Plasmodium sporozoites require the protein B9 to invade hepatocytes

Summary Plasmodium sporozoites are transmitted to a mammalian host during blood feeding by an infected mosquito and invade hepatocytes for initial replication of the parasite into thousands of erythrocyte-invasive merozoites. Here we report that the B9 protein, a member of the 6-cysteine domain protein family, is secreted from sporozoite micronemes and is required for productive invasion of hepatocytes. The N-terminus of B9 forms a beta-propeller domain structurally related to CyRPA, a cysteine-rich protein forming an essential invasion complex in Plasmodium falciparum merozoites. The beta-propeller domain of B9 is essential for sporozoite infectivity and interacts with the 6-cysteine proteins P36 and P52 in a heterologous expression system. Our results suggest that, despite using distinct sets of parasite and host entry factors, Plasmodium sporozoites and merozoites may share common structural modules to assemble protein complexes for invasion of host cells.

Infectious Plasmodium sporozoites express the 6-cys protein B9 B9 is secreted from a specific subset of micronemes B9 is required for sporozoite productive invasion of hepatocytes B9 contains an essential CyRPA-like beta-propeller domain

INTRODUCTION
Malaria is caused by Plasmodium spp. parasites and still remains a major health and socio-economic problem in endemic countries. 1 Sporozoites, the mosquito-transmitted forms of the malaria parasite, first infect the liver for an initial and obligatory round of replication, before initiating the symptomatic blood stages. Infection of the liver is clinically silent and constitutes an ideal target for a malaria vaccine. Until now, only a single antigen, the circumsporozoite protein (CSP), had been considered for clinical vaccine development against the extracellular sporozoite stage, with limited success. 2 Other sporozoite antigens, especially parasite proteins involved in host-parasite interactions, could be considered as potential vaccine targets to prevent sporozoite entry into hepatocytes. This highlights the need to better characterize the molecular mechanisms of sporozoite infection in order to identify new vaccine targets.
Like other Apicomplexan parasites, Plasmodium invades host cells using a unique mechanism that involves the sequential secretion of apical organelles, called micronemes and rhoptries, and the formation of a moving junction (MJ) through which the parasite actively glides to enter the cell and form a specialized parasitophorous vacuole (PV) where it further replicates. 3 Proteins released from micronemes onto the parasite surface are prime candidates to interact with host cell surface receptors, triggering subsequent secretion of the rhoptry content, formation of the MJ, and commitment to productive invasion. However, until now the ligand-receptor interactions mediating Plasmodium sporozoite invasion and the nature of the sporozoite MJ have remained enigmatic. 4 We previously characterized host entry pathways used by human (P. falciparum, P. vivax) and rodent (P. yoelii, P. berghei) parasites to infect hepatocytes 5,6 and showed that CD81 and the Scavenger Receptor class B type I (SR-BI) define independent entry routes for P. falciparum and P. vivax sporozoites, respectively. 6 Remarkably, this alternative usage of host cell receptors is also observed with rodent malaria model parasites, providing robust and tractable experimental systems. 6,7 Indeed, P. yoelii sporozoites, like P. falciparum, strictly require CD81 to infect liver cells, whereas P. berghei can alternatively use CD81 or SR-BI for productive invasion. 6 Only two parasite proteins, P36 and P52, have been identified as being specifically required for productive invasion of hepatocytes. 6,[8][9][10][11] Using interspecies genetic complementation in mutant P. berghei and P. yoelii lines, we showed that P36 is a key determinant of host cell receptor usage, establishing for the first time a functional link between sporozoite and host cell entry factors. 6 The molecular function of P36 remains unknown. One study proposed that P36 interacts with the ephrin receptor EphA2 on hepatocytes to mediate infection, 12 but direct evidence for such an interaction is lacking, and EphA2 was later shown to be dispensable for sporozoite productive invasion. 13 Interestingly, interspecies genetic complementation experiments showed that P. berghei Dp52Dp36 mutants complemented with PyP52 and PyP36 exhibit a P. yoelii-like phenotype as they preferentially infect CD81-expressing cells. 6 However, while P. yoelii sporozoites are unable to infect hepatocytes in the absence of CD81, complemented P. berghei mutants retain a residual invasion capacity in CD81-deficient cells. 6 Furthermore, genetic complementation with P. falciparum or P. vivax P52 and P36 does not restore infectivity of Dp52Dp36 P. berghei sporozoites. 6 These results strongly suggest that additional parasite factors contribute to receptor-dependent productive invasion. P36 and P52 both belong to the so-called 6-cysteine domain protein family, which is characterized by the presence of one or several 6-cysteine (6-cys) domains. 14 6-cys domains are $120 amino acid-long domains containing four or six conserved cysteine residues that respectively form two or three disulphide bonds resulting in a beta-sandwich fold. 14 Plasmodium spp. possess 14 members of the 6-cys protein family. 15 Plasmodium 6-cys proteins are typically expressed in a stage-specific manner and have been implicated in protein-protein interactions in P. falciparum merozoites, 16,17 gametocytes, 18,19 ookinetes, 20 and sporozoites. 11 Proteomic studies have shown that, in addition to P36 and P52, Plasmodium sporozoites express three other 6-cys proteins, P12p, P38, and B9. [21][22][23][24] While the contribution of P12p and P38 had not been studied until now, a previous study reported that the protein B9 is not expressed in sporozoites due to translational repression and is not required for sporozoite invasion of hepatocytes but is needed for early maintenance of the PV. 15 Here, we systematically analyzed the role of P12p, P38, and B9 during sporozoite invasion, using a reverse genetics approach based on our gene out marker out (GOMO) strategy. 25 We report that b9 gene deletion totally abrogates sporozoite infectivity, while p12p and p38 are dispensable for hepatocyte infection in both P. berghei and P. yoelii. We show that B9 is a sporozoite micronemal protein and that B9-deficient sporozoites fail to productively invade hepatocytes. Secondary structure analysis and protein structure modeling indicate that B9 is a hybrid protein containing a CyRPA-like beta-propeller domain in addition to noncanonical 6-cys domains. Structure-guided mutagenesis reveals that the propeller domain is not associated with host cell receptor usage but is essential for sporozoite infectivity, being required for adequate protein expression and/or function, possibly through the assembly of supramolecular protein complexes with the 6-Cys proteins P36 and P52.

