Virus-like particles displaying conserved toxin epitopes stimulate broadly reactive, polyspecific, murine antibody responses capable of snake venom recognition


 Antivenom is currently the first-choice treatment for snakebite envenoming. However, only a low proportion of antivenom immunoglobulins are specific to venom toxins, resulting in poor dose efficacy and potency. We sought to investigate whether linear venom epitopes displayed on virus like particles can stimulate a robust and focused antibody response capable of recognising venom toxins from diverse medically important species. Bioinformatically-designed epitopes, corresponding to predicted conserved regions of group I phospholipase A2 and three finger toxins, were engineered for display on the surface of hepatitis B core antigen virus like particles and used to immunise female CD1 mice over a 14-weeks. Antibody responses to all venom epitope virus like particles were detectable by ELISA by the end of the immunisation period, although total antibody and epitope specific antibody titres were variable against the different epitope immunogens. Immunoblots using pooled sera demonstrated recognition of various venom components in a diverse panel of six elapid venoms, representing three continents and four genera. Finally, pooled terminal sera was compared to conventional antivenom via quantitative immunoblot, and demonstrated superior recognition of lower-molecular weight elapid venom toxins. This study demonstrates proof-of-principle that virus like particles engineered to display conserved toxin linear epitopes can elicit specific antibody responses in mice which are able to recognise a geographically broad range of elapid venoms.

). These homology groups broadly corresponded to functional sub-family 121 annotation, although notable exceptions were observed. For example, type I alpha-neurotoxins (sNTX) splitting into 122 two distinct groups: GR17 & GR15, while GR7 consisted of type IA cytotoxins (CTX) and Orphan group XV (OGXV) (Supp. 123 Table S3). 124 Of the 21 3FTX groups, six (aminergic-type [ATX] GR1, CTX and OGXV GR7, non-conventional [NCX] GR10, orphan group 125 VIII [OG8] GR13, sNTX GR15 and sNTX GR17) were selected based on their represented frequency in the data set (Supp. 126 Table S2). Hereafter, these groups will be referred to by their toxin-type only (e.g. ATX, CTX, OGXV, NCX, OG8, or SNTX) 127 (Table 1). Combined, these six groups represented 70.6% (187/264) of the 3FTX sequences analysed in the study. 128 Sequence conservation within 3FTX groups (except for NCX) was generally high, with 49-66% of AA residues being 129 ≥80% conserved across all group sequences (Supp. Fig. S2 and Supp. Fig. S3). Group I PLA2 sequences were very 130 homogenous (68% and 48% AA residues at 80% and 100% conservation) across full-length sequences (Supp. Fig. S4). 131 Fifteen individual 3FTX epitopes were designed based on i) BepiPred predicted epitope regions, ii) conservation within 132 groups, iii) predicted accessibility and iv) their molecular location: either the hydrophobic core or 1 st or 3 rd finger of 133 3FTX, designated by "_F" or "_C", respectively (Supp. Figs. S2, S3, S5, Supp. Table S3, Supp. File S5). It was not always 134 possible to design epitopes corresponding to regions which reflected predicted epitopes due to limited sequence 135 conservation in predicted regions. In such cases epitopes were solely designed on sequence conservation and 136 accessibility (e.g., ATX_F, NCX_F, OG8_C [Supp. Fig. S2, S3, S5]). Epitope ATX_F was near identical (1 AA greater in 137 length) to epitope Pep604-B designed to elicit antibodies against Micrurus corallinus 3FTX 35 . 138 Three variants of a single group I PLA2 epitope were designed (Table 1)  suggests this region is readily recognised by current antivenoms. The selected PLA2 epitope region resides between 141 the calcium binding loop (Tyr 28, Gly 30, Gly 32) and Asp 49 (Supp. Fig. S4, S6) residues essential for calcium ion 142 positioning for hydrolytic activity 48 . 143 Expression of VLPs presenting snake venom 3FTX and PLA2 epitopes 144 A sub-selection of individual designed epitopes were chosen for expression on VLPs, herein referred to as venom 145 epitope VLPs (veVLPs), and were expressed and purified as described in the Supplemental Materials and Methods. 146 Despite the resulting recombinant veVLPs expressed in Escherichia coli proving to be largely insoluble, sufficient 147 quantities of soluble material were obtained for each veVLP. Attempts to improve solubility by varying incubation 148 temperature and inducing IPTG concentration resulted in little-to-no improvement. The methods detailed for the 149 expression of veVLPs resulted in a mean yield of 9.97 mg/L soluble veVLP from a 0.6 L culture (range 1.7 -16.4 mg/L). 