|Home | About | Journals | Submit | Contact Us | Français|
Three dengue virus type 4 (DENV-4) vaccine candidates containing deletions in the 3′ noncoding region were prepared by passage in DBS-FRhL-2 (FRhL) cells. Unexpectedly, these vaccine candidates and parental DENV-4 similarly passaged in the same cells failed to elicit either viremia or a virus-neutralizing antibody response. Consensus sequence analysis revealed that each of the three viruses, as well as the parental DENV-4 when passaged in FRhL cells, rapidly acquired a single Glu327-Gly substitution in domain III (DIII) of the envelope protein (E). These variants appear to have accumulated in response to growth adaptation to FRhL cells as shown by growth analysis, and the mutation was not detected in the virus following passage in C6/36 cells, primary African green monkey kidney cells, or Vero cells. The Glu327-Gly substitution was predicted by molecular modeling to increase the net positive charge on the surface of E. The Glu327-Gly variant of the full-length DENV-4 selected after three passages in FRhL cells showed increased affinity for heparan sulfate compared to the unpassaged DENV-4, as measured by heparin binding and infectivity inhibition assays. Evidence indicates that the Glu327-Gly mutation in DIII of the DENV-4 E protein was responsible for reduced infectivity and immunogenicity in rhesus monkeys. Our results point out the importance of cell substrates for vaccine preparation since the virus may change during passages in certain cells through adaptive selection, and such mutations may affect cell tropism, virulence, and vaccine efficacy.
Dengue virus (DENV) infects humans via the bite of infected mosquitoes, principally Aedes aegypti. DENV infections can be asymptomatic or cause a spectrum of illnesses that range from mild dengue fever to a severe, life-threatening disease characterized by dengue hemorrhagic fever/dengue shock syndrome (13, 38). The four DENV serotypes (DENV type 1 [DENV-1] to DENV-4) are the most important members of the genus Flavivirus in terms of morbidity, geographic distribution, and socioeconomic burden (1, 12). Several other members of the flaviviruses, including yellow fever virus (YFV), Japanese encephalitis virus (JEV), West Nile virus, and tick-borne encephalitis virus, are also important human pathogens.
The flavivirus virion is a spherical enveloped particle with icosahedral symmetry. It has a relatively simple structure, consisting of an inner nucleocapsid-virus RNA core and an outer lipid bilayer membrane into which a small ~9-kDa membrane protein (M) and a larger ~54-kDa envelope protein (E) are embedded. The E protein, which is approximately 500 amino acids in length, is the major antigen responsible for attachment to the cell surface, viral entry mediated by endocytosis, fusion with endosomal membranes, and the eliciting of host immune responses. There are 180 copies of E in the form of homodimers arranged in a tight array on the smooth virion surface without major spikes (21, 37, 48). Structural analysis indicates that each E monomer is folded into three structurally distinct domains, termed domains I, II, and III (DI, DII, and DIII, respectively). DIII has an immunoglobulin-like fold, a structural feature shared by many cell-adhesive molecules and receptor-binding proteins. DIII has been proposed to be responsible for binding interaction with cell surface receptors (16, 48). A number of mosquito-borne flavivirus E proteins contain a sequence motif in DIII that is recognized by integrin receptors. Mutations affecting cell attachment that cluster in this region are associated with attenuation of virulence and cell tropism (26, 27, 29, 43, 53).
A specific cell surface receptor has not been clearly identified for DENV or any other flavivirus. Studies focusing on the mechanisms of viral binding and entry in mosquito C6/36 cells (42, 56) or mammalian cells (35, 41) have suggested a number of proteins of various sizes that are capable of binding the DENV virion. Recently, the C-type lectin DC-SIGN was found to be capable of facilitating DENV infection of dendritic cells (51, 52). It has been proposed that flaviviruses could also utilize other less specific molecules on the cell surface as coreceptors for initial adsorption and infection. Infection of DENV-2 was first found to depend on heparan sulfate (HS), a major constituent of the extracellular matrix and a surface component of most mammalian cells, for binding interaction and infectivity of cultured cells (6). In that study the authors identified sequences of two HS binding sites in E, one in DIII and the other in the junction between DI and DIII. Although HS is essential for coordination of various cellular functions (10), the role of HS in mediating viral entry for infection of susceptible mammalian hosts or insect vectors is less defined. Studies have shown that propagation of wild-type strains of DENV adaptively selects variants to replicate in certain mammalian cell cultures, including rodent-derived BHK-21 (kidney fibroblast) and human SW-13 (adrenal carcinoma) cell lines (28, 33). These variants acquire mutations in E, creating new HS binding sites and facilitating interactions to gain an entry into these cells. Such adaptive selection of variants involving binding to HS also appears to be a common mechanism for other single-stranded RNA viruses, including alphaviruses and foot-and-mouth disease virus. Analysis of the biological properties of these HS binding variants demonstrated attenuation of virulence and restriction of cell tropism (2, 4, 15, 18, 26, 49).
