Attenuation of Chikungunya Virus Vaccine Strain 181/Clone 25 Is Determined by Two Amino Acid Substitutions in the E2 Envelope Glycoprotein

Abstract

Chikungunya virus (CHIKV) is the mosquito-borne alphavirus that is the etiologic agent of massive outbreaks of arthralgic febrile illness that recently affected millions of people in Africa and Asia. The only CHIKV vaccine that has been tested in humans, strain 181/clone 25, is a live-attenuated derivative of Southeast Asian human isolate strain AF15561. The vaccine was immunogenic in phase I and II clinical trials; however, it induced transient arthralgia in 8% of the vaccinees. There are five amino acid differences between the vaccine and its parent, as well as five synonymous mutations, none of which involves cis-acting genome regions known to be responsible for replication or packaging. To identify the determinants of attenuation, we therefore tested the five nonsynonymous mutations by cloning them individually or in different combinations into infectious clones derived from two wild-type (WT) CHIKV strains, La Reunion and AF15561. Levels of virulence were compared with those of the WT strains and the vaccine strain in two different murine models: infant CD1 and adult A129 mice. An attenuated phenotype indistinguishable from that of the 181/clone 25 vaccine strain was obtained by the simultaneous expression of two E2 glycoprotein substitutions, with intermediate levels of attenuation obtained with the single E2 mutations. The other three amino acid mutations, in nsP1, 6K, and E1, did not have a detectable effect on CHIKV virulence. These results indicate that the attenuation of strain 181/clone 25 is mediated by two point mutations, explaining the phenotypic instability observed in human vaccinees and also in our studies.

INTRODUCTION

Chikungunya virus (CHIKV) is a mosquito-borne human pathogen in the genus Alphavirus, family Togaviridae. First isolated by Ross during a 1952-1953 epidemic of arthralgic disease in Tanzania, CHIKV is now known to occur in many parts of Sub-Saharan Africa, as well as in Asia (33, 47). Human disease occurs through two different mechanisms: (i) direct spillover from enzootic transmission cycles in forests of Africa, where arboreal mosquitoes transmit CHIKV among nonhuman primates, and (ii) an urban transmission cycle involving humans as amplification hosts and the peridomestic mosquitoes Aedes (Stegomyia) aegypti and A. (Stegomyia) albopictus as vectors. Endemic disease caused by enzootic spillover is rarely recognized because chikungunya fever (CHIK) is not easily distinguished from many other tropical diseases such as dengue or malaria without laboratory diagnostics. The urban form of CHIK occurs in both Africa and Asia, and emergences of the urban cycle from African enzootic strains have probably occurred for centuries, aided initially by transport on sailing ships and more recently by air travel by infected persons (33, 47, 55). Unprecedented epidemic activity has been documented since 2004, when outbreaks first appeared in coastal Kenya (16), followed by transport of the virus to islands in the Indian Ocean (37) and then independently to India (50). These epidemics resulted in thousands of infected travelers returning to Europe, the Americas, and Southeast Asia, in some cases initiating urban transmission cycles (13, 34, 40, 42). Since 2004, millions of persons have been infected with CHIKV and epidemics continue to occur in India and Southeast Asia (41).

CHIKV infection is typically characterized by an abrupt onset of fever and polyarthralgia, often accompanied by a rash (38, 41). Although many patients recover completely within a few weeks of onset, the arthralgia, which is extremely painful and debilitating, can be persistent, and has been shown to be a major cause of morbidity over 1 year after the end of epidemics (11). Fatalities following CHIKV infection have also been documented (24), including neurologic disease (5, 9), and excess deaths reported during epidemics suggest that CHIKV infection is at least partially responsible for many fatal outcomes (15, 24, 43).

Like other alphaviruses, CHIKV has a single-stranded, positive-sense RNA genome of about 11.8 kb that contains two open reading frames (ORFs), (i) a nonstructural polyprotein ORF encoded in the 5′ two-thirds of the genome that is translated directly from genomic RNA to produce four nonstructural proteins involved in genome replication, capping, and polyprotein processing and (ii) a structural polyprotein ORF that is identical in sequence to the 3′ third of the genome, is encoded by a subgenomic RNA, and produces the capsid, as well as the E2 and E1 envelope glycoproteins (20). Virions are composed of the genomic RNA complexed with 240 copies of the capsid protein, which form nucleocapsids in the cytoplasm of infected cells, and 240 E2/E1 heterodimers, which are embedded in the plasma membrane following their transit through the secretory system. Thus, the 70-nm virion is enveloped, with 80 trimeric E2/E1 spikes projecting outward. The E2 protein forms the tips of the spikes and is believed to interact with cellular receptors (52). It is also the target of most alphavirus-neutralizing antibodies (35).

Although vaccine candidates for CHIK have been developed, only one has been tested in humans. Levitt et al. (21) used a wild-type (WT) CHIKV strain isolated in Thailand, AF15561, to perform 18 plaque-to-plaque passages in MRC-5 human lung fibroblast cultures and selected a plaque clone designated 181/clone 25 (hereinafter abbreviated 181/25) for further testing based on its small plaque size, attenuation in infant mice, and reduced viremia in nonhuman primates. Strain 181/25 elicited neutralizing antibodies after a single vaccination of mice or rhesus macaques and protected against both viremia and disease following a WT CHIKV challenge. This vaccine (called TSI-GSD-218 during human trials) also proved highly immunogenic in humans, with all but one vaccinee developing neutralizing antibodies (8). However, 5/59 vaccinees developed mild, transient arthralgia, suggesting insufficient and/or unstable attenuation, and no further clinical trials have been undertaken. Recent studies of another live-attenuated CHIK vaccine candidate that included direct comparisons with strain 181/25 also reported some residual pathogenicity of the latter (32).

Subsequent sequence comparisons between CHIKV parental strain AF15561 (GenBank accession no. EF452493) and derived attenuated strain 181/25 (GenBank accession no. L37661, listed as TSI-GSD-218) revealed 10 nucleotide differences in the genome, with 5 being nonsynonymous and none involving cis-acting genome regions known to be responsible for replication or packaging (Fig. 1a; Table 1). To further assess the attenuation of strain 181/25 and to determine the mutations responsible for its attenuation, we conducted reverse genetic studies to examine the attenuating effects of the amino acid changes encoded by these five nonsynonymous mutations that accompanied its derivation.