RESULTS
Analysis of the repertoire of Plasmodium sporozoite 6-cys proteins suggests that P36, P52, and B9 are employed by infectious sporozoites only In order to define the repertoire of 6-cys proteins expressed at the sporozoite stage, we first analyzed the proteome datasets of P. falciparum, 22,23 P. vivax, 24 P. yoelii, 23 and P. berghei 21 sporozoites. As expected, P36 and P52 were identified by mass spectrometry in sporozoites from all four species. Interestingly, three other 6-cys proteins, P12p, P38, and B9, were consistently identified across the datasets. Among this core of five 6-cys proteins, P12p and P38 have been identified in the surface proteome of P. falciparum sporozoites, with P12p being quantitatively enriched on the surface of activated parasites in the presence of bovine serum albumin. 26 Interestingly, P12p and P38 do not seem to be uniquely employed by sporozoites as they have been detected in P. falciparum asexual and sexual blood stages 27-31 and in P. berghei gametocytes, 32 respectively. In contrast, P36, P52, and B9 were only identified in sporozoites, and a recent study identified P36, P52, and B9 as upregulated in infectious sporozoites (UIS) proteins in P. falciparum and P. yoelii, while P12p and P38 were also detected in oocyst-derived sporozoites. 33 These observations suggest that B9, like P36 and P52, may play a role in mature sporozoites.
Reverse genetics analysis in rodent malaria parasites shows that b9 (but not p12p and p38) is essential for sporozoite infectivity A previous study reported that B9 is not expressed in sporozoites and is required for early liver-stage development but not host cell invasion. 15 The contribution of P12p and P38 during sporozoite invasion has not been investigated so far, although the p38 gene could be deleted in P. berghei without any detectable phenotypic defect during blood-stage parasite growth and transmission to mosquitoes. 34 iScience Article consistent detection of P12p, P38, and B9 proteins in sporozoites by mass spectrometry, we sought to determine the functional importance of these proteins in P. berghei and P. yoelii sporozoites using a reverse genetics approach. We used our GOMO strategy 25 to replace genes of interest, through homologous recombination, with a GFP expression cassette under the control of a constitutive HSP70 promoter, to facilitate monitoring of host cell invasion ( Figure S1A). Targeting vectors were assembled by inserting 5' and 3' homology fragments of P. berghei or P. yoelii p12p (PBANKA_0111100; PY17X_0112700), p38 (PBANKA_1107600; PY17X_1108700), and b9 (PBANKA_0808100; PY17X_0811300) genes in the GOMO-GFP plasmid 25 and used to transfect wild-type (WT) P. berghei (ANKA) or P. yoelii (17XNL) blood-stage parasites. We then applied the GOMO selection strategy, consisting of positive selection with pyrimethamine, negative selection with 5-fluorocytosine, and flow cytometry-assisted parasite sorting, as previously described. 25 Pure populations of GFP-expressing drug-selectable marker-free PbDp12p, PbDp38, PbDb9, PyDp12p, PyDp38, and PyDb9 parasite lines were obtained, confirming that none of the targeted genes are essential during blood-stage replication of the parasite. Genotyping by PCR confirmed gene deletion and excision of the drug-selectable marker cassette, as desired, in all parasite lines (Figures S1B-S1H). All the mutants could be transmitted to mosquitoes and produced normal numbers of salivary gland sporozoites, similar to Dp36 parasites ( Figures 1A and 1B). We then assessed the infectivity of the P. berghei and P. yoelii mutant lines in C57BL/6 and BALB/c mice, respectively. C57BL/6 mice injected with 10,000 PbDp12p or PbDp38 sporozoites all developed a patent blood-stage infection, like the parental PbGFP parasites ( Figure 1C). Similarly, BALB/c mice injected with 10,000 PyDp12p or PyDp38 sporozoites all developed a patent blood-stage infection ( Figure 1D). In sharp contrast, none of the animals injected with P. berghei or P. yoelii Db9 sporozoites developed parasitemia, phenocopying the Dp36 mutants ( Figures 1C and 1D). Abrogation of Db9 sporozoite infectivity was also observed in vitro in hepatocyte cell lines. Fluorescence-activated cell sorting (FACS) analysis 24 hours postinfection revealed a dramatic reduction in the number of PbDb9 exoerythrocytic forms (EEFs) in comparison to control PbGFP or PbDp12p and PbDp38 sporozoites in HepG2 cells, which was similar to the reduction observed with PbDp36 mutants ( Figure 1E). Using antibodies specific for UIS4, a marker of the PV membrane (PVM) that specifically labels productive vacuoles, 3,36 we confirmed that, in contrast to Dp12p and Dp38 mutants, Db9 parasites were not able to form productive vacuoles ( Figures 1F and S2). Together, these results show that b9 is essential for sporozoite infection of the liver both in vivo and in vitro, corroborating the results of a previous study, 15 and that p12p and p38 genes on the contrary are dispensable for parasite invasion and liver-stage development.

