What every neuropathologist needs to know: condensed protocol work-up for clinical dementia syndromes

Abstract

Concerns about the costs associated with autopsy assessment of Alzheimer disease and related dementias according to 2012 NIA-AA Guidelines have been expressed since the publication of those guidelines. For this reason, we designed and validated a Condensed Protocol for the neuropathologic diagnoses of Alzheimer disease neuropathologic change, Lewy Body disease neuropathologic change, as well as chronic microvascular lesions, hippocampal sclerosis of aging, and cerebral amyloid angiopathy. In this study, the Condensed Protocol is updated to include frontotemporal lobar degeneration [FTLD] tau (corticobasal degeneration, progressive supranuclear palsy, and Pick disease), FTLD-TDP, and limbic-predominant, age-related TDP-43 encephalopathy. The same 20 brain regions are sampled and processed in 5 tissue cassettes, which reduces reagent costs by approximately 65%. Three board-certified neuropathologists were blinded to the original Northwestern University Alzheimer’s Disease Research Center Original Protocol neuropathological diagnoses and all clinical history information. The results yielded near uniform agreement with the original comprehensive Alzheimer’s Disease Research Center neuropathologic assessments. Diagnostic sensitivity was not impacted. In summary, our recent results show that our updated Condensed Protocol is also an accurate and less expensive alternative to the comprehensive protocols for the additional neuropathologic diagnoses of FTLD Tau and TDP43 proteinopathies.

INTRODUCTION

Dementias including Alzheimer disease (AD) contribute to a substantial disease burden to society with over 55 million people currently living with dementia worldwide (1). The neuropathologic findings in patients with clinical dementia are not fully predicted by biomarkers during life, and it is common for multiple pathologic lesions to coexist in the brain in the setting of clinical dementia syndromes (2–4). We previously designed and comprehensively validated an alternative “Condensed Protocol” to streamline sampling and staining while reducing costs and maintaining the diagnostic performance of the original National Institute on Aging and Alzheimer’s Association (NIA-AA) guidelines. This protocol was initially validated for the evaluation of the neuropathologic diagnoses of Alzheimer disease neuropathologic change (ADNC), Lewy body disease, chronic microvascular lesions, and cerebral amyloid angiopathy (5). In addition to our original Condensed Protocol validation, our group further utilized the 2017 Condensed sampling and staining methods in a clinical hospital autopsy setting over a 12-month period (6). These results highlighted that more ADNC and Lewy body disease neuropathologic change (LBDNC) were detected using the Condensed Protocol than what would have been expected performing a clinical history-driven neuropathologic work-up alone, while saving approximately $900 per case (6). More recently, the extensive use of our original Condensed Protocol in the forensic setting has been described highlighting the ability of our original Condensed Protocol’s to accurately and efficiently diagnose additional neuropathologic entities such as FTLD Tau (e.g. progressive supranuclear palsy [PSP]) and multiple systems atrophy (7).

In addition to our original Condensed Protocol, other neuropathology groups have also published encouraging results using their own cost-effective sampling and staining approaches. For example, one group’s condensed protocol specifically suggested reducing the immunohistochemical amyloid-β assessment to include only 3 brain regions (inferior temporal, basal ganglia, and midbrain) based on findings highlighted in the neuroimaging literature and validated with a neuropathologic study (8). Another group suggests an alternative simplified protocol that limits sampling to 2 tissue sections per cassette using a targeted unilateral sampling approach, thereby preserving the anatomic relationships of the regions sampled (9). This alternative protocol was shown to reliably detect pathologic TDP-43 (9).