Fig 1.

Schematic representation of the CHIKV genome. (a) Positions of nonsynonymous mutations in vaccine strain 181/25. (b) Mutations cloned into the CHIKV LR genomic backbone. Numbers correspond to amino acid positions of the mutations in the respective viral protein. SG, subgenomic.

Table 1.

Mutations accompanying the generation of CHIKV vaccine strain 181/25 (AKA TSI-GSD-218) from its parent strain AF15561

Gene	Nucleotide position	Amino acid position	Amino acid substitution
nsP1	978	301	Thr→Ile
nsP2	3706a
nsP4	7259a
E3	8463a
E2	8576	12	Thr→Ile
	8760a
	8785	82	Gly→Arg
6K	9935	42	Cys→Phe
E1	10446a
	11204	404	Ala→Val

^aSynonymous mutations.

MATERIALS AND METHODS

Cell culture.

Vero African green monkey kidney cells (ATCC CCL-81; American Type Culture Collection, Manassas, VA) were maintained in Dulbecco's modified Eagle medium (DMEM) supplemented with 10% fetal bovine serum, nonessential amino acids, sodium pyruvate, and penicillin-streptomycin at 37°C under 5% CO₂.

Virus stocks.

CHIKV strain 181/25 (21) was passaged once on Vero cells. Stocks of all other WT and mutant CHIKV strains were generated by electroporation of Vero cells with in vitro-transcribed RNAs from cDNA infectious clones (see below). Briefly, subconfluent cell cultures from two T150 flasks were trypsinized, washed three times with Dulbecco's phosphate-buffered saline (PBS), resuspended in 700 μl of PBS, mixed with 10 μl of the in vitro-transcribed RNA suspension, and electroporated using a BTX ECM 830 instrument (Harvard Apparatus, Holliston, MA) in 4-mm cuvettes with three 10-ms pulses at 250 V at 1-s intervals. Cells were mixed with 10 ml of DMEM after standing at room temperature for 10 min and seeded into T75 flasks. Cell supernatants containing virus were harvested at 22 to 26 h postelectroporation, after the development of visible cytopathic effects (CPE), and clarified by centrifugation at 1,000 × g for 5 min, and titers were determined by plaque assay on Vero cells as described previously (1).

Viral infectious clone plasmids.

The CHIKV La Reunion (LR) cDNA infectious clone (45) was provided by Steven Higgs at the University of Texas Medical Branch (UTMB). The CHIKV strain AF15561 (21) sequence was retrieved from GenBank (accession no. EF452493), and a full-length cDNA of this viral sequence was synthesized in five fragments by GenScript (Piscataway, NJ). A SacI restriction enzyme site and an SP6 RNA polymerase promoter were added at the 5′ terminus, and a poly(A) tail, followed by a NotI restriction site, was added at the 3′ terminus. Ambiguous nucleotides present in the GenBank sequence were replaced as described in Results. These five cDNA fragments were assembled using standard DNA cloning techniques together with the NotI-SacI DNA fragment from the CHIKV LR infectious clone, which contained only the cloning vector, to generate the CHIKV 15561 cDNA infectious clone.

The CHIKV vaccine strain 181/25 sequence was obtained from GenBank (accession no. L37661, TSI-GSD-218). Compared to strain AF15561, its WT parent, there are five nonsynonymous mutations in its genome (Table 1 and Fig. 1a). Using PCR-mediated mutagenesis, these mutations were cloned separately or in combinations into the CHIKV LR and 15561 plasmids to generate cDNA infectious clones of the mutants (Fig. 1b; see also Fig. 5).

Fig 5.

Schematic representations of CHIKV AF15561 genomes with cloned mutations. Numbers correspond to amino acid positions of the mutations in the respective viral proteins.

Plasmid purification and in vitro transcription.

Large-scale preparations of infectious clone plasmids were made using standard methods, and plasmids were purified on CsCl gradients (36). All in vitro-synthesized and PCR-derived DNA sequences were confirmed using the BigDye Terminator v3.1 Cycle Sequencing kit on a 3500 Genetic Analyzer (Applied Biosystems, Foster City, CA). Plasmids were linearized using the NotI restriction enzyme prior to in vitro transcription using SP6 RNA polymerase (Ambion, Austin, TX) in the presence of an RNA cap analog using the manufacturer's protocol. The quality and integrity of RNA were confirmed by agarose gel electrophoresis under nondenaturing conditions, and it was stored at −80°C.

CD1 mouse infections.

All mouse experiments were approved by the Institutional Animal Care and Use Committees of UTMB or the University of Wisconsin (UW). Pregnant CD1 mice were purchased from Charles River Laboratories (Wilmington, MA). Cohorts (n = 15 to 36) of 5- to 6- or 6- to 7-day-old mice were infected subcutaneously (SC) in the back with 10⁵ PFU of virus. For experiments with mutations in the CHIKV strain LR backbone, body weights were measured daily and blood and hind limb and brain tissues were harvested (n = 2 or 3) on days 2, 4, 6, 8, and 10 postinfection. For experiments with mutations in the AF15561 backbone, blood and hind limb and brain tissues were harvested (n = 6) on days 1 to 5 and 7 postinfection. Viremia and viral loads were measured by plaque assay on Vero cells.

A129 mouse infections.

Strain A129 mice were bred at UW or UTMB from a breeding pair obtained from B & K Ltd. (Grimston, United Kingdom). Telemetric devices (Bio Medic Data Systems, Inc., Seaford, DE) were implanted SC into mice 1 week prior to infection. Cohorts (n = 4 to 9) of 5- to 8- or 8- to 10-week-old mice were infected intradermally (ID) in the left rear footpad with 10⁴ PFU of virus. Animals were monitored for 18 days for morbidity and mortality, and body weights and temperatures were measured daily. Mice were bled on days 1 to 4 (LR mutants) or on days 1 and 3 postinfection (AF15561 mutants), and viremia was determined by quantitative reverse transcription-PCR (qRT-PCR) as described previously (32) or by plaque assay on Vero cells, respectively. Measurements of the perimetatarsal height of the hind feet were taken at 48 h (LR mutants) or 2, 4, and 7 days (AF15561 mutants) postinfection.

