Increased serum vascular endothelial growth factor is associated with acute viral encephalitis in Bangladeshi children

Encephalitis causes significant global morbidity and mortality. A large number of viruses cause encephalitis, and their geographic and temporal distributions vary. In many encephalitis cases, the virus cannot be detected, even after extensive testing. This is one challenge in management of the encephalitis patient. Since cytokines are pivotal in any form of inflammation and vary according to the nature of the inflammation, we hypothesized cytokine levels would allow us to discriminate between encephalitis caused by viruses and other aetiologies. This pilot study was conducted in a tertiary care hospital in Dhaka, Bangladesh. Viral detection was performed by polymerase chain reaction using patient cerebrospinal fluid. Acute phase reactants and cytokines were detected in patient serum. Of the 29 biomarkers assessed using the Wilcoxon rank-sum test, only vascular endothelial growth factor (VEGF) was significantly higher (P = 0.0015) in viral-positive compared with virus–negative encephalitis patients. The area under the curve (AUC) for VEGF was 0.82 (95% confidence interval: 0.66–0.98). Serum VEGF may discriminate between virus-positive and virus-negative encephalitis. Further study will be needed to confirm these findings.

Cytokines constitute promising and relevant biomarker candidates for detecting viral encephalitis because they are central to any inflammatory process, and encephalitis is a brain parenchyma inflammation caused by infection, post-infection complication, or an autoimmune process 23 . Cytokines levels usually exhibit remarkable association with certain human diseases, which highlights their potential utility as encephalitis biomarkers. Many central nervous system (CNS) infections induce common pro-inflammatory cytokines, including interferon-γ (IFN-γ), tumour necrosis factor (TNF)-α, IFN-γ induced protein (IP)-10, monokine induced by gamma interferon (MIG) 23 . The cytokine spectrum in CNS-infections varied in different studies possibly due to the differences in etiology, age, stage of infection, degree of inflammation, source of sample tested [Cerebrospinal fluid (CSF), serum, plasma etc.], or method used for measuring cytokines (commercial kits versus in-house tests, standards used etc.). The pattern of cytokine elevation may differ depending on the type of viral encephalitis. For example, IFN-γ and TNF-α are elevated in HSV encephalitis, whereas interleukin (IL)-1b is increased in enterovirus encephalitis 23 . However, IFN-γ is not elevated in influenza encephalopathy and human herpesvirus (HHV)-6 23 .
To date, no cytokines that can differentiate between virus-positive and -negative cases have been identified. Such a test would likely have a considerable impact on patient management. Therefore, we hypothesized that differences between certain cytokines may differentiate between encephalitis induced by viruses compared with encephalitis induced by other causes. The purpose of the present pilot study was to identify cytokines and determine the serum levels that would allow us to discriminate between virus-positive and-negative encephalitis cases.

Patients.
A total of 120 children with acute encephalitis were enrolled in the study. CSF and serum samples were able to be collected from 102 children at the acute stage of infection. However, complete demographic and clinical data only were available for 97 samples; therefore, these were used in the present study. The male to female ratio was 1.7:1. The children ranged in age from 2-114 months, with a mean age of 46.9 months.
Virus detection in CSF and identification. A total of 18 (18.5%) infection causes were identified (Table 1) Analyses of serum cytokines. The total numbers of samples tested were 93 for C-reactive protein (CRP), 51 for procalcitonin, and 37 each for the other cytokines. The median and the 25th and 75th percentile values of the concentration of each biomarker in virus-positive and virus-negative samples are presented in Table 2. Regulated on activation, normal T cell expressed and secreted (RANTES) was excluded from calculations because levels were the same across samples. We believe sample dilution were not appropriate, disallowing confident measurement within the limit. Hierarchical clustering depicted as a heatmap showed that virus-positive and virus-negative samples separated into two distinct clusters. A dendogram showed that, except for vascular endothelial growth factor (VEGF), all cytokines separated into four clusters (Fig. 1), indicating that the dynamics of only VEGF may have been unrelated to the dynamics of the other cytokines. Furthermore, of the 29 biomarkers assessed using the Wilcoxon rank-sum test, only VEGF was significantly (P = 0.0015) associated with viral infection status, with a higher median in virus-positive samples (304.9 pg/ml) than in virus-negative samples (156.0 pg/ml).
Receiver operating characteristic (ROC) curve analysis was conducted to obtain the area under the curve (AUC) and validate the ability of VEGF to discriminate between virus-positive and virus-negative samples. The overall AUC for VEGF was 0.82 (Fig. 2) estimated optimal cut-off value (209.8 pg/ml) for this marker, the Youden index was 0.60, sensitivity was 85%, specificity was 75%, and the AUC was 0.80.

