CLAG3 Self-Associates in Malaria Parasites and Quantitatively Determines Nutrient Uptake Channels at the Host Membrane

ABSTRACT Malaria parasites increase host erythrocyte permeability to ions and nutrients via a broad-selectivity channel known as the plasmodial surface anion channel (PSAC), linked to parasite-encoded CLAG3 and two associated proteins. These proteins lack the multiple transmembrane domains typically present in channel-forming proteins, raising doubts about their precise roles. Using the virulent human Plasmodium falciparum parasite, we report that CLAG3 undergoes self-association and that this protein’s expression determines channel phenotype quantitatively. We overcame epigenetic silencing of clag3 paralogs and engineered parasites that express two CLAG3 isoforms simultaneously. Stoichiometric expression of these isoforms yielded intermediate channel phenotypes, in agreement with observed trafficking of both proteins to the host membrane. Coimmunoprecipitation and surface labeling revealed formation of CLAG3 oligomers. In vitro selections applied to these transfectant lines yielded distinct mutants with correlated changes in channel activity. These findings support involvement of the identified oligomers in PSAC formation and parasite nutrient acquisition.

While these molecular studies have convincingly implicated these parasite genes in nutrient uptake, fundamental questions remain about how the encoded proteins determine PSAC activity and the importance of increased permeability to intracellular parasite development. CLAG3.1 and CLAG3.2, related paralogs encoded by chromosome 3 in P. falciparum, are exported to the host membrane, where they are integral and susceptible to extracellular protease treatment (14). Nevertheless, the absence of unambiguous transmembrane domains and the lack of homology to known ion channels casts doubt on whether CLAG3 can form an ion channel pore. While association with RhopH2 and RhopH3 might facilitate pore formation, these other proteins do not appear to be exposed at the host cell surface (11). Molecular studies to define the role of CLAG3 have been complicated by epigenetic silencing, which allows infected cells to express only one clag3 paralog for several generations before switching to the other paralog (15,16). While these two genes encode proteins that are 90 to 95% identical, a short polymorphic motif is exposed at the erythrocyte surface and may affect function (8). This epigenetic switching confounds DNA transfection studies designed to examine channel phenotypes. Yet another difficulty is clag3's large size, which complicates full gene replacements through technologies such as CRISPR-Cas9.
Here, we overcame these problems by introducing a silent attB site in the last clag3 intron, allowing facile integrations through a novel application of Bxb1 integrase transposition technology. This not only allowed the efficient replacement of the downstream open reading frame (ORF) but permitted constitutive expression of two distinct CLAG3 isoforms at stoichiometric levels within individual parasites. Simultaneous expression of two isoforms yielded transport with intermediate properties. Pulldown experiments detected specific CLAG3 oligomerization, suggesting formation of an aqueous pore. We also used the resulting parasite as well as a distinct merodiploid parasite produced through random piggyBac insertional mutagenesis to examine how parasite cultures respond to selective pressure against individual CLAG3 isoforms. The molecular changes identified through these selections indicate that the intracellular parasite requires channel-mediated nutrient uptake and help define the roles served by these proteins. Our studies support direct involvement of CLAG3 in formation of PSAC, an essential nutrient uptake channel.

RESULTS
An intronic attB site facilitates production of tandem CLAG3 isoforms with separate epitope tags. While most lab and field isolates have two clag3 genes under epigenetic control (referred to as clag3.1 and clag3.2 [4,7,10,15]), some parasites have a single hybrid copy termed clag3h formed by crossover homologous recombination to yield copy number reduction (17). The KC5 parasite line carries a single clag3h gene and was chosen for our transfections to simplify interpretation; in KC5, the crossover occurred between conserved polymorphisms located 607 and 844 bp from the start of clag3h, with the 5= end derived from clag3.2 and the majority of the gene derived from an ancestral clag3.1 (see Fig. S1 in the supplemental material). We first introduced a silent attB site into the last intron of the clag3h gene by single homologous recombination (Fig. 1A). To facilitate preferential recombination upstream of this intron, the transfection plasmid carried a recodonized last exon whose product matches that of the clag3.1 gene from the Dd2 laboratory line (sLE Dd2 [Fig. S2A] in the supplemental material). PCR confirmed integration of this plasmid in the C3attB limiting dilution clone (Fig. S2B); DNA sequencing confirmed the expected introduction of the attB element flanked by intronic sequences with canonical splice sites. Immunoblotting revealed that the transfectant produces a CLAG3 protein of unchanged size (Fig. 1B). We then quantified PSAC activity by tracking osmotic lysis of infected cells in sorbitol (red trace, Fig. 1C), a sugar alcohol taken up primarily via the CLAG3-associated PSAC. A statistically insignificant increase in the osmotic lysis halftime indicates preserved sorbitol permeability ( Fig. 1D; P ϭ 0.09, n ϭ 5 to 9 each) and indicates that the silent attB element is not detrimental.
We predicted that replacement of the KC5 C-terminal sequence with the Dd2 clag3.1 last exon would produce channels inhibited by ISPA-28, a Dd2-specific inhibitor that enabled genetic mapping of PSAC phenotypes to clag3 (4,5). This prediction is based on the observation that inherited polymorphisms account for the specific action on Dd2 CLAG3.1 and that most of the polymorphisms are within a 10-to 30-residue motif encoded by the last exon. This hypervariable region (HVR) is exposed at the host erythrocyte surface (14). Consistent with our prediction, ISPA-28 potently blocked channels on C3attB but had negligible effect on uptake by wild-type KC5 parasites (Fig. 1E). Channels from the untransfected 3D7 parasite are also not blocked; while both KC5 and 3D7 carry genes that resemble Dd2 clag3.1, they have distinct HVR sequences. Dose-response studies revealed that the ISPA-28 block in C3attB matches that of untransfected Dd2 parasites (Fig. 1F). Thus, the last exon of clag3 is sufficient to fully define channel block and binding affinity for ISPA-28; these findings implicate the unique HVR sequence of Dd2 CLAG3.1 in ISPA-28 binding and block.
