Hypoglycemic Toxins and Enteroviruses as Causes of Outbreaks of Acute Encephalitis-Like Syndrome in Children, Bac Giang Province, Northern Vietnam

We investigated the cause of seasonal outbreaks of pediatric acute encephalitis-like syndrome associated with litchi harvests (May–July) in northern Vietnam since 2008. Nineteen cerebrospinal fluid samples were positive for human enterovirus B, and 8 blood samples were positive for hypoglycemic toxins present in litchi fruits. Patients who were positive for hypoglycemic toxins had shorter median times between disease onset and admission, more reports of seizures, more reports of hypoglycemia (glucose level <3 mmol/L), lower median numbers of leukocytes in cerebrospinal fluid, and higher median serum levels of alanine aminotransferase and aspartate transaminase than did patients who were positive for enteroviruses. We suggest that children with rapidly progressing acute encephalitis-like syndrome at the time of the litchi harvest have intoxication caused by hypoglycemic toxins, rather than viral encephalitis, as previously suspected. These children should be urgently treated for life-threatening hypoglycemia.

addition, a mapping strategy (Soap2 aligner) was used to produce coverage and depth of sequencing statistics for each complete genome.

PCR and Sanger Sequencing
PCR products were obtained and sequenced by using in-house-designed primer pairs and panenterovirus primer sets (1,(3)(4)(5). PCR amplifications were performed by using Taq DNA polymerase (Invitrogen). Conditions for amplification by using in-house primers are detailed above. Nucleotide sequences were obtained by using Big Dye Version 1.1 chemical analysis on an ABI 3730XL Apparatus (Applied Biosystems, Foster City, CA, USA).
According to corrected Akaike Information Criterion estimations, the most appropriate pattern of substitutions was as previously used and was complemented by a discrete gamma distribution of evolutionary rates among sites, assuming that certain sites remained invariable. An uncorrelated lognormal distribution with a default uniform prior ranging from 0 to 1 was used to model the rate of evolution. Several coalescent based models were tested as tree priors. The Markov chain Monte Carlo was set to 10,000,000 states, and 10,000 trees were sampled to obtain an adequate posterior effective sample size >200.

Identification of Recombinants
Virus genomes were included in recombination detection tests and molecular typing approaches. Phylogenetic analysis showed clustering of VP1 sequences from each sample with respective serotypes of human enterovirus B previously identified by the BLAST analysis (Technical Appendix Figure 1). Other parts of the genome, such as genes 2 (A, B, C) and 3 (A, C, D), clustered differently than VP1 and had robust statistical support, indicating that enteroviruses from 2008 were potential recombinants (Technical Appendix Figures 2, 3). Specific analyses of recombination statistically supported several putative recombination breakpoints (Technical Appendix Figure 2).

Toxicologic Analysis
Blood samples were tested for hypoglycin A (HGA), carnitine and glycine conjugates of methylenecyclopropylacetyl (metabolite of HGA) and methylenecyclopropylformyl (metabolite of methylenecyclopropylglycine [MCPG]), and short-to-medium length chain fatty acids by using a modification of a reported analytical method (17,18). Sample volume was increased and dilution of extracts was reduced to increase sensitivity. In brief, a methanolic internal standard solution (300 L) was added to 25 L of serum or urine for extraction, and the mixture was vortexed for 20 s and centrifuged for 10 min at a relative centrifugal force (RCF) of 17,000. From the clear supernatant, 250 L was removed and dried in a microtiter plate at 65°C for 30 min under a gentle stream of nitrogen.
The residue was treated with 50 L of 3N butanol-HCl for 15 min at 65°C and dried again at 65°C under nitrogen. The dry material was dissolved in in 70 L of methanol:water (80:20 vol/vol) and further diluted 1:2 with water. From this solution 90 L was transferred to a 384 microtiter plate, centrifuged at an RCF of 17,000 to sediment any particles, and then used for ultraperformance liquid chromatography-tandem mass spectrometry. From this solution, 5 L was injected onto an ACQUITY UPLC BEH C18 1.7 m, 2.1 × 50 mm column (Waters, Eschborn, Germany) for gradient chromatography. Tandem mass spectrometric analysis was performed with single-point calibration on a Xevo TQ-MS UPLC-MS/MS System (Waters). The lower limit of detection for all compounds was 1 nmol/L, and the lower limit of quantification was 10 nmol/L.
In addition, a spectrum of 24 carnitine esters of saturated and unsaturated fatty acids ranging from short-chain to long-chain molecules (C2-C18), including hydroxy and dicarboxylic acids, was quantified by using tandem mass spectrometry without preceding chromatographic separation, according to standard methods used in newborn screening for inborn errors of metabolism (19,20). Concentrations of acyl carnitines were deduced by comparing peak height to those of internal standards.

Methods
To study clinical and biologic profiles of the 58 patients, principal component analysis (PCA) was performed for age, temperature at admission, number of days between disease onset and symptoms, glycemia at admission, number of leukocytes in CSF, and blood levels of liver enzymes (aspartate aminotransferase and alanine aminotransferase). We determined which groups of patients were similar or different and characterized these groups by variables or groups of variables. Variables were normalized before performing PCA. We excluded from the analysis 1 person who had an excessive contribution to the first principal component (exceeding its weight), which could have caused unreliable results.

Results
We retained the first 3 principal components, which had eigenvalues >1, indicating that they accounted for more variance than accounted by 1 of the original variables. Together, they accounted for 68% of total variance.

Interpretation of Principal Components
We determined correlation coefficients between initial variables and the 3 principal components (Technical Appendix Table 3). For interpretation of PCA, coefficients of interest are those whose absolute value is closer to 1.

Three-Dimensional Representation of Patients
We obtained a projection of patients in 3-dimensional space formed by principal components (Technical Appendix Figure 4). Children with higher levels of HGA/MCPG (group 2) formed a distinct cluster in the projection space that differed from children infected with enteroviruses (group 1).
Children with higher levels of HGA/MCPG had increased levels of liver enzymes and more severe hypoglycemia (negative coordinates on the first axis), more rapid progression of the disease and standard levels of leukocytes in CSF (negative coordinates on the second axis), and a younger age (positive coordinates on the third axis). Children not infected with enteroviruses for whom HGA/MCPG levels were low (group 3) or not tested (group 4) had profiles more similar to those with enteroviruses. *Group ratio is the median of group 1 divided by the median of group 2. Children were divided into 2 groups: 9 had high (>100 nmol/L) values of HGA in serum (group 1), and 11 (group 2) had values below the limit of quantification (10 nmol/L), including 10 below the limit of detection (1 nmol/L) (group 2). Samples were sorted by decreasing order of HGA concentrations. Samples c1, c2, and c3 were from a healthy adult. HGA, hypoglycin A; ID, identification; MCPA, methylenecyclopropylacetyl; MCPF, methylenecyclopropylformy; MCPG, methylenecyclopropylglycine; ND, not determined. †Ultraperformance liquid chromatography-tandem mass spectrometry was performed with 40% of the serum volume.