Viral RNA extractions and sequencing.

Total RNA was extracted from 200 μl of CHIKV strain 181/25 viral stocks or 50 μl of blood collected from infected mice using TRIzol LS (Invitrogen, Carlsbad, CA) using the manufacturer's protocol. cDNA was synthesized using the SuperScriptIII First Strand kit (Invitrogen) with random hexamer primers. Three-microliter RT reaction mixtures were used for PCRs with Phusion Hot Start II High-Fidelity DNA Polymerase (New England BioLabs, Ipswich, MA) and CHIKV-specific primers (sequences are available upon request). Amplicons were purified from agarose gels with the QIAquick Gel Extraction kit (Qiagen, Valencia, CA) and used for sequencing as described above.

RESULTS

Construction of CHIKV LR clones with strain 181/25 mutations.

Alignment of sequences of CHIKV strain AF15561 and its derivative vaccine strain 181/25 revealed 10 nucleotide differences, all within ORFs of the genome. Five of these changes were synonymous, and the other five led to deduced amino acid differences, two in the E2 envelope glycoprotein and one each in the nsP1, 6K, and E1 proteins (Table 1 and Fig. 1a). The five synonymous mutations that accompanied the derivation of strain 181/25 were assumed to have little or no phenotypic effect, so our efforts focused on the nonsynonymous mutations. Therefore, they were incorporated individually into the genome of the CHIKV strain LR infectious clone (Fig. 1b) derived from a WT human isolate from La Reunion Island in the Indian Ocean (45).

Mutagenized cDNA clones, along with the WT LR clone, were used to produce infectious RNA in vitro, which was electroporated into Vero cells. All rescued mutant viruses were viable, producing CPE in Vero cells within 24 h of electroporation and replicating to 5 to 6 log₁₀ PFU/ml in cell supernatants, comparable to the WT virus (data not shown). One noticeable phenotype of the LR/E2-82 mutant was a reduced plaque size on Vero cells. This may be attributed to the addition of a positive charge (glycine→arginine) at a putative heparan sulfate binding site of E2, which can lead to increased interaction of the viral particles with the polyanions in agarose medium during plaque formation (3, 23). The phenomenon of positive charge acquisition in the E2 glycoprotein is commonly observed when alphaviruses are passaged on cultured cell lines and often correlates with the acquisition of an attenuated phenotype (7, 14, 19).

Infection of infant CD1 mice with CHIKV LR mutants.

Several mouse models have been developed to study the virulence of CHIKV (6, 10, 26, 56). One utilizes infant outbred CD1 mice, which develop a disease similar to that seen in humans (56). We tested the attenuation of the five single CHIKV LR mutants by SC infecting cohorts (n = 15 to 18) of 6- to 7-day-old CD1 mice with 10⁵ PFU. Cohorts were also infected with WT CHIKV LR and vaccine strain 181/25. Body weights were measured daily as indicated in Fig. 2, and all cohorts of mice continued to gain weight at similar rates, including those infected with WT LR virus. For this reason, body weights were not monitored further in CD1 mouse experiments.

Fig 2.

Mean body weights of 6- to 7-day-old CD1 mice (n = 10) after SC infection with 10⁵ PFU of CHIKV strain LR, mutants, or vaccine strain 181/25.

Groups of two or three mice from each cohort were sacrificed on days 2, 4, 6, 8, and 10 after infection, and blood and hind limb and brain tissues were collected to determine viral loads. No viremia was detected on days 2 and 4 postinfection in 181/25-infected mice, whereas 3/3 strain LR-infected mice were positive, with a mean viremia titer of 2.9 log₁₀ PFU/ml on day 2 and partial clearance by day 4, with only 1/3 mice viremic (Table 2). The LR/6K mutant was the most virulent, with all of the mice testing positive on days 2 and 4 after infection and with mean titers higher than those of the WT LR strain. Although it is possible that the 6K mutation slightly increases the virulence of CHIKV, a more likely explanation is that this cohort, which was tested independently of the others, was affected by maternal care or health or by the slightly lower age of the animals. The lack of any difference between the triple mutant described below, which includes the 6K mutation, and WT CHIKV supports this explanation. The other four single mutations produced intermediate viremia phenotypes (between WT and vaccine strains), in terms of the fraction of positive mice and mean titers, with LR/nsP1- and LR/E1-infected mice exhibiting higher-virulence phenotypes comparable to that of WT CHIKV at day 2 postinfection (Table 2). No viremia was detected in samples tested on day 6 after infection (data not shown).

Table 2.

Viremia in 6- to 7-day-old CD1 mice after SC infection with WT or mutant CHIKV strain LR or vaccine strain 181/25

Virus straina	Day 2 postinfection		Day 4 postinfection
	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD
LR	3/3	2.9 ± 0.6	1/3	1.3 ± 0.6
LR/nsP1	2/2	2.2 ± 0.2	0/2	≤ 0.9
LR/6K	3/3	5.0 ± 0.2	3/3	2.7 ± 0.8
LR/E1	3/3	2.5 ± 0.5	0/3	≤ 0.9
LR/E2-12	2/3	2.0 ± 1.1	0/3	≤ 0.9
LR/E2-82	1/2	1.6 ± 1.0	1/2	1.5 ± 0.8
181/25	0/2	≤ 0.9	0/2	≤ 0.9

^aSee Fig. 1b for strain designations.

^bTiters are in log₁₀ PFU/ml of serum, and the limit of detection of the assay was 0.9 log₁₀ PFU/ml of serum.

Analysis of the viral loads in hind limb tissues showed high viral titers in both WT- and vaccine strain-infected mice on days 2, 4, and 6 (Table 3), followed by clearance of the viruses by day 10, when only one mouse (infected with the WT LR strain) had virus present out of all sampled cohorts (data not shown). High viral loads in the knee tissues of the 181/25-infected mice might correlate with transient arthralgia in some of the vaccinees during human trials (8). The LR/6K mutant had the highest mean titers, and the LR/E1 and LR/E2-12 mutants also had phenotypes similar to that of WT CHIKV, with comparable mean hind limb titers on days 2 and 6 after infection. The single mutants, with the exception of LR/6K, exhibited lower-virulence phenotypes than the 181/25 strain, with lower titers at several time points and clearance by day 8 (Table 3). The LR/E2-82 mutant was the most attenuated, with no detectable virus in the hind limb at any time postinfection. The differences in viral titers between the vaccine strain and some of the individual mutants could be due to different genomic backgrounds (AF15561 versus LR, respectively).

