Arthritogenic Alphavirus Vaccines: Serogrouping Versus Cross-Protection in Mouse Models

Chikungunya virus (CHIKV), Ross River virus (RRV), o’nyong nyong virus (ONNV), Mayaro virus (MAYV) and Getah virus (GETV) represent arthritogenic alphaviruses belonging to the Semliki Forest virus antigenic complex. Antibodies raised against one of these viruses can cross-react with other serogroup members, suggesting that, for instance, a CHIKV vaccine (deemed commercially viable) might provide cross-protection against antigenically related alphaviruses. Herein we use human alphavirus isolates (including a new human RRV isolate) and wild-type mice to explore whether infection with one virus leads to cross-protection against viremia after challenge with other members of the antigenic complex. Persistently infected Rag1-/- mice were also used to assess the cross-protective capacity of convalescent CHIKV serum. We also assessed the ability of a recombinant poxvirus-based CHIKV vaccine and a commercially available formalin-fixed, whole-virus GETV vaccine to induce cross-protective responses. Although cross-protection and/or cross-reactivity were clearly evident, they were not universal and were often suboptimal. Even for the more closely related viruses (e.g., CHIKV and ONNV, or RRV and GETV), vaccine-mediated neutralization and/or protection against the intended homologous target was significantly more effective than cross-neutralization and/or cross-protection against the heterologous virus. Effective vaccine-mediated cross-protection would thus likely require a higher dose and/or more vaccinations, which is likely to be unattractive to regulators and vaccine manufacturers.

PGM platform, generating >500,000 reads. A consensus sequence was generated by performing an assembly with Geneious R8 software (Geneious, Auckland, NZ https://www.geneious.com) using RRV T48 (GenBank: GQ433359) as a reference. An alignment was performed with genome-length (and near-genome-length) RRV virus GenBank submissions using the MAFFT plugin of Geneious R8 with default parameters. A phylogenetic tree was constructed with the nucleotide sequence alignment in MEGA7 (Molecular Evolutionary Genetics Analysis version 7.0, Penn State University, Pennsylvania, USA) [67] using the Maximum Likelihood method and the General Time Reversible model with invariant sites (GTR + I) and 1000 bootstrap replicates. The tree was rooted using CHIKV as the out-group.

GETV MM2021 Sequencing
C6/36 cells were infected with GETV at MOI 0.01 and incubated for 2 days, and culture supernatant was centrifuged at 3000 rpm for 15 min. PEG6000 (Sigma Aldrich, Darmstadt, Germany) was added to culture supernatant to a final concentration of 10%, incubated on a rotor overnight at 4 • C, and then centrifuged at 12,000 rpm for 1 h at 4 • C. The pellet was resuspended in RAV1 lysis buffer from the Nucleospin RNA virus kit (Machery-Nagel, Duren, Germany), and viral RNA was purified as per manufacturer instructions. RNA was sent to Australian Genome Research Facility (AGRF) for Next Generation Sequencing on the MiSeq platform (Illumina, California, USA), and 1,143,407 paired-end reads (150 bp) were generated. Reads were aligned to the GETV MI-110-C2 strain sequence (LC079087.1) using STAR Aligner and the consensus sequence was obtained using Integrative Genomics Viewer (IGV). A complete genomic sequence of the isolate was obtained (GenBank MN849355). Alignment analyses showed a > 99.69% nucleotide identity with the available partial sequence of MM2021 (GenBank AF339484).
The formalin-inactivated whole virus JEV/GETV vaccine was purchased from Nisseiken Co. Ltd., Tokyo, Japan and was imported from NIID, Japan under Australian DAWR permit no. 0002721876, which restricted use to class 5 approved arrangement (AA) sites at QIMRB (quarantine facilities). The vaccine was developed using the MI-110 isolate (GenBank: LC079087.1) obtained from an infected horse during the 1978 GETV outbreak at the Miho training center, Ibaraki Prefecture, Japan [31].
The vaccines were administered intramuscularly (i.m.), with the dose split equally into both quadriceps muscles of restrained mice in 50 µL per muscle using an insulin syringe.

Antibody ELISA and Neutralization Assays
IgG responses were determined by standard ELISA using whole alphavirus as antigen and the mean plus 3 SD of naïve serum values as the endpoint as described [24]. Alphaviruses were purified from infected C6/36 cell supernatants by 40% polyethylene glycol precipitation (PEG6000) and ultracentrifugation (~134,000 rcf for 2 h at 4 • C) through a 20% sucrose cushion. Neutralization assays were performed essentially as described [24]. Heat-inactivated (56 • C, 30 min) mouse serum was incubated (in duplicate) with 150 CCID 50 of alphavirus for an hour at 37 • C before adding to Vero cells (except for ONNV where BHK cells were used) (10 5 cells/well, 96-well plate). After 5 days, cells were fixed and stained with formaldehyde and crystal violet and the 50% neutralizing titers interpolated from optical density (OD) versus dilution plots.

