Interaction of Yellow Fever Virus French Neurotropic Vaccine Strain with Monkey Brain: Characterization of Monkey Brain Membrane Receptor Escape Variants

Abstract

Binding of yellow fever virus wild-type strains Asibi and French viscerotropic virus and vaccine strains 17D and FNV to monkey brain and monkey liver cell membrane receptor preparations (MRPs) was investigated. Only FNV bound to monkey brain MRPs, while French viscerotropic virus, Asibi, and FNV all bound to monkey liver MRPs. Four monkey brain and two mouse brain MRP escape (MRP^R) variants of FNV were selected at pH 7.6 and 6.0. Three monkey brain MRP^R variants selected at pH 7.6 each had only one amino acid substitution in the envelope (E) protein in domain II (E-237, E-260, or E274) and were significantly attenuated in mice following intracerebral inoculation. Two of the variants were tested in monkeys and retained parental neurotropism following intracerebral inoculation at the dose tested. We speculate that this region of domain II is involved in binding of FNV E protein to monkey brain and is, in part, responsible for the enhanced neurotropism of FNV for monkeys. A monkey brain MRP^R variant selected at pH 6.0 and two mouse brain MRP^R variants selected at pH 7.6 were less attenuated in mice, and each had an amino acid substitution in the transmembrane region of the E protein (E-457 or E-458).

Wild-type yellow fever (YF) virus, the prototype member of the Flavivirus genus (family Flaviviridae), characteristically causes hepatitis and hemorrhagic fever (i.e., viscerotropic disease) and not neurotropic disease in humans and monkeys. YF is controlled by the use of live attenuated vaccines, known as 17D and FNV (for French neurotropic vaccine) (see reference 1 for a review), which demonstrate various degrees of neurotropism, but not viscerotropism, in humans and monkeys.

The attenuated FNV strain was derived by passage of wild-type strain French viscerotropic virus (FVV) in mouse brain (6). Although the FNV strain was attenuated for viscerotropism, it was found to have enhanced neurotropic properties, including lethality for monkeys following intracerebral inoculation and a high incidence of postvaccination complications in children under the age of 11 years (3). In comparison, the 17D vaccine strain (derived by passage of wild-type strain Asibi in chicken tissue) was found to be attenuated for both viscerotropism and neurotropism to the extent that 17D virus rarely kills monkeys following intracerebral inoculation, and the incidence of postvaccination complications is so low that the 17D vaccine can safely be given to 1-year-old children. The more attenuated phenotype of 17D relative to FNV resulted in 17D virus replacing FNV virus as the vaccine of choice to control YF. The molecular basis of the enhanced neurotropism and the cell receptor(s) for FNV are still unknown and are the subject of this communication.

The first step of a virus infection is attachment to the host cell (5). Binding of the virion to specific cellular receptor-binding site molecules is mediated by the viral spike protein(s), which in the case of the flaviviruses is the envelope (E) protein. Recent studies have elucidated the three-dimensional structure of a 50-kDa fragment of the ectodomain of the E protein of Central European tick-borne encephalitis (TBE) virus and assigned three domains (termed, I, II, and III) to the protein structure (8).

We have previously reported a methodology for investigation of the interaction of the flavivirus Japanese encephalitis (JE) virus with mouse brain cells by selection of mouse brain membrane receptor preparation (MRP) binding escape variants (MRP^R) and used this technique to identify potential E protein amino acids of JE virus that interact with mouse brain cell binding sites (7). This study proposed that E-306 in domain III was involved in the interaction of the E protein with the binding site in mouse brain cells.

This methodology (described in detail in reference 7) has now been used to investigate the interaction of FNV with monkey brain cells. Briefly, the brain and liver of a cynomolgus monkey and the brains of NIH Swiss mice were dissected, weighed, and then homogenized in 50 mM Tris buffer, pH 7.6. The homogenates were centrifuged at 35,600 × g for 10 min to obtain MRPs. The pellets were resuspended in Tris buffer, and the process was repeated twice. The final pellets were resuspended in Tris buffer at a final protein concentration of 20 to 40 mg (wet weight) of brain or liver tissue/ml and stored at −70°C. MRPs from monkey brain, monkey liver, and mouse brain will be referred to as MKB, MKL, and MS MRPs, respectively.

