We downloaded and mined the protein-protein association data from the STRING database 4 involving proteins from P. falciparum. Seventy-seven (77) out of 124 predicted proteases were found in this set and were associated with at least one other protein, constituting 2,431 associations (Additional Files 1 and 2). Each association between a pair of proteins has a confidence score (S) ranging from 0.15 to 0.999 that was inferred from the evidence used to establish the association: 221 associations (9.1%) have high confidence scores (S>0.7), 432 associations (17.8%) have medium confidence scores (0.4≤S≤0.7), and strikingly, 1,778 associations (73.1%) have relative low confidence scores (0.15≤S<0.4). The large proportion of low-scored associations arises from the paucity of annotation data. Before the genome of P. falciparum was sequenced, only about 20 proteins had been characterized; after genome sequencing this number increased by two orders of magnitude, but over 60% of the predicted gene products in the genome still had no functional assignment 18 and ten years of subsequent effort have reduced this number to roughly 45% 60. Consequently, information such as KEGG pathway assignments, PDB protein structures and reactome data, which tend to improve association scores, is scarce for P. falciparum. Therefore, our subsequent analysis will not exclude the associations with low confidence scores as they may well represent associations that have not been previously recognized.

The degrees of connectivity vary among the 77 predicted proteases with protein-protein associations, ranging from one to 143 (Additional File 1). Twenty-four (24) putative proteases have less than five association partners, 13 have 5-10 partners, and 40 are highly connected with more than 11 partners, suggesting that proteases are involved in complex cellular networks. Functional enrichment analysis 61 revealed that 120 Gene Ontology (GO) terms were over-represented in these protease associations (p < 0.05) (Additional File 3). Figure 1 shows the distribution of functional categories in a hierarchical order: proteolysis (GO 6508) is, not surprisingly, enriched (p = 8.29*10^-6), while the other most highly represented GO biological processes (p < 10^-5) are related to cellular catabolic processes (GO 44248), protein metabolic processes (GO 19538), macromolecule metabolic processes (GO 43170), and cofactor and coenzyme metabolic processes (GO 51186 and 6732). This result reflects the involvement of proteases in fundamental biological processes, many of which have been established in the wet lab. Other processes that are moderately enriched (10^-5<p < 0.05) included gene expression and response to endogenous and external stimuli such as heat, abiotic stimulus, organic substances, unfolded proteins, and protein stimuli. Five of the potentially most significant protease-associated sets of proteins are discussed in the following sections. They include the ubiquitin-proteasome system, the stress response system, the regulated intramembrane proteolysis system, the parasite egress network, and the signal peptidase network. These subnetworks were chosen because: (1) proteases are the central players in these networks; (2) These networks play crucial roles in parasite life cycle and are closely associated with adaptive phenotypes such as stress response, transcriptional regulation, pathogenesis, and virulence; (3) These networks are considered to be potential antimalarial targets as their disruption would cause deleterious effects on the growth or infectivity of the parasites.

The largest protease-associated network in P. falciparum is the ubiquitin-proteasome protein degradation system (UPS). The UPS is responsible for degrading unwanted or misfolded proteins and is believed to execute important roles in protein turnover and cell cycle regulation in a wide variety of organisms 62. We previously identified a group of threonine proteases that form α- and β- subunits of the proteasome complex and two families of ubiquitin-specific hydrolases (C12 and C19) 163 (Additional File 1). The UPS pathway in P. falciparum has been deduced by Dr. Hagai Ginsburg (http://sites.huji.ac.il/malaria/maps/proteaUbiqpath.html), and involves two consecutive steps: (1) tagging the ubiquitin molecules to target proteins and (2) degradation of the tagged protein by the proteasome complex with release and recycling of ubiquitin. The major components of the UPS in P. falciparum are conserved with other eukaryotes. However, a growing body of evidence suggests that the UPS plays a critical role in the parasite-specific life style and it is therefore intriguing to unveil the proteins and pathways that are associated with or regulated by the UPS 6465, as they may carry out functions specific to pathogenesis or virulence. We identified 1,148 associations in P. falciparum that involved 11 threonine proteases in the T1 family, two proteases in the C12 ubiquitin C-terminal hydrolase family, and six proteases in the C19 ubiquitin-specific protease family. One hundred and twenty-four (124) associations are protease-protease associations, and the remaining 1,024 associations involve non-protease partners. One hundred and sixty-four (164) of these associations have high confidence scores (S>0.7), the majority of which involve the association between catalytic components and regulatory components in the proteasome complex (Additional File 4).

