Patients
The study was approved by the ethics committee of the Oulu University Hospital and the Northern Ostrobothnia Hospital District (224/2013). The use of brain tissue samples obtained in the autopsy was approved by the National Institute for Health and Welfare (BB_2017_1003). This retrospective observational cohort study was conducted at the Oulu University Hospital, Oulu, Finland, an academic tertiary level referral hospital. Adult patients deceased with sepsis who underwent postmortem examination during the years 2007–2015 and had brain tissue specimens available were included in the study. Medical non-forensic autopsies are performed if antemortem data or clinical examination does not provide the clinician enough information to issue a death certificate and to better understand the disease course, or in cases in which the next of kin request it. The final decision is made at the discretion of the treating physician. All patients were treated by a multidisciplinary team of intensivists and infectious disease specialists in our 26 beds, closed adult intensive care unit. Sepsis was defined according to the American College of Chest Physicians/Society of Critical care medicine criteria [
14]. The intensive care treatment was performed according to normal intensive care unit (ICU) protocols and sepsis guidelines.
Clinical data collection
Clinical parameters were gathered from the ICU clinical data management system database (Centricity Critical Care, Clinisoft; GE Healthcare, Helsinki, Finland). On admission, the severity of illness was determined by the Acute Physiology and Chronic Health (APACHE II) score [
15], and daily total and individual organ group sequential organ failure assessment score (SOFA) was used as a measure of organ dysfunction [
16]. The Glasgow Coma Score (GCS) was employed for assessing the impairment of patients’ level of consciousness [
17]. The presence of multi-organ failure (MOF) during the ICU treatment was defined as more than two organs failing based on grade 3 or 4 sequential organ failure assessment [
16]. Data regarding age, sex, time to death, focus of infection, and blood culture positivity were collected. Procalcitonin (PCT) on admission and its highest value, as well as white blood cell count were retrieved. Clinical laboratory samples were analyzed by commercially available laboratory methods in the hospital accredited central laboratory (NordLab, Oulu University Hospital, Oulu, Finland).
Immunohistochemical stainings
The median time from the death to autopsy was 5 (3–6) days. At the autopsy, brain specimens were routinely collected, usually from a least two anatomical locations. A systematic neuropathological examination was performed when the pathologist in charge considered it necessary or by request of the clinician, for example, in cases when the patient had neurological symptoms. For each deceased subject, brain tissue specimens were categorized according to anatomical location (cerebrum, cerebellum), and one representative tissue block from each location was selected for the study. The exact location of the specimens had not been registered in all cases, but according to the suggested sampling routine of our hospital, likely followed in most cases, cerebral samples represented the precentral gyrus, and the cerebellar samples the lateral aspect of either hemisphere. The immunohistochemical stainings were performed in accordance with the manufacturer’s recommendation, and a set of samples was used to optimize the dilution of primary antibodies. We used the following antibodies and conditions: (1) human occludin (rabbit polyclonal, catalog no: 711500, Invitrogen, Frederick, MD, USA), with pretreatment using pronase, at a dilution of 1:800, incubation at room temperature (RT) 60 min; (2) ZO-1 (rabbit polyclonal; Zymed, cat no: 61-7300, Carlsbad CA, USA), pretreatment with 15 min boiling in a TRIS-EDTA buffer, dilution at 1:400, incubation for 60 min at RT); and (3) claudin-5 (mouse monoclonal, clone 4C3C2, Zymed), pretreatment with TRIS-EDTA, dilution at 1:50, incubation for 60 min at RT. For all antibodies, the detection was performed with a polymer-based kit (Envision, Dako, Copenhagen, Denmark). Diaminobenzidine (Dako basic DAB-kit, Dako) was used as a chromogen. All stainings were performed with the Dako Autostainer (Dako, Copenhagen, Denmark). Validation of our immunohistochemical analysis was performed through a series of negative controls by omitting the primary antibody.
Evaluation of immunohistochemical stainings
Stained sections were digitized (Leica-Aperio AT2; Leica Biosystems, Nussloch, Germany) and assessed using the Aperio ImageScope program independently by two researchers (KE and ME) supervised by an experienced neuropathologist (HT), all blinded for the clinical and outcome data. Endothelial cells of the capillaries were analyzed for each anatomic location. Expression was considered positive if 50% or more showed a positive reaction of any degree and were negative in the absence of any staining or if less than 50% showed a positive reaction.
Statistical analysis
Statistical analyses were performed with SPSS for Windows (2017 release, Version 25; IBM Corporation, Armonk, NY, USA). Data are expressed as the percentage of stained cells as median (25–75th percentile) in continuous variables. Categorical data were analyzed with Fisher’s exact test. Mann-Whitney U test was applied to distributions across the two groups. Spearman’s correlation coefficient (rho) was calculated. Two-sided p values are presented and a p < 0.05 was considered statistically significant. However, the level of statistical significance should be treated with caution, given the large number of statistical tests performed.