GFAP positivity in neurons following traumatic brain injuries

This study re-evaluated digitalized photographs of brain tissue samples taken from 50 TBI fatalities and 16 controls collected for our study on astrocytic GFAP expression following traumatic brain injuries (for individual case details see Trautz et al. [5]). In brief, human brain samples were retrieved during forensic autopsies; afterwards, they were submerged in neutral buffered 4% formaldehyde, which was changed once after 24 h. The duration of fixation was up to 3 weeks before embedding in paraffin. The control samples were collected from individuals who died of sudden natural cardiovascular causes of death with absence of acute or former injuries of the central nervous system. To investigate a potential region dependency of immunopositivity for GFAP in neurons within the brain in TBI fatalities, samples from the following four brain sites were evaluated: (i) the cerebral area (cortex and adjacent white matter) next to the either microscopically (haemorrhages with abundant necrotic tissue) or macroscopically verified contusions, (ii) a cerebral region (cortex and adjacent white matter) opposite to the contusion called the contralateral cortex (CLC), which was microscopically and macroscopically free of contrecoup haemorrhages, (iii) the CA4 region of the hippocampus (HC) and (iv) the cerebellum (CB). Hippocampal and cerebellar samples were taken from regions ipsilateral to the cerebral contusion as also performed previously [5, 12, 15].

Using semiquantitative analyses, the contusions were located in the prefrontal, temporoparietal and occipital regions of the brain in 22, 23 and 5 TBI fatalities, respectively. For controls, samples were retrieved from the following brain regions, which varied from TBI fatalities as no impact zone was present: (i) the frontal cortex and adjacent white matter (FC), (ii) the ipsilateral CA4 region of the HC and (iii) the ipsilateral CB.

In line with previous studies [5, 12], the TBI cases were further categorized into the following three groups according to their survival time following the traumatic event: an acute group (survival time < 2 h), a subacute group (survival time 2–72 h) and a delayed group (survival time > 72 h). The Ethics Committee of the University of Leipzig, Germany approved the study (local number 117–12-23,012,012).

The brain samples allocated to the subgroups and their respective sample characteristics are depicted in Table 1. Routine H&E stains were available to assess the overall morphology and potential trauma-related or inflammatory changes. The immunolabelled slices were stained with a polyclonal rabbit GFAP antibody (catalogue number Z0334; Agilent Dako, Santa Clara, USA), dilution 1:4000, performed with a fully automated immunostainer (Benchmark XT; Roche, Basel, Switzerland) using DAB as standard chromogen. For each brain region per case, a set of ten coherent “high-power field” (HPF) images was assessed, resulting in a standardized investigation area of 2.37 mm² per region [25]. The commercially available object counting application SIGMA (open source, developed by Karen Grigoryan) was used to evaluate the slides in all ten digital HPF images at × 200. Only cells with a distinct cellular immunostaining which were identified as neuronal in origin were counted to ensure a repeatability of the results. Here, nucleus morphology, nucleolus location and cell boarders were used to differentiate neurons from non-neuronal cells as established morphological criteria. After immunopositivity for GFAP was noted in the neurons, the consecutive slides were stained analogously per immunolabelling and counterstained with haematoxylin.

The overall staining intensity of the section was rated according to the scoring system of Remmele and Stegner [26] to relate the observed results of this study to the staining quality of the brain tissue. The scoring system implemented the following grades: (0) absence of noticeable staining, (1) slight staining with the marker being predominantly concentrated on cell bodies, (2) moderate staining with incidental spots that are more intensively coloured and some blunt processes and (3) strong staining with predominantly condensed central granulations and widespread processes. The final grade was generated by the arithmetic mean of the ten sections available per region.

To confirm or reject the immunohistochemical observations, an additional staining technique was performed employing immunofluorescence labelling on brain tissue of two additional TBI fatalities and one control case, which were not included in the dataset of Trautz et al. [5]. Both TBI cases were male, and were 18 and 32 years old at death. Both cases had a survival time of less than 10 min and post-mortem intervals of under 48 h. The control case was a 50-year-old male who died of an acute cardiac arrest. The procedure was performed twice using primary and secondary antibodies of different batches and manufacturers to exclude that the observations were attributed to a specific batch (or manufacturer) of antibodies (Table 2). Also, negative controls were made concomitantly to test for unspecific binding. Whereas the Agilent Dako GFAP antibody “reacts strongly with human GFAP”, the BD Pharmingen anti-GFAP cocktail “specifically detects GFAP and does not cross-react with other intermediate filaments” according to the details provided by both manufacturers. Since antibody “cocktails” are composed of monoclonal antibodies combining several targets, one could describe them as oligoclonal antibodies, combining the specificity of monoclonal and the sensitivity of polyclonal antibodies.

