Astrocyte reactivity influences the association of amyloid-β and tau biomarkers in preclinical Alzheimer’s disease

An unresolved question for the understanding of Alzheimer’s disease (AD) pathophysiology is why a significant percentage of amyloid β (Aβ)-positive cognitively unimpaired (CU) individuals do not develop detectable downstream tau pathology and, consequently, clinical deterioration. In vitro evidence suggests that reactive astrocytes are key to unleashing Aβ effects in pathological tau phosphorylation. In a large study (n=1,016) across three cohorts, we tested whether astrocyte reactivity modulates the association of Aβ with plasma tau phosphorylation in CU people. We found that Aβ pathology was associated with increased plasma phosphorylated tau levels only in individuals positive for astrocyte reactivity (Ast+). Cross-sectional and longitudinal tau-PET analysis revealed that tau tangles accumulated as a function of Aβ burden only in CU Ast+ individuals with a topographic distribution compatible with early AD. Our findings suggest that increased astrocyte reactivity is an important upstream event linking Aβ burden with initial tau pathology which might have implications for the biological definition of preclinical AD and for selecting individuals for early preventive clinical trials.

Clinical studies support that plasma measures of GFAP correlate with its CSF levels, and are increased in CU individuals with AD pathophysiology, representing a robust proxy of astrocyte reactivity in the brains of living individuals [20][21][22] . Based on this previous literature, we designed a multi-site biomarker study including three cohorts to test the hypothesis that the presence of astrocyte reactivity biomarker abnormality is a key element determining the association of Aβ with early tau phosphorylation and aggregation biomarkers in preclinical AD.
To this end, we investigated 1,016 CU individuals (mean age = 69.6 ± 8.9, CDR = 0) from two research (TRIAD, McGill University, Canada and Pittsburgh, University of Pittsburgh, USA) and one communitybased (MYHAT, Pittsburgh, USA) cohort with in vivo biomarkers. Individuals were classi ed as negative (Ast-) or positive (Ast+) for astrocyte reactivity biomarker according to their plasma GFAP levels (see Methods). Demographic and clinical characteristics of participants are summarized in Table 1. Overall, participants classi ed as Aβ+/Ast+ presented increased plasma p-tau181, p-tau231, and p-tau217 compared to other groups. No differences in Aβ levels were observed between CU Aβ+/Ast-and Aβ+/Ast+ in any cohort. Demographic characteristics of individuals segregated by cohort are presented in Supplemental Tables 1-3.
First, we z-scored biomarker levels inside each cohort and applied a robust local weighted regression to model the trajectory of plasma p-tau181, the only p-tau biomarker available in all cohorts, as a function of Aβ burden [plasma or positron emission tomography (PET)] in CU individuals classi ed as Ast-(n = 743) or Ast+ (n = 273). Notably, we observed that plasma p-tau181 levels increased as a function of Aβ only in CU Ast+ individuals (Fig.1a). Similarly, linear regression showed a signi cant association between Aβ burden and plasma p-tau181 in CU Ast+ (β = 0.34, t = 5.37, p < 0.0001; Fig.1b, Supplemental Table 4) but not in CU Ast-(β = 0.04, t = 1.06, p = 0.29; Fig.1b) individuals. A signi cant interaction between Aβ burden and astrocyte reactivity status on plasma p-tau181 (β = 0.31, t = 4.62, p < 0.0001; Fig.1b) further supported the presence of astrocyte reactivity was key to determining Aβ effects on tau phosphorylation.
Local weighted regression using only continuous values for Aβ, p-tau181, and GFAP levels con rmed that these results were not in uenced by biomarker thresholds (β = 0.10, t = 3.22, p = 0.0013, Fig.1c). Cohen's d analysis revealed that the presence of Aβ+ and Ast+ has a large magnitude of effect on tau phosphorylation (Cohen's d = 0.80), whereas Aβ+ in the absence of Ast+ presented a negligible effect size (Fig.1d). Voxel-wise analysis con rmed that Aβ levels in brain regions known to present early Aβ plaque accumulation in AD, including the posterior cingulate, precuneus, and insula 23 associated with plasma p-tau181 only in the presence of astrocyte reactivity (Fig.1e).
