Nrf2 activation in the human brain after stroke due to supratentorial intracerebral haemorrhage: a case–control study

Study design

In this case–control study, cases were patients with first-ever spontaneous supratentorial ICH between 2010 and 2016 ascertained by a prospective, community-based, inception cohort study of adults (aged ≥16 years old) in the Lothian health board region of Scotland,19 with authorisation for research autopsy as part of the Lothian Intracerebral Haemorrhage, Pathology, Imaging and Neurological outcome study. Controls were people who died suddenly in 2005–2014 of non-neurological causes, who underwent autopsy to determine the cause of death, and authorisation had been provided for research autopsy for the Sudden Death Brain and Tissue Bank. Both brain banks include adults from the Lothian health board region of Scotland and are part of the Medical Research Council Edinburgh Brain and Tissue Bank.19 20

Setting

Academic tertiary medical centre situated in Edinburgh and all other secondary and primary care centres around the health board region of Lothian, Scotland. The study described in this manuscript was conducted between 2016 and 2021.

Participants

Blinded to histological findings, EC and CS selected 26 cases with first-in-a-lifetime spontaneous ICH (not intracranial haemorrhage that was exclusively extra-axial or due to trauma, macrovascular causes, neoplasms or haemorrhagic transformation of an ischaemic stroke): 8 died <7 days after ICH onset, 9 died 7–60 days after ICH onset and 9 died >60 days after ICH onset. Cases were selected such that the similarity of the age, sex and ICH epicentre distribution (ie, lobar vs deep, as defined previously19) of the three time groups was preserved as much as possible. Blinded to histological findings, EC and CS selected eight controls so that they had a similar age and sex distribution to cases, without a clinical history or macroscopic pathological evidence of ICH, from the Sudden Death Brain and Tissue Bank. Written consent was provided by participants, or their legal representative if participants lacked mental capacity, in accordance with the Human Tissue (Scotland)Act 2006 and the Adults with Incapacity (Scotland) Act 2000. Written consent was provided by the participant’s nearest relative in all but one case; in this case, written consent was obtained from the patient themselves prior to their death. Further details about the consenting process were described in previous studies.19 20 Briefly, eligible patients and/or their families were identified and approached if appropriate by a doctor or highly trained research nurses about the possibility of tissue donation for research purposes. Information about the projects, their aims and the process of tissue donation was provided with full opportunities for questions.

Variables

We collected demographic details and potential confounders (such as neurological comorbidities and medication use at the time of ICH for cases (or time of death for controls), which might affect Nrf2 activation21 22) from patients’ medical records or their relatives. The assessment of brain tissue histology was blinded to these clinical variables. Our outcomes of interest are further described in the subsequent sections of the manuscript.

Data sources/measurement

For each case, a consultant neuropathologist (CS) identified brain areas to sample from formalin-fixed, paraffin-embedded tissue as ‘perihaematomal’ (adjacent to the ICH), ‘ipsilateral’ (distant from the ICH, but in the same cerebral hemisphere) and ‘contralateral’ (in the opposite cerebral hemisphere to the ICH) (online supplemental table 1). The4 µm sections of tissue were immunohistochemically stained for Nrf2 (rabbit polyclonal anti-Nrf2 antibodies, 1:50 (ab31163, Abcam, UK)) and CD68 to detect phagocytic activity in reactive myeloid cells (microglia and macrophages) (mouse monoclonal anti-CD68 antibodies, 1:100, citric acid antigen retrieval (M0876, Dako, Denmark)). We stained positive and negative controls for Nrf2 and CD68 antibodies to enable concurrent comparison when visually inspecting slides. The Novolink Max Polymer Detection System (RE7280-K, Leica, Germany) was used in accordance with the manufacturers’ instructions. Slides were developed with 3,3’-diaminobenzidine (DAB) and weakly counterstained with haematoxylin to facilitate higher sensitivity of image analysis. Image analysis was undertaken with an automated slide scanner (Axio Scan Z1, Zeiss, Germany) linked to the Zen software (Zeiss, Germany). Thirty-two randomly placed coloured images from each slide from ICH cases and controls were captured at ×40 magnification. Image capture settings and calibration parameters remained unchanged throughout and no manipulation was performed on the images. EC measured the % of each image area that stained positively with each antibody, using the Trainable Weka Segmentation feature of ImageJ (V.1.8.0_66; https://imagej.nih.gov/ij/), which achieves complete binary segmentation of images into areas stained or not stained with antibody. EC determined the overall % of each image area that are stained for Nrf2 +or CD68+using the ‘analyse particles’ feature of ImageJ without restrictions on particle size or circularity limit. EC was trained to identify positive and negative DAB nuclear staining by a consultant neuropathologist (CS). EC subsequently quantified Nrf2 nuclear localisation by counting the number of nuclei that were either positive or negative for Nrf2 staining using the Cell Counter feature on ImageJ. The percentage of nuclei that was positive was calculated by dividing the number of positive nuclei by the total number of nuclei in each image multiplied by 100 and the mean was calculated from the randomly placed images for each case and control. EC was blinded to the details of the image (eg, whether it was a case or a control, whether it was from a perihaematomal, ipsilateral or contralateral region) throughout and quantification was performed in a random order using online random sequence generator.