Passage of wild-type YFV in animals and in cell cultures was successfully employed to produce the live-attenuated 17D vaccine several decades ago, and, more recently, the live-attenuated Japanese encephalitis vaccine SA14-14-2 was similarly produced (17, 39). In an effort to develop a DENV vaccine, investigators have sought to attenuate the virus by serial passage in primary dog kidney (PDK) cells or selection of small plaque-forming viruses in cultured cells (9, 14). Depending on the DENV serotype, various passage levels in PDK cells have been empirically derived in order to produce attenuated live vaccines. Final passage in fetal rhesus lung (FRhL) cells is frequently used for virus seed and vaccine lot production (8). The FRhL cell strain is a normal diploid cell suitable for the production of vaccines for human use (55). These cells have been shown to support the replication of all four DENV serotypes to high titers (9, 31).
The availability of DENV cDNA clones has made it possible to modify the viral genome in order to derive growth-restricted and chimeric DENV mutants for the production of live vaccine candidates (7, 22, 23). Earlier, a series of DENV-4 mutants containing deletions in the 3′ noncoding region (NCR) was generated and shown to be attenuated for replication in cultured cells and in nonhuman primates (36). In an effort to develop DENV vaccine candidates with defined attenuating mutations, these viruses were propagated in FRhL cells for vaccine production and then tested in rhesus macaques. Unexpectedly, the animals failed to develop either antibody responses or viremia after inoculation. In the present report, we present evidence that passage of these DENV-4 constructs in FRhL cells rapidly selected for variants containing a single mutation in E that resulted in increased virus binding to heparin, a highly sulfated form of HS, and the loss of infectivity for primates.
The following cultured cells were used for passage of full-length DENV-4 virus and its derived constructs: simian Vero (epithelial) cells of passage 142 shown to be free of retroviruses, primary African green monkey kidney (AGMK) cells, diploid fetal rhesus lung DBS-FRhL-2 (FRhL) (fibroblast) cells of passage 25, and Aedes albopictus C6/36 cells. All the cells were grown in minimal essential medium (MEM; Invitrogen, Carlsbad, CA), supplemented with 10% heat-inactivated fetal bovine serum (Biowhittaker, Walkersville, MD), 0.05 mg/ml gentamicin (Invitrogen), and 2.5 units/ml fungizone (Invitrogen). Mammalian cells were propagated at 37°C, and mosquito cells were propagated at 32°C. C6/36 cells were purchased from the American Type Culture Collection (Manassas, VA), and FRhL cells were obtained from the Walter Reed Army Institute of Research Pilot Bioproduction Facility (Silver Spring, MD). Vero cells and AGMK cells were kindly provided by L. Potash (Dyncorp, Rockville, MD).
Preparations of DENV-4 strains or constructs grown in various cells at various passage levels are summarized in Table Table1.1. Cloned full-length DENV-4 strain 814669 (GenBank accession number M14931) recovered from cDNA and its derived mutants containing deletions of various lengths in the 3′ NCR, i.e., deletions of nucleotides 172 to 143 from the 3′ end (3′d 172-143), 172 to 113 (3′d 172-113), and 303 to 183 (3′d 303-183), were recovered from C6/36 cells transfected with RNA transcripts made by SP6 polymerase, as described previously (36). Subsequently, these virus constructs were passaged in FRhL cells to obtain preparations of pre-master seed, master seed, and production seed at passages 1, 2, and 3, respectively. Passage 3 preparations of these viruses were designated DENV-4(FRhL-3), 3′d 172-143(FRhL-3), 3′d 172-113(FRhL-3), and 3′d 303-183(FRhL-3). Full-length DENV-4 recovered from C6/36 cells was also passaged three times in Vero cells to prepare DENV-4(Vero) or in AGMK cells to prepare DENV-4(AGMK). Additionally, DENV-4(FRhL-3) was further passaged in mosquito C6/36 cells three times to obtain the preparation of DENV-4(FRhL-C6/36). Titers of these preparations were determined, and the preparations were used for inoculation of monkeys. The titer of each virus in PFU per ml (PFU/ml) was determined on C6/36 cells or Vero cells, as both cell types gave comparable titers. The DENV-4 vaccine parental strain 341750 (also called parent-2) that had been passaged five times in AGMK cells and four times in FRhL cells (34) was subsequently passaged four times in C6/36 cells for sequence determination.
