To obtain a broad overview of proteins localized to the parasite nucleus we identified the protein content of crude nuclear preparations followed by further biochemical fractionation from ring, trophozoite, and schizont stage parasites by high accuracy mass spectrometry (Additional file 1). Isolated nuclei were significantly increased in size compared to intact parasites, a phenomenon regularly seen in preparations of nuclei after hypotonic lysis 5758 (Figure 1a). Despite extensive washing of the nuclear pellet, contamination with free hemozoin crystals was obvious. Analysis by transmission electron microscopy (TEM) showed that the nuclear fraction mainly consisted of rounded and enlarged nuclei with varying degrees of intactness (Figure 1b). No consistent organellar impurities were apparent, although some haemozoin crystals were also visible throughout these samples. Importantly, immunolabeling with antibodies specific for histone 3 lysine 4 tri-methylation (H3K4me3) identified these structures as parasite nuclear material (Figure 1c).

We next performed protein fractionation to reduce sample complexity. Nuclei were serially extracted with 0.1 M KCl (fraction 2), DNAseI (fraction 3), 1 M KCl (fraction 4), and 2% SDS (fraction 5). Each fraction displayed a distinct 1D-SDS-PAGE protein pattern indicative of differential protein extraction (Figure 1dand Additional file 1). The cytosolic enzyme glyeraldehyde-3-phosphate dehydrogenase (GAPDH) and the mitochondrial heat shock protein 60 (HSP60) were exclusively detected in the NP40-soluble cytoplasmic extract (fraction 1) demonstrating efficient lysis of the parasite plasma membrane and the double membrane-bound mitochondrion (Figure 1e). In contrast, H4 and PfHP1 were soluble only after extraction of nuclear pellets with high salt and SDS, consistent with their intimate association with chromatin. In summary, the microscopy-, SDS-PAGE-, and immunodetection-based assessments show that the protocol applied here efficiently separated the cytoplasmic and nuclear compartments and yielded distinct subnuclear protein fractions suitable for mass spectrometry analysis.

Endoproteinase LysC- and trypsin-digested total proteins were analyzed by two-dimensional capillary liquid chromatography and tandem mass spectrometry (MudPIT). MS/MS spectra were searched against a combined P. falciparum/human proteome database using the SEQUEST algorithm 59. We identified 1,518 different parasite proteins that were represented by at least one peptide in any of the 30 samples analyzed (fractions 1 to 5 each for ring stages, trophozoites, schizonts; two biological replicate samples each) (Additional file 2). Similar numbers of proteins were identified for ring stages (1,050), trophozoites (1,017), and schizonts (1,092). A total of 649 proteins were shared by all three stages, and similar numbers of proteins were either unique to one stage or shared between any two of the three stages (Figure 2a). In the cytoplasmic and combined nuclear fractions 870 and 1,273 proteins were detected, respectively, and 625 were shared between both compartments (Figure 2b). Figure 2c shows the distribution of proteins in individual nuclear fractions (DNAseI- and high salt-soluble fractions were combined into one category potentially enriched in DNA/chromatin-associated proteins). The percentages of proteins measured in the low salt, DNAseI/high salt and SDS fractions were 20.1%, 25.5% and 43.6%, respectively, and were unique to the corresponding fraction.

On average, 69.7% (+/- 14.7 SD) of proteins measured in a given sample were also detected in its matching replicate. Furthermore, calculation of Spearman rank coefficients showed that each sample correlated best with its matching replicate (Additional file 2). This clearly underlines the repeatability of the fractionation and the mass spectrometric methods applied. The absolute correlations between replicates were moderate (0.50 +/- 0.09 SD; Spearman rank coefficients). This can be explained by the stochastic sampling of peptides in the mass spectrometer, a limitation inherently associated with shotgun proteomics of complex samples, and by biological and experimental sample-to-sample variation. Undersampling is a consequence of the limited speed with which MS/MS spectra can be acquired and is specific for the instrument used. Both factors contribute to the differential detection/identification of proteins, particularly for low abundance proteins.

To maximally cover the nuclear proteome, we accepted protein identifications based on single peptide matches. For reliable peptide identifications we used the decoy search strategy querying a reversed sequence database 60. This strategy resulted in very low mean false discovery rates of 0.0028 +/- 0.0024 SD per sample, reflecting the stringent search criteria applied for peptide identification and acceptance of single peptide hits (Additional files 1 and 2). Furthermore, random inspection of single peptide-based protein identifications confirmed the automated protein identification of the search engine. In light of these high confidence identifications, proteins represented by single peptides were included in all downstream analyses. The SEQUEST output data for all stages and fractions (Additional files 3, 4, and 5) and a comprehensive summary table (Additional file 6) are provided as supplementary information.