NeuN was used as a typical and specific neuronal marker of human brain tissue [27].

The tissue was cut into 12-μm-thick serial sections. Antigen retrieval was achieved using microwave pre-treatment in citrate buffer (pH = 6.0, 98 °C, 2 × 20 min). Tissues were washed and then blocked with 1% bovine serum albumin (BSA; A7906, Sigma-Aldrich, St. Louis, USA) in PBS-Triton (0.03%, PBS-T) to prevent further non-specific binding. Then, sections were incubated with the primary antibodies overnight in an adequate dilution in 0.5% BSA in PBS-T at 4 °C (Table 2). For the negative controls, the primary antibody was omitted and the section was incubated in 0.5% BSA in PBS-T overnight. After rinsing with PBS-T, the Alexa Fluor 488 and Alexa Fluor 568-conjugated secondary antibodies were applied and the slices were incubated for 60 min at room temperature at a dilution of 1:200. Counterstaining with 4′,6-diamidino-2-phenylindole dihydrochloride (DAPI, dilution 1:50.000; Invitrogen, Waltham, MA, USA) was done to detect cell nuclei. Slices were then washed and cover slipped. Images were taken using an Olympus BX40 fluorescent microscope (Olympus Deutschland, Hamburg, Germany). Due to the young age of the tested individuals, no lipofuscin quenching was performed to prevent a decrease of staining intensity.

Data analysis was conducted using Prism version 8 (GraphPad Software, La Jolla, CA, USA) and Microsoft Excel version 16.16 (Microsoft Corporation, Redmond, WA, USA). The Shapiro–Wilk test was used to test the Gaussian distribution of the data. Pearson’s r and Spearman’s ϱ were reported for parametric and non-parametric data, respectively. The occurrence of immunopositivity for GFAP in neurons of any brain region per cadaver was correlated with age at death, sex, survival time and PMI. The total number of immunopositivity for GFAP in neurons was correlated with the total number of GFAP-positive astrocytes of the respective brain region (metadata in Trautz et al. [5]) and the staining score mentioned above. P-values of 0.05 or less were considered to be statistically significant.

Immunopositivity for GFAP in neurons in any of the investigated brain regions was detected in 6 out of 50 TBI cases (12%), as well as 1 (44-year-old male who died of an acute coronary insufficiency) out of 16 controls (6%). A significant and moderately positive correlation was observed between the survival time and the occurrence of immunopositivity for GFAP in neurons in any of the four investigated sub-regions from the TBI cases (r = 0.306, P = 0.015). The occurrence of immunopositivity for GFAP in neurons in any of the sub-regions did not correlate with age at death (r =  − 0.016, P = 0.451), sex (r =  − 0.013, P = 0.459) and PMI (r = 0.011, P = 0.467) on a statistically significant level for TBI cases. Age, sex and PMI correlations were not performed for control cases due to the limited sample size.

Immunopositivity for GFAP in neurons was observed in the PCZ and HC of 13 and 17% of the subacute TBI fatalities (Fig. 1). Similarly, immunopositivity for GFAP in neurons was found in 17 and 18% of the HC and PCZ in delayed TBI fatalities, respectively (Fig. 1). The CB contained immunopositivity for GFAP in neurons in 14% of subacute TBI fatalities (Fig. 1). None of the acute TBI cases or CLC regions of subacute and delayed TBI fatalities revealed any immunopositivity for GFAP in neurons (Fig. 1). Control samples showed immunopositivity for GFAP only in Purkinje neurons in 6% of cerebellar samples (Fig. 1). The two TBI samples that presented immunopositivity for GFAP in Purkinje neurons had a contusion in the prefrontal cortex area. Immunopositivity for GFAP in neurons was found in the HC following both temporal contusions (n = 2) and a prefrontal cortical contusion (n = 1). No immunopositivity for GFAP in neurons was detected in the frontal cortices or hippocampi of controls.

The low absolute number of GFAP-labelled neurons correlated only weakly and insignificantly with the number of GFAP-positive astrocytes of the respective HPF (r = 0.247, P = 0.244). Also, the adapted Remmele and Stegner staining score did not influence the absolute number of GFAP-labelled neurons significantly (r = 0.086, P = 0.410). Figure 2 shows typical GFAP-positive astrocytes, as well as immunopositivity for GFAP in representative neuronal cells in the anti-GFAP immunolabellings that were evaluated in this study.