The silent attB introduced into the last intron should enable rapid transfection to insert distal epitope tags and regulatory elements. As proof of principle, we found that a C-terminal hemagglutinin (HA) epitope tag could be readily added (Fig. S3). To explore whether more-complex sequences can be inserted at this site, we designed and performed a transfection to allow simultaneous expression of tandem full-length clag3 isoforms under the control of the endogenous KC5 clag3h promoter (Fig. 1A). The pLN-attP-Dd2nLE-3D7FL plasmid consists of the attP sequence required for recombination with the genomic attB site, the distal intronic sequence, the native Dd2 clag3.1 last exon with a C-terminal HA tag, an in-frame viral T2A ribosome skip peptide, and a full-length 3D7 clag3.1 cDNA terminated with tandem miniSOG and FLAG epitope tags (18). Although the two clag3 genes are large (5.2 kb each) and we sought epitope tags of distinct sizes on the two isoforms, the compact design and use of attB-attP recombination allowed synthesis of a manageable transfection plasmid of 11.6 kb size (Fig. 1A). Successful integration and transcription by the native clag3h promoter should yield bicistronic expression of two distinct CLAG3 isoforms because the T2A skip peptide functions to cleave the nascent protein into two polypeptides during translation (19). This strategy is designed to allow simultaneous expression of two distinct CLAG3 isoforms, which is normally prevented by epigenetic silencing and monoallelic expression.
After this second transfection and limiting dilution, we obtained the TFLC3 clone (for "tandem full-length clag3") with the expected integration (Fig. S2C). Southern blotting with a probe mixture designed to interact promiscuously with all clag3 sequences yielded a single band in KC5, a smaller band that reflects two digested fragments of nearly equal size in C3attB, and multiple bands corresponding to 4 distinct sites of probe hybridization after the second transfection ( Fig. 1G and increased-exposure blot in Fig. S2D). Immunoblotting with antibodies to the HA and FLAG epitope tags identified the expected addition of both tags through a single transfection ( Fig. 1H and I). These large proteins were faithfully separated at the viral T2A skip peptide, yielding bands consistent with their calculated sizes (163 and 171 kDa for the tagged Dd2 and 3D7 isoforms, respectively); an~330-kDa band indicative of uncleaved fusion protein was not detected, suggesting efficient ribosome skipping at the T2A sequence and/or rapid degradation of residual fusion protein. Finally, both CLAG3 isoforms fractionated to the membrane pellet upon ultracentrifugation, as previously demonstrated for native CLAG3 (4).
Our transfection strategy predicted that TFLC3 parasites will simultaneously express the two CLAG3 isoforms under a single clag3h promoter. Indirect immunofluorescence microscopy indeed revealed that each infected cell was recognized by both anti-HA and anti-FLAG antibodies ( Fig. 2A). In schizont-infected cells, the two proteins colocalized as discrete puncta within developing merozoites ( Fig. 2A, top row), consistent with CLAG3 packaging into apical rhoptry organelles (20). As observed for CLAG3 and other rhoptry proteins (11), both isoforms were transferred to new erythrocytes after egress and reinvasion at the end of the intracellular cycle. Upon maturation of these daughter parasites to the trophozoite stage ( Fig. 2A, bottom row), both isoforms were observed to be exported into host cytosol and again colocalized at the cell periphery, suggesting association with the erythrocyte membrane.
Because proteolysis within its extracellular HVR is the hallmark of the CLAG3 pool integral to the host membrane (14), we then performed immunoblotting after extracellular protease treatment and found that both CLAG3 isoforms are integral at the host membrane in TFLC3. Pronase E yielded proteolytic fragments of 35 to 45 kDa in size when probed with antibodies to the HA and FLAG epitope tags present on the Dd2 and 3D7 isoforms, respectively ( Fig. 2B and C). Importantly, while immunoblotting with the C3attB parent produced a single cleavage product when probed with a polyclonal antibody to the conserved CLAG3 C terminus (4), TFLC3 produced two cleavage products ( Fig. 2D and E). The distinct sizes of these two products were anticipated by our transfection strategy because of the differing C-terminal epitope tags added to the two CLAG3 isoforms. A 7.3-kDa-larger size for the tagged 3D7 isoform is predicted based on protease-mediated cleavage at the distal ends of each CLAG3 HVR domain, the tandem miniSOG and FLAG epitope tag molecular weights, and the segregation site within the T2A skip peptide (19). Although this difference in size could not be easily confirmed for the full-length CLAG3 isoforms, it is apparent for the two proteaseinduced cleavage products because of clearly resolved migration of the smaller C-terminal fragments (Fig. 2D).
CLAG3 isoforms interact with each other in stable complexes. We next explored interactions between the two CLAG3 isoforms using coimmunoprecipitation (co-IP) experiments. Pulldown using anti-FLAG beads and the TFLC3 parasite revealed a stable association between the FLAG-tagged 3D7 CLAG3 protein and the HA-tagged Dd2 CLAG3 protein (Fig. 3A). A control transfectant (CLAG3-HA), also made using KC5 parasites to introduce an HA-tagged Dd2 clag3.1 terminal sequence but without a FLAG-tagged bait protein (Fig. S4A), did not yield detectable protein in pulldowns, excluding nonspecific binding of HA-tagged CLAG3 to anti-FLAG beads under our conditions. The reverse co-IP experiment using anti-HA beads confirmed the association in TFLC3, as the FLAG-tagged 3D7 allele could also be pulled down (Fig. 3B). Here, a separate control transfectant expressing these two CLAG3 alleles but without HA tagging of the Dd2 allele, miniSOG2 ( Fig. S4B to D), excluded nonspecific binding of either the FLAG tag or the 3D7 isoform to anti-HA beads. These experiments confirmed a stable association between the two CLAG3 isoforms and implicate formation of CLAG3 hetero-oligomers within trophozoites.