MATERIALS AND METHODS

Retrospective case selection

Formalin-fixed paraffin embedded tissue samples from 64 retrospective brain autopsies in the Northwestern University Alzheimer’s Disease Research Center Neuropathology Core and Northwestern Department of Pathology were selected for inclusion in our study. Appropriate approval was obtained from the Northwestern Institutional Review Board. Each brain autopsy case was previously evaluated according to the original comprehensive protocols for each entity of interest based on the established and published criteria (10–15). Brain autopsy cases were selected to include 10 FTLD-Tau corticobasal degeneration (CBD), 10 FTLD-Tau PSP, 10 FTLD-Tau Pick disease, 10 limbic-predominant age-related TDP-43 encephalopathy, and 24 FTLD-TDP (10 FTLD-TDP Type A, 10 FTLD-TDP Type B, 3 FTLD TDP Type C, and 1 FTLD-TDP type E). We additionally selected 10 random autopsy cases with changes consistent with limbic-predominant age-related TDP-43 encephalopathy cases from the brain bank that lacked corresponding comprehensive clinical information and had enough required tissue available for our condensed protocol sampling. Coexisting neuropathologic lesions such as ADNC (not, low, intermediate, and high) and LBDNC (brainstem predominant, amygdala predominant, limbic/transitional, and neocortical) were not exclusionary criteria in our study. This approach enabled us to comprehensively test the sensitivity and specificity of our “Condensed Protocol” for the neuropathologic diagnoses of FTLD-tau (CBD, PSP, and Pick disease), FTLD-TDP, and limbic-predominant, age-related TDP-43 encephalopathy in the setting of coexisting neuropathologic lesions of different severities.

Tissue and slide preparation

Established immunohistochemical staining protocols were used in the pathologic interpretation of each case, all of which were among the “preferred” methodologies in the NIA-AA guidelines (10, 11). Five tissue blocks were obtained with sufficient brain regions to permit the application of neuropathologic diagnostic criteria for ADNC, LBDNC, hippocampal sclerosis of aging, and chronic microvascular lesions, as previously described (5, 6). Additionally, in this study, the Condensed Protocol is updated and validated to include the additional neuropathologic diagnoses of FTLD-tau (CBD, PSP, and Pick disease) (12–14), FTLD-TDP (13, 14), and limbic-predominant, age-related TDP-43 encephalopathy (15) with sampling and staining optimized to accurately and consistently detect these entities neuropathologically. Histologic sections were cut at 5 microns in thickness and stained with hematoxylin and eosin (H&E). The following antibodies were used for immunohistochemical staining that was performed on various blocks from each brain autopsy case as outlined in Figure  1: (1) amyloid-β (Aβ) 4G8 clone from Biolegend, San Diego, CA; (2) phosphorylated Tau AT8 clone (Ser202, Thr205) from Thermo Fisher Scientific, Waltham, MA; (3) antialpha-synuclein (α-synuclein) LB509 clone from Thermo Fisher Scientific; and (4) antiphospho-TDP-43 (pS409/410) clone from Cosmo Bio USA, Carlsbad, CA. The combinations outlined in Figure 1 enabled the required initial immunohistochemical staining to be completed simultaneously rather than using a staged multi-tiered approach to the neuropathologic work-up.

Figure 1.

Updated condensed protocol sampling and staining protocol.

Neuropathologic evaluation

The validation of the neuropathologic diagnoses of FTLD-tau (CBD, PSP, and Pick disease), FTLD-TDP, and so-called limbic-predominant, age-related TDP-43 encephalopathy was performed as a collaborative effort among neuropathologists (M.E.F., R.J.C., and J.A.) at Northwestern University. Original neuropathologic interpretations followed the comprehensive Northwestern University Alzheimer’s Disease Research Center original protocol which adheres to the established diagnostic neuropathologic guidelines encompassing the full spectrum of neurodegenerative diseases included in our study. The additional neuropathologic entities included in our study included FTLD-tau (CBD, PSP, and Pick disease), FTLD-TDP, and limbic-predominant, age-related TDP-43 encephalopathy (12–15). Condensed Protocol evaluations were performed subsequently and independently. For the Updated Condensed Protocol, 14 slides (5 H&E-stained slides and 9 immunohistochemically stained slides) per case, along with score sheets were distributed to the 2 evaluating neuropathologists at Northwestern (Fig. 1). The final retrospective study cohort included a total of 64 cases as shown in Table 1. One case was excluded from analyses as all immunohistochemistry failed.

TABLE 1.