B9 is required for sporozoite invasion
After infection of HepG2 cell cultures with Db9 sporozoites, only very low numbers of intracellular parasites were observed, all of which were seemingly intranuclear and lacked a UIS4-labeled PVM, similar to the Dp36 mutants ( Figures 1F and S2). Our results contrast with previous reports where mutant EEFs devoid of a PVM were observed in the cytoplasm of infected cells. 15,37 This discrepancy is likely due to differences in the hepatoma cell lines that were used (Huh-7 versus HepG2). Intranuclear EEFs in HepG2 cells are known to result from cell traversal events. 38 Accordingly, a cell wound-repair assay confirmed that the cell traversal activity of Db9 sporozoites is not different to PbGFP parasites, in both HepG2 and HepG2/CD81 cells (Figures 2A and 2B). In contrast, direct quantification of invaded cells by FACS at 3 h postinfection revealed that host cell invasion by Db9 sporozoites is greatly impaired in both cell types ( Figures 2C and 2D). Further examination of the invasion kinetics revealed low invasion rates with both PbGFP and Db9 sporozoites in the early time points (15-60 min) ( Figure 2E), when sporozoites are in the traversal mode. 3,6 At the 2 hour time point, the percentage of PbGFP-invaded cells was markedly increased ( Figure 2E), reflecting commitment to productive invasion and accumulation of sporozoites inside PV. 3,6 A similar increase was not observed with Db9 parasites (Figure 2E), suggesting a defect in productive invasion, similar to P52/P36deficient sporozoites. 6 Productive host cell invasion is associated with discharge of the sporozoite rhoptries, resulting in depletion of the rhoptry proteins RON2 and RON4. 3,39 To visualize the rhoptries in B9-deficient parasites, we genetically modified the ron4 locus in the PbDb9 mutant line to replace the endogenous RON4 by a RON4-mCherry fusion by double homologous recombination ( Figures S3A  and S3D). In parallel, we also genetically modified parental PbGFP and mutant PbDp36 parasites, using the same RON4-targeting vector ( Figures S3B and S3C). Examination of PbGFP/RON4-mCherry, PbDb9/RON4-mCherry, and PbDp36/RON4-mCherry by fluorescence microscopy confirmed expression of the rhoptry marker in merozoites and sporozoites, as expected 39 ( Figure 2F). We then performed invasion assays in HepG2 cells and analyzed the presence of the RON4-mCherry rhoptry marker by fluorescence microscopy. As expected, the RON4-mCherry signal was lost in a vast majority of intracellular iScience Article PbGFP/RON4-mCherry sporozoites as a result of rhoptry discharge during productive invasion ( Figure 2G). In sharp contrast, RON4-mCherry was detected in all examined PbDb9 and PbDp36 intracellular sporozoites, indicating that sporozoites lacking B9 or P36 invade cells without secreting their rhoptries, i.e. through  iScience Article traversal mode only. Altogether, these data demonstrate that genetic deletion of B9 abrogates productive host cell invasion by sporozoites, phenocopying the lack of P36. Our data also show that B9, like P36, is essential for both CD81-dependent and CD81-independent sporozoite entry.  The phenotype of Db9 mutants, combined with proteomic data, implies that the protein B9 is expressed in P. berghei sporozoites and plays a crucial role during host cell productive invasion, unlike previously thought. 15 In order to confirm the expression of B9 at the protein level and define its localization, we genetically modified the endogenous b9 locus in P. berghei (PbGFP) to insert a triple Flag epitope in the proteincoding sequence, through double homologous recombination ( Figure S4A). Because B9 is predicted to be glycosylphosphatidylinositol (GPI) anchored, we inserted the 3xFlag tag towards the C-terminus of the protein, downstream of the putative 6-cys domains but upstream of the predicted omega site (aspartate residue at position 826). Correct integration of the construct was confirmed by PCR on genomic DNA from B9-Flag blood-stage parasites ( Figure S4B). Importantly, we observed no defect in sporozoite development ( Figure S4C) and infectivity ( Figure S4D) in the B9-Flag line, demonstrating that the insertion of a 3xFlag epitope in B9 sequence had no detrimental effect on the protein function.
Immunofluorescence with anti-Flag antibodies revealed that B9 is readily detected in B9-Flag salivary gland sporozoites, with a distribution pattern typical of a micronemal protein ( Figure 3A). As a control, parental PbGFP sporozoites showed no signal with the Flag antibody, confirming the specificity of the labeling ( Figure 3A). Super-resolution microscopy using stimulated emission depletion (STED) showed that B9 distributes in numerous vesicles localized on each side of the nucleus, consistent with B9 being a micronemal protein ( Figure 3B). Interestingly, B9 colocalized in part with P36 (Figures 3C and S5A) but not with the thrombospondin-related anonymous protein (TRAP) ( Figure S5B) or the apical membrane antigen 1 (AMA1) ( Figure S5C), suggesting that B9 is present in a specific subset of micronemes in salivary gland sporozoites. We next analyzed the fate of B9 upon activation of sporozoite microneme secretion, by western blot. In nonactivated control parasites, B9 was detected as a single band between 75 and 100kDa, in both reducing and nonreducing conditions, consistent with the expected size of the protein ($95 kDa) ( Figure 3D). Upon stimulation of microneme secretion, B9 was also recovered in the supernatant fraction as a slightly smaller band, indicating that B9 is secreted from sporozoites upon activation, possibly after enzymatic processing ( Figure 3D). We failed to detect B9 on the surface of B9-Flag sporozoites by immunofluorescence, irrespective of parasite activation, suggesting that following microneme secretion, B9 is mainly released as a shed protein.