We also explored the impact of Ast+ in the associations of Aβ burden with plasma p-tau231 (available for Pittsburgh and TRIAD cohorts, n = 502) and p-tau217 (available for the TRIAD cohort, n = 136) levels in subsets of individuals that had these markers available. Plasma p-tau231 increased as a function of Aβ only in CU Ast+ individuals (Fig.1l). Additionally, we found a signi cant association between Aβ and plasma p-tau231 in CU Ast+ (β = 0.36, t = 4.62, p < 0.0001; Fig.1m, Supplemental Table 5) but not in CU Ast-individuals (β = 0.10, t = 1.87, p = 0.06). We also observed a signi cant interaction between Aβ and astrocyte reactivity status on plasma p-tau231 (β = 0.26, t = 2.84, p = 0.004; Fig.1m). Cohen's d analysis suggests that the presence of both Aβ+ and Ast+ also had a strong effect on the levels of p-tau231 (Cohen's d=0.91), whereas pathologies independently did not have a signi cant effect (Fig.1n). Similarly, plasma p-tau217 presented a steeper increase as a function of Aβ burden in Ast+ compared to Ast-  Table 5). An association between Aβ burden and plasma p-tau217 was observed in Ast-(β = 0.19, t = 3.48, p = 0.0008, Fig.1p), but with a much larger magnitude in Ast+ (β = 0.74, t = 5.62, p < 0.0001, Fig.1p) individuals. The stronger association in CU Ast+ individuals was further evidenced by a signi cant interaction between Aβ burden and astrocyte reactivity status on plasma p-tau217 (β = 0.52, t = 3.61, p = 0.0004, Fig.1p). The presence of both Aβ+ and Ast+ had the largest effect size on plasma p-tau217 increase (Cohen's d = 1.41; Fig.1q) compared to p-tau181 and p-tau231. Importantly, a sex effect was observed in the association between Aβ and plasma p-tau epitopes in CU Ast+ individuals in all cohorts, with the association being stronger in men than women (Fig.1r, Supplemental Fig.2). The greater effect of Aβ burden on tau phosphorylation in the presence of Ast+ in men than in women may prove to play a role in the larger magnitude of effect of anti-Aβ therapies in man 24 . Finally, the presence of astrocyte reactivity did not impact the association between Aβ burden and NfL levels in any of the three cohorts (Supplemental Table 6), supporting that astrocyte reactivity unleashes Aβ effects on early tau pathology.
We used PET imaging available in the TRIAD cohort to determine the topographic localization of p-tau protein aggregates in the form of tangles (n = 147). Tau-PET deposition occurred as a function of Aβ burden only in CU Ast+ and in regions expected to present the earliest tau deposition (Fig.2a), affecting 100% and 62% of the extension of the Braak I and II regions, respectively (Fig.2b). As expected, in later Braak regions tau-PET did not increase as a function of Aβ in either group (Fig.2b). Finally, we investigated the link of baseline Aβ and astrocyte reactivity status with future tau-PET burden (n = 71; mean follow-up = 2.3 years; Supplemental Table 7). We observed that the annual rate of tau-PET accumulation was higher in the CU Ast+ group (Fig.2c) and predicted by baseline Aβ burden only in CU Ast+ (Fig.2d). Interestingly, while the baseline association was con ned to the mesial temporal cortex, the longitudinal tau-PET accumulation as a function of Aβ/Ast presented initial tau spread over the neocortex in Braak III-IV regions (Fig.2e), further supporting the notion that these individuals are following a tau accumulation pathway consistent with AD progression 25 .
