Amyloid beta and tau are associated with the dual effect of neuroinflammation on neurodegeneration

Abstract

INTRODUCTION

Current literature presents conflicting results regarding the impact of neuroinflammation on Alzheimer's disease (AD)‐related neurodegeneration. While some studies suggest that neuroinflammation potentiates neurodegeneration, others indicate a protective effect.

METHODS

We evaluated 145 individuals with positron emission tomography (PET) for amyloid beta (Aβ), tau, and translocator protein (TSPO), a proxy of neuroinflammation, to test the hypothesis that Aβ and tau are associated with the dual effect of neuroinflammation on neurodegeneration across the AD continuum.

RESULTS

The detrimental effects of neuroinflammation on gray matter density occurred in two waves. The first neuroinflammation‐related detrimental wave was associated with brain Aβ deposition, while the second was with widespread tau tangle pathology. Furthermore, the concomitant presence of neuroinflammation, Aβ, and tau was associated with faster cognitive decline over 2 years.

CONCLUSIONS

Our results support a model in which Aβ‐ and tau‐associated neuroinflammation are related to two waves of deleterious effects on AD‐related neurodegeneration.

Highlights

• Two waves of detrimental neuroinflammation effects on brain density associated with Aβ or tau.

• Aβ associated with deleterious effect of neuroinflammation on brain density in early AD.

• Tau associated with deleterious effect of neuroinflammation on brain density in late AD.

• Interactions of Aβ, tau, and neuroinflammation are associated with cognitive decline.

1. BACKGROUND

Brain deposition of amyloid beta (Aβ) and tau proteins is the hallmark pathological feature of Alzheimer's disease (AD). ¹ A growing body of evidence underscores that neuroinflammation, often associated with microglial activation, plays a role in the interplay between AD protein aggregates and neurodegeneration. ² When microglia are activated, they undergo morphological and functional changes, including altered phagocytic activity and the release of a wide array of inflammatory mediators. ³ The nature of this response is highly stimulus‐dependent and results in detrimental or protective microglial phenotypes that can distinctly influence AD progression. ³ In animal models, neuroinflammation has been shown to accelerate Aβ deposition ⁴ , ⁵ and tau spreading ⁶ and directly contribute to neurodegeneration. ⁷ On the other hand, there is evidence suggesting that neuroinflammation can hinder neurodegeneration in AD models. ⁸ , ⁹ , ¹⁰ Together, these seemingly contradictory experimental findings point to a complex relationship between neuroinflammation and neurodegeneration during AD progression.

The intricacy in the results of neuroinflammation studies has also been observed in the literature on human patients, with findings supporting both beneficial and deleterious roles in the progression of AD. ² Specifically, while some biomarker studies point to neuroinflammation as a contributor to tau phosphorylation and accumulation, brain atrophy, and cognitive decline, ¹¹ , ¹² , ¹³ , ¹⁴ others report that neuroinflammation may mitigate these outcomes. ¹⁵ , ¹⁶ , ¹⁷ These divergent results have led researchers to hypothesize that neuroinflammation has complex and context‐dependent effects on the progression of AD patients. ² , ³ , ¹⁸ , ¹⁹ However, the existence of these context‐dependent effects of neuroinflammation in AD patients remains speculative, and their determinants are still unknown. Identifying these determinants is essential for designing the next generation of clinical trials targeting neuroinflammation.

Using positron emission tomography (PET) for the spatial quantification of neuroinflammation, Aβ, and tau burden in the brain, we tested the hypothesis that the effect of neuroinflammation on neurodegeneration is associated with the hierarchical progression of Aβ and tau pathologies across the AD continuum.

