Neuroimaging Compendium of Microglia, Amyloid, Tau, and Neurodegeneration Across Clinical Variants of Alzheimer Disease

Abstract

Alzheimer disease and its clinical variants have characteristic spatial and temporal progression patterns of amyloid and tau driving symptomatology, but the distribution of microglia density, as measured by 18kDa translocator protein (TSPO) PET, is unknown. Baseline TSPO, amyloid, and tau PET as well as T1 MRI from the longitudinal imaging of microglial activation in different clinical variants of Alzheimer disease study were adjusted for age, sex, body mass index, APOE4 status, TSPO genotype, and intracranial total volume. Imaging outcomes were standardized against controls, visualized across the brain, and placed along a pseudo-longitudinal timeline using disease duration. Microglia density follows the spatial distribution of tau in amyloid-positive individuals and that of neurodegeneration in amyloid-negative individuals. The magnitude, location, and timing of elevated microglia density relative to amyloid, tau, and neurodegeneration is specific to different clinical subtypes of Alzheimer disease.

There is an ongoing debate in the field whether biomarkers for Alzheimer disease (AD) can be used to define the disease in the absence of clinical information or whether AD biomarkers should only be interpreted in the clinical context.¹ Regardless of the biological-clinical framework versus the clinical-biological framework, biomarkers for disease processes beyond amyloid plaque and tau tangle accumulation, including inflammatory, cerebrovascular, metabolic, and synaptic mechanisms, are increasingly being recognized as part of AD pathogenesis. These nonspecific disease processes may add heterogeneity to the biological context and clinical presentations of AD and related dementias (ADRD).

Within AD, various biological subtypes based on their spatial pattern of tau and neurodegeneration have been reported as typical AD (55%), limbic-predominant (21%), hippocampal-sparing (17%), and minimal atrophy (15%).² Similarly, various clinical subtypes³ based on their symptomatology have been reported where amnestic presentations, including typical AD dementia and limbic-predominant age-related TDP-43 encephalopathy (LATE) are more common than nonamnestic presentations, including posterior cortical atrophy (PCA), logopenic variant of primary progressive aphasia (lvPPA), and frontal variant of AD. Amnestic presentations differ in their rate of clinical progression, whereas nonamnestic presentations had earlier age of onset and greater pathologic burden.³

The 18kDa translocator protein (TSPO) is expressed in many immune and nonimmune cells, but is most highly expressed in microglia. TSPO is further upregulated during microglia activation in rodents, but not humans⁴ where a greater TSPO PET signal can be interpreted as greater microglia density and/or recruitment. TSPO, as a general marker of inflammatory alterations, can provide spatial and temporal specificity, particularly when anchored in amyloid, tau, and neurodegeneration. In partnership with the Columbia University Alzheimer Disease Research Center, we have compiled a compendium of clinical variants with TSPO, amyloid, tau, and structural neuroimaging. Our objective was to show the spatial distributions of TSPO, amyloid, tau, and neurodegeneration by amyloid positivity and clinical presentation along a pseudo-longitudinal timeline based on disease duration in a biologically and clinically heterogeneous sample.

METHODS

Participants in the longitudinal imaging of microglial activation in different clinical variants of Alzheimer disease study (R01AG063888) were co-enrolled in the Columbia Alzheimer Disease Research Center (P30AG066462). Participants underwent TSPO PET imaging with the third-generation tracer, 11C-ER176. Participant TSPO binding affinity (rs6971 SNP, TaqMan assay) included high affinity binders (HAB; 54%), mixed affinity binders (MAB; 33%), and low affinity binders (LAB; 13%). APOE genotyping (rs7412 and rs429358 SNPs, LGC Genomics assay) was used to categorize participants as APOE4 carriers if they had at least one ε4 allele. Participants also underwent amyloid PET (18F-Florbetaben), tau PET (18F-MK6240), and structural T1 MRI; analytic methods were described previously.⁵ Amyloid positivity was determined through visual read following vendor specifications. Following Columbia ADRC protocols, participants were evaluated with the Mini-Mental State Examination (MMSE) as a measure of overall performance and cognitive domain-specific tests adjusted against normative data from the National Alzheimer’s Coordinating Center (NACC) Uniform Data set. At Columbia ADRC case consensus, participants were categorized as cognitively unimpaired (controls, n = 20) or as having Alzheimer disease or related dementias, including mild cognitive impairment (MCI; n = 8), AD dementia (n = 7), PCA (n = 5), lvPPA (n = 1), LATE (n = 1), and frontotemporal dementia (FTD; n = 1). Biomarker data was used in case consensus when available. The Columbia Institutional Review Board gave ethical approval, participants (or their legally authorized representatives) provided informed consent, and study procedures were conducted according to the Declaration of Helsinki.