Study design (n = 64)

n	Disease groups
n = 10	FTLD-Tau, corticobasal degeneration
n = 10	FTLD-Tau, progressive supranuclear palsy
n = 10	FTLD-Tau, Pick disease
n = 10	Limbic-predominant age-related TDP43 encephalopathy
n = 24	FTLD-TDP (A = 10; B = 10; C = 3; E = 1)

Statistics

We calculated sensitivity, specificity, and agreement between the 2 raters (Cohen’s kappa) for the neuropathologic diagnoses of FTLD (FTLD)-tau encompassing FTLD-CBD, FTLD-tau-PSP, and FTLD-tau-Pick disease under the single heading of “FTLD-Tau.” We additionally calculated each statistic for CBD, and PSP combined, and Pick disease, separately. Finally, we calculated sensitivity, specificity, and interrater agreement for TDP-43 proteinopathies which included FTLD-TDP types A-C, E, and limbic-predominant, age-related TDP-43 encephalopathy. Sensitivity and specificity for FTLD-Tau and TDP-43 proteinopathy were calculated from 2 × 2 tables and were each evaluated for the pathologic protein of interest (e.g. Tau or TDP-43). Tau and TDP-43 were recorded as being either “present” or “absent” for each case that was assessed using our updated Condensed Protocol.

Ninety-five percent confidence intervals for sensitivity, specificity, and kappa were calculated using a bootstrap approach when possible. To construct the 5000 bootstrap samples, data from the 63 cases in this study were randomly sampled with replacement; the size of the bootstrap samples was equal to the study n. For each bootstrap sample, sensitivity, specificity, or kappa was computed, from which percentile-based confidence intervals were constructed (16). For cases of perfect agreement between raters, 100% sensitivity, or 100% specificity, bootstrap samples could not be used to construct confidence intervals because the estimates do not vary across the bootstrap samples under those conditions. In those cases, analytic confidence intervals were computed for sensitivity and specificity using the Wilson (1927) method (17, 18). When kappa is 1, the standard errors for computing confidence intervals are equal to zero, so no confidence intervals are provided (19). All analyses were performed using Stata 17.0 (StataCorp, College Station, TX). All raw data from reviewers 1–3 are included in the Supplementary Materials.

RESULTS

The data presented here add to our prior Condensed Protocol work, which was previously shown to have adequate sensitivity and specificity for the neuropathologic diagnoses of ADNC, LBDNC, chronic microvascular lesions, hippocampal sclerosis of aging, and cerebral amyloid angiopathy (5, 6). The original Condensed Protocol demonstrated the feasibility of eliminating tiered immunohistochemical staining by placing 4 separate brain regions in a single cassette strategically chosen for streamlining the diagnostic work-up of a given entity. The particular regions chosen for each cassette are based on the established criteria for determining the extent of the disease (5, 6). The sensitivity for all the additional neuropathologic entities included in our recent validation study was >88% and the specificity for each of the additional neuropathologic entities was >89% (Table 2). Sensitivity for Pick disease using the updated Condensed Protocol was the highest (100%), with the other subtypes of FTLD-Tau having lower sensitivity (90.5%). As the consensus neuropathologic diagnostic criteria for differentiating FTLD-TDP from limbic-predominant age-related TDP-43 encephalopathy remain unclear, we evaluated pathologic TDP-43 as simply being “present” or “absent.” Thirty-three cases of TDP-43 proteinopathy (both FTLD-TDP and LATE) were included in our study and our protocol detected the presence of pathologic TDP-43 with excellent sensitivity (88.6%–100%) and specificity (89%–100%) (Table 2). Additionally, Cohen’s kappa indicated excellent interrater agreement (kappa ≥0.85) for all conditions (Table 2).

TABLE 2.

Sensitivity, specificity, and Cohen’s kappa statistic with 95% confidence intervals for 3 reviewers

Disease	Sensitivity	Specificity	Cohen’s Kappa*
Any TDP-43+	97.14 (90.62, 100) #1	100 (87.94, 100)† #1	0.8506 (0.7390, 0.9529)
	100 (90.11, 100)† #2	100 (87.94, 100)† #2
	88.57 (76.47, 97.37) #3	89.29 (76.00, 100) #3
Non-AD Tauopathy	93.55 (83.87, 100) #1	100 (89.28, 100)† #1	0.9362 (0.8513, 1)
	93.55 (83.87, 100) #2	100 (89.28, 100)† #2
	93.55 (83.56, 100) #3	96.88 (89.70, 100) #3
CBD/PSP	90.48 (76.10, 100) #1	100 (91.62, 100)† #1	0.9254 (0.8256, 1)
	90.48 (76.10, 100) #2	100 (91.62, 100)† #2
	90.48 (76.19, 100) #3	97.62 (92.31, 100) #3
Pick disease	100 (72.25, 100)† #1	100 (93.24, 100)† #1	1.00‡
	100 (72.25, 100)† #2	100 (93.24, 100)† #2
	100 (72.25, 100)† #3	100 (93.24, 100)† #3

^*Confidence intervals (CIs) are percentile-based bootstrap CIs unless otherwise noted.

