The Development of Resistance to Passive Therapy by a Potently Neutralizing Humanized West Nile Virus Monoclonal Antibody

Abstract

Previous studies have established the therapeutic efficacy of humanized E16 (hE16) monoclonal antibody against West Nile virus (WNV) in animals. Here, we assess the potential for WNV strains encoding mutations in the hE16 epitope to resist passive immunotherapy and for the selection of neutralization escape variants during hE16 treatment. Resistance to hE16 in vivo was less common than expected, as several mutations that affected neutralization in vitro did not significantly affect protection in mice. Moreover, emergence of resistant variants after infection with fully sensitive virus occurred but was relatively rare, even in highly immunocompromised B and T cell-deficient RAG mice.

West Nile virus (WNV) is a mosquito-borne flavivirus that has emerged as the primary cause of epidemic encephalitis in the United States. Although case numbers have declined somewhat since the peak activity of 2002 and 2003, virus transmission continues throughout the continental United States and thousands of cases of neurological disease have been reported each year since 2004. At present there is no WNV vaccine or antiviral therapy approved for use in humans, although promising treatment results were observed in case reports using purified immunoglobulin derived from Israeli donors [1, 2].

The potential of passive immunotherapy for treatment of WNV infection has led to the development and evaluation of potently neutralizing human or humanized mouse monoclonal antibodies (mAbs) [3, 4]. One therapeutic candidate, hE16, binds to a highly conserved epitope on the upper lateral ridge of WNV envelope (E) protein domain III (E-DIII) and neutralizes WNV at low stoichiometric occupancy apparently by inhibiting conformational changes in E that are required for fusion of the virus with host cell membranes [3, 5, 6]. However, mapping studies by several groups have identified mutations in the hE16 epitope that significantly reduce or abolish binding of this and other neutralizing WNV mAbs [3, 7–9]. A few of these mutations are found in WNV strains isolated in the field [8] suggesting that the efficacy of hE16 could be limited. Furthermore, flaviviruses have an error-prone genome replication, which results in significant genetic diversity within any individual isolate. Indeed, this property has been used as the basis for in vitro selection of variants resistant to a particular selective pressure, such as that imposed by antiviral inhibitors or neutralizing mAbs. The potential for in vivo selection of mAb-resistant variants of flaviviruses has not yet been examined in detail.

In this study, we assessed (1) the potential for WNV strains encoding engineered mutations in the hE16 binding site to resist passive immunotherapy with hE16 in two mouse models of WNV neuroinvasive disease, and (2) the potential for neutralization resistant variants to be selected in vivo during hE16 treatment.

Methods

All WNV strains and infectious clone-derived variants used in this study (Table 1 and Results) were grown and plaque titrated on Vero cells. Neutralization assays were performed on Vero or BHK-21 cells, as previously described [3]. RNA extractions, RT-PCR and nucleotide sequencing of the pre-membrane (prM) and E coding regions of WNV genomes were performed using protocols and primers as previously described [7].

Table 1.

Outcomes of infections with West Nile virus strains/variants following pre- or post-exposure treatment with neutralizing mAb hE16.

			hE16 treated		PBS only
Mouse strain	hE16 Treatment*	Virus	% survivors	AST±sd†	% survivors	AST±sd	Survival P value‡
Swiss Webster	100µg day −1	NY99 wt	100	n/a	20	8.4±1.2	0.0003
		NY99ic T332A	100	n/a	10	8.1±0.3	<0.0001
		NY99ic T332M	90	8.0±0.0	0	8.4±0.8	<0.0001
		NY99ic T332K	30	7.4±0.8	10	8.8±1.5	0.8984
		NY99ic T330I	90	8.0±0.0	0	7.8±0.4	0.0008
		SA58	0	7.2±0.4	0	7.5±0.5	0.1704
		SA58 K332T	100	n/a	0	7.5±0.7	<0.0001
C57BL/6	100µg day +2	NY99ic wt	100	n/a	30	11.0±1.7	<0.0001
		NY99ic T332A	100	n/a	20	9.3±1.9	0.0003
		NY99ic T332M	100	n/a	30	10.7±1.6	<0.0001
		NY99ic T330I	100	n/a	10	9.4±1.8	<0.0001
		SA58	10	10.8±1.6	0	9.3±1.1	0.0207
		SA58 K332T	100	n/a	10	10.2±1.4	<0.0001
Swiss Webster	100µg day −1	NY99 K307E	10	8.8±1.3	0	7.8±0.4	0.0773
		NY99 T332M	70	13.3±3.8	0	8.0±0.7	0.0001

