APOE ε4 gene dose effect on imaging and blood biomarkers of neuroinflammation and beta-amyloid in cognitively unimpaired elderly

The number of persons affected by Alzheimer’s disease (AD) across its pathological continuum was recently estimated to be as high as 416 million [1]. From this global estimate, 3/4 of individuals were classified as preclinical AD, characterized by the presence of beta-amyloid (Aβ) plaques but absence of clinical symptoms [1]. In addition to the hallmark pathologies, i.e., Aβ plaques and neurofibrillary tangles, inflammation in the central nervous system (CNS) is recognized to have an important, partly independent, role in Alzheimer’s continuum [2]. In the brain, inflammation is mainly mediated by microglia and astrocytes [3], and in AD, compiling evidence suggests that increased microglial and astrocytic reactivity could be present already during early, possibly protective processes [4‐7].

Apolipoprotein E (APOE) ε4 allele is the strongest genetic risk factor of sporadic AD; it increases the risk of disease and decreases the age of onset when compared with the most common APOE ε3 or the protective APOE ε2 alleles [8]. APOE ε4 gene dose-related increase in brain Aβ load is present already in cognitively normal individuals [9‐11], and it has been suggested to be caused by impaired degradation and clearance of Aβ, a task which is performed by glial cells and affected by APOE isoforms [12, 13]. In neuropathological studies, APOE ε4 has been seen to associate with increased microglial number in the brains of individuals with AD [14] and higher microglial cell reactivity around Aβ plaques in a mouse model of Aβ deposition and human APOE alleles [15]. Still, clinical investigations concerning the relationship between early ongoing neuroinflammatory processes and Aβ in cognitively unimpaired APOE ε4 carriers are scarce.

Investigation of regional neuroinflammation in AD in vivo has been enabled by PET imaging and specific ligands such as ¹¹C-PK11195 that target 18-kDa translocator protein (TSPO) as a proxy for microglial reactivity. TSPO is present in the outer mitochondrial membranes of microglia and elevated in the brain in relation to injuries or pathology [16]. In humans, increased TSPO ligand-binding has recently been suggested to represent changes in cell density rather than protein overexpression [17], and to be mostly covered by microglia, and to a lesser extent astrocytes and endothelial cells [18, 19]. Previous studies using TSPO PET imaging have shown increased regional ligand-binding in patients with AD [20‐23], mild cognitive impairment [4, 24, 25], and some also in Aβ-positive compared with Aβ-negative controls [7, 26]. However, results are partly inconclusive since also minor or no differences between diagnostic groups have been reported [27‐29].

In addition to imaging, more easily accessible biomarkers for AD pathology measured in blood have become available recently thanks to the development of more sensitive methods [30]. Unfortunately, since proteins expressed by microglia in the CNS are also present in peripheral macrophages, the development of blood-based assays targeting microgliosis is demanding, and interpretation of measurements from blood is complicated [31]. However, one interesting fluid biomarker for glial reactivity, glial fibrillary acidic protein (GFAP, a marker of reactive astrocytosis), is measurable from blood using the single molecule array (Simoa) technology and has been recently shown to be associated with Aβ deposition and increased already in early stages of AD [32‐34].

Based on previous literature suggesting an early involvement of neuroinflammation during the AD continuum, we hypothesized that in vivo TSPO binding and plasma GFAP concentrations would be elevated in cognitively normal APOE ε4 homozygotes or APOE ε4 heterozygotes, representing different genetically increased risk for Aβ accumulation and sporadic AD, compared to age-matched non-carriers. To test this hypothesis, we evaluated differences in (i) regional Aβ and TSPO PET and (ii) plasma Aβ_42/40 and GFAP concentrations and their associations primarily among cognitively normal APOE ε4 homozygotes, heterozygotes, and non-carriers. In addition, we performed secondary analyses between Aβ-positive (representing Alzheimer’s pathological change or preclinical AD [35]) and Aβ-negative individuals and investigated the association between imaging and fluid biomarkers of neuroinflammation and Aβ deposition and markers of disease progression (cognitive performance and volumetric brain changes) in our cohort comprised by cognitively unimpaired participants enriched with APOE ε4 carriers.

Microglial and recently also astrocytic reactivity has been suggested to be early events during the long AD continuum [3, 4, 6]. In this study, we used TSPO PET imaging and plasma GFAP to investigate early neuroinflammatory differences in cognitively unimpaired individuals stratified by their APOE ε4 gene dose and thus risk for AD. Despite the verified presence of Aβ pathology in APOE ε4 carriers, we found no significant regional differences in TSPO-binding between APOE ε4 gene doses or cognitively normal Aβ-positive subjects (presenting Alzheimer’s pathological change or preclinical AD) compared with Aβ-negative individuals. However, in the whole cognitively unimpaired sample, higher levels of both neuroinflammatory markers were associated with lower hippocampal volume, used as a proxy for future disease progression.