Statistics
Statistical analyses of experimental data were performed using IBM SPSS Statistics for Windows, Version 19.0 (IBM Corp., Armonk, NY, USA). The t-test was used when the difference in variances was <4, skewness was >2 and kurtosis was <2. Otherwise, the non-parametric Kolmogorov-Smirnov test was used. Correlation analyses used the non-parametric Spearman's rank-order correlation.

Characterisation of RRV TT , a Contemporary Human Isolate from an RRV Disease Patient, in Wild-Type Mice
The RRV T48 strain has been extensively used in mouse models of RRV musculoskeletal disease [75][76][77]. RRV T48 is a mosquito-derived, mouse-adapted isolate, with a largely undocumented passage history [60]. A contemporary clinical RRV isolate was recently obtained from a transfusion case, where the donor had both symptoms of RRV (fatigue and arthralgia) and tested positive for RRV by serology [53]. We sequenced this isolate (RRV TT ; GenBank accession number: KY302801) with phylogenetic analyses illustrating that it clustered with another human isolate (98% nucleotide identity) but was relatively distant from RRV T48 (96% nucleotide identity) ( Figure S3a), with both conservative and non-conservative amino acid differences in structural and non-structural proteins evident between RRV T48 and RRV TT ( Figure S3b,c). Nevertheless, the binding of a small panel of monoclonal antibodies was the same (Figure S3d), and in vitro replication was largely comparable, ( Figure S3e) for these two virus isolates.
To the best of our knowledge, no human RRV isolate has been tested for its ability to infect and cause disease in mice. Adult (6-10 week old) female C57BL/6J mice were thus infected s.c. in the hind feet with RRV T48 or RRV TT at a dose of 10 4 CCID 50 . Viremia was significantly higher for RRV T48 -infected mice compared with RRV TT (Figure 1a, days 1-3), with the age of the mice (6-versus 24-week-old) not having a significant effect on peak viremia (Figure 1a, inset bar graph). Both the viral titers on days 2 and 6 ( Figure 1b) and viral RNA levels on days 6 and 30 post-infection (Figure 1c) in the feet were not significantly different for the two RRV isolates. Foot swelling was, however, significantly higher for RRV TT, although it only reached a maximum mean increase of≈12% (Figure 1d), substantially lower than the >60% increase seen in the adult wild-type mouse model of CHIKV arthritis [42,75], perhaps consistent with the substantially lower overall pathogenicity of RRV when compared with CHIKV in humans [4]. The RRV TT-associated swelling, although evident in 6-week-old mice, was absent in 24-week-old mice (Figure 1d, inset bar graph). Age dependence for alphavirus mouse models is well described [78], with the traditional RRV T48 mouse model using 3-week-old weanling mice so that an overt musculoskeletal phenotype can be observed [75].
H&E staining of arthritic feet from 6-weeks-old RRV-infected mice day 6 post-infection illustrated the characteristic mononuclear cellular infiltrates [4,79] that were clearly evident in muscle tissues ( Figure 1e, black ovals), with some also evident around joint tissues (Figure 1f, black ovals). Quantitation using blue (nuclear)/red (cytoplasmic) staining ratios showed that cellular infiltrates were not significantly different for RRV T48 and RRV TT (Figure 1g). (Leukocytes tend to have higher nuclear to cytoplasm ratios than resident cells, so blue/red ratios represents a measure of leucocyte infiltration [56]). Subcutaneous edema was, however, more evident in RRV TT -infected mice (Figure 1h, edema indicated by asterisks), perhaps explaining the foot swelling differences seen in Figure 1d. CD4+ CD3+ T cells and monocytes/macrophages are intimately associated with alphaviral arthritides [5,80]. Immunohistochemistry (IHC) for anti-CD3 on day 6 post-infection illustrated that infiltrating CD3+ T cells were often found in muscle tissues (Figure 1i, black ovals). RRV TT arthritis was associated with more infiltrating CD3+ T cells when compared with RRV T48 arthritis ( Figure 1j). F4/80+ monocytes/macrophages were often found in subcutaneous connective tissues ( Figure 1k) and were not significantly different for the two RRV strains ( Figure 1l). Thus, both RRV TT and RRV T48 infections of adult wild-type mice show the inflammatory infiltrates characteristic of alphaviral arthritides [4].
