Background
Recent studies have shown blood phosphorylated tau (p-tau) forms to be reliable biomarkers in supporting a diagnosis of Alzheimer’s disease (AD) [
1‐
5] and in screening for individuals with biomarker evidence of the disease in the absence of cognitive impairment [
6‐
10]. Blood-based p-tau biomarkers increase according to amyloid beta (Aβ) pathology and disease severity, associate well with established cerebrospinal fluid (CSF) and neuroimaging biomarkers, and differentiate biomarker-positive AD dementia from other dementias as well as Aβ-negative controls [
1,
2,
5,
6,
11]. Importantly, blood p-tau is highly accurate at detecting brain amyloidosis and in predicting those who will progress to cognitive impairment and neurodegeneration, often at similar magnitude as CSF p-tau [
2,
6,
8,
12‐
16]. These findings support the integration of blood p-tau analyses into routine clinical assessments and population screening programs to identify biological evidence of AD, especially in the face of the recent approval of the anti-amyloid drug aducanumab (Aduhelm®) and the ongoing consideration of other anti-amyloid therapies by the US Food and Drug Administration (FDA) and equivalent regulatory agencies elsewhere [
17‐
21].
Despite the rapid progress, several analytical hurdles need to be addressed to allow for large-scale adoption of blood p-tau in clinical and research settings. For example, recent head-to-head comparison studies have demonstrated that blood p-tau biomarkers from independent research laboratories and biotechnology/pharmaceutical companies show high inter-assay correlations as well as analytical and diagnostic robustness for clinical use [
22,
23]. However, these studies have been limited to the use of blood processed into ethylenediaminetetraacetic acid (EDTA) plasma [
22,
23]. Despite EDTA plasma being the most commonly used blood matrix type in the dementia biomarker field, it is unclear if other matrices are equally viable given previous reports of large matrix-dependent deviations in biomarker concentrations [
24‐
26]. With blood p-tau already being included in anti-amyloid clinical trial programs and with planned diagnostic applications expected in several clinics, it is vital to ensure the widespread use of these biomarkers in multiple settings including research and medical centers that preferably process blood into other matrix types [
18‐
20,
27,
28].
So far, only p-tau181 has been shown to be measurable in paired samples and in matrices other than EDTA plasma [
1,
24,
26,
29]. Despite strong inter-matrix correlations, p-tau181 concentrations varied significantly between paired samples in different matrices [
24,
26]. This was also true for total-tau, suggesting that variable matrix-dependent concentrations may be common to tau biomarkers and not just to p-tau [
24,
26]. These results also point to a need for thorough verification of tau-based biomarkers in non-EDTA plasma matrices prior to clinical use.
Blood biomarker verification in serum is essential given its widespread use in clinical settings. However, direct comparison of the diagnostic performances of different p-tau forms in serum is limited. In this proof-of-concept study, we investigated if p-tau231 and p-tau181 can be reliably measured in serum versus paired plasma samples to distinguish biomarker-positive AD cases from biomarker-negative controls, as previously shown for plasma and CSF in multiple independent cohorts [
1,
5,
6,
9,
10,
13,
30‐
34]. We then compared inter-matrix agreements between p-tau measures, and further validated the serum performance by evaluating associations with paired CSF samples.
Discussion
The results of this pilot study showed that serum may be a viable blood matrix for the assessment of the novel p-tau231 AD biomarker that has recently been verified for use in plasma [
5,
9,
22,
23,
34]. Increases in serum p-tau231 discriminated between biomarker-positive AD vs. biomarker-negative controls with AUCs of up to 88.2%, equivalent to 90.2% for plasma. Moreover, we extended earlier results that plasma p-tau181 is measurable in serum [
1,
24,
26,
29], demonstrating that diagnostic performance is similarly high when measured in serum or plasma (89.6% and 89.8% vs. 85.4%). We further validated the utility of serum p-tau231 and p-tau181 by showing (i) similar correlations vs. the same biomarkers measured in paired plasma samples and (ii) significant correlations with biomarkers measured in CSF samples from the same individuals. Together, the results suggest that p-tau biomarkers in serum reflect brain pathophysiological changes and may be employed to support clinical and research-based evaluation of AD. Nonetheless, we recorded biases of 15–84% between the absolute concentrations of paired p-tau measures to suggest that biomarker levels are not interchangeable between serum vs plasma. Furthermore, the consistently lower absolute concentrations of p-tau231 in serum compared with plasma may pose analytical challenges that need addressing especially for Aβ-negative individuals, as illustrated in the lower concentration ranges, where the relative differences between the two matrices were larger (Fig.
