Pfcerli2, a duplicated gene in the malaria parasite Plasmodium falciparum essential for invasion of erythrocytes as revealed by phylogenetic and cell biological analysis

Merozoite invasion of host red blood cells (RBCs) is essential for survival of the human malaria parasite Plasmodium falciparum. Proteins involved with RBC binding and invasion are secreted from dual-club shaped organelles at the apical tip of the merozoite called the rhoptries. Here we characterise P. falciparum Cytosolically Exposed Rhoptry Leaflet Interacting protein 2 (PfCERLI2), as a rhoptry bulb protein that is essential for merozoite invasion. Phylogenetic analyses show that cerli2 arose through an ancestral gene duplication of cerli1, a related cytosolically exposed rhoptry bulb protein. We show that PfCERLI2 is essential for blood-stage growth and localises to the cytosolic face of the rhoptry bulb. Inducible knockdown of PfCERLI2 led to an inhibition of merozoite invasion after tight junction formation. PfCERLI2 knockdown was associated with inhibition of rhoptry antigen processing and a significant elongation of the rhoptries, suggesting that the inability of merozoites to invade is caused by aberrant rhoptry function due to PfCERLI2 deficiency. These findings identify PfCERLI2 as a protein that has key roles in rhoptry biology during merozoite invasion.


ABSTRACT 23
Merozoite invasion of host red blood cells (RBCs) is essential for survival of the human 24 malaria parasite Plasmodium falciparum. Proteins involved with RBC binding and invasion 25 are secreted from dual-club shaped organelles at the apical tip of the merozoite called the 26 rhoptries. Here we characterise P. falciparum Cytosolically Exposed Rhoptry Leaflet 27 Interacting protein 2 (PfCERLI2), as a rhoptry bulb protein that is essential for merozoite 28 invasion. Phylogenetic analyses show that cerli2 arose through an ancestral gene duplication 29 of cerli1, a related cytosolically exposed rhoptry bulb protein. We show that PfCERLI2 is 30 essential for blood-stage growth and localises to the cytosolic face of the rhoptry bulb. 31 Inducible knockdown of PfCERLI2 led to an inhibition of merozoite invasion after tight 32 junction formation. PfCERLI2 knockdown was associated with inhibition of rhoptry antigen 33 processing and a significant elongation of the rhoptries, suggesting that the inability of 34 merozoites to invade is caused by aberrant rhoptry function due to PfCERLI2 deficiency. These 35 findings identify PfCERLI2 as a protein that has key roles in rhoptry biology during merozoite 36

PfCERLI2 knockdown inhibits merozoite invasion 241
As PfCERLI2 was most highly expressed at schizont-stages, and its knockdown 242 inhibited growth, we hypothesised that PfCERLI2 could be involved in schizont development, 243 rupture, or merozoite invasion. To quantify invasion, we transfected PfCERLI2 HAGlmS parasites 244 with a plasmid that expresses cytosolic GFP (PfCERLI2 HAGlmS/GFP ), which allows reliable 245 detection of newly invaded ring stages by flow cytometry. Treating PfCERLI2 HAGlmS/GFP 246 parasites with increasing concentrations of GLCN from trophozoite-stages for 24 hours until 247 early ring-stages the following cycle, we saw that PfCERLI2 knockdown inhibited invasion in 248 a dose-dependent manner (Figure 3a). The extent of invasion inhibition caused by PfCERLI2 249 knockdown was lower than growth inhibition over 72 hours, however, it is likely that this   To determine if this invasion inhibition was specifically due to a defect in merozoite The flow cytometry-based assay does not allow discrimination between free merozoites 290 inhibited prior to or after tight-junction formation. Therefore, we analysed Giemsa-stained thin 291 smears of GLCN treated and untreated PfCERLI2 HAGlmS ring-stage parasites and quantified the 292 number of bound merozoites following schizont rupture (Figure 3 f&g). PfCERLI2 knockdown 293 led to a 49% (± 12% SEM) increase in the percentage of merozoites bound to RBCs, suggesting 294 that knockdown inhibits invasion following tight-junction formation. Newly invaded ring-295 stage parasites ( Figure 3h) were also quantified, with GLCN treatment leading to a 41% (± 296 5.9% SEM) reduction in newly invaded ring stages. These data indicate that the GLCN 297 inducible knockdown of PfCERLI2 inhibits merozoite invasion both at and prior to formation 298 of the tight-junction. 299