RRV TT therefore represents a contemporary (non-mouse-adapted) human isolate of RRV that is suitable for comparison, in adult wild-type mouse models, with other (non-mouse-adapted) human arthritogenic alphavirus isolates.

Protection and Cross-Protection Mediated by Arthritogenic Alphaviruses Infection
Antibody responses directed at E1 and/or E2 proteins are believed to be the primary mediators of protection against arthritogenic alphaviruses [72,81], although other factors may play a minor role [56,82]. The phylogenetic relationships between the E1/E2 amino acid sequences for alphaviruses used herein are shown in Figure 2a.
Using the new human RRV TT isolate and human isolates of CHIKV, ONNV and MAYV (Figure 2a), adult C57BL/6J mice were infected with one alphavirus and were then challenged with the same (homologous) or different (heterologous) alphavirus ( Figure 2b). As expected, in all four homologous combinations, complete protection against viremia was observed ( Figure S4a,d, middle top graphs, red lines). Complete protection is defined herein as no post-challenge viremia detected on any day in any mouse; (limit of detection 2 log 10 CCID 50 /mL of serum).
As might also be expected amongst members of the same antigenic complex [41], in all heterologous combinations, some level of significant cross-protection was always evident (Figure 2c, see p-values). However, for half of the heterologous combinations, cross-protection was partial ( Figure 2c, middle column of graphs and third row of graphs). Thus, although all these viruses are clearly antigenically related, complete cross-protection was not universally assured. MAYV infection completely protected against delectable RRV TT viremia, whereas RRV TT infection only partially cross-protected against MAYV viremia ( Figure 2c). Although this might be interpreted as some form of one-way cross-protection [83], it should be noted that the "immunizing" viremias were (at their peak)~100-fold higher for MAYV than for RRV TT ( Figure S4).

Conservation of Receptor Contact Residues in E1 and E2
Although antibody-based virus neutralization can involve a range of mechanisms [5,46,85], most neutralizing antibodies target E1/E2, and many block receptor binding [81]. The 48 contact residues between the arthritogenic alphavirus receptor, MXRA8, and the E1/E2 heterotrimeric envelope glyoproteins of CHIKV were recently identified by cryo-electron microscopy [72]. These residues are relatively well conserved between CHIKV and ONNV (77% amino acid identity) but are less well conserved between CHIKV and MAYV (60% amino acid identity) and CHIKV and RRV (48% amino acid identity) (Figure 3a,b). The contact residues are represented on cryo-electron microscopy images, with color-coding (as in Figure 3a) illustrating the levels of conservation (Figure 3b). conserved between CHIKV and MAYV (60% amino acid identity) and CHIKV and RRV (48% amino acid identity) (Figure 3a,b). The contact residues are represented on cryo-electron microscopy images, with color-coding (as in Figure 3a) illustrating the levels of conservation (Figure 3b). Convalescent CHIKV, ONNV, MAYV and RRVTT sera from individual mice were used to determine the neutralization and cross-neutralization titers against CHIKV. When these titers were plotted against the percentage of amino acid identity in the E1/E2 receptor-binding residues, a highly significant relationship became evident (Figure 3c, left graph). Both the significance (p-value) and the correlation coefficient (rho) were lower when the same analysis was undertaken using the percentage of amino acid identity for all of the amino acids in E1/E2 (Figure 3c, middle graph), or the percentage of amino acid identity for all of the amino acids encoded by the viral genomes (Figure 3c, right graph). The level of cross-protection mediated by a polyclonal antibody response would thus appear to be closely related to the level of conservation in the E1/E2 receptor-binding residues.
Another key observation here is that reductions in the percentage of amino acid identities resulted in large drops in cross-neutralization titers. For example, an approximate halving of the percentage amino acid identity from 100% to 48% resulted in a ~100 fold drop in cross-neutralization titers (Figure 3c, left graph), likely explaining why only partial cross-protection was frequently observed in Figure 2c.  (a) The 48 residues on the E1/E2 Chikungunya virus (CHIKV) viral spike glycoproteins reported to make contact with the arthritogenic alphavirus receptor, MXRA8 [72] are aligned against the corresponding residues in o'nyong nyong virus (ONNV), Mayaro virus (MAYV) and RRV. Conservative and non-conservative substitutions are defined as described [86]. The complete sequence for ONNV IMTSSA 2004 was not available so residues E1 39-84 and E2 18-123 were derived from the closely related ONNV SG650 [62]; (b) The crystal structure of CHIKV E1/E2, with the receptor contact residues colored as in (a) for each of the indicated alphaviruses. Percentages indicate the percentage of identical contact residues relative to CHIKV; (c) Wild-type C57BL/6J mice were infected with CHIKV, ONNV MAYV or RRV. After 5 weeks, sera from each mouse was tested for its CHIKV neutralization titer. The neutralization titers were plotted against the % identity in E1/E2 receptor binding amino acids (left), the % identity in E1/E2 amino acids (middle) and % amino acid identity in the total genome (right). Statistics by Spearman's correlations, p and rho indicated.