The protease with the highest connectivity is PF10_0111, a putative 20S proteasome beta subunit, which has 143 association partners (Figure 2). In addition to the proteasome components and ubiquitin conjugation enzymes, the other interacting proteins appear to be involved in a variety of activities (Table 1): (1) a nucleotide binding activity involving a Tat binding protein homolog (PFL2345c) which has an AAA ATPase domain; (2) cell cycle regulation involving MAL13P1.337, which is a putative protein in the Skp1 (S-phase kinase-associated protein 1) family. The Skp1 protein is an important component of the cyclin A-CDK2 S phase kinase complex in baker's yeast (Saccharomyces cerevisiae) 66 and directs cell cycle-regulated proteins to the kinetochore; (3) translation involving a number of ribosomal proteins such as 60S ribosomal proteins L40/UBI (PF13_0346) and L10 (PF14_0141), a putative translation initiation factor eIF-1A (PF11_0447), and a putative elongation factor 1 (EF-1) (PFC0870w); (4) transcriptional regulation involving a putative multiprotein bridging factor type 1 (MBF1) (PF11_0293). MBF1 is a transcriptional cofactor that bridges the TATA box-binding protein (TBP) and its specific regulatory proteins for transcriptional activation 67; (5) membrane traffic regulation involving a putative rab specific GDP dissociation inhibitor (PFL2060c) 68.

Moreover, the yeast 2-hybrid assay using PF10_0111 as a bait revealed 15 PPI preys (Table 1), confirming that it is associated with (1) transcriptional regulation involving a CCAAT-box DNA binding protein subunit B (PF11_0477) containing a histone-like transcription factor domain, and (2) translation involving a putative translation elongation factor EF-1 subunit alpha (PF11_0245), a putative 60S ribosomal protein L4 (PFE0350c), and a putative ribosomal protein L15 precursor predicted to localize to the apicoplast (PF14_0270), a specific organelle of prokaryotic origin found in Apicomplexa parasites. PF10_0111 may also be associated with protein modifications involving a putative ubiquitin transferase (MAL7P1.19) 69 and chromatin fluidity involving a putative nucleosome assembly protein (PFI0930c).

Interestingly, PF10_0111 is shown to have PPI with three predicted surface antigens: (1) merozoite surface protein 3 (PF10_0345), which was shown by global RNA decay and nuclear run-on assays to serve a role in transcriptional regulation and RNA stabilization 7071; (2) a merozoite surface protein (PF10_0348). Domain analysis revealed a N-terminus Duffy binding domain that is present in the Duffy receptors expressing blood group surface determinants and a C-terminus SPAM (secreted polymorphic antigen associated with merozoites) domain, both of which have been implicated in parasite immune evasion, cytoadherence and pathogenesis 7273; (3) a parasite-infected erythrocyte surface protein (PFE0060w). The microarray and proteomics assays show that these three surface proteins are expressed at the invasive merozoite stage 681011.

These results reflect much that is known about the UPS, but also suggest that it may also be associated with a variety of processes ranging from transcriptional regulation, translation, cell cycle progression, invasion, protein trafficking, and immune evasion. Not surprisingly, the UPS has become a promising antimalarial target. Various independent studies have shown that inhibition of proteasome activity can arrest parasite growth, and yet show limited toxicity to human cell lines 647475.