Discussion
Using only clinical features without laboratory support, the specific etiologic agents of encephalitis cannot be determined. Furthermore, the microbiology of encephalitis is not static, and findings from one study period may not be reproducible among the same population in future years 24 . Therefore, a test that can differentiate between virus-positive and virus-negative encephalitis may be valuable at the initial stage of treatment until a specific cause can be identified in the laboratory. Among the currently available CSF biomarkers, pleocytosis, neopterin, and oligoclonal bands are less sensitive and non-specific 23 . MRI abnormalities and evidence of viral infection by PCR or serology may aid in the diagnosis of viral encephalitis 23 , but some of these are time consuming or beyond the reach of clinicians in resource-poor settings. Although acute-phase reactants such as CRP and procalcitonin are increasingly used to identify infections, we could not find significantly higher levels of serum CRP and procalcitonin in virus-positive compared with virus-negative encephalitis cases. Similar results questioning the value of erythrocyte sedimentation rate (ESR) and procalcitonin in meningitis diagnosis have been previously reported 23 . Many CNS infections induce common pro-inflammatory cytokines such as Th1-related (predominantly IFN-γ, TNF-α, IP-10, MIG) and other cytokines, including interleukin (IL)-1ra, IL-1b, IL-6, and IL-8, are often elevated, suggesting activation of the lymphocytes in the CNS that participate in viral clearance 23 . In the present study, these cytokines were also elevated compared with normal physiological levels (http://www.bio-rad.com/ webroot/web/pdf/lsr/literature/Bulletin_6029.pdf), indicating inflammation and host immune reaction to disease. The novel finding of this study is that VEGF serum level is significantly higher in patients with virus-positive compared with virus-negative encephalitis.
It remains unclear whether the VEGF increase in viral encephalitis is induced by viruses as a part of the pathogenesis. However, in vitro, VEGF increases the permeability of brain microvascular endothelial cell monolayers by downregulating tight junction proteins, disrupting cell-cell adhesion, and inducing endothelial fenestrations 25 . In addition, it remains unclear whether increased VEGF during viral encephalitis is a host induced protective mechanism for blood-brain barrier disruption because VEGF also plays important roles in wound healing and tissue cytoprotection by stimulating vascular angiogenesis, permeability, and remodelling 26 . To our knowledge, the utility of serum VEGF in differentiating between virus-positive and virus-negative encephalitis has not been previously reported. However, compared with serum from uninfected controls, VEGF has been reported to be significantly increased in tick-borne encephalitis 27 , which is consistent with our observation. Other evidence indicates that VEGF in CSF significantly increases in bacterial meningitis, including active tubercular meningitis 28,29 .
One limitation of our study was that, due to the insufficient number of samples we were unable to determine the level of cytokines in CSF. Therefore, the dynamics of VEGF in CSF during encephalitis remain unclear. Contrary to our finding of viral encephalitis, serum VEGF is decreased in bacterial meningitis 28 . Another limitation was that we did not use a next-generation sequencer to detect novel viruses. However, we expect that the number of novel viruses among the total cases would have been too small to affect our overall findings 30 .

Patients.
We conducted an observational, non-interventional study of paediatric patients from April 2010 through August 2012 at the Institute of Child and Mother Health (ICMH) Hospital in Matuail, Dhaka, Bangladesh. This tertiary care hospital serves approximately 4.8 million people in its catchment area and has 200 beds, among which, 85 are dedicated to paediatrics. The study was approved by the ICMH ethics committee. All experiments were performed in accordance with relevant guidelines and regulations. Children with encephalitis admitted to the paediatric ward during the study period were enrolled. Encephalitis was defined as presence of fever, convulsion, and unconsciousness with or without signs of meningeal irritation. Samples were collected from children whose guardians provided verbal consent for their participation. Only samples remaining after the completion of routine tests were used for this study.
Routine laboratory investigations. CSF samples were subjected to macroscopic examination, total and differential white blood cell (WBC) counts, bacterial culture, Gram staining and measurement of protein and glucose levels. Blood was cultured for bacteria and examined for total and differential WBC counts, ESR, and haemoglobin and CRP levels.

CSF virus and bacteria detection and identification.
Genomic DNA and RNA were extracted using a QIAmp viral RNA mini kit (Qiagen Company Ltd., Tokyo, Japan) and Trizol reagent (Invitrogen, Carlsbad, CA, USA) according to the manufacturers' instructions.

Statistical analysis.
The concentrations of 27 of the 29 biomarkers tested were non-Gaussian distributed.
Therefore, we summarized the concentrations as the median and the 25th and 75th percentile values and used the nonparametric Wilcoxon rank-sum test to assess whether the concentration of acute phase reactants and cytokines differed significantly by viral infection status (positive vs. negative). For Wilcoxon rank-sum tests, we considered a difference significant if the obtained P-value was smaller than the Bonferroni-adjusted α of 0.0017.
The ROC curve was used to obtain the AUC in order to evaluate whether each cytokine can be used to make a distinction between the two etiological groups. In addition, for significant candidates, the Youden index was calculated to identify the point of maximum sensitivity and specificity and to describe its potential effectiveness. All statistical analyses were performed using Stata 14.0 (StataCorp LP, College Station, TX, USA).