Functional enrichment analyses using the David and the GOstat tools 6162 revealed that in all cases proteins detected in the combined nuclear fractions only were statistically enriched in annotations consistent with nuclear functions, whereas proteins found in the cytoplasmic fraction only were enriched in functions known to be cytosolic (Additional files 1 and 7). As shown in Table 1, all annotated histones were detected and their distribution was marked by a clear enrichment in the nuclear fractions. We measured 10 out of 12 subunits of the DNA-directed RNA polymerase II (RNApolII) complex, all of which were detected in fraction 4. Peptides derived from ApiAP2 factors (13 out of 27 detected) were exclusively present in nuclear fractions. Furthermore, we detected 111 out of 247 in silico predicted P. falciparum transcription-associated proteins (TAPs) 3863, 108 of which (97.3%) were present in any of the four nuclear fractions (Additional file 6).

To further assess the enrichment in nuclear proteins, we carried out comparisons to curated lists of proteins with known or likely nuclear localization. First, lists were derived from ApiLoc, a database of apicomplexan proteins previously localized by microscopy 64. At the time of analysis, it held information on 424 P. falciparum proteins, 60 of which were nuclear-localized at any stage during the life cycle (14.3%). A simple algorithm, designed to decide whether the annotation recorded in ApiLoc is consistent with a protein being found in the nucleus (Additional file 1), showed that 37 of these 60 nuclear proteins were detected in the nuclear preparation (Additional file 6). Proteins in the cytoplasmic and any of the four nuclear fractions were enriched in nuclear-localized proteins, whereas those detected exclusively in the cytoplasmic fraction were depleted (one nuclear protein detected) (Figure 3a). These enrichments were statistically significant under the assumption that the proteins with recorded localization annotation were representative (P value = 1.4 × 10^-7; Fisher's exact test). Similarly, accuracy was tested by comparison to a curated list of proteins whose localization has been inferred in peer-reviewed articles, largely through their having an experimentally localized predicted orthologue in model organisms such as Saccharomyces cerevisiae or Homo sapiens (Additional file 8). Of 465 proteins in this list, 296 (63.7%) are linked to nuclear localization, 136 of which were detected in the nuclear preparation (enrichment P value = 3.3 × 10^-5), and only eight were present exclusively in the cytoplasmic fraction (Figure 3a and Additional file 6). In summary, this overall assessment of the proteomic output highlights the enrichment of known and probable nuclear proteins in the nuclear fractions.

Next, we applied two measures, positive predictive value (PPV) and the number of proteins remaining in each set, to assess differences in content between the nuclear fractions. The pool of detected proteins annotated as nuclear in the ApiLoc database 64 and/or by literature review was used as reference set (159 proteins). As expected, the sets of proteins found in individual fractions showed varying PPVs with the least and most predictive observed in the cytoplasmic lysate and high salt nuclear extract, respectively (Figure 3b). The sum of proteins found in the combined nuclear fractions gave the largest set at a high PPV. Considering only those proteins in this set that were not also detected in the cytoplasm increased the PPV by 8% but came at too steep a cost by removing 49% of all proteins including many of nuclear and unknown localization. This shows that the cytoplasmic fraction is not especially deficient in nuclear proteins, consistent with the presence of many nuclear proteins in both compartments. In contrast, proteins found exclusively in the cytoplasmic fraction were markedly deficient in predicted nuclear proteins.

A variety of bioinformatic filters were applied to remove components inconsistent with nuclear localization. Each filter applied was motivated by a hypothesis grounded in current knowledge of the properties of nuclear proteins. We accepted filters to remove proteins: (1) carrying a predicted signal peptide (SP); (2) carrying a predicted transmembrane (TM) domain; (3) predicted to be exported to the RBC based on the presence of a PEXEL motif 6566 and/or the association of the encoding genes with PfHP1 35; and (4,5) found in two previously published P. falciparum organellar proteomic studies of Maurer's clefts 53 and the food vacuole 52. We rejected filters to remove proteins: (1) predicted to localize to the mitochondrion by PlasMit 67; (2) detected by a single peptide only; and (3) detected in the cytoplasmic fraction. Benchmarked results for each filter are shown in Figure 4, and full rationales for accepting or rejecting filters are available in Additional file 1.

