN1 were amplified in a two-step PCR with the first round primer pairs indicated between brackets and the second round with universal AttB adapter primers Att-B1-adopt and Att-B2-adopt to complete the AttB-sites. To clone full-length HAD2a and HAD2b YTH constructs we used first round primer pairs AttB1-F-HAD2a/AttB1-R-HAD2a and AttB1-F-HAD2b/AttB1-R-HAD2b respectively. The amplicons were cloned by BP recombination into the pDONR201 vector. Subsequently, the inserts were transferred into plasmids pGBK-Att or pGAD-Att vectors by LR recombination. Plasmids pGBK-Att and pGAD-Att were derived from pGBKT7 and pGADT7 plasmids (Clontech), respectively, and converted into Gateway compatible vectors by cloning the reading frame B cassette (Invitrogen) into the SmaI site of each plasmid.

Yeast two-hybrid screen

Saccharomyces cerevisiae strain AH109 (MATa, trp1-901, leu2-3, 112, ura3-52, lys2-181, his3-200, ade2-101, gal4-952, gal80-538, LYS2::GAL1uas-GAL1tata-HIS3, GAL2uas-GAL2tata-ADE2, URA3::MEL1uas-MEL1tata-lacZ) (Clontech) was used. Yeast was transformed using the LiC₂H₃O₂ method with either 2.5 µg of bait plasmid mixed with 1.0 µg of a pGAD-‘GOI’-AD prey plasmid or the RH two-hybrid cDNA library (Fig. S1). The library was generated from random cycling tachyzoites. Oligo-dT primed polyA+ RNA was directionally cloned by EcoRI/XhoI digestion into plasmid pGBK-Att. The library complexity was 2.4 × 10⁶ recombinants with an average insert size of 1.65 kb and an empty plasmid background of under 1%.

Transformed yeast were plated on SD/Trp⁻/Leu⁻ plates and incubated at 30°C until appearance of colonies. Plates were then replica-plated to a quadruple-dropout selective medium (QDO) (SD/His⁻/Ade⁻/Trp⁻/Leu⁻). Growing colonies were patched onto new QDO plates, which were subsequently replica plated to QDO/X-α-Gal and QDO/3-aminotriazole (3-AT) at 5, 10, 25 and 50 mM. Colonies growing on QDO/X-α-Gal and at least QDO + 10 mM 3-AT plates were selected as positive clones. Prey plasmids were isolated from positive colonies grown overnight in 5 ml of QDO/Leu⁻ medium at 30°C using published methods (Hoffman and Winston, 1987) and transformed into E. coli DH5α. Minipreps were screened by HindIII digestion and distinct clones sequenced from the 5′-end insertion site using the standard T7 primer. Sequence results were analysed by BLAST searching the Toxoplasma genome (www.toxodb.org). For clones of interest the library insert was sequenced using primer T7 (refer to Supplementary Data file for all collected sequences).

Because full-length MORN1 resulted in autoactivation, two MORN1 baits (N-terminal: start till MORN domain 6; C-terminal MORN domains 13–15; Gubbels et al., 2006) were used independently to screen the tachyzoite cDNA Y2H library (Fig. S1). Approximately 120 000 total yeast transformants were screened for YTH interactions. Forty four total cDNAs were isolated following these screens and subjected to DNA sequencing numbered with MNIC or MCIC prefixes representing MORN1 N- or C-terminal interacting clone respectively. Nine genes were chosen for secondary screening by YFP-tagging and fluorescent microscopy. Results are summarized in Table S1.

Sequence analysis

HAD2a-d sequences were aligned with ClustalW (standard settings), manually adjusted and coloured with BoxShade (50% identity and similarity settings). For phylogram construction, the extended HAD domain as shown in the alignment in Fig. 1C was BLASTP searched against the NCBI Nr database. Representative sequence fragments covering the spectrum of hits were picked and aligned by ClustalW (standard settings). The phylogram was assembled in ClustalW (standard settings). The 3D-Model was generated by model searches against structures deposited on Swiss-model (http://swissmodel.expasy.org) using automated mode (Arnold et al., 2006). The obtained coordinates were visualized using PyMOL (www.pymol.org).