Previous imaging and cell fractionation experiments have revealed distinct subcellular pools of CLAG3 within maturing trophozoites that presumably reflect protein in transit to the host membrane (11,21). As CLAG3 monomers may self-associate at only some of these sites, we next used immunoblotting to estimate the fraction of HAtagged protein that is coprecipitated on anti-FLAG beads in TFLC3. By comparison to serial dilutions of the input lysate and measurement of band intensities, we estimated that between 1 and 10% of the HA-tagged Dd2 isoform is in stable association with the FLAG-tagged 3D7 CLAG3 protein (Fig. 3C). Because this experiment identifies only complexes that contain at least one copy of each CLAG3 isoform, it likely underestimates the fraction of HA-tagged protein in oligomers; the actual value will depend on whether the two isoforms assort randomly into oligomers and the presently unknown number of monomers in each membrane complex. CLAG3 self-association may be further underestimated if our detergent and immunoprecipitation conditions compromise the integrity of protein complexes.
We also performed blue native PAGE (BN-PAGE) experiments to examine formation of complexes containing more than one CLAG3 monomers but found that BN-PAGE conditions yield anomalous migration of these proteins. n-Dodecyl ␤-D-maltoside (DDM) or a low 0.05% concentration of FC-12 detected CLAG3 within a high-molecularweight complex that migrated near the 720-kDa marker (Fig. S5A); a similar band detected when probing for RhopH3 suggests that this represents the intact RhopH complex (Fig. S5B). Addition of SDS, a denaturing detergent, to either of these samples disrupted the complex and yielded a single CLAG3 band migrating near the 480-kDa marker. Increasing the FC-12 concentration to 1% also disrupted the complex and yielded similar CLAG3 migration; addition of SDS to this sample did not further increase mobility. Thus, this band likely represents a CLAG3 monomer that migrates at 400 to 500 kDa instead of 160 kDa as expected from its gene model; RhopH3 also exhibited anomalous migration in BN-PAGE (Fig. S5B), as has been observed for many membraneassociated proteins (22). SDS-PAGE followed by silver staining also suggested that 1% FC-12 disrupts the complex, as CLAG3 could not coimmunoprecipitate either RhopH2 or RhopH3 from total cell lysates, whereas these proteins were readily recovered in experiments using 1% DDM (Fig. S5C). Finally, coimmunoprecipitation (co-IP) using the TFLC3 merodiploid parasite after solubilization with 1% FC-12 revealed that the FLAG and HA isoforms did not remain associated (Fig. S5D), in contrast to the results with either Triton X-100 or DDM. Each of these observations confirms that 1% FC-12 denatures the RhopH complex and that CLAG3 migration is anomalous in BN-PAGE. In light of these findings, caution is required in interpreting the recent proposal of CLAG3 homotrimers based solely on FC-12 solubilization and the apparent molecular weight in BN-PAGE (23). Co-IP using TFLC3 parasites expressing two CLAG3 isoforms, as we have used here (Fig. 3), provides more direct experimental evidence supporting CLAG3 self-association.
Both CLAG3 isoforms contribute to PSAC-mediated nutrient uptake. The tandem CLAG3 isoforms expressed in TFLC3 are separated into two proteins during translation, and traffic to the host membrane; our co-IPs suggest that they assemble into complexes containing two or more CLAG3 monomers, presumably still in association with other RhopH proteins. Depending on subunit stoichiometries and random assortment into aggregates containing the tagged Dd2 and 3D7 CLAG3 isoforms, we propose that there may be at least three distinct complexes at the host cell surface in TFLC3 (Fig. 4A). To explore this model and how these complexes contribute to formation of functional channels, we next measured organic solute uptake and quantified block by ISPA-28, an inhibitor specific for the Dd2 CLAG3.1 isoform. While this blocker potently inhibits sorbitol uptake by cells infected with Dd2 parasites (K 0.5 of 56 Ϯ 5 nM [4]), it has negligible activity against PSAC associated with 3D7 parasites ( Fig. 1E and 4B, K 0.5 Ͼ 15 M). Importantly, TFLC3 produced PSAC activity intermediate between these two wild-type parasites ( Fig. 4B and F), indicating that each of the isoforms is functional at the host membrane.
We next examined CLAG3 trafficking and channel phenotype using the miniSOG2 parasite, which is also a CLAG3 merodiploid parasite that instead expresses the Dd2 and 3D7 isoforms under separate promoters ( Fig. 4C and S4B to D). Because this parasite was generated using piggyBac transfection (24), the 3D7 CLAG3 isoform is expressed from a distinct genomic site after transposase-mediated integration of the inverted terminal repeat (ITR) cassette. Southern blotting confirmed a single integration event without residual episomes in the miniSOG2 clone (Fig. S4D). In this parasite, the 3D7 allele carried the same tandem miniSOG-FLAG epitope tag at its C terminus as in TFLC3 but with expression under a heterologous msp2 promoter. We selected this promoter because msp2 exhibits a stage-specific transcription profile similar to that of clag genes in P. falciparum (25). This choice yielded CLAG3 expression and faithful trafficking to the rhoptry within schizonts and export to the host compartment in trophozoites (Fig. 4D). Despite use of a heterologous promoter and expression from a distinct genomic locus, the tagged 3D7 isoform successfully integrated at the host membrane, as confirmed by susceptibility to extracellular protease (Fig. 4E). As observed with TFLC3, this merodiploid also produced two cleavage products when probed with anti-CLAG3 antibody (Fig. 4F) and an intermediate ISPA-28 inhibition phenotype ( Fig. 4B and G). Thus, CLAG3 protein contribution to channel-mediated solute uptake does not depend on genomic site, the specific promoter used, or polypeptide separation within the T2A linker. Notably, the ISPA-28 susceptibilities of miniSOG2 and TFLC3 parasites were indistinguishable (P ϭ 0.97 at 15 M inhibitor, n ϭ 12 to 14 trials each). Thus, findings with two distinct merodiploid parasites suggest that the identified CLAG3 oligomers are physiologically significant.