DENV genomic RNA was prepared from a 200-μl aliquot of each virus by extraction using Trizol LS reagent (Invitrogen). Reverse transcription was subsequently performed with a ThermoScript RT-PCR system (Invitrogen), followed by RNase H treatment. DNA amplification by PCR was performed using appropriate primers and Takara LA Taq polymerase (Takara, Madison, WI). DNA products were gel purified (QiaQuick gel extraction kit; Qiagen, Valencia, CA), and the entire viral genome was sequenced using a BigDye Terminator, version 3.1, Cycle Sequencing mixture (Applied Biosystems, Foster City, CA), in three independent experiments. The 5′ and 3′ noncoding sequences of the DENV-4(FRhL-3) viral genome were determined following RNA ligation and PCR amplification as described elsewhere (7, 32). In the case of identifying DENV-4 in monkey serum, only the C-prM-E region was sequenced using specific primers. Sequencing reactions were performed on a 3730 DNA analyzer (Applied Biosystems), and the sequences of overlapping fragments were assembled and evaluated using Sequencher, version 4.8 (GeneCodes Corp., Ann Arbor, MI). In the case where the electropherogram indicated the presence of two nucleotides at the same position in both plus- and minus-strand sequencing reactions, nucleotide assignment was made by direct examination using the software Sequencing Analysis, version 5.2 (Applied Biosystems). The percentage of each population was calculated using the following formula, described by Kapoor et al. (20): [MPH/(MPH + WPH)] × 100, where MPH and WPH are the arbitrary fluorescence intensity peak heights of the mutant and wild-type nucleotides, respectively, in the sequencing electropherograms.
Groups of DENV-seronegative, juvenile, Indian-origin rhesus macaques that weighed 3 to 6 kg were inoculated with the virus preparations indicated in Table Table1.1. Each monkey was injected subcutaneously with 105 PFU (titers determined on C6/36 cells) of each virus construct diluted to a final volume of 1.0 ml in MEM plus 0.025% human serum albumin, administered as two doses of 5 × 104 PFU, one dose in each side of the upper shoulder area. Ketamine at 10 mg/kg was administered to anesthetize the animals prior to virus inoculation. In the first study, serum samples from inoculated monkeys were collected daily for 2 weeks and then at 4 weeks after infection for analysis of viremia and antibody response, respectively. The antibody response was measured by a plaque reduction-neutralization test (PRNT) using DENV-4, and the 50% DENV-4 neutralizing titer (PRNT50) was determined as previously described (36). Viremia was detected by inoculation of Toxorhynchites mosquitoes, and then the head tissues of infected mosquitoes were examined for DENV antigens by an indirect immunofluorescence assay. In the second monkey experiment, serum samples from inoculated monkeys were collected daily from day 1 to day 10 for viremia assay and 8 weeks after inoculation for serologic analysis. For viremia analysis, serum specimens were inoculated onto C6/36 cells to amplify any virus present. Briefly, an 80-μl serum sample diluted with 220 μl of MEM plus 2% heat-inactivated fetal bovine serum (MEM-2%) was added to C6/36 cells in a T-25 flask for 1 h, and the cells were then fed with 5 ml of MEM-2%. Following a medium change at day 7, the fluid medium at day 14 was collected for detection of any amplified virus by immunofluorescence assay or plaque assay on C6/36 cells (36).
DENV-4 heparin inhibition and heparin-Sepharose binding assays were performed essentially as described by Lee et al. (24, 28). Briefly, 1 × 104 focus-forming units (FFU) of parental DENV-4 or its variants were incubated with soluble heparin (from porcine intestinal mucosa; purchased from Sigma, St. Louis, MO) at 0.02, 0.2, 2, 2,0 or 200 μg/ml in Hanks' balanced salt solution (Invitrogen) containing 0.2% bovine serum albumin (HBSS-BSA) (Roche Diagnostics, Indianapolis, IN). Controls were prepared by preincubation of the same amount of virus in HBSS-BSA without heparin. The mixture was incubated for 30 min at 37°C prior to the addition to Vero cell monolayers in 24-well plates (105 cells/well). After an absorption period (1 h at 37°C), cells were washed twice with HBSS-BSA buffer prior to the addition of maintenance medium (MEM-2%). The same heparin concentrations were maintained in the medium during the incubation for 4 days at 37°C and 5% CO2. Supernatants were collected at 24, 48, 72, and 96 h postinfection and titrated by focus-forming assay (FFA) on Vero cells (45). Inhibition assays were performed in duplicate in two independent experiments. Percentage of inhibition was calculated as follows: 100 − (100 × [number of FFU (heparin treatment)/number FFU (control)]). For the heparin binding assay, heparin-Sepharose and control protein A-Sepharose beads (Pharmacia, Uppsala, Sweden) were suspended in phosphate-buffered saline (30%, wt/vol) and equilibrated before use by pelleting and washing three times in HBSS-BSA plus 10 mM HEPES (pH 8.0). Parental full-length DENV-4, DENV-4 harboring a Glu327-Gly mutation (DENV-4, Gly327), and DENV-4 harboring a Glu327-Ala mutation (DENV-4, Ala327) (105 PFU in 100 μl of HBSS-BSA) plus 100 μl of HBSS-BSA with or without Sepharose beads were mixed in Eppendorf tubes and held at 4°C for 6 h with repeated mixing. Virus-bead mixtures were then centrifuged for 5 min at 6,000 × g at 4°C to pellet the Sepharose beads, and infectious titers in supernatants were determined by FFA on Vero cells.