Table 3.

Viral loads in hind limb tissue from 6- to 7-day-old CD1 mice after SC infection with WT or mutant CHIKV strain LR or vaccine strain 181/25

Virus straina	Day 2 postinfection		Day 4 postinfection		Day 6 postinfection		Day 8 postinfection
	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD
LR	2/3	3.3 ± 2.2	2/3	4.3 ± 3.0	2/3	2.3 ± 1.2	1/3	2.1 ± 2.0
LR/nsP1	2/2	3.1 ± 0.1	1/2	2.6 ± 2.3	0/2	≤ 0.9	0/2	≤ 0.9
LR/6K	3/3	5.0 ± 1.4	3/3	6.8 ± 0.2	3/3	6.9 ± 0.9	1/3	2.3 ± 2.5
LR/E1	3/3	4.3 ± 1.3	3/3	5.8 ± 1.1	1/3	1.5 ± 1.1	0/3	≤ 0.9
LR/E2-12	3/3	4.3 ± 1.0	2/3	3.5 ± 2.4	1/3	1.9 ± 1.8	0/3	≤ 0.9
LR/E2-82	0/2	≤ 0.9	0/2	≤ 0.9	0/2	≤ 0.9	0/2	≤ 0.9
181/25	1/2	3.7 ± 4.0	2/2	5.2 ± 1.2	2/2	4.2 ± 0.4	1/2	3.1 ± 3.0

^aSee Fig. 1b for strain designations.

^bTiters are in log₁₀ PFU/g of tissue, and the limit of detection of the assay was 0.9 log₁₀ PFU//g of tissue.

CHIKV was detected in the brains of strain LR-infected CD1 mice up to day 6, while only one mouse infected with vaccine strain 181/25 was positive (on day 4) during the time course of the experiment (Table 4). Again, strains LR/6K and LR/E1 were the most virulent mutants, behaving the same way as WT CHIKV. Similar to the results from the limb tissues, there was no virus in the brains of any LR/E2-82 mutant-infected mice at any time point. Also, the LR/E2-12 mutant was detected in the brain of only one mouse on day 4 after infection. All brains sampled were virus negative at day 10 postinfection (data not shown).

Table 4.

Viral loads in brain tissue from 6- to 7-day-old CD1 mice after SC infection with WT or mutant CHIKV strain LR or vaccine strain 181/25

Virus straina	Day 2 postinfection		Day 4 postinfection		Day 6 postinfection		Day 8 postinfection
	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD
LR	3/3	3.9 ± 1.3	2/3	2.3 ± 1.3	2/3	3.1 ± 2.1	0/3	≤ 0.9
LR/nsP1	1/2	2.3 ± 2.0	0/2	≤ 0.9	1/2	2.8 ± 2.6	1/2	2.2 ± 1.8
LR/6K	3/3	4.2 ± 0.4	2/3	3.9 ± 2.8	3/3	4.3 ± 1.2	2/3	3.7 ± 2.6
LR/E1	2/3	3.0 ± 1.8	2/3	4.6 ± 3.3	2/3	2.4 ± 1.3	1/3	1.7 ± 1.3
LR/E2-12	0/3	≤ 0.9	1/3	1.8 ± 1.6	0/3	≤ 0.9	0/3	≤ 0.9
LR/E2-82	0/2	≤ 0.9	0/2	≤ 0.9	0/2	≤ 0.9	0/2	≤ 0.9
181/25	0/2	≤ 0.9	1/2	3.7 ± 3.9	0/2	≤ 0.9	0/2	≤ 0.9

^aSee Fig. 1b for strain designations.

^bTiters are in log₁₀ PFU/g of tissue, and the limit of detection of the assay was 0.9 log₁₀ PFU//g of tissue.

In summary, infections of infant CD1 mice with the single mutants showed the attenuation phenotype of LR/E2-82 to be close to that of vaccine strain 181/25. To confirm that the remaining four nonsynonymous mutations did not strongly influence virulence, we constructed an additional clone, LR/4x (Fig. 1b), with these four amino acid substitutions (in nsP1, E2-12, 6K, and E1) combined in the LR genomic background. The LR/4x mutant virus was successfully rescued in Vero cells and was used to infect 5- to 6-day old CD1 mice as described above, along with WT LR and 181/25 controls. As expected, the LR/4x mutant behaved like the WT LR strain, showing high viremia titers up to day 4 after infection (Table 5). In contrast to the lack of viremia in the previous experiment, 33% of the 181/25-infected mice showed viremia on days 2 and 4 postinfection, which may be attributed to the slightly younger age of the mice.

Table 5.

Viremia in 5- to 6-day-old CD1 mice after SC infection with WT or mutant CHIKV strain LR or vaccine strain 181/25

Virus straina	Day 1 postinfection		Day 2 postinfection		Day 3 postinfection		Day 4 postinfection
	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD
LR	3/4	3.3 ± 2.0	4/4	3.7 ± 0.9	4/4	2.6 ± 1.0	2/4	2.2 ± 1.6
LR/4x	4/4	4.1 ± 1.0	4/4	3.2 ± 0.4	4/4	3.7 ± 1.5	2/4	1.9 ± 1.2
181/25	3/3	4.3 ± 0.2	1/3	1.5 ± 1.0	0/3	≤ 0.9	1/3	1.5 ± 1.0

^aSee Fig. 1b for strain designations.

^bTiters are in log₁₀ PFU/ml of serum, and the limit of detection of the assay was 0.9 log₁₀ PFU/ml of serum.

The 6K protein mutation was viewed as unlikely to be involved in attenuation; this protein is not known to affect viral replication, it only acts as the signal peptide for the E1 glycoprotein, and only small amounts of 6K remain associated with the viral particle (20). Indeed, incorporation of the 6K protein mutation in the LR strain backbone, in the form of either LR/6K or LR/4x, did not produce any detectable attenuation. For these reasons, the 6K single mutation was not tested further.

Infection of A129 mice with CHIKV LR mutants.

