Perhaps, the greatest medical challenge facing the developed world, and of increasing importance in the developing world, is the problem of late onset neurodegenerative disease. As life expectancies have increased, and smoking and cholesterol risk factors for heart disease and stroke have become understood and combatted, more and more of us are living to old age, where the scourge of neurodegenerative disease still remains. Most prevalent amongst these primary neurodegenerative diseases is Alzheimer's disease, but Lewy body dementia, frontotemporal lobar degeneration, motor neuron disease, progressive supranuclear palsy, prion disease and multiple system atrophy all afflict significant numbers of people. In all these cases, and in the ataxias and Huntington's disease, we have made enormous strides in understanding disease aetiologies – gene mutations have been found; cells have been transfected; transgenic mice have been made and, in some cases, these mice have developed ‘symptoms’ and pathologies and have had treatments tested upon them (Hardy and Gwinn-Hardy 1998). In no case, however, treatments based on these mechanistic understandings have yet yielded clinically useful treatments.

The reasons for this lack of useful clinical outcomes have been debated, especially with respect to Alzheimer's disease and motor neuron disease (see Golde et al. 2011 and Ludolph et al. 2007 for a discussion of some of the issues). However, undoubtedly, amongst the problems are those of uncertainty in diagnosis, slowness of progression (meaning that drug trials need to be very long and costly to determine and effect changes in the slope of decline), the likelihood of a long prodrome during which multiple irreversible cascades of harmful events could be initiated, and the possibility that the undoubted aetiologic heterogeneity within some diagnostic categories could lead to pathogenetic heterogeneity. For all the study of models of these diseases, we need a deeper understanding of the disease itself, and for this, we need accurate and informative biomarkers.

The most useful biomarkers in neurodegenerative disease to date have been imaging biomarkers, both amyloid imaging (Klunk et al. 2004) and MRI (Ridha et al. 2006) and CSF biomarkers (Villemagne et al. 2011). In all these cases, while cross-sectional data on cases and controls are valuable, of far greater value are longitudinal data (Bateman et al. 2012) – how does the biomarker in the patient change over time and is that change altered by intervention?

The use of CSF in clinical practice and in biomarker studies is surprisingly culturally dependent and subject to changes in fashion. In Sweden, in the 1980s when I was a postdoc, CSF was routinely taken from essentially all patients with neurodegenerative disease. In the United States, in the 1990s, taking CSF was regarded as a very serious issue and would never be taken as a routine. Of course, the more often a procedure is done, the safer it becomes. I would be happy to be someone's 100th CSF donor, less sanguine about being their first.

Among CSF biomarkers of neurodegeneration, tau and Aβ have been the most useful, with the combination of low Aβ levels during the clinical phase of disease and high tau levels being clinically useful in the diagnosis of Alzheimer's disease (Fagan et al. 2007). Tau more generally has a central role in neurodegenerative disease being implicated pathologically in Alzheimer's disease, PSP and CBD with the tau locus being genetically implicated in PSP, CBD and Parkinson's disease (Vandrovcova et al. 2010): increased CSF tau levels usually being associated more generally with the process of neurodegeneration.

As of the complex involvement of tau in many neurodegenerative diseases, clearly distinguishing different tau species in CSF could be critical and informative, both for diagnostic and prognostic reasons. In this respect, the report by Luk et al. (2012) marks a significant advance. Tau has six common isoforms in the human brain distinguished by alternate splicing of exons 2 and 3 together, 3 alone and 10 (Goedert and Spillantini 2011). In different diseases, different isoforms are deposited in different proportions (Goedert and Spillantini 2011). And in part, these differences are genetically determined by the now famous MAPT inversion haplotype (Baker et al. 1999; Stefansson et al. 2005; Fig. 1) and additionally by genetic variability in the promoter region of the microtubule-associated protein tau (MAPT) (Pittman et al. 2005). Of these differences, the difference between exon 10 in and exon 10 out (four repeat and three repeat) tau has been thought to be critical because four repeat tau is deposited selectively in PSP, CBD and in some forms of FTD (Goedert and Spillantini 2011). In this report, Luk and colleagues, who had previously developed antibodies selective for the two isoforms, which have been extensively used in histopathology and in ELISA's (de Silva et al. 2003; Luk et al. 2009), now adapt an immuno-PCR procedure to allow the selective detection of these isoforms in CSF down to levels at which they can be used diagnostically and then tested the processes in samples from patients with different diseases collected cross-sectionally. The antibodies clearly work very well in the assay with an impressive dynamic range, and though no significant differences between diseases are noted from this first rather small study, clearly the antibodies and the assay hold real promise in large sample series and in longitudinal sample series.


Figure 1. The H1/H2 haplotype polymorphism on chromosome 17. The H2 haplotype is, in Europeans, derived from a single founder (Fung et al. 2005), and is protective against Parkinson's disease probably because of increased inclusion of exon 3 (Trabzuni et al. 2012). A variant in the MAPT promoter found on some H1 haplotypes (rs242557) (Pittman et al. 2005; Höglinger et al. 2011) is additionally associated with both PSP and CBD and is associated with increased expression of tau, possibly selectively of four repeat isoforms (Myers et al. 2007; Vandrovcova et al. 2010).

Download figure to PowerPoint

What the article also illustrates very clearly is the difficulty of quality control in CSF analyses, which need to be overcome before data can really be generated and shared between laboratories. Variability between sampling sites and between labs is something which needs very careful control and standardisation before the CSF biomarker field can really move on (Mattsson et al. 2010).

With respect to tau in CSF, we have now really good and sensitive assays for four repeat and three repeat tau. Recent genetic data, especially in Parkinson's disease, show that we need to also understand exon 2 and 3 isoforms in genetically defined individuals (Trabzuni et al. 2012; see Fig. 1). We also have little idea what proportions of tau in CSF are there because of neuronal loss and lysis, synaptic damage and or active secretion. Clearly, given the now overwhelming evidence that tau pathology can spread through the CNS (Clavaguera et al. 2009; Liu et al. 2012; de Calignon et al. 2012; Hardy and Revesz 2012), and the importance of the extracellular compartment for tau protein becomes an extremely important area of study. The study by Luk and colleagues opens one window on to this compartment.


  1. Top of page
  2. Acknowledgements
  3. References

Work in my laboratory has been supported by the Peacock Foundation, the MRC and the Wellcome Trust. I have no conflicts of interest related to this work.


  1. Top of page
  2. Acknowledgements
  3. References