In vitro selections against CLAG3 merodiploids: phenotype changes linked to distinct gene modifications. In contrast to their expression in TFLC3 and miniSOG2, the two clag3 genes in most wild-type parasites exhibit monoallelic expression through epigenetic control (8), yielding two subpopulations expressing one or the other allele (4). As predicted for subpopulations in steady-stage cultures, ISPA-28 has been previously used for purifying selections either for or against cells expressing the Dd2 clag3.1 isoform. Positive selection for this isoform has been demonstrated with brief daily incubations in sorbitol with ISPA-28 to produce osmotic lysis of cells that express clag3 isoforms linked to ISPA-28-insensitive channels; cells expressing Dd2 clag3.1 survive this treatment, as sorbitol uptake is blocked by the inhibitor (4). The reverse selection, which eliminates Dd2 clag3.1-expressing cells, was achieved through cultivation with ISPA-28 in a modified culture medium termed PGIM (PSAC growth inhibition medium); PGIM is based on the standard RPMI 1640 formulation but contains lower, more physiological levels of three key nutrients (5). Under these conditions, cells expressing Dd2 clag3.1 are killed because ISPA-28 prevents nutrient uptake required for parasite growth.
As TFLC3 and miniSOG2 express two clag3 isoforms without epigenetic control, their cultures do not contain subpopulations amenable to purifying selections. We initially predicted that they would therefore not tolerate stringent selections with ISPA-28 because they are constrained to express both sensitive and resistant CLAG3 isoforms simultaneously. We subjected both transfection clones to either the sorbitol or PGIM selection with ISPA-28, observed marked attrition of each culture, and were surprised as resistant mutants were obtained in all four cases (Fig. 5). After limiting dilution cloning, the two mutants obtained after daily sorbitol treatments, called TFLC3.sorb and miniSOG2.sorb, exhibited significantly improved PSAC block by ISPA-28 (P Յ 0.001 for comparisons to unselected merodiploid at a 15 M concentration; n ϭ 3 trials each); inhibition for both selected clones matched the dose responses for wild-type Dd2  Circles represent clones generated after treatment with sorbitol, and ISPA-28; triangles reflect clones obtained after cultivation in PGIM with this inhibitor. These clones exhibit potencies that match those of Dd2 or 3D7 (reflected by the solid lines, reproduced from the best fits in Fig. 4G).
parasites (lower solid line, Fig. 5), suggesting complete loss of the 3D7 CLAG3 protein's contribution to PSAC activity.
Because these changes were unexpected, we explored the possible molecular mechanisms. PCR and DNA sequencing revealed that TFLC3 survived our in vitro selections through growth of two distinct transfection revertants. The TFLC3.pgim clone underwent ectopic recombination to loop out both the pCC1-attB-Dd2sLE and pLN-attP-Dd2nLE-3D7FL plasmids, presumably as a single 21.6-kb concatemer carrying both the attL and attR sites (Fig. 6A and S6); DNA sequencing of the clag3 locus did not reveal any differences relative to the untransfected KC5 parent, suggesting homologous recombination between the identical upstream clag3 sequences (gray elements at the two ends of the TFLC3 ribbon schematic, Fig. 1A).
Although integrase-mediated recombination between attB and attP sites is generally considered irreversible (26), PCR revealed that TFLC3.sorb attained high ISPA-28 affinity by looping out only the pLN-attP-Dd2nLE-3D7FL plasmid ( Fig. 6A and S6). DNA sequencing revealed full restoration of the gene's last intron and attB site without modification, identical to its C3attB parent.
Despite exhibiting quantitatively identical changes in PSAC pharmacology, selections applied to the miniSOG2 merodiploid resulted from distinct molecular changes. Because it carries the Dd2 genetic background, the miniSOG2 parasite has three clag3 genes: the clag3.1 and clag3.2 genes native to Dd2 as well as the epitope-tagged clag3.1 of 3D7 introduced through transfection (Fig. 6B). To examine possible contributions from each gene, we began with real-time quantitative PCR (qRT-PCR). As observed with its Dd2 parent (4), miniSOG2 preferentially expressed the Dd2 clag3.1 gene, maintaining the native clag3.2 in a silenced state. Selection with PGIM and ISPA-28 largely reversed this preference (Fig. 6C, red bars showing a 40-fold-reduced expression of the Dd2 clag3.1 and a 7-fold increase for Dd2 clag3.2); these changes can account for the loss in ISPA-28 block and permit growth of miniSOG2.pgim despite restricted nutrient access. The 3D7 clag3.1 transgene did not exhibit altered expression, consistent with its expression under the msp2 promoter incapable of epigenetic upregulation.
As qRT-PCR did not reveal changes in the miniSOG2.sorb parasite (Fig. 6C, blue bars), we next used immunoblotting and found that the FLAG-tagged transgene product was rendered undetectable upon selection for improved ISPA-28 block (Fig. 6D). Immunoblotting with anti-CLAG3 revealed the native Dd2 CLAG3.1 protein (Fig. 6E); the observed faster migration of CLAG3 in miniSOG2.sorb than in miniSOG2.pgim is consistent with a lower molecular weight of this allele than that of the epitope-tagged 3D7 CLAG3.1 (Fig. 6B). DNA sequencing of the 3D7 transgene revealed a nonsynonymous point mutation that changes glutamine 271 to a stop (Fig. 6F), yielding a truncated 31-kDa protein not detectable by antibodies that recognize downstream epitopes. Thus, both increased and reduced ISPA-28 affinity were obtained through selections applied to the TFLC3 and miniSOG2 merodiploids. The parasite may also use other mechanisms, such as increases in clag3 copy number (5), to survive selections that target this essential nutrient uptake channel.