Applying the five filters resulted in a set of 841 proteins with a PPV of 70% (Figure 4b). After removal of 55 proteins experimentally localized to non-nuclear compartments, re-addition of 16 known nuclear proteins, and adjustments due to altered SP annotations (Additional file 9) we present a set of 802 proteins, whereof we estimate that 76% represent true nuclear proteins (compared to 46% in the unfiltered dataset) (Figure 4b). This set is henceforth referred to as the 'core nuclear proteome' and all further analyses were carried out on this set (Additional file 10).

An overview of the contents of the core nuclear proteome is shown in Figure 4c where all proteins are categorized into broad generic classes based on functional annotation. Apart from proteins engaged in typical nuclear processes such as transcription, chromatin remodeling, DNA replication/repair, splicing, and nucleolar functions, the core nuclear proteome contains many putative RNA-binding proteins and factors involved in RNA processing and metabolism, protein folding, modification and degradation, and cytoskeletal organization. Notably, one-third of the core nuclear proteome consists of proteins with no annotated function, and only 5% of all proteins are reliably allocated to non-nuclear compartments.

To test if the core nuclear proteome will be useful in identifying true nuclear proteins we experimentally validated the subcellular localization of 28 proteins by indirect immuno-fluorescence assays (IFA). We generated transgenic parasite lines expressing full-length C-terminally tagged proteins from episomally maintained plasmids. We chose 22 nuclear protein candidates (NuProCs1-22) that were included in the core nuclear proteome, and six non-nuclear protein candidates (n-NuProCs1-6) that were excluded from the list based on the presence of predicted TM domains and/or SPs (Additional file 11). Of the 22 NuProCs, seven carried annotations indicative for nuclear localization and 15 were annotated as hypothetical proteins at the time of selection on PlasmoDB 68. Amongst the negative control set, n-NuProC1 (PF11_0099) is annotated as heat shock protein DnaJ homolog Pfj2 whereas all other n-NuProCs are hypothetical proteins with unknown function.

Sixteen of the 22 NuProCs co-localized exclusively with the DAPI-stained area of the nucleus in trophozoites (Figure 5a) (detailed IFA localization results throughout the IDC for all NuProC cell lines are available in Additional file 12). Within this group of confirmed nuclear proteins are six of the seven candidates annotated as putative nuclear proteins. A diffuse pattern was observed for NuProC1 (putative regulator of chromosome condensation) and NuProC5 (putative fork head domain protein). NuProC7 (structure-specific recognition protein) displayed a more condensed appearance throughout the IDC. NuProC2 (putative nucleolar preribosomal assembly protein) and NuProC3 (putative splicing factor) both localized to a distinct subnuclear region. NuProC6 encodes a putative bromodomain protein with a more restricted and peripheral localization in ring stages and a diffuse nuclear pattern in trophozoites and schizonts.

Ten candidates annotated as hypothetical proteins (NuProC9, 10, 11, 12, 14, 15, 17-20) were also unambiguously localized within the parasite nucleus. A diffuse pattern was evident for NuProC12, NuProC15, NuProC17, and NuProC20. NuProC18 showed a cytosolic staining in late schizont and ring stages but was clearly nuclear in trophozoites. NuProC9 changed from a rather ubiquitous to a more condensed nuclear localization upon transition from the early to late schizont stage. Both NuProc10 and NuProC11 are localized to unknown nuclear subcompartments. NuProC14 and NuProC19 were preferentially associated with the nuclear periphery in ring stages but showed a diffuse pattern in trophozoite nuclei.

Two additional proteins, NuProC4 (putative multiprotein bridging factor type 1 MBF1) and NuProC13 localized outside but in close proximity to the DAPI signal (Figure 5b). These patterns were clearly distinct from the cytosolic GAPDH signal and reminiscent of that obtained for the nuclear porin PfNUP100 28 suggesting that these two factors may be associated with the nuclear periphery. In addition, NuProC4 appeared to accumulate in the cytosol in mature trophozoites.