2. MATERIALS AND METHODS

2.1. Study population

The participants were from the Translational Biomarkers in Aging and Dementia (TRIAD) cohort and underwent comprehensive neuropsychological assessments, including the Clinical Dementia Rating (CDR) evaluation. Cognitively unimpaired (CU) individuals exhibited no objective cognitive impairment with a global CDR score of 0. Cognitively impaired (CI) patients were characterized by global CDR scores higher than 0. To include CI participants in the AD continuum, only Aβ+ participants were included. ¹ We genotyped study participants for the 18‐kDa translocator protein (TSPO) gene's Ala147Thr polymorphism (rs6971), which predicts the binding affinity of [¹¹C]PBR28 to the TSPO. Among 501 individuals successfully genotyped for the rs6971 polymorphism, 42 (8.4%) were classified as low‐affinity binders, 196 (39.1%) as mixed‐affinity binders, and 263 (52.5%) as high‐affinity binders. To ensure reliable results, only individuals identified as high‐affinity binders underwent [¹¹C]PBR28 imaging and were included in the study. Participants underwent PET scans for Aβ ([¹⁸F]AZD4694), tau ([¹⁸F]MK6240), TSPO ([¹¹C]PBR28), and magnetic resonance imaging (MRI).

2.2. MRI acquisition and processing

T1‐weighted MRI scans were obtained using a 3T Siemens Magnetom scanner. High‐resolution structural images were acquired using the magnetization prepared rapid acquisition gradient echo sequence. These images were corrected for non‐uniformity and field distortion, coregistered, and spatially normalized to Montreal Neurological Institute (MNI) space. Voxel‐based morphometry (VBM) was used to investigate gray matter density. Preprocessing used the standard VBM pipeline in SPM12, with T1 images segmented and aligned to MNI template space using DARTEL. Gray matter maps were smoothed with an 8‐mm full width at half maximum (FWHM) Gaussian kernel.

2.3. PET acquisition and processing

PET images were acquired using a Siemens high‐resolution research tomograph, corrected for attenuation, motion, dead time, decay, and scattered and random coincidences. [¹⁸F]AZD4694, [¹⁸F]MK6240, and [¹¹C]PBR28 images were acquired at 40 to 70 min, 90 to 110 min, and 60 to 90 min after injection, respectively. Images were aligned with native T1 via linear transformations and registered to MNI space. Images were smoothed to a final resolution of 8 mm FWHM. Standard uptake value ratios (SUVRs) were calculated using the whole cerebellum gray matter for [¹⁸F]AZD4694 and [¹¹C]PBR28 and the inferior cerebellum gray matter for [¹⁸F]MK6240. Neocortical Aβ‐PET SUVR was converted to the Centiloid scale. ²⁰ A cutoff of Centiloid > 12, which identifies moderate to frequent neuritic plaques as classified by the Consortium to Establish a Registry for Alzheimer's Disease, ²¹ was used to determine Aβ positivity. Meta‐temporal tau‐PET SUVR was used as global tau load. Tau‐PET Braak‐like stages were calculated from corresponding brain regions, and positivity was defined based on the CU young group. ¹¹ Early tau positivity was based on Braak stages I to IV and late tau positivity on Braak stages V to VI. ¹¹ Regions of interest (ROI) were defined using the Desikan—Killiany–Tourville (DKT) atlas.

RESEARCH IN CONTEXT

1. Systematic review: We reviewed the literature using traditional sources. Neuroinflammation has been associated with both detrimental and protective effects in AD progression. Evidence from animal and human studies remains conflicting, linking neuroinflammation to neurodegeneration as well as neuroprotection. Although these inconsistencies in the literature suggest a context‐dependent effect of neuroinflammation across the AD continuum, the underlying factors of this variability remain unclear.

2. Interpretation: In this PET imaging study, we identified a biphasic pattern in which the deleterious effects of neuroinflammation on brain density were linked to the emergence of Aβ in early stages of the disease and widespread tau accumulation later in the disease continuum.

3. Future directions: Our findings support a two‐wave model of deleterious effects of neuroinflammation in the AD continuum and may inform future therapeutic strategies to target neuroinflammation at the right time.