Convalescent CHIKV, ONNV, MAYV and RRV TT sera from individual mice were used to determine the neutralization and cross-neutralization titers against CHIKV. When these titers were plotted against the percentage of amino acid identity in the E1/E2 receptor-binding residues, a highly significant relationship became evident (Figure 3c, left graph). Both the significance (p-value) and the correlation coefficient (rho) were lower when the same analysis was undertaken using the percentage of amino acid identity for all of the amino acids in E1/E2 (Figure 3c, middle graph), or the percentage of amino acid identity for all of the amino acids encoded by the viral genomes (Figure 3c, right graph). The level of cross-protection mediated by a polyclonal antibody response would thus appear to be closely related to the level of conservation in the E1/E2 receptor-binding residues.
Another key observation here is that reductions in the percentage of amino acid identities resulted in large drops in cross-neutralization titers. For example, an approximate halving of the percentage amino acid identity from 100% to 48% resulted in a~100 fold drop in cross-neutralization titers (Figure 3c, left graph), likely explaining why only partial cross-protection was frequently observed in Figure 2c.

Cross-Protection Provided by SCV-ZIKA/CHIK Vaccination in Wild-Type Mice
We have previously shown that a single vaccination of adult female C57BL/6J mice with 10 6 pfu of the SCV-ZIKA/CHIK vaccine i.m. induced CHIKV-specific ELISA and neutralizing antibody responses and provided complete protection against CHIKV viremia after CHIKV challenge [16].
To determine whether such SCV-ZIKA/CHIK vaccination could provide cross-protection against other human arthritogenic alphaviruses in the same antigenic complex, C57BL/6J mice were vaccinated with 10 6 pfu SCV-ZIKA/CHIK, and antibody responses and protection against ONNV, RRV TT and MAYV infection were determined (Figure 4a). Vaccination with 10 6 pfu of SCV-ZIKA/CHIK generated ELISA and neutralizing responses specific for both CHIKV and ONNV, although ELISA responses were slightly lower (1.6 fold) (Figure 4b), and neutralization titers were~10-fold lower for ONNV when compared with CHIKV ( Figure 4c). Nevertheless, the mean cross-neutralization titer against ONNV of 40.2 + SE 9.6 ( Figure 4c) was sufficient for complete cross-protection against detectable viremia post-ONNV challenge in this model (Figure 4d, red arrow). Although SCV-ZIKA/CHIK vaccination did induce a mean reciprocal endpoint anti-MAYV ELISA titer of 111 + SE 30 (Figure 4e), no cross-neutralization titers were detected (Figure 4f), and no cross-protection was observed (Figure 4g). No cross-reactive (Figure 4h) or cross-neutralizing (Figure 4i) antibody responses against RRV TT were observed, and a single vaccination with 10 6 pfu of SCV-ZIKA/CHIK was unable to provide any cross-protection against viremia (Figure 4j).
To determine whether increased SCV-ZIKA/CHIK vaccination dose and/or boosting might provide cross-protection against RRV TT , mice were vaccinated twice with 10 6 pfu SCV-ZIKA/CHIK, once with 10 7 pfu SCV-ZIKA/CHIK ( Figure S5) or twice with 10 7 pfu SCV-ZIKA/CHIK (Figure 4k). Only the latter dose and schedule induced RRV-specific antibody responses detectable by ELISA, although mean levels were~15-fold lower than CHIKV-specific ELISA titers (Figure 4l, p = 0.005). No cross-neutralizing antibody responses were detected against RRV TT (Figure 4m); nevertheless, partial cross-protection against RRV TT viremia after RRV TT challenge was evident (Figure 4n), with 3 out of 6 mice showing no detectable viremias. Thus even when SCV-ZIKA/CHIK was given at a 10-fold higher dose and twice, only 50% of the mice were fully protected against a detectable RRV TT viremia.