^†95% CIs constructed using Wilson’s method for proportions. CIs are truncated at 1 if the calculated value exceeds 1. Bootstrap samples do not vary because agreement is 100%.

^‡Unable to calculate standard errors for kappa = 1; analytic standard error is 0, bootstrap samples do not vary. #1, #2, #3: refers to the 3 separate board-certified neuropathologists who independently reviewed the condensed protocol slides.

DISCUSSION

We have updated our original neuropathologic Condensed Protocol methodology to include an additional 6 immunohistochemical stains while adhering to the original Condensed Protocol sampling (5) in order to comprehensively validate our Condensed Protocol for the detection of the additional neuropathologic diagnoses of FTLD-Tau (CBD, PSP, and Pick disease) and TDP-43 proteinopathies (FTLD-TDP types A-C, and E and limbic-predominant age-related TDP43 encephalopathy). Of note, no FTLD-TDP type D cases were included in this study. With our updated validation studies, we demonstrate how the addition of 6 additional immunohistochemical stains while using the same original Condensed Protocol sampling (5) can reliably and efficiently detect FTLD-Tau subtypes, in addition to various subtypes of TDP43 proteinopathies including FTLD-TDP Types A-C, FTLD-TDP Type E, and limbic-predominant age-related TDP-43 encephalopathy. Our recent studies further highlight the accuracy and efficiency of our strategic Condensed Protocol sampling and staining in the neuropathologic work-up of various clinical dementia syndromes. As anticipated, the sensitivity, specificity, and Cohen’s kappa were better for the neuropathologic diagnoses of FTLD-Tau and TDP-43 proteinopathy than for ADNC or LBDNC. We expected that this would be the case since ADNC and LBDNC analyses are performed using much more detailed subjective semiquantitative scoring systems to determine the final classification. By contrast, FTLD-tau and FTLD-TDP in particular tend to have very pathognomonic appearances microscopically combined with the fact that a detailed collection of more variable semiquantitative data points for final categorization is not necessary (e.g. ADNC intermediate vs low).

It is important to note that the Condensed Protocol was designed with the clinically practicing neuropathologist in mind. Therefore, the Condensed Protocol was specifically designed to include a straight-forward brain-cutting sectioning protocol as a first step. However, this certainly does not limit 1 from applying the principals Condensed Protocol to retrospective brain autopsy cases for cost-effective research diagnoses should 1 wish to do so. For example, if no wet brain tissue is available from a retrospective brain autopsy case, it would be reasonable to consider dissecting out brain regions from retrospective blocks to create each Condensed Protocol targeted tissue block.

Pros and cons of the Condensed Protocol must be weighed and considered by neuropathologists in different practice settings. Additionally, because neuropathological diagnoses at autopsy are not limited to targeted neurodegenerative diseases, the ultimate approach and determination of regions to sample should remain at the discretion of the neuropathologist. Therefore, neuropathologists should still always consider supplementing and adding additional sampling and staining on a case-by-case basis. For example, although this condensed protocol was sensitive and specific for FTLD-Tau, the distinction between certain FTLD-Tau subtypes (CBD and PSP) was less precise. Therefore, additional samples with more interrogation of glial tau (i.e. threads characteristic of CBD, tufted astrocytes, and astrocytic plaques) may be indicated. Examples of such glial Tau morphology are shown in Figures 2 and 3. By contrast, the distinctive pathologic Tau seen in Pick disease is shown in Figure 4.

Figure 2.

Comparison of FTLD Tau entities. (A) FTLD-Progressive supranuclear palsy (PSP). (B) FTLD-corticobasal degeneration (CBD) (B).

Figure 3.

Microscopic features of PSP and CBD. FTLD, frontotemporal lobar degeneration.

Figure 4.