B9 contains a CyRPA-like beta-propeller domain
To get more insights into B9 properties, we investigated sequence and structural features of the protein using P. falciparum B9 as the reference sequence. Both hydrophobic cluster analysis and secondary structure prediction of B9 suggested that the whole sequence contains some strand and helix structures ( Figures S6A and S6B). However, no annotated conserved domain was detected at the sequence level using InterPro. In sharp contrast, three domains were predicted at the structural level using HHpred: a N-terminus propeller domain similar to that of CyRPA (e-value: 5.4e-03) encoded by the first exon and two putative but poorly supported 6-cys domains encoded by the second exon (e-value > 1) ( Figure 4A). CyRPA is a cysteine-rich protein expressed in P. falciparum merozoites, where it forms a protein complex that is essential for invasion of erythrocytes. 40,41 B9 is enriched in cysteines, nine being located in the predicted propeller domain that we suppose are involved in the formation of disulphide bonds in a similar manner to CyRPA, 42 to stabilize the protein structure ( Figure 4A).
To explore the structural features of the B9 propeller, we predicted the tertiary structure of PfB9 propeller (covering positions 26 to 386) by homology modeling using CyRPA as a template structure 42 (PDB ID: 5TIH; Figure S7). As expected, PfB9 adopted a six-bladed propeller structure, with each blade being composed of four-stranded antiparallel beta-sheets ( Figure 4B). Four disulphide bonds were predicted within the blades which may stabilize each individual blade of the PfB9 propeller (C47-C64, C171-C181, C239-C248, and C328-C352; Figure 4B). Furthermore, a long loop connecting blades 5 and 6 and containing three putative short helices was observed in the PfB9 propeller, which was not found in CyRPA and in most Plasmodium B9 proteins (such as PbB9 and PyB9; Figure S8). This partially structured region is supported by intrinsic disorder prediction ( Figure S6C), in line with another characteristic of CyRPA, where the loop located on blade 5 likely becomes disordered to accommodate occupancy by a helix of Rh5. 41 The model superimposed well with the CyRPA structure, except for some blade-and strand-connecting loops (root-mean-square deviation [RMSD]: 3.8 Å ; Figure 4C). This similar fold, in addition to the binding activities of CyRPA (targeting Rh5 and Ripr 41  iScience Article Because CyRPA is functionally annotated and its binding properties are known, we checked whether the B9 propeller and CyRPA shared a common evolutionary history, which could help to predict the functional sites in the B9 propeller. For this, we generated two datasets consisting of distinct Plasmodium B9 (n = 23) or CyRPA (n = 18) sequences (Table S1). Multiple sequence alignments and corresponding phylogenetic trees of these datasets ( Figure S9) were then used concomitantly with their respective tertiary structures to estimate spatially correlated site-specific substitution rates using the GP4Rate tool (Table S2). The six blades were found to be heterogeneously conserved over time for both B9 and CyRPA (Kruskal-Wallis H test: B9: p = 0.01; CyRPA: p = 2.4e-8; Figure 4D). Interestingly, we noticed distinct patterns of evolution  The propeller domain of B9 is required for sporozoite infectivity We next sought to define the functional importance of the predicted propeller domain, using a structureguided genetic complementation strategy to evaluate the functionality of truncated B9 proteins (Figure 5A). We assembled various constructs encoding the entire or partially deleted B9, all containing an intact signal peptide and C-terminus sequences to ensure correct secretion and GPI anchoring of the protein ( Figure 5B). Constructs were used for transfection of the drug-selectable marker-free PbDb9 mutant line. After confirmation of correct integration by genotyping PCR (Figure S10), genetically complemented parasites were transmitted to mosquitoes, and sporozoites were tested for infectivity in cell cultures. Complementation of PbDb9 sporozoites with a construct encoding the entire PbB9 fully restored sporozoite infectivity in HepG2 cell cultures ( Figure 5C), validating the genetic complementation approach. In contrast, parasites complemented with a truncated B9 lacking the propeller domain, alone or in combination with the first 6-cys domain, were not infectious, phenocopying the parental B9-deficient parasites ( Figure 5C). These results show that the propeller domain is required for sporozoite infectivity.