Subconfluent FRhL cells in 24-well plates were infected with parental DENV-4 strain 814669 or its heparin-binding variants at a multiplicity of infection (MOI) of 0.01. After an adsorption period of 1 h at 37°C, cells were washed twice with warmed phosphate-buffered saline, and then 1 ml of MEM-2% was added. Aliquots of the culture fluids were collected daily for 10 days and kept at −70°C. The infectivity in FFU/ml was determined in Vero cells, and quantitative analysis of viral RNA was performed by real-time quantitative reverse transcription-PCR (qRT-PCR) essentially as described previously (11). Briefly, viral RNA was extracted from 140 μl of the culture fluid sample with a QIAmp Viral RNA Mini Kit (Qiagen, Valencia, CA). RT-PCR was performed by using a TaqMan RNA-to-CT 1-Step Kit (Applied Biosystems, Foster City, CA) and DENV-4 M-specific primer pairs and probe in an ABI Prism 7900HT Fast Real-Time PCR System (Applied Biosystems). Parental DENV-4 RNA was used to generate a standard curve with 10-fold dilutions of RNA isolated from a known amount of virus, covering a 7-log10 dynamic range. The results were expressed as genome equivalents per ml (GEq/ml).
Previously, DENV-4 mutants containing deletions in the 3′ NCR were prepared in LLC-MK2 cells and shown to exhibit the predicted attenuation phenotype, as demonstrated by growth restriction in cultured cells and reduced viremia and immunogenicity in rhesus monkeys, compared to the parental DENV-4 (36). In order to produce virus seeds suitable for vaccine production, these attenuated virus constructs were recovered in C6/36 cell cultures and then passaged in diploid FRhL cells. After three FRhL cell passages, the DENV-4 vaccine candidates, i.e., the 3′d 172-143(FRhL-3), 3′d 172-113(FRhL-3), and 3′d 303-183(FRhL-3) viruses were evaluated for infectivity and immunogenicity in rhesus macaques, together with the full-length cDNA-derived DENV-4(FRhL-3) that had been similarly recovered from C6/36 cells and propagated in FRhL cells and the uncloned DENV-4 strain 814669. Table Table22 shows that the uncloned DENV-4 elicited virus-neutralizing antibodies in all four animals, with titers ranging from 1:122 to 1:420 (geometric mean titer, 1:289) as well as viremia lasting 2 to 5 days in three of four animals. In contrast, neither of the three vaccine candidates nor the full-length DENV-4(FRhL-3) that had been passaged in FRhL cells induced a detectible antibody response or viremia in the inoculated animals, results apparently inconsistent with those of an earlier study (36). We then sought to determine the effect on virus phenotype of passage of the full-length DENV-4 in other cell lines and the molecular mechanism responsible for the unexpected loss of virus infectivity and immunogenicity after passage in FRhL cells.
An experiment was performed to determine if the cDNA-derived full-length DENV-4 recovered from C6/36 cell culture and then propagated in AGMK or Vero cells maintained monkey infectivity at a level similar to that of virus propagated in LLC-MK2 cells (36). Table Table33 shows that DENV-4(C6/36) recovered from C6/36 cells without additional passage in mammalian cells replicated in rhesus macaques and elicited high antibody titers (geometric mean titer, 1:1,330). Similarly, viruses passaged in Vero cells [DENV-4(Vero)] or AGMK cells [DENV-4(AGMK)] also induced viremia and neutralizing antibodies in the animals. All six monkeys inoculated with DENV-4(Vero) or with DENV-4(AGMK) developed similarly high titers of neutralizing antibodies, with geometric mean titers of 1:796 and 1:721, respectively, comparable to the antibody titers measured in monkeys inoculated with DENV-4(C6/36). These results demonstrate that the full-length DENV-4 following passage in Vero cells or primary AGMK cells maintained satisfactory levels of immunogenicity, unlike viruses passaged in FRhL cells.
We then sought to test the hypothesis that passage of DENV-4 in FRhL cells (but not in LLC-MK2, Vero, or AGMK cells) adaptively selected for genetic variants responsible for the loss of infectivity and immunogenicity for monkeys. If such adaptive selection of variants occurred in FRhL cells, then only three cell passages might be sufficient for the variant genotype to supplant the consensus parental DENV-4 genotype. We passaged DENV-4(FRhL-3) three times in C6/36 cells in an attempt to reverse select for virus that was infectious and immunogenic for monkeys, similar to parental DENV-4. The resulting virus, DENV-4(FRhL,C6/36), was inoculated into six monkeys. All animals developed moderate to high titers of neutralizing antibodies, and three of six animals developed viremia of 1 to 3 days of duration, demonstrating that infectious and immunogenic DENV-4 rapidly reemerged as a result of adaptive selection in C6/36 cells (Table (Table33).