2.4. Statistical analysis

We categorized our cohort into four disease severity stages: CU Aβ−, CU Aβ+, CI Aβ+ with early tau deposition, and CI Aβ+ with late tau deposition. ¹¹ , ²¹ Analysis of variance (ANOVA) with a Tukey correction for multiple comparisons explored differences between groups for continuous variables. For categorical or ordinal variables, Kruskal–Wallis test was used to identify group differences, followed by post hoc Mann–Whitney U tests for contrasts between groups. TSPO‐PET and VBM values were normalized by centering them around the group mean. Linear regression models were constructed to adjust TSPO‐PET and VBM values for age and sex for all participants, with an additional adjustment for global tau load for the CU group. We extracted residuals for each individual from TSPO‐PET and VBM models and used them in a Spearman's rank cross‐correlation analysis to evaluate the association between TSPO‐PET and VBM in the 37 brain DKT regions. All associations are displayed in the correlation matrixes, regardless of statistical significance. We highlighted significant correlations (p < 0.05) between TSPO‐PET regions with at least five associations with VBM regions in Circos plots and three‐dimensional brain surfaces. We also conducted a sensitivity analysis for these correlations using a stricter p value threshold (p < 0.01). For longitudinal analyses, we calculated the annual rate of change for CDR sum of boxes (CDR‐SB) as (Follow‐up–Baseline)/ΔTime, where ΔTime represents the time interval between baseline and follow‐up measurements in years. We constructed a voxel‐wise interaction model to test the interactive effects of global Aβ/tau load (Aβ Centiloid for CU and tau in the meta‐temporal for CI) and neuroinflammation on the annual rate of change in the CDR‐SB. From the voxel‐wise analysis, we identified the region with the peak t‐value for both the CU and CI groups as the ROI for TSPO‐PET. Using this region, we performed an interaction model between Aβ or tau load and the TSPO‐PET ROI on longitudinal changes in CDR‐SB, accounting for age, sex, and years of education. The statistical analyses were performed in the R statistical software package.

3. RESULTS

3.1. Participants

We assessed 145 individuals (101 CU and 44 CI) with cross‐sectional TSPO, Aβ, and tau PET, and longitudinal clinical assessments (mean follow‐up =  2.05 [0.72] years). Demographic and clinical characteristics of the population at baseline are summarized in Table 1 and at follow‐up in Table S1.

TABLE 1.

Demographics and key characteristics of participants.

Characteristics	Young (n = 21)	CU Aβ− (n = 49)	CU Aβ+ (n = 31)	CI early tau (n = 27)	CI late tau (n = 17)
Age, mean (SD)	23.1 (2.39)	70.1 (8.63) a	73.4 (5.05) a	73.0 (8.58) a	67.7 (8.02) a
Sex, n (percentage female)	12 (57.1)	38 (77.6)	25 (80.6)	12 (44.4) b , c	10 (58.8)
Years of education, mean (SD)	16.8 (1.64)	15.5 (3.95)	14.8 (2.79)	15.4 (2.83)	15.1 (3.40)
Tau load SUVR, mean (SD)	0.825 (0.088)	0.842 (0.086)	0.923 (0.118)	1.08 (0.253) a , b	2.47 (0.802) a , b , c , d
Aβ Centiloid, mean (S.D.)	−0.752 (5.35)	3.96 (5.14)	44.5 (31.5) a , b	68.5 (35.5) a , b , c	103 (19.6) a , b , c , d
APOE ε4, n (% of carriers)	2 (9.5)	9 (18.4)	13 (41.9)	14 (51.9) a , b	9 (52.9) a
CDR‐SB, mean (SD)	0.0 (0.0)	0.0 (0.0)	0.0 (0.0)	0.444 (0.289) a , b , c	0.765 (0.400) a , b , c

Abbreviations: Aβ, amyloid beta; CU, cognitively unimpaired; CI, cognitively impaired; SUVR, standard uptake value ratio; CDR‐SB, Clinical Dementia Rating‐Sum of Boxes; Missing APOE ε4, 1 CI early tau.

^a Different from Young.

^b Different from CU Aβ−.

