Fluid Biomarkers in Clinical Trials for Alzheimer’s Disease: Current and Future Application

Core AD CSF biomarkers

The three established core CSF biomarkers, including amyloid-β 1–42 (Aβ₄₂), phosphorylated tau (P-tau), and total tau (T-tau), are employed increasingly to support the improved diagnosis of AD in clinical and research settings [40, 41]. The CSF levels of Aβ₄₂ are markedly reduced in AD patients, which has been reproduced in numerous studies [12]. Use of the CSF Aβ₄₂/Aβ₄₀ ratio may compensate for inter-individual differences in total Aβ production or normalize for pre-analytical confounders affecting both Aβ species, and thereby more precisely identify AD-related Aβ₄₂ reduction [42, 43]_. Both CSF T-tau and P-tau are consistently elevated in AD patients [12]. CSF T-tau can be used as a biomarker of disease progression in AD [44], and the higher the increase, the more intense the neurodegenerative process [45]. CSF P-tau is believed to be a specific marker of tangle pathology and high CSF P-tau is only found in AD rather than other tauopathies [44, 46].

Novel CSF biomarkers

Besides the classical biomarkers, there is a growing number of less well-studied but promising novel biomarkers that relate to different aspects of AD pathology. Aβ oligomers (AβOs) are believed to be the toxic forms of Aβ peptide [47], and novel ultra-sensitive techniques hold the promise of providing clinically useful tests for AβOs [48]. Neurofilament light (NfL) has served as a biomarker of neuroaxonal injury in research [49], and CSF levels of NfL are increased in the early stage of AD and increased over time as the disease progresses [32, 50]. Visinin-like protein 1 (VILIP-1) is a neuronal calcium sensor protein, and elevated CSF levels may be a potential indicator for neuronal loss or neurodegeneration in AD [51]. Neurogranin (Ng) is a post-synaptic protein that is increased in CSF of patients with AD [52, 53], and is supposed to be an effective predictor of cognitive decline [54]. Both Chitinase-3-like protein 1 (also known as YKL–40) [36, 37] and soluble triggering receptor expressed on myeloid cells 2 (sTREM2) [38, 39] are elevated in CSF of AD, and can help in tracking microglial and astroglial-related response along the disease trajectory [15].

Blood biomarkers

In contrast to earlier reports [12], recent studies have demonstrated that plasma Aβ₄₂ levels are reduced in AD and the Aβ₄₂/Aβ₄₀ ratio may be useful to predict brain amyloid status with high accuracy, supporting the use of plasma Aβ measures as screening tools in the detection of AD [21–23, 55]. In addition, several studies have reported a strong correlation between plasma P-tau and CSF P-tau or tau-PET, which indicates the potential of plasma P-tau as a blood-based biomarker for tau pathology [27–30]. So far, NfL is the only marker that has been shown to be directly transferrable from CSF to blood and virtually all CSF findings have been replicated in blood [56]. Blood NfL may be the most promising noninvasive biomarker of active neurodegeneration in the detection and tracking of AD [57, 58].

Fluid biomarkers for participant selection

Most recently, the proposed research framework has defined AD as pathologic process that is identified primarily by biomarker evidence of both amyloid and tau pathology [41]. This approach will not only enable a more precise approach to clinical trials where drug candidates can be targeted on the appropriate people, but also allow for recruitment of individuals in the very early stage of the disease [11]. Thus, an increasing number of clinical trials for AD now use biomarkers of amyloid pathology for the purposes of identifying eligible subjects [62]. A review of 2019 AD drug development pipeline shows that 60% DMT trials in phase III and 39% DMT trials in phase II use Aβ-PET and/or CSF Aβ as inclusion criteria [62]. Another review of current ongoing clinical trials suggests that 62 studies use biomarkers in the selection of participants, and 48.4% of those use CSF Aβ₄₂ or Aβ₄₂/Aβ₄₀ ratio in their protocols [63]. Because CSF Aβ has been shown to have high concordance with Aβ-PET [64–67], there is an emerging use of CSF biomarkers to provide reliable alternatives to PET in AD diagnosis. CSF may be preferred for the demonstration of amyloid pathology, since measurement of CSF biomarkers is considerably cheaper and more widely available than PET scan [68].

However, despite the advantages of confirming the diagnosis of AD with Aβ biomarkers, its diagnostic specificity is still suboptimal. The rising prevalence of amyloid pathology with advancing age can be attributed to a high rate of comorbid Aβ positivity, which will reduce the specificity of Aβ biomarkers [69]. The addition of tau biomarkers may enhance the diagnostic efficiency, since the age-associated increase in tau pathology is less prevalent [70]. Recent studies have suggested that CSF P-tau is able to discriminate AD from other neurodegenerative disease with high specificity, which highlights its potential as a diagnostic biomarker [71, 72]. Interestingly, CSF P-tau abnormality may begin with the initial increases in aggregate Aβ pathology as early as decades before symptoms [26], Thus, CSF P-tau can be used with CSF Aβ as inclusion criterion for clinical trials that target AD, even in its earliest stage.

