A Critical Evaluation of Wet Biomarkers for Huntington’s Disease: Current Status and Ways Forward

Biomarker definition

The term biomarker is defined as “a characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention” [8]. Therefore, a biomarker can be related to the disease itself or to a treatment effect. Various biomarker categories are defined in Table 1 (based on BEST (Biomarkers, EndpointS, and other Tools) Resource [9]). An ideal disease biomarker should be closely linked to pathophysiology, reliable, accurate, sensitive, specific, reproducible, inexpensive, non-invasive and acceptable for the patient [10].

Huntington’s disease biomarkers: Unmet needs

At present there are no effective treatments in clinical practice that will prevent the disease, halt its progression or delay its onset [11–13], but as the future holds promise for disease modifying strategies [14], there is a need for reliably assessable disease progression biomarkers, as well as markers to evaluate therapeutic interventions. As a chronic progressive disorder, HD has a premanifest phase during which no motor signs are present, but the underlying pathological processes are ongoing [15, 16]. The possibility of genetic testing in HD opens up the possibility of evaluating disease-modifying treatments already in the premanifest phase. Thus, biomarkers that reflect underlying disease processes or progression are needed to monitor changes in clinically asymptomatic individuals [17]. Biomarkers providing an objective measurement for the assessment of HD severity would also be valuable in the clinical care of existing HD patients and might prove particularly useful in a clinical trial context, potentially serving as surrogate endpoints. In 2004, the U.S. Food and Drug Administration (FDA) emphasised the importance of biomarkers in improving drug development efficiency [18] and continues to promote research on potential new biomarkers to use as surrogates in future trials [19]. Since then, there has been an increase in the use of biomarkers in all phases of drug development. Impaired proteasome activity [20], transcriptional dysregulation [21], oxidative stress [22], mitochondrial and metabolic dysfunction [23], neuroinflammation [24, 25], and microglial activation [26–28] have been shown to play important roles in the pathogenesis of HD and many biomarker studies have been undertaken that reflect these processes (clinical, imaging and biofluid). Although much data has been generated, the road to clinical application is long and validation of the intended use in independent cohorts is challenging. To help identify the most promising candidate biomarkers, we reviewed “wet biomarker” literature in HD published to date. Here, we are defining a “wet biomarker” as a potential biomarker that is objectively measured in a bodily fluid. We have limited our focus to studies in blood, urine, saliva and CSF and have excluded DNA, RNA and microRNA studies. While these studies are also of great interest, they require a different set of criteria from markers discussed here and are beyond the scope of this review. We have built our discussion around a table giving an overview of wet biomarker studies to date. The supplementary table is an unbiased summary including all studies regardless of methodology and reproducibility, describing the markers evaluated, the number of subjects, disease stage and the main finding of the study. At the outset, our intention was to also include a table showing “validated” wet biomarkers that have been consistently reproduced and validated longitudinally; however, only one of the markers reviewed here met these criteria. None of the studies to date have yielded biomarkers for accurately predicting either age at onset or disease progression for HD. Since the first step towards clinical implementation of a newly discovered biomarker is independent replication, we focus on biomarkers that have been validated in at least two independent cohorts and we discuss recent data challenging earlier findings. The goal of this critical review is to highlight the most promising markers for future validation studies and to help investigators design future biomarker studies in HD.

Immune markers (Table 2)

Both innate and adaptive immune systems have been suggested to play a role in HD pathology [29] and a number of studies have highlighted the existence of a peripheral immune response in HD patients. An unbiased proteomics screen of HD plasma identified immune proteins that are elevated in HD compared to healthy controls [30], including proinflammatory cytokine IL-6, acute-phase protein alpha-2-macroglobulin, complement factors and a complement inhibitor clusterin (Table 2). The authors hypothesised that IL-6 induces the release of acute-phase proteins and this in turn leads to the activation of the complement cascade and modulating factors such as clusterin [30]. A follow-up study demonstrated that immune changes are apparent even during the preclinical stage of HD [24] with IL-6 increased even in premanifest subjects, presenting the possibility that inflammatory markers in plasma can be used to track HD. The findings of increased IL-6 levels in HD plasma were replicated in three independent studies [31–33], however, no change has also been reported [34–36] (Table 2).