Standard uptake value ratio (SUVR) and volumetric images were averaged bilaterally and adjusted for age, sex, body mass index, APOE4 status, and TSPO genotype. Standardized differences from controls (ie, in units of SDs for comparison across biomarkers) were placed along a pseudo-longitudinal timeline based on disease duration. We inverted volume such that more positive values indicate less volume and more neurodegeneration compared with controls to align with amyloid, tau, and TSPO. Inferential statistics were not performed given the small sample sizes for the groups by amyloid positivity and diagnosis groups.

RESULTS

Amyloid-positive and amyloid-negative diagnostic groups did not differ in age, but amyloid-positive MCI, amyloid-positive AD, amyloid-positive lvPPA, amyloid-negative MCI, amyloid-negative AD, and amyloid-negative LATE tended to be older, while amyloid-positive PCA, amyloid-negative PCA, and amyloid-negative FTD tended to be younger (Table 1), following previous reports. APOE4 carrier status was different across groups, with APOE4 carriers being more common in amyloid-positive diagnostic groups and APOE4 noncarriers being more common in amyloid-negative diagnostic groups. Disease duration was greatest in the MCI groups, reflecting individuals who may not progress clinically. Overall and domain-specific cognition differed across groups as expected.

TABLE 1.

Demographic and Clinical Characteristics for Each Amyloid-Positive (A+) and Amyloid-Negative (A−) Diagnostic Group

