The P. vivax genome database shows that the pvmsp3-α, pvmsp3-β and pvmsp3-γ genes encoding the originally identified members of the MSP3 family [13], [14] are located in chromosome 10 within a segment of DNA spanning close to 60 kb (www.plasmodb.org). Our analyses of this region and the surrounding DNA has resulted in eight additional open reading frames (ORF) within the ∼60 kb segment being classified as members of the pvmsp3 family and the complete set of eleven msp3 genes being annotated as pvmsp3.1 through pvmsp3.11 [30] (Fig. 1A, Table 1). Based on their position in the series, the pvmsp3-α, pvmsp3-β, and pvmsp3-γ genes have been renamed pvmsp3.10, pvmsp3.3 and pvmsp3.1, respectively. Each gene is contained in a single exon, without intervening sequences, and they are positioned head-to-tail within the multigene locus. The ATG start codon of each member of the pvmsp3 family is preceded by an adenine at position -3, forming the expected consensus start sequence described in Plasmodium [33].

Three characteristic protein features have been confirmed for the proteins encoded by all eleven members of the pvmsp3 family: an NLRNG motif immediately following the putative signal peptide cleavage site, a predominant central alanine-rich region with predicted coiled-coil domains, and a C-terminal motif containing a glutamate-rich domain (Fig. 1B, 2, S1, S2 and S3).

To arrive at this conclusion, the MEME (Multiple Em for Motif Elicitation) algorithm was used to search for significantly overrepresented amino acid residues in each predicted PvMSP3 family member. Our MEME analysis also included the full-length sequences of the following MSP3 orthologs from P. falciparum, P. knowlesi and P. cynomolgi, respectively: PfMSP3.1 (FC27 and CSL2 strains), PkMSP3₁₄₀, PkMSP3₁₇₂, and PcyMSP3.10 (Berok strain) (Fig. 2, sequence accession numbers in Table S2). It is of special interest that the NLRNG motif is present in each deduced MSP3 sequence, despite otherwise quite divergent MSP3 sequences in all of these species, whether closely (P. cynomolgi and P. knowlesi) or distantly (P. falciparum) related to P. vivax [34].

Using the Multicoil algorithm (http://groups.csail.mit.edu/cb/multicoil/cgi-bin/multicoil.cgi), the central alanine-rich region of all predicted PvMSP3s scores significantly above the cut-off value for the formation of coiled-coils. The central domain spans the majority (66.1%–84.2%) of the total amino acid sequences, and this domain differs in size for each member of the family (Table 1, Fig. 1B, S1 and S2). Though the alanine residues are often in positions a and d of the tandemly repeated heptad AxxAxxx of other MSP3 proteins, we found that the predicted heptad repeat in the PvMSP3 family is better described as having any hydrophobic residue, such as Valine, Isoleucine, Serine or Alanine in the a and d positions (Fig. 2B, S1 and S2).

A complete protein sequence analysis revealed that Ala, Glu, and Lys account for more than 50% of the amino acid composition of the entire PvMSP3 family, although there is low amino acid sequence identity between family members (Table 1). They also have different calculated molecular weights ranging from 41.1 kDa to 134 kDa. The isoelectric point (pI) of 10 of the 11 predicted proteins is close to 5.0, with the one exception being the smallest predicted member, PvMSP3.11, with a pI of 4.2 (Table 1). It is worth noting that in further experimental studies (below) we did not detect any protein expressed from this gene. Thus, the predicted molecular weights of the PvMSP3s that were found to be expressed range from 90.6 kDa to 134 kDa, although each of them consistently migrates significantly higher by SDS-PAGE (see below), as reported earlier for the PvMSP3α and β family members [13], [14].

Global pairwise alignment of the putative PvMSP3 sequences results in similarity scores well below 50%; however, identity scores are much higher between PvMSP3.6 and PvMSP3.8 (81.8%), PvMSP3.4 and PvMSP3.5 (72.6%), and PvMSP3.7 and PvMSP3.9 (63.2%) (Table 2). Alignment of the predicted amino acids of these proteins showed varying identity and similarity percentages, for example, PvMSP3.6 and PvMSP3.8 scored 88 and 82%, for similarity and identity, respectively (Table 2), suggesting a common ancestry and occurrence of gene duplications, compared to the highly polymorphic central domains.