^c Different from CU Aβ+.

^d Different from CI early tau.

3.2. Aβ is associated with deleterious effect of neuroinflammation on brain density in early AD stages

In CU older individuals, an exploratory cross‐correlation matrix showed a mixed pattern of positive and negative associations between neuroinflammation and gray matter density within and across brain regions (Figures 1A–C). The distribution of rho coefficients was symmetric and centered near zero (skewness = 0.039, kurtosis = 3.11, Figure 1A), suggesting there was no predominance of negative or positive associations. Upon segregation of CU individuals based on Aβ status, we observed two different association patterns. While CU Aβ− showed only positive associations (Figure 1B), CU Aβ+ showed only negative associations between neuroinflammation and brain density (Figure 1C). In CU Aβ−, associations skewed to the positive side (skewness = −0.154, kurtosis = 3.146, Figure 1B), whereas in CU Aβ+, associations skewed to the negative side (skewness = 0.125, kurtosis = 2.784, Figure 1C). Consistent with the pattern observed for gray matter density associations, the CU Aβ+ group showed an overall pattern of negative associations between neuroinflammation and Aβ burden (Table S2). In CU Aβ−, neuroinflammation in the medial orbitofrontal cortex, and subcortical regions such as the hippocampus, nucleus accumbens, and lingual gyrus was significantly associated with preserved gray matter density (p < 0.05, Figures 1E,H; Figures S1 and S2). In CU Aβ+, the paracentral and precentral gyrus, superior frontal cortex, and thalamus were the main regions where neuroinflammation was significantly negatively associated with gray matter density (p < 0.05, Figures 1F,I; Figures S1 and S2).

FIGURE 1.

Aβ associates with detrimental effects of neuroinflammation on gray matter density in early AD stages. (A–C) Cross‐correlation matrix showing regional associations of neuroinflammation with gray matter density in (A) all CU, (B) CU Aβ−, and (C) CU Aβ+ individuals and histogram showing frequency distribution for regional correlation coefficients between neuroinflammation and gray matter density in (A) all CU, (B) CU Aβ−, and (C) CU Aβ+. (D and F) Circos plots showing statistically significant regional associations between neuroinflammation and gray matter density in (D) all CU, (E) CU Aβ−, and (F) CU Aβ+. (G—I) Brain maps showing brain regions where neuroinflammation is positively or negatively associated with gray matter density in (G) all CU, (H) CU Aβ−, and (I) CU Aβ+. Estimates were adjusted by age, sex, and tau load. Green and purple represent positive and negative associations between neuroinflammation and gray matter density, respectively. Aβ, amyloid beta; ACUM, accumbens area; AG, amygdala; cACC, caudal anterior cingulate; cMFC, caudal middle frontal; CUN, cuneus; ENTH, entorhinal; FG, fusiform; HC, hippocampus; infTC, inferior temporal; insula, insula; IP, inferior parietal; isthCC, isthmus cingulate; latOFC, lateral orbitofrontal; LING, lingual; medOFC, medial orbitofrontal; midTC, middle temporal; OC, lateral occipital; PALL, pallidum; paraCent, paracentral; paraHC, parahippocampal; parsOPC, pars opercularis; parsORB, pars orbitalis; parsTRI, pars triangularis; PCC, posterior cingulate; PCUN, precuneus; periCalc, pericalcarine; postCent, postcentral; preCent, precentral; PUTA, putamen; rACC, rostral anterior cingulate; rmidFC, rostral middle frontal; SF, superior frontal; SP, superior parietal; supraMG, supramarginal; supTC, superior temporal; THAL, thalamus proper; transTC, transverse temporal.