While CSF biomarkers provide accurate diagnosis of the disease, blood biomarkers are desired for large-scale screening. Ideally, to enrich studies for subjects who are likely to have AD pathology, blood-based biomarkers optimized for sensitivity may be part of a screening funnel that can reduce costs and improve participant experience [73]. Many candidate blood biomarkers have failed in replication studies, but some promising results have been shown for plasma Aβ₄₂/Aβ₄₀ ratio [21–23], P-tau [27–30, 74], and exosomal biomarkers [75] during the past few years. These encouraging results suggest that blood test of Aβ and tau pathology may enable screening of large populations to identify those who need a more specific CSF- or PET-based test. While further studies are needed, the excellent diagnostic performance of blood Aβ and tau also highlights that they can potentially be useful surrogates for CSF/PET biom-arkers as diagnostic procedures in the future.

Fluid biomarkers for target engagement

It is becoming increasingly clear that target engagement biomarkers should be used to identify the downstream effects of drugs on pathogenic mechanisms in early-phase trials [9]. Such biomarkers are especially useful in selecting the most promising drug candidates and their most optimal dose to move forward to late-phase clinical trials. At present, many potential DMTs for AD have focused on the amyloid plaques, so Aβ peptides in CSF are essential for providing evidence of target engagement [76]. For example, β-secretase inhibitions have further reduced the CSF levels of Aβ₄₀ and Aβ₄₂, indicating a strong enzymatic inhibition [77]. Work in animal models indicates that even partial reduction of Aβ₄₂ concentration in CSF is associated with a substantial lowering of amyloid plaque formation and growth [78, 79]. However, the effects of other Aβ-directed therapeutics on CSF Aβ monomers are less predictable, since the successful clearance of amyloid deposits will increase the monomeric but decrease the oligomeric Aβ [68]. Furthermore, with the deepening understanding of the role of AβOs in the neurotoxic etiology of AD, ultrasensitive immunoassays for AβOs in CSF will be necessary to demonstrate target engagement in clinical trials of anti-Aβ agents [48]. In addition, plasma Aβ measures may also be used as evidences to support claims of target engagement [74], but further research is needed to clarify their association with the mechanisms of candidate therapies.

For use in clinical trials of tau-targeted drugs, which are expected to modify tau production or clearance rates, tau biomarkers may be useful in providing evidence of positive target engagement. However, despite a handful of tau-targeted drug candidates have progressed to clinical testing [17], whether the treatment reduce tau levels or prevent the aggregation of tau is seldom tested. In this regard, CSF P-tau may be a well-developed biomarker to measure target engagement, regarding that CSF P-tau levels mirror abnormal tau metabolism (e.g., increased phosphorylation and release of tau from dying neurons) in the brain [72]. It is important to note that tau has > 30 potential phosphorylation sites and site-specific phosphorylation changes occur at different stages of disease progression, so P-tau assays need to be designed specifically for the therapeutic agent and disease stage [26, 80].

While emerging evidence suggests that neuroinflammation plays a causal role in AD pathogenesis, interest in anti-inflammatory treatments is increasing [81]. Biomarkers mirroring neuroinflammatory component of the disease may have a potential for application in the evaluation of target engagement of drug candidates. CSF sTREM2 may be useful to monitor the target effects of therapeutic attempts that aim to modulate microglial function [38, 39]. YKL–40 could be another attractive biomarker candidate for the tracking of microglia activity in future clinical trials [82]. However, the link between these CSF biomarkers and microglia activation in the brain is not firmly established [83]. Therefore, more data are needed to understand how sensitive sTREM2 and YKL–40 are to detect downstream effects on microglia activation of immunomodulatory treatments in AD.

The use of target engagement biomarkers in clinical trials is supposed to improve the success rate of candidate DMTs. However, it is worth noting that demonstration of target engagement does not guarantee efficacy in late-phase trials [84]. In contrast, target engagement in phase II or phase Ib trials provides the basis for deciding if the treatment candidate is viable for further development [84]. This should be considered in explaining the failure of the phase III clinical trials, which have shown effective target engagement in early-phase trials [85]. Of note, a biomarker valuable for target engagement may not have value as a surrogate end point to measure effectiveness. For example, while amyloid marker may be an indicator of target engagement, demonstration of an effect on neurodegeneration is likely to require evidence of changes in the other biomarkers [68].