The results obtained thus far supported the hypothesis that passage of DENV-4 strain 814669 in FRhL cells selected for variants unable to replicate sufficiently in monkeys to induce viremia and neutralizing antibody responses. In order to further test this hypothesis, the consensus genomic sequences of the three FRhL cell-passaged DENV-4 deletion mutants and the full-length parental DENV-4 also passaged in FRhL cells were determined and compared with the sequences of corresponding viruses recovered from C6/36 cells. Table Table44 shows that a variant DENV-4 that contained a nucleotide change from GAA to GGA at position 1918 (underlined), resulting in a Glu327-Gly substitution in E, was detected in each of the preparations after three passages in FRhL cells. Based on the nucleotide signal at position 1918 on the electropherogram, there was a predominance of the variant population in DENV-4(FRhL-3) and 3′d 303-183(FRhL-3), where the presence of the parental DENV-4 genotype was too low to be detected. In contrast, both the 3′d 172-143(FRhL-3) and 3′d 172-113(FRhL-3) viruses contained mixed populations, with the variant representing a majority estimated at 60 to 80%, according to one calculation (20). Consensus analysis of 3′d 172-113(FRhL-3) also detected substitutions outside the E protein, which, however, were not present in the other FRhL cell-passaged viruses and therefore were not likely to be associated with infectivity for monkeys (Table (Table44).
To gain an insight into the kinetics of the adaptive selection of variants during passages in FRhL cells, consensus sequence analysis was also performed on the earlier passages of the parental DENV-4 and its derived deletion mutants. Figure Figure11 shows typical sequence electropherograms of these viruses at passage levels 1, 2, and 3. There was no detectible G1918 substitution at passage 1. At passage 2, both A and G nucleotides appeared at the position as a mixture in the full-length DENV-4 and the 3′d 172-143, and 3′d 303-183 mutants, indicating that the Glu327-Gly variant of each virus rapidly emerged. The Glu327-Gly variant population increased further and became predominant at passage level 3. In the case of 3′d 172-113, the Glu327-Gly variant was already predominant at passage 2. The predominance of the variant suggests that this virus had a clear selective advantage over the parental genotype for growth in FRhL cells.
It is interesting that an identical nucleotide substitution was detected in all four DENV-4 constructs following adaptive selection in FRhL cells. It is possible that the mutation was present in quasispecies, generated by error-prone SP6 polymerase used to generate the infectious RNA for transfection (47). To test this possibility, the uncloned DENV-4 strain 814669 was also passaged in FRhL cells three times to determine if rapid acquisition of mutations occurred. The result showed that a variant containing an A-to-C nucleotide mutation at nucleotide position 1918 resulting in an Glu327-Ala substitution emerged after two passages and became predominant after three passages (data not shown). Thus, a substitution occurred at the same site at amino acid position 327 but involved a different amino acid in E. The detection of the variant DENV-4 containing the Glu327-Ala substitution was associated with the appearance of small plaques on Vero cell monolayers (Fig. (Fig.2),2), morphologically similar to those observed for variants containing the Glu327-Gly substitution.
The replication capacity of DENV-4 variants relative to the parental DENV-4 in FRhL cells was studied by growth analysis. Infection was initiated at an MOI of 0.01 and continued for 10 days. Both variants yielded peak virus titers 10- to 25-fold higher than the parental virus 4 days after infection. Similarly, higher viral yields of the variants were also observed when the RNA GEq was measured by qRT-PCR (Fig. 3A and B). Based on these results, the ratio of infectivity to GEq (FFU/GEq) was calculated for the DENV-4 parent and its derived variants. Figure Figure3C3C shows that the FFU/GEq ratio was approximately twofold lower for the parental DENV-4 than for the variants. This finding further indicates that the parental DENV-4 was less infectious, i.e., probably contained more defective particles, than the variants similarly propagated in FRhL cells. Taken together, these results support the notion that DENV-4 mutants Glu327-Gly and Glu327-Ala exhibited beneficial growth in FRhL cells compared with the parent virus.
