Cerebral white matter rarefaction has both neurodegenerative and vascular causes and may primarily be a distal axonopathy

Abstract

Cerebral white matter rarefaction (CWMR) was considered by Binswanger and Alzheimer to be due to cerebral arteriolosclerosis. Renewed attention came with CT and MR brain imaging, and neuropathological studies finding a high rate of CWMR in Alzheimer disease (AD). The relative contributions of cerebrovascular disease and AD to CWMR are still uncertain. In 1181 autopsies by the Arizona Study of Aging and Neurodegenerative Disorders (AZSAND), large-format brain sections were used to grade CWMR and determine its vascular and neurodegenerative correlates. Almost all neurodegenerative diseases had more severe CWMR than the normal control group. Multivariable logistic regression models indicated that Braak neurofibrillary stage was the strongest predictor of CWMR, with additional independently significant predictors including age, cortical and diencephalic lacunar and microinfarcts, body mass index, and female sex. It appears that while AD and cerebrovascular pathology may be additive in causing CWMR, both may be solely capable of this. The typical periventricular pattern suggests that CWMR is primarily a distal axonopathy caused by dysfunction of the cell bodies of long-association corticocortical projection neurons. A consequence of these findings is that CWMR should not be viewed simply as “small vessel disease” or as a pathognomonic indicator of vascular cognitive impairment or vascular dementia.

INTRODUCTION

Cerebral white matter rarefaction (CWMR) in the brains of elderly humans has been known to neuropathologists since at least the 1890s, when Binswanger and Alzheimer conducted their seminal studies (1, 2). Alzheimer gave it the term “arteriosclerotic white matter atrophy” while Binswanger called it “encephalitis subcorticalis chronica.” For Alzheimer, the cause was sclerosis of the long penetrating arterioles supplying the deep white matter and he named the condition after Binswanger. The white matter changes, together with cerebral infarcts, fell under the term “cerebral arteriosclerosis,” which throughout the first 6 decades of the 20th century was assumed to be the major cause of presenile and senile dementia (3–7). In 1962, Olszewski suggested the generic term, “subcortical arteriosclerotic encephalopathy” (4). DeReuck noted the similarity to neonatal periventricular leukomalacia and suggested that the periventricular area is a “watershed” between ventriculofugal and ventriculopetal arterial systems, and that chronically decreased perfusion could cause incomplete ischemic infarcts with loss of myelinated fibers (8).

The advent of brain imaging with CT and MRI, however, led some to question an exclusively vascular origin of white matter rarefaction (7). Both modalities showed characteristic periventricular changes, lucencies on CT and hyperintensities on T2 MRI. Hachinski coined the term “leukoaraiosis” (Greek for white matter rarefaction) and linked it to multiple infarct dementia (9), but the changes were quickly found to be common in cognitively and neurologically unimpaired older people as well as those with clinical diagnoses of Alzheimer’s disease (AD) (7). Brun and Englund corroborated the imaging findings in a landmark autopsy study; they found that in 60% of AD cases there was symmetrical loss of white matter staining, most severe in periventricular regions and “…tapering off towards the cerebral cortex” (10). They endorsed the chronic underperfusion theory but also raised the possibility of Wallerian degeneration secondary to the gray matter changes of AD. Braak and Braak conjectured CWMR to be a primary oligodendroglial disorder (11), while Fisher suggested, “The white matter hypomyelination of congophilic angiopathy and Alzheimer's disease should provide clues” but “a unifying hypothesis has not been attained” (6). The relative contributions of cerebrovascular disease and AD to CWMR are still being debated today (12). Impeding a resolution has been the relatively infrequent usage in neuropathological studies, of large-format sections that allow visualization of entire cerebral lobes, resulting in statistically limited studies with small subject numbers.