Fluid biomarkers for supporting disease modification

In current drug development, biomarkers for providing evidence in support of disease modification are in great request, which will evaluate drug efficacy in a more definitive, quick and efficient manner. These biomarkers can be served as surrogate end points in phase III clinical trials to support the approval of drug candidates [59]. Unfortunately, there are no validated surrogate biomarkers known to support a disease modification claim in AD clinical trials. However, this situation does not impede the use of reasonably likely surrogate markers in clinical trials of promising new drug candidates. Indeed, there are good prospects for novel biomarkers of neuronal and synaptic integrity, which are supposed to strongly correlate with cognitive function in AD [86].

As neuronal loss is assumed to correlate with disease progression in AD, being able to monitor neurodegeneration will be an important advantage in clinical trials to detect treatment efficacy [87]. By providing fluid surrogates of neurodegeneration, T-tau, NfL, and VILIP-1 in CSF may assist in tracking clinical response to disease-modifying therapies. T-tau is a well-established CSF biomarker and has been used in clinical trials as indicators of disease modification, which could reflect the intensity of neuronal degeneration in patients with AD [51]. To some extent, NfL has served as a biomarker for neurodegeneration in clinical practice for decades [88]. Recent work has shown that CSF levels of NfL correlate with treatment response in children with spinal muscular atrophy [89]. NfL may, therefore, be a novel biomarker to monitor treatment response also in AD clinical trials. Given that NfL is not specifically involved in AD pathophysiology, it may give more unbiased information than T-tau [56]. Previous findings also highlight the potential usefulness of VILIP-1 as a biomarker surrogate for neurodegeneration in clinical trials of DMTs [87].

Importantly, synaptic damage or loss has been suggested as the best anatomical correlate of cognitive deficits in AD [90]. Recent studies have shown that, in the absence of significant synaptic dysfunction, individuals positive for AD neuropathology may remain cognitively unimpaired [86]. Therapeutic strategies that halt or reduce synaptic damage or loss, with the hope that delay or stop the cognitive impairment of AD, have a strong rationale as DMTs of the disease [90]. Thus, monitoring treatment-related changes of synaptic function may provide suitable substitute end point in clinical trials of DMTs [91]. Currently, a number of emerging CSF biomarkers of synaptic alteration, such as Ng [53, 54], synaptosomal associated protein-25 (SNAP-25) [92], growth associated protein 43 (GAP–43) [93], and neuronal pentraxin 2 (NPTX2) [94], may serve as tools to test for the disease-modifying effects of novel drugs. Ng is the most promising biomarker candidate for synaptic dysfunction in neurodegenerative diseases [95]. However, a combination of synaptic CSF biomarkers may be a more reliable for synapse loss, since it covers many different aspects of synapse pathology (pre-synaptic, post-synaptic, and dendritic) [90].

Preventing or halting the progress of neuronal and synaptic injury is obviously the primary goal of disease modifying AD trials, and this can be ass-essed using CSF biomarkers. Nevertheless, owing to the invasiveness of CSF sampling, peripheral blood biomarkers are desired for repeated sampling to mon-itor the efficacy of therapies. However, few of blood-based biomarkers having a strong link with disease modification has been developed thus far. Encouragingly, recent studies have suggested that plas-ma/serum NfL may be considered as a candidate tool to monitor effects of drug candidates on the intensity of neurodegeneration in other neurological disorders [96, 97], and the same may also be true for AD [57, 99]. Likewise, although there is not enough evidence to support the use of plasma T-tau as monitoring tool in trials of disease-modifying drugs, this biomarker may be suited for use as outcome measures. Furthermore, recent studies have identified that synaptic proteins in blood neuronal-derived exosomes may reflect pathological changes in the AD brain [100, 101]. These findings will provide new opportunities to facilitate minimally invasive evaluation of DMTs in clinical trials, but further investigation is required before widespread use.

Standardization of pre-analytical and analytical variables

Issues with currently available fluid biomarkers, including but not limited to between-laboratory and lot-to-lot variability, greatly limit their potential application in clinical practice and clinical trials [102, 103]. Thus, standardization of pre-analytical protocols and analytical methods is essential. Particularly, standardization can facilitate direct comparisons across studies and improve the potential to measure biological changes resulting from therapeutic intervention [10]. Fortunately, in recent years there have been significant improvements in the standardization efforts for fluid biomarkers in AD research. There are now pre-analytical sample handling recommendations for measuring AD CSF [103–105] and plasma [106, 107] biomarkers that will contribute greatly to the generation of comparable results. Moreover, the core AD CSF biomarkers are available on fully automated immunoassay methods, which has been shown to dramatically reduce the real-world variability [108] and have high concordance with Aβ-PET [64–67]. The production of certified reference materials (CRMs) for CSF Aβ₄₂ is another important progress on standardization, and CRMs for other core AD biomarkers are under development [109].