Previous analysis showed that after further passage of DENV-4(FRhL-3) in C6/36 cells, the resulting virus, DENV-4(FRhL,C6/36), was infectious and immunogenic in monkeys. Reverse selection of parental DENV-4 in mosquito C6/36 cells was most likely responsible for restoration of infectivity and immunogenicity in monkeys. To provide evidence, consensus sequence analysis of DENV-4(FRhL,C6/36) was performed, and the result showed that the parental DENV-4 containing Glu327 in E increased from undetectable levels in the DENV-4(FRhL-3) virus to approximately 50% and that the variant containing Gly327 in E was correspondingly reduced to 50% of the population (data not shown). This finding prompted us to further analyze the virus present in the serum samples of monkeys inoculated with this preparation (Table (Table3).3). Among six serum samples determined to be viremia positive by virus focus assay, four were shown to be DENV-4 positive by PCR DNA amplification. The E sequence corresponding to the parental DENV-4 genotype, but not the variant, was detected in all four samples. This experiment provides additional evidence that the DENV-4 variant containing the Glu327-Gly substitution failed to replicate in monkeys and that the antibody response was elicited in response to infection with parental DENV-4 present in the inoculum.
Glu327 in DENV-4 E is located within the BC loop at the upper lateral ridge of DIII and is accessible on the virion surface, based on the DENV-2 structural model determined by cryo-electron microscopy (Protein Data Bank code 1THD) (57). Molecular modeling showed that a Glu327-Gly or Glu327-Ala substitution is predicted to increase the net positive charge at the local area, as shown by surface mapping of the electrostatic field generated by the Adaptive Poisson-Boltzmann Solver program using the DENV-4 DIII nuclear magnetic resonance reconstruction (54) (Fig. (Fig.4A4A).
Since a single Glu327-Gly or Glu327-Ala substitution was detected in the entire genome of DENV-4, together with small-plaque morphology after three passages in FRhL cells, the virus was appropriate for further biologic characterization. The ability of heparin to inhibit the infectivity of the DENV-4 variants was compared with that of the parent virus by focus assay on Vero cells. Figure Figure4B4B shows that the DENV-4(FRhL-3) variant containing Glu327-Gly and the uncloned 814669(FRhL-3) variant containing Glu327-Ala were highly sensitive to heparin inhibition even at a low concentrations, whereas the infectivity of parent DENV-4 was only slightly affected by heparin. An additional experiment was carried out to examine the binding of the parent and both DENV-4 variants to heparin-Sepharose beads. The results showed that the fraction of the variant containing either the Glu327-Gly or Glu327-Ala substitution retained by heparin beads was significantly higher than that of the parental DENV-4 (Fig. (Fig.4C).4C). The Glu327-Ala variant appeared to show a higher level of binding to heparin than the Glu327-Gly variant although its biological properties were not investigated. As a control, less than 10% of the parental and variant viruses bound to protein A-Sepharose beads without heparin (data not shown). Taken together, these experiments demonstrated that the Glu327-Gly mutation in the DENV-4 variant provided a binding site for a high-affinity interaction with heparin, consistent with high-affinity binding of virus to the mammalian cell surface via HS. As shown for variants of many other viruses adaptively selected in mammalian cells (4, 24, 25, 28, 33, 44, 50), DENV-4 variants that acquired a positive surface charge were possibly selected for their ability to bind to negatively charged glycosaminoglycans, especially HS, abundantly present on the surface of most mammalian cells. Likewise, increased binding of DENV-4 variants with cell surface HS was also associated with small-plaque morphology and reduced infectivity in monkeys.
This study resulted from an earlier attempt to prepare DENV-4 vaccine candidates containing deletions in the 3′ NCR by passage in FRhL cells suitable for live vaccine production. We found that each of the vaccine candidates and the parental DENV-4, when similarly passaged in FRhL cells and inoculated into rhesus macaques, failed to induce viremia or antibody responses in the animals. Consensus sequence analysis showed that each of three vaccine candidates and the parental DENV-4 acquired a single Glu327-Gly substitution in E during passage in FRhL cells. Evidence suggests that these variants were selected in response to adaptation to the specific cell environment. First, the mutation was not present in either viruses recovered from C6/36 cells or viruses passaged in LLC-MK2 (36), AGMK, or Vero cells. Second, an analysis of virus replication showed that the DENV-4 yield was low after the first and second FRhL cell passages and increased after the third passage, consistent with the kinetic profile of mutation acquisition (G. Añez, unpublished observations). Adaptation to growth in FRhL cells was further supported by growth analysis in which variants Glu327-Gly and Glu327-Ala yielded titers 10- to 25-fold higher than the titer of the parental DENV-4.