^*Swiss Webster mice were treated with 100µg hE16 or PBS only, then challenged 1 day later with 100 PFU of indicated WNV strains (1000 pfu for NY99ic T332K). C57BL/6 mice were inoculated with 100 PFU of indicated WNV strains, then treated 2 days later with 100 µg hE16 or with PBS only. n=10 for each virus/treatment group, except for NY99 K307E/T332M “PBS only” groups, where n=5.

^†AST – average survival time; sd – standard deviation; n/a – not applicable

^‡Significant differences in survival curves for hE16 treated vs. PBS only groups determined by log rank test using the Prism application (Graphpad Software).

Swiss Webster mice (female, 3–4 weeks of age) were obtained from Harlan Laboratories; C57BL/6 wild type and congenic RAG mice (female, 5 weeks of age) were obtained from The Jackson Laboratory. All mice were housed in AAALAC accredited animal biosafety level 3 (ABSL3) facilities and experiments were conducted under protocols approved by the Animal Care and Use Committee of the University of Texas Medical Branch or Washington University School of Medicine. Details of individual passive protection experiments are described below.

Results

Previous crystallographic and epitope mapping studies suggested that hE16 had key contacts at residues 307, 330, and 332 of the WNV E protein [3, 5]. The ability of hE16 to neutralize selected WNV strains and NY99 infectious clone-derived variants encoding single amino acid changes at these residues was initially assessed by a plaque reduction neutralization assay on Vero cells. These viruses were shown previously to variably escape neutralization by other anti-WNV E-DIII specific neutralizing mAbs [7, 8]. Notably, only a mutation T332K resulted in substantial loss of hE16 neutralizing activity, whereas other mutations (K307R, T330I, T332A/M) showed only modest changes in neutralization compared to the wild-type lineage 1 NY99 virus (Figure 1a). Lineage 2 South African strain H442 (SA58), isolated in 1958 from a human patient, normally encodes a lysine at residue 332 and was previously reported to be resistant to neutralization by several E-DIII-reactive antibodies raised against NY99 [7]; this virus was also resistant to neutralization by hE16, whereas an SA58 variant encoding threonine at 332 [7] was efficiently neutralized (Figure 1b).

Figure 1.

Neutralization by mAb hE16 of: (a) lineage 1 WNV strain NY99ic and K307R, T330I and T332A/K/M variants; (b) lineage 2 WNV strain SA58 and K332T variant; and (c) K307E escape variants selected in vivo from hE16-treated RAG mice. The data are an average of two to four independent experiments performed in triplicate on (a, b) Vero or (c) BHK21–15 cells.

Two independent mouse challenge models were employed to assess the protection provided by hE16 against the neutralization sensitive and resistant WNV strains and variants. Groups of outbred Swiss Webster mice, which are highly susceptible to peripheral challenge with neuroinvasive WNV strains, were pre-treated with 100 µg doses of hE16 or PBS only and challenged 24 hours later with 10² plaque forming units (PFU) of each WNV strain/variant (equivalent to approximately 100 LD50s in each case). Alternatively, groups of inbred C57BL/6 mice, which are more resistant to WNV NY99 and have been used in previous evaluations of hE16 [3, 10] were challenged with 10² PFU of each virus and treated at two days post-infection with 100 µg hE16 or PBS only. The hE16 mAb afforded significant protection (90–100% survival) for mice in either the pre- or post-exposure model against NY99 and variants that were efficiently neutralized in vitro by hE16 (T330I, T332A/M). However, little or no protection was observed after challenge with the NY99ic T332K mutant or SA58 (Table 1).

Although mutagenesis of an infectious clone or in vitro selection can readily generate resistant viruses against hE16 and other mAbs, we questioned whether this occurred commonly under selective pressure during the course of treatment. To assess the potential for selection of resistant variants in vivo during mAb treatment, Swiss Webster mice pre-treated with 100 µg hE16 were exposed to a high dose (10⁴ PFU) of NY99 virus. This challenge dose was chosen based on the known frequency of mAb escape variants selected from this virus population during earlier studies [7] and because this represents a virus dose reliably introduced during feeding by Culex spp. mosquitoes [11]. All untreated control animals succumbed by day 9 post-infection (average survival time 7.4±1.0 days). Two of ten treated mice showed signs of neuroinvasive disease and were euthanized on days 8 and 9. Virus was isolated from the brain of each mouse and these encoded single nucleotide changes in E coding for mutations K307E and T332M, respectively. Both retained the highly neuroinvasive phenotype of the parental NY99 virus (data not shown).