Interestingly, chimeric B9 versions where the propeller domain of PbB9 was replaced by the equivalent sequence from PyB9 (Pyprop, Pyprop6cys1; Figure 5B) restored sporozoite infectivity ( Figure 5C). In contrast, substitution of the PfB9 propeller domain for the PbB9 propeller (Pfprop; Figure 5B) did not restore infectivity in complemented parasites ( Figure 5C). Complementation with the PyB9 propeller domain restored infection in both HepG2 cells, which express SR-B1 but not CD81, and HepG2/CD81 cells, which express both receptors, 6 suggesting that the B9 propeller domain does not restrict host cell receptor usage ( Figure 5D). To exclude a defect in protein expression with the nonfunctional constructs, we generated two additional parasite lines expressing Flag-tagged version of the Dprop and Pfprop B9 proteins ( Figures 5B and S10). The Dprop-Flag and Pfprop-Flag proteins were detected in transgenic sporozoites by immunofluorescence ( Figure 5E) and western blot ( Figure S11), indicating that truncation of the propeller domain does not totally impair protein expression. However, both proteins were expressed at low levels as compared to PbB9-Flag, as evidenced by western blot ( Figure S11). Both Dprop-Flag and Pfprop-Flag constructs failed to restore infectivity in PbDb9 sporozoites ( Figure 5F), as observed with untagged proteins. This confirms that the beta-propeller domain is essential for sporozoite infectivity, by enabling adequate protein expression and/or regulating B9 function.
The propeller domain of B9 interacts with P36 and P52 in a heterologous system Our structural modeling revealed that B9 contains an N-terminus beta-propeller domain structurally similar to CyRPA. In P. falciparum merozoites, CyRPA interacts with Rh5 and Ripr to form a complex that is essential for invasion of erythrocytes. 40,41,43 While Ripr is conserved among Plasmodium species, CyRPA is found in primate but not rodent parasites, and Rh5 is restricted to P. falciparum and other Laverania species. 44 As Rh5 and Ripr are not expressed by sporozoites, 21,23,24 we hypothesized that B9 might be involved in the formation of distinct protein complexes in sporozoites. To test this hypothesis, we first performed coimmunoprecipitation (coIP) experiments with anti-Flag antibodies, using protein extracts from B9-Flag sporozoites, followed by protein identification by mass spectrometry. PbGFP sporozoites were used as a control. B9 was the only protein consistently identified in five independent biological replicates by mass spectrometry (Table S3). We considered that B9 might interact with other sporozoite proteins only at the time of host cell invasion, similarly to CyRPA, which interacts with Rh5 following secretion of merozoite apical organelles. 40 To test this hypothesis, we performed coIP experiments on supernatants of HepG2 cell cultures incubated with B9-Flag sporozoites, using uninfected cultures as a control. Again, B9 was the only protein specifically identified in supernatants from infected cultures (Table S3). While these results confirm that B9 is secreted during infection, it is likely that protein amounts released in the culture supernatants are not sufficient to identify interacting proteins by mass spectrometry after immunoprecipitation.
Therefore, we opted for an alternative strategy based on heterologous expression of sporozoite proteins in mammalian cells, to test for potential interactions between B9 and the 6-cys proteins P36 and P52 as candidate partners, a choice motivated by the shared phenotype of gene-deletion mutants. For this purpose, we used a surface display approach to express P. berghei proteins on the surface of Hepa1-6 cells after transient transfection. 45   iScience Article domain of glycophorin A, followed by mCherry, C-Myc, and 6xHis tags ( Figure 6A). As a control, we used an mCherry construct containing all elements except the B9 or P36 sequences. Codon-optimized versions of the tandem 6-cys domains of P. berghei P36 and P52 (amino acids 33-302) were expressed either as transmembrane proteins with 3xFlag and GFP tags or as soluble secreted proteins (sol), with a 3xFlag epitope tag only at the C-terminus ( Figure 6A). Following transient transfection of Hepa1-6 cells, all protein constructs distributed mainly intracellularly (likely in the endoplasmic reticulum [ER]), but a fraction was correctly targeted to the cell plasma membrane, as evidenced by immunolabeling of non-permeabilized cells with anti-V5 and anti-Flag antibodies ( Figure 6B), suggesting correct folding. Interestingly, the soluble forms of PbP36 (P36-Sol) and PbP52 (P52-Sol) were also detected on the surface of transfected cells, indicating that both are secreted and could interact with host cell membrane factors ( Figure 6C). Interaction between proteins was then tested in co-transfection experiments in Hepa1-6 cells, by immunoprecipitation followed by western blot. Both P52-GFP ( Figure 6D) and P52-sol ( Figure 6E) proteins were co-immunoprecipitated with P36-mCherry but not with the control mCherry protein, validating the strategy and confirming the interaction between P. berghei P36 and P52 proteins. More importantly, these experiments showed that P36 and P52 co-immunoprecipitated with B9-mCherry, in both transmembrane ( Figure 6D) and soluble ( Figure 6E) configurations. Although the interactions were only observed using a heterologous expression system, these results suggest that B9, P36, and P52 may form a supramolecular protein complex. When considering our functional data, such a complex could mediate productive invasion of hepatocytes by sporozoites.

DISCUSSION
Productive invasion of hepatocytes is a crucial step following transmission of the malaria parasite by a mosquito; however, the molecular mechanisms involved remain poorly understood. Until now, only two sporozoite-specific proteins, the 6-cys proteins P36 and P52, have been associated with productive host cell invasion. 6,8 Here we identify another member of the 6-cys family, B9, as a crucial entry factor. Our data confirm that B9 is required for sporozoite infectivity, as reported previously. 15 However, in that study, the authors concluded that B9 is not expressed in sporozoites and is not involved during parasite entry into hepatocytes. This conclusion was based on an indirect promoter assay in P. berghei and immunofluorescence assays in P. falciparum using antibodies generated against a 152 aa recombinant protein (233Asn-384Glu), representing a truncated propeller domain that may not reproduce the native protein conformation of the entire PfB9 propeller (26Leu-384Glu). Here, we demonstrate through genetic tagging that B9 is expressed in P. berghei sporozoites, corroborating mass spectrometry data. [21][22][23][24] Furthermore, direct quantification of invasion by flow cytometry established that PbDb9 parasites have an invasion defect. In addition, PbDb9 sporozoites do not discharge their rhoptries upon contact with host cells, similar to PbDp36 sporozoites, indicating that both proteins are acting at an early step during invasion. We further provide evidence that B9 interacts with P36 and P52 using a heterologous expression system, suggesting that the three proteins could participate in an invasion complex required for productive invasion of hepatocytes.