The remaining four of the 22 tagged NuProCs showed a fluorescence pattern inconsistent with nuclear localization (Figure 6a). NuProC21 (nucleosome assembly protein PfNAPL; PFL0185c), NuProC8, and NuProC22 co-localized to various degrees with GAPDH throughout the IDC suggesting a cytosolic/cytoplasmic localization. Interestingly, NuProC22 occurred in close proximity to the nuclear periphery in late and segmented schizonts indicating that this protein may be transiently associated with the parasite nucleus. NuProC16, currently annotated as putative PFMNL-1 CISD1-like iron-sulphur protein, localized to a cytoplasmic structure distinct from the cytosolic and ER compartments, which likely represents the mitochondrion. This conclusion is supported by the presence of an iron-binding zinc finger domain CDGSH 69 found in the outer mitochondrial membrane protein MitoNEET in vertebrates 70.

None of the six non-nuclear protein candidates localized to the nucleus. Five co-localized with PfBIP throughout the IDC, and n-NuProC3 was associated with the ER in early stages and exported into the host cell in trophozoites and schizonts (Figure 6b) (detailed IFA localization results throughout the IDC for all n-NuProC cell lines are available in Additional file 13). The strong representation of ER-localized proteins in the negative control set indicates that the crude nuclear preparation was enriched in ER-associated proteins. However, we cannot exclude the possibility that ER association of some of these proteins may be artefactual due to over-expression of the tagged proteins.

In summary, our experimental validation of the predicted core nuclear proteome tested 16 out of 22 nuclear protein candidates (72.7%) positive for true nuclear localization, plus another two that are associated with the nuclear periphery. This figure is consistent with the estimated precision of the bioinformatically reduced core set of nuclear proteins and validates this inventory as an important platform for the identification and characterization of novel parasite nuclear proteins.

To test if the two distinct subnuclear domains delineated by NuProC2 and NuProC3 represented the same compartment we generated a parasite line expressing both proteins simultaneously. IFA analysis revealed that both proteins co-localized within the same intra-nuclear compartment that was preferentially located towards the nuclear periphery in a DAPI-negative area (Figure 7a). A monoclonal antibody directed against the human nucleolar marker fibrillarin 71 that is cross-reactive with Toxoplasma gondii fibrillarin 72, specifically recognized the same compartment as NuProC2 or NuProC3 (Figure 7b). Hence, these co-localization experiments identified two novel nucleolar proteins in P. falciparum.

Using the core nuclear proteome as an input collection, we sought to detect novel functional domains enriched in nuclear proteins using an all-versus-all BLAST search followed by manual inspection (Additional file 1). Six novel domains comprised of 17 proteins warranted further inspection, all of which were conserved across Apicomplexa providing further evidence that these modules are functional.

One of these domains (approximately 80 aa) was found exclusively in ApiAP2 proteins 407374. It occurred in nine P. falciparum, 15 T. gondii, and two Babesia bovis proteins, and in one protein from each of the sequenced Cryptosporidium spp. (Additional files 14, 15 and 16). Unlike AP2 domains, this novel domain was only ever found once in each protein. Interestingly, this domain was located at the C-terminus in all but one of the P. falciparum ApiAP2s where it is found at the N-terminus (Figure 8). This relative position appears to be at least moderately conserved evolutionarily with 42 of 56 proteins having this domain located <50 amino acids away from the C-terminus. Hence, we termed this novel domain ACDC (

A

c

d

C

Four of the five other novel domains are associated with either cleavage of mRNA 3' UTRs (partial cleavage stimulation factor (CSTF) domain), transcriptional regulation (extended ELM2 domain, MYND domain), or the cytoskeleton (alveolin domain) (Additional files 1, 14, 17, 18, and 19). A further domain identified in our search was encoded by two proteins annotated as PfNUP100 (PFI0250c) and a hypothetical protein (PF14_0442). The localization of NUP100 to the nuclear membrane 28, and the fact that much of the conservation between the two proteins lies in the phenylalanine-glycine (FG) pairs of amino acids, suggested that both are FG-repeat nuclear pore components 75 (Additional file 20). To verify if PF14_0442 encodes a nuclear pore protein we generated a transgenic cell line expressing endogenously tagged PF14_0442-GFP (Additional file 21). Indeed, IFA analysis revealed that this protein localized to the nuclear rim, internal to the ER, with several foci of higher intensity, a staining pattern reminiscent of nuclear pores (Figure 9).

Classical nuclear localization signals (cNLSs) have been the subject of several bioinformatic prediction algorithms, and our core nuclear proteome provided an opportunity to test their utility in identifying P. falciparum nuclear proteins. We used three different bioinformatic tools to predict cNLSs: NLStradamus 76, predictNLS 77, and cNLS mapper 78. All three algorithms attempt to predict nuclear localization based solely on the presence of cNLSs and not other protein domains.