Recombinant protein expression and polyclonal antisera generation

His₆-HAD2a expression was induced for 4 h with 1 mM IPTG at 37°C in BL21-DE3-pLysS Escherichia coli grown in 2xYT medium. The protein was purified from inclusion bodies as follows: the E. coli pellet is resuspended in lysis buffer (20 mM Tris HCl pH8.0, 25% sucrose, 1 mM EDTA) and sonicated. An equal volume of wash buffer 1 (200 mM NaCl, 1% deoxycholic acid, 1% NP-40, 20 mM Tris pH 7.5, 2 mM EDTA) was added and then pelleted at 5000*g for 10 min at 4°C. The pellet was resuspended in wash buffer 2 (0.5% TX-100, 1 mM EDTA), followed by addition of an equal volume of wash buffer 1 and centrifugation (repeated a total of two times). The purified and pelleted inclusion bodies were then refolded and solubilized using 8 M urea in 50 mM Tris pH 7.6, 1 mM β-mercaptoethanol, 0.025% Tween 20 for 3 h by end-over-end rotation at 4°C. The supernatant was collected following centrifugation (1500*g for 20 min at 4°C) and dialysed for 8–16 h periods at 4°C subsequently against 5, 4, 3, 2, 1, 0 M urea in 50 mM Tris pH 7.6, 1 mM β-mercaptoethanol, 0.025% Tween 20, followed by a final 24 hr dialysis against 0 M urea. His₆-HAD2b expression and bacterial lysis was identical to His₆-HAD2a. An end-concentration of 10 mM imidazole was added to the cleared lysate and applied to pre-a equilibrated Talon column (Clontech). The column was washed subsequently with 10 and 20 mM imidazole in wash buffer and the protein eluted with the supplied elution buffer containing 150 mM imidazole. All purified proteins were snap frozen and stored at −80°C till further use.

Polyclonal antisera were generated by rabbit immunizations (Covance, Denver, PA). HAD2a antisera were affinity purified against His₆-HAD2a recombinant protein using published methods (Gubbels et al., 2006).

Toxoplasma immunofluorescence microscopy

LoxP-HAD2a parasites were incubated with 50 nM rapamycin for ~16 h to induce Cre-recombination and fixed for immunofluorescence after ~24 h with methanol or 4% PFA (permeabilized with methanol) as described previously (Anderson-White et al., 2011). The following antibodies were used: rat α-IMC3 (1:1000; Anderson-White et al., 2011); mouse α-RNG1 (1:1000, Tran et al., 2010; kindly provided by Naomi Morrissette, UC Irvine); MAb α-cMyc (1:50, Santa Cruz Biotech); MAb α-Ty (1:10, Bastin et al., 1996), kindly provided by Markus Meissner, University of Glasgow, Chris de Graffenried, Brown University and Peter Bradley UCLA; rabbit α-CherryRFP (1:200 000, kindly provided by Iain Cheeseman, Whitehead Institute); rabbit α-GFP (1:1000, Torrey Pines Biolabs); mouse α-GFP (1:500, Roche Diagnostics); rabbit α-GAP45 (1:1000, Con Beckers University of North Carolina); mouse α-IMC1 (1:1000, Garry Ward, University of Vermont); mouse α-ISP1 (1:500, Peter Bradley, UCLA); Alexa594 labelled α-SAG1 (1:500, J.-F. Dubremetz, Université de Montpellier) and Streptavidin labelled Alexa 594 (1:1000, Invitrogen) in blocking solution. Secondary antibodies were conjugated to the Alexa Fluor dyes (Invitrogen) and DNA was stained with 4′,6-diamidino-2-phenylindole (DAPI). Imaging was performed on a Zeiss Axiovert 200 M wide-field fluorescence microscope equipped with standard DAPI, FITC, YFP and TRTC filter sets, a α-Plan-Fluar 100x/1.45 NA oil objective or with a Zeiss Axio Imager Z2 with ApoTome.2. Images were analysed and processed using Volocity (Improvision/Perkin-Elmer), Zeiss ZEN and ImageJ software.

Western blot

Western blots were performed as previously reported (Gubbels et al., 2006). Briefly, LoxP-HAD2a and DiCre control parasites were harvested by needle-lysing host cells ~24 h after host cells were infected. For each sample, 15 × 10⁶ parasites were loaded on a 12% NuPAGE Bis-Tris gel (Invitrogen) and proteins transferred to a PVDF membrane (BioRad). After blocking with a 6% milk solution, blots were probed with mouse anti-Ty, 1:500 (Peter Bradley, UCLA) or mouse anti-IMC1, 1:6000 ((Mann and Beckers, 2001); kindly shared by Gary Ward, University of Vermont). After incubation with goat anti-mouse HRP, 1:10000 (Molecular Probes) membranes were developed with chemiluminescent HRP substrate (Millipore) and exposed to X-ray film. The same membranes were stripped and re-probed with rabbit anti-HAD2a or rabbit anti-HAD2b, 1:5000, with subsequent decoration of goat anti-rabbit HRP, 1:3000 (Molecular Probes).