Our data show that two other sporozoite 6-cys proteins, P12p and P38, are dispensable for infection of the liver, in both P. berghei and P. yoelii. Interestingly, there was a slight delay in the onset of blood-stage patency in mice following inoculation of PbDp12p or PyDp12p mutant sporozoites, associated with reduced numbers of PbDp12p EEFs in HepG2 cell cultures. This suggests that P12p, while nonessential, could nevertheless contribute to optimal sporozoite infection in the liver, a possibility that deserves further investigation.
Comparison of profile hidden Markov models between PfB9 and tertiary structure database identified an N-terminus beta-propeller domain structurally similar to CyRPA, a cysteine-rich protein expressed in P. falciparum merozoites, where it forms a protein complex that is essential for invasion of erythrocytes. 40,41 Our data suggest that the propeller domain of B9 could directly interact with both P36 and P52. We speculate that blades 3 and 4 of the propeller, which are the most conserved, might be involved in these interactions. Importantly, the interaction of B9 with P36 and P52 was detected using a heterologous expression system but not by coIP from sporozoite protein extracts. Our data are consistent with a previous study performed with P. yoelii sporozoites, where P52 but not B9 was identified by mass spectrometry after immunoprecipitation of P36. 11 We speculate that B9 could interact with P36 and P52 only after parasite activation, similar to CyRPA, which forms a complex with Rh5 and Ripr only at the time of merozoite invasion in P. falciparum. 40  iScience Article failed, possibly due to a lack of sensitivity. Alternatively, the presence of 6-cys domains in the native B9 protein may impact the binding properties of the propeller domain. B9 was secreted from sporozoites upon stimulation of microneme exocytosis, as described previously with P36 in P. yoelii. 11 B9 shedding could be associated with enzymatic processing, as suggested by the differential migration pattern in western blots. This suggests two possible models, where B9 may bind to P36/P52 either as a membrane-bound iScience Article or as a free form (Figure 7). Using STED super-resolution microscopy, we could visualize individual micronemes in sporozoites. Interestingly, the distribution of B9 partially overlapped that of P36, suggesting that a subset of micronemes may contain both proteins. While a previous immuno-electron microscopy study showed that a fraction of P36 and P52 colocalizes with TRAP in micronemes of P. yoelii sporozoites, 11 we did not observe colocalization of B9 with TRAP or AMA1 in P. berghei sporozoites. Our data support the hypothesis that sporozoites contain discrete subsets of micronemes, associated with specific functions. 4 In line with this hypothesis, Db9 sporozoites show a similar invasion phenotype to Dp36 parasites, without alteration of sporozoite migration (which requires TRAP). Using RON4-mCherry as a rhoptry marker in the invasion assays, we observed that Db9 and Dp36 sporozoites do not discharge their rhoptries, suggesting that B9 and P36 act upstream of AMA1-dependent MJ formation. AMA1 is also required upstream of liver infection, during sporozoite invasion of the mosquito salivary glands. 46 This temporal functional difference between AMA1 and B9 is consistent with the observation that the two proteins are contained in distinct secretory compartments.
P. berghei and P. yoelii sporozoites use different pathways to invade hepatocytes, with the latter being strictly dependent on CD81, like P. falciparum. 5,7 Interspecies complementation experiments have shown that P36 (but not P52) is a key determinant of this differential usage of host receptors. 6 Using a similar approach, we show that the propeller domain of PyB9 can functionally replace the homologous sequence in PbB9, however, without altering host receptor usage. This suggests that the B9 propeller does not directly participate in interaction with host receptors. Rather, we hypothesize that B9 may regulate the trafficking and/or binding of P36 to host cells, possibly by concentrating P36-P52 complexes at the surface of the parasite. In contrast, substituting the PfB9 propeller for the P. berghei domain abolished protein function, possibly due to impaired protein expression, as suggested by our western blot data, or as a result of altered interactions with P. berghei P36 and/or P52. In this regard, the PfB9 and PbB9 propeller domains show only 48% identity at the amino acid level, versus 90% between PyB9 and PbB9 domains ( Figure S8). Our data are consistent with a recent study showing that chimeric P. berghei sporozoites where the entire PbB9 has been replaced by PfB9 are not infective. 47 Reciprocally, the essential role of B9 in assembling invasion complexes with P36 and P52 could also explain why P. falciparum and P. vivax P36 and P52 failed to compensate for the absence of their counterparts in P. berghei 6 as these proteins may not associate with PbB9 to form functional complexes.
Interestingly, an improved version of the neural network-based model AlphaFold 48 predicts that the C-terminus portion of B9 is organized in three beta sandwiches rather than two (https://alphafold.ebi.ac. uk/). The structures of these domains and their function remain to be experimentally determined. While iScience Article our data suggest that B9 6-cys-like domains are not required for interaction with P36 and P52, they might regulate the activity of the propeller and/or participate in interactions with host cell surface molecules.