Cross-Protection Against RRV TT Mediated by the JEV/GETV Vaccine in Wild-Type Mice
Of the arthritogenic alphaviruses described herein, GETV is the most closely related to RRV (Figure 3a). A formalin-inactivated JEV/GETV vaccine is available in Japan for use in horses (Nisseiken, Tokyo, Japan) [31][32][33] and currently represents the only arthritogenic alphavirus for which a vaccine is commercially available. RRV also infects horses [37,38], but no commercial RRV vaccine is currently available for veterinary or human use.
We determined that the JEV/GETV vaccine contains 172 µg/mL of protein using a standard Bradford protein assay. The equine 3 mL dose thus represents~500 µg of protein, with two i.m. doses separated by 1 month recommended for horses by the manufacturer. To explore whether the JEV/GETV vaccine could provide cross-protection against RRV, mice were vaccinated once i.m. with 10 µg of the JEV/GETV vaccine ( Figure 5a); assuming 50% of this vaccine is GETV, this represents a vaccination dose of~5 µg of GETV proteins. An experimental formalin-inactivated RRV vaccine at a single dose of 2.5 µg was previously shown to induce mean reciprocal neutralizing titers of~30 and provided protection against viremia in 93% of CD-1 mice [48]. Reciprocal 50% neutralizing antibody titers for RRV of >10 were deemed a conservative estimate of protection in a human vaccine trial [22].
Mice vaccinated once with 10 µg of JEV/GETV vaccine generated a mean reciprocal anti-GETV ELISA antibody titer of 4141 + 1365, with the mean reciprocal anti-RRV TT ELISA titers~100 fold lower at 37.7 + SE 23.1 (Figure 5b). The mean reciprocal anti-GETV neutralizing titer was 35 + SE 20, with anti-RRV TT cross-neutralizing titers below the limit of detection (Figure 5c). PBS vaccination induced no detectable antibody responses (Figure 5b  Following challenge with RRV TT , significant, but only partial, protection against viremia was observed ( Figure 5h). (As we do not as yet have a mouse model of GETV, we were unable to verify that the commercial JEV/GETV vaccine can protect against GETV in mice.)  The E1/E2 receptor binding residues are relatively well conserved for both RRV TT and GETV  (the vaccine strain), and RRV TT and GETV MM2021 (the GETV isolate used for the ELISA and neutralization assays) (Figure 5i). Given the 75% amino acid identity between RRV TT and GETV MI-110 (Figure 5i), the poor level of cross-protection (Figure 5d,h) might be viewed as somewhat unexpected. However, the JEV/GETV vaccine is formalin-fixed, a process known to reduce immunogenicity [87], potentially by altering the antigenic structure [88]. Formalin irreversibly modifies certain epitopes, particularly those containing lysine (K) and (to a lesser extent) tryptophan (W) residues [89]. Induction of cross-reactive antibodies that are reliant on conserved epitopes that (for instance) contain E1 K 130 and/or W 89 (Figure 5i, underlined) may thus be compromised.

Cross-Protection Against RRV TT after SCV-ZIKA/CHIK Vaccination in IRF3/7 -/-Mice
Mice defective in type I interferon (IFN) responses, such as IFN alpha-receptor-deficient (IFNAR -/or A129) mice, have been widely utilized to evaluate alphaviral vaccines and antibodies, with protection against mortality often used as an indicator of efficacy [45,52,[90][91][92]. Such mice have also been used to demonstrate cross-protection against ONNV by a CHIKV vaccine [47]. Herein we used Interferon Response Factor 3-and 7-deficient (IRF3/7 -/-) mice rather than IFNAR -/mice, as severe disease has a slower onset [55], which facilitates compliance with ethically defined endpoints for euthanasia, as monitoring of animal well-being can be reduced to twice daily. The behavior of CHIKV in IRF3/7 -/and IFNAR -/mice is otherwise very similar; both IFNAR -/and IRF3/7 -/mice are unable to mount protective type I IFN responses and generally show a lethal phenotype associated with hemorrhagic shock after CHIKV infection [55,93]. The latter represents a severe disease manifestation occasionally seen in CHIKV and MAYV patients [5]. To our knowledge, RRV infection of IRF3/7 -/mice has not previously been reported, with infection (as might be expected [55]) resulting in high viremia, foot swelling (edema) and mortality ( Figure S6a-d). Subcutaneous monocyte/macrophage infiltration was also illustrated by immunohistochemistry ( Figure S6e); a feature not previously reported for alphavirus infections in type I IFN response defective mice.