DISCUSSION
CLAG3 was identified as a critical determinant of infected-cell nutrient uptake through classical genetic mapping with ISPA-28, an isolate-specific PSAC inhibitor found through high-throughput screening (4). While DNA transfections, transport mutants, and functional proteolysis of channel-mediated uptake have provided independent evidence for CLAG3 involvement (6,7,9,14), exactly how this protein contributes to channel activity is unclear. As CLAG3 localizes to the host membrane, either it may directly contribute to pore formation or it may function as a modulator A Protein Complex for Plasmodium Nutrient Uptake ® that activates solute uptake through other pore-forming proteins (3). To explore these and other possible mechanisms, we used a novel transfection approach to produce parasites expressing two distinct CLAG3 paralogs with separate epitope tags. This TFLC3 transfectant and a separate miniSOG2 transfectant expressing these two paralogs under separate promoters show that CLAG3 can traffic normally to the host membrane when expression is driven under the msp2 promoter or when two isoforms are produced with an intervening T2A ribosome skip peptide. These fortuitous observations allowed us to determine that CLAG3 oligomerizes within infected cells. Functional studies with these transfectants then revealed that the relative expression of the Dd2 and 3D7 CLAG3 isoforms quantitatively defines PSAC block by the sequence-specific inhibitor ISPA-28. In vitro selections for resistance to PSAC-mediated osmotic lysis or blocked nutrient acquisition independently confirm the relationship between CLAG3 sequence and channel phenotype; these selections also revealed the molecular mechanisms available to parasites to evade selective pressures while preserving essential channel-mediated nutrient uptake.
So how does CLAG3 contribute to PSAC formation? Because CLAG3 is integral to the host membrane and has an amphipathic transmembrane domain (9), one model is direct CLAG3 contribution to the nutrient pore. Arguing against this model is the paucity of predicted transmembrane domains: most ion channels require a minimum of 3 to 5 alpha-helices to create a stable aqueous pore (27). Our finding that CLAG3 undergoes self-association provides a conservative explanation for this: higher-order stoichiometries with each monomer providing a pore-lining transmembrane domain could provide the required number of transmembrane domains. Indeed, many ion channels are formed by oligomerization of identical or highly related subunits (28)(29)(30). As CLAG3 associates with RhopH2 and RhopH3 to form a stable complex (15,16,31), these other proteins may also contribute to the channel and provide additional transmembrane domains (11). This complicated heteromeric arrangement may also account for the many unusual features of this nutrient uptake channel (32,33).
Further studies will be required to conclusively determine whether the RhopH complex forms the nutrient pore directly. While heterologous expression could resolve this, the member proteins are large and do not express well in standard expression systems. This endeavor will be further complicated by required cotranslational assembly of the three-protein complex, proteolytic processing of RhopH3 that may be necessary for activation of the pore, and complex trafficking and eventual insertion at the host membrane (11). As CLAG3 must transition from peripheral association with membranes to an integral form at the host cell surface to contribute to PSAC formation (14), each of these steps may require parasite-specific proteases and chaperones that are absent from heterologous expression systems.
Our findings suggest a higher-order stoichiometry to the RhopH complex, with at least two CLAG3 monomers in association with an unknown number of RhopH2 and RhopH3 proteins. Native PAGE, a method that could in principle determine the precise subunit stoichiometry (34), unfortunately cannot be used with confidence here. We determined that these proteins exhibit anomalous migration: under conditions that dissociate the complex and impose migration as monomers, native PAGE overestimates the molecular weights of both CLAG3 and RhopH3 monomers (see Fig. S5 in the supplemental material). As incomplete cleavage of the T2A skip peptide between CLAG3 monomers in TFLC3 might yield higher-than-expected molecular weights, our native PAGE experiments were performed exclusively with HB3 3rec , a transfectant line that expresses a single CLAG3 isoform and does not utilize skip peptides. While the reasons for anomalous migration are unclear, a survey of membrane proteins revealed that a 2-fold-increased apparent molecular weight could be attributed to Coomassie blue binding during migration in native PAGE (35); the same study found substantial effects of detergent choice also. In the case of CLAG3, the apparent molecular weight may also be affected by this protein's highly charged C terminus (36). Another important question is whether this complex also contains CLAG paralogs from other chromosomes (termed CLAG2, CLAG8, and CLAG9). A previous study found that immuno-precipitation with CLAG9-specific antibody does not pull down either CLAG3 or CLAG2 (10), but the subsequent determination that these proteins are integral to membranes and that the complex is unstable in some detergents requires cautious interpretation of early findings. The low levels of CLAG3 self-association identified by these coimmunoprecipitation experiments may result from several factors. Because trophozoite-stage immunofluorescence assays (IFAs) and protease susceptibility studies reveal that a small fraction of CLAG3 is successfully delivered and integrated at the host membrane (e.g., Fig. 2), one explanation is that the intracellular pools may be present as monomers and that CLAG3 oligomerization occurs shortly before or upon insertion at the host membrane, a model with good precedent (37). Another possibility is that solubilization with detergents and available coimmunoprecipitation procedures, despite our efforts to optimize these steps, may compromise RhopH complex integrity and underestimate CLAG3 oligomerization (Fig. S5). Yet another consideration is whether the two CLAG3 isoforms that we used, derived from the clag3.1 products of Dd2 and 3D7 parasites, associate with each other randomly; if these proteins preferentially self-associate, our approach would again underestimate CLAG3 oligomerization. Despite these uncertainties, our reverse immunoprecipitations with two negative-control transfectants suggest that the observed low levels of oligomerization do not represent an experimental artifact.