Another mouse model of CHIKV infection is the A129 inbred mouse (6). This strain has impaired type I interferon signaling due to inactive alpha/beta interferon receptors. Cohorts (n = 4 or 5) of 8- to 10-week-old A129 mice were infected ID in the footpad with 10⁴ PFU of vaccine strain 181/25 or the four single mutants (Fig. 1b) and monitored for morbidity and mortality. All 181/25-infected mice survived to day 18 postinfection, when the experiment was terminated. However, the animals that received WT LR and the LR/nsP1 and LR/E1 mutants succumbed to infection by day 4 postinfection (Fig. 3). The LR/E2-12 and LR/E2-82 mutant-infected mice had significantly extended survival times (P < 0.01, Mantel-Cox test), but all died by day 7 postinfection. All six groups of mice gradually lost weight until day 7. However, mice infected with vaccine strain 181/25 began to recover after that time point (Fig. 4a). Morbidity was also reflected in body temperature changes; in moribund mice, there was a slight hyperthermia for the first 1 to 3 days postinfection, followed by distinct hypothermia 1 to 2 days before death (P < 0.05 on day 6 and P < 0.001 on day 7 for the E2 single mutants, one-way analysis of variance [ANOVA] with a Tukey-Kramer posttest, compared to 181/25) (Fig. 4b). Slight hyperthermia in 181/25-infected mice was sustained until day 5 postinfection, and then body temperatures returned to normal (Fig. 4b).

Fig 3.

Survival of 8- to 10-week-old A129 mice (n = 4 or 5) after ID infection with 10⁴ PFU of CHIKV strain LR, mutants, or vaccine strain 181/25. Animals were monitored for 18 days.

Fig 4.

Morbidity in 8- to 10-week-old A129 mice (n = 4 or 5) after ID infection with 10⁴ PFU of CHIKV strain LR, mutants, or vaccine strain 181/25. (a) Body weight changes. (b) Body temperature changes. (c) Footpad swelling at day 2 postinfection. Asterisks indicate P values (determined by one-way ANOVA with a Tukey-Kramer posttest) of differences from 181/25 or PBS control groups. *, P < 0.05; **, P < 0.01.

Swelling of the footpad, the location of ID injection, is another indication of CHIK disease in murine models. Measurement of the perimetatarsal height of the infected limb is used to assess swelling (10). Such measurements taken at day 2 postinfection indicated significant (P < 0.05 or less by one-way ANOVA with a Tukey-Kramer posttest) swelling induced by mutants LR/nsP1 and LR/E1 compared to 181/25- or sham-infected mice (Fig. 4c). Footpad swelling in the WT LR-infected mice was also overt. However, because of one animal with an outlier value, it was not significantly different by one-way ANOVA compared to 181/25-infected mice. Animals infected with the LR/E2-82 or LR/E2-12 mutant did not exhibit significant swelling (P > 0.05 by one-way ANOVA) compared to controls. Analysis of virus content in the feet should be examined in future CHIK pathogenesis experiments.

A129 mice were bled daily for the first 4 days postinfection, and viremia was determined using qRT-PCR due to the small volume of blood collected. Mice infected with the LR/nsP1 and LR/E1 mutants consistently showed a viremia that did not differ significantly (P > 0.05, one-way ANOVA, except on day 1 postinfection), up to 6.8 log₁₀ PFU/ml for LR/E1 on day 3. The LR/E2-82 mutant- and 181/25-infected mice showed lower and similar viremia levels (P > 0.05, one-way ANOVA), except on day 4 postinfection, while the LR/E2-12 mutant-infected mice developed intermediate viremia titers (Table 6).

Table 6.

Viremia in 10-week-old A129 mice after ID infection with a CHIKV strain LR mutant or vaccine strain 181/25

Virus straina	Day 1 postinfection		Day 2 postinfection		Day 3 postinfection		Day 4 postinfection
	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD
LR/nsP1	5/5	2.9 ± 0.8	5/5	5.2 ± 0.3	4/4	6.7 ± 0.4	NSc	NS
LR/E1	5/5	4.1 ± 0.2	5/5	6.3 ± 1.0	2/2	6.8 ± 0.2	NS	NS
LR/E2-12	5/5	2.5 ± 0.7	5/5	4.2 ± 0.6	5/5	4.7 ± 0.8	5/5	4.6 ± 0.4
LR/E2-82	2/5	0.2 ± 0.1	5/5	2.4 ± 0.4	5/5	3.1 ± 0.4	5/5	3.8 ± 0.2
181/25	4/4	0.5 ± 0.4	4/4	3.3 ± 0.4	4/4	2.9 ± 0.4	4/4	2.6 ± 0.2

^aSee Fig. 1b for strain designations.

^bTiters are in log₁₀ PFU/ml of serum, and the limit of detection of the assay was 0.1 log₁₀ PFU/ml of serum.

^cNS, no survivors.

In summary, these A129 mouse experiments revealed partial attenuation of CHIKV caused by expression of the E2-12 and E2-82 mutations in the LR strain background. The LR/nsP1 and LR/E1 mutants showed little or no attenuation and exhibited the virulent WT phenotype.

Construction of the CHIKV AF15561 infectious clone and mutants thereof.

Different CHIKV strains have been shown to differ in their responses to mutations that can enhance mosquito vector infectivity (46, 48). Therefore, to eliminate the possibility that the CHIKV genomic background affected the expression of the attenuation phenotype of the 181/25 strain mutations, we constructed an infectious cDNA clone of strain AF15561, the parental strain of 181/25. This strain shows 92.9% nucleotide and 97.7% amino acid identity with the LR strain. The AF15561 strain sequence was retrieved from GenBank (accession no. EF452493), furnished with the 5′ SP6 RNA polymerase promoter and a 3′ poly(A) tail, and divided into five fragments according to the availability of convenient restriction sites. These five DNA fragments were synthesized in vitro and ligated together in two steps using the cloning vector plasmid fragment from the CHIKV LR infectious clone. The GenBank sequence of the AF15561 strain contains 4 ambiguous nucleotides that do not affect the amino acid sequence, and these were assigned in the clone, based on the 181/25 sequence, as follows: K₉→G, Y₄₁₂→T, Y₁₆₁₂→C, and Y₆₉₉₄→T (nucleotide numbering is from the 5′ end of the CHIKV genome). Electroporation of Vero cells with in vitro-synthesized RNA of strain AF15561 caused CPE within 24 h, and the infectivity of this RNA was 1.4 × 10⁵ PFU/4 μg. The viral stock titer was 1.5 × 10⁷ PFU/ml and produced medium-size plaques on Vero cells. These data indicated that the AF15561 virus generated was viable and did not require compensatory mutations to be rescued in Vero cells.

