A screen for Plasmodium falciparum sporozoite surface protein binding to human hepatocyte surface receptors identifies novel host-pathogen interactions

Sporozoite invasion of hepatocytes is a necessary step prior to development of malaria, with similarities, at the cellular level, to merozoite invasion of erythrocytes. In the case of the malaria blood-stage, efforts to identify host-pathogen protein-protein interactions have yielded important insights including vaccine candidates. In the case of sporozoite-hepatocyte invasion, the host-pathogen protein-protein interactions involved are poorly understood. Here, we performed a systematic screen to identify such interactions. We substantially extended previous Plasmodium falciparum and human surface protein ectodomain libraries, creating new libraries containing 88 P. falciparum sporozoite protein coding sequences and 182 sequences encoding human hepatocyte surface proteins. Having expressed recombinant proteins from these sequences, we used a plate-based assay capable of detecting low affinity interactions between recombinant proteins, modified for enhanced throughput, to screen the proteins for interactions. We were able to test 7540 sporozoite-hepatocyte protein pairs under conditions likely to be sensitive for interaction. We report and characterise an interaction between human fibroblast growth factor receptor 4 (FGFR4) and the P. falciparum protein Pf34, and describe an additional interaction between human low-density lipoprotein receptor (LDLR) and the P. falciparum protein PIESP15. Strategies to inhibit these interactions may have value in malaria prevention, and the modified interaction screening assay and protein expression libraries we report may be of wider value to the community.


Introduction
, and no clear role in hepatocyte invasion has been shown for the interaction of P. 81 falciparum thrombospondin-related anonymous protein (TRAP) with integrin αvβ3 [5]. 82 An interaction between host Ephrin type-A receptor 2 (EphA2) and parasite P36 has been 83 suggested but not biochemically demonstrated, and EphA2 has subsequently been shown 84 not to be required for sporozoite invasion [13,14]. 85 We hypothesized that, like merozoite-erythrocyte invasion, sporozoite-hepatocyte 86 invasion involves multiple protein-protein interactions, identification of which would 87 enable improved vaccine strategies. Biologically important extracellular protein-protein 88 interactions are often of low affinity and can be very transient (for example, the PfRH5-89 basigin interaction has 0.8 μM affinity and a half-time of 2.7 s). Some of us previously 90 reported a plate-based assay which can be used to identify such interactions, the AVidity-91 based EXtracellular Interaction Screen (AVEXIS) [15]. We therefore set out to perform an 92 AVEXIS screen to identify sporozoite-hepatocyte protein-protein interactions, modifying 93 the assay method to enhance throughput to a level adequate to investigate a large 94 proportion of the surface protein repertoire of P. falciparum sporozoites and human 95 hepatocytes. 96

98
AVEXIS assay development for systematic high-throughput screen 99 The AVEXIS assay has previously been used to identify protein-protein interactions, 100 including those with low-affinity and fast dissociation rates [15]. Subsequently we have 101 modified the method to enhance sensitivity by using luciferase rather than β-lactamase 102 as the reporter assay (Galaway et al, manuscript in preparation). Here, given the limited 103 prior information about candidate sporozoite ligands and hepatocyte receptors and the 104 large number of expressed candidate proteins, we wished to perform a broad screen and 105 therefore further modified the method to enhance throughput while preserving or 106 improving sensitivity ( Figure 1A-B). The assay modifications allowed expression of most 107 proteins in 24-well plates, removed dependency on streptavidin-biotin-mediated capture 108 and so eliminated the need for bait dialysis and expensive streptavidin-coated plates, and 109 retained the sensitive and immediate luciferase-generated luminescence readout. 110 The modified assay remained capable of the detection of four known protein-protein

133
Of these constructs, we were able to clone and attempt expression of 88 sporozoite and 134 182 hepatocyte proteins as AVEXIS baits and preys respectively (i.e. constructs as shown 135 in Figure 1B). Synthesis of the remaining coding sequences was unsuccessful. Full details 136 of constructs produced are provided in Supplementary Tables 1 and 2. 142 nM which we regarded as optimal. To achieve these levels, spin filter concentration was 143 required for 54 baits.