In conclusion, this study reveals that the 6-Cys protein B9 is required for productive host cell invasion by sporozoites. B9 contains a functionally important beta-propeller domain that is required for proper protein expression and could be involved in the formation of a supramolecular protein complex with P36 and P52.
Our results suggest that Plasmodium sporozoites and merozoites, despite using distinct sets of parasite and host entry factors, may share common structural modules to assemble protein complexes for invasion of host cells. The complex formed by B9, P36, and P52 proteins may represent a potential target for intervention strategies to prevent the initial stages of malaria liver infection.

Limitations of the study
One of the limitations of this study is use of the rodent malaria model parasite P. berghei to dissect the function of B9 beta-propeller domain through genetic approaches. The propeller domain of P. falciparum B9 did not allow proper protein expression and/or function in P. berghei; therefore, other approaches will be required to determine the function of this domain in human-infecting malaria parasites.
The interactions between the beta-propeller domain of B9 and P36 or P52 were only observed in a heterologous expression system but not with endogenous sporozoite proteins. Such interactions may occur in the parasite in a transient manner, possibly during host cell invasion, which is a rare event and difficult to address experimentally. An additional caveat with our heterologous system is that the bulk of parasite proteins is trapped inside transfected Hepa-16 cells, so we cannot rule out interactions occurring between misfolded proteins in the ER. Finally, the interactions between sporozoite 6-cys proteins and host receptors are not addressed in the study.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:   Experimental animals, parasites, and cell lines P. berghei and P. yoelii blood stage parasites were propagated in female Swiss mice (6-8 weeks old, from Janvier Labs). We used wild type P. berghei (ANKA strain, clone 15cy1) and P. yoelii (17XNL strain, clone 1.1), and GFP-expressing PyGFP and PbGFP parasite lines, obtained after integration of a GFP expression cassette at the dispensable p230p locus. 25 Anopheles stephensi mosquitoes were fed on P. berghei or P. yoelii-infected mice using standard methods, 72 and kept at 21 C and 24 C, respectively. P. berghei and P. yoelii sporozoites were collected from the salivary glands of infected mosquitoes 21-28 or 14-18 days post-feeding, respectively. P. berghei and P. yoelii sporozoite infections were performed in female C57BL/6 or BALB/c mice, respectively (6 weeks old, from Janvier Labs), by intravenous injection in a tail vein. HepG2 (ATCC HB-8065), HepG2/CD81 38 and Hepa1-6 cells (ATCC CRL-1830) were cultured at 37 C under 5% CO 2 in DMEM supplemented with 10% fetal calf serum and antibiotics (Life Technologies), as described. 7 HepG2 and HepG2/CD81 were cultured in culture dishes coated with rat tail collagen I (Becton-Dickinson).

METHOD DETAILS
Gene deletion of p12p, p38 and b9 in P. berghei and P. yoelii Gene deletion mutant parasites were generated using our ''Gene Out Marker Out'' (GOMO) strategy. 25 For each target gene, a 5' fragment and a 3' fragment were amplified by PCR from P. berghei (ANKA) or P. yoelii (17XNL) WT genomic DNA, using primers listed in Table S4, and inserted into SacII/NotI and XhoI/KpnI restriction sites, respectively, of the GOMO-GFP vector, 25 using the In-Fusion HD Cloning Kit (Clontech). The resulting targeting constructs were linearized with SacII and KpnI before transfection. All constructs used in this study were verified by DNA sequencing (Eurofins Genomics). Purified schizonts of P. berghei ANKA or P. yoelii 17XNL WT parasites were transfected with targeting constructs by electroporation using the AMAXA Nucleofector TM device, as described, 73 and immediately injected intravenously in mice. GFPexpressing parasite mutants were then isolated by flow cytometry after positive and negative selection rounds, as described. 25 Parasite genomic DNA was extracted using the DNeasy Blood & Tissue Kit (Qiagen), and analyzed by PCR using primer combinations specific for WT, 5' or 3' recombined and marker excised loci (listed in Table S4).