150 To ensure confidence in the assembly of veVLPs (as opposed to monomers resulting from non-assembly), all affinity-151 purified veVLPs were concentrated with a 100 kDa MWCO centrifugal filter to deplete any sub-100 kDa proteins whilst 152 retaining assembled veVLPs 49 . All purified veVLPs were visually assessed for purity using anti-His fluorescent immuno 153 blots, comparing total protein stain (Supp. Fig. S7A) to anti-His signal (Supp. Fig. S7B). As shown in Supp. Fig. S7, the 154 major bands in purified samples contained His-tagged proteins corresponding to the individual expected molecular 155 weight of the reduced monomeric constituent proteins of each veVLP monomer. Larger bands (presumably) corresponding to dimeric complexes of individual monomers were also visible for the majority of veVLPs. The veVLPs 157 were subsequently probed with SAIMR Polyvalent antivenom to determine if antibodies raised against crude venom 158 could recognise venom epitopes displayed on VLPs. Results demonstrated antivenom recognition of four veVLPs -159 sNTX_F1 and sNTX_F2, 3FTX Core-string and 3FTX Finger-string. No recognition of the other veVLPs by SAIMR 160 Polyvalent was apparent (Supp. Fig. S7D). 161 VLPs presenting snake venom epitopes induce antibody responses in mice 162 Female CD1 mice were immunised with different veVLPs over a 14-week immunisation schedule (see Materials and 163 Methods). As described previously, we were forced to humanely euthanise 20 individuals (Table 2). This severely 164 restricted monitoring of responses in several immunogen groups. 165 The antibody response of mice to veVLP immunisation was monitored at specific points (at weeks 3, 6, 10 and at the 166 end of experiment at week 14) via ELISA using pooled sera consisting of equal volumes from each individual in each 167 experimental group (Fig. 1, Supp. File S4). Antibody responses to veVLPs were detected at week 3 for all groups (OD405 168 ≥ 1 at 1/500 dilution of sera), with the exception of mice receiving Core-string (Group K) and CTX_C (Group M -without 169 adjuvant) veVLP immunogens, whose signal was indistinguishable to that of naïve serum or negative controls. The 170 OD405 for 7 of the 13 veVLP groups continued to increase until reaching a peak at week 10, which then subsequently 171 declined modestly by week 14. Due to a processing error, we unfortunately lost week 10 sera corresponding to animals 172 immunised with CTX_C (Group Awith adjuvant), CTX_C (Group M without adjuvant) and CTX_F (Group B), thus it is 173 not possible to infer if a similar response profile occurred with CTX_C or CTX_F veVLPs. 174 CTX_C , CTX_F and sNTX veVLP immunised groups (A/M, B and E) provided the lowest overall titres at terminal bleed 175 (week 14) (Fig. 1). Both CTX_C immunised groups (groups A and M with and without adjuvant, respectively) resulted 176 in near identical mean titres (1/500 OD450 1.16 and 1.15, respectively), however, CTX_C with adjuvant (Group A) titres 177 remained stable throughout the schedule, whereas CTX_C without adjuvant (Group M) titres slowly reached this titre 178 by week 14. CTX_F (Group B) titres declined after week 6, possibly reflective of the early euthanasia of 80% of the 179 individuals in this group on humane grounds (Table 2), thus this data only reflects n=1 from week 6 onwards. sNTX_C 180 (Group E) titres remained stable but relatively low throughout the immunisation period (maximum mean 1/500 OD405 181 = 0.97). 182 VLPs presenting snake venom epitopes elicit antibodies against the displayed epitope and against the VLP carrier 183 To ascertain the proportion of the antibody response directed towards the displayed epitope as opposed to the VLP 184 carrier, pooled sera from each group of veVLP immunised mice was also used to probe nativeVLPs (Fig. 1, Panel O). 185 Results broadly demonstrated three distinct profiles; i) veVLP generated sera recognised nativeVLPs at slightly lower 186 or equivalent titres compared to recognition of respective veVLPs, throughout the immunisation period (nine groups), 187 ii) veVLP sera initially recognised nativeVLP before rapid declining of nativeVLP-specific titres to baseline (3 groups), 188 or iii) veVLP generated sera displayed negligible recognition of nativeVLP (1 group). These results demonstrate that 189 the majority of veVLPs used in these immunisations are capable of eliciting polyspecific antibodies towards the carrier 190 molecule and the venom epitope.