NLStradamus, PredictNLS, and cNLS mapper suggest 51%, 37%, and 22% of proteins found in the core nuclear proteome contain a cNLS, respectively (Additional file 10), compared to 45%, 31%, and 17% of the entire P. falciparum proteome (P values 0.008, 0.006, and 0.006, respectively; Fisher's exact test). In contrast, only 47%, 32%, and 18% of proteins found in the whole cell trophozoite proteome 17 were predicted by these algorithms (P values 0.17, 0.4, and 0.18, respectively, relative to all P. falciparum proteins; Fisher's exact test). While this result adds some evidence to the hypothesis that cNLS-mediated nuclear import may operate in P. falciparum similarly to model eukaryotes, the difference in the percentage of cNLS-containing proteins in the core nuclear proteome compared to the set of all P. falciparum proteins is only marginal. Likewise, the same three predictors identify only 50% (NLStradamus), 37% (PredictNLS), and 28% (cNLS mapper) of proteins in the combined ApiLoc and literature review reference set of 317 P. falciparum nuclear proteins to contain a cNLS (P values 0.08, 0.07, and 0.0004, respectively, relative to all P. falciparum proteins; Fisher's exact test). Using more stringent bioinformatic filters on the core nuclear proteome resulted only in a slightly higher percentage of proteins with predicted cNLS (Additional file 22). Together, the low percentage of nuclear proteins being predicted suggests that either P. falciparum cNLSs may be too divergent for current bioinformatic predictors, or that the major mode of nuclear import may not occur through classical import signals, or a combination of both.

We detected 299 proteins (37.3%) implicated in nuclear processes such as transcription, chromatin remodeling, DNA replication/repair, RNA binding and processing, ribosome biogenesis, but also in more general processes such as protein folding and modification, protein degradation, translation, cytoskeleton organization, and metabolism. This classification is based on direct experimental evidence, literature review, and/or sequence similarity to known S. cerevisiae nuclear proteins (Additional file 10). A total of 118 proteins (14.7%), for which such evidence is missing, carry annotations similarly consistent with roles in DNA/chromatin interactions, cell cycle control, mitosis, RNA binding and processing, protein folding and modification, protein degradation, cytoskeleton organization, and metabolic processes. Ninety-five proteins (11.8%) represent ribosomal subunits or translation-associated factors, and 35 proteins (4.4%) are predicted to localize to other compartments such as the mitochondrion, ER/Golgi, protein sorting vesicles, or the vacuole. Finally, the largest fraction of the core nuclear proteome consists of 255 proteins of unknown function (31.8%).

To date, only three proteins have been localized to the parasite nucleolus (RNA polI, PFE0465c; fibrillarin/PfNOP1, Pf14_0068; PfNOP5, PF10_0085) 8384. Further, only six P. falciparum proteins are annotated with the GO term 'nucleolus', and 36 are annotated by GeneDB 85 as putative nucleolar proteins. Here, we detected 12 of these proteins including fibrillarin, PfNOP5, RNA methyltransferase (PF11_0305), putative ribonucleoprotein (RNP) components such as LSM homologs (PFL0460w, PF08_0049), and several predicted pre-ribosomal assembly proteins, as well as another 19 putative homologs of S. cerevisiae nucleolar proteins (Additional file 10). Notably, two of these candidates, NuProC2 (PF10_0278) and NuProC3 (PF11_0250), co-localized with fibrillarin to the parasite nucleolus. These novel identifications expand our current knowledge of the P. falciparum nucleolus and provide a basis for detailed analyses of this essential nuclear compartment.

Plasmodium protein kinases (PK) play central roles in growth, development, and differentiation throughout the life cycle, and are intensely studied as a class of promising antimalarial targets 8687. We identified kinases or their accessory factors of seven PK systems, six of which have been implicated in nuclear roles in Plasmodium or other eukaryotes: PfMAT1 (PFE0610c) 88; the α- and β-subunits of casein kinase 2 (PF11_0096, PF11_0048) 899091; casein kinase 1 (PF11_0377) 9293; the catalytic and regulatory subunits of cAMP-dependent protein kinase (PFI1685w, PFL1110c) 94; NIMA-related protein kinase PfNEK-1 (PFL1370w) 9596; and mitogen-activated protein kinase 2 PfMAP2 (PF11_0147) 9798. PfMAP2 is a substrate of PfNEK-1 96 and both kinases are essential for completion of the IDC in P. falciparum 9599, whereas in P. berghei PbMAP2 is implicated specifically in the development of male gametes 97. Interestingly, while the targets of these and other nuclear kinases remain largely unknown, we find that 70% of proteins in the core nuclear proteome (562/802) were recently shown to be phosphorylated 100 (Additional file 10) suggesting an important role for kinase signalling in the regulation of nuclear processes in P. falciparum.

