Article Text


Tracking and predicting disease progression in progressive supranuclear palsy: CSF and blood biomarkers
  1. Edwin Jabbari1,
  2. Henrik Zetterberg1,2,
  3. Huw R Morris1
  1. 1 Department of Clinical Neuroscience, UCL Institute of Neurology, London, UK
  2. 2 Institute of Neuroscience and Physiology, Goteborgs Universitet, Gothenburg, Sweden
  1. Correspondence to Prof Huw R Morris, Department of Clinical Neuroscience, UCL Institute of Neurology, Queen Square, London WC1N 3BG, UK; h.morris{at}


Progressive supranuclear palsy (PSP) is a rare and progressive neurodegenerative condition characterised pathologically by neuronal cell loss due to abnormal tau deposits. Clinically, the condition manifests as parkinsonism with the addition of progressive balance, speech, swallowing, eye movement and cognitive impairment, ultimately leading to death. Measuring change over time in neurodegenerative conditions is central to defining the effects of therapeutic intervention and disease biology. The current gold standard for measuring clinical disease progression in PSP is the PSP Rating Scale score. However, such scales may be affected by intrarater and inter-rater variability. In addition, their use in clinical trials may be hindered by differences in the time interval between pathological disease progression/response to therapeutics and change in clinical state. Therefore, the need for reliable disease progression biomarkers to complement clinical rating scales is clear. Here we discuss the benefits of using biomarkers to predict and track disease progression in both clinical and research settings. Through reviewing the literature to date on the role of cerebrospinal fluid (CSF) and blood biomarkers, we highlight data that reveals the ability of CSF and plasma neurofilament light chain (NF-L) to predict and track clinical disease progression in PSP. We also discuss the need for large-scale longitudinal studies to identify novel biomarkers.

  • Biomarkers
  • Progressive Supranuclear Palsy
  • Disease Progression
View Full Text

Statistics from


Progressive supranuclear palsy (PSP) is a progressive neurodegenerative condition and the most common cause of atypical parkinsonism, with an estimated prevalence of 5–7 per 1 00 000.1 The predominant clinical phenotype of PSP is Richardson’s syndrome, which presents with an akinetic-rigid syndrome, falls, a vertical supranuclear gaze palsy and cognitive impairment with executive dysfunction. In recent years, the clinical heterogeneity of PSP has been highlighted by the description of two rarer clinical phenotypes in pathologically confirmed cases of PSP: PSP-Parkinsonism, characterised by asymmetrical onset, tremor and a moderate initial therapeutic response to levodopa2; and pure akinesia with gait freezing, characterised by gradual onset of freezing of gait or speech, absent limb rigidity and tremor, no sustained response to levodopa and no dementia or ophthalmoplegia in the first 5 years of disease.3

Diagnosing PSP, particularly in the early stages of disease, may be challenging as there is a clinical overlap with Parkinson’s disease (PD), other atypical parkinsonian conditions such as multiple system atrophy (MSA) and other tauopathies including frontotemporal dementia (FTD) and corticobasal degeneration (CBD). Furthermore, most patients with PSP present with falls or non-specific visual symptoms, both of which are common in elderly subjects.4 A clinicopathological study from 2002, which included 20 cases of PSP, found that the positive predictive value of a clinical diagnosis of PSP was 80%, in contrast to the positive predictive value of a clinical diagnosis of PD which was 98.6%.5 However, a recent meta-analysis of 11 clinicopathological studies showed that the pooled diagnostic accuracy of PD was 80.6%,6 therefore similar to PSP. Furthermore, clinical criteria for the diagnosis of PSP according to the National Institute for Neurological Disorders and Society for PSP7 have low diagnostic sensitivity, such that many patients with pathologically proven PSP will not have been identified according to clinical diagnostic criteria. New diagnostic criteria for PSP are expected to be published by the Movement Disorder Society in due course. Pathology at postmortem remains the gold standard for diagnosing PSP.