While the model of CLAG3 oligomerization is appealing, many important questions remain. Because the quantitative correlations that we found between CLAG3 sequence and channel phenotype could also be observed with channels that contain only CLAG3 monomers, we cannot conclusively assign a role to the identified oligomers. Because the measured oligomerization is low relative to the total cellular pool, our various transfectant lines and biochemical studies could also not address whether the CLAG3 oligomers remain associated with RhopH2 and RhopH3, proteins additionally linked to PSAC formation (11). If the resulting heteromeric complex does form the channel, the absolute stoichiometries and functional contributions to pore formation will require exploration for each subunit. Definitive resolution of these uncertainties will require a combination of molecular, structural, and biochemical studies. The large size of RhopH member proteins, epigenetic regulation of clag paralogs, and compensatory changes in the parasite genome upon DNA transfection should all be considered carefully in these studies.
From the perspective of host-pathogen interactions, our study provides insights into genome-level changes available to parasites when subjected to strong selective pressures such as osmotic destabilization or target-specific drugs. Based on prior selections targeting PSAC (5), we initially predicted that small-volume cultures of either TFLC3 or miniSOG2 would be unable to survive selection either for or against ISPA-28-mediated block because our transfections imposed intermediate ISPA-28 affinity through forced coexpression of two alleles. In contrast to our predictions, both clones rapidly acquired genome-level changes that conferred a quantitatively complete change in ISPA-28 affinity and allowed mutant progeny expansion. We identified epigenetic silencing of the native Dd2 allele, truncation of the 3D7 allele, and loop-out of attB and/or attP plasmids as strategies available to the parasite.
The loop-out of the attP plasmid alone, as observed when TFLC3 was pressured to increase ISPA-28 block of sorbitol uptake (Fig. 6A and S6), may be the most surprising of these rapidly acquired genome-level changes. This loop-out corresponds to recombination between the attL and attR sites and is the reverse of the attB-attP recombination mediated by Bxb1 integrase. As Bxb1 integrase is generally considered incapable of this reverse reaction in isolation (26), we used PCR to determine that the TFLC3 clone no longer carries the pINT helper plasmid encoding the integrase. Diagnostic PCR also excluded low-level retention of parasites that failed to integrate the attP plasmid in the TFLC3 limiting dilution clone (lane 4, Fig. S6B). A more likely scenario is that the attP plasmid was removed through an alternative form of nonhomologous end joining after a coincidental genome break. Although canonical nonhomologous end joining is not thought to occur in P. falciparum due to lack of critical enzymes (38), some alternative mechanisms that utilize very short regions of homology termed "microhomologies" have been observed infrequently (39,40). While such mechanisms are typically associated with a residual indel or inversion (41), we could not identify any CLAG3 sequence differences between this selected mutant and the original C3attB clone, implicating a clean loop-out of the attP transfection plasmid.
Allelic exchange transfection via a silent attB site engineered into a gene's intron, as we demonstrate here, represents a useful tool for molecular and cellular studies of gene function. While similar introduction of loxP sites into an intron has been used in parasite transfections (42), the present approach is more versatile as a single initial transfection to engineer the silent attB enables the full range of downstream modifications-gene truncation, epitope tagging, and insertional and site-directed mutagenesis. While CRISPR-Cas9 transfection will be the method of choice for most gene editing experiments in P. falciparum, it is worth noting that CRISPR may not be ideal for insertion of large sequence elements as it requires homologous recombination on both sides of the insert. Another recently described approach, selection-linked integration (43), should also work for most targets but requires an AT-rich homology arm for upstream recombination; it may also fail for proteins with low or undetectable expression in asexual-stage cultures because it depends on the target gene's promoter to drive selectable marker expression. For cases such as the Ͼ5-kb insert that we required to produce TFLC3 (Fig. 1A), integration distal to an engineered silent attB site may be the most robust approach.
Plasmid construction and DNA transfection. Plasmids were prepared by standard methods and confirmed by DNA sequencing and restriction digestion. Primers used are tallied in Table S1 in the supplemental material.
The pCC1-120w-attB plasmid was used to introduce a silent attB site within the last intron of the KC5 clag3h gene by single-crossover homologous recombination. A codon-optimized sequence for the last exon facilitated recombination upstream of the desired attB sequence. While the upstream homologous sequence was obtained by PCR amplification from KC5, the last exon's sequence was synthesized from the Dd2 CLAG3.1 sequence (17). Parasite transfection was performed by plasmid electroporation into uninfected erythrocytes prior to addition of trophozoite-infected cells for in vitro culture (44). After selection with 2 nM WR99210 (Jacobus), parasite growth was detected with Giemsa-stained slides after 3 weeks; PCR was then used to detect integration prior to limiting dilution cloning to obtain C3attB (45). The Bxb1 integrase system was used for efficient integration at the engineered attB (46); this second transfection utilizes two plasmids. The pLN-attP plasmids carry the attP site and downstream sequences to be integrated, while the pINT helper plasmid encodes Bxb1 integrase. C3attB parasites were transfected with these two plasmids and selected with 2.5 g/ml blasticidin S and 125 g/ml G418 (Thermo Fisher). Parasite growth and integration were typically detected in~3 weeks.