To our knowledge, this study is the first one to compare TSPO PET findings in cognitively normal at-risk individuals grouped by their APOE ε4 gene dose. In previous human in vivo studies, most robust increases in TSPO-binding have been found in Alzheimer’s dementia in comparison to controls [20‐22, 49], but also in Aβ-positive MCI [4, 24, 25] and Aβ-positive controls [7, 26]. Thus, we had hypothesized that early Aβ related neuroinflammatory changes should be present also in either cognitively normal APOE ε4 homozygotes or APOE ε4 heterozygotes, both representing a genetically increased risk for Aβ accumulation and sporadic AD. Here, despite clearly increased fibrillar Aβ load, we did not find associated increased TSPO binding in APOE ε4 homozygotes. In addition, our secondary analyses were not able to replicate the reported increased TSPO-binding in Aβ-positive “at-risk” individuals using ¹¹C-PK11195 PET, even with a larger sample size compared with previous reports. Our study included approximately 20 participants in each APOE ε4 gene dose group and 34 cognitively unimpaired Aβ-positive individuals, whereas the previous studies included only six [7] or seven [26] Aβ-positive controls. However, it should be remembered that our cohort, and especially its Aβ-positive participants, was highly enriched for APOE ε4 carriers. Neuritic Aβ plaques are known to be surrounded by reactive microglia in AD; however, APOE is also directly liked to immune responses [50, 51], and APOE ε4 genotype has been suggested to attenuate the response of “amyloid related microglia” towards AD pathology [52]. Here, using TSPO imaging, we are unable to differentiate microglial phenotypes and cannot exclude a direct down modulation of microglial response in APOE ε4 carriers that could explain the lack of hypothesized increased TSPO-binding in regions with abundant Aβ pathology.

In addition to TSPO PET, we also performed ¹¹C-PiB PET for all participants to verify the presence of Aβ pathology in our cognitively unimpaired sample. As expected, Aβ deposition in the brain increased in an APOE ε4 gene dose dependent fashion; significantly elevated cortical ¹¹C-PiB retention was present in APOE ε4 homozygotes compared with both heterozygotes and non-carriers in all evaluated regions. These findings are in line with previous PET studies [9‐11, 53, 54] as well as with a recent study by the Amyloid Biomarker Study Group summarizing APOE ε4 gene dose-related effects on temporal course of Aβ accumulation [55]. In our study, all APOE ε4/ε4 participants over the age of 63 were already Aβ-positive, whereas approximately 50% of the heterozygotes were still Aβ-negative, and 40% of non-carriers were classified as Aβ-positive. Activated microglia are known to be located in the proximity of Aβ plaques in AD, and using PET imaging in vivo, Aβ pathology has been shown to correlate with TSPO-binding in some, although not all studies [7, 25, 27, 56‐58]. Our voxel level analysis showed significant positive correlations between 11C-PiB and 11C-PK11195 both in cognitively normal APOE ε4 homozygotes and heterozygotes. However, significant clusters were found also in regions outside our a priori chosen regions of interest, such as the white matter, suggesting that these effects might not all be due to higher TSPO binding associated with Aβ plaques.

Here, we also found that Aβ positivity modulated the effect of APOE ε4 gene dose on ¹¹C-PK11195-binding in regions known for early tau deposition, where Aβ-positive APOE ε4 homozygotes seemed to have increased TSPO binding compared with Aβ-positive non-carriers. In addition to Aβ, APOE ε4 is known to accelerate tau pathology that again has been suggested to be closely associated with microglial reactivity [59], and increased tau PET signal in the entorhinal cortex has been reported for cognitively unimpaired Aβ-positive APOE ε4 homozygotes and heterozygotes compared with Aβ-positive non-carriers [53]. Since Aβ build up starts earlier in APOE ε4 carriers, we could hypothesize that increased tau deposition and related inflammatory processes in APOE ε4 carriers would be driving this interaction. Unfortunately, lack of tau PET or CSF tau measurements in our cohort prevented us from investigating the interaction with TSPO-binding and tau further in our cohort.