The adaptation of the malaria parasite to the host environment requires a rapid and effective response to diverse physiological signals and stress conditions, such as changes in temperature within hosts, nutritional challenges, host immune responses, antimalarial administration, and so on. One such adaptive network in the malaria parasite is the robust heat shock response system. During its life cycle, the parasite is transmitted from the mosquito vector (~25°C) to the human host (37°C), resulting in heat shock. Periodic fever, the patient's response to infection, also presents recurrent heat shock to the parasite. A comprehensive chaperone system has been identified in P. falciparum genome, accounting for 2% of the open reading frames (ORFs) 76. The system is comprised of various chaperone proteins 77 and includes proteases that degrade misfolded proteins. We identified 344 associations involving five putative proteases in the ClpP endopeptidase family (S14), a lon protease PF14_0147 (S16), and an hslV protease PFL1465c (T1B). As shown in Figure 3, these proteases are associated with a large number of heat shock proteins (HSPs) including Hsp90, Hsp70, Hsp40, and DnaJ proteins. The protease having the highest degree of connectivity (80) in the heat shock response network is PFL1465c, a threonine protease hslV. In addition to the classical HSPs, it is associated with a wide variety of enzymes such as ubiE/COQ5 methyltransferase, rRNA methyltransferase, multiple tRNA synthetases, various phosphate isomerases, amino transferase, aldolase, and a number of kinases, suggesting it may have an important role in parasite metabolism. It is also associated with three other heat-shock response proteases (PFC0310c and PF08_0063 in the S14 ClpP endopeptidase family, and PF14_0147 in the S16 ATP-dependent protease lon family), an organelle processing peptidase in the M16 pitrilysin family (PFI1625c), a leucyl aminopeptidase in the M17 family (PF14_0439), and aminopeptidase P in the M24 family (PF14_0517); together, they form a complex protease degradation network. It is also interesting to note that this putative hslV protease appears to be linked to a second very important stress system in the malaria parasite that acts against oxidative challenges: Protease hslV is associated with the thioredoxin 1 protein (PF14_0545), a member of the thioredoxin system which controls cell redox homeostasis, and a putative Fe-superoxide dismutase (Fe-SOD, PF08_0071) which is critical for antioxidant defense. Because the malaria parasite is sensitive to oxidative stress, both the thioredoxin system and SOD have been considered as potential antimalaria targets 78. Finally, our phylogenetic analysis revealed that this hslV protease (PFL1465) is of prokaryotic origin and there is no homolog in the human host, a desirable feature for drug targets 798081. A second heat shock response protease PfClpP (PFC0310c) was recently characterized 8283; protease inhibition assays have shown that it, along with other ATP-dependent chaperones, plays a crucial role in parasite growth and development. Furthermore, PfClpP is localized to the apicoplast, which is of cyanobacterial origin, making this protein an apicoplast-targeting antimalarial candidate. This protein is indeed highly connected with 69 association partners, including Hsp70, Hsp60, Hsp40, co-chaperones, and proteins involved in proteasome acitivities, replication, translation, protein biosynthesis, metabolism, and heat shock response, implying that its inactivation would have devastating consequences for the parasite.

The common belief that proteases cleave peptide bonds in a water environment was challenged by the discovery of a set of proteases that conduct hydrolysis in the hydrophobic environment of cellular membranes 84. During RIP, intramembrane proteases cleave transmembrane-spanning helical (TMH) segments of the substrates and release soluble effectors, many of which are signaling molecules, thereby triggering cascades of signal transduction pathways 8586. RIP is now believed to be a ubiquitous signaling mechanism in a wide variety of organisms from bacteria to humans 87. The roles of RIP in the parasite life cycle have begun to be unraveled. Three families of membrane-tethered proteases involved in RIP have been identified in P. falciparum, including an aspartic signal peptide peptidase (PfAPP, PF14_0543) in the A22 presenilin family, eight rhomboid serine proteases (PfROMs) in the S54 family, and two putative Site-2 metallo proteases (S2Ps, PF13_0028 and PF10_0317) in the M50 family1888990919293.