The AF15561 clone was mutated to incorporate the 181/25-specific amino acid substitutions (see clone designations in Fig. 5): E2-12 and E2-82 were mutated individually and together in one clone, and the remaining three nonsynonymous mutations (nsP1, 6K, and E1) were included in one construct. Viral stocks of the mutants were generated in Vero cells as described above, with titers of 6 to 7 log₁₀ PFU/ml (data not shown). As was observed with the LR/E2-82 virus, the 15561/E2-82 mutant produced smaller plaques on Vero cells.

Infection of infant CD1 mice with CHIKV AF15561 mutants.

Cohorts (n = 36) of 5- to 6-day-old CD1 mice were infected SC with 10⁵ PFU of the 15561/3x, 15561/E2-12, 15561/E2-82, or 15561/2xE2 mutant, WT CHIKV AF15561, and vaccine strain 181/25 to test the attenuation effects of the vaccine-specific substitutions in the background of the parental AF15561 genome. Six mice from each group were sacrificed on days 1 to 5 and 7 postinfection to harvest hind limb and brain tissues and collect blood. The AF15561 parent virus and the 15561/3x mutant produced the highest mean viremias over the course of the experiment, which did not differ significantly (P > 0.05, one-way ANOVA), reaching 5 to 6 log₁₀ PFU/ml on days 1 and 2 postinfection, and were not cleared from the bloodstream until day 7 (see Table S1 in the supplemental material). From day 3 postinfection, the number of viremic animals began to decline, and on day 5, no detectable viremia was present in mice infected with 181/25 or with the single and double E2 mutants. The mean titers were similar (P > 0.05, one-way ANOVA, except for 15561/E2-12 on day 1) and 100- to 1,000-fold lower than in WT strain 15561-infected mice on days 1 to 4 after infection.

All six groups of mice had similarly high viral loads in hind limb tissue up to day 4 postinfection, with mean peak titers of 6 to 7 log₁₀ PFU/g on day 2 (see Table S2 in the supplemental material). On days 5 and 7 postinfection, there was a decline in the virus titers and the number of positive animals. The absence of a difference in viral loads in hind limb tissue of mice infected with the vaccine strain versus those infected with WT CHIKV suggests that this phenotype is not a good indicator of attenuation.

Viral loads in the brains of mice infected with WT AF15561, 15561/3x, and 15561/E2-12 were the highest and similar to each other (P > 0.05, one-way ANOVA) during the time course of the experiment (see Table S3 in the supplemental material). In contrast, 181/25- and 15561/2xE2-infected mice had the lowest titers, which did not differ significantly (P > 0.05, one-way ANOVA), mostly having zero to two positive animals out of six. The 15561/E2-82 mutant generated intermediate viral loads in the brain.

In summary, infection of infant CD1 mice with AF15561-derived viruses demonstrated a general trend of the 15561/3x mutant exhibiting a phenotype that was similar to WT AF15561 virus and the 15561/2xE2 mutant resembling that of the 181/25 vaccine strain, with single E2 mutants showing intermediate levels of virulence.

Infection of A129 mice with CHIKV AF15561 mutants.

To study the combined effect of the E2-12 and E2-82 mutations using adult mice, we infected cohorts (n = 9) of 5- to 8-week-old A129 mice ID with 10⁴ PFU of the 15561/2xE2, 15561/3x, 15561/E2-12, and 15561/E2-82 mutants, as well as WT CHIKV strain AF15561 and derived vaccine strain 181/25. All of 15561/2xE2- and vaccine strain-infected mice survived, but animals infected with the WT and 15561/3x viruses began to die on day 4 after infection and succumbed to infection by day 8, with no difference from the WT strain (P > 0.05, Mantel-Cox test) (Fig. 6). The 15561/E2-82-infected mice died by day 9, and 56% of the 15561/E2-12-infected mice survived, with the last death on day 9 postinfection.

Fig 6.

Survival of 5- to 8-week-old A129 mice (n = 9) after ID infection with 10⁴ PFU of CHIKV strain AF15561, mutants, or vaccine strain 181/25. Animals were monitored for 14 days.

Body weights declined gradually until death in mice infected with the WT, 15561/3x, and 15561/E2-82 viruses, and there was no significant difference between 15561/3x and WT (P > 0.05, one-way ANOVA). The 15561/E2-12-infected mice also lost weight until day 9 postinfection, and then the surviving mice began to gain weight (Fig. 7a). The 15561/2xE2- and vaccine strain-infected mice exhibited a slight weight loss until days 7 and 9, respectively, and then recovered, with no difference from those infected with the 181/25 strain (P > 0.05, one-way ANOVA), except on day 9 postinfection, when there was an increase in the mean weight of the 15561/2xE2-infected cohort.

Fig 7.

Morbidity in 5- to 8-week-old A129 mice (n = 9) after ID infection with 10⁴ PFU of CHIKV strain AF15561, mutants, or vaccine strain 181/25. (a) Body weight changes. (b) Body temperature changes. (c) Footpad swelling. ***, P < 0.001 (determined by one-way ANOVA with a Tukey-Kramer posttest) compared to the 181/25 group. p.i., postinfection.

Patterns of body temperature changes were nearly identical to those produced by the LR-derived CHIKV strains. In mice that suffered fatal infections, there was slight hyperthermia during the first few days, followed by hypothermia preceding death (Fig. 7b). The 15561/2xE2- and 181/25-infected mice had the same pattern (P > 0.05, one-way ANOVA) of slight hyperthermia until day 5 postinfection as the 181/25 cohort, followed by a return to normal values.