	A− controls (%)	A+ MCI (%)	A+ AD (%)	A+ PCA (%)	A+ lvPPA (%)	A− MCI (%)	A− AD (%)	A− PCA (%)	A− LATE (%)	A− FTD (%)	Test statistic, P
N	20 (46.5)	5 (11.6)	4 (9.3)	4 (9.3)	1 (2.3)	3 (7)	3 (7)	1 (2.3)	1 (2.3)	1 (2.3)	—
Age	54–80 70.9±5.4	67–86 74 ±7.9	60–78 70.8 ± 7.7	59–77 66.2 ±7.6	75	66–84 75 ± 9	73–76 74.7±1.5	53	73	60	F(9,33) = 1.9, P = 0.09
Sex	10 (50) women 10 (50) men	3 (60) women 2 (40) men	1 (25) Women 3 (75) Men	3 (75) Women 1 (25) Men	1 (100) Men	2 (66.7) Women 1 (33.3) Men	1 (33.3) Women 2 (66.7) Men	1 (100) Women	1 (100) Men	1 (100) Men	χ2(9) = 6.8, P = 0.65
BMI	21–31.3 25.1 ±3.1	23.8–27.5 25.2±1.5	20.8–29 24.5±3.8	20.1–31.5 27.2±5	22.5	19.1– 29.3 24.2 ± 5.1	20.1– 25.7 22.9 ±2.8	40.6	23.6	34.2	F(9,33) = 3.5, P = 4e-03
Education	12–20 16.4 ±2.1	11–17 14.4 ±2.4	13–22 16.5 ±3.9	16–20 18±1.6	20	12–22 16 ± 5.3	12–22 16.7±5	15	19	18	F(9,33) = 0.8, P = 0.66
Race	17 (85) white 3 (15) black	4 (80) White 1 (20) Black	4 (100) White	4 (100) White	1 (100) White	2 (66.7) White 1 (33.3) Asian	3 (100) White 0 (0) Black 0 (0) Asian	1 (100) White	1 (100) White	1 (100) White	χ2(18) = 16.8, P = 0.54
Ethnicity	18 (90) Non-Hispanic 2 (10) Hispanid	4 (80) Non-Hispanic (20) Hispanic	4 (100) Non-Hispanic	4 (100) Non-Hispanic	1 (100) Non-Hispanic	2 (66.7) Non-Hispanic 1 (33.3) Hispanic	3 (100) Non-Hispanic 0 (0) Hispanic	1 (100) Non-Hispanic	1 (100) Non-Hispanic	1 (100) Non-Hispanic	χ2(9) = 4.3, P = 0.89
APOE genotype	13 (65) E3/E3 7 (35) E3/E4	1 (20) E3/E3 1 (20) E3/E4 2 (40) E4/E2 1 (20) E4/E4	1 (25) E3/E3 2 (50) E3/E4 1 (25) E4/E4	3 (75) E3/E3 1 (25) E3/E4	1 (100) E3/E4	3 (100) E3/E3	2 (66.7) E3/E3 1 (33.3) E3/E4	1 (100) E3/E3	1 (100) E3/E2	1 (100) E3/E4	χ2(36) = 75.6, P = 1e-04
TSPO affinity	2 (10) low affinity 7 (35) mixed affinity 11 (55) high affinity	2 (40) low affinity 1 (20) mixed affinity 2 (40) high affinity	1 (25) low affinity 2 (50) mixed affinity 1 (25) high affinity	2 (50) mixed affinity 2 (50) high affinity	1 (100) high affinity	1 (33.3) mixed affinity 2 (66.7) high affinity	1 (33.3) mixed affinity 2 (66.7) high affinity	1 (100) high affinity	1 (100) mixed affinity	1 (100) high affinity	χ2(18) = 11.6, P = 0.87
Disease duration	—	3–18 8.4±7.5	1–9 3.8 ± 3.6	3–3 3±0	3	2–10 5 ± 4.4	3–7 5±2	0	5	3	F(8,14) = 0.6, P = 0.74
MMSE	24–30 29 ±1.7	22–29 25.6 ±2.7	14–24 19.2 ± 5.5	21–28 24.5 ±3.1	18	25–29 27 ± 2	24–28 25.3 ±2.3	24	26	24	F(9,33) = 7.2, P = 1e-05
Delayed episodic memory Z-score	−1.3 to 1 0.1 ± 0.7	−1.4 to −0.9 −1.2±0.3	−3.4 to −1.7 −2.6 ±0.7	−3.4 to −2.1 −2.8 ± 0.7	—	−2.5 to −1.6 −2.1 ±0.5	−2.8 to −1.9 −2.2± 0.5	−2.9	−3.3	−2.4	F(8,29) = 16.9, P = 5e-09
Immediate episodic memory Z-score	−1 to 1.9 0.3 ± 0.8	−2.2 to −1.7 −1.9 ± 0.4	−3 to −1.1 −2.2 ±0.8	−3.1 to −1.1 −1.8±1.1	—	−1.6 to −1.1 −1.3 ±0.3	−3.3 to −1.5 −2.3 ± 0.9	−3	−2.3	−1.8	F(8,29) = 11, P = 5e-07
Attention Z-score	−2.2 to 1.2 −0.1 ± 0.9	−2.9 to 0.2 −1.2±1.3	−2.7 to −0.6 −1.9±1	−6.2 to −2.3 −4.7±1.7	−2	−1.2 to 0.5 −0.1 ±1	−1.3 to −0.6 −1 ±0.4	−5.8	−1.5	−3.2	F(9,33) = 10.6, P = 2e-07
Executive Z-score	−1.4 to 1.5 0±0.8	−2.3 to 0 −1.2±1.3	−4 to −2.8 −3.4±0.9	−3.4 to −2.2 −2.8 ± 0.9	—	−1.3 to 0.6 −0.3 ±0.9	−1.4 to −1.1 −1.2 ± 0.1	−4	−0.8	−3.7	F(8,28) = 9.6, P = 3e-06
Language Z-score	−0.6 to 1.6 0.5 ± 0.6	−2.6 to 0.5 −0.8±1.2	−4.7 to −0.9 −2.7±1.6	−2 to −0.4 −1.1 ± 0.7	−2.7	−1.4 to 0.5 −0.7±1	−1.7 to −0.5 −1.2± 0.6	−1.3	−0.8	−2.5	F(9,33) = 8.2, P = 3e-06
Visuospatial Z-score	−2.8 to 0.4 −0.4 ±0.9	−8 to 0.7 −1.9 ± 3.6	−10.2 to 0.4 −4 ±4.7	−11.4 to −8.1 −10.1 ±1.6	−1.3	−0.6 to 0.6 −0.2 ±0.6	−1.3 to −0.2 −0.6± 0.6	−10.4	−0.6	−1.4	F(9,33) = 10.8, P = 1e-07