The increase in the net positive charge on the local surface of E DIII resulting from the Glu327-Gly substitution was consistent with the creation of a new site for binding interaction with HS. The finding that all three mutants and the parental DENV-4 independently acquired the same amino acid mutation at the same site suggests the possibility that the mutation was present in the viral quasispecies generated by the error-prone SP6 polymerase used to produce RNA for transfection. It has been reported that SP6 polymerase introduced an approximately 1,000-fold greater sequence error rate of mutations in the YFV/DENV chimeric genome than the YFV RNA-dependent RNA polymerase did in YFV-infected cells (47). However, quasispecies generated as a result of DENV RNA-dependent RNA polymerase during viral replication in the infected cells could not be ruled out. Also, the number of potential sites for HS binding on DENV-4 E appeared to be limited, perhaps due to the sequence or structural constraints. In any event, the acquisition and accumulation of mutation were rapid during passages in FRhL cells. Glu327 is conserved among DENV-1, DENV-2, and DENV-4 Es, but this position is occupied by Lys in DENV-3 E. Molecular modeling showed that a similar Gly substitution at this position would also increase the positive surface charge on DENV-1 and DENV-2 E, thus potentially generating an HS-binding site (data not shown). It remains to be investigated whether passage of DENV-1, DENV-2, or DENV-3 in FRhL cells also selects HS-binding variants containing a similar mutation in DIII or other locations on E. The current view is that flaviviruses as well as members of a number of other single-stranded RNA virus families undergo adaptive mutations involving HS binding for growth selection in cultured mammalian cells (2, 25, 33, 50). As part of this adaptation strategy, DENV-2 variants selected after passage in SW-13 cells or in BHK-21 cells contained a number of HS-binding substitutions clustered in DII but none in DI or DIII of E (28). On the other hand, a large panel of tick-borne encephalitis virus variants has been selected after passage in BHK-21 cells and shown to contain HS-binding mutations on each of the three domains of E. It has been suggested that the locations of HS-binding sites on the E protein of flaviviruses may be influenced by different cell environments (33).
Evidence indicates that HS-binding variants of a number of flaviviruses exhibited a small-plaque morphology and reduction of virulence in animals and possibly in humans. Molecular analysis of YFV 17D and JEV SA14-14-2 vaccines, each of which has been passaged numerous times in cultured chicken cells (for YFV) or hamster kidney cells (for JEV) and in mice, has revealed HS-binding mutations that presumably contribute to virus attenuation (25, 26, 44). Interestingly, Lys326 in the YFV 17D virus E sequence aligns with Glu327 of DENV-4 E. In the virulent YFV Asibi strain, position 326 is also occupied by Glu. Mutagenesis analysis using cloned YFV DNA revealed that viruses with a Lys or Arg substitution at this position, resulting in a net gain of positive charge, exhibited increased heparin sensitivity and reduced neuroinvasiveness in mice (44). Thus, it has been suggested that heparin-binding activity involving position 326 plays a role in modulating YFV virulence phenotypes. Studies also showed that a heparin-sensitive mutation at Glu306 leads to loss of neuroinvasiveness of the JEV SA14-14-2 vaccine (26). In both cases, satisfactory attenuation of virulence was probably achieved in combination with numerous other mutations elsewhere in the viral genomes, some of which possibly even facilitate viral replication, thus compensating for possible overattenuation due to HS-binding-mediated mutations. These studies suggest that binding interactions with HS play a principal role in determining cell tropism and the attenuation of viruses. Conversely, DENV-2 mutants that were selected for increased neurovirulence in mice contain negatively charged Glu at positions 124/128 in DII of E, which is responsible for reduced interactions with HS (46). Studies have shown that HS-binding variants of a number of encephalitic flaviviruses and alphaviruses are rapidly cleared from the bloodstream (2, 5, 24, 26, 28). Thus, the ability of such viruses to cause disseminated infection and their potential for neuroinvasiveness in animals are limited. The attenuation phenotype of DENV-4 HS-binding variants in monkeys, as characterized by reduced replication and loss of antibody response in the present study, could be explained by the same clearance mechanism.
While passage of parental DENV-4 in Vero cells did not lead to the selection of a HS-binding variant, the DENV-4 HS-binding variant selected in FRhL cells remained stable at least after four passages in Vero cells, as demonstrated by maintenance of the small-plaque phenotype and the presence of the Glu327-Gly substitution, which was determined by consensus sequence analysis (data not shown). The question can be raised as to whether the HS-binding mutation in DENV-4 is equally stable in vitro and in vivo. DENV-4(FRhL-3) preparations which contained mainly a DENV-4 variant with little or no detectible parent virus, as determined by consensus sequence analysis, exhibited approximately equal quantities of variant and parental viruses after three passages in C6/36 cells, suggesting that selection for the parental DENV-4 genotype was strongly favored in these cells. Monkeys inoculated with the mixed virus preparation developed viremia containing only the parental DENV-4 genotype but not the HS-binding variant, which is consistent with the notion that only the parental DENV-4 replicated sufficiently to cause viremia and the production of neutralizing antibodies in response to infection. The relative capacity of the parental and HS-binding variant of DENV to replicate in mosquitoes and in primates has potentially important implications: (i) in mosquitoes, the parental genotype will be rapidly amplified through selection for a growth advantage in the insect vector; and (ii) in the mammalian host only the parental virus, but not HS-dependent variants, should be amplified. Both processes likely play an important role for the survival of the virus and maintenance of its virulence phenotype in nature. HS could serve as an important modulator of DENV replication and virulence in the infected host to maintain its transmission cycles in the mosquito vector and vertebrate hosts. Along this line, studies have also shown that inoculation of cattle with the attenuated, HS-binding variant of foot-and-mouth disease virus resulted in recovery of the virulent wild-type virus during infection in animals (50).
