Many investigations have focused on the molecular epidemiology of EV71 
, but few have attempted to identify in-host adaptation and the potential viral determinant of neurotropism or neurovirulence. Mutations in EV71 5′ UTR 
, VP1 gene (including the BC loop) 
, and 3D polymerase 
have been shown to result in attenuation in cynomologus monkeys and in mice, but they do not change the tissue specificity in the CNS of these experimental animal models 
. These models have many intrinsic limitations, namely, the use of adapted EV71 mouse strains and/or direct intracranial, intramuscular, or intraperitoneal inoculation. These experimental models are thus unable to mimic the natural route of infection in humans.
In this study, we analyzed the genomic differences in the EV71 genogroup C1 virus during a disseminated human disease that included meningitis. EV71 serotypes are divided into three major genetic lineages; lineage A whose prototype is the BrCr strain, and lineages B and C 
that are further subdivided into subgenogroups B1 to B5 and C1 to C5 
. Studies have suggested that the C1 genogroup is rarely a cause of CNS infection 
. Our goal was to identify viral signatures that could account for dissemination or site-specific adaptation. The comparison of five full-length genomes, sequenced directly from respiratory, gastrointestinal, CSF and blood specimens, revealed a drastic non-synonymous L97R substitution in the BC loop of the VP1 capsid protein that significantly modified the resulting viral phenotype. This mutation was specifically present in the blood and CSF, but not in the respiratory tract, and was present as a mixed population in the gastrointestinal tract. In addition to the VP197R
substitution, a conservative amino acid substitution at position 38 of protein 2B (V38A) was also observed in the blood, CSF, and gastrointestinal tract. Finally, a mixture of two nucleotides, both translated into a leucine in the 3D gene (3D175
), was also observed in the sequence of the stool specimen.
The BC loop region of VP1 is a known dominant immunogenic site as evidenced from experimental models using laboratory-adapted poliovirus or coxsackie virus strains 
. Seroneutralization experiments with the patient's serum failed to highlight the presence of an antibody-mediated, selective immune pressure promoting the VP1 L97R substitution. Thus, it is unlikely that the VP197R
sequence present in the immunocompromised patient's blood and CSF played a critical role in viral dissemination to the nervous system via immune escape. This has to be related to the immunosuppressed condition of the patient who was previously treated with anti-CD20 antibodies, although we cannot rule out the presence of low level antibodies or cellular immunity directed against the VP197L
strain that have gone undetected.
Apart from its immunogenic role, the BC loop region of VP1 was also identified as a determinant of poliovirus host adaptation 
. In vitro
cell tropism assays revealed that the VP197R,
conferred a significant advantage to the ability of EV71 to grow in neural-derived cells, independent of the virus lineage. The results of competition experiments suggest that the advantage is probably at the cell entry step. Indeed, introduction of a positively charged amino acid in the BC loop may have a substantial impact on the interaction of host cell surface receptors with this epitope. The EV71 VP1 structure models and the virus binding assay further support this hypothesis by revealing that residue 97 is close (~12 Å) to the interface of other known picornavirus VP1 receptors (), and by illustrating how the positively charged arginine side chain of VP197R
on the viral capsid surface may be more accessible to certain host cell receptors than the smaller side-chain of leucine of VP197L
(). Of note, backbone carbon atom alignment of our models to the EV71 VP1 structures 
that were released while this manuscript was under review shows less than 1.0 angstrom RMSD between models and structures (data not shown), thus further validating our modelling approach.
Interestingly, although the improved receptor-binding capacity of VP197R might be sufficient to confer a growth advantage in neuroblastoma cells, it cannot be ruled out that this substitution also confers potential advantages at various other stages (e.g., during the virion assembly process). This binding advantage was also observed to a lesser extent in Vero cells. Therefore, compensatory events must occur at one or multiple steps during virus amplification (viral genome replication, assembly, recruitment of cellular factors, others) to explain the VP197L advantage observed in the competition assay in this cell line. Of note, inoculation of the patient's stool specimen (that presented both VP197R and VP197L populations) in Vero cells resulted in virus isolate containing only the VP197R sequence. While this may seem surprising in light of the results from the competition experiment, one must note that Vero cells were transfected with RNA transcripts from pCI derivatives, thus bypassing the viral entry step. According to the binding assay, VP197R also provides a binding advantage in Vero cells that may partially explain these contradictory observations. Another possible explanation is that while the presence of a mixed population in stools was shown at the RNA level, there is no indication about the viability of the corresponding viral species in the sample. Thus, the presence of a leftover, potentially defective, viral VP197L genome that is unable to grow in culture cannot be ruled out.
Notably, in neuroblastoma cells, VP197R was frequently associated with a second mutation located in the EF loop at position 167 (E167G) of VP1 (VP1167G). In our structure models, position 167 is situated in the receptor-binding canyon near the base of the BC loop and may serve to limit the conformational flexibility of the BC loop (). Substitution of a negatively charged glutamate by the smaller neutral glycine may alleviate steric and/or electrostatic interference created by the VP197R at the VP1-receptor interface, thus serving to stabilize the VP1 interaction with the host cell receptor. This E167G substitution was absent in all of the patient's specimens analyzed, suggesting that this position is either not clinically relevant and only reflects cell-type adaptation under our experimental conditions, or that there was not enough time for it to appear during the course of infection. If more time had elapsed before treatment was administered, it is possible that the VP1167G substitution would have appeared and, in turn, exacerbated the patient's symptoms. This second hypothesis is favored by the finding of a publicly available sequence that contains both VP197R and VP1167G residues (GenBank accession number: AAF13503.1).
studies in mouse models 
and a comparative analysis of all EV71 complete genome sequences with identified clinical backgrounds available in the Genbank database 
both identified amino acid positions in VP1 associated with EV71 virulence, such as VP1145
in the DE loop situated on the rim of the surface canyon, or VP1164
in the EF loop situated on the slope of the canyon (). Furthermore, after amplification of infectious clones harboring EV71 subgenogroup B3 in SH-SY5Y and RD cell lines, VP194
was recently identified and is postulated to be important for cell-type adaptation 
. Taken together, the region surrounding the VP1 L97R mutation identified in this study likely plays an important role in cell-type adaptation and potentially neurotropism, independent of the EV71 genogroup.
In addition to the VP197R substitution, a conservative amino acid substitution at position 38 of protein 2B was observed, 2B38A. This substitution is uncommon and described in only one case among all available GenBank sequences. Whether this conservative V38A change in protein 2B may also confer new viral tropism was not substantiated in our experiments and remains an unsupported hypothesis.
Taken together, the sequence of clinical events, the genome characterization, our in vitro experiments, and our comparative VP1 structure models support the following scenario: the virus could have initially infected the respiratory tract, leading to a first viremic phase followed by invasion of the gastrointestinal tract. Alternatively, the virus may have entered simultaneously by oro-fecal and respiratory routes. High replication in the gastrointestinal tract may then have given rise to the appearance of a mixed viral population. The reduced immune response of the host then allowed a prolonged viremia, originating from a subspecies generated during the replication within the gastrointestinal tract that conferred a selective advantage for certain cell types, including neural cells. This resulted finally in neuro-invasion. In conclusion, this study provides the first genome-wide analysis of EV71 evolution and dissemination within a single human host over the course of an infection, and highlights how emergence of mutations at critical regions of the viral genome can lead to new phenotypes and neurovirulence. Further studies are underway to better define the target of the VP197R substitution and to investigate any potential effects of the associated mutation, VP1167G.