The dynamics of plasma biomarkers across the Alzheimer’s continuum

As a leading cause of dementia, Alzheimer’s disease (AD) is quickly becoming one of the most expensive, lethal, and burdening diseases of the 21st century [1]. China has the largest population of patients with AD, with an estimated 9.83 million individuals aged 60 years or older [2]. Unfortunately, numerous drug candidates for AD have not been successful, and no approved therapies exist at present. This grim situation reinforces the necessity of accurately detecting the neuropathological events (especially the earliest events) in AD, which is a basis for the selection of individuals eligible for treatments [1].

The pathophysiological changes of AD, notably amyloid β (Aβ)-containing senile plaques and tau-containing neurofibrillary tangles, occur years before symptoms arise [3–5]. Current cerebrospinal fluid (CSF) and positron emission tomography (PET) assays enable the robust assessment of classic pathophysiological hallmarks of AD [1, 6]. However, the roll-out of these technologies into clinical practice has been hampered by their high costs, invasiveness, and unavailability in routine settings [7]. Blood tests, which circumvent the aforementioned drawbacks of CSF and PET biomarkers, are gaining increasing attention due to their great potential for real-world implementation [8]. Indeed, several research groups have investigated blood measures of Aβ pathology (Aβ42/Aβ40, Aβ42, and Aβ40), tau deposition (phosphorylated tau181 (p-tau181)), neurodegeneration (neurofilament light (NfL)), and glial activation (glial fibrillary acid protein (GFAP)), and discovered important clues for AD diagnosing, screening and monitoring [1, 9–18].

Nevertheless, the patterns of dynamic changes of various blood biomarkers across the whole Alzheimer’s continuum remain to be explored. On the one hand, recruiting individuals from the preclinical to symptomatic phases of AD is particularly challenging. Several previous studies, restricted by unavailable CSF or PET data, could only focus on the clinical spectrum of AD [11, 19, 20] or simply classify the Alzheimer’s continuum into A− and A+ stages [10, 12]. On the other hand, many studies included a single or very low number of markers but ignored the head-to-head comparison with other biomarkers [11, 21], hence the relative orderings of biomarker changes couldn’t be derived. Recent studies overcame both of the above shortcomings, but they examined only the preclinical stage of AD [22] or preclinical and prodromal stages [23] and lacked an understanding of the whole AD continuum. In addition, these former studies are primarily from western countries, how early the blood biomarkers begin to change, their characteristics of alterations, and whether these biomarkers reflect AD pathologies within the Chinese population is unclear. Because of recent reports on potential racial differences in AD biomarkers [24–26], our study of Chinese cohorts may provide novel insights.

Herein, we aimed to perform a comprehensive comparison of changes in plasma Aβ42/Aβ40, Aβ42, Aβ40, p-tau181, NfL, and GFAP in the context of established CSF biomarkers from the asymptomatic preclinical stage to symptomatic phases of AD. Moreover, we investigated the abilities of these plasma biomarkers to indicate AD pathophysiology. By addressing the above issues, we expected to provide a reference for understanding the main pathophysiological changes and, concomitantly, for better informing the design of clinical trials of AD.

This is the first study to explore the dynamic changes of plasma biomarkers throughout the Alzheimer’s continuum in the Chinese Han population. The main findings are as follows: (1) there was a sharp and sustained increase in plasma GFAP as soon as CSF Aβ became positive, followed by CSF p-tau181 and t-tau, and, to a lesser extent, then plasma p-tau181. However, changes in plasma NfL, Aβ42/Aβ40, Aβ42, and Aβ40 were less pronounced; (2) plasma biomarkers exhibited smaller dynamic ranges than their CSF counterparts, except for GFAP which was the opposite; (3) plasma GFAP and p-tau181 outperformed other plasma biomarkers in tracking AD neuropathology and increased across age only in Aβ-positive individuals.