We detected 32 of the 33 annotated P. falciparum proteasome components and all, except two of the 19S regulatory particle (RP) subunits, were represented in the cytoplasmic fraction. Intriguingly, all subunits of the RP were also associated with the insoluble nuclear fraction in ring stage parasites (Additional file 10). In other species, the RP interacts with chromatin and is involved in transcriptional regulation 101. For example, the S. cerevisiae RP participates in SAGA histone acetyltransferase complex recruitment 102, interacts with the FACT (facilitates chromatin transcription) complex 103, and influences H3 methylation and gene silencing 104. Hence, we speculate that the 19S RP may have similar non-proteolytic roles in P. falciparum transcriptional regulation.

The nuclear proteome also contained several orthologues of the mRNP complex implicated in translational repression during P. berghei gametocytogenesis 105106, such as the RNA helicase PfDOZI (PFC0915w), PfCITH (PF14_0717), the RNA-binding proteins PfHOBO (PF14_0096) and PfHOMU (PFI0820c), and a homolog of a yeast poly(A)-binding protein (PFL1170w). Although DOZI and CITH were described as cytoplasmic proteins in P. berghei, their localizations do not appear to exclude the nucleus. Furthermore, homologs of PfHOMU and PFL1170w shuttle between the cytosol and nucleus in mammalian cells and S. cerevisiae, respectively 107108. The detection of these proteins in the asexual nuclear proteome suggests the presence of translational repression machinery during the IDC, although P. berghei DOZI and CITH loss-of-function mutants have no apparent phenotype in asexual parasites 105106.

We also discovered novel domains that are likely linked to previously unrecognized nuclear functions of parasite proteins, and we envisage their roles will be more thoroughly recognized in future studies. Of particular interest is the identification of the ACDC domain, which we identified exclusively in members of the ApiAP2 family. Moreover, we experimentally validated the identification of PF14_0442 as a novel subunit of P. falciparum nuclear pores. Testing this in more detail will be an important future task given that Plasmodium parasites lack identifiable orthologues of most nuclear pore components 109.

We made several interesting observations regarding stage- and fraction-specific protein profiles that allow us to speculate about potential temporal and spatial expression patterns of nuclear proteins. In light of the stochastic undersampling of complex proteomes in shotgun proteomics approaches and possible sample-to-sample variation, however, these data have to be interpreted with caution.

A total of 58, 90, and 105 proteins were found exclusively in ring, trophozoite, and schizont stages, respectively. In trophozoites and schizonts, expression of these proteins occurred roughly in line with transcription of the encoding genes (Additional files 1 and 23). mRNA expression of ring stage-specific nuclear proteins was somewhat surprising, with a collective profile similar to that observed in schizonts, and an additional smaller peak at 15 to 20 hpi. This suggests that some proteins were newly synthesized in the ring stage while some remained from the preceding schizont stage. A total of 159 of the 253 stage-specifically detected proteins carry no functional annotation and several others carry annotated domains that indicate little about their function. Hence, a large number of novel proteins have been assigned a potential stage-specific role in nuclear biology where previously no informative annotation was available. Further, 13 predicted TFs 63 were detected specifically in a single IDC stage only, indicating a role for these factors as cell cycle-specific regulators of transcription and genome regulation (Additional file 24).