For in situ RNA hybridisation to assess HMOX1 and NQO1, which are canonical Nrf2-regulated genes through ARE binding motifs,23 two cases who died <7 days after ICH, two cases who died between 7 and 60 days after ICH, and one case who died later than 60 days after ICH were randomly selected from the existing cases. Two controls were also randomly selected from the existing controls. Sections of perihaematomal and distant ipsilateral tissue were selected from ICH cases. The 4 µm thick sections of formalin-fixed paraffin-embedded tissues were stained with fast red targeted either to HMOX1 or NQO1 mRNA using BaseScope probes (ACDBio) in one batch. Tissue was deparaffinised and treated with proteinase 3 before the probe solution was applied. Eight rounds of amplification at 40°C were performed prior to chromogen application. All sections were lightly counterstained with haematoxylin. Ten non-overlapping x40 fields of view were acquired per slide using a Zeiss AxioImager 2 upright microscope. Fast red staining was quantified on Fiji (V.2.0. 2018) using the default FastRed colour deconvolution vector, followed by serial minimal thresholding and Gaussian blur subtraction (σ=60). Smoothing of the final image was achieved with a further Gaussian blur (σ=2), residual background subtracted with a rolling ball method. The image was autothresholded using ‘Moments’ and then watershed. Fast red positive dots and clusters were quantified using ‘analyse particles’ with circularity=0.1–1.

Bias

We blinded the selection of cases and controls to histological findings and blinded the collection of histological data to clinical variables. Throughout the study, cases and controls were deidentified and coded such that assessments were blind to their identity. The cases’ and controls’ clinical variables and histological data were kept in separate spreadsheets and were only linked for data analysis purposes. Cases and controls for in-situ RNA hybridisation were randomly selected.

Study size

Given the absence of published studies on Nrf2 pathway activity after ICH in human ICH cases or sudden death controls and activity of the Nrf2 pathway over time after ICH, we did not have prior estimates on which to base a sample size calculation. Therefore, we included all controls available in our brain bank with age and sex distributions similar to those of patients incident with first-ever ICH in our population,24 and used a case–control ratio of approximately 3:1 to maximise power. Human brain tissue is a scarce resource, and our case–control ratio is the best our tissue bank allows without compromising comparability of cases and controls.

Statistical methods

EC and JL used GraphPad Prism (V.8.4.1) to summarise distributions of variables (mean per patient, with error bars representing the 95% CI of the mean) and perform statistical tests. Differences in distributions between the means of controls and cases were compared using the Kruskal-Wallis test and corrected Dunn’s post hoc test, which corrects for multiple comparisons. All ICH time point groups (ie, died <7 days, 7–60 days and >60 after ICH onset) were included in each analysis. Differences in distributions between the means of each group of cases (according to time from ICH onset and brain location—using contralateral tissue as the comparator) were tested for statistical significance by using matched two-way analysis of variance with the Geisser-Greenhouse correction and the Dunnet test, which corrects for multiple comparisons. All statistical tests were two tailed. Differences were considered statistically significant at p<0.05.

Clinical characteristics of cases and controls

We selected 26 ICH cases (median age 82 (IQR 76–86); 13 (50%) male; 14 deep ICH epicentre and 12 lobar ICH epicentre; online supplemental table 2) and 8 sudden death controls (median age 79 (IQR 77–80) years; 3 (37.5%) male; online supplemental table 3). One of the cases had hepatocellular cancer diagnosed 1 month before ICH, and two controls died from complications of metastatic cancers. One of the controls was taking prednisolone at the time of death but none of the other controls or cases took immunosuppressive medication. Ten (38.5%) cases, but none of the controls, had dementia at the time of death.

Nrf2 tissue and nuclear staining

There were no differences in the percentage total area staining positive for Nrf2 overall in brain tissue sections between cases and controls, nor between brain areas in ICH cases, nor between ICH cases according to their time of death after ICH onset (table 1; figure 1). When quantifying only nuclear staining for Nrf2 (figure 2), we found that the mean % of nuclei staining for Nrf2 seemed higher in peri-haematomal regions compared with contralateral regions, although this was only statistically significant >60 days after ICH onset (25% (95% CI 17% to 33%) vs 14% (95% CI 11% to 17%), p=0.029; table 1). The percentage of nuclei staining for Nrf2 was lower in ICH cases in ipsilateral and contralateral brain regions at all times after ICH onset compared with sudden death controls (figure 3).

