Translational evidence for two distinct patterns of neuroaxonal injury in sepsis: a longitudinal, prospective translational study

All experiments were performed according to local ethics committee (University College London) and Home Office (UK) guidelines under the 1986 Scientific Procedures Act. Adult male Wistar rats (approximate body weight 300 g, 12–14 weeks old) were used to generate a 3-day in vivo sepsis model of fecal peritonitis. Sepsis induction and experimental procedures were performed as described in detail before [25]. Four naive (noninstrumented) rats, eleven sham-operated, and fourteen septic rats were included. All rats were instrumented under brief anesthesia with tunneled carotid arterial (left side) and jugular venous (right side) catheters to monitor hemodynamics, sample blood, and infuse fluids. The catheters were mounted onto a swivel-tether system, allowing the rat, on recovery from anesthesia, to have free movement in its cage and ad libitum access food and water. Sepsis was induced 24 h later by an intraperitoneal injection of fecal slurry (0.63 mg/100 g body weight, prepared from bowel contents of rats from the same batch). From 2 h postinsult, fluid resuscitation using a 1:1 solution of colloid and 5% glucose was administered at a rate of 10 ml/h for the first 24 h and then halved on successive days to ensure normovolemia and normoglycemia. All septic animals showed clinical signs of sepsis by decreased movements, decreased alertness, hunched posture, and piloerection from about 12 h postinjection of fecal slurry. Rats were killed by cervical dislocation under deep isoflurane anesthesia either 24, 48, or 72 h after sepsis induction. After craniectomy, brains were removed within a few minutes of death. The brain was dissected through the midline into halves. One half was snap-frozen in liquid nitrogen; the other half was placed in formalin.

Neuropathologists (MG, FS) were masked to the condition of rat and human brain samples. Human and rat brain tissue from frontal lobe areas and the cerebellum were processed in paraffin wax using a standard 7-day dehydration and paraffin-embedding protocol on an automated tissue processor. Five-micrometer paraffin sections were cut, mounted onto Superfrost^TM glass slides and dried overnight at 37 °C. For the general histologic examination, sections were dewaxed, rehydrated, and stained with H&E using a standard protocol. For the immunohistochemical detection of neurofilaments (Nf) and glial fibrillary acidic protein (GFAP), the sections were dewaxed and rehydrated before being placed in 600 ml of 0.1 M citrate buffer (pH 6.0) and microwaved at full power for 15 minutes in a 850-W microwave oven. The four sections were allowed to cool before being rinsed in 0.05 M PBS, pH 7.4, and incubated overnight at room temperature in either a mouse monoclonal antibody directed against an epitope common to the 70 and 200 kDa Nf proteins (clone 2 F11 diluted 1:20; MP Biomedicals Inc., Santa Ana, CA, USA), β-tubulin (1:200; Sigma-Aldrich, Gillingham, UK), β-APP (1:500; Dako, Ely, UK), or a rabbit polyclonal antibody directed against GFAP (diluted 1:1500; Dako). After being washed in PBS, the sections were incubated in a biotinylated secondary antibody for 1 h (Dako), followed by a washing step and incubation in peroxidase-conjugated streptavidin for 1 h (diluted 1:300; Sigma-Aldrich) All dilutions were in PBS with 0.1% Triton-X. Antibody localization was visualized by incubating the sections for 10 minutes at room temperature in 0.05% diaminobenzidine with 0.04% NiCl₂ and 0.01% hydrogen peroxide added. The sections were then counterstained with hematoxylin, dehydrated, cleared, and mounted.

The dry weight of the snap-frozen rat brain tissue was 0.56 to 1.94 g. Barbitone ethylenediaminetetraacetic acid (EDTA) buffer (pH 9.6) containing a protease inhibitor cocktail (P8340; Sigma-Aldrich) was added to 1:2 wt/vol. The samples were homogenized on ice using an ULTRA-TURRAX T 25 instrument (IKA-Werke GmbH & Co., Staufen, Germany) for 1 minute, followed by sonication on ice for 1 minute. Samples were refrozen at −70 °C and then thawed at 30 °C, and 2 ml of sample was added to 5 ml of diisopropyl ether and 2 ml of barbitone EDTA buffer. After a mixing step, the samples were spun at 25,000 × g and 4 °C for 30 minutes, and the protein soluble fraction was collected. Tissue levels of Nf heavy chain (NfH^SMI35) and GFAP were measured by enzyme-linked immunosorbent assay (ELISA), and total protein was measured using the Lowry method [26, 27].