The piggyBac method was used to integrate a full-length clag3.1 gene at a distant genomic site through random insertion at one or more transposon target sites (24). The clag3.1 gene from the 3D7 parasite, an in-frame tandem miniSOG-FLAG epitope tag, and the clag3.1 3= untranslated region (UTR) were cloned behind the msp2 promoter. This cassette was inserted into the pXL-BacII-hDHFR vector between the inverted terminal repeats (ITRs), which are recognized by the piggyBac transposase. The Dd2 parasite line was transfected with this plasmid and the pHTH helper plasmid encoding piggyBac (24). Selection with WR99210 for the hDHFR cassette present between the ITRs yielded a transfection pool that was then cloned to obtain the miniSOG2 line. Southern blotting was used to evaluate retention of pXL-M2-120w-miniSOG episomes and the number of genomic integration events.
Southern blot hybridization. Genomic DNA was extracted using the Quick-gDNA MiniPrep kit (Zymo Research), digested with indicated enzymes, resolved on an 0.7% agarose gel, acid depurinated, and blotted overnight onto an Amersham Hybond N ϩ charged membrane (GE Healthcare). A digoxigenin (DIG)-labeled-DNA probe complementary to hdhfr was prepared using primers 5=-ATTTCCAGAGAA TGACCACAAC-3= and 5=-TTAAGATGGCCTGGGTGATTC-3= and used to evaluate integration in miniSOG2. A separate DIG-labeled DNA probe directed against a conserved region of the clag3 last exon was prepared with 5=-TTTTTCATGGAACTTGCAAATGGTTTCATGTATGC-3= and 5=-GACATTTTGTTTTTGAAAGCA TTTTCCATATC-3=. PCR was performed with KC5 genomic DNA, pCC1-120w-attB, and pLN-attP-Dd2nLE-3D7FL as the templates. The amplified probes were mixed at equal concentrations and used; these probes were designed to bind clag3.1-like sequences from KC5, Dd2, and 3D7 as well as the codonoptimized fragment present in C3attB parasites without recognizing clag paralogs from other chromosomes. After prehybridization of cross-linked membrane with DIG-Easy Hyb (Roche), the labeled probe was added and hybridized overnight at 42°C. The blot was washed with low-stringency buffer (0.1ϫ SSC [1ϫ SSC is 0.15 M NaCl plus 0.015 M sodium citrate]-0.1% SDS) at 50°C and blocked. Probe binding was then detected with antidigoxigenin-alkaline phosphatase (AP) Fab fragments at a dilution of 1:10,000 and CDP-Star substrate (Roche).
In vitro selections of parasite transfectants. Synchronous cultures of transfected parasite lines were subjected to osmotic lysis selection by incubation in 5% sorbitol with 0.6 M ISPA-28 at room temperature for 30 min on 4 consecutive days with cultivation in standard RPMI 1640-based medium as described previously (4). Subsequent cultivation and limiting dilution cloning yielded the TFLC3.sorb and miniSOG2.sorb lines.
Reverse selection to reduce ISPA-28 sensitivity was performed by cultivation of synchronous cultures at initial parasitemias of 1 to 2% in PGIM supplemented with 5 M ISPA-28 continuously for 4 days. The culture was then washed and cultivated in standard medium prior to limiting dilution cloning to obtain TFLC3.pgim and miniSOG2.pgim.
Healthy parasites could not be identified immediately after completion of each of these selections, but regular microscopic examination revealed recovery after approximately 2 weeks. Lines were cryopreserved shortly thereafter. Selection pools were subjected to initial phenotyping and molecular characterization prior to limiting dilution cloning (45); each limiting dilution clone was phenotypically identical to the corresponding selection pool.
Solute transport measurements. Channel-mediated transport was quantified with a kinetic, transmittance-based assay that tracks osmotic lysis of infected cells in sorbitol, a sugar alcohol with high PSAC permeability (47). Trophozoite-stage-infected erythrocytes were enriched with the Percoll-sorbitol method, washed, and resuspended at 37°C in 280 mM sorbitol, 20 mM Na-HEPES, 0.1 mg/ml bovine serum albumin, pH 7.4, to initiate sorbitol uptake and osmotic lysis; indicated concentrations of ISPA-28 were added without preincubation. Lysis was continuously tracked by measuring transmittance of 700-nm light (DU640 spectrophotometer with Peltier temperature control; Beckman Coulter). Inhibitor dose-response relationships were calculated from the time to a fractional lysis threshold. ISPA-28 dose-response data were fitted to the sum of two Langmuir isotherms with the following equations: where P represents the normalized sorbitol permeability in the presence of inhibitor at concentration x and a, b, and c are constants. ISPA-28 block contrasts with that of other potent inhibitors, whose dose responses are adequately fitted by a single Langmuir isotherm (5); this difference corresponds to a lack of saturating block despite use of high ISPA-28 concentrations.
Protease susceptibility and membrane fractionation. Enriched trophozoite-infected erythrocytes were suspended in phosphate-buffered saline (PBS) supplemented with 0.6 mM CaCl 2 and 1 mM MgCl 2 (PBS2) at 5% hematocrit prior to treatment with 1 mg/ml pronase E (Sigma-Aldrich) for 1 h at 37°C. Protease treatment was terminated by adding an equal volume of ice-cold PBS2 with 2 mM phenylmethylsulfonyl fluoride (PMSF) and 2 mM EDTA. After extensive washing with PBS2, the treated cells were lysed in ice-cold hypotonic lysis buffer (7.5 mM NaHPO 4 , 1 mM EDTA, pH 7.5) with 1 mM PMSF and centrifuged (100,000 ϫ g, 4°C, 1 h). The supernatant and pellet were collected as the soluble and membrane fractions. The membrane fraction was then resuspended in 100 mM Na 2 CO 3 , pH 11, for 30 min on ice before centrifugation (100,000 ϫ g, 4°C, 1 h). The resulting supernatant and pellet fractions represent the peripheral and integral membrane protein fractions; both were neutralized with a 1/10 volume of 1 M HCl and solubilized and denatured in Laemmli sample buffer containing 6% SDS.