The double immunofluorescence stains with primary anti-GFAP and anti-NeuN antibodies revealed GFAP-positive staining within vital (DAPI-positive) neuronal cell bodies, which are also positively labelled with NeuN (Fig. 3a). However, a comparable, more peripheral and granular reaction of the secondary antibody was observed in the negative control (Fig. 3b). These positive stainings were present even when another antibody batch was used for both the primary and secondary antibodies, respectively. Furthermore, the GFAP-positive stainings were detected for both of the secondary antibodies used in this study (Fig. 4). However, comparable GFAP-positive stainings were also present in the 488-channel for the negative control.

Even though remarkable evidence exists that supports the neuronal potential to synthesize GFAP mRNA [23, 28], it has to be considered that the GFAP immunoreaction might solely be an artefact and, therefore, not truly depict neuronal GFAP. Regarding the former, the strongest argument in this given study forms the observation of fluorescent neuronal staining in the negative control of the anti-GFAP stain. The immunopositivity of the anti-GFAP-negative control suggests a non-specific binding of the secondary antibody, which contradicts the observation of neuronal GFAP. Regarding this, Middeldorp and colleagues were able to show that a neuronal staining with different GFAP antibodies was caused by a cross-reaction with neurofilament-L (NF-L). This is most likely due to a homology of the GFAP peptide with several amino acids in the tail domain of NF-L [24]. Even though the antibodies used are described to have increased specificity in epitope binding, the exact epitope the antibody was binding to is unknown. Therefore, it is possible that the antibody cross-reacted with similar proteins, hence binding different epitopes than the intended GFAP ones [24]. Interestingly, in this given study, TBI cases were more likely to contain GFAP-positive neuronal deposits, similar to the increased findings in AD brains [28] and, therefore, could be a hint for the trauma-associated neuronal damage. Alternatively, it could be suspected that the observed neuronal staining reflects lipofuscin, as mentioned previously; in this study, no lipofuscin quenching was performed due to the young age of the tested individuals. A proteomics study was able to detect GFAP as a major lipofuscin protein in humans, but not in rats [29]. In addition, lipofuscin was described to have a yellow–brown or translucent appearance under visible light [30]. However, an age-related increase would be expected for lipofuscin [31], which was not observed for the neuronal GFAP staining in this study. Furthermore, the increase of GFAP-positive staining in cases with traumatic brain injury could reflect an increase in lipofuscin and, therefore, an accelerated neurodegeneration following a traumatic brain injury. The former correlation was already described in several studies before [32].

A third possible explanation is the probability of metabolic alterations, such as acidosis [33] of brain tissue after severe traumatic events. The authors hypothesize that this might alter epitopes of intraneuronal proteins, thereby increasing the likelihood of a non-specific antibody binding. However, the altered neuronal homeostasis and its respective impact on neuronal proteins might just lead to a non-specific binding of secondary antibodies, if the vital neuron was exposed to the metabolic change for at least 2 h.

GFAP is known to be a glia-specific biomarker [17, 34]. Being an intermediate filament, it is thought to contribute to cell stability [24]. Furthermore, recent studies determined GFAP plays an important role in myelination, synaptic plasticity, scar formation or the integrity of the blood–brain barrier [35‐38]. Studies using a targeted deletion of the GFAP gene in embryonic stem cells creating GFAP knockout mice failed to distinguish a specific phenotype [38‐40]. However, immunopositivity for GFAP in neurons was observed in pathologically altered human brain tissue, which was observed at the protein level [22, 28, 41], as well as at the RNA level [23]. Furthermore, patients suffering from AD presented out-of-frame splice forms of GFAP mRNA [28], which might result in impaired protein interaction [24]. A potential cell damage or stress might be visualized by the here observed satellitosis, which surrounded GFAP-positive labelled neurons. The neuronal origin of these immunopositive cells was proven by the double immunolabelling with the neuronal marker NeuN. The current study observed the potential immunopositivity for GFAP in few scattered neurons, rather than in all neurons of the investigated area. However, this may be due to GFAP being present in most but not all neuronal cell types, as shown with NeuN [42]. The authors hypothesize that the former might reflect an adaptation to a decreased energy supply following the traumatic event and provide an advantage to survive an acute stressful situation. The adaptation mechanism might be similar to the proposed temporary retro-differentiation of the neurons of AD patients to a GFAP-positive precursor, which differentiate into neurons after the stress survival [28].