Immune dysfunction is a significant risk factor for development of severe WNV disease [reviewed in 12], suggesting that therapeutic antibodies and other treatments may be especially important for immunodeficient patients. To assess the potential for emergence of resistant variants during treatment of an immunocompromised host, a group of 30 B and T cell-deficient RAG mice was exposed to a low dose (10² PFU) of strain WNV-NY2000 (3000-0259), which has an E gene sequence identical to that of NY99 [13], and then treated with 500 µg doses of hE16 administered one day post-infection and at 14 day intervals thereafter. Whereas 24 mice remained healthy over a period of several months, six mice became ill and were found dead or were euthanized on days 18 (2 mice), 20 (2 mice), 32 and 34. Brains recovered from the latter two animals each yielded WNV encoding a K307E E protein mutation.

Consistent with previous studies [3], the recovered K307E variants were resistant to in vitro neutralization (Figure 1c) or in vivo protection (Table 1) by hE16, whereas the recovered T332M variant was only modestly resistant to E16 neutralization in vitro (not shown) or in vivo (Table 1), similar to the NY99ic T332M variant.

Discussion

The use of polyclonal or monoclonal immunotherapy is currently being explored for treatment of infections caused by many families of viruses and clearly offers a promising approach for post-exposure treatment of flavivirus infections. In the case of hE16 for treatment of WNV infections, the presence of naturally occurring mutations in some WNV strains at the mAb binding site suggests that treatment coverage will not be universal, as evidenced by resistance of lineage 2 strain SA58 in vitro and in vivo. Nonetheless, hE16 still afforded strong protection against variants encoding other mutations that increased resistance to neutralization but did not abolish binding. Analysis of more than 650 derived amino acid sequences in the NCBI Protein database representing human, avian, equine or mosquito North American WNV isolates during 1999–2005 identified none with amino acid variation at positions 307, 330, or 332, which could impair hE16 recognition and neutralization.

These studies have also demonstrated the potential for in vivo selection of mAb escape variants. Although surprisingly infrequent, in vivo resistance may occur via selection of a pre-existing mutant subpopulation following exposure to higher doses of virus, a phenomenon which has also recently been reported in passively immunized monkeys exposed to 10⁶ PFU doses of dengue virus type 2 [14]. Alternatively, resistance can develop through the emergence of variant viruses during prolonged treatment.

The length of time required for the emergence of the resistant K307E variants following 10² PFU challenge in RAG mice (up to 34 days) compared with the rapid selection of variants from the 10⁴ PFU challenge in Swiss Webster mice (8–9 days), was somewhat surprising. Based on the inherent error rate of the flavivirus RNA polymerase (approximately one mutation per genome replication), the occurrence of resistant mutations at the genetic level should be expected to occur rapidly. The delay in emergence of resistance in RAG mice could be explained by packaging of early variant genomes in virions that also contain sufficient levels of wild-type E protein that allow neutralization by hE16. This is possible because flavivirus RNA can be encapsidated by structural proteins in trans [15] and complete neutralization (>99%) of WNV virions by hE16 requires occupancy of less than 50% of the available epitopes [6]. Although further studies are necessary, high levels of circulating hE16 with a long half-life may adequately neutralize virions comprised of wild type and resistant E protein variants to prevent rapid or eventual emergence of pathogenic escape variants, even in a highly immunocompromised subject.

Overall, these findings suggest that resistance to neutralizing anti-WNV antibody therapy can occur in vivo, although at frequencies that are perhaps lower than anticipated based on analogous studies on resistance to antiviral drugs that target virus replication. However, our data suggest that clinical evaluation of hE16 and other mAb therapeutics for flavivirus infections should monitor for selection or emergence of resistant variants, especially in immunocompromised patients who are a likely target population for treatment of severe WNV infections. Such studies should include characterization of viral populations for detection of resistance and possibly, in some patients, the use of combinations of neutralizing antibodies that recognize distinct epitopes.