The specificity of binding of FNV to MKB MRPs was determined by comparison of wild-type and vaccine strains of YF virus with other flaviviruses, namely, JE virus strain P3 and dengue virus serotype 2 (DEN-2) strain New Guinea C (NGC). Neurotropic JE virus strain P3 bound to MKB MRP but bound less well to MKL MRP (Table 1), while nonneurotropic, nonviscerotropic DEN-2 NGC strain bound poorly to both MKB and MKL MRPs (i.e., there was a correlation with virus tissue specificity in vivo). Wild-type YF virus strains Asibi and FVV bound well to MKL MRP but bound poorly to MKB MRP, while, interestingly, vaccine strain 17D-204 bound poorly to both MKB and MKL MRPs. This may be indirectly involved in the attenuated phenotype of 17D virus. FNV was the only strain of YF virus to bind well to MKB MRPs.

TABLE 1.

Comparison of binding of YF, JE, and DEN-2 to MKB and MKL MRPs

Virus	Strain	Infectivitya (log10 PFU)		Binding indexb
		Virus + buffer	Virus + MRP
JE	P3	5.7	3.6 (MKB)	2.1 (MKB)
			5.0 (MKL)	0.7 (MKL)
DEN-2	NGC	5.8	5.5 (MKB)	0.3 (MKB)
			5.6 (MKL)	0.2 (MKL)
YF	FVV	5.6	5.0 (MKB)	0.6 (MKB)
			3.8 (MKL)	1.8 (MKL)
	FNV-Yale	6.5	2.3 (MKB)	4.2 (MKB)
			5.0 (MKL)	1.5 (MKL)
	Asibi	6.3	5.8 (MKB)	0.5 (MKB)
			5.0 (MKL)	1.3 (MKL)
	17D-204	6.8	6.3 (MKB)	0.5 (MKB)
			6.0 (MKL)	0.8 (MKL)

^aInfectious virus remaining in the supernatant following incubation of virus with MRP or buffer and pelleting of the virus-MRP complex as described previously (7). 

^bLog₁₀ reduction in infectivity titer (log₁₀ infectivity in buffer/infectivity in MRP supernatant). 

Four MKB MRP binding-resistant (MRP^R) variants were generated for FNV “strain” Yale (10) as described previously (7), except that MRPs were derived from the brain of a cynomolgus monkey. Briefly, MRP^R variants were generated by incubating virus with MRPs for 30 min at 37°C in 50 mM Tris, pH 7.6 (or pH 6.0 in some experiments) and then centrifuging at 13,000 × g for 10 min to remove MRP and bound virus. Residual infectious virus in the supernatant was titrated in Vero cell monolayers, and individual MRP^R variant plaques were picked and amplified in Vero cells. Putative MRP^R variants were examined for binding to MRPs, and lack of reduction in infectivity following incubation of the virus with fresh MRPs confirmed that they were true MRP^R variants. Using this procedure, three MKB MRP^R variants were selected in buffer at pH 7.6 and designated MKB MRP^R I, MKB MKRP^R II, and MKB MRP^R IV. A fourth variant was selected in 50 mM Tris buffer, pH 6.0, to investigate the potential effect of pH-induced conformational changes in the E protein; it was termed MKB MRP^R (pH 6.0). Two MRP^R variants, selected from FNV-Yale after incubation with mouse brain MRPs at pH 7.6, were designated MS MRP^R I and MS MRP^R II. The plaque morphology of the six MRP^R variants was indistinguishable from that of parental FNV. Similarly, there were no differences in growth characteristics or infectivity titers in Vero cell cultures.