Tau in PSP

The pathology of PSP is centred on the structural microtubule associated protein tau, encoded by MAPT. In PSP there is neuronal and glial accumulation of hyperphosphorylated tau aggregates, with neurodegeneration.8 There are several possible links between tau pathology and neurodegeneration including the formation of toxic oligomers, disruption of normal microtubule function9 and cell to cell spread of pathogenic tau.10

Autosomal dominant mutations in MAPT leads to FTDP-1711 and common variation in MAPT predisposes to PSP.12

Tau pathology in PSP also involves an alteration in tau isoform homoeostasis. Tauopathies can be classified according to the predominant isoform of tau that accumulates through alternative splicing of MAPT, leading to tau protein with 3 (3R) or 4 (4R) repeats of ~32 amino acids in the carboxy-terminus microtubule binding domain. 4R-tau is the dominant tau isoform in both PSP and CBD, in contrast to mixed 3R/4R-tau in Alzheimer’s disease (AD).13 Tau hyperphosphorylation and microtubule dysfunction have been targeted in two recent clinical trials14 15 although both trials showed no effect on disease modification in PSP.

CSF biomarkers used in the diagnosis of PSP

Many cross-sectional CSF studies have assessed whether CSF protein biomarkers can reliably differentiate PSP from healthy controls and patients with relevant differential diagnoses (see supplementary table for mean biomarker concentrations in PSP and control groups, and p values for differences between groups).

Tau and modified tau

Based on the pathology, one would expect that increased levels of CSF tau could be detected in PSP. The majority of CSF total tau (t-tau) and phosphorylated tau (p-tau) data reported in the literature are based on commercially available ELISAs, such as the INNOTEST assay (Fujirebio). In these assays, t-tau and p-tau measurements are dependent on antitau capture antibodies, such as AT120 and AT270 (pT181), specific for the mid-domain region of the protein, encoded by exons 4–8 (see figure 1).

Figure 1

Modified figure from Wagshal and colleagues24 highlighting tau antibody binding sites and differences in capture antibodies between standard (INNOTEST) and novel tau ELISAs.

However, in established PSP disease there has been no observed consistent elevation of CSF t-tau or p-tau compared with healthy controls.16 17 In addition, although Borroni and colleagues have shown lower levels of the CSF 33 kDa/55 kDa tau fragment ratio compared with healthy controls and other neurodegenerative conditions including FTD and AD,18 19 this has subsequently been attributed to assay artefact.20 Regarding tau isoforms, Luk and colleagues showed decreased levels of CSF 4R-tau in PSP and AD groups compared with healthy controls,21 although further replication data is required.

In contrast, raised CSF tau levels form a robust part of the diagnostic criteria for AD.22 However, it is uncertain as to why there is a difference between the CSF tau profile in PSP and AD considering that both conditions involve extensive neurofibrillary tangle pathology related to hyperphosphorylated tau protein. A potential explanation for the divergent tau profiles described above is that standard CSF tau ELISA might not detect elevations of specific tau species that are elevated in primary tauopathies such as PSP. In addition, the tau fragment profile is different in AD and PSP, which will determine epitopes and protein levels identified by ELISA. Importantly, most of tau in CSF appears to be in fragments containing N-terminal and/or mid-domain epitopes; virtually no full-length tau containing C-terminal sequences can be detected.23 Whether the concentrations of particular fragments differ between different tauopathies is currently unknown. To investigate this possibility, Wagshal and colleagues used novel ELISAs24 (see figure 1) targeting N-terminal and mid-domain epitopes to compare CSF tau concentrations in patients with severity-matched AD and PSP. This revealed that the PSP group had lower mean CSF N-terminal and mid-domain tau concentrations compared with the AD and control groups. In addition, receiver operating characteristics analysis revealed that the N-terminal fragment tau12-BT2 assay was best at differentiating PSP from AD (area under the curve=0.948, p<0.001). It is important to note that these novel ELISAs do not distinguish between 3R-tau and 4R-tau isoforms. It is hypothesised that the tau isoform profile may lead to an alteration in tau detectable in CSF in PSP and AD as 4R-tau is released less readily from cells compared with 3R-tau.25

Non-tau related CSF biomarkers

Non-tau related CSF biomarker candidates have also been studied extensively. The most promising of these have been neurofilament light and heavy chain (NF-L and NF-H). Similarly to tau, neurofilaments are found in axons of neurons and are important components of the cytoskeleton. However, it appears that neurofilaments are highly expressed in large-calibre myelinated axons in white matter, while tau is predominantly expressed in thin unmyelinated axons of the cortex.26 Significantly higher CSF levels of both NF-L17 ,27 and NF-H28 have been shown in patients with atypical parkinsonian conditions, including PSP, compared with controls and patients with PD. However, there were no significant differences between the individual atypical parkinsonian conditions. These results are thought to reflect the more extensive and rapid neuronal loss that occurs in atypical parkinsonian conditions compared with PD. The level of NF-L has also been shown to correlate with disease severity in PD, PSP,17 vascular dementia29 and FTD.30