Using ELISA data generated from pooled sera probed against veVLPs and nativeVLP (Fig. 1), we compared the 192 proportion of apparent epitope-specific response of individual group sera by subtracting the 1/500 OD405 response 193 against nativeVLP from that of the response against veVLP (Supp. Fig. S8). Apparent epitope-specific antibody 194 responses were identified across all experimental groups by week 14. The apparent epitope specific antibody 195 response to CTX_C veVLPs (Groups A and M, with and without adjuvant, respectively) and sNTX_C (group E) was the 196 greatest of all immunogens examinedwith 60-80% of the overall antibody response (Supp. Fig. S8). However, this 197 could ultimately reflect the poor/modest seroconversion of these groups to these antigens. Sera from the remaining 198 veVLP immunised groups typically demonstrated ~20% of their anti-veVLP antibody response could be considered 199 specific towards the heterologous displayed epitopes (Supp. Fig. S8). This proportion of epitope specific response is overall recognition of 3FTX from the three Naja spp. and O. scutellatus venoms had further increased, whilst a slight 216 decrease in overall fluorescence intensity was observed for B. candidus and D. polylepis (Fig. 2). To test the specificity 217 of the veVLP sera towards the toxins against which they were designed, sera from the best responding animal per 218 immunisation group (as described below) were used in immune-blot to probe a panel of purified 3FTXs, consisting of 219 muscarinic toxin 3 from D. angusticeps (an ATX representative), cytotoxic 3FTX from N. nigricollis (a CTX 220 representative) and short chain 3FTX from N. haje (a sNTX representative), and a basic group I PLA2 from N. nigricollis. 221 Across all immune-blots against purified toxins (Supp. File S7), all sera showed some degree of binding towards the 222 purified PLA2, therefore data are shown as fold-difference over naïve signal for normalisation purposes. As shown in 223 Supp. File S7, only one of the three sera raised against veVLP-CTX_C (Group A animal 5) recognised the purified 224 cytotoxic 3FTX; however, these sera did not demonstrate strong recognition of other purified toxins, thereby 225 suggesting toxin specificity. Sera raised against veVLP-ATX_C and veVLP-ATX_F (Groups C and D) and veVLPs displaying 226 3FTX-strings (Core-string and Finger string, Groups K and L, respectively) showed strong and specific recognition of the 227 purified muscarinic toxin (ATX representative), whilst sera raised against PLA2 (Groups H, I and J) displayed limited non-specific recognition of other toxins. Sera raised against sNTX veVLPs (Groups E, F and G) did not demonstrate 229 recognition towards any of the toxins tested. 230 Dotblots using crude venoms showed specific recognition of venom by the veVLP sera, with no recognition by naïve 231 sera, thus demonstrating that the generated antibodies were able to recognise venom components in their native 232 conformation, as well as in reduced and denatured conditions (i.e. in immune-blot). Our results (  Fig. S8). The inability of CTX_F (group B) sera to recognise 274 venom components was unsurprising due to near-baseline levels of veVLP recognition in pooled ELISA results (Fig. 1). 275 However, the results from these groups need to be interpreted in the context of several influencing factors. by group C sera was more variable (mean 1/500 OD405 = 1.69, range 1.29-1.92) ( identity, 80% coverage, mismatches ≤ 1) with members of this toxin subclass (Table S3). 296 In veVLP immunisation groups where multiple individuals produced antibodies that could recognise venom proteins 297 (groups D, H, K and M, immunised with ATX_F, PLA2_1, Core-string and CTX_F, respectively), the number of venoms 298 recognised was either consistent across individuals (same number of venoms, similar intensity of recognition, e.g. 299 Group D) or variable (recognising different numbers of venoms with differing intensity of recognition, e.g. Group H) 300 ( This study demonstrates proof of principle that VLPs decorated with rationally selected conserved linear venom-309 epitopes can be used to stimulate the production of murine antibodies that are able to recognise a geographically and 310 taxonomically diverse range of elapid venoms (Table 2, Fig. 2, File. S6). Additionally, the promising results and 311 resources generated from this study enable the further progression of this research, as splenocytes isolated from the 312 best-responding individual mice identified in this study are being investigated as a resource for therapeutic anti-toxin 313 monoclonal antibody discovery. 