Relevance of biomarkers across different neurodegenerative diseases

Alzheimer’s disease (AD) is the most common cause of dementia with an incidence of 1700–2900/100,000 individuals per year in the USA [4]. AD is a dual proteinopathy characterized by the accumulation of tau neurofibrillary tangles and extracellular Aβ plaques [5]. Postmortem studies indicate that AD features a long preclinical phase where tau lesions and associated neuronal loss first appear in the subcortical nuclei and begin to involve limbic regions with associated subjective cognitive decline and neuropsychiatric symptoms [1, 6‐9]. Cortical tau lesions appear in later stages [7], which correlate with the prototypical amnestic Alzheimer’s clinical syndrome [1, 10]. Also in the preclinical phase, Aβ lesions initially appear in the neocortical regions followed by the allocortical, then subcortical, and cerebellar involvement [11]. In contrast to tau lesions, the distribution of Aβ is not significantly associated with symptoms [12]. Besides age, risk factors include cerebrovascular diseases, diabetes, hypertension, obesity, dyslipidemia, and genetic factors such as the APOE-ε4 allele and TREM2 mutations [13, 14]. Approximately 5% of cases have an age of onset under 65 and are referred to as early-onset AD cases. Familial AD, accounting for about 20% of early-onset cases and less than 1% of all AD cases, is mainly caused by rare, dominantly inherited PSEN1, PSEN2, or APP mutations (Table 5) [15‐17].

Synucleinopathies are the second most common neurodegenerative disease [18]. The most common clinical presentation of a synucleinopathy, Parkinson’s disease (PD), has an estimated incidence of 10–18/100,000 individuals per year in the USA [19]. Synucleinopathies feature the aggregation of α-synuclein in the form of Lewy bodies starting subcortically, spreading into limbic and neocortical regions [20]. Generally, synucleinopathies present as PD, PD dementia, or dementia with Lewy bodies, which collectively feature varying degrees of behavioral, cognitive, and autonomous symptoms on top of motor dysfunction. Mutations in a number of genes (Table 5) have been linked to synucleinopathies [10, 21], as has pesticide exposure and traumatic brain injury [22].

FTLD is nearly as prevalent as early-onset AD in dementia cases below the age of 65 with an incidence of 1.6–4.1/100,000 individuals annually in the USA and Europe [23, 24]. FTLD describes a spectrum of diseases characterized by vacuolation, gliosis, and neuronal loss in the external cortical layers of predominantly frontal and temporal neocortices and features the accumulation of tau, TDP-43, or fused-in sarcoma, which define the FTLD subtype [25, 26]. FTLD is associated with a spectrum of frontotemporal dementia (FTD) syndromes ranging from behavioral, language, to motor variants. Amyotrophic lateral sclerosis (ALS) and other motor neuron diseases can be considered part of the FTLD spectrum due to common neuropathologic and genetic features [27]. Roughly one third of all FTD cases are familial [28].

Huntington’s disease (HD) is a rare, autosomal dominant disease with an incidence of 0.16–0.94/100,000 individuals annually [29]. The age of onset of symptoms is typically in the mid-40s but can range from as young as 2 to over 80 and features chorea with behavioral and cognitive symptoms [30]. While there is a correlation between CAG repeat length and age of onset [31, 32], there is a high degree of variability making the predictive power of CAG repeat size poor, illustrating the complexity involved with treating HD. HD features progressive neuronal loss and astrogliosis in the striatum [33] as well as significant degeneration of the cortical regions, cerebellum, and brainstem nuclei [34]. While the huntingtin polyglutamate expansion leads to protein aggregation and transcriptional dysregulation, the mechanisms tying this to neuron loss are unclear.

Prion diseases are characterized by spongiform encephalopathy with underlying neuronal loss, gliosis, and aggregation of the prion protein (PrP) encoded by the PRNP gene [35]. PrP distinguishes prion diseases from other neurodegenerative diseases that may feature prion-like mechanisms, whereby misfolded amyloidogenic proteins spread throughout the brain via seeding aggregation processes. Several genetic, sporadic, and acquired/zoonotic forms of prion diseases exist with sporadic Creutzfeldt-Jakob disease (sCJD), accounting for roughly 85% of cases [36]. With a mortality rate of 1.5–2 individuals per million, sCJD is an extremely rare disease with an incidence rate that is difficult to ascertain [36]. Relevant clinical features include rapid cognitive decline, myoclonus, akinetic mutism, visual and cerebellar difficulties, and pyramidal or extrapyramidal features. A definitive identification requires neuropathological confirmation with postmortem or biopsied tissue. Probable identification requires either an electroencephalogram, magnetic resonance imaging (MRI), or cerebrospinal fluid (CSF)-based biomarkers with the use of real-time quaking-induced conversion (RT-QuIC) preferred over 14-3-3 detection [37].