Assessing the independent associations of AD and cerebrovascular insufficiency to CWMR may now be approached in vivo with multimodal imaging including PET amyloid, PET tau, and MRI (13–15). However, these also have not yet been adequately powered and without neuropathological examination the effects of comorbid neurodegenerative and cerebrovascular lesions would not be addressed. As the routine neuropathological examinations done by the Arizona Study of Aging and Neurodegenerative Disorders (AZSAND) (16) have included large-format sections encompassing complete or near-complete cerebral lobes as well as comprehensive standardized histopathological assessments of AD and other comorbid conditions, we used these data to analyze vascular and neurodegenerative correlates and predictors of CWMR.

MATERIALS AND METHODS

Research subjects and clinical assessments

Subjects included in this study were volunteers enrolled in AZSAND and its Brain and Body Donation Program ([BBDP]; www.brainandbodydonationprogram.org), at Banner Sun Health Research Institute (BSHRI) in metropolitan Phoenix, Arizona (16). All subjects signed informed consents, approved by BSHRI Institutional Review Boards, for both clinical assessment and autopsy for research purposes. Subjects are clinically characterized with annual standardized test batteries, consisting of general neurological, cognitive, and movement disorders components done by cognitive/behavioral neurologists, movement disorders’ neurologists, and neuropsychologists. Private medical records are obtained for all subjects, and these are reviewed in a standardized manner for additional clinical information. Generally, the records from the 2 most recent years are obtained at initial enrollment and again after death.

Most subjects are recruited directly from the surrounding communities through public speaking events, media reports, and monthly public tours of the Institute. Additional subjects are referred by community neurologists. Over the history of the program, recruitment has been directed at independently living, cognitively unimpaired (CU) retired people residing in northwestern Maricopa County, Arizona, especially in the Sun Cities. Reflective of the general population characteristics as obtained from US census data, the population consists mainly of Caucasian, middle income individuals. In addition to cognitively and movement-unimpaired subjects, those with dementia and parkinsonism are also actively recruited while those who have had major strokes prior to enrollment have been excluded.

Subjects for the current study were chosen by searching the BBDP database for those who died between January 1, 1997 and December 31, 2021 (n = 1753), who had an autopsy and were assessed for CWMR scores in all 4 lobes. As we wished to focus this study on CWMR without an obvious cause, we excluded subjects with grossly apparent infarcts at autopsy as well as those with neuropathologically defined acute or subacute infarcts or hemorrhages, traumatic contusions, or history of traumatic head injury, primary or metastatic brain neoplasms, meningitis, or encephalitis. These exclusions effectively restricted our study to those subjects with idiopathic CWMR, i.e. CWMR without an apparent cause other than advanced age. After these exclusions, 1181 subjects remained who were included for the initial analyses (presented in Fig. 2A–C). A subsequent analysis (presented in Fig. 2D) focused on the continuum between CU subjects and those with mild-cognitive impairment (MCI) and clinicopathologically defined Alzheimer disease dementia (ADD) (n = 646); for this set we excluded those subjects with clinicopathologically diagnosed vascular dementia (VaD) as well as all other major neurodegenerative diagnoses. For the final logistic regression models, we added back in only the VaD group (n = 66), giving a total subject number of 712 that were used for modeling.

Figure 2.

Results of CWMR grading in cerebral lobes (A, B) and in groups defined by sex (B), clinicopathological diagnoses (C) and Braak neurofibrillary stages (D). See Supplementary Data Table S1 for all data and Supplementary Data Table S2 for all statistical results. All error bars = 95% confidence intervals. Front, frontal lobe; Temp, temporal lobe; Par, parietal lobe; Occ, occipital lobe; TOT, sum of regional scores; Norm, cognitively unimpaired; MCI, mild-cognitive impairment without a major neuropathological diagnosis; AD, intermediate or high NIA-Reagan classification; VaD, vascular dementia (49 also had AD); DLB, dementia with Lewy bodies (110 also had AD); PD, Parkinson disease (includes PD with dementia; 56 also had AD); PSP, progressive supranuclear palsy (36 also had AD); CBD, corticobasal degeneration (3 also had AD); MSA, multiple system atrophy (1 also had AD); Pick, Pick disease (2 also had AD); FT-TDP, frontotemporal lobar degeneration with TDP-43 proteinopathy (15 also had AD). Diagnoses are not mutually exclusive. (A) Regional and total CWMR scores for all cases. (B) Regional and total CWMR scores for all cases subdivided by sex. (C) Total CWMR scores for all cases by major clinicopathological diagnoses. (D) Total CWMR scores by Braak stage, excluding all major clinicopathological diagnoses shown in Figure 1A–C, other than AD but including the Normal and MCI categories, Cases with nondiagnostic α-synuclein and TDP-43 pathology were not excluded and not separately analyzed.

