Deep proteomics network and machine learning analysis of human cerebrospinal fluid in Japanese encephalitis virus infection

Japanese encephalitis virus (JEV) is a mosquito-borne flavivirus, and leading cause of neurological infection in Asia and the Pacific, with recent emergence in multiple territories in Australia in 2022. Patients may experience devastating socioeconomic consequences; JEV infection (JE) predominantly affects children in poor rural areas, has a 20-30% case fatality rate, and 30-50% of survivors suffer long-term disability. JEV RNA is rarely detected in patient samples, and the standard diagnostic test is an anti-JEV IgM ELISA with sub-optimal specificity; there is no means of detection in more remote areas. We aimed to test the hypothesis that there is a diagnostic protein signature of JE in human cerebrospinal fluid (CSF), and contribute to understanding of the host response and predictors of outcome during infection. We retrospectively tested a cohort of 163 patients recruited as part of the Laos central nervous system infection study. Application of liquid chromatography and tandem mass spectrometry (LC-MS/MS), using extensive offline fractionation and tandem mass tag labelling, enabled a comparison of the CSF proteome in 68 JE patient vs 95 non-JE neurological infections. 5,070 proteins were identified, including 4,805 human proteins and 265 pathogen proteins. We incorporated univariate analysis of differential protein expression, network analysis and machine learning techniques to build a ten-protein diagnostic signature of JE with >99% diagnostic accuracy. Pathways related to JE infection included neuronal damage, anti-apoptosis, heat shock and unfolded protein responses, cell adhesion, macrophage and dendritic cell activation as well as a reduced acute inflammatory response, hepatotoxicity, activation of coagulation, extracellular matrix and actin regulation. We verified the results by performing DIA LC-MS/MS in 16 (10%) of the samples, demonstrating 87% accuracy using the same model. Ultimately, antibody-based validation will be required, in a larger group of patients, in different locations and in field settings, to refine the list to 2-3 proteins that could be harnessed in a rapid diagnostic test. Author summary Japanese encephalitis virus (JEV) is a leading cause of brain infection in Asia and the Pacific, with recent introduction in multiple territories in Australia in 2022. Patients may experience devastating socioeconomic consequences; JEV infection (JE) predominantly affects children in poor rural areas, has a 20-30% case fatality rate, and 30-50% of survivors suffer long-term disability. The disease is difficult to diagnose, and there are no rapid tests that may be performed in remote areas that it exists such that we remain unclear of the burden of disease and the effects of control measures. We aimed to apply a relatively novel method to analyse the proteins in patients with JE as compared to other neurological infections, to see if this could be useful for making a diagnosis. We tested the brain fluid of 163 patients recruited as part of the Laos central nervous system infection study. We used a method, ‘liquid chromatography mass spectrometry’ that does not require prior knowledge of the proteins present, that is you do not target any specific protein. Over 5,000 proteins were identified, and these were analysed by various methods. We grouped the proteins into different clusters that provided insight into their function. We also filtered the list to 10 proteins that predicted JE as compared to other brain infections. Future work will require confirmation of the findings in a larger group of patients, in different locations and in field settings, to refine the list to 2-3 proteins that could be harnessed in a rapid diagnostic test.

(1, 2). It is an emerging disease, with recent evidence of JEV in multiple territories in Australia (3).

104
Patients may experience devastating socioeconomic consequences; JE predominantly affects children 105 in poor rural areas, has a 20-30% case fatality rate, and 30-50% of survivors suffer long-term 106 disability (4). Although no specific treatment is available, several vaccines are available and 107 recommended by the WHO (5, 6). Although recent efforts have strengthened JEV vaccination 108 programs, still only 15 of 24 endemic countries include JEV vaccine in routine immunisation 109 policies, and even then, it is not uniformly nationwide, with vaccine coverage in targeted areas 110 reported to be as low as 39% (7). JEV is a zoonosis, and sustained vaccine coverage is essential to 111 control disease.

112
A fundamental limitation in the control of JE is the poor accuracy of existing diagnostic tests, 113 requirement for lumbar puncture and laboratory capacity for diagnosis (8) . Surveillance data suggest 114 that only 11 of 24 countries meet the minimum surveillance standards, equivalent to diagnostic 115 testing in a sentinel site (7). This is a threat to vaccine implementation, as accessible and accurate 116 diagnostics are essential to understand epidemiology, effectiveness of vaccination, identify associated 117 research knowledge gaps and facilitate public engagement. This also has implications for appropriate 118 risk-assessment for travellers. Aside from JEV control, diagnosis is crucial for patients, families and 119 health-workers, to be able to institute appropriate supportive and rehabilitation care, stop unnecessary 120 antibiotics, or if the test is negative to prompt further investigation.

121
The gold-standard JEV test is a neutralisation assay. However this requires paired acute and 122 convalescent sera, is laborious, time-consuming, requires specialist skills, high-level isolation with neutralisation assays. The manufacturer of the only available commercial kit for clinical 127 diagnosis (InBios) quotes a sensitivity of >90% for well-characterised CSF samples, but sensitivity in 128 the field is as low as 53% (9). There are also increasingly recognised problems with specificity 129 related to prior vaccination and cross-reactivity with other flaviviruses (10,11). Reported specificity is 130 >90%, however a study by our group demonstrated that 13% of patients with JE IgM detected in CSF 131 by MAC-ELISA had another pathogen detected that may have explained the presentation (10).

132
Detection of JEV RNA would be highly specific, but the period of viraemia is brief and hard to 133 capture clinically, often occurring before the onset of neurological symptoms and signs. RT-qPCR 134 remains insensitive irrespective of the analytical sensitivity or gene targets (12) Table 2 and ROC in supplementary data S12. 411 412   predicting outcome using the Boruta algorithm and 2 proteins using Lasso, such that 2 proteins were 427 identified by both Boruta and Lasso, see supplementary data S13. In view of the small sample size, 428 the data were not split into a training and test set. These proteins were used to train different models 429 with five-fold CV repeated ten times evaluated on ROC, and then combined in an ensemble model 430 with cross-validation scores reported in Table 2, see the list of proteins in supplementary data S13 and 431 ROC in S14. There were five JE patients in the DIA LC-MS analysis of which 3 had outcome data, 432 and this was considered too small to report test metrics.