3.3. Widespread tau is associated with the deleterious effect of neuroinflammation on brain density in late AD stages

In CI Aβ+ individuals, the exploratory cross‐correlation matrix showed a mixed pattern of positive and negative associations between neuroinflammation and brain density within and across brain regions (Figures 2A–C). The distribution of rho coefficients is slightly skewed toward a negative association (skewness = 0.479, kurtosis = 3.001, Figure 2A). Upon segregation of CI Aβ+ individuals based on the topographical distribution of tau pathology (into early [Braak I to IV] vs late [Braak V to VI] tau stages), we identified two distinct patterns of association. While CI Aβ+ with early tau deposition showed mainly positive associations (Figure 2B), CI Aβ+ with late tau burden showed mainly negative associations between neuroinflammation and gray matter density (Figure 2C). Cumulative distribution showed that in CI Aβ+ with early tau deposition, the distribution skewed to the positive side (skewness = −0.108, kurtosis = 2.478, Figure 2B), while in CI Aβ+ with late tau pathology, the distribution skewed to the negative side (skewness = 0.165, kurtosis = 2.842, Figure 2C).

FIGURE 2.

Widespread tau associates with detrimental effects of neuroinflammation on gray matter density in late AD stages. (A–C) Cross‐correlation matrix showing regional associations of neuroinflammation with gray matter density in (A) all CI Aβ+ individuals, (B) CI Aβ+ with early tau pathology, and (C) CI Aβ+ with late tau pathology and histogram showing frequency distribution for regional correlation coefficients between neuroinflammation and gray matter density in (A) all CI Aβ+ individuals, (B) CI Aβ+ with early tau pathology, and (C) CU Aβ+ with late tau pathology. (D)–(F) Circos plots showing statistically significant regional associations between neuroinflammation and gray matter density in (D) all CI Aβ+ individuals, (E) CI Aβ+ with early tau pathology, and (F) CI Aβ+ with late tau pathology. (G)–(I) Brain maps showing the neuroinflammation brain regions that are positively or negatively associated with gray matter density in (G) all CI Aβ+ individuals, (H) CI Aβ+ with early tau pathology, and (I) CI Aβ+ with late tau pathology. Estimates were adjusted by age, sex, and tau load. Blue and red represent positive and negative associations between neuroinflammation and gray matter density, respectively. Aβ, amyloid beta; ACUM, accumbens area; AG, amygdala; cACC, caudal anterior cingulate; cMFC, caudal middle frontal; CUN, cuneus; ENTH, entorhinal; FG, fusiform; HC, hippocampus; infTC, inferior temporal; insula, insula; IP, inferior parietal; isthCC, isthmus cingulate; latOFC, lateral orbitofrontal; LING, lingual; medOFC, medial orbitofrontal; midTC, middle temporal; OC, lateral occipital; PALL, pallidum; paraCent, paracentral; paraHC, parahippocampal; parsOPC, pars opercularis; parsORB, pars orbitalis; parsTRI, pars triangularis; PCC, posterior cingulate; PCUN, precuneus; periCalc, pericalcarine; postCent, postcentral; preCent, precentral; PUTA, putamen; rACC, rostral anterior cingulate; rmidFC, rostral middle frontal; SF, superior frontal; SP, superior parietal; supraMG, supramarginal; supTC, superior temporal; THAL, thalamus proper; transTC, transverse temporal.

Positive associations between neuroinflammation and tau PET were observed in both early and late tau groups (Table S2). In individuals with early tau deposition, significant associations were limited to early Braak regions. In contrast, individuals with late‐stage tau deposition showed significant associations between neuroinflammation and tau only in late Braak regions, suggesting a spatial progression of neuroinflammatory involvement alongside tau pathology (Table S2). In the early tau stage, the significant associations were predominantly in the temporal cortex and subcortical regions such as the hippocampus, amygdala, and lingual gyrus (Figures 2E,H; Figures S1 and S2). In the late tau stage, the occipital cortex and subcortical regions such as the pallidum were the main regions showing significant negative associations (Figures 2F,I; Figures S1 and S2). Importantly, after excluding the early‐onset AD participants from the late tau group (n = 4), the pattern of associations between neuroinflammation and gray matter density remained similar (Figure S3).