Acute-phase proteins alpha-1 antitrypsin [37], alpha-2-macroglobulin [30, 37] and C-reactive protein (CRP) [31, 37–40] have been investigated as potential biomarkers in HD. One study found decreased levels of CRP in one HD cohort and no significant difference in another [37], whereas several other studies reported raised CRP levels in HD subjects [31, 40] (Table 2). Increased CRP level may reflect an acute-phase response due to mutant huntingtin expression in HD. However, as HD mutation carriers that use antipsychotics are prone to develop an acute phase response [39], it is possible that the CRP increase reflects the use of antipsychotics by HD subjects studied. It still remains to be confirmed whether CRP is indeed increased in HD and if yes, whether this increase tracks with disease progression.

Neopterin is synthesised by macrophages following interferon gamma stimulation and is a marker of cellular immune system activation. Several studies have consistently reported increased neopterin levels in late stages of HD [40–43]; however, no follow-up studies in early stages of the disease have been performed.

Metabolic markers (Table 3)

It is reported that there is a negative energy balance in HD [44] and many of the peripheral manifestations of HD, such as weight loss [45] and muscle wasting are indicative of metabolic alterations. This is reflected in the fact that many of the analytes tested as potential biomarkers for HD are metabolites (Table 3). Advances in metabolomics provide the opportunity to exploit this catabolic phenotype to discover novel biomarkers for HD. An untargeted metabolomic study using gas chromatography and mass spectrometry to analyse serum from control subjects, premanifest and early stage HD found metabolic alterations associated with a catabolic phenotype [46], thus confirming longstanding reports that catabolic changes in amino acid metabolism occur before onset of symptoms [47]. Levels of branched chain amino acids (valine, leucine and isoleucine) have been consistently reported to be correlated with weight loss, disease progression and abnormal triplet repeat expansion [36, 46–55] (Table 3). However, there are also a number of studies that report no change in branched chain amino acids [38, 56–58].

Many studies have investigated markers of lipid metabolism, in particular, cholesterol metabolism in HD. Most studies have shown no alteration in peripheral levels of total cholesterol, HDL-cholesterol and LDL-cholesterol [57, 59–64], with the exception of Wang et al. [34] that showed a reduction of all three analytes in premanifest as well as manifest HD subjects. The major brain cholesterol metabolite 24(S) hydroxycholesterol (24OHC) reduction has consistently been observed in plasma of individuals with HD [59, 66] and was paralleled by the reduction of caudate volume, suggesting that the reduction of 24OHC may reflect progressive neuronal loss [59] (Table 3). Additionally, HD patients’ plasma also showed reduced levels of cholesterol precursors lanosterol and lathosterol, and of the bile acid precursor 27-hydroxycholesterol [65]. Future studies need to follow subjects longitudinally to examine the rate of change of metabolic markers as the disease progresses.

Endocrine markers (Table 4)

In the tuberal nucleus of the lateral hypothalamus of HD patients there is progressive neuronal death that could have an impact on the function of most of the pituitary axes [67–72]. Consequently, numerous neuroendocrine studies have been carried out in HD, although sometimes with conflicting results (Table 4).

Oxidative stress markers (Table 5)

Evidence from studies in both humans and animal models suggests the involvement of energy metabolism dysfunction and oxidative stress in the pathogenesis of HD [98, 99]. It is thought that impairment in the electron transport chain and mitochondrial dysfunction are the major mechanisms involved in increased reactive oxygen species (ROS) production in HD [100, 101]. It has been reported that HD individuals have an increased level of oxidative stress markers (Table 5) accompanied by a reduction in antioxidant systems compared to healthy subjects [102, 103]. Several studies have reported enhanced lipid peroxidation in individuals with HD [40, 102–106] (Table 5). Chen et al. [103] detected a correlation between lipid peroxidation products in plasma and degree of severity in patients with HD, while Klepac et al. [105] reported an increase in plasma lipid peroxidation accompanied by reduced glutathione content (Table 5). An increase in Cu/Zn-superoxide dismutase (Cu/Zn-SOD), an important antioxidant enzyme, has also been reported in HD [62, 64] (Table 5). In addition, alterations in the kinetics of the kynurenine pathway (a major route of tryptophan catabolism) have been reported in patients with HD [40, 107] (Table 5), although some studies find no alterations [43, 108].