PfCERLI2 localises to the rhoptry bulb 300
The homology to PfCERLI1 as well as its functional implications in RBC invasion 301 suggested PfCERLI2 may have a rhoptry localisation. After failing to get consistent PfCERLI2 302 localisation using conventional antibody staining, likely due to the low peak expression level 303 of Pfcerli2 17 , we investigated the cellular distribution of PfCERLI2 using a high affinity anti-304 HA biotin system 18 . Despite its peak transcription late in schizont development, we observed 305 that PfCERLI2 foci were brightest in early schizonts (~44 hrs post invasion) (Figure 4a showed that PfCERLI2 forms the donut shape characteristic of membrane associated rhoptry 315 bulb proteins (Figure 4c). 316 Repeat immunofluorescence experiments consistently showed that mature schizonts 317 had little PfCERLI2 staining. Parallel western blots, however, showed no significant decrease 318 in PfCERLI2 detection between early and late schizonts (Supplementary Figure 11). Given the 319 lack of change in detectability by Western blot, it is possible that the HA epitope is masked 320 and therefore undetectable using anti-HA antibodies in microscopy with mature schizont stage 321

parasites. 322
PfCERLI2 is peripherally associated with the cytosolic face of the rhoptry bulb 323

membrane. 324
The absence of a signal peptide suggested that PfCERLI2 is a peripheral membrane 325 protein that localises to the rhoptries but is exposed to the merozoite cytosol. To test this, we 326 subjected PfCERLI2 HAGlmS schizont lysates to a proteinase protection assay. Following lysis 327 with saponin and digitonin, which lyse the RBC membrane, PVM, and PPM, but not organellar 328 membranes, PfCERLI2 was sensitive to proteinase K degradation ( Figure 5a); suggesting it is 329 exposed to the cytosol. To assess membrane association of PfCERLI2, we performed a 330 solubility assay using lysates from PfCERLI2 HAGlmS schizonts (Figure 5b   antibodies. Images representative of 3 independent experiments. 369 PfCERLI2 knockdown inhibits rhoptry bulb antigen processing but not secretion of 370 rhoptry neck proteins 371 PfCERLI1 has previously been shown to be involved with rhoptry secretion 9,10 , and 372 given the similarities between PfCERLI1 and PfCERLI2 we sought to determine whether 373 PfCERLI2 was also involved in rhoptry secretion. To test the effect of PfCERLI2 knockdown 374 on rhoptry secretion, ring-stage PfCERLI2 HAGlmS parasites were treated with 2.5 mM GLCN 375 or left untreated. The RBCs of each culture had been treated with trypsin, chymotrypsin, and 376 neuraminidase to prevent reinvasion. Parasites were cultured until schizont rupture when the 377 supernatant was collected (containing secreted proteins) and a parasite lysate (containing 378 merozoites that had failed to invade) were prepared for Western blot ( Figure 6a). As expected, 379 there was no change in secretion of the micronemal antigen erythrocyte binding antigen 175 380 (EBA-175) with knock-down of the rhoptry associated PfCERLI2 (Figure 6b). There was also 381 no noticeable defect in the secretion of the rhoptry neck antigens reticulocyte binding 382 homologue 4 (RH4) and rhoptry neck protein 4 (RON4). These data indicate that PfCERLI2 383 knock-down does not block secretion of rhoptry neck antigens in merozoites that fail to invade 384 enzyme treated RBCs. 385 Parasite lysates from the secretion assay, containing free merozoites, were also probed 386 with antibodies to the rhoptry bulb antigen rhoptry associated protein 1 (RAP1) (Figure 6a). 387 RAP1 has previously been shown to be processed by Plasmepsin IX and subtilisin-like protease   likely Toxoplasma gondii homologue of PfCERLI2 (TGME49_315160, also known as TgRASP2), the PH domain (degenerated in PfCERLI2) is involved in, but not essential for, 499 rhoptry localisation and secretion 10 . Based on evidence from its T. gondii homologue and the 500 lack of a functional PH domain and palmitoylation site, we suggest that membrane attachment 501 of PfCERLI2 is mediated by its C2 domain, a structure involved in targeting many proteins to 502 various membranes 25 . 