Neuropathological protocol and examinations

Details of the neuropathological protocol have been previously described (16). Rapid autopsies were facilitated by obligatory 24-7 response teams, enabling a median postmortem interval of 3.2 hours. Fixation of freshly cut 1-cm left-sided coronal brain slices was carried out for 2 days at 4°C with a commercial, neutral-buffered formalin solution containing 4% formaldehyde. Following fixation, diagnostic tissue blocks were taken from 28 brain regions for embedding in paraffin wax; additionally, 6–8 large tissue blocks, representing most or all of each cerebral lobe as well as cerebellum, were cryoprotected in 2% dimethyl sulfoxide/20% glycerol and then sectioned at 40 and 80 μm on a sledge-type freezing microtome. These large-format sections, after mounting on 5 × 7.5-cm glass slides, offer the opportunity for grading of CWMR as well as an extensive survey for microscopic and lacunar infarctions.

Assignments were made for Braak neurofibrillary stage (17) and CERAD neuritic plaque density (18). Neuropathological AD diagnoses were assigned to those cases that had “intermediate” or “high” criteria according to NIA-Reagan Institute criteria (19); the more recent NIA-Alzheimer’s Association criteria have only been used since 2012 so for consistency across all cases were not used for statistical analyses. Vascular dementia was diagnosed following the NINDS-AIREN outline (20); after exclusion of cases with gross infarctions, all cases were of the Binswanger subtype. The Unified Staging System (21) for Lewy Body Disorders (USSLBD) was used to stage and grade Lewy-type α-synuclein pathology. Other neurodegenerative conditions were diagnosed using published clinicopathological consensus or expert-recommended criteria (22–26).

CWMR was graded in frontal, temporal, parietal, and occipital lobes on the large-format sections stained with hematoxylin and eosin (H&E). Comparative sections were stained in 14 and 3 cases, respectively, with the myelin-sensitive dyes Luxol fast blue (14 cases) and eriochrome cyanine R (solochrome cyanine R) (27, 28), in all cases with Campbell-Switzer silver (29) and in 42 cases with immunohistochemistry for neurofilament (Abcam; Cat. No. ab8135) and myelin-associated glycoprotein (Abcam Cat. No. ab89780). On these thick sections, the H&E stain gave the best contrast between rarefied and nonrarefied regions. Rarefaction restricted to the immediate periventricular region, occupying 25% or less of the centrum semiovale, was termed “mild,” while “moderate” was used when rarefaction extended to between 25% and 50%, with “severe” being reserved for rarefaction that involved more than 50%. For statistical purposes, these were transformed to a 0–3 scale. A summary “total” CWMR score was obtained by adding the scores from all 4 lobes, giving a maximum score of 12. Using the same slides as well as the paraffin-embedded slides, brain infarcts were classified by estimated age (acute, subacute, old, or chronic), location (cerebral cortex, centrum semiovale, deep diencephalic nuclei, infratentorial), and size (large are more than 27 cc, small are 1–27 cc, lacunar are 1 cc or less and microscopic infarcts are not grossly apparent and are assigned a volume of 0.1 cc each).

Statistical analyses included, as appropriate, nonparametric Mann-Whitney U-tests, unpaired, 2-tailed t-tests, analysis of variance with Bonferroni post hoc paired comparisons, and Fisher exact tests. A series of logistic regression models were used to determine the independence and strength of selected variables for predicting the total CWMR score.