The cascade of biomarker changes in AD has been hypothesized to follow specific orderings from amyloid to tau, then to neurodegeneration [6], whereas this is not true for plasma biomarkers. Specifically, the earliest dynamic changes in plasma levels occurred in GFAP, which already increased early in the CU A+T− stage and continued to elevate as the disease progressed. This is consistent with previous studies that merely simplified the Alzheimer’s continuum into distinct stages [13, 17, 33], and we further reinforced the notion that plasma GFAP may be an early and sensitive biomarker of AD pathological changes by assessing the alterations of biomarkers as a function of CSF Aβ42/Aβ40 and p-tau181. Synthesizing our and former findings [13, 34], a model of biomarker trajectories in AD may therefore be updated (Fig. 4E). Among the plasma biomarkers studied, GFAP was the one that changed most prominently. However, the magnitude of its increases reported in a recent ALFA+ study was not as large as what we observed [22]. The heterogeneity in modeling approaches and biomarker cutoffs may help account for the discrepancy. Whether this discrepancy could be explained by ethnic heterogeneity remains to be verified by future large-scale studies. Supporting previous views [13, 17, 33, 35], plasma GFAP exerted a desirable accuracy in detecting the presence of Aβ burden and was strongly associated with AD pathologies, with no plasma biomarkers performing better than it. This further emphasizes the value of plasma GFAP in uncovering AD-related neuropathological changes. As a growing number of studies demonstrated the great promise of plasma p-tau231 and p-tau217 in the early identification of AD pathophysiology [12, 21–23], their inclusion in the analysis may yield intriguing results. However, limited by the current technical conditions, we are unable to quantify their concentrations. And this is one of the directions of our future research.

Plasma p-tau181 didn’t rise significantly until the latter CU A+T+ and symptomatic stages. It had a relatively steep increase and wide dynamic range throughout the Alzheimer’s continuum, closely following plasma GFAP. Decades of research have converged to demonstrate the specificity of plasma p-tau181 as an AD blood biomarker [9, 16, 36–40]. As a proof-of-principle, our plasma p-tau181 assay in Chinese people showed identical performances as previously described in other populations [10, 11, 36–39]. Specifically, it had similar but smaller dynamic ranges as CSF p-tau181, as well as close associations with AD pathologies and high accuracy in identifying amyloidosis. The consistent results obtained from diverse populations indirectly prove the reliability of our findings in Chinese cohorts. Plasma NfL barely increased during the preclinical AD stage, while increased markedly in later phases, closer to symptoms onset. This was in agreement with preceding observations [10, 41, 42], and with the fact that plasma NfL had modest accuracy in detecting advanced Aβ PET deposition while low accuracy in detecting earlier CSF Aβ pathology. Plasma Aβ42/Aβ40 decreased from the CU A+T+ stage, in a relatively flat trend. Its associations with AD pathologies were apparent in preclinical and prodromal stages, but not in advanced symptomatic stages. Given its slightly worse performance in discriminating Aβ status than previously reported [9], larger Chinese studies are needed to elucidate whether this difference is related to ethnic inheritance. In contrast, plasma Aβ42 and Aβ40 didn’t show pronounced changes throughout the entire Alzheimer’s continuum and lacked correlations with AD pathologies. This is in line with the previously reported dynamics across the clinical spectrum of AD [19] and the comparable concentrations of plasma Aβ in AD and controls [43], supporting that plasma Aβ levels might reflect peripheral Aβ generation more than they reflect AD brain pathology [19]. However, other studies demonstrated the potential of plasma Aβ markers in capturing early cerebral Aβ changes [10, 22, 23]. With the development of technology, recent studies have revealed that mass spectrometry-based methods outperform most immunoassays in the precise assays of plasma Aβ [44–46]. We thus hypothesize that the lack of significant changes in plasma Aβ levels may be an artifact attributable to the technique employed.