Of 145 proteins predicted to interact with DNA and/or chromatin, 137 (94.5%) were identified in the chromatin-containing fractions (DNAseI-, high salt-, and/or SDS-soluble fractions) (Additional file 10). Some of these factors, such as histones, SNF2 helicase (PFF1185w), chromodomain-helicase-DNA-binding protein 1 (CHD1) (PF10_0232), putative chromosome assembly factor 1 (PFE0090w), and the nuclear peroxiredoxin PfnPRX (PF10_0268) 110 were detected in all three chromatin-associated fractions. In contrast, nine out of the 10 RNA pol II subunits identified were exclusively detected in the high salt nuclear extract. Both FACT components (PFE0870w, PF14_0393) extracted almost identically only after high salt and SDS extraction. The two recently described high mobility group box proteins (PFL0145c, MAL8P1.72) 111 were identified in DNAse1- and high salt-soluble extracts but not in the SDS fraction. These examples suggest that at least some regulatory complexes were extracted as interacting entities. Interestingly, all ApiAP2 factors showed a noticeable association with the insoluble nuclear fraction. While the reason for this remains unknown our observation hints at possible functions of these DNA-binding proteins. Most ApiAP2 proteins are large and, apart from the short and well-defined DNA-binding AP2 domains, consist of extensive uncharacterized regions. It is possible that these non-AP2 regions may mediate the formation of regulatory complexes involved in diverse processes such as DNA replication, transcriptional regulation, or functional organization of the genome. Such complexes are often resistant to extraction with DNAseI and high salt buffers, as observed for many RNA- or DNA-binding proteins associated with the nuclear matrix 112. In case of PfSIP2, the only P. falciparum ApiAP2 characterized in vivo, the detection of PfSIP2-derived peptides in the insoluble nuclear fraction is consistent with the association of this factor with condensed heterochromatin 46.

A large number of proteins in the core nuclear proteome (379) were also detected in the cytoplasmic fraction. This pool of proteins contained members of all functional classes but was clearly enriched in distinct pathways associated with the various functions of the nucleolus in other eukaryotes 113114115. These include the majority of ribosomal subunits (90.5%), RNA-binding proteins (78.9%), factors involved in protein degradation (69.0%), and protein folding and modification (65.1%). Furthermore, 84.4% of translation-related factors were identified in the cytoplasmic and nuclear fractions, a finding consistent with the existence of nuclear translation 116117. We also noticed that 42% (62 proteins) of confirmed or likely nuclear proteins in the core nuclear proteome were detected in the cytoplasmic fraction. Moreover, six proteins experimentally localized to the nucleus in this study had peptides detected in the cytoplasmic fraction. Hence, while some of the dually detected proteins may represent cytoplasmic contaminants such as abundant cytosolic proteins or macromolecular complexes, our results show that many nuclear proteins in P. falciparum shuttle between the nucleus and cytosol and/or perform their tasks at multiple destinations.

Transport of proteins into the nucleus remains poorly understood in apicomplexan parasites 118. In yeast and mammalian model systems short, arginine-, and lysine-rich cNLSs are thought to be the major mediators of nuclear import, though alternative and redundant mechanisms have been described 119. The poor enrichments of cNLSs in both the core nuclear proteome and the set of 317 curated nuclear proteins show that bioinformatic discrimination of nuclear vs. non-nuclear P. falciparum proteins via prediction of cNLSs remains impractical. Notably, however, cNLS predictors are also problematic in the reliable identification of nuclear proteins in the model systems they were designed for 120. Nevertheless, differences between P. falciparum nuclear proteins and these current computational models of cNLSs appear prevalent but remain unclear. It has not been determined whether the majority of P. falciparum nuclear proteins are imported independently of importin α, the protein that binds cNLSs 118, or alternately by similar but unrecognized cNLSs that do mediate importin α-dependent translocation. The fact that only 22% to 51% of nuclear proteins are predicted to contain cNLSs (depending on the predictor used) suggests that in its current definitional form the cNLS is not the major mode of nuclear import in P. falciparum. Notably, the important insight that most verifiable Plasmodium nuclear proteins lack a recognizable nuclear localization sequence, and need thus be identified through empirical strategies, reinforces the value of an experimentally robust nuclear proteome for understanding Plasmodium nuclear biology.

Using a combination of OrthoMCL and synteny analyses (Additional files 1 and 25) only around 10 proteins are genuinely falciparum-specific. Unsurprisingly, nearly all of them have no functional annotation, though one is the gametocyte-specific protein Pfg27 (PF13_0011). Pfg27 binds RNA and has previously been localized to the cytoplasm and nucleus of gametocytes 121. Interestingly, we detected another nuclear P. falciparum-specific protein of unknown function (PFB0115w) that is expressed during the IDC and in gametocytes and contains C-terminal homology to the Pfg27-specific fold 122 suggesting a functional relation between these two proteins.

The pool of around 100 genus-specific proteins represents a promising group from which to characterize features of nuclear biology peculiar to Plasmodium. Several are potentially worthy of prioritized treatment; these include some of the previously characterized ApiAP2s 40, as well as a large number of uncharacterized proteins with nucleic acid-binding and chromatin-interacting domains. Several others in this group have kinase or phosphatase domains and may be involved in expression regulation cascades. One of the Plasmodium-specific nuclear proteins is a putative metacaspase (PF14_0363). Interestingly, unrelated trypanosomatid parasites also possess nuclear localized metacaspases that are required for proliferation, possibly through modulation of cell cycle progression 123.