RESULTS

The mean age of all 1181 included subjects (see Supplementary Data Table S1 for all subject sets) was 82.2 years (SD 9.67). Female subjects (n = 495) constituted 41.8% of the total. Divided by clinicopathological diagnoses, the largest group was those with AD (n = 646), followed by those who were CU at death, hereafter termed Normal (n = 184), and those with Parkinson disease (PD; n = 174), dementia with Lewy bodies (DLB; n = 121), progressive supranuclear palsy (n = 101), MCI (n = 77), vascular dementia (VaD; n = 66), hippocampal sclerosis (HS; n = 55), frontotemporal lobar degeneration with TDP-43 proteinopathy (FT-TDP; n = 41), multiple system atrophy (MSA; n = 12), corticobasal degeneration (CBD; n = 11), and Pick disease (PICK; n = 7). Due to the common presence of more than one of these conditions in the same subject, these diagnoses are not mutually exclusive.

Figure 1 shows patterns and grades of CWMR, as seen in the large-format sections stained with H&E. Some cases have no CWMR and are thus graded 0 (Fig. 1A, I). For most cases with CWMR, the involved area is centered on the immediate periventricular region, with more extensive changes progressively extending concentrically toward the overlying cerebral cortex. Grade 1 is CWMR confined to a periventricular rim extending peripherally to occupy not more than 25% of the entire centrum semiovale (Fig. 1B, E, M). Grade 2 encompasses the grade 1 pattern as well as a peripheral expansion involving up to 50% of the centrum semiovale (Fig. 1C, F, J, N, O). Grade 3 subsumes the grade 1 and 2 areas and extends closer to the gray-white cortical junction (Fig. 1D, H, K, L, P, U–X), in extreme cases sparing only the U-fiber zone.

Figure 1.

Photographs of large-format sections of cerebral lobes, showing cerebral white matter rarefaction patterns and grading. (A–U) Stained with hematoxylin and eosin (H&E). (A–P) The typical, periventricular-centered pattern that is most common in aging humans. (Q–T) Asymmetrical white matter rarefaction patterns associated with small, local cerebral cortex infarcts. (U–X) The pattern of white matter rarefaction in the temporal lobe of 1 case appears consistent with different stains including H&E, silver (Campbell-Switzer), myelin-associated glycoprotein (MAG), and neurofilament (NF), supporting loss of both axons and myelin sheaths.

A different CWMR pattern was seen in cases with localized small cortical infarcts (Fig. 1Q–T). The periventricular region is often not affected at all but rather the zone of rarefaction is broad-based against the involved segment of cortex, affecting the U-fiber zone and extending into deeper white matter as irregular, triangular, or tongue-like areas. In contrast to the usual concentrical periventricular pattern, the infarct-associated pattern is off-center and is usually cortical-based. These cases were excluded at initial case selection.

Immunohistochemical stains that selectively illustrate myelin or axons, such as myelin-associated glycoprotein and neurofilament (Fig. 1W, X) show that within regions showing CWMR, staining for both of these usually showed depletion similar to that seen with the H&E and Campbell-Switzer stain.

Results of CWMR grading in all 1181 subjects by cerebral lobes and in groups defined by sex, clinicopathological diagnoses and Braak neurofibrillary stages are shown in Fig. 2A–C (see Supplementary Data Table S1 for all data, and Supplementary Data Table S2 for complete statistical group data and results). The most severe CWMR was seen in the frontal lobe, followed by the parietal and occipital lobes, with the temporal lobe having the least average severity (Fig. 2A). Females had significantly more severe CWMR scores than males, in paired comparisons of frontal and parietal lobe scores as well as for total WMR scores (Fig. 2B). Of different clinicopathologically defined conditions (Fig. 2C), the VaD group had significantly more severe CWMR than almost all other groups (except Pick disease and CBD, with both of these latter groups being limited by low case numbers). The Normal group had significantly less severe CWMR than all other groups except MSA and CBD (again with limited numbers in both of the latter groups). Figure 2D shows how CWMR varied by Braak stage across the Normal, MCI, and AD continuum; for this analysis VaD and all major clinicopathological diagnoses other than AD were excluded (n = 646). The graph shows how CWMR score increases in a stepwise manner from Braak III through IV, V, and VI, but with no discernible differences between Braak I, II, and III.