To evaluate cross-protection in type I interferon-response-deficient mice, IRF3/7 -/mice were vaccinated once with 10 6 pfu SCV-ZIKA/CHIK and were then challenged with RRV TT (Figure 6a). Vaccination induced CHIKV-specific ELISA and neutralizing antibody responses, but anti-RRV TT antibody responses were below the level of detection (Figure 6b,c). After RRV TT challenge, both SCV-ZIKA/CHIK and SCV-control vaccinated mice developed high viremias; although on days 4 and 5, post-challenge SCV-ZIKA/CHIK-vaccinated mice showed significant 0.8-1.4 log lower viremia titers (Figure 6d). After RRV TT challenge, both SCV-ZIKA/CHIK and SCV-control vaccinated mice showed similar levels of foot swelling ( Figure 6e); images of typical foot swelling in control mice are shown in Figure 6f. Mice were also monitored for a series of disease signs (1 = mild, 2 = moderate, 3 = severe); any animal scoring 2 in two or more disease parameters (or 3 in any one parameter) was deemed to have reached an ethically defined endpoint requiring euthanasia. No SCV-ZIKA/CHIK-vaccinated animals reached this criterion, whereas all SCV-control vaccinated mice scored 2 for Joint swelling and 2 for Posture on day 6 and were euthanized (Figure 6g). Presented as a Kaplan-Meier survival plot, SCV-ZIKA/CHIK vaccinated animals showed significantly better survival than SCV-control vaccinated mice (Figure 6h).
SCV-ZIKA/CHIK vaccination of IRF3/7 -/mice thus provided no cross-protection against RRV TT foot swelling, marginal cross-protection against viremia and disease, but complete cross-protection against mortality. The same vaccination also provided no protection against RRV TT viremia in wild-type mice (Figure 4j). Survival in the type I interferon-deficient mouse model is thus a poor indicator of (or poor surrogate marker for) cross-protection against infection and arthritic disease, with low-level cross-reactive anamnestic responses perhaps only just able to generate sufficient immune responses in time to prevent mortality. (g) Mice were scored for 1 = mild, 2 = moderate, 3 = severe for the indicated 6 disease manifestations. The mean scores for each disease manifestation for 6 mice per group are shown for SCV-ZIKA/CHIK (left) and SCV-Control vaccinated mice (right); (h) Kaplan-Meier survival curves for mice described in (g), with mice scoring 2 or more for two disease manifestations deemed to have reached ethically defined endpoints requiring euthanasia. An additional group (n = 3) were mock-vaccinated with PBS. Statistics by log rank test.
Rag -/-mice have no B or T cell responses and are unable to clear CHIKV or RRV infections resulting in long-term steady-state viremias [56,82]. Rag1 -/-mice were infected with CHIKV, ONNV, MAYV or RRVTT (n = 4 per group), and after the persistent viremias were established, mice were
Rag -/mice have no B or T cell responses and are unable to clear CHIKV or RRV infections resulting in long-term steady-state viremias [56,82]. Rag1 -/mice were infected with CHIKV, ONNV, MAYV or RRV TT (n = 4 per group), and after the persistent viremias were established, mice were treated on day 10 with pooled CHIKV convalescent sera (Figure 7a). The sera were pooled from 18 CHIKV-infected mice and had a reciprocal CHIKV IgG ELISA endpoint titer of~50,000. With the exception of RRV TT , all viremia levels dropped to levels below detection within a day, with the MAYV viremia reappearing by day 20 in three mice and by day 30 in the remaining mouse (Figure 7b; Figure S7). The ONNV and CHIKV viremia levels remained depressed for >47 days and were not significantly different from each other (on days 34 and 39) (Figure 7b). The results are consistent with the ability of SCV-ZIKA/CHIK vaccination to protect against CHIKV [16] and cross-protect against ONNV, but not MAYV or RRV (Figure 4). This pattern (Figure 7b) also correlates well with amino acid identity in receptor contact residues [81] shown in Figure 3b (Figure S7); with the cross-protection activity of anti-CHIKV sera against persistent viremia, best for ONNV (77% identity), less for MAYV (60% identity) and least for RRV TT (48% identity) (Figure 7b). Taken together, these data implicate cross-neutralizing IgG responses as playing an important role in mediating cross-protection. (These data also illustrate the utility of Rag -/mice for evaluating anti-viral treatments for arthritogenic alphaviruses, with further examples shown in Figure S9.) Vaccines 2020, 8, x 18 of 26 treated on day 10 with pooled CHIKV convalescent sera (Figure 7a). The sera were pooled from 18 CHIKV-infected mice and had a reciprocal CHIKV IgG ELISA endpoint titer of ~50,000. With the exception of RRVTT, all viremia levels dropped to levels below detection within a day, with the MAYV viremia reappearing by day 20 in three mice and by day 30 in the remaining mouse (Figure 7b; Figure  S7). The ONNV and CHIKV viremia levels remained depressed for >47 days and were not significantly different from each other (on days 34 and 39) (Figure 7b). The results are consistent with the ability of SCV-ZIKA/CHIK vaccination to protect against CHIKV [16] and cross-protect against ONNV, but not MAYV or RRV (Figure 4). This pattern (Figure 7b) also correlates well with amino acid identity in receptor contact residues [81] shown in Figure 3b ( Figure S7); with the crossprotection activity of anti-CHIKV sera against persistent viremia, best for ONNV (77% identity), less for MAYV (60% identity) and least for RRVTT (48% identity) (Figure 7b). Taken together, these data implicate cross-neutralizing IgG responses as playing an important role in mediating crossprotection. (These data also illustrate the utility of Rag -/-mice for evaluating anti-viral treatments for arthritogenic alphaviruses, with further examples shown in Figure S9.)