3.4. Aβ, tau, and neuroinflammation interaction is associated with cognitive decline

Neuroinflammation activation alone did not predict future decrease in gray matter density or cognitive decline (CDR‐SB) in either CU Aβ− or Aβ+ individuals (Table S3). In CU, voxel‐wise interaction models revealed that a high global Aβ burden combined with high neuroinflammation levels were synergistically associated with longitudinal changes in CDR‐SB, particularly in parietal and temporal regions (peak t‐value = 9.19, Figure 3A). We also found a synergistic interaction between global Aβ‐PET levels and local neuroinflammation PET on future changes in CDR‐SB (β = 0.679, p < 0.0001, Figure 3B). Similarly, in CI Aβ+, we observed a synergistic effect between tau burden and neuroinflammation on longitudinal changes in CDR‐SB across widespread cortical regions (peak t‐value = 4.52, Figure 3C). Additionally, we found a synergistic interaction between global tau burden and local neuroinflammation PET on future changes in CDR‐SB (β = 0.200, p = 0.0052, Figure 3B).

FIGURE 3.

Aβ− and tau‐associated neuroinflammation effects on cognitive decline across Alzheimer's disease spectrum. (A) Brain maps show voxel‐wise interaction between TSPO‐PET and Aβ on longitudinal changes in CDR‐SB in CU individuals. (B) Graphical representation of interaction between global Aβ and TSPO‐PET (ROI‐based) on longitudinal changes in CDR‐SB in CU individuals. (C) Brain maps show voxel‐wise interaction between TSPO‐PET and Aβ on longitudinal changes in CDR‐SB in CI Aβ+ individuals. (D) Graphical representation of interaction between global tau burden and TSPO‐PET (ROI‐based) on longitudinal changes in CDR‐SB in CI Aβ+ individuals. Aβ, amyloid beta; CDR‐SB, Clinical Dementia Rating‐Sum of Boxes; CU, cognitively unimpaired; CI, cognitively impaired; ROI, region of interest; PET, positron emission tomography.

4. DISCUSSION

Our results suggest that the emergence of Aβ and tau pathologies are associated with the two distinct deleterious effects of neuroinflammation on neurodegeneration across the AD continuum. Specifically, Aβ associated with early deleterious effects of neuroinflammation on gray matter density, while widespread tau potentiated neuroinflammation effects on late degeneration, setting the stage for cognitive decline.

We identified a context‐dependent effect of neuroinflammation on neurodegeneration, modulated by the progression of AD. Previous hypothetical frameworks have proposed a two‐wave effect of neuroinflammation in AD by suggesting that neuroinflammation levels would increase and decrease across the AD spectrum. ² , ¹⁸ Conversely, our group, along with others, showed that neuroinflammation gradually increased with disease progression. ²² , ²³ , ²⁴ Bridging these seemingly conflicting findings, our results indicate that neuroinflammation increases progressively across the AD disease spectrum, exhibiting two waves of detrimental effects associated with the emergence of Aβ deposition or widespread tau pathology. The two waves demonstrated here are supported by the concept that neuroinflammatory states are transient and vary during disease progression. ³ For instance, experimental studies suggest that microglial activation may have an early protective phenotype, mitigating the emergence of brain pathologies such as soluble Aβ pre‐plaque conformations. ²⁵ With the increase in Aβ fibrillar deposition, the phagocytic capacity of microglia becomes dysfunctional and might assume a deleterious phenotype. ²⁶ , ²⁷ Under this construct, Aβ deposition primes activated microglia, leading to the release of inflammatory mediators that potentiate neurodegeneration and memory decline. ²⁸ , ²⁹ Similarly, microglia internalize and degrade toxic tau, ³⁰ with an initial protective mechanism that mitigates tau propagation. ³¹ , ³² As tau tangles deposit, microglia become dysfunctional and release tau seeds that further spread tau, ³³ , ³⁴ marking the beginning of the deleterious neuroinflammation effect associated with tau identified in our study. Together, these results add to current knowledge, suggesting a two‐wave effect of neuroinflammation in AD, modulated hierarchically by the deposition of Aβ and tau proteins throughout the disease's progression.