P. falciparum 3D7 parasites were cultured as described previously 125. Growth synchronization was achieved by repeated sorbitol lysis 126. Transfections were performed as described 127 and selected on 5 μg/mL blasticidin-S-HCl and/or 4 nM WR99210. Parasites transfected with pH_0442-GFP were cycled on/off WR three times to select for integration of the plasmid into the endogenous locus.

Nuclei from ring stages, trophozoites, and schizonts were isolated and fractionated as previously described with minor modifications 35. A detailed protocol is provided in Additional file 1. For western blot analysis protein fractions were separated on 10% SDS-PAGE gels and transferred to nitrocellulose membranes (Schleicher&Schuell). Primary antibody dilutions were: anti-H4 (Abcam, ab10158) 1:10,000; anti-HSP60 (kind gift from Geoff McFadden, University of Melbourne) 1:2,000; mAb anti-PfGAPDH 1:2,000 128; rabbit anti-PfHP1 (kind gift from Mike Duffy, University of Melbourne) 1:2,000; and anti-GFP (Roche Diagnostics, 11814460001) 1:1000.

Nuclei were incubated with DAPI (10 μg/mL) in CLB for 10 min on ice and directly observed by light and fluorescence microscopy. Images were obtained using a Leica DM 5000B microscope with a Leica DFC 300 FX camera and acquired via the Leica IM 1000 software and processed and overlayed using Adobe Photoshop CS2. For TEM analysis isolated nuclei or whole asynchronous 3D7 cultures were fixed, prepared for microscopy and imaged as described 129. For immunolabelling, nuclei were incubated with rabbit anti-H3K4Me3 (Abcam ab8580) diluted 1:300, then detected with 18 nm colloidal gold-conjugated goat anti-rabbit secondary antibody (Jackson ImmunoResearch, Baltimore, MD, USA) diluted 1:20.

Detailed protocols for sample preparation, MudPIT analysis, and protein identification are provided in Additional file 1. All proteomics data are publicly available on PlasmoDB 68 and in Additional files 3, 4, and 5.

Bioinformatic methods used to assess and analyze the nuclear proteome are presented in detail in Additional file 1.

To express epitope-tagged candidate proteins in P. falciparum, full-length genes were amplified from gDNA or cDNA and cloned into pBcam-3xHA 35 using BamHI (or BglII) and NcoI (or NheI) restriction sites. The PFL0635c gene (NuProC6) was cloned into BamHI/NheI-digested pHcam-2xTy 46. To generate the double transfectant co-expressing NuProC2-3xHA/NuProC3-2xTy the PF11_0250 gene (NuProC3) was cloned into BamHI/NcoI-digested pHcam-2xTy and transfected into BSD-resistant 3D7/NuProC2-3xHA parasites. All forward primers included five wild-type nucleotides directly upstream of the natural ATG start codon. pH_0442-GFP was obtained by cloning the 3' end of PF14_0442 (551 bp) into PstI/NotI-digested pHcam-GFP 130. Successful 3' replacement integration of pH_0442-GFP into the endogenous locus was verified by PCR and western blot (Additional file 21). Primer details are listed in Additional file 26.

IFAs were performed with iRBCs fixed in 4% formaldehyde/0.01% glutaraldehyde as described elsewhere 131. Primary antibody dilutions were: rat mAb anti-HA 3F10 (Roche Diagnostics) 1:100; mouse mAb anti-Ty BB2 1:2,000 (kind gift from Keith Gull); mouse mAb anti-GAPDH 1:500 128; rabbit anti-PfBIP 1:500 132133 (kind gift from Tim Gilberger); and mouse mAb anti-fibrillarin 1:500 71 (kind gift from Michael Terns, University of Georgia). Secondary antibody dilutions were: Alexa-Fluor^® 568-conjugated anti-rat IgG (Molecular Probes) 1:500; Alexa-Fluor^® 488-conjugated anti-rabbit IgG (Molecular Probes) 1:500; and FITC-conjugated anti-mouse IgG (Kirkegaard Perry Laboratories) 1:250. Images were taken on a Leica DM 5000B microscope with a Leica DFC 300 FX camera and acquired via the Leica IM 1000 software and processed and overlayed using Adobe Photoshop CS2. Observed protein localizations have been deposited in ApiLoc 64.