Axonal and glial degeneration markers (Table 6)

HD is associated with a variety of pathological changes affecting both glial and neuronal brain tissue and these changes may be mirrored in the release of proteins into the CSF, and to lesser extent into the blood. Markers that reflect CNS pathology are needed for disease-modifying agents that target the CNS specifically. Currently, the most promising marker of HD onset and progression is neurofilament light protein (NF-L), a subunit of neurofilaments that make up the neuronal cytoskeleton (Table 6). Four studies have consistently found raised NF-L in CSF of individuals with HD [109–112] and several studies have found CSF NF-L to correlate with clinical measures [109, 111–114]. A recent, longitudinal study has found raised plasma NF-L in HD individuals, a finding confirmed in a separate, cross-sectional cohort [112] (Table 6). This is a very important finding because plasma NF-L levels reflect CNS pathology and may be a useful marker for trials involving CNS-delivery of disease-modifying agents. Importantly, both CSF and plasma NF-L levels have been found to correlate with disease onset in HD [112, 113] and brain atrophy [115]. A neuronal damage marker, NSE, is also increased in serum of HD patients [62, 64] (Table 6).

Several other markers of axonal and glial degeneration have been reported to be increased in HD individuals’ CSF including myelin basic protein (MBP) [111], total tau (T-tau) [110, 116], and chitinase-3-like-protein 1 (CHI3L1) / YKL-40 [110, 117]. T-tau and YKL-40 have also been found to correlate with clinical measures in some, but not all, of the studies [110, 118].

These studies suggest that the profile of glial-related inflammatory CSF biomarkers (such as YKL-40, GFAP, MCP-1, sCD14) and cytoskeletal and myelin markers of neurodegeneration (such as NF-L, T-Tau, P-Tau181, MBP, NSE, UCHL1), and their relation to disease severity should be further investigated.

CSF mutant huntingtin (Table 7)

Unsurprisingly, mutant huntingtin (mtHTT) has been found to be absent in healthy controls and increased in manifest HD CSF compared to premanifest HD CSF [114, 119] (Table 7). It is promising to note that CSF mtHTT levels have shown associations with clinical phenotype [114, 119].

Why are studies not reproducible?

When conducting a biomarker study, it is important to consider the future purpose of the biomarker (its context-of-use) [120, 121] (see Fig. 1). For instance, if the purpose of the biomarker is to track disease progression, it should alter with advancing disease. Therefore, it is important to note that although many of the analytes in Supplementary Table 1 are not useful markers for HD in terms of disease progression, that does not mean that they would not be useful as, for example, pharmacodynamic markers.

OPEN IN VIEWER

Fig.1

Phases of biomarker development and recommendations for biomarker discovery studies. Biomarker development from target selection and biomarker discovery to the product launch of a clinical assay that has been validated in large cohorts is a time-consuming, expensive process. It is, therefore, important that makers put forward for clinical validation are based on solid evidence. Here, we suggest factors to consider during biomarker study design to ensure meaningful data are produced. Adapted from [273].

This review of published HD biomarker studies to date reveals that very few wet biomarkers have been consistently and reliably reproduced; there are many conflicting studies that could be due to a myriad of variability factors. In biomarker discovery studies, great attention needs to be paid to several aspects of study design, sample collection, sample measurement and data analysis [122–125]. Factors to consider during biomarker discovery are outlined in Fig. 1. Below we discuss potential sources of variability in biomarker studies and suggest considerations for further developments and harmonisation of standard operating procedures (SOPs).

Biological variation

Within-subject (intra-individual) biological variation is the inherent change in analyte results around a homeostatic setting point within an individual over time [126, 127]. Individual homeostatic setting points are different and this difference between individuals is termed between-subject (inter-individual) biological variation [126, 127]. Some analytes have predictable biological rhythms that may cycle daily (e.g. cortisol, GH) [128], monthly (e.g., luteinizing hormone (LH), follicle-stimulating hormone (FSH), progesterone) [129, 130] or seasonally (e.g., vitamin D, cholesterol) [131–133]. With advancing age blood levels of some hormones increase (e.g., cortisol [134, 135], FSH [136], LH [137], noradrenaline [138]); decrease (e.g., melatonin [139], aldosterone [140, 141], GH [142], oestrogen in women [143], prolactin in women [144]); or change only slightly (e.g., adrenaline [145], insulin [146], thyroid hormone T4 [147]). In addition, some hormones are metabolised more slowly with age [147].

Some analytes can even be affected by activity levels; for instance, exercise prior to sample collection can result in higher cytokine and myokine levels in plasma [148]. Many HD individuals require medication for their symptoms and it is often difficult to get a medication-free cohort. Medication can influence analyte levels, e.g., [149, 150], and thus needs to be taken into consideration when interpreting results.

A large biological variability requires studies with larger subject cohorts. Lessons learned from cardiovascular biomarker research illustrate the importance of effect size; the prognostic effect was shown to be significantly stronger in datasets from observational studies than in datasets from randomised controlled trials [151]. As can be observed from Supplementary Table 1, many HD biomarkers studies to date have used less than 20 subjects per group, making it likely that the studies are underpowered.