In the present study, we explored plasma biomarker changes in the context of established CSF biomarkers, which enabled us to make a direct comparison of their trajectories. Intriguingly, unlike plasma GFAP, CSF GFAP did not change noticeably across the AD continuum. The finding is surprising as neurologically related blood biomarkers are generally regarded as proxies of their CSF counterparts, but a recent study supports this idea [13]. The divergent clearance mechanisms in biofluids may be one possible interpretation [13]. In addition, plasma GFAP is very stable to freeze-thaw cycles [47], whereas CSF GFAP is far more sensitive over time [48], which may help explain the greater dynamic ranges of the former. In line with previous findings [10], other plasma biomarkers we tested exhibited smaller dynamic ranges than the corresponding CSF biomarkers, suggesting that the former are not sufficiently sensitive to reflect the pathological brain changes when compared to the latter. In this scenario, the early change of plasma GFAP before CSF p-tau181 and t-tau further underscores its potential as an ideal biomarker for tracking AD neuropathology. It is worth noting that although both NfL and t-tau are neurodegenerative biomarkers, plasma and CSF NfL had considerably smaller dynamic ranges than CSF t-tau. Therefore, they may provide different information concerning neurodegeneration. NfL may reflect age-related neuronal and axonal injury independent of Aβ pathology, while t-tau mainly reflects Aβ-dependent neurodegeneration [32, 42].

Plasma GFAP or p-tau181 increases as a function of age only occurred in Aβ-positive but not in negative individuals, suggesting that Aβ pathology underlies the glial or tau-related response. Of note, glial activation is not specific to AD pathophysiology, elevated blood GFAP levels could be observed in other neurological diseases (e.g., FTD, VaD), as our and other studies have shown [49–53]. The present study favors the finding that men had higher levels of plasma p-tau181 [54], whereas other plasma biomarkers were not found to differ by sex. Whether this indicates that men may have a greater susceptibility to tau deposition remains unclear. The effects of sex, often overlooked, do need further investigation to better understand AD pathogenesis and to design precision medicine strategies.

The early and prominent elevation of plasma GFAP, as observed in this study, has important implications for prevention trials of AD, which are increasingly seeking to enroll participants at initial neuropathological events. Substantial neurodegeneration and cognitive decline occur many years (even decades) after overt brain amyloidosis [55, 56], thus providing a window of opportunity for the initiation of disease-modifying therapies. Plasma p-tau181 has been proposed as a prescreening tool to detect Aβ pathology [11, 37, 38]. The present study shows that plasma GFAP may be superior at this task, like the current study [17], and has earlier and more marked changes than plasma p-tau181. This supports the idea of targeting glial markers early in the disease, whenever Aβ pathology emerges and before tau deposition. Moreover, plasma GFAP continued to increase along with disease progression, which implies that there may be late stages where treatment interventions can be effective. Such a biomarker could also be applied to track AD progression up to advanced disease stages. Presently, no effective treatments can cure AD or halt its progression. Elucidating the power of plasma GFAP as an estimator of target engagement is an urgent and arduous task for future clinical trials.

Strengths of our study include using well-characterized cohort data with multimodal biomarkers to determine the trajectories of plasma biomarkers from preclinical to symptomatic phases of AD and compare them directly with established CSF biomarkers. Besides, we tested the relations of plasma biomarkers with AD pathophysiology in different manners. Several caveats should be mentioned. First, this study was limited by its cross-sectional plasma data and could not provide direct insights into the longitudinal dynamics of plasma biomarker changes. Longitudinal studies are necessary to address this issue. Second, the sample size was relatively small. Large studies are needed to further validate and generalize our findings. Third, the hospital-based population selection in this study may include participants with more severe AD-related pathologies or comorbidities. Future studies of Asian populations in community or multicenter cohorts, as well as populations of other races or ethnicities, are warranted. Fourth, due to technical constraints, the definition of Aβ status in the present study is not uniform, that is, by CSF or PET assays. Future well-designed studies are warranted to circumvent this.

In summary, we delineated the trajectories of plasma biomarkers in the context of established CSF biomarkers, in which plasma GFAP and p-tau181 indicated the most pronounced changes and were tightly linked to AD pathologies. As soon as CSF Aβ positivity has been attained, plasma GFAP began to continually increase across the entire Alzheimer’s continuum, followed by CSF p-tau181 and t-tau, and then plasma p-tau181. However, the changes in plasma NfL, Aβ42/Aβ40, Aβ42, and Aβ40 were less prominent. This study gave a new and comprehensive understanding of how plasma biomarkers changed and performed in the Chinese Han population, which has critical practical implications for future trials of AD prevention and treatment.