Gross and microscopic features of pick disease. Left panel shows asymmetric “knife like” atrophy in Pick disease grossly contrasting with more diffuse and posterior atrophy in Alzheimer disease. Right panel: (A) Coronal plane showing asymmetric severe temporal atrophy in Pick disease. (B) Characteristic eosinophilic Pick bodies within neurons on H&E-stained slide. (C) Characteristic round Pick bodies on ubiquitin stain. (D) Characteristic Pick bodies on AT8 immunohistochemical stain. (E) Severe neuronal loss, gliosis, and microvacuolation in Pick disease on H&E-stained slide. (F) AT8 stain highlighting both Pick bodies (solid black arrows) and Tau inclusions within ramified astrocytes (open arrows).

Similarly, although our updated Condensed Protocol was sensitive and specific for the detection of pathologic TDP-43, we noted that differentiating between FTLD-TDP and limbic-predominant age-related TDP-43 encephalopathy was problematic. This may be due to the current lack of validated and clear diagnostic neuropathologic criteria that can consistently and accurately differentiate between the 2 entities in a blinded setting. Some additional interesting observations were noted in our study. For example, pathologic TDP-43 in the setting of absent or low levels of coexisting ADNC was almost invariably interpreted as FTLD-TDP by blinded review. By contrast, whenever pathologic TDP-43 was identified in the setting of high coexisting ADNC, cases were instead almost invariably interpreted as limbic-predominant age-related TDP-43 encephalopathy by blinded review. This suggests that the extent of coexisting ADNC may be considered as a coexisting helpful feature to aid in consistently being able to differentiate between FTLD-TDP and limbic-predominant age-related TDP43 encephalopathy on a neuropathologic basis alone however this remains unclear. Although our results raise the possibility of the severity of coexisting ADNC being helpful when differentiating LATE versus FTLD-TDP, it is important to note that recent reports suggest that LATE may occur in up to 27% of individuals with up to low ADNC and coexisting ADNC severity may not be a reliable feature to use for making the distinction of LATE versus FTLD-TDP (20). Nevertheless, it is clear that certain neuropathologic features may be helpful moving forward as the field continues to establish the necessary criteria to reliably differentiate between TDP-43 proteinopathy subtypes. For example, another interesting although not unexpected finding was that FTLD-TDP was more consistently diagnosed when coexisting features of amyotrophic lateral sclerosis were identified by H&E (e.g. Bunina bodies in the neurons of the hypoglossal nucleus, and pyramidal tract degeneration) versus limbic-predominant age-related TDP-43 encephalopathy.

Taken together, our observations suggest that there may be consistent neuropathologic criteria to consider utilizing in the future in order to establish the necessary framework for an updated reproducible diagnostic classification system that is based solely on the neuropathologic findings (e.g. FTLD-TDP neuropathologic change vs limbic-predominant age-related TDP-43 encephalopathy neuropathologic change). It should finally be noted that our updated Condensed Protocol does not include sampling of the spinal cord and therefore does not sufficiently detect amyotrophic lateral sclerosis. As noted above, the overarching purpose of our protocol and other simplified protocols is to address the broadest spectrum of neurodegenerative diseases possible, while doing so efficiently and conserving resources. As brain regions sampled and sectioned in our condensed protocol are much smaller than our “Original Protocol” tissue sections, this may make it more challenging to examine adjacent structures for lesional context (e.g. adjacent white matter changes). However, it is crucial to remember that this condensed protocol is not meant to be a “one size fits all” for autopsy brain interpretation. As always, the extent and choice of samples should be at the discretion of the individual neuropathologist based on the issues confronted in any individual brain autopsy case.

Supplementary Data

Supplementary Data can be found at academic.oup.com/jnen.

FUNDING

This study was supported by grants from the National Institute on Aging: P30 AG013854 (PI Vassar, Neuropathology Core Leader Flanagan), P30AG020506 (PI Vassar, Neuropathology Core Leader Flanagan), and K08AG065463 (M.E.F), RF1 AG072080 (M.E.F.). Other sources of support: Rogavin Neuroscience Award (M.E.F.).

CONFLICT OF INTEREST

R.A. Medeiros consults for the Stroke Thrombectomy and Aneurysm Registry (STAR) and Acorai; they have also been previously employed by CorEvitas and consulted for Biomedial Statistical Consulting and Neurogrow Brain Fitness Center. None of the other authors have any disclosures.