Quantitative neuropathology: an update on automated methodologies and implications for large scale cohorts

Brain tissue from 146 donors (mean age 79.91, SE ±0.72 years; male 89; female 57; AD 36; LBD (inclusive of DLB, PDD and PD) 56; mixed AD/DLB 14 and non-demented controls 40 (Table 1)], was obtained from Newcastle Brain Tissue Resource (NBTR) as part of a consecutive case series in accordance with the approval of the joint Ethics Committee of Newcastle and North Tyneside Health Authority and following NBTR brain banking procedures. During life, patients underwent clinical assessments including Mini-mental state examination (MMSE) (Folstein et al. 1975) by board certified Old Age Psychiatrists or Neurologists and clinical diagnoses were reviewed by AJT and IGM post-mortem, blinded to neuropathological diagnosis and checked against relevant standard international clinical criteria (McKhann et al. 1984, 2011; McKeith et al. 2005; Emre et al. 2007).

At autopsy the right hemisphere, brainstem and cerebellum were immersion fixed in 4% buffered aqueous formaldehyde for 4–6 weeks. Following fixation, the right hemisphere was dissected in coronal planes approximately 0.7 cm intervals and subjected to standard macroscopic examination, and brain regions required to determine the neuropathological diagnosis were sub-dissected and processed through increasing concentrations of alcohol then chloroform to paraffin wax. Subsequently, all brains underwent standard neuropathological assessment using internationally accepted criteria including neuritic Braak stages (Alafuzoff et al. 2008), Thal amyloid phases (Thal et al. 2002), CERAD scores (Mirra et al. 1991), NIA-AA scores (Montine et al. 2012) and McKeith criteria (McKeith et al. 2005).

Each case was then sampled to compose a TMA block. Areas that were sampled for the TMA were taken from paraffin embedded (donor) blocks containing: pre-frontal cortex [Brodmann area 9 (BA), 10/46], mid-frontal cortex, (BA8, 9), cingulate gyrus (BA24, 32), caudate, putamen, external globus pallidus, amygdala, insular cortex, motor cortex (BA4), thalamus, entorhinal cortex, temporal cortex (BA21, 22, 41/42), parietal cortex (BA22, 40) and occipital cortex (BA17, 18, 19, 19/37) (Fig. 1).

Fixed paraffin donor blocks were warmed for 1 h at 37 °C to aid tissue removal. 3 mm cylindrical tissue cores were taken from predefined positions using a hand held tissue sampler (Tissue-Tek^® Quick-Ray™ TMA system, Sakura, CA, USA). The tip of the hand held punch was inserted at the correct position through the depth of the tissue in the donor block and removed along with the cylindrical tissue core (Fig. 2a). Each tissue core was then inserted into the correct ‘hole’ in a single, regular sized pre-made recipient TMA paraffin block (4 cm × 3 cm—made to perfectly match the Tissue-Tek^® Quick-Ray™ TMA system) in numerical order (Fig. 2b). If any donor blocks are missing, empty ‘holes’ in recipient block were filled with molten wax. Each of the punches was pushed securely into the recipient block by hand and the block incubated at 37 °C to reduce the wax–tissue interface when sectioning.

An in-house silicone mould, made specifically to fit the TMA recipient block, was preheated to 60 °C for 1 h and 2–4 ml of molten paraffin wax placed into the base. The recipient TMA block was placed tissue face down onto the molten wax and left for 5 min before a further 15 min incubation at 37 °C to anneal (Fig. 2c). The block was allowed to cool fully before the silicone mould was removed and excess wax removed. TMA sections were then cut and mounted onto glass slides (Fig. 2d) and immunohistochemically stained for HP-_T (AT8 clone, Innogenetics, Belgium), β amyloid (4G8 clone, Covance, UK) and α-syn (Leica, UK) as described previously (Walker et al. 2015).