Significant results for univariable correlation statistics between CWMR total scores, age, and postmortem measures of neurodegenerative, cardiovascular, and cerebrovascular pathology are shown in Table 1 (complete results are shown in Supplementary Data Table S3). For these calculations, the VaD group was added back into the Normal, MCI, and AD groups (n = 712) but all other major clinicopathological conditions were excluded. Nondiagnostic amounts of Lewy body pathology or TDP-43 pathology were not excluded. Significant correlations were present for age and brain weight as well as several AD-associated measures including CERAD neuritic plaque density, Braak neurofibrillary stage and amyloid angiopathy summary score. Several measures of cardiovascular or cerebrovascular pathology were also significant correlates, including body mass index (BMI), circle of Willis atherosclerosis score, cortical and deep nuclei microscopic infarct number, deep nuclei lacunar infarct number, heart weight (in males only), and kidney weight (in females only).

Table 1.

Results for significant Spearman correlations of cerebral white matter rarefaction summary scores with postmortem cardiovascular, Alzheimer-related and cerebrovascular severity scores (n = 712)

Ante/postmortem cerebrovascular/cardiovascular condition	N; Spearman rho; p value
Age at death	693; 0.15; <0.0001
Brain weight	643; –0.16; <0.0001
Body mass index mean antemortem score	614; –0.109; 0.007
CERAD neuritic plaque cortical maximum density	691; 0.25; <0.0001
Braak neurofibrillary stage	693; 0.26; <0.0001
Amyloid angiopathy cortical summary score	682; 0.12; <0.001
Circle of Willis atherosclerosis score (large arteries)	681; 0.24; <0.0001
Deep nuclei lacunar infarct quantity (small arteries/arterioles)	693; 0.10; 0.007
Deep nuclei microinfarct quantity (arterioles)	693; 0.22; <0.0001
Cortical microinfarct quantity (arterioles)	693; 0.15; <0.0001
Heart weight males	285; 0.13; 0.026
Kidney weight left females	183; −0.25; 0.0006

Cases include only the Normal, MCI, AD, and VaD groups.

See Supplementary Data Table S2 for full results including nonsignificant correlations.

Multiple clinical history conditions were assessed for their contribution to total CWMR score; only hyperlipidemia (synonymous in BBDP database with hypercholesterolemia and dyslipidemia) was significantly associated. Histories of hypertension, type 2 diabetes, carotid artery disease, carotid endarterectomy, coronary artery disease, coronary angioplasty or bypass, myocardial infarction, atrial fibrillation, cardiac pacemaker, cardiac valvular disease, cardiac valvular replacement, congestive heart failure, aortic aneurysm, peripheral vascular disease, and chronic kidney disease were not significantly associated with CWMR score. Of these, a history of carotid endarterectomy was the only condition to approach the significance level (p = 0.073). See Supplementary Data Table S3 for full statistical results.

A series of multivariable logistic regression models (Table 2) were used to explore the independence and strength of association of those variables that had significant associations on univariable statistics (all those shown in Table 1). Again, clinicopathological conditions other than Normal, MCI, AD, and VaD were excluded. With the binary dependent variable designated as positive when the total CWMR score was 4 or greater, only age, sex, BMI, Braak stage, cortical microinfarct number, deep lacunar infarct number, and deep microinfarct number were independent significant predictors. Of these, Braak stage had the most significant p value as well as the strongest effect size (odds ratio), followed by age, deep nucleus microinfarct number, BMI, female sex, cortical microinfarct number, and deep nuclei lacunar infarct number.

Table 2.