The first family, PfAPP (PF14_0543), has 54 association partners (Figure 4 andTable 2). The association partner with the highest confidence score is a putative Rer1 (retrieval receptor for endoplasmic reticulum (ER) membrane proteins, PFI0150c) that is important for localizing proteins to the ER. Another related partner for PfAPP is a putative ER lumen protein retaining receptor (PF13_0280), which contains a signal sequence that facilitates the protein transport between the cis side of the Golgi apparatus and the ER 94. It is believed that parasite invasion of erythrocytes requires the export of proteins to the ER and the cell surface. They then traverse the parasitophorous vacuole membrane (PVM) into the erythrocyte or parasite-derived membranous structures known as Maurer's clefts. PfAPP, along with the ER-localization proteins, may play a role in protein trafficking, cell-cell communication and remodeling of the host erythrocyte for parasite entry. Other proteins that are associated with PfAPP include secretory proteins, translation initiation and elongation factors, splicing factors and the spliceosome unit, peptide chain release factor, and various enzymes, suggesting it is involved in diverse networks related to transport, translation, posttranslational processing and metabolism. Recent gene disruption assays showed that PfAPP is essential for merozoite invasion and parasite growth 9293; the versatile associations of this protease underscore its potential as a drug target.

The second family, PfROM, includes a group of serine proteins with demonstrated roles in parasite invasion 90919596. Only one out of the ten rhomboid protease homologs in P. falciparum, PfRom1 (PF11_0150), was predicted to have protein-protein associations. Most interestingly, all the six proteins associated with it are antigens that have been considered as vaccine candidates; they belong to three families of adhesins that are essential for parasite invasion (Figure 4 andTable 2): (1) the apical membrane antigen 1 (AMA1, PF11_0344) is an adhesin required for merozoite invasion and it plays an indispensible role in the proliferation and survival of the malaria parasite 97. PfRom1 was shown to be able to cleave AMA1 88; (2) the erythrocyte binding-like (EBL) family is involved in binding to a host chemokine receptor, the Duffy antigen 98. Among the four EBAs with predicted association with PfRom1, EBA-175(MAL7P1.176) is proven a natural substrate for PfRom1 88, but it remains unclear whether PfRom1 can cleave EBA-140 (MAL13P1.60), EBA-181 (MAL1P1.16), and a putative merozoite adhesive erythrocytic binding protein (PF11_0486); (3) a reticulocyte binding protein 2 homolog b protein (MAL13P1.176) in the reticulocyte binding-like (RBL) family. PfRom1 is able to cleave the RBL proteins 88. Apparently, PfRom1 plays a central role in the RIP network that is tightly linked to the invasion process 86 and as such merits further investigation as a drug target.

S2Ps in the third family, PF10_0317 and PF13_0028, have two and one associations, respectively (Figure 4 andTable 2). PF10_0317 is associated with a proteasome 26S regulatory subunit and a cell division cycle (CDC) protein 48 homolog, which is implicated by GO analysis in ER localization and cell cycle regulation. Our previous domain analysis showed that PF10_0317 contains a Der-1 like domain, which was implicated in proteolysis associated with the ER 99100101102. PF13_0028 is associated with an adenylosuccinate synthetase AdsS (PF13_0287), which is important for the de novo biosynthesis of purine nucleotides. This association was predicted based on the genome synteny analysis, which revealed that the homologs of S2P and AdsS are located in the same chromosomal neighborhood in a variety of Actinobacteria. The functions of these S2Ps in malaria parasites are yet to be defined.

Egress, the parasite's emergence from host erythrocytes, is a well-coordinated process involving the rupture of the parasitophorous membrane (PVM) and the erythrocyte membrane (EM). Proteases that have been implicated in parasite egress 3136 include (1) aspartic proteases (plasmepsins PMI, PMII, and PMIII, also known as histo-aspartic protease (HAP)) in the A1 family, (2) cystein proteases in the A1 papain family including falcipain 2a, 2b, and 3, dipeptidyl peptidase 3 (PfDPAP3), and a series of Serine Repeat Antigens (SERAs), and (3) a serine protease subtilase 1 (PfSUB1) in the subtilisin S8 family. We analyzed the protein-association network (Figure 5) involving proteases mediating egress and found that a central player in the network is SERA5 (PFB0340c), which has 28 associations. SERA5 is associated with PfSUB1 (PFE0370c) and PfDPAP3 (PFD0230c). Both these proteases can proteolytically activate SERA5, which triggers downstream processing of cellular substrates 103104. SERA5 is also associated with several erythrocyte membrane antigens such as PfEMP2 and EBA-175. It is abundantly expressed in the blood stage, especially in the schizont stage, as revealed by microarray and proteomic analysis. SERA5 has an in vitro catalytic activity and it is refractory to gene disruption 105, suggesting its vital role in the parasite life cycle.