The third characteristic motif of PvMSP3 members is a hydrophilic region in the C-terminus, which contains a high content of acidic amino acids in clusters, particularly glutamic acid and aspartic acid (Fig. 1B, 2C and S3), and lacks a C-terminal transmembrane motif, or any site for post-translational GPI modifications. Notably, all PvMSP3 members also lack a leucine-zipper, which is typical at the terminal end of MSP3 proteins in P. falciparum [16,17 31,35].

An additional ORF was identified between pvmsp3.9 and pvmsp3.10 with the gene ID: PVX_097715 (Fig. 1A). The putative protein sequence encoded by this ORF lacks the typical coiled-coil region, the N-terminal NLRNG motif and the C-terminal motif containing the glutamate-rich domain; therefore, it has not been included as a member of this family. Likewise, the 5′ (PVX_097665) and 3′ (PVX_097730) ORFs flanking the pvmsp3 cluster do not have characteristic features of pvmsp3 genes [13], [14]. These are annotated in PlasmoDB database as the 4-diphosphocytidyl-2c-methyl-D-erythritol kinase gene and a hypothetical gene, respectively.

We set out to determine whether all eleven putative pvmsp3 genes were transcribed in blood-stage infections, or only a subset. If each of these genes proved to be transcribed, it was also important to know when, during the life cycle of the parasite, these transcripts were being produced and their abundance. To achieve this aim, gene-specific primer sets based on the P. vivax Salvador I strain sequences (Table S1) were tested in PCR and RT-PCR experiments using P. vivax Salvador I strain gDNA and cDNA as templates, respectively. The primer pairs were strategically designed to generate products of different sizes to distinguish and unambiguously confirm the specific expression of individual gene family members. The cDNA was produced from total RNA representing predominantly P. vivax Salvador I strain schizonts from a S. boliviensis blood-stage infection. The expected product sizes for all eleven genes were obtained in all reactions performed with gDNA and cDNA (Fig. 3). The melting curve generated with SYBR Green after PCR and RT-PCR amplifications showed similar, sometimes identical, melting temperatures (Tm) between the two templates (Fig. 3). These primer pairs were also evaluated for their ability to amplify the individual pvmsp3 genes from P. vivax gDNA of the Belem strain. No amplicons were detected for pvmsp3.1, 3.2, 3.3, 3.5, and 3.6 from this strain, while amplicons were detected for the rest of the genes (data not shown). In these reactions, the retrieval of all pvmsp3 genes would not be expected, due to polymorphisms between the respective gene sequences represented by the oligonucleotide primer pairs, or possible differences in the number of pvmsp3 genes maintained in each strain.

Next, we sought to determine whether each of the pvmsp3 transcripts from the P. vivax Salvador I strain was similarly expressed, or not. Real-time quantitative qRT-PCR assays were carried out with Taqman probes for specificity to quantify and compare the pvmsp3 transcript levels of each family member present in predominantly trophozoite or schizont total RNA samples, purified from P. vivax Salvador I strain infected RBCs generated from S. boliviensis blood-stage infections (Table S3 and Fig. 4). The real-time qRT-PCR results confirmed that all eleven pvmsp3 genes from the Salvador I strain of P. vivax were transcribed during the blood-stage cycle, but at different levels. The detected transcript levels of the different members of the gene family showed between 0.7- to 25.5-fold changes relative to the gDNA calibrator sample. The overall transcription level of the pvmsp3 genes was below a 6-fold change relative to gDNA in both the trophozoite and schizont stages. However, in the trophozoite-stage sample, transcript levels for the pvmsp3.1 (pvmsp3-γ) and pvmsp3.9 genes showed 25.5- and 9.9-fold changes, respectively, and in the schizont stage, pvmsp3.2, pvmsp3.7 and pvmsp3.9 transcripts showed >6-fold increases relative to gDNA. The comparison of the levels of any pvmsp3 gene transcript between the trophozoite and schizont stages revealed two genes with noticeable transcriptional distinctions: pvmsp3.1 (pvmsp3-γ), which decreased 4.5-fold, and pvmsp3.2, which increased 6.9-fold. The products generated from these experiments were also separated by agarose electrophoresis to confirm the real-time PCR results (data not shown).