TMA slides were analyzed using an automated system consisting of a Nikon Eclipse 90i microscope, DsFi1 camera and NIS Elements software v 3.0 (Nikon). Sections were placed onto the microscope in the correct orientation to position tissue core 1 in the top left field of view. To capture images of each of the 40 tissue cores on the TMA slide, the microscope was positioned in the center of the first tissue core at 20× magnification for guidance and brought into focus at 100× magnification. The co-ordinates of this first tissue core were then mapped using a macro designed to take large images comprised of 3 × 3 small images measuring 1.7 mm². The microscope was then positioned over the center of the second tissue core and the process repeated until the positions of all 40 tissue cores had been mapped. If any of the tissue cores were missing or too damaged to be included in the analysis the tissue core number was noted and this was factored into the analysis. Once all 40 tissue cores were mapped, the microscope was directed to the co-ordinates of the first tissue core and the images of all tissue cores were automatically taken in sequence. Regions of interest (ROI) were applied to individual images if necessary to exclude any white matter or abnormalities in the tissue (e.g., folded tissue or tears). Restriction threshold was applied to capture all immunopositive signals. The measurement of immunopositivity and subsequent calculation of the percentage area covered by immunopositivity was performed using an automated methodology. Red, Green and Blue (RGB) thresholds that determine the pixels that are included in the binary layer used for measurement were standardized separately for each AT8, 4G8 and α-syn immunopositivity and thresholds were set at a level that was reached by immunopositive pathological structures only (except APP; see below). RGB intensity values are measured on a scale between 0 and 255 (see NIS elements version 3.0, user guide, 2008, Nikon, Surrey UK) and were set as follows; AT8: R25–170, G27–156, B11–126; 4G8: R50–180, G20–168, B8–139, α-syn: R15–161, G7–139, B4–133 (Fig. 3). Thereby, unspecific background staining did not reach the threshold and was not included into the measurement. In addition to RGB thresholds, we set a restriction threshold for the assessment of 4G8 immunopositivity that excluded the measurement of immunopositive signals of a size below 100 μm²; this was necessary to ensure that physiological, cellular APP that is stained with 4G8 antibody was not included in the measurement. Of note, the exclusion of areas below 100 μm² implies that pathological β amyloid depositions of less than 100 μm² were not included into the measurement. However, diffuse β amyloid depositions and β amyloid plaques are typically larger than 100 μm² (Duyckaerts et al. 2009). Of note, only immunoreactive neurones harboring HP-_T positive NFTs and NTs were quantified in sections stained with AT8 antibody. Glial HP-_T pathology such as for example seen in progressive supranuclear palsy (Dickson et al. 2007), corticobasal degeneration (Dickson et al. 2002), and aging-related tau-astrogliopathy (ARTAG) (Kovacs et al. 2016) was identified on visual inspection and excluded when quantifying pathology as part of TMA. In addition, TMA punches that were devoid of β amyloid plaques but had severe cerebral amyloid angiopathy on 4G8 stained sections were disregarded in quantitative analysis. Percentage area of the tissue covered by immunopositivity was subsequently calculated and for brain regions that had more than one tissue core, mean values were calculated.

Statistical analysis was conducted using the Statistical Package for Social Sciences (SPSS v 22, IBM). Data was tested for normality using Kolmogorov–Smirnov test followed by visual inspection of variable histograms. Kruskal–Wallis was used to determine overall differences between groups and a non-parametric t test (Mann–Whitney U) to assess post hoc differences between individual groups. Spearman’s correlation coefficients (two tailed) were used to assess associations between pathology load with pathological stages and MMSE scores. Exploratory linear regression analyses were conducted to investigate predictors of disease progression and cognitive decline.

Mean neocortical HP-_T load increased significantly (p < 0.001) with increasing NFT Braak stage in all neocortical regions (Fig. 4a), as did neocortical β amyloid (p < 0.001) in line with increasing Thal phases (Fig. 4b). Although α-syn load was higher in all regions classified as neocortical LBD compared to limbic LBD, only the increase in temporal lobe α-syn load was significant (p < 0.05) (Fig. 4c). For post hoc statistics see Table 2.

Of the cases classified as Braak stage VI, the median HP-_T load in the entorhinal cortex was 20.96%; however, the range of HP-_T load was extensive (minimum 0.43%, maximum 71.48% and interquartile range 4.13–44.03%) (Fig. 5a). In cases classified as Thal phase 5 the median β amyloid load in the frontal cortex was 11.45% (minimum 1.03%, maximum 51.03% and interquartile range 6.75–17.03%) (Fig. 5b). Cases that fulfilled McKeith criteria for neocortical LBD harbored a median α-syn load of 0.2% (minimum 0.001%, maximum 1.85%, interquartile range 0.04–0.58%) in the cingulate cortex (Fig. 5c).

We investigated whether HP-_T load in the entorhinal cortex was associated with disease progression as measured by NFT Braak stage and cognitive decline as measured by MMSE. AD and control cases were selected to display a full range of NFT Braak stages. HP-_T load positively correlated with NFT Braak stage (r _s = 0.714, p < 0.01) (Fig. 6a). In cases where MMSE scores were available (n = 37), HP-_T load negatively correlated with MMSE score (r _s = −0.420, p < 0.01) (Fig. 6b). In AD and control cases β amyloid load in the frontal cortex positively correlated with Thal phase (r _s = 0.818, p < 0.01) (Fig. 6c) and frontal β amyloid load negatively correlated with MMSE score (r _s = 0.676, p < 0.01) (Fig. 6d). Whilst in LBD and control cases, α-syn load in the cingulate positively correlated with disease progression as measured by McKeith criteria (r _s = 0.875, p < 0.01) (Fig. 6e). In cases where MMSE scores were available (n = 39) α-syn load in the cingulate negatively correlated with MMSE score (r _s = 0.690, p < 0.01) (Fig. 6f).

To address whether pathology load could predict disease progression and cognitive decline as measured by MMSE, we performed exploratory linear regression analyses in AD, LBD and control cases. HP-_T load in the entorhinal cortex predicted both NFT Braak stage (model R ² = 0.324, F ₍₁₎ = 35.423, p < 0.001) and MMSE score (model R ² = 0.116, F ₍₁₎ = 4.873, p < 0.05). In AD and control cases, β amyloid load in the frontal cortex predicted Thal phase (model R ² = 0.421, F ₍₁₎ = 48.025, p < 0.001) and MMSE score (model R ² = 0.413, F ₍₁₎ = 25.277, p < 0.001). Whilst in LBD and control cases α-syn load in the cingulate predicted disease progression as classified by McKeith criteria (model R ² = 0.260, F ₍₁₎ = 43.959, p < 0.001) and MMSE score (model R ² = 0.119, F ₍₁₎ = 4.983, p < 0.05).