There were 9 amyloid-positive individuals with MCI or AD that represented the typical clinical subtype (Fig. 1A). Amyloid-positive individuals with MCI had mildly elevated amyloid that was widespread in the frontal, temporal, parietal, and cingulate cortices, as well as the striatum, compared with controls. Subtly elevated tau was more focal in the medial and lateral temporal cortex as well as limbic structures. TSPO was slightly lower compared with controls across the brain with the exception of subcortical, limbic, and brainstem regions. Neurodegeneration was observed in temporal, parietal, frontal, and limbic regions, but not as a function of disease duration. As expected, amyloid-positive individuals with AD showed progressive involvement of Thal phase regions. However, tau was greatest in the frontal and temporal cortices for individuals with shorter disease duration, which was mirrored by TSPO. Neurodegeneration followed a typical progression from medial temporal and limbic structures to lateral temporal and frontal cortices. There were 5 amyloid-positive individuals with atypical, nonamnestic clinical subtypes (Fig. 1B). In PCA, there was a consistently elevated pattern across biomarkers in the parietal, occipital, and temporal cortices, although less pronounced in amyloid. In lvPPA, there was elevated amyloid in Thal phase regions, elevated tau in early and middle Braak stage regions, and focal neurodegeneration in the superior temporal cortex.

FIGURE 1.

Regional amyloid, tau, TSPO, and neurodegeneration for amyloid-positive individuals with (A) mild cognitive impairment or typical Alzheimer disease or (B) clinical variants of Alzheimer disease, and for amyloid-negative individuals with (C) mild cognitive impairment or typical Alzheimer disease or (D) clinical variants of Alzheimer disease. Note, volume was inverted so hotter colors represent more neurodegeneration and grayed out brains do not have available imaging data.

Individuals who clinically presented with MCI or typical AD were also found to be amyloid-negative (n = 6; Fig. 1C). Even in the absence of elevated amyloid and tau, there was elevated TSPO and neurodegeneration in individuals with shorter disease duration for MCI. Similarly, in AD with low amyloid and tau, there was elevated TSPO and neurodegeneration that progressed with longer disease duration. There were 3 individuals with atypical, nonamnestic clinical presentations who were amyloid-negative and had low tau (Fig. 1D). In PCA, TSPO and neurodegeneration were elevated in parietal, frontal, temporal, and brainstem regions. LATE demonstrated elevated limbic TSPO and neurodegeneration that extended to a lesser extent to the occipital cortex, brainstem, and precentral and postcentral gyrus. In FTD, TSPO and neurodegeneration were elevated in the frontal cortex.

DISCUSSION

Elevated TSPO and neurodegeneration in characteristic spatial patterns were present with or without elevated amyloid and tau, converging on similar clinical presentations. TSPO PET may offer similar benefits to FDG PET in terms of differential diagnosis;⁶ however, TSPO PET offers additional specificity to microglia density and/or recruitment, which would be beneficial as its own therapeutic target or for monitoring the efficacy of treatments that leverage microglia mechanisms, including antiamyloid treatments.

Our previous study demonstrated that TSPO and tau colocalize on the group level.⁵ Within individuals, we report that TSPO generally colocalizes with tau when present (ie, amyloid-positive individuals). TSPO often colocalized with neurodegeneration in amyloid-positive and amyloid-negative individuals, but there were cases when there was elevated TSPO without neurodegeneration (eg, amyloid-negative MCI for 3 y; amyloid-negative AD for 5 y) or elevated neurodegeneration without TSPO (eg, amyloid-positive MCI at 15 and 18 y). Therefore, individual level biomarker information may provide critical spatial and temporal context for microglia involvement during the disease process. Specific measures from CSF, blood, or tissue samples related to surveillance, phagocytic capability, surface receptors, and effector molecule expression can further provide functional information.⁷

We observed an early onset AD case (65 y or younger) with unexpectedly high tau and TSPO burden after only 1 year of disease duration, suggesting that early onset dementia is accompanied by more aggressive pathology.⁸ In addition, the pattern of tau did not necessarily follow Braak staging in the early-onset AD case. Tau spreading in atypical AD is more likely to follow individual functional connectivity patterns⁹ rather than strict Braak staging. We also observed unexpectedly low tau and TSPO in MCI cases with a non-progressive clinical syndrome (ie, > 10 y since initial diagnosis).

While our sample size is too small for inferential statistics, this compendium of baseline scans encompasses a series of clinically relevant case reports. Visual classification of individual scans identified similar subtypes in tau burden, clinical features, and longitudinal outcomes as the Subtype and Stage Inference (SuStaIn) algorithm.¹⁰ We extended this approach to include TPSO PET, but longitudinal within-subject data is needed to validate these pseudo-longitudinal observations. Ultimately, using neuroimaging, biofluid, genetic, and cognitive measures together can provide complementary information about an individual’s disease progression.