Biological variation due to factors such as age, gender, diet, ethnicity, smoking, alcohol consumption, or medication may, at least in part, explain the contradicting results seen in different studies [152]. For example, for haemoglobin, the main sources of between-subject variation is clinical indication, but following that gender, age and nationality greatly influence variation [153], whereas CRP is not only affected by age, but also, to a large extent, gender [154, 155]. Confounding factors may be controlled for by selecting the most appropriate time of day, month or year for sample collection or may be taken into consideration in the interpretation of results. A study examining classification of control subjects from Alzheimer’s disease patients showed that removal of causes of variability such as age and gender can significantly improve classification accuracy [156]. Therefore, it is paramount to have a well-characterised cohort and a balanced design that minimises the number of uncontrolled covariates such as age and sex. If they are not matched, confounding factors should at the very least be controlled for in the statistical analyses [157, 158].

Study design: Considering preanalytical variables

Preanalytical variables such as time of day that the sample is obtained, anticoagulant choice, feeding state and acute stress of the patient, and method and duration of sample storage can influence reported analyte levels and should be documented [159, 160]. Below we discuss how pre-analytical variables contribute to variation in biomarker studies.

Analytical variables

Importance of replication and validation

As can be seen in Supplementary Table 1, many biomarkers have been proposed for HD, however, most have been evaluated in only a few studies. For a biomarker to be recognised, there is a need for validation and independent replication [217, 218]. Validation generally refers to a replication experiment performed by the same research laboratory with a different technology or a different sample cohort, whereas an independent replication of results is usually performed by an outside laboratory (Table 8). Validation for the intended purpose, at both technical and clinical levels, is essential in biomarker studies. The best biological validation of a biomarker is via its clinical correlations and clinical relevance, showing consistent data in cross-sectional and longitudinal studies and repeatability across different clinical studies.

The need for replication of biomarker data is best exemplified by 8-hydroxy-2-deoxyguanosine (8-OHdG), a marker of DNA damage produced when guanine is oxidised by reactive oxidative species, that was initially thought to be a promising disease severity marker for HD [reviewed by [219]]. 8-OHdG levels were reported to be elevated serum [220], plasma [102] and leukocytes [103] of HD individuals and plasma levels were found to correlate with proximity to projected clinical diagnosis in premanifest HD [221]. However, subsequent studies failed to replicate these findings [222, 223] bringing into question the validity of 8-OHdG as a marker of disease state in HD.

Comparisons can be “statistically significant” (i.e., have a p < 0.05) even if no true relationship is present. 8-OHdG has a high biological variability so it is unsurprising that 8-OHdG measurements yielded false positive results. Elevated 8-OHdG is associated with confounding factors that are common in HD patients including stress [224], depression [225], smoking and low weight [226]. 8-OHdG levels are also influenced by exposure to pollution [227] and polyphenol-rich foods such as vegetables and red wine [228, 229].

The study by Borowsky et al. [223] challenging previous claims that 8OHdG is a useful clinical biomarker for HD progression is a good example of a well-designed biomarker study that included good practice of carrying out blinded sample analysis, use of independent analytic methods, estimation of accuracy of analytic methods and stringent collection and storage of samples.

Blinded analysis

Interpretation of data is seldom completely objective and is, as such, vulnerable to prior convictions and conflict of interest [230]. To reduce research bias in biomarker studies, blinded sample analysis and data interpretation is recommended [231, 232].

Biomarker discovery platforms

In general, there are two principal approaches for biofluid biomarker discovery: targeted and untargeted. The targeted, “pathophysiologic” approach tests a prespecified hypothesis. It involves knowledge of the relevant human physiology and disease processes, allowing the discovery of novel biomarkers that are produced, released, or cleared during the disease process. In contrast, the untargeted approach is an unbiased approach where a range of “omics” profiling technologies are generally used to systematically screen body fluids for novel biomarkers, paving the way to panels of multiple biomarkers that reflect multiple disease mechanisms. Proteomics and metabolomics are increasingly being used for biomarker discovery in neurodegenerative diseases [reviewed by [233] and [234]], including HD [30, 46]. Since the pathophysiology of HD is heterogeneous, with multiple mechanisms involved at every clinical or pathological stage, until the mechanism of HD fully understood, we must be open to novel mechanisms, and untargeted discovery using various “omics” technologies may lead to the discovery of novel markers. An integrated, multi-omics approach combining genomics, transcriptomics, proteomics, and metabolomics technologies can simultaneously detect hundreds of molecules simultaneously, some of which can be altered in the disease state which should result in a more complete biomarker fingerprint and give insight to fundamental mechanisms of HD.