The study was approved by the local ethics board at Rostock University (A 2012-0058) and registered as a clinical trial (ClinicalTrials.gov, NCT02442986). The patient recruitment period was from November 2012 to May 2015. All patients or their legal representatives signed written informed consent forms before study inclusion. Inclusion criteria for participants were aged ≥ 18 years and the presence of severe sepsis or septic shock according to the criteria used at the time [28]. Exclusion criteria were preexisting cerebrovascular diseases, including dementia, preexisting neuromuscular disease, high-dose glucocorticoid administration (>300 mg hydrocortisone or equivalent per day), preexisting renal replacement therapy, coagulopathy with active bleeding, and frequent administration of neuromuscular blocking agents (more than three times per week). Twenty patients with septic shock were included prospectively in the study. Seven participants without magnetic resonance imaging (MRI) examinations were excluded for the following reasons: death before MRI performed (n = 1), only cranial computed tomographic scan available owing to contraindication for MRI (n = 2), disclaimer for MRI from patient/legal representative after study inclusion (n = 2), and repeated surgery/unstable patient (n = 2). In total, 13 patients were enrolled prospectively in this single-center, longitudinal, observational study.

All statistical analyses were performed using SAS software (version 9.4; SAS Institute, Inc., Cary, NC, USA). Because of non-Gaussian distribution, the nonparametric Wilcoxon test was used for comparing two independent variables. Two-way unbalanced analysis of variance (general linear model) was used for comparing more than two independent variables, followed post hoc analysis if significance was achieved. The F values providing the degrees of freedom and the number of samples included in each particular analysis are shown. The linear correlation between continuous variables was evaluated using the Spearman correlation coefficient. Linear regression analysis was performed using the least squares method. A p value < 0.05 was considered significant.

Average total protein levels were comparable between groups (naive 7.4 ± 2.6 g/L, sham 8.7 ± 3.9 g/L, and sepsis 8.9 ± 2.9 g/L). We found that brain tissue levels for GFAP were not statistically different when we compared sham-treated (0.27 ± 0.19 ng/g total protein) and septic (0.29 ± 0.21 ng/g total protein) rats with naive rats (0.34 ± 0.11 ng/g total protein). Average brain tissue levels of NfH^SMI35 were higher in sham-treated (2.6 ± 2.2 ng/g total protein) and septic (1.8 ± 1.7 ng/g total protein) rats than in naive rats (0.8 ± 0.6 ng/g total protein), but this difference failed to reach statistical difference (p = 0.094 and p = 0.356, respectively) (Table 1). In septic rat brain tissue, there was a mild correlation between NfH^SMI35 levels and time from sepsis induction (R = 0.53, p = 0.045) that was not seen for either GFAP (R = −0.39, p = 0.154) or total protein (R = −0.05, p = 0.854). No such correlations were observed in either sham-treated or naive rat brain samples. The immunohistochemistry of brain tissue from septic compared with naive and sham-treated rats showed two types of lesions. The normal white matter appearance in a sham-treated rat brain is shown in Fig. 1a. Typically, βAPP staining is restricted to the neuronal cell soma and the proximal axonal hillock. For comparison, in septic brain tissue, diffuse, more widespread axonal staining is seen (Fig. 1b). This pathologic staining was most marked for long white matter tracts (Fig. 1c). Spinal tissue was not available. The second lesion type seen comprised scattered small ischemic lesions (Fig. 1d). The early inflammatory component suggests a septic embolic etiology. These pockets of ischemic, inflammatory lesions stained intensely for βAPP, extending peripherally from the core lesion (Fig. 1e). An alternative marker for early axonal injury is β-tubulin. In sham-treated rats, the axonal staining for β-tubulin is neat and continuous, as expected on the basis of preserved neuronal (Fig. 1f) and axonal (Fig. 1g) integrity. This is best appreciated in the magnified insets in Fig. 1g and h. In contrast to the diffuse axonal pathology present in septic rat brains, axonal β-tubulin is subjected to proteolysis and broken up, giving the impression of a structurally disorganized axonal (Fig. 1h) and neuronal cytoskeleton (Fig. 1i).

Postmortem brain tissue was available from five patients who died as a result of sepsis (mean age 64 years, one male patient). The three control subjects (mean age 37 years, all male) died as a result of other causes: suicide, assault, and road traffic accident (RTA) without concurrent brain injury.

Causes leading to death in sepsis were directly related to multiorgan failure, but one patient had additional complications in the form of a gastrointestinal and intracranial hemorrhage. Careful examination of this patient’s brain did not reveal any evidence of an amyloid angiopathy. The immunohistochemical results of brain tissue from the patients with sepsis are summarized in Fig. 2. Patient 1, a 56-year-old male control subject, died after having an RTA. There was diffuse staining of axons for βAPP restricted to the axonal hillock (Fig. 2a, arrows). More extensive axonal βAPP staining was seen in patient 2, a 55-year-old male patient with sepsis (Fig. 2b). There was beadlike swelling indicating axonal pathology. In addition, staining for nonphosphorylated NfH (SMI32) demonstrated the presence of a large amount of axonal endbulbs, a sign of axonal degeneration (Fig. 2c). Next, scattered ischemic lesions were visible in patient 3, a 67-year-old female who died as a result of multiorgan failure due to sepsis (Fig. 2d). Amyloid plaques were found only in a 79-year-old female (Fig. 2e). Her medical history was not suggestive of cognitive impairment such as that seen with a neurodegenerative dementia; however, a formal neuropsychological assessment done before occurrence of sepsis was not available. Of note, the degree of diffuse deep white matter axonal injury was the most severe of all patients with sepsis (Fig. 2f).