144
Of the preys, 165 were obtained at concentrations >1x10 7 LU/mL which we regarded as 145 potentially informative, of which 139 preys (including CD81, SR-BI, EphA2, and integrin 146 αvβ3) were obtained at >4x10 8 LU/mL which we regarded as optimal. Results of Western 147 blotting of preys are shown in Figure 2D   To provide additional assurance regarding the quality of key bait and prey proteins, and 149 in particular the activity of the folded proteins in a plate-format assay, we tested whether 150 the full-length CSP bait, and CD81, SR-BI and EphA2 preys could be captured onto 96 well 151 plate using cognate monoclonal antibodies. Captured baits and preys were detected using 152 ELISA and luciferase assay respectively, demonstrating the expected antibody reactivity    Table 3). Of these, 7540 candidate interactions were tested using protein 164 concentrations we would regard as optimal (bait concentration ≥7 nM, prey 165 concentration 4x10 8 LU/mL, and good protein quality as assessed by Western blotting), 166 and a further 4718 were tested using protein concentrations which our assay validation 167 data ( Fig 1C) suggested would provide a signal:noise ratio of >10 for any of our 'test set' 168 of four known interactions.  were obtained in a similar screen investigating P. berghei [26].

197
To confirm the Pf34-FGFR4 interaction, the AVEXIS assay was repeated using 198 reciprocally-oriented constructs, with FGFR4 expressed as dimeric bait and probed with 199 pentameric Pf34 prey. Again, clear and reproducible binding was observed ( Fig 2B).  Having identified the interaction between human FGFR4 and P. falciparum Pf34, we 211 examined whether this interaction is conserved across species by testing murine FGFR4 212 for interaction with the Pf34 orthologs found in the rodent malaria parasites P. yoelii and 213 P. berghei. We found no evidence of interaction of either of these protein pairs, despite 214 expression of all proteins at levels in the range expected to give optimal AVEXIS 215 sensitivity ( Figure 3E).

216
Pairwise host-pathogen protein-protein interactions are frequently components of larger 217 multi-molecular complexes, and so we proceeded to investigate possible additional 218 interacting partners of Pf34.

219
Binding of many FGF family members to their receptors is enhanced by heparin and/or 220 heparan sulfate [27]. Using SPR, we tested whether heparin and heparan sulphate (HS) 221 may have a similar affect upon the Pf34 -FGFR4 interaction. We used a similar design to 222 that used in our experiment measuring Pf34 -FGFR4 kinetics, assessing whether pre-223 incubation of soluble monomeric FGFR4 with heparin or HS had any effect upon binding 224 to Pf34 immobilised on the chip. Despite using high heparin/HS concentrations 225 (1mg/mL), we did not observe any enhancement of FGFR4 binding (data not shown).

227
Invasion of hepatocytes by P. falciparum. sporozoites is a bottleneck in the malaria 228 parasite lifecycle. Inhibition of this process, particularly by vaccine-induced antibodies, 229 is a major focus in efforts to develop means of malaria prevention. This effort is hindered 230 by limited knowledge of the host-parasite interactions involved in hepatocyte invasion.

231
This study has therefore sought to improve understanding in this area.

232
Our approach, expressing human and P. falciparum proteins in a human cell line and 233 testing them for interaction using a modified AVEXIS assay, was designed to achieve the 234 best sensitivity which we could achieve in a broad, high-throughput screen. Our selection     interpolating from a standard curve of samples of known protein concentration (Fig 2A).  Proteins on the probed membrane were detected using SuperSignal West Pico 423 Chemiluminescent substrate (ThermoFisher) (Fig 2C-D  was removed and normalised baits were immobilised onto the plates overnight at 4ᴼC.

443
The next day, plates were washed as above, and normalised preys were added and 444 incubated for 2h at room temperature. Plates were washed and bound preys were 445 detected by using the Nano-Glo Luciferase Assay System (as described for prey 446 quantification).

447
Given that background levels of signal were observed to vary between preys, the screen        Table S1. Sporozoite protein ectodomain library details 593 Details include bait index number (corresponding to numbering in Figure 3A), construct 594 boundaries and sequence, and expression levels (corresponding to Figure 2A).   Table S3.

602
First worksheet presents complete AVEXIS screen results, presented in terms of double-603 corrected signal:noise ratio (see Methods). Color scale denotes the signal:noise ratio of 604 each interaction, ranging from dark green (low) through yellow to dark red (high).

605
Second and third worksheets show results with FGFR4 prey against all sporozoite baits 606 (as shown in Figure 3A) and with Pf34 prey against all sporozoite baits (as shown in 607 Figure 4B).  Table S4. 610 All protein pairs with a signal:noise ratio exceeding 5 in the initial AVEXIS screen.