As an adaptive survival strategy, the malaria parasite harbors a powerful secretion system that transports parasite-encoded virulence proteins to their subcellular locations. The central players in this secretion system are a group of signal peptidases that are capable of cleaving signal sequences from the target proteins that can then be routed to their destinations. Five signal peptidases have been predicted and characterized, constituting the signal peptidase complex (SPC) in P. falciparum 3106107. Three of these peptidases have association partners: PfSPC21 (MAL13P1.167) has 120 associations; the putative microsomal signal peptidase (PF14_0317) has five associations; and the putative SPC22 (PFI0215c) has five associations (Figure 6). The associated proteins are part of the secretion pathway and include secretory complex protein 61 (Pf61) alpha and gamma subunits, a signal recognition particle (SRP) and an SRP receptor, an ER lumen protein retaining receptor, and a transport protein particle (TRAPP) component. These signal peptidases are also associated with members of the ubiquitin-proteasome system and the heat shock response system, with the translational machinery, and with metabolic networks.

Proteases in P. falciparum may play other roles important for parasite biology. We previously identified a single copy of calpain PfCalp (MAL13P1.310) in P. falciparum genome 3106107. Calpain is crucial for signal transduction, cell cycle regulation, differentiation, development, and cell-cell communication from bacteria to humans. Very little is known about its role in P. falciparum. Only four proteins seemed to be associated with calpain: including a putative protein with a C3HC4 type zinc finger, the motif commonly present in transcriptional regulators, a ribosomal protein, and two proteins with unknown function. However, partial knockdown assays recently suggested that PfCalp is essential for the parasite's optimal growth and cell cycle progression 108. Phylogenetic analysis revealed that PfCalp is a unique type of calpain confined to alveolates (a group of protists) with distant relatedness to human calpains 63108, adding it to a new line of promising drug target. Another class of proteases that mediate cell cycle regulation and programmed cell death is comprised of the three metacaspases from the C14 protease family 63109. Only one association partner was identified for PF13_0289 and PF14_0363, (polyubiquitin and a hypothetical protein with unknown function respectively), and no associations were found for PF14_0160, reflecting our limited knowledge about their functions in malaria parasite.

The proteases in P. falciparum were predicted using a comparative genomics approach and a support vector machine (SVM)-based, supervised machine learning approach 123. The classification and annotation were according to the MEROPS protease nomenclature, which is based on intrinsic evolutionary and structural relationships 110.

The complete set of protein-protein associations for P. falciparum was extracted from the downloaded STRING database 4; each association between a pair of proteins has a confidence score (S) ranging from 0.15 to 0.999 that was inferred from the evidence used to establish the association, such as homology transfer, KEGG pathway assignments, conserved chromosome synteny, phylogenetic co-occurence, and literature co-occurence 111. This set of associations was visualized in Cytoscape 112 and converted to an undirected weighted graph, where there is a single edge between any pair of proteins and the S value is used as the weight. The network was characterized using NetworkAnalyzer 113 and significant modules were detected using MINE 114 and MCODE 115. The default values were used for all three plugins. The set of proteins directly associated with the 77 proteases in the association set were screened using BiNGO 116 to determine if any categories of proteins, as identified by their Gene Ontology terms, were over-represented. The hypergeometric test was used with the Benjamini and Hochberg false discovery date correction. A significance level of 0.05 was selected.

We downloaded the P. falciparum genomic sequence and annotation data 18, transcriptomic microarray data 678, mass-spectrometry proteomic data 9101112, and protein-protein interactome 5 data for network associated proteins from PlasmoDB, the Plasmodium Genome resource center (http://www.plasmodb.org) 117. Conserved domains/motifs in P. falciparum sequences were identified by searching InterPro 118. Multiple alignments were obtained using the ClustalX program 119 and T-coffee 120, followed by manual inspection and editing. Phylogenetic trees were inferred by the neighbor-joining, maximum-parsimony and maximum-likelihood methods, using MEGA5 121.