Interpretation of biomarker results remains problematic

While the application of fluid biomarkers relies heavily on interpretation of data, a robust and validated cut-points that can accurately stratify patient populations and quantity drug effects are required. However, there is no general cut-point and the existing cut-points show high variability among laboratories [110]. Uncertainty about the appropriate cut-point for defining the biomarker-positive population may complicate the clinical trial. Moreover, biomarker results in some cases are difficult to interpret, especially borderline or conflicting results. For example, recent studies have showed that nondemented subjects with lower CSF Aβ₄₂ levels within the normal range is associated with an intermediate risk of clinical progression to AD [111, 112]. This finding suggests the existence of a potential “gray zone” of biomarkers, within which the ability to predict the subsequent cognitive trajectories is lost [113]. In this setting, the use of two thresholds that will allow for three group has been proposed to account for some limitation of the dichotomous classification [41]. Nevertheless, moving beyond and improving the current interpretation and classification of fluid biomarkers are necessary to their use in clinical trials.

Searching for biomarkers as surrogates of assessing treatment response

There is an urgent need for developing reliable and sensitive novel biomarkers that may offer accurate assessment of response to treatment that target underlying pathogenic mechanisms [10]. Identifying biomarkers that strongly associate with cognitive decline and disease progression is especially critical as more clinical trials of potential DMTs shifting the focus towards the preclinical stage of AD. To date, there are no surrogate biomarkers that have been shown to correlate with clinical outcome and supporting a disease modification claim [59]. Whereas several possible biomarkers of treatment response have been proposed, as above mentioned, none has been validated. In addition, there is a need for fluid biomarkers that allow for repeated sampling as surrogate end points to monitor changes over time [114]. Thus, further research focused on identifying pathophysiological biomarkers that are minimally invasive and widely accepted is needed. Ideally, a blood-based biomarker as a reflection of molecular pathogenic mechanisms of AD, could be used to predict treatment response like serum cholesterol is used to predict drug effect in clinical trials for coronary heart disease.

Identifying biomarkers for potential AD subtypes

Beyond the identification of reliable biomarkers for AD overall, there is a need to define biomarkers for potential AD subtypes. The multifactoriality of AD contributes to the heterogeneity of patient populations and makes it difficult to test drugs in clinical trials without pre-selecting appropriate patient groups [115]. Recent studies in cognitive and atrophy subtypes of AD have suggested that clinical heterogeneity has pathophysiological basis [116, 117]. Therefore, one potential future scenario is that fluid biomarkers could be used to sub-classify the clinical syndromes in individual patients according to their pathological signature. In this context, a biomarker-guided pathophysiological stratification approach in clinical trials may assist with participant selection by enriching study populations with individuals more likely to benefit from a given treatment or intervention [118]. Such successful enrichment may improve the efficiency of a trial design by increasing the power of the study and minimizing the required sample size or duration.

Knowledge of biomarker changes improves AD clinical trials

Obviously, understanding the temporal evolution of changes in fluid biomarkers will enable us to better integrate them in the design and interpretation of clinical trials [119, 120]. For instance, studies have reported longitudinal decreases in several CSF biomarkers of neuronal injury in individuals who had symptomatic AD [31]. Knowledge of such within-person patterns of change has important implications for clinical trials in terms of the use of biomarker concentrations as outcome measurements that are dependent on the disease stage [31]. Moreover, substantiating the temporal models of AD biomarker evolution provide insight into the pathologic disease cascade, which can be utilized to initiate targeted therapeutic strategies at an appropriate stage in the disease [121, 122]. For fluid biomarkers to be clinically useful, it is critical to elucidate the dynamics of AD biomarkers during the course of the disease. However, there are few studies that evaluated the longitudinal change in biomarkers in well-defined and large cohorts over a long period of time [123]. So more work is needed to better understand the pattern and time course of biomarker changes in AD, both at the individual and the population level [10].

Biomarker panels may be superior over single biomarker

Fluid biomarkers offer the possibility to detect many pathophysiological processes simultaneously at a relatively affordable cost. Thus, a collaborated panel of fluid biomarkers with multiple components will likely be most useful for clinical trials. Importantly, such biomarker panels will be valuable in enabling the accurate definition of AD subtype and disease onset, and for the monitoring of disease progression and treatment response [15]. For example, combination of fluid biomarkers may be sufficient for the prediction of midterm progression to AD dementia form its prodromal state, which is important for patient selection and stratification in clinical trials [124, 125]. The data indicates that biomarker panels, compared to CSF- or imaging-based single biomarker, may be more powerful in the prediction of the rate of AD progression [124]. Another example, panels of fluid biomarkers can be employed for monitoring the pathological processes phenotypic of AD [86], which will allow for the effect of putative treatments to be evaluated with unprecedented detail [68]. However, biomarker panels for AD pathology are an ongoing pursuit and the efforts for the best combination of biomarkers is still unfolding.