During recent years, significant efforts have been made to measure various biomarkers of AD pathology in plasma that would provide a less invasive and more easily accessible alternative to brain imaging and lumbar puncture [30]. Plasma GFAP has been recently reported to be an early marker of astrocytosis in response to AD pathology that strongly correlates with Aβ pathology [34, 60] but not with tau when accounting for Aβ [32]. In our cohort, plasma GFAP levels showed elevated concentrations in the most Aβ positive individuals and correlated with composite amyloid PET SUVRs. Interestingly, plasma GFAP was the only biomarker showing significant associations with both cognitive performance and hippocampal volume that could be considered as markers for progression in the Alzheimer’s continuum. On the contrary, plasma GFAP concentration did not correlate with composite TSPO-binding (Fig. 4). This is not surprising, considering that the two inflammatory markers present different targets; plasma GFAP is expected to reflect more astrocytic reactivity associated with Aβ pathology [34], whereas TSPO PET is thought to reflect microglial density [17]. Our results with GFAP support the previous findings suggesting that reactive astrocytosis is present already in cognitively normal individuals and related to Aβ pathology [32‐34]. Despite clear differences in fibrillar Aβ levels measured by PET, we did not see significant differences between the APOE ε4 gene doses in plasma Aβ_1-42/1–40 measured by previously described IP-MS method [47]. Plasma Aβ_1-42/1–40 was previously reported to correlate with global cortical Aβ PET signal in another study including cognitively normal individuals using the same IP-MS method [48]. We could not replicate this finding in our cohort, comprised of slightly older and highly APOE ε4-enriched cognitively normal participants, although a trend towards a negative association could be seen in the whole cohort.

Last, we also wanted to compare all the biomarkers and their associations with cognitive and structural variables that could serve as proxies for disease progression in our “at-risk” cohort. We found a negative association between composite cortical TSPO-binding and hippocampal and entorhinal volumes, suggesting that more global elevation in TSPO-binding, and thus microglial density, could be present in individuals with subtle neurodegeneration. Interestingly, higher plasma GFAP associated with both lower cognitive performance and lower hippocampal volume in our cognitively normal cohort. Previously, Hamelin and colleagues reported a positive correlation with both hippocampal volume and MMSE score, suggesting that higher glial reactivity associated with higher hippocampal volume would likely be protective [7]. However, our study population is composed of only cognitively unimpaired individuals highly enriched for APOE ε4 carriers, and all having MMSE scores > 25, thus likely presenting more subtle structural brain changes compared with the population of the previous study. In addition, we did not find any association with TSPO-binding and MMSE, CERAD total score, or the preclinical cognitive composite, in line with other studies performed with ¹¹C-PK11195 [25]. Based on our results, increased TSPO-binding in the preclinical phase, at least in APOE ε4 carriers, could be more related to a later preclinical phase when subtle neurodegeneration already starts to be present.

The strength of this study is our well characterized and balanced cohort of cognitively unimpaired participants stratified by their APOE ε4 gene dose and related risk for sporadic AD and a relatively large number of rare homozygotic carriers of the APOE ε4 allele. In most studies, group comparisons are done between APOE ε4 carriers against non-carriers and especially cognitively unimpaired homozygotes are rarely studied. In addition, we have measured both imaging and more easily available fluid biomarkers for amyloid deposition and neuroinflammatory processes from the same individuals and were able to combine them with additional information concerning brain structure and cognitive data. However, this study does not go without limitations. First, it would have been optimal to compare TSPO PET findings with CSF biomarkers specific to microglial responses, but unfortunately, we did not have CSF samples available from the whole cohort. Thus, plasma GFAP analysis were included. However, we are fully aware that TSPO PET and plasma GFAP reflect different neuroinflammatory processes. Second, we were not able to include tau PET or CSF tau measurements to further investigate association of glial biomarkers and early tau pathology in our cohort. The lack of tau biomarkers prevented us also from evaluating possible early tau pathology as a confounding factor that could have contributed to the reported variation in ¹¹C-PK11195 retention seen in this study. However, since the cohort was comprised of cognitively unimpaired individuals, we did not expect wide-spread tau pathology. Third, even though ¹¹C-PK11195 has shown robust changes in primary inflammatory conditions such as multiple sclerosis, it has been suggested that its sensitivity is limited and outperformed by the second generation TSPO ligands, such as ¹¹C-PBR28. However, affinity of the second generation TSPO ligands is affected by a single nucleotide polymorphism rs6971 in the TSPO gene, leading to division of people into high, mixed, and low affinity binders. Due to the difficulty of recruiting rare homozygotic APOE ε4 carriers, we wanted to avoid the unfortunate scenario of having multiple homozygotic participants excluded due to low-binding TSPO genotype.

Our study on cognitively unimpaired “at-risk” individuals carrying either one or two copies of the APOE ε4 gene showed clear differences in fibrillar Aβ load in the brain, but the changes were not accompanied by higher glial reactivity as measured with TSPO PET either in APOE ε4 carriers, or in Aβ-positive individuals, representing preclinical AD. However, cortical TSPO binding was associated with lower hippocampal and entorhinal volumes in the whole cognitively unimpaired sample. These findings suggest that in our cognitively unimpaired cohort enriched by APOE ε4 carriers, neuroinflammatory processes measured by TSPO PET are related to a more advanced preclinical phase of AD where Aβ accumulation is accompanied by subtle structural changes.