314 Of the 10 individual epitopes designed and used in this study, six were shown to elicit antibodies capable of binding 315 specific venom components (Table 2, Supp. File S6). veVLPs decorated with 3FTX epitopes corresponding to finger 316 regions were more likely (38%, n=13) to elicit venom binding sera than core epitope veVLPs (13%, n=8) at the end of 317 the immunisation period. Additionally, sera against PLA2 epitopes PLA2_1 and PLA2_2 (immunisation groups H and I, 318 respectively) demonstrated similar abilities in recognising respective immunogens at the end of the immunisation 319 period (Fig. 1), but provided considerable differences in their ability to recognise venom components, despite differing 320 by only a single amino acid (KGTPVDLDD and KGTAVDDLD, respectively) ( Table 1 Antivenoms have long been reported to have poor dose efficacy in neutralising small molecular weight venom toxins, 328 including 3FTX and group I PLA2, which is frequently attributed to these toxins being poorly immunogenic 21,22 . Notably, 329 in this study, a pool of experimental veVLP sera demonstrated a substantial improvement in recognition of small 330 molecular weight compounds as compared to the comparative SAIMR Polyvalent antivenom at an equivalent 331 concentration (Fig. 3). While the results confidently demonstrate improved recognition of these medically important 332 toxins, we are currently unable to demonstrate if this increased recognition translates into more potent neutralising 333 efficacy. This is due to the quantities of antibodies recoverable for each animal not being sufficient to perform 334 informative in vitro neutralisation assays of toxin activity, with yields in this study ranging from approximately 45 µg 335 to 240 µg per animal. 336 Despite the success in demonstrating the ability of veVLPs to elicit anti-toxin antibodies, this study was subject to 337 several limitations. Notably, a large proportion of animals during immunisation were euthanized early due to 338 presumed adverse reactions to adjuvant (20 out of 65). Notable local inflammation and irritation (redness and local 339 swelling) were typically observed at all dosing locations in animals that received VLPs and adjuvant, which usually 340 resolved 1-2 weeks post immunisation. Animals that received veVLP without adjuvant (Group M) generally did not 341 develop any local reaction at any dose sites, which leads us to hypothesise the adjuvant, Alhydrogel (alum), was 342 contributing to the observed adverse effects. This is surprising as alum based adjuvants are routinely considered as 343 safe and are widely used in human vaccines 56 , and suggests that the combination of the self-adjuvating nature of 344 VLPs 41 and the adjuvant might adversely exacerbate local inflammatory responses. Unfortunately, acute local 345 inflammation resulting in a non-resolving lump at the inoculation site was observed in 20 animals dosed at the rump 346 on week 2 (2 nd immunisation), which necessitated euthanasia of affected animals as per our institutional and national 347 license conditions. We suspect that the tighter skin around the rump, as compared to the scruff and flanks of the mice, 348 in combination with irritation and local swelling may have exacerbated local conditions. Due to the substantial adverse 349 reactions observed in this study, we advise against immunising mice in the rump area in similar immunisation 350 experiments conducted in the future. Furthermore, the groups that ultimately were unable to recognise venom 351 components were also the groups most affected by the losses in numbers at week 2 (Table 2). For example, groups A, 352 B and E representing CTX_C, CTX_F and sNTX_C, were the most severely affected, losing 60-80% of their representative 353 animals. Based on the results obtained from groups less affected by animal loss, we currently cannot say if the inability 354 of these epitopes to elicit an immune response is due to poor candidate epitopes or individual variation in response 355 to immunisation. We speculate that the large amount of variation in venom reactivity observed between individuals 356 within a group is due in part to immunological heterogeneity in the experimental animals, as the mouse strain selected, 357 CD1, is outbred and thus reflective of antivenom manufacturing animals. Similar highly variable results in responses 358 to immunisation have been observed in antivenom manufacturing animals 33 and camels 57 . 359 Our results demonstrate that venom epitopes fused to VLPs can induce robust anti-toxin antibody responses. 