FNV-Yale, which is known to be highly neurovirulent for mice after intracerebral inoculation (10), was found to have a Vero cell PFU/50% lethal dose (LD₅₀) ratio of 0.08 in 4-week-old female NIH-Swiss mice, while MKB MRP^R variant viruses selected at pH 7.6 were attenuated at least 400-fold (e.g., 32 PFU/LD₅₀ for MKB MRP^R II) (Table 2). MKB MRP^R (pH 6.0) was attenuated 87-fold compared to parent FNV-Yale virus and was less attenuated than MKB MRP^R variant viruses selected at pH 7.6. The two MS MRP^R variants were attenuated approximately 50- to 100-fold compared with parental FNV-Yale virus (Table 2). Two of the MKB MRP^R variant viruses selected at pH 7.6 were indistinguishable from the parental FNV-Yale when tested in monkeys by the World Health Organization neurovirulence test inoculation, dose, and scoring procedure (Table 2) (12), but just one monkey per preparation was tested. Sequencing of premembrane and E protein genes of parent and MRP^R variant viruses revealed that each variant had only a single nucleotide change, resulting in one amino acid substitution in the E protein, compared with the parental FNV-Yale virus. No changes in the M protein gene were identified. The MKB MRP^R (pH 6.0) variant had one amino acid substitution at E-458 (G→R), and the three MKB MRP^R variants selected at pH 7.6 each had a single amino acid substitution at E-260 (G→A), E-274 (Y→H), or E-237 (P→Y) for MKB MRP^R I, MKB MRP^R II, and MKB MRP^R IV, respectively. These substitutions were found to be unique; they have not been found in the E protein of any other YF virus strain sequenced to date (2, 11). The two MS MRP^R variants had identical amino acid substitutions at E-457 (M→I). Note that unlike the JE MS MRP^R variants reported previously (7), none of the YF MRP^R variants had amino acid substitutions in domain III of the E protein. This suggests that the interaction of YF and JE virus E proteins with MS MRPs may be different.

TABLE 2.

Comparison of mouse and monkey neurovirulence following intracerebral inoculation, and amino acid differences between FNV-Yale and MRP^R variants

Virus	PFU/mouse LD50	Mouse avg survival timeb (days ± SEM)	Amino acid substitution (parent→variant)	Monkey results
				Actual survival time (days)	Clinical scorea	Histology scorea
						Discriminator area	Discriminator plus target area
FNV-YALE	0.08	5.4 ± 0.4		5	3.5	1.75	2.72
MKB MRPR I (pH 7.6)	>100	NAc	E-260 (G→A)
MKB MRPR II (pH 7.6)	32	6.5 ± 0.3	E-274 (Y→H)	6	3.4	1.52	2.49
MKB MRPR IV (pH 7.6)	80	6.3 ± 0.3	E-237 (H→Y)	5	3.5	1.31	2.12
MKB MRPR (pH 6.0)	7	7.2 ± 0.2d	E-458 (G→R)
MS MRPR I (pH 7.6)	5	7.6 ± 0.4d	E-457 (M→I)
MS MRPR II (pH 7.6)	8	7.2 ± 0.2d	E-457 (M→I)

^aThe variants were indistinguishable from the parental virus on the basis of clinical score and histology score in either the discriminator areas or target plus discriminator areas, which are the usual criteria for evaluation of the World Health Organization neurovirulence test (12). 

^bASTs were calculated based on an inoculum of 100 PFU given by the intracerebral route to 4-week-old female NIH Swiss mice. 

^cNA, not applicable. 

^dAST statistically longer than that of parental FNV-Yale virus, P < 0.2. 

Previous studies have shown that the E protein of wild-type FVV differed from four strains of FNV in having three common amino acid substitutions at E-54, E-227, and E-249 (10). E-227 is not considered important to the phenotype of FNV, since FNV has an E residue at this position, like all YF virus strains sequenced except for FVV. Figure 1 shows a diagrammatic representation of the three-dimensional structure of the soluble portion of YF virus E protein based on the structure of the TBE virus E protein reported by Rey et al. (8), with the amino acid substitutions found in the FNV viruses (E-54, E-227, and E-249) highlighted in white. The single amino acid substitutions in the E proteins of MKB MRP^R I, MKB MRP^R II, and MKB MRP^R IV give rise to mutations that are clustered in domain II at E-237, E-260, and E-274, respectively, and are shown as gray-shaded amino acids in Fig. 1. This clustering is adjacent to the E-54 substitution found in the FNV viruses, suggesting that this region of the E protein is involved in binding of virus to monkey brain cells and may be in part responsible for the enhanced neurotropism of FNV viruses in monkeys.