GFAP is the major intermediate filament protein in the adult brain and characteristic for mature astrocytes. During development, GFAP is expressed by radial glia. These bipolar cells are located in the ventricular zone and have the ability to act as neural stem cells [43‐45]. However, this expression only lasts during gestation. In contrast, neural precursor cells of the subventricular zone display an expression of GFAP [46, 47], which persists into adulthood [48]. With regard to TBI, immunopositivity for GFAP in neurons were detected in the CA2, CA3 and CA4 regions of the HC previously and were thought to be the result of brain swelling and brain stem haemorrhages [22]. The former study also used immunofluorescence to highlight GFAP-positive staining in neurons [22]. The results confirm the previous observation of hippocampal immunopositivity for GFAP in neurons following traumatic brain injuries [22]. Furthermore, immunopositivity for GFAP in neurons was also observed in the PCZ and the CB ipsilateral to the traumatic brain injury. Remarkably, immunopositivity for GFAP in neurons was absent in all the investigated CLCs, but was observed in the CB only in prefrontal contusions. Therefore, if the given study truly depicts intraneuronal GFAP, the data presented in this study supports that immunopositivity for GFAP in neurons occur in an area adjacent and ipsilateral to the contusion, except for some Purkinje neurons. This implies that Purkinje neurons might be more susceptible to show GFAP-positive staining in contrecoup injuries compared to cerebral neurons. However, this observation should be confirmed in a future study with a larger sample size, in order to determine its statistical significance. Contrary to the neuronal observations, the number of GFAP-positive astrocytes was shown to increase in the cerebral cortex contralateral to the contusion [5]. Only one of the 16 controls presented immunopositivity for GFAP in neurons, and this was located in the CB. Therefore, it can be hypothesized that either a small number of immunopositivity for GFAP in neurons may be expected in the CB in an atraumatic patient cohort or the person had an unknown history of a traumatic brain injury during their lifetime, or undiagnosed brain injury at death. Also, as all control cases died from sudden cardiovascular diseases, the occurrence of GFAP in neurons might be related to this disorder.

The results of this study revealed that neuronal immunopositivity for GFAP is influenced by the trauma survival time. Immunopositivity for GFAP in neurons was only observed in subacute and delayed TBIs, corroborating previous observations on immunopositivity for GFAP in hippocampal neurons [22]. This was explained by an induced expression of GFAP or a structurally related protein that is stained by the anti-GFAP immunolabelling, as well as a consequence of post-traumatic brain swelling [22]. Similarly, S100 + in neurons was mainly detected in subacute and delayed TBIs previously, with only few positive cases in acute deaths [12, 13]. Several studies and the GFAP expression pattern in certain neuronal stem cells provide evidence of the neuronal ability to express mRNA for GFAP [28, 46]. Recently, a study researching the effects of traumatic brain injury using scRNA sequencing showed an increase in neuronal GFAP expression following a traumatic event in rodents ([23] Supplemental Data). This given study reveals that supratentorial neuronal GFAP can be expected 2 h after the traumatic event, at the earliest. It has to be noted that the post-injury behaviour of the neuronal plasma membrane is not fully understood to date. It is also unclear whether GFAP with a molecular size of 50 kDa [49] can pass the damaged neuronal plasma membrane, as it is 40 kDa larger in size of the molecule that was shown to pass the disrupted membrane, previously [50]. GFAP concentrations in the cerebrospinal fluid and serum were shown to significantly increase immediately following the traumatic event [10]. However, in this study, no statistically significant changes with increasing survival times were observed [10].

With regard to forensic questions such as cause of death determination, time since death estimation or survival time estimation, the usability of the findings presented in this study is limited. The GFAP immunolabelling of neurons in patients suffering from AD [28] or tuberous sclerosis [41] highlights that immunopositivity for GFAP in neurons is not TBI-specific. At most, immunopositivity for GFAP in neurons indicates a pathological state of the brain, which does not allow for a distinct cause of death determination. The immunopositivity for GFAP in neurons lacks useful information for forensic time since death estimations as the PMI did not significantly correlate with the presence of immunopositivity for GFAP in neurons in this study. However, the presented findings may be useful for TBI survival time estimations in certain cases. If a GFAP immunostained neuron is present in the PCZ or the HC, the survival time after the traumatic brain injury was likely exceeding 2 h. However, it also needs to be mentioned that the “minimum survival time” determination based on supratentorial immunopositivity for GFAP in neurons can be biased, if the deceased suffered from neurodegenerative diseases. The absence of immunopositivity for GFAP in neurons lacks useful information on the survival time as an absence of neuronal immunopositivity for GFAP is the “expected” staining result and was found regardless of the survival time or the mechanism of death.