503 Knockdown of PfCERLI2 led to reduced parasite growth and merozoite invasion. 504 Morphology and rate of schizont rupture in PfCERLI2 knockdown parasites were 505 indistinguishable compared to the controls, indicating merozoite development was not 506 impacted by knockdown of PfCERLI2. Instead, we observed increased numbers of both free 507 merozoites in the culture medium and merozoites bound to the RBC surface in PfCERLI2 508 knockdown parasites. Based on this evidence, the growth inhibition seen with PfCERLI2 509 knockdown is most likely to be a result of merozoite invasion being interrupted at or before 510 formation of the irreversible tight-junction with the RBC surface. The increase in RBC bound 511 merozoites following PfCERLI2 knockdown suggests that rhoptry neck and micronemal 512 protein contents were secreted for PfCERLI2 deficient merozoites, and the TJ formed for some 513 of these, but these merozoites were then unable to proceed with invasion. The fact that we saw 514 no reduction in secretion of the rhoptry neck proteins RON4 and RH4 following CERLI2 515 knockdown supports the idea that in many cases the glucosamine inducible ribozyme 516 PfCERLI2 knockdown parasites can secrete enough protein to form the TJ, but these 517 merozoites fail to invade due to a later-occurring defect. 518 The presence of a prominent decapeptide repeat region at the C-terminus of PfCERLI2 519 is a differentiating feature compared to PfCERLI1. Based on its differential evolution relative 520 to the rest of the PfCERLI2 protein, and its exclusive presence in Laverania, we hypothesise 521 that the decapeptide tandem repeat number may be under positive selection, whereas the rest 522 of the protein is under negative selection that strongly favours amino acid conservation across Plasmodium spp. A previous study identified a growth inhibitory monoclonal anitibody 524 (known as M26-32 15 ) that bound to PfCERLI2, as well as a number of other proteins 26 , and 525 was reported to localise this protein to the merozoite surface. This study also reported that the 526 decapeptide tandem repeat contained highly immunogenic antigens that bind malaria-exposed 527 serum antibodies. A potential mechanism for the positive selection of the PfCERLI2 repeat 528 could therefore be classical antibody selection. However, our data using endogenously tagged 529 PfCERLI2 shows that the protein localises to the cytosolic face of the rhoptry bulb membrane 530 and has a role in the function of this organelle, making it unlikely that PfCERLI2 is 531 significantly exposed to the host immune system and antibody selection. We therefore suggest 532 that the PfCERLI2 repeat has an as yet unidentified function in P. falciparum and its evolution 533 is under positive selection based on its Ka/Ks value of 1.43 across Laverania. Amongst P. 534 falciparum isolates, repeat number varies and it is currently unclear whether the same occurs 535 among different isolates of all Laverania, or whether this is a unique property of PfCERLI2. 536 Given the conservation of the repeat sequence in P. falciparum, but the variation in repeat 537 number, this variation in repeat number may have an important role in this protein's function 538 between isolates. 539 Due to the presence of two paralogous genes in all coccidians (e.g. Toxoplasma gondii 540 and Besnoitia spp.) and haematozoa (e.g Plasmodium spp.), it is likely that an ancestral cerli1 541 sequence was duplicated in the most recent common ancestor of these related parasite lineages. 