To support investigations regarding the expression of the PvMSP3 family, and to specifically distinguish potential differences in protein expression between family members, recombinant proteins and specific antisera were produced representing each member of the family. The recombinant proteins produced included the unique central coiled-coil regions of PvMSP3.1 through PvMSP3.9, the coil-coiled region plus the C-terminal of PvMSP3.10 (PvMSP3-α) and the near full-length sequence of PvMSP3.11 (Fig. 1B).

The antiserum produced against each recombinant PvMSP3 showed extensive cross reactivity as determined by immunoblot (Fig. S5), preventing the use of these antisera to identify the expression of the individual proteins. The cross-reactive antibodies from each specific rPvMSP3 antiserum were eliminated by affinity purification. Individual rPvMSPs were selectively bound to Affi-Gel beads and used to adsorb antibody reactivities in a strategic manner, starting with the strongest cross-reactivities (Fig. S6). For example, cross-reactive antibodies present in the rPvMSP3.1 antiserum were removed by serial passage through affinity columns containing rPvMSP3.2, rPvMSP3.8, and rPvMSP3.9 proteins. We found that it was not necessary to run each antiserum through all ten protein-coupled resins. Antibodies specific to each PvMSP3 family member were attained after three to seven sequential passages over different selected affinity columns. These results indicate that each rabbit antiserum contained antibodies that were specific for each rPvMSP3 immunogen, as well as cross-reacting antibodies, which could be removed by strategic adsorption procedures.

Antibodies recognizing the central domain of most PvMSP3s (rPvMSP3.1 - rPvMSP3.9) did not recognize rPvMSP3.10 or rPvMSP3.11, expressed as full-length proteins. In contrast, the anti-rPvMSP3.10 and anti-rPvMSP3.11 sera recognized all of the other rPvMSP3s but did not cross-recognize each other. All anti-rPvMSP3 sera did not recognize unrelated His-tagged recombinant proteins cloned in the same expression vector or E. coli lysates (Fig. S5).

The expression of the PvMSP3 family was further evaluated by quantitative immunoblot assays. Affinity-purified, PvMSP3-specific antisera were used to probe immunoblots containing P. vivax proteins extracted from infected RBCs containing late trophozoites and schizonts of different stages of maturity from P. vivax (Salvador I strain) and Belem strain parasites (Fig. 5). Ten of the eleven predicted PvMSP3 proteins (PvMSP3.1 through PvMSP3.10) were detected in the Salvador I strain, but at quantitatively different levels; PvMSP3.11 was the only protein not detected in these preparations. An especially large (47.3%) amount of the total amount of PvMSP3 detected corresponded to PvMSP3.9. By comparison, <0.8% of the total PvMSP3 protein corresponded to PvMSP3.4 or PvMSP3.8. The comparative levels of the originally described PvMSP3 family members PvMSP3.1 (PvMSP3-γ), PvMSP3.3 (PvMSP3-β) and PvMSP3.10 (PvMSP3-α) were found to be approximately 15.3%, 6.5%, 14.2%, respectively. Antisera reactive with PvMSP3.1, 3.2, 3.3, 3.9 and 3.10 also recognized small fragments, presumably breakdown products. As noticed in previous studies, likely due to aberrant migration of coiled-coil proteins [13], [14], the apparent molecular weights of PvMSP3s after polyacrylamide gel separation were at least 50% greater in size than their predicted value based on deduced protein sequences (Fig. 5, Table 1).

Eight antisera showed cross reactivity with the Belem strain rPvMSP3 proteins, in support of the likelihood that there is cross-reactivity of antibody immune responses against different MSP3 family members expressed in human infections. Predicted counterpart PvMSP3.1, 3.5, 3.6, 3.7, 3.8, 3.9 and 3.10 proteins were recognized, but not 3.2, 3.4 and 3.11. PvMSP3.11 was not expressed in either strain. Of potential significance in terms of understanding the function and relative immunodominance of these proteins, PvMSP3.9 was detected in the greatest abundance in both strains.