We found that the presence of Aβ pathology was associated with the first deleterious effect of neuroinflammation in the progression of AD. When we examined older CU individuals as a single group, we found both negative and positive associations between neuroinflammation and neurodegeneration, which aligns with the mixed findings of previous literature. ¹³ , ¹⁵ For example, some studies that evaluated heterogeneous populations (i.e., CU/MCI, Aβ−/Aβ+ together) suggested deleterious effects, ¹³ while others suggested a protective effect of neuroinflammation in AD. ¹⁵ On the other hand, we could argue that previous biomarker studies showing that Aβ modified the effects of neuroinflammation on cognition supported the Aβ phase of neuroinflammation proposed here. ²² , ²³ , ³⁵ Our results show that neuroinflammation in CU Aβ+ associates with degeneration in brain regions known to be associated with early AD, further corroborating this notion. ³⁶ These findings support the idea that Aβ pathology seems to unleash the detrimental effects of neuroinflammation in the early progression of AD‐related neurodegeneration.

Widespread tau deposition is associated with the late deleterious effect of neuroinflammation on neurodegeneration and cognitive decline. Specifically, increased neuroinflammation was associated with preserved gray matter density in individuals in the early stages of tau pathology. Conversely, we found that widespread tau pathology was negatively associated with neuroinflammation‐related gray matter density in regions comprising late Braak stages. Interestingly, it was previously shown that individuals with MCI, who are more likely to be at the early stages of tau deposition, present a potential protective effect of neuroinflammation on gray matter volume. ³⁷ On the other hand, in AD dementia patients, in which widespread tau pathology is more likely, ³⁸ neuroinflammation was associated with reduced cortical volume ³⁹ and worse cognition. ⁴⁰ , ⁴¹ Furthermore, previous TSPO‐PET studies have shown that the presence of tau pathology potentiates the effects of neuroinflammation on cognitive decline in symptomatic AD. ¹² It was shown that inhibition of microglial activation protected against tau toxicity ⁶ and brain atrophy ⁴² in animal models of tau pathology. Similarly, chronic microglial activation in the presence of both Aβ and tau pathology promoted neuritic dystrophy. ⁴³ This supports the notion that the concomitant presence of Aβ, tau, and neuroinflammation abnormalities facilitates forthcoming neurodegeneration and cognitive decline.

Our findings may have implications for clinical trials targeting neuroinflammation in AD. Trial designs for drugs targeting neuroinflammation might have been impacted by the limited knowledge about the natural history of neuroinflammation in different AD stages. The presence of antagonistic effects of neuroinflammation across the disease, as demonstrated here, supports the idea that it is imperative to stage AD individuals using Aβ and tau biomarkers before testing drug effects on inflammatory processes. For instance, elevated TSPO PET is observed in the inferior temporal cortex and lingual gyrus of CI Aβ+ individuals; however, they can exhibit both positive and negative associations with gray matter density, depending on the degree of tau pathology (i.e., early vs late tau stages). These divergent patterns may reflect dynamic changes in microglial phenotype driven by the high tau load in the brain, from protective to potentially detrimental states, within the same anatomical regions. These state transitions have been reported in single‐nucleus transcriptomic analyses of post mortem AD brains. ⁴⁴ , ⁴⁵ In this sense, integrating complementary modalities, such as fluid‐based proteomics, to accurately characterize the dynamic and region‐specific roles of microglia in neurodegenerative processes can further help to increase the chances that the right process will be targeted at the right time. Specifically, participants should be recruited based on their levels of inflammatory markers and Aβ/tau stages to maximize the likelihood that mitigating the neuroinflammation‐related target at that particular stage will be beneficial.