Multivariable logistic regression model showing all variables independently and significantly predictive of a total CWMR total score of 4 or greater

Variable	p value	Odds ratio	95% CI
Age	0.00010	1.037	1.018–1.057
Female sex	0.0097	1.53	1.11–2.12
BMI	0.0089	1.006	1.001–1.010
Braak stage	<0.0000001	1.71	1.47–1.98
Cortical microinfarct #	0.011	1.26	1.054–1.52
Deep nuclei microinfarct #	0.0070	1.45	1.11–1.91
Deep nuclei Lacunar infarct #	0.025	1.31	1.034–1.65

BMI, body mass index; CI, confidence intervals; CWMR, cerebral white matter rarefaction.

DISCUSSION

This study was focused on the periventricular-centered type of CWMR that is most common in elderly subjects, excluding the off-center foci of WMR that are usually associated with infarcts of the cerebral cortex. Using these very thick sections, we found that the H&E stain best illustrated the contrast between rarefied and nonrarefied regions, perhaps because this general stain may better reflect the general loss of tissue density that is also imaged with MR. Our staining results, including those for immunohistochemical stains that selectively illustrate myelin or axons, suggest that this most common form of CWMR involves both of these white matter structural elements, as both seem to be equally depleted. It is not possible with our results to be sure which of these might be first affected. Most investigators have assumed a primary axonal loss but an early or critical myelin and/or oligodendrocyte role have also been advanced (1, 11, 30–32). We have previously published on periventricular gliosis in elderly brains as well on periventricular white matter proteomics and capillary and myelinated axon density estimates, all of which mainly support conjugate axonal and myelin loss with astrocytic and microglial reaction accompanied by complex and diverse biochemical changes (33–35). The current study did not directly assess a primary pathogenic role for oligodendroglia, but this seems possible in conditions with oligodendroglial tau pathology, such as the coiled bodies that are often frequent in PSP and CBD, and in globular glial tauopathy. Conclusions are difficult, however, as with either primary axonal or myelin degeneration, the remaining component is usually lost as a secondary change (36–39). It is likely that CWMR occurs very gradually over long time-spans, obscuring the temporal order. More investigation needs to be done on this issue, however.

The periventricular primacy of CWMR has remained a mystery but there are at least 2 hypotheses. The earliest of these was inspired by neonatal periventricular leukomalacia and postmortem vascular anatomical studies (8, 40, 41). These posited a deep borderzone between arterioles penetrating from the cortical surface (“centripetal”) versus those originating from the ventricular surface (“centrifugal”). The existence of this border zone has been questioned (32, 42, 43), but it remains likely that the periventricular white matter oligodendrocytes and axons receive a tenuous blood supply due simply to being at the very end of a single centripetal arterial distribution “end-zone.” Against a particular susceptibility of the centrum semiovale is the relative rarity of infarcts isolated to the region; out of all 1181 cases included in this study, only 23 (1.9%) had lacunar or microinfarcts that selectively involved the centrum semiovale. Another possibility is that CWMR is at least partially secondary to cell-body energy or axonal transport failure of long-association cortico-cortical projection neurons in the neocortical gray matter, whose axons account for most of the centrum semiovale white matter volume, and are those most susceptible to neurofibrillary degeneration (44). This is likely to be at first mainly a “dying-back” or distal axonopathy, rather than Wallerian degeneration (45, 46) as numerically significant neocortical neuronal death probably does not occur with either arteriolosclerosis or early AD (47). By end-stage AD, marked by the appearance of numerous neocortical ghost tangles, or when there are very frequent microscopic cortical infarcts, frank Wallerian degeneration would probably occur when these neurons die. As corticocortical axons are widely and diffusely spread throughout the bulk of the centrum semiovale but are bundled more tightly in periventricular white matter tracts (32, 48–50), their loss might first become apparent in these, only later peripherally extending to involve larger areas. The characteristic U-fiber zone sparing might occur because it is mainly derived from short-association fibers that have lesser energy demands. Chronic cortical hypoperfusion and Braak neurofibrillary changes may act alone or synergistically to produce corticocortical neuronal energy and axonal transport failure at the cell body level, resulting in distal axonopathy.