Other investigators have described passage of the same DENV-4 strain 814669 construct in MRC-5 cells (a line of diploid fetal human lung fibroblast cells) also used for live vaccine production (30). In that study, the authors identified a Glu345-Lys substitution resulting in another change producing a gain of positive charge that could create a potential HS-binding site on E although the heparin sensitivity of the variant was not determined. To further gain an insight into selection of variants upon passage of other DENV-4 strains in cultured cells, we chose DENV-4 strain 341750 that had been used to derive candidate vaccines. Since the wild-type DENV-4 strain 341750 and its derived live-attenuated vaccine candidates were not available for analysis, we used the same DENV-4 strain (also called parent-2) that had been passaged in AGMK cells five times and then in FRhL cells four times. Consensus sequence analysis of DENV-4 strain 341750 (parent-2) detected a mixed population containing Lys along with Glu at position 345, calculated at 82% and 18%, respectively. To provide evidence that the population containing Glu345 in E represented the wild-type virus, passage of DENV-4 strain 341750 (parent-2 virus) in C6/36 cells was performed in an effort to reselect the wild-type virus. Subsequent sequence analysis of the C6/36 cell-passaged virus showed that the DENV-4 population containing Glu345 in E increased from 18% to 36%, supporting the notion that favorable selection of wild-type DENV-4 had occurred (data not shown). To further corroborate this finding, a sequence search of the GenBank database also showed that Glu345, but not Lys345, was present in all the E sequences of wild-type DENV-4 isolates. This finding is consistent with the notion that Lys345 emerged during initial passages of DENV-4 strain 341750 in AGMK cells or, most likely, in FRhL cells. DENV-4 strains 814669 and 341750 were isolated between 1981 and 1982 in the Caribbean area (Dominica and Colombia, respectively). Both strains belong to genotype II and contain an otherwise identical amino acid sequence in E (G. Añez, unpublished observation).
Recently, Monath et al. reported the preparation of a chimeric JEV/YFV 17D virus in FRhL cells, noting that five consecutive passages in these cells selected a predominant (>85%) small-plaque variant, with the Met279 in the attenuated JEV SA14-14-2 reverting to the wild-type Lys in DII of E (40). The chimeric JEV/YFV variant was shown to exhibit small-plaque morphology, reduced viremia, and antibody responses in monkeys, but its HS binding and inhibition profiles were not investigated.
FRhL cells have been used for the manufacture of several licensed vaccines including rabies (RVA) (3) and a rotavirus vaccine (Rotashield) (19). These cells are also being used for final passages of tetravalent live-attenuated DENV vaccines currently in phase 2 trials being developed by the Walter Reed Army Institute of Research and GlaxoSmithKline. The DENV-4 strain 341750 component of the tetravalent vaccine formulation was initially passaged in PDK cells and then passaged in FRhL cells four times, and the attenuation phenotype was assessed in rhesus monkeys prior to human trials (8). A comparison of viremia and antibody responses in monkeys may be possible between DENV-4 strain 341750 (parent-1) and its FRhL cell-passaged virus (parent-2) (34). Both the parent-1 and parent-2 viruses showed a similarity of infectivity in monkeys, as detected by viremia assay. There appeared to be a reduction of neutralizing antibody titer in monkeys infected with parent-2, but its significance was difficult to assess because only two monkeys were used. A reduction of monkey infectivity and immunogenicity was seen with a DENV-3 vaccine candidate (PDK-30); however, this virus has not been investigated to look for changes associated with heparin sensitivity (8).
Regardless of the attenuation strategy used, whether it is by serial passage or by introduction of defined mutations in the virus genome, careful evaluation is required to assess the genetic stability and biological properties of vaccine candidates in every step of the production process. Our current results point out the importance of choosing cell substrates for vaccine preparation since the virus may change during passage in certain cells through adaptive selection and since such mutations may affect its attributes of virulence, cell tropism, and immunogenicity.
We thank Robert H. Purcell and Ana P. Goncalvez for helpful discussions, Robert Putnak for critical reading of the manuscript, Doria R. Dubois for expert help in the preparation of DENV-4 constructs in FRhL cells, Masayuki Tadano and Sakae Arakaki for determination of neutralizing antibody titers in serum samples, and Ronald E. Engle for technical assistance with the real-time qRT-PCR experiments.
This work was supported by the Intramural Research Program of the National Institute of Allergy and Infectious Diseases, National Institutes of Health. This research project received partial financial support from the WHO Global Programme for Vaccines.
Published ahead of print on 5 August 2009.