We also tested the reactivity of the panel of antibodies against P. cynomolgi (Berok strain) and P. knowlesi (H strain) schizont extracts in immunoblots. P. cynomolgi extracts were detected by all rPvMSP3 antisera except anti-rPvMSP3.4 and PvMSP3.5. Fewer positive reactions were detected in P. knowlesi (Fig. S7). Only two MSP3 proteins are present in P. knowlesi. They are encoded on different chromosomes and have a high degree of homology in the regions flanking the central coiled-coil domain ([36] and unpublished data).

The expression of native PvMSP3 family members in the infected RBCs was then studied using the affinity purified rPvMSP3 antibody reagents. Indirect immunofluorescence assays were conducted using each of the affinity purified rPvMSP3 antibodies, DAPI nuclear stain and a well-characterized PvMSP1 monoclonal antibody [37] for co-localization studies on blood smears containing P. vivax (Salvador I strain) trophozoites, schizonts and free merozoites (Fig. 6). Several of the PvMSP3 fluorescent patterns appeared to fill the area surrounding the differentiated merozoites within the parasitophorous vacuole. Specifically, eight of the proteins (PvMSP3.2, 3.3, 3.5, 3.6, 3.7, 3.8, 3.9 and 3.10) were detected in mature schizonts. Three of the proteins (PvMSP3.6, PvMSP3.7 and PvMSP3.10) were also detected with free merozoites. Only two of the proteins (PvMSP3.7 and PvMSP3.10) were detected by IFA in trophozoites, schizonts and free merozoites. Three proteins, PvMSP3.1, PvMSP3.4 and PvMSP3.11 were not detected in these assays (not shown). The fluorescent patterns observed by IFA for the various PvMSP3 antisera resemble that of a merozoite surface fluorescence pattern exemplified by PvMSP1 antibodies, included in these assays for comparison (Fig. 6).

Unexpectedly, PvMSP3.7 specific antiserum produced a distinct fluorescent pattern in both mature schizonts and free merozoites that was not typical of a MSP fluorescent pattern and was not suggestive of localization in the PV space. Rather, the observed pattern was typical for antigens localized in organelles at the apical end of merozoites, in this case as a large single dot. Moreover, immunoblot experiments distinctly detected a protein of about 200 kDa in both Salvador 1 and Belem strain parasites without any other bands detected. These data suggest that this gene, despite being denoted as pvmsp3.7, based on the criteria established and summarized above for MSP3s, encodes a unique protein that becomes specifically expressed at the apical pole of merozoites (Fig. 6).

All animal experiments in the current study were conducted in AAALAC-accredited facilities at the Yerkes National Primate Research Center in accordance with the Animal Welfare Act and the Guide for the Care and Use of Laboratory Animals. All experimental protocols were approved by Emory University’s Animal Care and Use Committee. Adult Bolivian squirrel monkeys (Saimiri boliviensis) were housed indoors under conditions of controlled temperatures (67 to 77F) in same-sex pairs. Each cage was equipped with perches and at least the minimum amount of floor space required by federal rules, regulations, and guidelines. Animals were fed a diet of LabDiet New World Monkey Diet (Richmond, IN) twice daily. Peanuts and at least 1/4 orange were offered as enviromental enrichment. All animals are monitored daily by veterinary staff for potential health problems. All animals were trained by experienced personnel to voluntary present at the front of the cage for skin prick using positive reinforcement. For large blood collection volumes, animals were sedated with Ketamine (dose/route) or Telazol (dose/route) to minimize distress. No animals were sacrificed in this study.

Plasmodium vivax (Salvador I and Belem strains) [61], [62] parasites were obtained from blood-stage infections in Saimiri boliviensis monkeys. The P. vivax late-stage trophozoite-infected erythrocytes or early schizonts were purified as described previously [61] and aliquoted samples were stored as frozen stocks. Genomic DNA (gDNA) was extracted as described previously [63]. Total RNA was extracted from the Salvador I strain samples using a RiboPure™-Blood kit (Ambion), followed by two rounds of DNase I digestion to remove any remaining gDNA. Since the pvmsp3 genes have a one-exon gene structure, trace contamination of gDNA in RNA samples would dramatically affect the results of quantitative real-time Reverse Transcription - Polymerase Chain Reactions (qRT-PCRs). DNA contamination was evaluated by PCR amplification of total RNA samples without reverse transcription using Taq DNA polymerase (Invitrogen), either a weak or no amplicon was detected, thus, we concluded that our RNA samples contained little or no gDNA.