360 However, difficulties in production of veVLPS in this specific VLP format (HBcAg), encountered by ourselves and others 361 expressing heterologous antigens 49,58 , may prove challenging for application if this approach were to be applied to 362 producing a rationally designed antivenom at commercial scale. However, research to circumvent production 363 obstacles has been actively undertaken over the past decade. Developments include alternative methods of genetic-364 fusion for decorating VLPs with heterologous antigens, such as SpyCatcher-SpyTag 'plug and play' systems 59 , and the 365 development of computationally designed hyper-stable and soluble synthetic VLPs 49,60 . Use of such particles to 366 generate rationally designed antivenoms, or as a tool to rapidly discover monoclonal antibodies to specific venom 367 targets, may be a more cost-effective approach that would also increase translational viability. 368 The use of linear snake venom epitopes for rationally targeted anti-toxin antibody generation now has substantial 369 background 26  pathologies. Whilst considered to be inferior to conformational epitopes in terms of potency, linear epitopes have the advantage that they are easier to identify and cheaper to produce than conformational antigensvitally important 374 considerations when proposing improvements to a therapeutic which is already prohibitively expensive to the majority 375 of people who need it most 2 . Furthermore, as snake venoms consist of multiple toxin families and sub-families, it is 376 highly likely that such a strategy will require multiple epitopes to ensure adequate protection against all medically 377 important venom components 1 . Recent publication of high-throughput antivenom-venom peptide arrays 47,63 , 378 alongside transcriptomic and proteomic characterisation of venoms 29,64 , means there is now a wealth of resources 379 available for informative venom epitope prediction. 380 However, substantial research questions remain when considering whether the approach of employing rationally 381 designed linear epitope antigens to elicit anti-toxin antibodies is a genuinely practical method for producing more 382 efficacious antivenoms. Questions include: will the results, all generated so far in mice and rabbits, be translatable in 383 manufacturing antivenom-manufacturing animals? Will antisera produced in this manner possess potencies which at 384 least match existing conventionally produced antivenoms? To date, demonstration of neutralising efficacies of anti-385 epitope antibodies have been performed against relatively low challenge doses of venom. Could alternative 386 immunisation strategies, such as combined epitope and crude venom approaches, substantially increase efficacy, 387 especially against lower molecular weight toxins? Notably in this study, the improvement in recognition of small 388 molecular weight compounds as compared to current antivenoms is promising and it may be possible to use a handful 389 of veVLPs decorated epitopes of toxins as an additive to crude venom during manufacture to improve potency of the 390 resulting antivenom. Additionally, through this work we have been able to preserve splenocytes from individual veVLP 391 immunised animals displaying the most promising antibody responses. We hope to investigate this valuable resource 392 with a view to developing mAbs, which have been raised against specific, rationally designed venom epitopes, into 393 potential next generation antivenom therapies 65 . 394 Female CD1 mice (18 -20 g) were purchased from Charles River and allowed to acclimatise for one week before first 407 immunisation. Mice were housed in groups of five with ad libitum access to certified reference materials irradiated 408 food (Special Diet Services) and reverse osmosis water (in automatic water system), along with enrichment, and kept at approximately 22 °C at 40-50 % humidity, with 12/12 hour light cycles. Mice were housed in Techniplast GM500 410 cages with Lignocel bedding (JRS, Germany) and zigzag fibres nesting material (Sizzlenest, RAJA), and cages were 411 changed fortnightly. Mice were kept in specific-pathogen-free facilities. All experiments were performed by mixed 412 gender experimenters. Humane endpoints were weight loss (>10% loss of body weight within one week, or >20% 413 within one month [despite remedial actions such as wet food]), or observation of the following animal behaviour or 414 appearance signsreduced activity, physiological impairments, pallor, or ulceration following immunisation. 415

Materials and Methods