Parasite sorting (FACS)

Cre-recombination in LoxP-HAD2a parasites was induced for 16 h by addition of 50 nM rapamycin to parasites inoculated in HFF cells. After incubation, parasites were mechanically released by passing them through a 26G needle and filtered through a 3 µm polycarbonate filter. YFP⁻ and YFP⁺ parasites were sorted using a Becton Dickinson FACSAria cytometer (BD Biosciences) and recovered in PBS containing 15% FBS. YFP was excited using a 488 nm Argon blue laser and detected with a 530/30 nm band pass filter. Parasites were gated on FSC/SSC. Parasites were washed with ED1 media before plaque assays were inoculated. For microscopy of YFP⁺ parasites four days after FACS, parasites were inoculated in Lab-Tek II Chambers (Thermo Fisher) confluent with HFF cells and kept under standard conditions until observed with a Zeiss Axiovert 200 M wide-field fluorescence microscope and α-Plan-Fluar 63x/1.40 NA oil objective.

Basal complex diameter measurement

To measure the basal complex diameter, we induced had2a knockout in LoxP-HAD2a parasites co-expressing mCherry-MORN1 with 50 nM rapamycin overnight. Parasites were fixed in host cells after ~24 h with 4% PFA and permeabilized with methanol. mCherry-MORN1 was stained with anti-mCherry antibody, and 3D-images of vacuoles harbouring 2–4 parasites were taken with a Zeiss Axio Imager Z2 with ApoTome.2. mCherry-MORN1 signal was measured in total projection of 3D-Images using ImageJ software.

Phosphatase activity assays

Recombinant HAD2a^D262A and HAD2b^D171A were purified in parallel with wild type proteins under identical conditions (refer to recombinant protein expression). Proteins were separated with 12% NuPage SDS-PAGE (Invitrogen) and stained with Coomassie Blue. Phosphatase activity was assessed by conversion of the phosphatase substrate p-NPP (Sigma), based on published methods (Takai and Mieskes, 1991; Zhuo et al., 1993). One microgram of recombinant protein (His₆-HAD2a, His₆-HAD2b and respective D- > A mutants) was incubated with 27 mM p-NPP in 50 mM HEPES (pH 7.5), 1 mM DTT for 4 h at RT and OD_{405 nm} of triplicate reactions was measured in a M5 SpectraMax spectrophotometer (Molecular Devices). Reactions were performed in presence of either 5 mM MgCl₂, 0.5 mM MnCl₂ or a combination of both salts and background corrected values are depicted as percentage of control (in absence of cation).

Plasmodium nucleic acids and construct

To tag Plasmodium had2 with GFP via homologous recombination, we amplified 864 bp of the had2 ORF 3′end from Plasmodium 3D7 genomic DNA and cloned it into the NotI/AvrII restriction sites of the pARL-1a-GFP expressing plasmid releasing the crt promoter (Struck et al., 2005). Single cross-over integration was confirmed by PCR. All primers are listed in Table S2.

Plasmodium parasite strain and transfection

P. falciparum 3D7 parasites at asexual stages were cultured in human O⁺ erythrocytes as previously published (Trager and Jensen, 1976). Parasites were transfected at the ring stage with 100 µg of purified plasmid DNA as described previously (Fidock and Wellems, 1997). We selected for transfected parasites using 2.5–5 nM WR99210 (Jacobus Pharmaceuticals) and obtained parasites with single cross-over integration in the had2 locus by growing them in alternating cycles of WR99210 presence/absence.

Plasmodium microscopy

Images of unfixed Plasmodium schizonts expressing PfHAD2-GFP were captured using a Zeiss Axioskop 2 Plus microscope with a Hamamatsu camera (Model C4742-95, Zeiss axiovision). DNA was stained with 1 µg ml⁻¹ DAPI (Roche). To co-localize PfHAD2-GFP with the IMC marker protein GAPM2, we fixed schizont stage parasites with 4% formaldehyde and 0,0075% glutaraldehyde as previously published (Tonkin et al., 2004). Parasites were permeabilized with 0.1% Triton X100 and blocked with 3% bovine serum albumin and then incubated with anti-GAPM2 antibody (1:500; Kono et al., 2012). DNA was stained with DAPI.