FIG. 1.

Diagrammatic representation of the structure of the ectodomain of the flavivirus E protein (modified from reference 8). (A) Top view; (B) side view. White-highlighted amino acids are different in wild-type FVV and FNV; gray-shaded amino acids are substitutions found in MKB MRP^R variants.

The substitutions at E-260, E-274, or E-237 in the MKB MRP^R variants selected at pH 7.6 were distinct from those of the MKB MRP^R variant selected at pH 6.0, which had an amino acid substitution at E-458 in the proposed transmembrane region. However, selection of an MRP^R variant at pH 6.0 did not affect the viability of the virus, since both MKB MRP^R (pH 7.6) and MKB MRP^R (pH 6.0) variants and parental FNV were not sensitive to pH, as evidenced by their infectivity at pHs from 7.6 to 5.0. Infectivity titers varied by twofold or less. We speculate that the E-458 substitution is due to the conformational change in the E protein that takes place because of the decrease in pH and is consistent with the proposal for TBE virus (9). Also, the substitution found in the two MS MRP^R variants of YF FNV-Yale strain selected at pH 7.6 was at E-457 in the transmembrane region of the E protein. It is considered unlikely that the proposed transmembrane region of the E protein interacts directly with a cell receptor. Rather, the amino acid substitutions at E-457 and E-458 may induce a conformational change in the E protein that affects the interaction of the E protein with the cell receptor. Support for this proposal comes from comparison of levels of virus binding to MRPs. MS MRP^R and MKB MRP^R (pH 6.0) variants did not bind to either MKB or MS MRPs, while MKB MRP^R (pH 7.6) variants bound to MS MRPs (albeit poorly compared to parental FNV) but not to MKB MRPs (Table 3). Presumably, mutations at E-237 to E-274 in domain II affect the conformation of the E protein less than mutations in the transmembrane region, and MKB MRP^R variants can still bind to MS MRPs. Interestingly, the three MRP^R variants with amino acid substitutions at E-457 or E-458 had statistically longer average survival times (ASTs) in mice than did parental FNV-Yale and MKB MRP^R variants I, II and IV, even though these last three MKB MRP^R variants had higher PFU/LD₅₀ ratios than the other variants (Table 2). Thus, mutations in domain II appear to be associated with attenuation in terms in PFU/LD₅₀ ratios, while mutations in the transmembrane region appear to alter both the PFU/LD₅₀ ratios and the ASTs. The differences in neurotropism among the MRP^R variants may be associated with the binding of these variants to MS MRPs; namely, variants with amino acid substitutions in domain II still bind to MS MRPs, while the variants with substitutions in the transmembrane region do not (Table 3). Also, mutations in the transmembrane region may affect the uptake of virus into cells in the mouse brain and may be in part responsible for the increased ASTs of mice infected with these MRP^R variants. The above proposals remain to be directly tested using infectious clones and site-directed mutagenesis.

TABLE 3.

Comparison of binding indices of FNV-Yale and its MO and MKB MRP^R variants to MO or MKB MRPs

Virus strain	Binding indexa
	Virus + MS MRP	Virus + MKB MRP
Parent
FNV-Yale (pH 7.6)	2.9	4.6
FNV-Yale (pH 6.0)	2.9	5.1
MKB variants
MRPR I	1.3	0.1
MRPR II	0.9	0.2
MRPR IV	0.9	0.1
MRPR (pH 6.0)	0.1	0.1
MS variants
MRPR I	0.2	−0.2
MRPR II	0.5	0.3

^aLog₁₀ infectivity in buffer/infectivity in MRP supernatant (see legend to Table 1 for more details). 

Overall, these results suggest that, at the very least, the interaction of the E protein of FNV with mouse brain cell binding sites differs from that with monkey brain cell binding sites, and this may indicate that FNV recognizes different cell receptors on mouse brain and monkey brain cells. This suggestion is consistent with biological studies showing that all strains of YF virus are lethal for mice while only FNV kills monkeys following intracerebral inoculation (for a review, see reference 4).