First strand complementary DNA (cDNA) was synthesized using the SuperScript™ III First-Strand Synthesis System (Invitrogen) with an oligo(dT)₂₀ primer at 50°C, and resulting cDNA samples were stored at −20°C and thawed as needed for RT-PCR studies. PCR and RT-PCR were performed using gDNA or single strand cDNA as templates, respectively, the SYBR Green PCR Master Mix system (Applied Biosystems), the iCycler machine (Bio-Rad) and gene-specific primers (Table S1). Amplification conditions consisted of one cycle of 15 min at 95°C followed by 45 cycles of 95°C for 30 sec, 58°C for 15 sec, and 72°C for 60 sec. At the end of the reactions, using SYBR Green dye, the melting curve of each sample was acquired and analyzed. After amplification, the size of the PCR products amplified from both gDNA and cDNA were analyzed by electrophoresis on 1.5% agarose gels.

Real-time qRT-PCR was performed using TaqMan® Universal PCR Master Mix Kit (Applied Biosystems) and the iCycler (Bio-Rad) detection system. An initial incubation at 55°C for 2 min activated the AmpErase UNG (degradation of potential carry-over DNA contamination). Then, amplification conditions consisted of one cycle of denaturation (95°C, 10 min) and 45 cycles of amplification (95°C, 15sec, and 60°C, 1 min). 25 ul reactions contained 1 ng of gDNA or cDNA, 0.3 µM each primer, 0.2 µM Taqman probe and 12.5 µl of TaqMan® Universal PCR Master Mix. Real-time qRT-PCR experiments using pairs of primers (Table S1) representing each of the 11 pvmsp3 genes were repeated in duplicate including the P. vivax seryl-tRNA synthetase gene as an internal control and both stage-specific cDNA and gDNA as templates. Negative controls without primers were also included in the qRT-PCR amplifications.

The relative quantification method was used to measure gene transcription levels. First, the amplification efficiency of primers pairs representing the 11 pvmsp3 genes were validated using dilutions of gDNA [64]. The P. vivax seryl-tRNA synthetase gene served as an internal control. The validation experiment, using gDNA diluted in a 100-fold range, demonstrated that the efficiencies of all primer sets amplifying the pvmsp3 gene family members, and the internal control reference gene, were approximately equal (data not shown). P. vivax Salvador I gDNA, known to harbor single copies of each msp3 gene family, was used as a calibrator in the qPCR, and data was analyzed by the 2^−ΔΔCT method [64]. To calculate the relative quantification of each pvmsp3 gene, the ΔΔCt formula was used as follows ΔΔCt = (Ct, pvmsp3 − Ct, pvseryl-tRNA synthetase)χ – (Ct, pvmsp3 - Ct, pvseryl-tRNA synthetase)y, where χ = cDNA and y = gDNA.

The alanine-rich coiled-coil central domains of all but one PvMSP3 were expressed as recombinant fusion protein (rPvMSP3) in E. coli using the Champion™ pET200 Directional TOPO® expression vector (Invitrogen), the PvMSP3.10 (PvMSP3α) gene was alternatively expressed in the vector pET-24d(+) expression vector (Novagen). All fragments were amplified from P. vivax (Salvador I) gDNA using high fidelity KOD DNA polymerase (Novagen) and directional ligation. Positive clones were confirmed by DNA sequencing and subsequently transformed into BL21 Star™ (DE3) E. coli cell (Invitrogen) for expression. Fusion proteins were induced by the addition of 1 mM isopropyl-b-D-thiogalactopyranoside (IPTG) at 37°C for 3 hr. Recombinant proteins (rMSP3s) were expressed as 6X-His tagged fusion proteins and were purified using a combination of HisTrap™ HP columns (Amersham) and gel filtration or anion exchange columns. New Zealand White female rabbits were immunized with the PvMSP3 recombinant proteins for production of a set of specific polyclonal antisera following a three-dose immunization schedule (Covance).