Genetic tagging of RON4, P36 and B9
Fusion of mCherry at the C-terminus of RON4 was achieved through double crosser homologous recombination. For this purpose, 5' and 3' homology fragments, consisting of a 1.2 kb terminal RON4 fragment (immediately upstream of the stop codon) and a 0.6 kb downstream fragment were amplified by PCR using primers listed in Table S4, and cloned into NotI/SpeI and HindIII/KpnI sites, respectively, of the B3D+mCherry plasmid. 51 The resulting construct was linearized with NotI and KpnI before transfection of PbGFP, PbDb9 or PbDp36 purified schizonts. Recombinant parasites were selected with pyrimethamine and cloned by limiting dilution and injection into mice. Integration of the construct was confirmed by PCR on genomic DNA using specific primer combinations listed in Table S4. P36 fused with mCherry at the C-terminus was expressed from a centromeric episomal plasmid. For this purpose, we first introduced ama1 promoter and 3' UTR fragments in the centromeric pCEN-SPECT2 plasmid, 52

Fluorescence microscopy
To visualize RON4-mCherry in transgenic parasites, purified schizonts and sporozoites were deposited on poly-L-lysine coated coverslips and fixed with 4% FA. GFP and mCherry images were captured on a Zeiss Axio Observer.Z1 fluorescence microscope equipped with a Plan-Apochromat 633/ 1.40 Oil DIC M27 objective. Images acquired using the Zen 2012 software (Zeiss) were processed with ImageJ 53 or Photoshop CS6 software (Adobe) for adjustment of contrast. To quantify rhoptry discharge, RON4-mCherry expressing PbGFP, PbDb9 or PbDp36 sporozoites were incubated with HepG2 cells for 3 h at 37 C. After extensive washes to remove extracellular parasites, cultures were trypsinized and cells were examined under a fluorescence microscope to assess for mCherry fluorescence in GFP-expressing intracellular sporozoites. At least 50 intracellular parasites in triplicate wells were examined for each parasite line. The percentage of rhoptry discharge was defined as the proportion of intracellular sporozoites without detectable RON4-mCherry signal. For immunofluorescence analysis of Flag-tagged parasites, sporozoites collected from infected mosquito salivary glands were deposited on poly-L-lysine coated coverslips, fixed with 4% FA and permeabilized with 1% Triton X-100. Parasites were labelled with anti-Flag mouse antibodies (M2 clone, Sigma) and AlexaFluor 594-conjugated secondary antibodies (Life Technologies). Nuclei were stained with Hoechst 33342. For double labelling of B9 and AMA1, we used anti-Flag mouse antibodies (M2 clone, Sigma) and anti-AMA1 rat antibodies 75 (clone 28G2, Bei Resources), followed by at-to647N-conjugated anti-mouse and Alexa-594-conjugated anti-rat antibodies. For double labelling of B9 and P36-mCherry or TRAP, we used anti-Flag mouse antibodies (M2 clone, Sigma), anti-mCherry rat antibodies (Invitrogen) or anti-TRAP rabbit antibodies, 50 followed by STAR RED-conjugated anti-mouse and Alexa-594-conjugated anti-rat antibodies or atto647N-conjugated anti-mouse and Alexa-594-conjugated anti-rabbit antibodies, respectively. Coverslips were mounted on glass slides with ProLongä Diamond Antifade Mountant (Life Technologies). STED imaging was carried out with a 93x glycerol-immersion objective (NA 1.3) on a Leica TCS SP8 STEDX microscope equipped with a White Light Laser. AlexaFluor 594 and atto647N-or STAR RED-labelled compartments were excited at 590 or 644 nm, respectively, and depleted with a pulsed 775 nm STED laser. Image frames were acquired sequentially frame by frame at a scan speed of 200 lines/s with an optimal pixel size and a line average of 4 to 8. Deconvolution of STED data was performed using the default deconvolution settings in Huygens Professional Deconvolution software v18. 10 (Scientific Volume Imaging) that were estimated from the metadata. Brightness and Contrast were adjusted using Fiji. 76 Western blot B9-Flag sporozoites were isolated from the salivary glands of infected mosquitoes and resuspended in 1X PBS. Microneme secretion was stimulated by incubation for 15 min at 37 C in a buffer containing 1% BSA and 1% ethanol, as described. 77 Pellet and supernatant fractions were then isolated from activated and non-activated (control) sporozoites, resuspended in Laemmli buffer and analyzed by SDS-PAGE under non-reducing conditions. For the Dprop-Flag and Pfprop-Flag parasites, only pellet fractions were analyzed. Western blotting was performed using primary antibodies against the Flag epitope (M2 clone, Sigma) or against GFP (loading control), and secondary antibodies coupled with Alexa Fluor 680. Membranes were then analyzed using the InfraRed Odyssey system (Licor).

Heterologous expression of Plasmodium proteins in Hepa1-6 cells
Two vectors for mammalian cell expression were first assembled in a pEF1a-AcGFP1-N1 backbone. The first one (mCherry) encodes a cassette consisting of the signal peptide from bee venom melittin (BVM), a V5 epitope, the transmembrane and C-terminus of mouse Glycophorin A (GYPA), mCherry, Myc and 6xHis tags. In the second one (GFP), the cassette encodes the signal peptide from BVM, a 3xFlag epitope, the transmembrane and C-terminus of mouse GYPA, and GFP. Codon-optimized versions of PbB9 propeller domain (amino acids 31-348), PbP36 (amino acids 67-352) or PbP52 (amino acids 33-302) were inserted in the mCherry and/or GFP plasmids between the signal peptide and the Flag or V5 epitope tag. Two additional constructs for expression of soluble PbP36 and PbP52 were obtained by adding a stop codon immediately after the 3xFlag epitope. The construct cassette sequences are indicated in Table S5. High concentration plasmid solutions were produced using XL1-Blue Competent Cells (Agilent) and plasmid extraction was performed using Qiagen Plasmid Maxikit (Qiagen) according to the manufacturer's recommendations. Plasmid transfection was performed in Hepa1-6 cells using the Lipofectamine 2000 reagent ll OPEN ACCESS iScience Article QUANTIFICATION AND STATISTICAL ANALYSIS Statistical significance of infection data was assessed by one-way ANOVA followed by Dunnett's multiple comparisons test, two-way ANOVA followed by Sidak's multiple comparisons test, or two-tailed ratio paired t test, as indicated in the figure legends. Survival curves were analyzed using the Log rank Mantel-Cox test. All statistical tests were computed with GraphPad Prism 7 (GraphPad Software). In vitro experiments were performed with a minimum of three technical replicates per experiment. Quantitative source data are provided in Table S6. Statistical analyses for structural modelling were performed using the computing environment R version 3. 5