In summary, we provide biomarker evidence across multiple cohorts that increased astrocyte reactivity, measured by a plasma GFAP assay, plays a key role in the association of Aβ with early tau pathology in preclinical AD. The fact that the presence of abnormal astrocyte reactivity potentiates Aβ-triggered tau pathology may prove to favor the inclusion of astrocyte reactivity biomarkers in the biomarker modeling 1 and biological de nitions 26 of AD. The strengths of our study include large sample size and the use of well-characterized research and population-based cohorts. While our cohort represents signi cant socioeconomic diversity, the main limitation is that our cohorts are composed mainly of White participants, which limits the generalizability of our ndings to a more diverse world population. As biomarkers are naturally continuous, dichotomizing thresholds are invariably subject to conceptual and analytical idiosyncrasies and may change depending on the method used.
Furthermore, our ndings support recent observations suggesting that plasma p-tau is a state marker of Aβ in preclinical AD 3 , but also add that this occurs mainly with the concomitant presence of astrocyte reactivity biomarker abnormality. As preventive clinical trials have increasingly focused on individuals in the earliest preclinical phases of AD, our results highlight that the selection of CU individuals Aβ+/Ast+ without overt p-tau abnormality may offer a time window very early in the disease process but with increased risk of AD-related progression. Finally, based on our results, we may speculate that a combination of drugs targeting both Aβ and astrocyte activation can potentiate the prevention of early tau pathology in preclinical AD.

Con icts of interest
HZ has served at scienti c advisory boards and/or as a consultant for Abbvie, Acumen, Alector, Alzinova, ALZPath, Annexon, Apellis, Artery Therapeutics, AZTherapies, CogRx, Denali, Eisai, Nervgen, Novo Nordisk, Optoceutics, Passage Bio, Pinteon Therapeutics, Prothena, Red Abbey Labs, reMYND, Roche, Samumed, Siemens Healthineers, Triplet Therapeutics, and Wave, has given lectures in symposia sponsored by Cellectricon, Fujirebio, Alzecure, Biogen, and Roche, and is a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program (outside submitted work). KB has served as a consultant, at advisory boards, or at data monitoring committees for Abcam, Axon, Biogen, JOMDD/Shimadzu. Julius Clinical, Lilly, MagQu, Novartis, Prothena, Roche Diagnostics, and Siemens Healthineers, and is a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program, all unrelated to the work presented in this paper. SG has served as a scienti c advisor to Cerveau Therapeutics. ERZ serves in the scienti c advisory board of Next Innovative Therapeutics (Nintx). PRN has served at scienti c advisory boards and/or as a consultant for Roche, Novo Nordisk, Eisai and Cerveau radipharmaceuticals. NJA has given lectures in symposia sponsored by Lilly and Quanterix. The other authors declare that they have no con ict of interest.
3 . Jack, C.R., Jr., et al. De ning imaging biomarker cut points for brain aging and Alzheimer's disease.
Alzheimers Dement 13, 205-216 (2017 29 , the Normal Aging study 30 , and the MsBrain 31 . CU individuals were classi ed using either CDR = 0 or MoCA > 25 (for the MsBrain study). Individuals were selected according to cognitive status and plasma biomarker availability. Details of each cohort recruitment are reported in the Supplemental Table 8.

Plasma Biomarkers
For Pittsburgh and TRIAD cohorts, plasma biomarkers (except for plasma p-tau217) were measured using Single molecule array (Simoa) methods on an HD-X instrument (Quanterix, Billerica, MA, USA), at the Clinical Neurochemistry Laboratory at the University of Gothenburg, Sweden. Plasma Aβ42, Aβ40, GFAP, and NfL were quanti ed with the Neurology 4-Plex E (#103670) commercial assays from Quanterix in CU individuals 35 . As younger individuals are expected to present less AD-related pathology 36,37 , cutoffs for astrocyte reactivity were generated using plasma GFAP mean of the 15% youngest Aβ-negative individuals plus 2 standard deviations (s.d).