Mice were briefly anaesthetised with 5% isofluorane to enable shaving at the injection site for subsequent monitoring 416 of adverse reactions. Mice were subcutaneously immunised with 1 µg immunogen in a total volume of 40 µL at each 417 immunisation, according to the following schedule; week 0: Injection at one site on the scruff (with adjuvant), week 2: 418 Injection at one site on the rear, week 4: Injection over two sites (20 µL/site) on the right flank (without adjuvant), 419 week 8: Injection over two sites (20 µL/site) on the left flank, week 12: Injection over two sites (20 µL/site) at the 420 scruff. An additional group received GR7_c (Group M) immunisations which were always performed without adjuvant. 421 A total of 12 veVLP immunogens were used for immunisation, with the specific epitope immunogens assigned to each 422 group of five mice listed in Table 1. 423 Following the second immunisation at week 2, 20 animals developed large non-resolving lumps at the rear dose site 424 and were euthanised on humane grounds to prevent pain, harm and distress. Subsequent immunisations were given 425 over two dose sites in a refinement of the immunisation, from which all animals developed mild, small, self-resolving 426 lumps at the injection sites. Animals were monitored twice per week throughout the course of immunisation for 427 adverse reactions and general health, and no animals were culled due to weight loss or behavioural endpoints being 428 met. 429

Sera isolation 430
Approximately 50 µL venous blood samples were collected at week 3, 6 and 10 by tail bleed. Whole blood was allowed 431 to clot for a minimum of 2 hours at room temperature, and sera was obtained by centrifugation at 2000 x g for 10 432 minutes at 10 °C. Sera was immediately stored at -20 °C. Remaining animals were euthanised by rising concentrations 433 of carbon dioxide at week 14 (end of experiment). Following confirmation of death, cardiac punctures were performed 434 to collect whole blood and sera was processed as above. 'Naïve' unimmunised mouse sera controls (strain matched) 435 were sourced commercially from Charles River UK and Sigma. Additionally, splenocytes were collected and preserved 436 for future work.  D. jamesoni, D. polylepis, N. nivea, N. melanoleuca, N. annulifera and N. mossambica, was used as a  447 control comparator to the generated murine samples. Representative toxins for short chain 3FTX, PLA2 and cytotoxic 448 3FTX were purified in-house as described in Supplemental Materials and Methods. A representative aminergic-type 449 toxin (muscarinic toxin 3, from D. angusticeps) was bought from Alomone Labs (Israel). 450

Immunoassays 451
To assess the toxin recognition and specificity of antibodies from the immunised mice we performed immunoassays 452 comprising of immunoblotting (immuno-blots and dotblots) and end-point ELISAs. Tween-20 for 2 hours at room temperature on an orbital shaker. Membranes were washed a further three times in 468 TBS-T and once in TBS, prior to imaging for 2 minutes in the 700 and 800 nm channels on an Odyssey Fc Imaging 469 System. All images were obtained using the Image Studio software (Version 5.2, LI-COR Biosciences) 470 To determine ELISA end-point titres of veVLP immunised mice, antigens, either veVLPs or control VLP displaying no 471 heterologous epitopes (nativeVLP), were coated at 100 ng per well on Nunc MaxiSorp plates (ThermoScientific) in 50 472 mM carbonate-bicarbonate coating buffer (pH 9.6) and allowed to bind overnight at 4 °C. Plates were washed six 473 times with TBS-1% Tween20, and then blocked with 5% rabbit serum in TBS-1% Tween20 for 8 hours at room 474 temperature. Pooled mouse sera was diluted 1 in 100 in blocking solution, added to the plate and five-fold serial 475 diluted to 1 in 500 and 1 in 2500, and incubated overnight at 4 °C. The following day, plates were washed as above 476 and secondary antibody (anti-mouse IgG-HRP, Abcam) at 1 in 2000 in TBS was added for 2 hours at room temperature. 477 Plates were washed as above and developed with 3% ABTS in citrate buffer pH 4.0 with 0.1% hydrogen peroxide. 478 Developing solution was added to each well (100 µL per well) and developed for 10 minutes at room temperature. 479 Reactions were stopped with 100 µL 1% SDS and immediately read at an optical density of 405 nm (OD405) on an Infinite 480 F50 plate reader (Tecan).
All measurements were performed in triplicate, except where indicated (due to limited amounts of sera). Control wells 482 of no protein (with mouse sera and secondary antibody), and no mouse sera (immunogen and secondary antibody) 483 were included. Raw data is available in Supp. File S4. 484

Data availability 485
All data generated or analysed during this study are either included in this published article (and its Supplementary 486 Information files) or, in the case of raw data files for fluorescent immunoblots, available from the corresponding author 487 upon request. 488 29.