We found that CWMR is on average most severe in the frontal lobe, followed by equivalent severity in the parietal and occipital lobes, with the temporal lobe being the least severely affected. This is in general agreement with prior reports and a possible explanation is that the frontal lobe is also the lobe most often affected by both microscopic and macroscopic cerebral infarcts, providing some support for the hypothesis that CWMR is due to chronic hypoperfusion (51–55). Although we have shown in this study that Braak neurofibrillary stage is the strongest and most significant predictor of CWMR, it does not explain this lobar distribution, as neurofibrillary changes are earliest and most severe in the temporal lobe (17).

A strong finding was that CWMR was significantly more severe in females, in agreement with some previous reports (55–60). This sex difference does not seem to be explained on the basis of differences in the distribution of traditional ischemic lesions, as males are generally thought to more often be affected by these (59, 60). In the current study, males and females did not differ in their numbers of cortical microinfarcts or deep nuclei lacunar or microinfarcts. Instead, the sex difference we found in CWMR could possibly be due to a greater severity of neurofibrillary changes in females, as females in this study had significantly higher Braak stages than males (means of 4.57 vs 4.30, unpaired, 2-tailed t-test, p = 0.0089), as we have previously reported (61). However, the logistic regression analyses showed that female sex retained its independent significance as a CWMR severity predictor even when Braak stage was included in the model, so another possibility is that the female brain may suffer more damage from a given amount of AD pathology, as compared to men, due to the generally heightened female immune response (62, 63), resulting in more “bystander damage” as a consequence of the immune reaction to AD histopathology.

Of different clinicopathologically defined conditions, the group with VaD had the most severe CWMR. This is confounded, however, by CWMR being part of the diagnostic criteria we used for VaD (20). Of the included neurodegenerative conditions, Pick disease had the highest total CWMR score but conclusions are tentative due to the small number of subjects (n = 7). All of the other neurodegenerative conditions studied, except MSA but including AD, DLB, PD, FTLD-TDP, PSP, and CBD, had significantly more severe CWMR than the Normal group, as generally reported by a few prior small studies (64–68); they did not greatly differ from each other, however. We did not account for concurrent diagnoses in these analyses so it is possible that comorbid AD may have been a major influence. The group with the lowest score, MSA, is likely explained by its predominantly subcortical neuraxis concentration. After largely excluding non-AD neurodegenerative conditions, it was seen that CWMR increases in a stepwise manner from Braak III through IV, V, and VI, but with no discernible differences between Braak I, II, and III (Fig. 2D). Others have also noted this general relationship with different methodologies (32, 46, 48, 69–71). Although not the focus of the current study, it might be conjectured that, for all of the non-AD neurodegenerative conditions, the cause of CWMR could be a primary neocortical neuron factor, a primary cerebrovascular factor, or both.

The univariable statistical analyses showed significant correlations of CWMR total score with age and brain weight as well as several measures of neurodegenerative, cerebrovascular, and cardiovascular pathology. Representing AD changes, there were significant correlations with CERAD neuritic plaque density, Braak neurofibrillary stage, and amyloid angiopathy severity, as expected (10, 12–15, 32, 46, 55, 59, 68–71). Cardiovascular and cerebrovascular factors and lesions were also significant and not unexpected independent predictors, given the prior knowledge. These included BMI, circle of Willis atherosclerosis score (73–75), cortical and deep nuclei microscopic infarct number, deep nuclei lacunar infarct number (46, 53, 54, 76–80), heart weight (81), kidney weight (82), and a clinical history of hyperlipidemia. It was somewhat surprising that more types of clinically recorded cardiovascular disease markers, other than hyperlipidemia, were not significant CWMR correlates, but this may be due to the documented tendency for reversal of cardiovascular disease at the onset of AD (83–90).