542 By contrast, all more divergent Apicomplexans and Chromerids contain only a single cerli 543 gene, suggesting that there was originally only a single cerli gene in the more distant ancestors 544 of coccidia and haematozoa. Cryptosporidium is more divergent than coccidia and haematozoa 545 and contains only a single CERLI homologue. As the gene structure of the cerli homologue of 546 Cryptosporidium is highly similar to that of cerli1 across other Apicomplexa, we suggest that 547 cerli1 is most similar to the ancestral gene that underwent duplication and gave rise to the more divergent cerli2 gene. From studies of both PfCERLI1 and PfCERLI2, it is clear that both 549 proteins have functional similarities due to their stage of expression and localisation, along 550 with their knockout and knockdown phenotypes 9,10,13 . Additionally, in Toxoplasma gondii it 551 has been reported that TgRASP1 and TgRASP2 form a complex 10 , but it not clear if this occurs 552 in other Apicomplexa. For the coccidia and haematozoa containing cerli paralogues, it is likely 553 that the duplicated cerli2 locus that arose from gene duplication of an ancestral cerli1 locus 554 underwent subfunctionalisation. This is supported by shared characteristics of PfCERLI1 and 555 PfCERLI2, along with the strong negative selection CERLI2 has been under across 556 Apicomplexa. A potential driver for this subfunctionalisation of the paralogous genes may be 557 a direct binding of CERLI1 to CERLI2, which has been observed for RASP1 and RASP2 in T. This study highlights that CERLI2, like its paralogue CERLI1, has an important role in 571 Plasmodium invasion and rhoptry biology that is likely shared broadly across Apicomplexa. In 572 P. falciparum, loss of PfCERLI2 led to elongated rhoptry structures that were evident in both 2D and 3D cell analysis and was associated with reduced merozoite invasion. The localisation 574 of PfCERLI2 puts it in the right place at the right time to undertake overlapping functions with 575 its paralogue PfCERLI1 on the cytosolic face of the rhoptry bulb, and this possibility may be 576 resolved in future studies. The modified structure of CERLI2 in Plasmodium spp. suggests that 577 the requirements of malaria parasite host cell invasion has led to further evolution of the 578 protein, highlighting the ongoing specialisation of these proteins as the apicomplexan parasites 579 Both pairwise and multiple sequence alignments were generated using Geneious 9.1.3 604 (Biomatters) using a Geneious global alignment with free end gaps (gap open penalty = 12, 605 gap extension penalty = 3) algorithm with the Blosum62 cost matrix. Multiple sequence 606 alignments were also constructed using Muscle 29 in an R environment 30 . Phylogenetic trees 607 were constructed from nucleotide and amino acid sequences using the Jukes-Cantor 31 DNA or 608 LG 32 amino acid substitution models, respectively, genetic distance model and were built using 609 the unweighted pair group method with arithmetic mean (UPGMA) tree build method using 610 Geneious or the R package phangorn v. 2.5.5 33 . A tanglegram visualisation was generated using 611 the cophyloplot function of the R package ape 34 . 612 To identify P. falciparum proteins that contained a PHIS, or PHIS-like motif, the PHIS 613 sequence from PfCERLI1, PfCERLI2, or the consensus sequence generated from their 614 Apicomplexan homologues was used in a BLASTp search against the P. falciparum 3D7 615 proteome. 616 Ka/Ks ratios were calculated using the online Ka/Ks Calculation tool 617 (http://services.cbu.uib.no/tools/kaks) 35,36 using the maximum likelihood tree method, 618 discrete_Grantham submatrix, and a LI rate of moderate. Where Ka/Ks ratios are reported, they 619 were generated using the same coding sequences as the corresponding phylogenetic tree. 620

Protein structure prediction 621
The protein structure of PfCERLI1, PfCERLI2, along with their homologues in T. 622 gondii (TGME49_235130, TGME49_31516) and C. muris (CMU_007720) was predicted 623 using the online protein structure prediction tool Phyre2 on intensive mode 37 . All predicted 624 structures were visualised, presented and imaged using either Jmol 38 or EzMol 39 . To generate the Pfcerli2 HAGlmS/GFP parasite line, which expresses cytosolic GFP, the 644 pHGBrHrBl-1/2 GFP plasmid was used, without modification 42 . 645 For disruption of Pfcerli2 using selection linked integrationtargeted gene disruption (SLI-646 TGD) system, a source SLI-TGD vector (Pf3D7_1463000 SLI-TGD) was used 16 . A 772 bp 647 homology region of Pfcerli2 was PCR amplified using the primers Pfcerli2 SLI-TGD F