4D–F).
Blood p-tau biomarkers have been validated in EDTA plasma to show excellent diagnostic and analytical performances [
1,
2,
5,
22,
23,
34,
38], but their utility in other blood matrices is unclear. In the present study, we have demonstrated that the biomarker performances of plasma p-tau231 and p-tau181 are replicable in serum, expanding the repertoire of blood matrix types that are suitable for evaluating these biomarkers. Additionally, p-tau231 and p-tau181 correlated strongly with each other when measured in serum, plasma, and CSF, demonstrating corresponding elevations of both biomarkers in AD that are quantifiable in multiple bodily fluids [
5,
10,
31]. Notably, there were similar correlations of p-tau231 and p-tau181 when both were measured in serum (rho = 0.92) vs
. in plasma (rho = 0.88) in cohort 1. Additionally, the serum-based correlation of p-tau231 and p-tau181 in cohort 3 (rho = 0.93) was similarly as high as in cohort 1. The serum biomarker correlations in each cohort were stronger than previous results in plasma (rho = 0.6) [
5].
While there were strong correlations between paired serum and plasma p-tau levels and either modality accurately differentiated between AD and controls, Bland-Altman plots support a bias in serum vs. plasma concentrations in paired samples, with the disagreements being higher at lower average values. This is explained by the observed differences in absolute levels of p-tau measures in paired samples from the same individuals. For example, paired p-tau231 levels were twofold higher in plasma compared with serum in cohort 1. Similarly, p-tau181 levels were higher in plasma vs
. serum in both cohorts 1 and 2, although these differences reached statistical significance only in cohort 2. These consistently lower p-tau concentrations in serum vs
. in plasma are in agreement with recent reports for p-tau181 and total-tau [
1,
24,
26]. Thus, we propose that the measured concentrations of plasma and serum p-tau biomarkers are not interchangeable despite values in either matrix showing strong correlations and excellent diagnostic performances. In effect, it is important to use either plasma or serum samples independently for an entire study, including longitudinal monitoring, without switching between matrix types.
Translation of our results that serum is equally good for blood-based p-tau analyses as plasma will require careful standardization of how whole blood is processed into serum. At present, the standard operation procedures for serum generation are somewhat vague and are likely to vary between hospital systems and research institutions. For instance, most manufacturers recommend that whole blood should be incubated at room temperature for 30 min to 1 h to form clots that separate the liquid fraction from cellular components. Deviating from these recommendations can have consequences. For instance, removal of cellular elements from the liquid portion is less complete in samples incubated for less than 30 min [
39]. Additionally, samples incubated or transported for over 60 min tend to experience cell lysis to release cellular materials that are not usually found in serum [
39]. For patients or research participants known to be on anticoagulant treatments, longer incubation period may be necessary for clot formation. Standardization of preanalytical factors (e.g., time of clotting, centrifugation, and aliquoting) between and within studies is important to ensure reproducibility of p-tau results collected using serum samples.
Furthermore, p-tau measurements in serum are likely to result in more values below the lower limit of quantification of several p-tau assays given the lower biomarker concentrations in this matrix versus EDTA-plasma. Potential disadvantages include challenges in differentiating cognitively normal individuals with preclinical evidence of AD from those without since the biomarker levels between these groups tend to be marginally different [
19].
Assuming that p-tau secreted to serum is a fraction of the same pool of p-tau molecules that are secreted to the CSF from the brain, we estimated that <1% of CSF p-tau231 concentrations reflect in serum, with comparable results found for p-tau181 (<2%) in cohort 2. The similarities in CSF-to-serum fractions between controls and AD suggest that p-tau transport/release between CSF and blood in normal aging remains unchanged in AD. These results are also in agreement with our previous finding of 5% CSF to plasma fraction [
1].