Discussion
We illustrate herein that some level of cross-protection, mediated by CHIKV, ONNV, MAYV and RRV infections, was universal in adult wild-type mice, as might be envisaged given these viruses all belong to the Semliki Forest virus antigenic complex. However, in half the combinations, crossprotection was only partial (Figure 2c, Figure S4). The live attenuated recombinant poxvirus-based SCV-ZIKA/CHIK vaccine was able to mediate complete cross-protection against ONNV challenge but was unable to provide such protection against MAYV or RRV (Figure 4). The formalin-inactivated JEV/GETV vaccine was unable to provide complete cross-protection against RRV challenge, despite the high E1/E2 sequence identities between these two viruses (Figures 2a,5i). Both infection-mediated and vaccine-mediated cross-protection were thus often partial or absent.
Although SCV-ZIKA/CHIK vaccination provided effective cross-protection against ONNV challenge (Figure 4d), the mean neutralizing antibody response was ~10-fold lower for ONNV than for CHIKV (Figure 4c). In addition, after SCV-ZIKA/CHIK vaccination, the ratio of neutralizing titers over ELISA titers was significantly lower for ONNV than for CHIKV ( Figure S8a, p < 0.001), further illustrating that the SCV-ZIKA/CHIK vaccine is less ideal for ONNV than it is for CHIKV. In humans, Figure 7. Antibody-mediated cross-protection in Rag -/mice. (a) Timeline of infections, viremia determinations and injection of convalescent CHIKV mouse serum in Rag1 -/mice; (b) Viremias for the indicated alphaviruses before and after injection of anti-CHIKV convalescent serum (n = 4 mice per group). (Limit of detection for each mouse 2 log 10 CCID 50 /mL.) For all reductions to undetectable viremia levels for each day relative to day 10, p = 0.037 by Kolmogorov-Smirnov tests (as differences in variance were infinite).

Discussion
We illustrate herein that some level of cross-protection, mediated by CHIKV, ONNV, MAYV and RRV infections, was universal in adult wild-type mice, as might be envisaged given these viruses all belong to the Semliki Forest virus antigenic complex. However, in half the combinations, cross-protection was only partial (Figure 2c, Figure S4). The live attenuated recombinant poxvirus-based SCV-ZIKA/CHIK vaccine was able to mediate complete cross-protection against ONNV challenge but was unable to provide such protection against MAYV or RRV (Figure 4). The formalin-inactivated JEV/GETV vaccine was unable to provide complete cross-protection against RRV challenge, despite the high E1/E2 sequence identities between these two viruses (Figures 2a and 5i). Both infection-mediated and vaccine-mediated cross-protection were thus often partial or absent.