The main strength of our study is the in‐depth phenotypic characterization of participants using multiple PET scans and detailed clinical and cognitive assessments. Furthermore, all individuals were assessed for rs6971 polymorphism in the TSPO gene, which impacts the affinity of the PBR28 tracer TSPO protein, and only high‐affinity binders were included, avoiding corrections for mixed‐affinity cases. Importantly, while this approach minimizes binding variability, it may limit the generalizability of our findings to individuals with diverse TSPO binding profiles. Limitations include that it is uncertain whether TSPO expression indirectly captures microglial activation or microglial density. ⁴⁶ , ⁴⁷ Nonetheless, currently, no technology is better validated for assessing brain microglial activation in living individuals than TSPO‐PET ligands. ⁴⁸ In addition, this imaging modality does not allow for the differentiation of specific neuroinflammation phenotypes, and therefore, microglial phenotypes cannot be directly inferred from this analysis. As a result, while regional variations in TSPO‐PET signals presented in our study may suggest underlying heterogeneity in microglial responses, definitive conclusions regarding microglial phenotype (e.g., protective vs detrimental) cannot be drawn from TSPO imaging alone. Finally, it is important to acknowledge that neuroinflammation is a multifaceted and complex process that extends beyond microglial activation. Thus, other neuroinflammatory players may also contribute to the findings observed in our study.

To conclude, our results support the existence of two waves of deleterious effects of neuroinflammation on AD‐related neurodegeneration that are modulated by the stereotypical brain deposition of Aβ and tau proteins in the patient's brain.

CONFLICT OF INTEREST STATEMENT

S.G. has served as a scientific advisor to Cerveau Therapeutics. V.L.V. received consulting fees from Eli Lilly and Life Molecular Imaging and speaker honoraria from ACE Barcelona and BRI Japan. T.K.K. has consulted for Quanterix Corporation, SpearBio Inc., Neurogen Biomarking LLC., and Alzheon, has served on advisory boards for Siemens Healthineers and Neurogen Biomarking LLC, outside the submitted work. He has received in‐kind research support from Janssen Research Laboratories, SpearBio Inc., and Alamar Biosciences, as well as travel support from the AA and Neurogen Biomarking LLC, outside the submitted work. T.K.K. has received royalties from Bioventix for the transfer of specific antibodies and assays to third‐party organizations. He has received honoraria for speaker/grant review engagements from the NIH, UPENN, UW‐Madison, the Cherry Blossom symposium, the HABS‐HD/ADNI4 Health Enhancement Scientific Program, Advent Health Translational Research Institute, Brain Health Conference, Barcelona‐Pittsburgh Conference, the International Neuropsychological Society, the Icahn School of Medicine at Mount Sinai, and the Quebec Center for Drug Discovery, Canada, all outside the submitted work. T.K.K. is an inventor on several patents and provisional patents regarding biofluid biomarker methods, targets, and reagents/compositions that may generate income for the institution and/or self should they be licensed and/or transferred to another organization. These include WO2020193500A1: Use of a ps396 Assay to Diagnose Tauopathies; US 63/679,361: Methods to Evaluate Early‐Stage Pre‐Tangle TAU Aggregates and Treatment of Alzheimer's Disease Patients; US 63/672,952: Method for the Quantification of Plasma Amyloid‐Beta Biomarkers in Alzheimer's Disease; US 63/693,956: Anti‐tau Protein Antigen Binding Reagents; and 2450702‐2: Detection of Oligomeric Tau and Soluble Tau Aggregates. E.R.Z. serves on the scientific advisory board of Next Innovative Therapeutics (Nintx). P.R.‐N. has served on scientific advisory boards and/or as a consultant for Roche, Novo Nordisk, Eisai, and Cerveau Radiopharmaceuticals. The other authors declare that they have no conflicts of interest. Author disclosures are available in the supporting information.

CONSENT STATEMENT

All study participants provided written informed consent for all study procedures.

Supporting information

Supporting Information

Supporting Information

Povala G, Bellaver B, Bastiani MAD, et al. Amyloid beta and tau are associated with the dual effect of neuroinflammation on neurodegeneration. Alzheimer's Dement. 2025;21:e70746. 10.1002/alz.70746