Reactivity of each rabbit polyclonal serum was tested against its own recombinant and the other 10 recombinant proteins by immunoblot. All rabbit antisera showed variable but selective cross-reactivity against some rPvMSP3. Thus, the antisera were adsorbed with highly reactive PvMSP3 recombinants coupled to Affi-Gel 15 (Bio-Rad). Briefly, 1.5 mg of purified recombinant protein was coupled to 2 ml Affi-Gel 15 using the recommended protocol. After blocking the active protein binding ester sites with 1 M ethanolamine HCI (pH 8) and equilibration of the resin with 1xTBS containing 10% dry milk, 200 µl of each antiserum was run through the Affi-Gel 15 column. The flow-through was then run through columns containing other selected antigens that were also causing cross-reactivity. All antisera showed no cross-reactivity by immunoblot after performance of these adsorption procedures (Fig. S6). The affinity purified antisera were stored at −20°C in 10% dry skim milk.

Indirect immunofluorescence assays (IFA) were conducted on air-dried thin smears of P. vivax (Salvador I strain) iRBCs purified from blood-stage infections of S. boliviensis, using Institutional Animal Care and Use Committee approved protocols. When necessary, the desired stage of parasite development was obtained by maturation in short-term in vitro culture [61], and slide collections were stored at −80°C as frozen stocks. Upon removed from −80°C storage, the slides were immediately fixed with 0.25% paraformaldehyde in 1x DPBS at room temperature for 20 minutes. Primary antibodies, adsorbed rabbit anti-PvMSP3 sera and PvMSP1 monoclonal antibody (3F8.1A2), were mixed and diluted at 1∶250 and 1∶1000 respectively with 1x DPBS containing 0.2% fraction V bovine serum albumin (BSA), and added to each slide and incubated at room temperature for 1.5 hr, followed by three washes with 1x DPBS containing 0.2% BSA. Goat anti-rabbit IgG (highly cross-absorbed by human and mouse IgG) conjugated with Alexa 488 dye (Invitrogen) and goat anti-mouse IgG (highly cross-absorbed by bovine, goat, rabbit, rat, and human IgG and human serum) conjugated with Alexa 555 dye (Invitrogen) were diluted at 1∶100 and used as secondary antibodies. Finally, the slides were washed four times with 1x DPBS containing 0.2% BSA, mounted with ProLong® Gold antifade reagent with DAPI (Invitrogen) and covered with a cover slip. Antibody binding and DNA staining were assessed using a Zeiss Z.1 fluorescence microscope.

Whole cell lysates of P. vivax Belem (trophozoite and early schizonts) or Salvador I strains (schizonts) iRBCs were solubilized with reducing (2-β-Mercaptoethanol) sample buffer. Ten µl of sample were separated by SDS-PAGE and transferred to nitrocellulose membranes using a Bio-Rad Mini Trans-Blot Cell (Bio-Rad). The membranes were probed with adsorbed polyclonal rabbit antisera (1∶10,000) against each rPvMSP3 followed by goat anti-rabbit serum 1∶5000) conjugated to horseradish peroxidase (Promega, Madison, WI, USA). The band intensities were quantified using the Quantity One software (Bio-Rad) with the adjusted volume method and expressed as a percentage of total MSP3 protein. Simultaneously, duplicate samples were stained with mixture antibodies of all eleven rabbit anti-PvMSP3 or Ponceau S reagent after SDS-PAGE separation and nitrocellulose membrane transfer for normalization of the protein concentration in the parasite lysates.

The Multicoil software (http://groups.csail.mit.edu/cb/multicoil/cgi-bin/multicoil.cgi ) was used to compare the input amino acid sequence to a database of known parallel two-stranded coiled-coils. A cutoff of 0.5 was applied for scoring the coiled-coil structures present in the PvMSP3 predicted proteins. The putative signal peptide cleavage site was predicted using SignalP 3.0 (http://www.cbs.dtu.dk/services/SignalP/) which uses the neural network model [65]. The TMHMM Server v. 2.0 (http://www.cbs.dtu.dk/services/TMHMM-2.0/) was used to search for transmembrane domains. The GPI Modification Site Prediction algorithm by Eisenhaber’s method [66] was used to evaluate the presence of post-translational modifications (http://mendel.imp.ac.at/gpi/gpi_server.html ). Conserved motifs were evaluated using the MEME package (http://meme.sdsc.edu/meme4/cgi-bin/meme.cgi).