When combining biomarkers in a study, it is also important to account for differences in how they represent respective disease hallmarks. Importantly, the biomarkers must have a comparable target engagement with their respective hallmarks or, at the least, empirical measures of target engagement to adjust for different sensitivities. For fluid biomarkers, the kinetics of how biomarkers appear in the fluid samples (i.e., how a protein might be digested, processed, or passed into the CSF or blood) may differ and must be accounted for. While expensive and labor-intensive, kinetics can be measured through in vivo labeling techniques [47] that track protein production and clearance. Significant technological gaps exist with the detection of soluble biomarkers in vivo. Of note, lipid peroxidation and metabolism have been implicated in neurodegenerative diseases [48], yet current tools limit the degree to which this can be monitored in vivo. Similarly, robust fluid biomarkers of blood-brain barrier function have been lacking, with several under development.

Idiosyncratic properties of pathophysiological processes can manifest misleading results in biomarker-based studies of pathophysiology. For PET, there is a heterogeneous uptake of different ligands throughout the brain, and differences in this uptake, off-target binding, and imaging sensitivity should be accounted for. In AD, early tau lesions begin in the brainstem nuclei followed by limbic involvement, whereas Aβ lesions begin in the neocortical regions [7, 8, 49]. Several limitations with PET may lead to biases in detecting these two patterns. PET cameras are susceptible to partial volume effects and lack resolution to measure tau ligand binding in the subcortical regions, which are small but can have profound neurobiological and behavioral effects [6, 8, 9, 50‐54]. Lowe and colleagues demonstrated that tau PET positivity denotes individuals who, at autopsy, are all already at Braak stage IV [55], at which point many subcortical structures and limbic regions would already feature significant degeneration [8, 9, 51]. Further, a tau PET imaging study [56], with relatively low SUVR values for positivity, estimated the prevalence of Braak stage 0 individuals as roughly 55% of their study population (n = 161) while a population-based autopsy study [6] featured just 28% of the study population (n = 455) as Braak stage 0. Depending on the thresholds used, tau PET recapitulates the histological progression of tau lesions seen at autopsy; however, the thresholds set may render the tau PET “Braak stage” an underestimate of the true Braak stage at autopsy, particularly at early stages, as seen in these studies [56, 57]. Additionally, some of the first regions to feature protein aggregation in neurodegenerative diseases contain neuromelanin, which many PET ligands will non-specifically bind to [58, 59]. By contrast, the signal from ¹¹C-PiB and other ligands targeting Aβ can detect early, neocortical involvement of plaques [60], making Aβ easier to detect.

Other biological factors, including the significant impact of low-abundance hallmarks such as mutant huntingtin, synaptic proteins, and neurogranin should also be considered, as their low abundance may lead to underestimation of the roles these hallmarks play using event-based modeling approaches. To illustrate, differences in relative abundances of tau versus Aβ lesions, due to tau tangles accumulating within cells and Aβ occurring extracellularly, make the relative determination of proteinopathy onset difficult in vivo. The strengths of Aβ biomarkers may outweigh the weaknesses of tau biomarkers, potentially overrepresenting the relative involvement of Aβ versus tau lesions at different AD stages, especially in early stages. The development of biomarkers and studies with multidisciplinary, postmortem observations are needed to reconcile these issues.

Biomarkers are invaluable for elucidating pathophysiology with appropriate caution. Specific biological questions may call for different implementations of biomarkers, and establishing universal guidelines for their research use would be presumptuous. Biomarker status may be sensitive and specific to certain neuropathological diagnoses; however, one should not assume that clinical utility implies broad-sweeping research utility or vice versa. In vivo, labeling approaches and high-resolution biomarker-neuropathology correlations are critical for assessing the validity of individual biomarkers as they relate to underlying lesions [47]. Unless the extent of the factors connecting lesions to the detection of their corresponding biomarkers is elaborated, one must be cautious when employing biomarkers in studies of disease etiology, particularly when comparing biomarkers with varied biases.

In addition to the array of targets that can be detected by standard fluid and imaging biomarkers, other features of neurodegenerative diseases have been proposed as viable biomarkers of disease progression. One technique, the Multimer Detection System-Oligomeric Aβ, examines the tendency of plasma proteins to oligomerize, circumventing issues in measuring concentrations of Aβ itself [83]. Another uses biophysical properties associated with the propensity of the amyloid protein to form β-sheets to measure blood levels with an immune-infrared sensor assay [84, 85]. In vivo measurement of an early-affected region, the locus coeruleus, is being explored as a viable early biomarker using specialized MRI sequences and analysis [52]. In line with this, pupillometry is being explored as a measure of noradrenergic activity and, thus, locus coeruleus integrity [52]. Signal analysis from transcranial magnetic stimulation, electroencephalography, or magnetoencephalography has also shown promise as a biomarker of functional connectivity and may be sensitive to early changes [86‐88]. Other functional readouts, such as sleep polysomnography and app-based digital phenotyping assessments have been proposed as early biomarkers as well [89‐91].