The height of the injected footpad was measured on days 2, 4, and 7 postinfection. Footpad swelling was more severe on day 2 or 4 in mice infected with the WT, 15561/3x, and 15561/E2-82 viruses than in 181/25-infected mice by one-way ANOVA with a Tukey-Kramer posttest (Fig. 7c). Footpad swelling peaked on day 2 after infection in WT- and 15561/3x-infected mice and then gradually decreased. Generation of virulent revertants (see below) by day 4 postinfection might explain the sharp peak in footpad swelling in 15561/E2-82-infected mice. The mean footpad heights of 15561/E2-12- and 15561/2xE2-infected mice were not significantly different (P > 0.05) from those of vaccine strain-infected mice on all of the days tested.

Viremia was evaluated by plaque assay on days 1 and 3 postinfection. Most of the infected mice showed viremia, with AF15561 and 15561/3x having the highest titers, about 5 log₁₀ PFU/ml on day 1 and about 8 log₁₀ PFU/ml on day 3 (Table 7), which did not differ significantly (P > 0.05, one-way ANOVA). Mice infected with the 15561/2xE2 and 181/25 strains had 100- to 1,000-fold lower viremia on days 1 and 3, respectively, with no significant differences (P > 0.05, one-way ANOVA). Intermediate levels were observed in mice infected with single E2 mutants, especially on day 3 postinfection.

Table 7.

Viremia in 5- to 8-week-old A129 mice after ID infection with WT or mutant CHIKV strain AF15561 or vaccine strain 181/25

Virus straina	Day 1 postinfection		Day 3 postinfection
	No. of animals positive/total no. inoculated	Mean titerb of positive animals ± SD	No. of animals positive	Mean titerb of positive animals ± SD
15561	8/9	5.0 ± 0.8	9/9	8.2 ± 1.0
15561/3x	7/9	4.6 ± 2.2	9/9	7.7 ± 0.9
15561/E2-82	8/9	3.3 ± 1.1	9/9	7.0 ± 0.8
15561/E2-12	8/9	2.9 ± 0.9	9/9	6.5 ± 1.3
15561/2xE2	7/9	2.8 ± 1.2	8/9	4.9 ± 1.5
181/25	7/9	2.5 ± 1.0	9/9	5.0 ± 0.5

^aSee Fig. 5 for strain designations.

^bTiters are in log₁₀ PFU/ml of serum, and the limit of detection of the assay was 0.9 log₁₀ PFU/ml of serum.

In summary, infections of A129 mice with AF15561-derived viruses demonstrated that simultaneous expression of E2 substitutions at amino acid positions 12 and 82 is necessary and sufficient to recapitulate the attenuation phenotype of the 181/25 vaccine strain, as shown by the survival of infected mice, body weight and temperature changes, footpad swelling, and viremia.

Sequencing of 181/25 viral stock and E2 mutants isolates.

To confirm that our stock of the 181/25 vaccine strain corresponded to the sequence published in GenBank, we extracted RNA from 0.2 ml of the stock and performed RT-PCR to produce six overlapping amplicons covering the full genome of the virus. Direct sequencing of the amplicons revealed no amino acid differences. However, nucleotide position 978 contained a mixture of 181/25-specific T and AF15561-specific C, indicating a mixed threonine/isoleucine nsP1-301 population (Table 1). The lack of predominance of the isoleucine residue in the 181/25 population we tested further suggests that it is not an important attenuation determinant. However, we cannot rule out the possibility that this mutation exerts an effect on the quasispecies population even though it is not present in the majority of CHIKV genomes. Two other nucleotide changes were also found in our 181/25 stock, both synonymous: R₃₇₀₆→A and T₆₀₄₃→Y. These mutations may have occurred during additional cell culture passages of the 181/25 stock available in our laboratory.

Since our investigation indicated the importance of the E2-12 and E2-82 substitutions in the attenuation of strain 181/25, we looked for evidence of reversion in infected mice. Blood samples collected on day 3 postinfection from A129 mice injected with the 15561/E2-12, 15561/E2-82, and 15561/2xE2 viruses were used to extract total RNA and amplify genome fragments covering both of these mutations via RT-PCR. The resultant amplicons were sequenced directly, and the presence of revertants was investigated by examining sequence electropherogram peaks. All eight samples of the 15561/2xE2 mutant showed the presence of both E2 mutations, as indicated by clear nucleotide peaks. Two of the eight strain 15561/E2-12 blood samples and all of the nine strain 15561/E2-82 samples had partial reversions to the WT C at position 8576 or WT G at position 8785, respectively (presence of mixed nucleotide peaks). These data indicate that the WT sequences were selected in vivo, partially explaining the high-virulence phenotype of these two mutants in the A129 mice model, as well as the inadequate attenuation stability suggested by previous human trials of strain 181/25 (8).

DISCUSSION

For decades, the generation of live-attenuated viral vaccine strains by passaging virulent WT viruses in cell culture has been the mainstay of vaccine development (2, 44). These passages are usually accompanied by the acquisition of several adaptive mutations in the viral genome, with one or a few of these mutations responsible for the attenuated phenotype that is desired in a vaccine strain (12, 18, 49). This cell culture virus-passaging procedure was utilized by Levitt et al. to produce the 181/25 CHIKV vaccine strain to protect against CHIK (21), a debilitating illness that has affected millions of people during recent outbreaks (41). The vaccine exhibited reduced virulence and proved to be highly immunogenic. However, it caused transient arthralgia in 8% of the vaccinees (8). Sequence comparisons of CHIKV strains 181/25 and AF15561, the vaccine's parent, identified five synonymous and five nonsynonymous mutations, all in ORFs of the genome (Table 1).

We sought to identify the attenuation determinants of strain 181/25 to understand its reactogenicity in humans and to better understand the limitations of attenuation based on cell culture adaptation. Using a reverse genetic approach, we generated an array of CHIKV mutants containing different combinations of 181/25-specific amino acid substitutions in the genetic backbone of two WT CHIKV strains and tested their levels of attenuation in two mouse models.

The first set of mutants was based on the CHIKV LR genome (Fig. 1b) derived from a French patient returning from La Reunion Island during the 2005 epidemic (45). For a more accurate assessment of the significance of the 181/25 mutations—in their WT genomic background—we also assembled an infectious clone of the vaccine's parent strain, AF15561 (21), and used it to generate the second set of comparable mutants (Fig. 5). The viruses were tested in infant outbred CD1 mice, a model of transient CHIKV infection that resembles human disease histopathologically (56). The mutant, WT, and vaccine strains were compared for the generation of viremia and viral loads in hind limb and brain tissues. Body weight change proved to be uninformative in this model. Inbred A129 mice, which lack type I interferon receptors and are sensitive to differences in CHIKV virulence (6, 29, 32), were also used and monitored for survival, body weight and temperature, footpad swelling, and viremia. Strain 181/25 infection was well tolerated by both mouse strains and ages, and the animals survived with only transient signs of illness (29, 32).

