Between December 2017 and April 2018, 38 YF cases, defined by at least one of the Brazilian Ministry of Health definition criteria—isolation of the YF virus, detection of the viral genome, detection of IgM class antibodies by the MAC-ELISA technique in unvaccinated individuals, or with a fourfold or greater increase in antibody titers by the hemagglutination inhibition (HI) technique, in paired samples [24], underwent autopsy, at University of Sao Paulo, after informed consent from first-degree relatives or legal representatives. The cases obtained were those whose autopsy had been requested by the attending physicians during the YF epidemic. The autopsies, including brain removal, followed established protocols [25, 26] and were performed by a pathologist (FPF). Of these cases, 34 were men aged between 16 and 71 years, and four were women aged between 31 and 74 years old (mean age = 48.35 years). Thirty-one presented acute neurologic disturbances reported in medical records—drowsiness, confusion, lethargy, agitation, reduced level of consciousness and/or coma. In three of them the descriptive term “encephalopathy” was used, unspecified, and in another three the designation “hepatic encephalopathy” (HE) was used. Convulsive episodes were present in 10 of them. In seven cases, detailed neurological changes were not described in the medical records.

It was possible to perform at least one electroencephalography in 21 YF cases. In 13 of them, the patient was in a coma (spontaneous or induced by sedation). In one case who had no clinically reported seizures, epileptiform paroxysms were identified. In another case alterations suggestive of toxic-metabolic and infectious encephalopathies (occasional paroxysms of acute waves, of generalized projection, outlining a triphasic aspect) were described in a report, although he did not have a detailed description of neurological signs and symptoms during evolution.

In the evaluation by in vivo computed tomography exams, performed in nine cases, three presented diffuse brain edema, one of them after a seizure episode, another presenting, in addition to edema, signs of transforaminal herniation.

All cases had as final cause of death fulminant hepatitis. Of these, 12 had an established, co-concurrent septic shock related to secondary infections, according to definition criteria [27], with at least one infectious agent isolated in cultures – Staphylococcus spp, Klebsiella pneumoniae, Acinetobacter spp, Pseudomonas spp, Stenotrophonomas maltophilia, Escherichia coli, Micrococcus spp, Listeria monocytogenes, Candida krusei.

For means of comparison, 21 individuals from the autopsy routine, aged between 18 and 74 years (mean age = 52 years), without YFV infection were selected by chance as controls, 10 of them with infection and septic shock, information obtained from clinical data and/or from autopsy material. This selection was based on the availability of cases without YF in the period, whose autopsy was performed after informed consent from family members or legal representatives. The inclusion and exclusion criteria of cases and controls are described in details [see Additional files 1].

After brain removal, in 33 cases 100 mm2 of the right frontal pole was sectioned and stored for reverse transcription-polymerase chain reaction (qRT-PCR) assays to detect YFV-RNA. Briefly, the tissue samples were macerated and nucleic acid extraction was performed using the TRIzol® reagent (Life Technologies). Molecular detection of YF virus was performed with the use of the AgPath-ID One-Step qRT-PCR Reagents (Ambion, Austin, TX, USA) with specific primers/probes previously described [9]. qRT-PCR reactions consisted of a step of reverse transcription at 45 °C for 10 min for enzyme activation, at 95 °C for ten minutes, and 40 cycles at 95 °C for 15 s and 60 °C for 45 s for hybridization and extension with the use of ABI7500 equipment (Thermo Fisher Scientific, Waltham, MA, USA).

Then, the brains were stored in buffered formaldehyde solution for twenty-one days, after which they were sectioned in 10 mm thickness slices. One sample on each side (symmetrical) were obtained from the frontal, parietal, occipital, temporal (mesial and lateral), basal nuclei, thalamus, centrum semiovale and cerebellar hemispheres, and one sample of each of vermis, midbrain, pons, and medulla oblongata. The samples, 30 mm large and 5 mm thick, were transferred to a histological cassette for further processing. Briefly, the samples were dehydrated in alcohol, cleaned with xylene and paraffin-embedded. The paraffin was then removed, the blocks were cooled and 5 µm-thick slices were obtained in the microtome. The slices were transferred to slides in a warm bath, dried and hematoxylin and eosin staining were performed.

Cases and controls were evaluated for perivascular and parenchymal edema, eosinophilic neuronal alterations, chronic small vessels changes, parenchymal hemorrhage and inflammatory infiltrate, microglial nodules, and periventricular inflammatory infiltrate, as absent (grade 0) or present (grade 1). Perivascular hemorrhage, hemosiderin deposition, and inflammatory infiltrate, meningeal hemorrhage and inflammatory infiltrate were semi-quantitatively evaluated as absent (grade 0) or present (grades 1 to 3, detailed in Table 1), an approach similar to that undertaken by Sharshar and colleagues to assess the neuropathology of septic shock [28].

For the statistical analysis four groups were established: YF cases without septic shock (n = 26), YF cases with septic shock (n = 12), controls without septic shock (n = 11), and controls with septic shock (n = 10). For each YF case and each control, 13 graduation values representing the 22 mapped areas (only the highest value being considered when bilateral) were obtained. For comparison purposes, YF cases and the control groups were paired separately, and then all YF cases were compared with all controls. p values < 0.05 were considered statistically significant.

To verify the existence of difference (or relationship) between the YF cases groups, between the control groups, and between the YF cases and the controls, we applied the statistical tests Fisher's exact test and chi-square.