The multivariable logistic regression models were able to pare the independently significant variables down to age, sex, BMI, Braak stage, cortical microinfarct number, deep lacunar infarct number, and deep microinfarct number. Of these, Braak stage had the strongest predictive value, followed by age, deep nucleus microinfarct number, BMI, sex (females more strongly associated with CWMR score), cortical microinfarct number, and deep nuclei lacunar infarct number. The significant association of infarcts with CWMR implicates ischemia as a cause. It seems unlikely that infarcts of the deep diencephalic nuclei would cause CWMR with a periventricular-centered pattern but rather they are indicative of a generalized occurrence of cerebral hypoperfusion. Cortical microinfarcts may be the sole sufficient substrate for an ischemic origin of CWMR, as the limited histological sampling done in autopsy studies, even with large-format sections, greatly underestimates their numbers (78), or cortical microinfarcts may just be a marker of susceptibility to chronic or episodic cortical noncritical ischemia. The univariable significant association of amyloid angiopathy with CWMR did not survive in the multivariable models, and the apolipoprotein E4 genotype was not associated in either univariable or multivariable comparisons; others have had similar findings (56, 58, 71, 90–93). Although a few prior studies have also concluded that neurofibrillary degeneration contributes more to CWMR than vascular changes (12, 46, 68–70, 72), these involved relatively few subjects, or were not autopsy studies and so could not exclude confounding neuropathological conditions.

It appears that while AD and cerebrovascular pathology may be additive or synergistic in causing CWMR, both may be solely capable of this. In our study, there are 9 cases clinicopathologically diagnosed with VaD (and excluding other major neurodegenerative diseases or gross cerebral infarcts), that had severe CWMR (total score of 8 or greater) without any neuritic plaques (CERAD neuritic plaque density 0), supporting the conclusion that AD is not required. Also, there are 59 cases with similarly severe CWMR but no brain infarcts at all. Excluding these, there are 65 cases with similarly severe CWMR that have both diagnostic-level AD neuropathological changes as well as cerebral infarcts. The presence of severe CWMR in early-onset genetically determined AD, such as that associated with APP and PS1 mutations (48) as well as Down syndrome (32), when there is usually no substantial cerebrovascular disease, supports the possibility that AD can be solely responsible, while the severe CWMR in CADASIL and other arteriolar vasculopathies (94) gives credibility to a “pure” arteriolosclerotic cause. Against a pure AD cause is the lobar distribution of CWMR, being most common in frontal, parietal and occipital lobes, where watershed macro- and microinfarcts are most common, while the neurofibrillary tangles of AD occur first and are heaviest in the mesial and inferolateral temporal lobe. Contributions from both processes may always be present to some degree from middle age onwards.

A consequence of these findings is that CWMR should not be viewed simply as “small vessel disease” or as a pathognomonic indicator of vascular cognitive impairment or vascular dementia; imaging evidence of deep lacunar infarcts would seemingly be a more dependable diagnostic sign, particularly if multiple. The presence of microscopic infarcts would also be likely to be useful, although reliably detecting them appears to be beyond the capability of current clinical MRI technology. Another conclusion is that vascular cognitive impairment is most often at least partially dependent on concurrent AD neurofibrillary changes at Braak stage IV or greater. Multiple prior studies have concluded that most cases of clinically relevant vascular cognitive impairment are “mixed” in that both vascular and AD pathology is present at significantly contributory thresholds but the current study suggests that it is the AD pathology that is more often the more significant contributor.

FUNDING

This study has drawn on the accumulated data of the Arizona Study of Aging and Neurodegenerative Disorders/Brain and Body Donation Program, which has been supported by the National Institute of Neurological Disorders and Stroke (U24 NS072026 National Brain and Tissue Resource for Parkinson’s Disease and Related Disorders), the National Institute on Aging (P30 AG19610 and P30AG072980, Arizona Alzheimer’s Disease Center), the Arizona Department of Health Services (contract 211002, Arizona Alzheimer’s Research Center), the Arizona Biomedical Research Commission (contracts 4001, 0011, 05-901, and 1001 to the Arizona Parkinson's Disease Consortium) and the Michael J. Fox Foundation for Parkinson’s Research.

CONFLICT OF INTEREST

Dr Beach has had recent unrelated personal paid consultancies with Aprinoia Therapeutics, Vivid Genomics, and Acadia Pharmaceuticals.

SUPPLEMENTARY DATA

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