Although SCV-ZIKA/CHIK vaccination provided effective cross-protection against ONNV challenge (Figure 4d), the mean neutralizing antibody response was~10-fold lower for ONNV than for CHIKV (Figure 4c). In addition, after SCV-ZIKA/CHIK vaccination, the ratio of neutralizing titers over ELISA titers was significantly lower for ONNV than for CHIKV ( Figure S8a, p < 0.001), further illustrating that the SCV-ZIKA/CHIK vaccine is less ideal for ONNV than it is for CHIKV. In humans, the minimum vaccine dose to achieve protection against its designated target is generally used to avoid excessive side effects such as reactogenicity [100] and to reduce the cost of goods. In humans, reciprocal neutralization titers of >10 have been deemed to represent a conservative estimate of the protective titer for RRV [22] and for CHIKV [101,102]; a contention not inconsistent with the mouse data presented herein ( Figure S8b). Thus, if the SCV-ZIKA/CHIK vaccine dose was reduced in humans to provide such levels of anti-CHIKV neutralization, the lower cross-neutralization titers against ONNV may provide suboptimal cross-protection against ONNV. Testing in clinical trials and ultimate licensing of a vaccine that requires higher or more vaccine doses in order to also provide cross-protection might present regulatory hurdles when the technology (if not the market) likely exists to generate an appropriately targeted vaccine. Vaccine manufacturers are also likely to focus on the primary target and not complicate registration and licensing by risking increased side effects (and countenancing increased cost of goods) in an attempt to capture an additional, economically less viable, cross-protection market.
Despite the high level of sequence identity between GETV and RRV, JEV/GETV vaccination was unable to induce anti-RRV cross-neutralizing antibody responses (Figure 5c,g) or to provide complete protection against RRV TT challenge (Figure 5d,h). Anti-RRV TT ELISA responses were also 100-and 14-fold lower than the anti-GETV ELISA titers (Figure 5b,f, respectively), again suggesting that cross-protection against RRV would require a higher dose of the JEV/GETV vaccine than that needed for protection against GETV. Even the anti-GETV neutralizing responses generated by the GETV vaccine were relatively poor, perhaps suggesting that a very high dose or multiple vaccinations would be needed before cross-neutralizing responses to RRV are seen. The suboptimal induction of neutralizing antibody responses by a formalin-fixed vaccine [87][88][89] is perhaps supported herein by the observation that the ratios of neutralizing titers over ELISA titers were significantly higher for SCV-ZIKA/CHIK than for JEV/GETV vaccinations (Supplementary Figure S8c, p < 0.0001). Generating cross-neutralizing responses with a formalin-fixed vaccine may thus represent a particularly difficult hurdle.
Herein, protection against viremia was used as the readout, rather than protection against disease, which primarily manifests in humans as acute and often chronic polyarthralgia/polyarthritis [4,5]. After injection of 10 4 CCID 50 of CHIKV s.c. into the foot of adult wild-type mice, the arthritic foot swelling reached a maximum mean~60%-80% increase in foot height x width [56,103]. For RRV TT , such foot swelling only reached~12% in 6-week-old mice and was essentially below detection in 24-week-old mice (Figure 1d). Although a higher dose (10 5 or 10 6 PFU) and different strain of MAYV (BeH407) has been reported to induce foot swelling comparable to that seen for CHIKV after s.c. injection into the feet [104], no significant foot swelling was seen herein in control mice challenged with 10 4 CCID 50 MAYV (or ONNV). These alphaviruses thus show distinct differences in their propensity to induce foot swelling in these wild-type mouse models, making cross-protection against arthritis difficult to standardize. No mouse models have so far been developed that measure polyarthralgia. The adult wild-type CHIKV mouse model recapitulates many aspects of human disease [42,57], with recent RNA-Seq comparisons further illustrating that this mouse model shares many inflammatory processes with those seen in humans [105,106]. However, such comparisons are not yet available for ONNV, MAYV or RRV. Further development and characterization of alphavirus mouse models would thus be required before one could formally test the inherent assumption herein that suppression of viremia correlates with suppression of rheumatic disease.
Overall the data presented herein does not support the concept of effective mediation of cross-protection by arthritogenic alphavirus vaccines. A vaccine capable of protecting against multiple arthritogenic alphaviruses would likely require a polyvalent vaccine [54] or a mixture of individual vaccines, as was recently described for the three major encephalitic alphaviruses [107]. However, the current small market size and low mortality rates for RRV, ONNV and MAYV would likely make such endeavors commercially unattractive. New outbreaks may clearly change this contention [10,14]; until then, mosquito control measures are likely to remain at the forefront of public health measures against these latter alphaviruses [108].

Conclusions
A CHIKV vaccine is deemed commercially viable, and several vaccine candidates are progressing into, and through, human clinical trials. Although CHIKV-vaccine or CHIKV-infection provided a level of cross-protection against infection by ONNV, MAYV and RRV, cross-protection and/or cross-neutralizing antibody responses were significantly lower than protection and/or neutralizing antibody responses against CHIKV. Increasing CHIKV vaccine dosing and/or scheduling to try and capture the much smaller cross-protection markets is unlikely to be attractive to regulators or vaccine manufacturers.