Request permission

Request permission

Request permission

RNA in situ hybridisation

Consistent with increased perihaematomal Nrf2 nuclear translocation, we found evidence of target gene induction in HMOX1 and NQO1 downstream of Nrf2; staining of both transcripts was higher in perihaematomal versus distant ipsilateral brain tissue obtained <7 days from onset of ICH (figure 4).

Request permission

CD68 tissue staining

We found differences in the % total area staining positive for CD68 overall in perihaematomal tissue >60 days after ICH (6.75% (95% CI 2.78% to 10.73%)) compared with contralateral brain tissue (1.45% (95% CI 0.93% to 1.96%), p=0.027) and compared with brain tissue from sudden death controls (1.08% (95% CI 0.20% to 1.97%), p=0.0008; table 1, figure 5 and online supplemental figure 2).

Request permission

Limitations

Time intervals from ICH onset to death were quite broad and future studies with narrower time intervals might generate a more precise temporal profile of Nrf2 pathway and MNP activities after ICH. Our finding of increased perihaematomal Nrf2 nuclear localisation in patients which died >60 days after ICH onset suggests that Nrf2 continues to play a role well into the chronic recovery phases of ICH. However, more than half of the group of patients which died >60 days after ICH onset died more than 6 months following their ICH. Exploratory analysis of this subgroup showed a slight trend towards a decline in nuclear localisation of Nrf2 with increasing time. This suggestion towards reduced Nrf2 nuclear localisation after 6 months from ICH onset needs to be confirmed in dedicated, adequately powered, analyses of long-term Nrf2 response to ICH.

We were unable to determine whether Nrf2 activity was less after ICH compared with control tissue because of patients’ age or other modifiers of response to ICH, or whether Nrf2 activation in controls was due to an immediate response to sudden death. High levels of nuclear translocation of Nrf2 after sudden death may reflect the oxidative stress generated by acute global hypoxia.31 In electrophilic environments, dissociation of Nrf2 from KEAP1 is a passive process and may occur postmortem, contrasting with the energy-demanding processes of DNA transcription to produce mRNA.32 33 Thus, sudden death may result in nuclear translocation of Nrf2 with a failure to affect a transcriptional response. Furthermore, evaluating the ‘normal’ resting expression and activation of Nrf2 and CD68 in human brains is challenging; data from animal models and human studies suggest that expression of both is linked to ageing (Nrf2 activity declines whereas CD68 expression increases with age34–36) and is associated with common neurodegenerative disorders such as Alzheimer’s disease and Parkinson’s disease.37 38 Even fresh sections of ‘normal’ brain tissue obtained during neurosurgery are subject to oxidative and inflammatory stress, and this is particularly challenging when evaluating metabolic and inflammatory sensors such as Nrf2.39 We attempted to minimise confounding factors by matching our cases for age, sex, and choosing equal numbers of deep and lobar ICH locations. Although baseline characteristics were similar between cases and controls, there was an imbalance in dementia which was unavoidable due to the limited availability of tissue and the frequency of cerebral small vessel disease as an underlying cause of both ICH and vascular dementia. Since nearly 40% of our cases had dementia, this might have increased the expression of Nrf2 target genes in cases,37 although we did not find evidence of this in an exploratory analysis of Nrf2 nuclear localisation stratified by severity of cerebral amyloid angiopathy (online supplemental figure 3). Future studies should try to match cases and controls for dementia or burden of cerebral small vessel disease if possible.

The meaning of the apparently ‘delayed’ Nrf2 nuclear staining following ICH observed in our study is unclear. A type 1 error due to imprecision is the most likely explanation for the lack of significance at earlier time points, given that the point estimates of the mean in perihaematomal tissue are greater than in other regions at all other time points (figure 3A). In addition, the 95% CI around the means of the perihaematomal regions are notably wider than ipsilateral or contralateral regions; there was greater variability in perihaematomal regions compared with other regions, which may be another reason for the lack of significance. Further work with larger samples is needed to identify which subgroups have greater Nrf2 nuclear localisation.

Finally, although most deaths (73.1%) in our ICH cases were directly attributed to the ICH, the remaining deaths (26.9%) were directly attributed to pneumonia (online supplemental table 2). Research into the effects of systemic infection or metabolic derangements on Nrf2 activity in the human brain is currently extremely limited and their influence on our findings thus remains uncertain.