Other CSF proteins of interest are soluble amyloid precursor protein α (sAPPα) and β (sAPPβ), and chitinase-3-like protein 1 (YKL-40). sAPP is bound to the mitochondrial outer membrane and therefore may be implicated in mitochondrial dysfunction, which contributes to the development of neurodegenerative conditions. Similarly, YKL-40 has been implicated in neuroinflammation and neurodegeneration and is considered a marker of glial activation. A large cross-sectional CSF study by Magdalinou and colleagues showed higher median levels of both NF-L and YKL-40, and lower median levels of both sAPPα and sAPPβ, in PSP compared with PD, AD, FTD and controls.31 The reason for such differences in the levels of sAPP and YKL-40 in PSP is unclear. Finally, there is limited evidence that CSF levels of glial fibrillar acidic protein (GFAP), a protein predominantly expressed in fibrillar astrocytes, are higher in patients with PSP compared with controls.16 Replication in larger cohorts is required to further explore the significance of these alternative CSF biomarkers. For now, it should be reiterated that there is a lack of reproducible data for a biomarker that can reliably differentiate PSP from other neurodegenerative conditions.

Blood biomarkers used in the diagnosis of PSP

In contrast to the large number of CSF based diagnostic studies discussed above, we found only four blood based biomarker candidates that have been analysed in diagnostic studies involving patients with PSP. Serum uric acid levels have been shown to be lower in patients with PSP compared with controls,32 but this was not confirmed in a similar cross-sectional study.33 Patients with MSA showed significantly higher levels of serum insulin-like growth factor-1 compared with controls in a study that also included patients with PD and PSP.34 However, one study did reveal significantly elevated levels of both methylmalonate and homocysteine in all patient groups (PD, PSP, Amyotrophic lateral sclerosis) compared with controls.35 The significance of this is unclear but it has been hypothesised that elevated levels of methylmalonate and homocysteine are neurotoxic.

Disease progression in PSP

Longitudinal studies assessing the natural history of PSP reveal that the clinical disease progresses over time. Measuring change in neurodegenerative conditions is central to defining the effects of therapeutic intervention and disease biology. In assessing the PSP Rating Scale (PSPRS) score, Golbe and colleagues followed up 162 patients with PSP over 11 years. They showed that the mean rate of progression was 11.3 points per year and the median actuarially corrected survival was 7.3 years from symptom onset to death. In addition, the PSPRS score was a good independent predictor of survival.36

However, such scales may be affected by intrarater and inter-rater variability. In addition, their use in clinical trials may be hindered by differences in the time interval between pathological disease progression/response to therapeutics and change in clinical state.

Therefore, the need for reliable disease progression biomarkers to complement clinical rating scales is clear. Predicting disease progression provides early prognostic information and may enable better powered clinical trials with more homogenous groups, and give insights into the mediators of disease progression. From a therapeutic trials perspective, measuring non-clinical biomarkers of disease progression may enable early demonstration of target engagement. However, no studies have shown that individual biomarkers are better than clinical scales at tracking disease progression and so, for now, clinical scales remain the only reliable option available. In addition, there are certain limitations encountered with CSF and blood as sources of biomarkers as lumbar puncture is an invasive procedure with potential complications, while protein concentrations in blood may not reflect pathological changes in the brain due to variations in blood-brain barrier permeability.

Here we review the reliability of CSF and blood biomarkers in predicting and tracking clinical disease progression in PSP. As well as summarising the literature to date, we also discuss the road ahead.


We performed a PubMed/Medline search and limited searches to studies reported in English, comparing PSP with other neurodegenerative conditions and/or healthy controls (see figure 2) up until February 2017. Studies predicting disease progression were defined as longitudinal, prospective studies with baseline blood and/or CSF analysis and clinical assessment followed by at least one subsequent clinical assessment, in which the baseline measures were correlated with subsequent progression. Studies tracking disease progression are defined as longitudinal, prospective studies with baseline blood and/or CSF analysis and clinical assessment followed by at least one subsequent blood and/or CSF analysis and clinical assessment, in which longitudinal measures were assessed for disease progression.