Coimmunoprecipitation. Enriched trophozoite-infected cells were lysed in 50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 5 mM EDTA with 1% Triton X-100; incubated for 30 min on a shaker at room temperature (RT); and centrifuged (12,000 ϫ g, 10 min, 4°C); other detergents were used in place of Triton X-100 as indicated. After reserving an aliquot as the input, the solubilized protein was incubated with washed anti-FLAG or anti-HA affinity agarose beads (Sigma-Aldrich) with gentle mixing for a minimum of 2 h at 4°C. We optimized the ratio of solubilized protein to affinity beads and confirmed complete bait protein binding with control immunoblot assays of the unbound supernatant after this incubation. The resin was then washed 5 times with IP wash buffer (50 mM Tris-HCl, pH 7.4, 50 mM NaCl, 5 mM EDTA) and the corresponding detergent. After an additional wash in SDS-free elution buffer, bound proteins were eluted into 100 mM Tris, pH 6.8, with 4% SDS; elution was also performed using 3ϫ FLAG peptide (Sigma-Aldrich) in some experiments. After electrophoresis, proteins were evaluated by immunoblotting or silver staining (Pierce).
Blue native PAGE experiments used enriched HB3 3rec parasites freeze-thawed into PBS with 1 mM PMSF. The membrane fraction was solubilized in 150 mM NaCl, 10 mM Tris, 1 mM EDTA, pH 7.5, with 1 mM PMSF and indicated detergents at 4°C for 15 min with gentle mixing. After centrifugation at 150,000 ϫ g to remove insoluble material, a fraction was treated with 0.1% SDS for 30 min at 4°C to disrupt protein complexes. Native PAGE sample buffer was added to a 1ϫ concentration; Coomassie blue G-250 was then added to 25% of the total detergent concentration prior to electrophoresis at 4°C (native PAGE 4 to 16% Bis-Tris gel; Thermo Fisher). A voltage of 150 V was applied for 30 min with Dark Blue Cathode buffer containing 0.02% Coomassie blue G-250 prior to replacement with Light Blue Cathode buffer (0.002% Coomassie blue G-250) and resumption of electrophoresis at 150 V for 30 min and 250 V for 60 min based on the manufacturer's recommendations. The gel was then incubated with transfer buffer (25 mM Tris-HCl, 192 mM glycine, 10% methanol, and 0.1% SDS) for 30 min prior to transfer to polyvinylidene difluoride (PVDF) at 150 mA for 90 min. After fixation with 5% acetic acid and Coomassie blue removal with methanol, blots were blocked with 150 mM NaCl, 20 mM Tris-HCl, pH 7.4, and 0.1% Tween 20 with 3% skim milk prior to antibody application and detection as described above.
Immunofluorescence microscopy. Indirect immunofluorescence confocal microscopy was performed with thin smears of parasite cultures. Dried slides were fixed with a chilled 1:1 mixture of acetone and methanol before blocking in 3% skim milk in PBS for 1 h at RT. Slides were then incubated in blocking buffer with indicated primary antibodies at a 1:100 dilution for 1 h at RT (mouse anti-HA [Sigma-Aldrich] and rabbit anti-FLAG [Cell Signaling]). After washing, the slide was incubated with 4=,6-diamidino-2-phenylindole (DAPI) (2 g/l) and goat anti-mouse Alexa Fluor 488 or goat anti-rabbit Alexa Fluor 594 (Thermo Fisher) for 30 min at RT. Slides were then washed in ice-cold PBS for 5 min before mounting with ProLong Diamond antifade mountant (Molecular Probes). Images were collected under a 64ϫ oil immersion objective on a Leica SP5 confocal microscope and processed using Leica LAS X software.
qRT-PCR. Normalized expression levels for the three copies of clag3 in miniSOG2 parasites before and after in vitro selection were measured with real-time PCR. Total RNA was harvested from cultures 27 h after sorbitol synchronization with the Ambion PureLink RNA minikit. Contaminating DNA was removed with RNase-free DNase I (Ambion) prior to first-strand cDNA synthesis using reverse transcription with SuperScript III (Invitrogen), 1.7 g RNA, and oligo(dT) primers. The constitutively expressed parasite gene PF07_0073 was used for normalization of gene expression among the three examined parasite lines as described previously (6).
Real-time PCR was performed with the synthesized cDNA, primers listed in Table S1 in the supplemental material, and the QuantiTect SYBR Green kit (Qiagen). Primers for real-time quantitative PCR (qRT-PCR) were designed based on specificity for individual clag3 paralogs and a desired amplicon size of~120 bp. qRT-PCR was carried out using the iCycler IQ multicolor real-time PCR system (Bio-Rad) according to the following procedure: denaturation at 95°C for 15 min followed by 40 cycles of denaturing at 94°C, annealing at 55°C, and extension at 62°C (30 s each). After cycling, the reaction mixture was gradually heated from 55°C to 95°C in 0.5°C increments over 30 s to confirm the specificity of primer binding and product synthesis. Each qRT-PCR mixture was accompanied by a negative control without reverse transcriptase to exclude genomic DNA contamination. All reactions were performed in triplicate wells for each of three separate harvests. The average threshold cycle (C T ) values were used to obtain normalized gene expression according to 2^(mean C T value of the test clag3 gene Ϫ mean C T value of Pf07_0073 gene). Expression changes relative to the miniSOG2 parent were calculated with the 2 ϪΔΔCT method (48).
Statistical analysis. Numerical data are presented as mean Ϯ standard error of the mean (SEM). Statistical significance was determined with the unpaired Student t test or one-way analysis of variance (ANOVA) with Dunnett's multiple-comparison post hoc analysis. A threshold of P Ͻ 0.05 was used for significance.