CSF Aβ
42/Aβ
40 is an early AD biomarker, often becoming abnormal ahead of Aβ-PET [
40]. The accessibility, cost-effectiveness, and simplicity advantages of blood make it a highly attractive biofluid for clinical chemistry evaluation of biological changes in disease. Hence, identifying a blood-based biomarker that performs closely to CSF Aβ
42/Aβ
40 for brain amyloidosis is crucial to enable large-scale screening in population and epidemiological studies for potential preclinical AD candidates for inclusion in anti-amyloid therapeutic trials. This is made even more relevant and urgent by the recent approval of the anti-amyloid drug Aduhelm® and the ongoing consideration of other promising candidate drugs by the FDA [
41]. However, measuring Aβ in blood is analytically challenging due to small dynamic ranges even when measured with immunoprecipitation-mass spectrometry (IPMS) assays [
19]. This challenge is even more acute in serum where concentrations are much lower, correlate poorly with plasma Aβ, and are sensitive to freeze-thaw cycling [
24,
26]. It has therefore been recently recommended that serum be avoided for blood Aβ measurements due to unreliable results [
24]. Even in plasma, some immunoassay methods have poor performance while IPMS techniques that have superior diagnostic utility are time- and resource-intensive [
18,
42]. While plasma IPMS Aβ can sometimes detect amyloidosis equally or slightly better than plasma p-tau particularly in preclinical AD [
7], current technical difficulties limit its throughput and widespread adoption [
17,
18]. Moreover, plasma p-tau217, which has shown substantial potential for brain amyloidosis in blood and CSF [
2,
10,
31,
43], is currently only reliably measurable in plasma; quantification in serum looks less promising given very low concentrations even in plasma, often below the detection limit [
2,
44‐
46]. Since plasma p-tau biomarker levels in preclinical AD are only marginally increased compared with biomarker-negative controls [
1,
2,
6‐
8,
10,
11,
32], robust and reproducible measures are essential. Together, there is presently a limited toolbox of accessible blood biomarkers to screen for and to longitudinally monitor preclinical AD participants undergoing clinical trials. To this end, we reported that plasma p-tau231 starts to increase in the “pre-amyloid phase” before Aβ-PET abnormality thresholds are reached [
5]. Furthermore, plasma p-tau231 outperformed plasma p-tau181 and CSF p-tau217 for preclinical AD, for which reason plasma p-tau181 performed better at separating AD dementia and Aβ+ cognitively unimpaired elderly [
5]. The slightly lower diagnostic accuracy of serum p-tau231 for AD vs
. controls compared with p-tau181 in the present study seems to replicate the preclinical capacity of p-tau231 in brain, CSF, and plasma [
5,
10,
31,
47], since the biomarker-negative control group most likely included those with emerging AD pathology. A significant finding in the present study is that by showing that p-tau231 has good performance in serum that match its demonstrated performance as an early amyloid marker in plasma, we show that the field can expand blood p-tau analyses to include matrices like serum which is preferred in some medical centers and clinical studies.
Major strengths of this study include the evaluation of p-tau231 and p-tau181 in paired plasma and serum (cohorts 1 and 2) as well as paired serum and CSF (cohort 3) samples, providing insights into p-tau levels in the central nervous system and the periphery. Moreover, we used two variations of the same p-tau181 method, the in-house Gothenburg p-tau181 assay and its commercially adapted variant available from Quanterix, meaning that the results obtained herein can be verified in other cohorts or research settings with access to the commercial Simoa method.
Limitations
The study had a decent sample size, given inherent difficulties to obtain paired samples provided at the same patient visit. Future studies should verify the results in larger cohorts and determine any relevant diagnostic advantages or disadvantages between plasma and serum. Regrettably, the small sample size and a lack of CSF Aβ42/Aβ40 or Aβ-PET data prevented in-depth analyses of the diagnostic performance of serum p-tau231 in preclinical AD.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.