Figure 2

Flow diagram outlining the selection procedure to identify the five studies which were included in this review.

We used searches including the terms ‘progressive supranuclear palsy’, ‘PSP’, ‘parkinsonism’, ‘neurodegenerative diseases’, ‘atypical Parkinsonism’, ‘blood’, ‘plasma’, ‘serum’, ‘CSF’, ‘CSF biomarkers’, ‘biomarker’, ‘NFL’, ‘tau’, ‘tauopathies’, ‘prospective’, ‘longitudinal’, ‘consecutive’, ‘serial’, ‘predicting’, ‘tracking’, ‘progression’ and ‘disease progression’. Further references were found manually from identified publications.

All identified studies met a minimum quality score defined by the presence of: at least five patients with PSP included in the study, with clearly defined diagnostic criteria; a clearly defined primary outcome of the study; a minimum period of 10 months between the first and last clinical assessments, lumbar puncture and/or blood test; clearly defined components of the clinical assessments used in the study; clearly defined CSF sampling protocol and use of assays; appropriate recording of statistical analysis used to interpret results.


We identified five longitudinal studies in total. Four of the identified studies tracked clinical disease progression in patients with PSP using CSF and/or blood biomarkers, including one study that was derived from the davunetide clinical trial.14 One study predicted clinical disease progression using blood biomarkers. We did not identify any studies that used CSF biomarkers to predict progression.

Tracking disease progression in PSP


Most, but not all of the identified CSF studies (see table 1 for statistics) showed that the levels of CSF NF-L increased over time and correlated with clinical disease progression. The one exception was the study by Constantinescu and colleagues who showed that consecutive CSF analysis revealed relatively stable levels of NF-L and GFAP over time in all of the investigated parkinsonian groups.37

In contrast, Backstrom and colleagues found that CSF NF-L levels were stable in patients with PD but rose by almost 30% in patients with PSP over 1 year when compared with baseline CSF levels.38 As the majority of study participants consisted of patients with PD, follow-up clinical assessment and analysis of whether CSF biomarkers could predict clinical disease progression was limited to this group and revealed that the baseline triad of high NF-L, low beta amyloid 42 (Aβ42) and high heart fatty acid-binding protein (HFABP) indicated a high risk of developing Parkinson’s disease dementia. Similarly, a small subgroup of patients, who were part of a larger study by Magdalinou and colleagues, underwent serial CSF analysis and clinical examination to assess disease progression over the course of 1 year. Clinical examination consisted of measuring disease severity using the Hoehn and Yahr (H&Y) staging system and assessing clinical rating scores using the PSPRS and Mattis Dementia Rating Scale (DRS-2). At 1 year they found an increase in H&Y and PSPRS scores, and a decrease in DRS-2. These changes were associated with a mean increase in levels of NF-L by 540 ng/L.31

A major source of longitudinal data has come from recent therapeutic trials in PSP. A small subset of patients in the davunetide trial had serial CSF analysis, measuring levels of CSF Aβ42, t-tau, p-tau and NF-L. Although there were no significant differences in the rates of disease progression between the davunetide and placebo groups, NF-L was the only CSF biomarker that showed a statistically significant change over time, with a mean increase in concentration by 755 ng/L when davunetide and placebo patients were grouped together. Although the mean increase in CSF NF-L was lower in the davunetide group compared with the placebo group (494 ng/L vs 922 ng/L), this difference was not statistically significant (p=0.43). However, the role of CSF NF-L in tracking disease progression was once again highlighted by the fact that the 1-year change in CSF NF-L levels correlated with an increase in the oculomotor subscale of PSPRS.14

Table 1

CSF biomarkers used to track disease progression in PSP


Longitudinal fluid biomarker analysis in the davunetide study14 also included plasma NF-H which did not show any significant change in concentration over 1 year, with a median baseline concentration of 760 ng/L.