To develop accurate markers of disease progression or therapeutic response that are clinically useful, it is necessary to combine data from wet biomarker studies with experimental, clinical and imaging data using systems biology approaches [235].

There are many articles discussing how systems biology and chemometrics can be used for biomarker discovery [236–239]. For example, Feala et al. integrated protein biomarker candidates for traumatic brain injury with publicly available canonical pathways and human protein-protein interaction networks to illustrate how to systematically generate new, testable hypotheses and identify candidate biomarkers [240]. Several tools are available that provide both pathway analysis and multi-omic integration including Open MS from KNIME [241], MetaboAnalyst [242] and XCMS online [243].

Longitudinal studies

One thing that became evident during the review work conducted was the lack of longitudinal studies. Only a small proportion of original-data studies evaluated HD biomarkers longitudinally. Most of the studies conducted so far are cross-sectional studies, which is a limitation if the aim is to examine the relationship between a change in a clinical measure and the change in a biomarker over time [244, 245]. A longitudinal design, in which the putative biomarker and clinical measure are both measured at least twice, will be useful when developing a disease progression marker. In addition, the duration needs to be long enough to observe a clinically significant change in the criterion used to draw associations with the putative biomarker. For a biomarker to be disease specific, it does not necessarily have to be different from its baseline levels in matched controls. However, levels of a disease specific marker would change with disease progression or in response to specific disease-modulating therapy while remaining constant in the controls during the same time period. Future longitudinal studies that assess the biomarker and clinical measures at several time points over a sufficient follow-up period are warranted to provide sufficient evidence of a biomarker’s potential validity [244].

It is encouraging to note that lack of longitudinal studies has already been realised by HD researchers; two large prospective, longitudinal, observational studies PREDICT-HD [246] and TRACK-HD [205] aiming to identify early HD biomarkers should address this issue.

Importance of publishing negative data (Table 9)

It is important that the biomarker research community publish all data, significant or not. Publication bias has been systematically investigated, particularly in clinical trials [247–249]. It has been demonstrated that studies reporting positive, or statistically significant, results are more likely to be published, and have higher odds of being fully reported [250]. Conversely, negative results are more likely than positive results to be published in journals with lower impact factors [251].

In this review, we have included a table for analytes unlikely to be useful markers for HD, as these markers have been assessed more than once, showing no difference in between HD and control subjects (Table 9). Of limited news value, but important knowledge to the HD biomarker field. Although these analytes do not differ between HD and control cohorts, some have been shown to correlate with specific disease measures as shown in Table 9.

Biobanks

Biomarker studies require sufficient numbers of properly annotated, well-characterised biospecimens. Good-quality biobanks that collect, store and distribute biospecimens under stringent quality control and assurance measures will be valuable for biomarker research [252]. Development of novel biomarkers is always a long process and optimal use of available resources is required. International collaboration is essential for standardisation of all aspects of biomarker studies, including biobanking procedures and study design, and to enable construction of sufficiently powered cohorts and optimally replicate findings. Lessons can be learned from other research fields. For example, the BioMS network has taken steps to standardise all aspects of biomarker research and validation to optimise biomarkers for multiple sclerosis [253]; the resulting guidelines can be utilised to benefit biomarker studies in HD [215, 254–262].

Biomarker qualification

Biomarker qualification is a mechanism that integrates the use of biomarkers into drug development programs to improve the efficiency and safety of clinical trials testing novel therapeutics [263–265]. The Food and Drug Administration (FDA) and The European Medicines Agency (EMA) have developed regulatory pathways for biomarker integration in clinical trials: the “Biomarker Qualification Program” [120] and “Qualification of Novel Methodologies for Drug Development” [266], respectively. If a biomarker lacks sufficient data to achieve full qualification, the FDA and EMA can publish a Letter of Support (LOS) [267], which provides support that the biomarker has demonstrated promise based on the level of evidence that has been formally provided to regulators.

At present, there are no biofluid biomarkers that are validated as outcome parameters for HD. Biomarker development and subsequent integration into drug development is critical to accelerating effective treatments for HD. Lessons on biomarker qualification can be learned from AD research [263] where new diagnostic criteria have been proposed that integrate pathophysiological biomarkers (imaging and/or biofluid) into all phases of the diagnostic approach to improve on the diagnostic specificity [268–272].