We found no evidence that vaccine-specific amino acid substitutions in the nsP1, 6K, or E1 proteins, when expressed individually or together, mediated attenuation. By all measures, these constructs exhibit virulence comparable to that of WT viruses, whether expressed in the LR or in the AF15561 genomic background. We also tested another construct with four combined 181/25-specific mutations, all except E2-82, and when infected with this mutant, CD1 mice developed a viremia similar to that produced by the WT. Sequencing of our 181/25 vaccine stock, which revealed a ca. 1:1 mixture of the vaccine strain- and WT-specific nucleotides at the nsP1 position, also suggested that it did not affect virulence.

The attenuated phenotype of the 181/25 vaccine strain was recapitulated by the 15561/2xE2 virus, which incorporated both E2 glycoprotein substitutions at amino acid positions 12 and 82. CD1 and A129 mice infected with this double mutant exhibited survival and viremia indistinguishable from those of mice infected with the vaccine strain, as well as all of the phenotypes tested. When these E2 mutations were expressed individually, each exhibited an intermediate level of attenuation in most experiments, with the E2-82 mutation demonstrating a higher degree of attenuation than E2-12 overall. Sequencing of isolates of the 15561/E2-12 and 15561/E2-82 viruses derived from the serum of A129 mice 3 days postinfection revealed a high rate of reversion in these single mutants: mixed mutant-WT populations were found in 9/9 mice infected with the E2-82 virus and in 2/8 infected with the E2-12 virus. The ability of these point mutations to rapidly revert to the WT may explain, at least in part, the intermediate-virulence phenotypes observed for these single E2 mutants. In contrast, the mouse isolates of the 15561/2xE2 double mutant showed no evidence of any reversion, underscoring the inherent stability advantage of multiple mutations. The stabilizing effect of multiple mutations in attenuation has been described for several other viruses, including Venezuelan equine encephalitis virus (VEEV) (7, 18), Ross River virus (25), Sindbis virus (SINV) (22), and poliovirus (27).

The importance of the Gly→Arg E2-82 mutation in the attenuation of CHIKV was not surprising, considering findings in other alphaviruses. Cell culture passage typically selects for an increased positive charge on the surface of alphavirus spikes composed principally of the E2 envelope glycoprotein (3, 7, 14, 19, 23). These changes typically increase the binding of alphaviruses to heparan sulfate and other glycosaminoglycans present at high concentrations on the cultured cell surface. However, they are believed to be attenuating in vivo because they prevent the spread of virus through tissues rich in negatively charged extracellular components. Based on the crystal structure of the mature CHIKV E3-E2-E1 glycoprotein complex (52), amino acid position 82 lies on the outer surface of E2 domain A (Fig. 8). Substitutions in the E2 protein of SINV responsible for heparan sulfate binding (4, 19) map to the same area, further suggesting that this substitution was selected during cell culture passage for glycosaminoglycan binding. The E2-12 position maps to the E2 beta ribbon and is buried between E2 domain A and the beta ribbon (Fig. 8). However, this polar-to-hydrophobic (Thr→Ile) mutation may affect the secondary structure of the beta ribbon and thereby affect nearby E2 domain A, possibly stabilizing the effect of the E2 arginine-82 substitution. These putative glycosaminoglycan binding properties of these mutations should be confirmed experimentally.

Fig 8.

Crystal structure of the mature envelope glycoprotein complex of CHIKV fitted into a cryoelectron microscopic reconstruction map of Western equine encephalitis virus glycoprotein spikes on the surface of the viral particle (39), prepared using UCSF Chimera software (30). (a) Cryoelectron microscopic map of Western equine encephalitis virus. (b) Top view of the three E2-E1 heterodimers forming the spike, magnified from a circle in panel a, with the CHIKV E2-E1 crystal structure fitted into one of the heterodimers. (c) Same spike with the CHIKV crystal structure as in panel b, rotated 90°. (d) Ribbon diagram of the CHIKV heterodimer (Molecular Modeling Database identification no. 86612, chains B and F [52]) in the same orientation as in panel b. E2 is yellow, and E1 is blue. E2 amino acid positions T12 and G82 are shown as red spheres. E2 positions 76 and 114 are residues described in the literature as responsible for heparan sulfate binding in SINV (4, 19) and are shown as magenta and cyan spheres, respectively.

Our work emphasizes the limitations of the traditional cell culture passaging method for vaccine development, which leads to virus attenuation based on either single or multiple point mutations alone. Due to the high rate of mutagenesis observed in RNA viruses, there is a high rate of reversion to WT sequences, restoring the virulence phenotype. The sequence of a virus isolate from a viremic human vaccinee who received the 181/25 vaccine, TSI-GSD-218-VR1 (GenBank accession no. EF452494), showed reversion to WT glycine at position 82 of the E2 glycoprotein, in addition to other, novel, mutations. Such instability might explain the occasional reactogenicity that was observed in 181/25 vaccine trials (8). A similar situation is observed with the TC-83 live-attenuated VEEV vaccine strain, which also depends on only two point mutations for its attenuation and also causes a high rate of adverse reactions in vaccinees (18, 31).

Recent progress in alphaviral genetics has led to the development of more complex and stable approaches for vaccine strain attenuation. Several alternative strategies based on directed manipulation of the alphaviral genome, such as chimeras consisting of two distinct alphaviruses or internal ribosome entry site-dependent translation of structural proteins, have proven to produce stably attenuated, safe, and effective strains in recent studies (17, 28, 32, 51, 53, 54).

In conclusion, our results indicate that attenuation of the CHIKV 181/25 vaccine strain is mediated by two mutations in the E2 glycoprotein, at amino acid positions 12 and 82. The presence of both mutations is required and sufficient to attain a well-attenuated phenotype. The identification of these two attenuating E2 mutations may be useful for the generation of other live-attenuated CHIK vaccines.