Predicting disease progression in PSP


We identified a recent longitudinal study of patients with PSP by Rojas and colleagues which assessed the ability of baseline plasma NF-L to predict a change in clinical measures using age, gender and baseline mini-mental state examination (MMSE)-adjusted mixed linear models. Similarly, age-controlled and gender-controlled Pearson’s partial correlations were determined between baseline plasma NF-L levels and changes in regional and whole brain volume. The study showed that high baseline plasma NF-L levels (see table 2 for statistics) predict more severe neurological, cognitive and functional decline at 1 year follow-up, as well as changes in whole brain and SCP volume loss.39 In addition, plasma and CSF NF-L were significantly correlated (r=0.74, p=0.002).

Table 2

Blood biomarkers used to predict disease progression in PSP


There is very limited evidence that CSF and blood biomarkers may have a role in predicting and/or tracking disease progression in PSP. Overall, the studies identified suggest that CSF NF-L tracks clinical disease progression over time, and that this correlates with the PSPRS score. A particularly interesting question, that none of the identified studies looked at, is whether or not there are differences in the change in NF-L concentration over time in early versus late stage disease. Other potential markers of tracking disease progression that had favourable results included CSF sAPPβ and GFAP.

Of interest, a recent study analysed NF-L changes in plasma, CSF and brain of a variety of mouse models of proteopathic (α-synuclein, tau and β-amyloid) neurodegenerative diseases. Plasma and CSF NF-L levels were strongly correlated, and NF-L increases coincided with the onset and progression of corresponding proteopathic lesions in the brain. In addition, experimental induction of α-synuclein lesions increased plasma and CSF NF-L levels, while blocking β-amyloid lesions attenuated the NF-L increase.40 The study therefore concluded that plasma and CSF NF-L may serve as both a marker of disease progression and also as a biomarker for treatment response in proteopathic neurodegenerative diseases.

With regards to predicting disease progression, Rojas and colleagues showed that high baseline plasma NF-L is able to predict more severe functional decline.39

However, when compared with the volume of cross-sectional diagnostic studies that have been carried out (a recent review by Magdalinou and colleagues on the diagnostic use of CSF biomarkers in parkinsonian conditions yielded 78 studies),41 there is a clear lack of longitudinal fluid biomarker data on the topic of disease progression in PSP. Of the five studies that were identified, most had small subject numbers, were limited to only 1 year of follow-up and exhibited heterogeneity in both the disease and control groups. In contrast, there appears to be a greater depth of studies looking at longitudinal brain changes in patients with PSP using a variety of conventional and non-conventional MRI modalities.42–45 Another major drawback is the lack of combined longitudinal CSF and blood studies—the davunetide study was the only longitudinal study identified to achieve this.

Other potential future directions on the topic in question include the use of new imaging modalities, including PET imaging with tau ligands46 and further studies on the role of blood based biomarkers. Of note, future studies may be able to achieve ultrasensitive measurement of fluid biomarkers by using novel technologies such as single molecule array,47 single molecular counting48 and proximity extension assay.49 Novel assays, such as the real time quaking-induced conversion assay, can be used to study misfolded protein biomarkers including tau and β-amyloid.50

This review has highlighted the need for longitudinal studies with large subject numbers that use combined (CSF, blood and imaging) biomarkers. Such studies will first of all aim to replicate the findings of existing longitudinal data on the role of biomarkers in tracking and predicting disease progression. There is also the potential to identify the first reliable PSP-specific biomarker, which may further improve our understanding of the underlying disease mechanisms. We hope that such biomarkers can be used in clinical trials of novel therapeutic targets, where the role of biomarkers as an objective measure of treatment response remains important. The role of genetics in predicting disease progression and its association with fluid biomarkers of interest is also important to consider. We will carry out these studies using patient data from the Progressive Supranuclear Palsy Corticobasal Syndrome Multiple System Atrophy Longitudinal UK (PROSPECT-M-UK) study, a multicentre longitudinal study that tracks patients over 5 years.

Supplementary Material

Supplementary Table 1


View Abstract


  • Contributors EJ performed the literature review and wrote this manuscript.

    HZ and HRM individually reviewed and edited the manuscript.

  • Funding EJ has been supported by the PSP Association. HZ has been supported by the Swedish and European Research Councils and the Wolfson Foundation. HRM has been supported by the PSP Association, Drake Foundation, Parkinson’s UK and the Medical Research Council.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.