Measuring tau and other neuronal proteins in cerebrospinal fluid gives insight into the state of the brain during its long run-up to Alzheimer’s dementia. However, cross-sectional studies yield but a snapshot, which can be misleading. Researchers need the “movie version,” i.e., serial measurements that show change over time. Now, just such a study from Anne Fagan and colleagues at Washington University in St. Louis finds that while CSF tau climbs in early stages of sporadic AD—as everyone knows by now—levels actually decline once symptoms set in. What’s more, the same is true for three other presumptive biomarkers of neuronal or synaptic injury, including the calcium sensor VILIP-1 and synaptic proteins neurogranin and SNAP-25. The findings appeared March 24 in Alzheimer’s and Dementia online.

  • Cross-sectional studies don’t catch full picture for some biomarkers.
  • Largest study of its kind looked at within-person changes in multiple CSF biomarkers of neuronal injury in late-onset AD.
  • Phospho-tau, VILIP-1, SNAP-25, and neurogranin all increase up to symptom onset, then decline.

“We’ve needed these types of longitudinal studies with repeated sampling of individuals for a long time. This is the largest study with the broadest range of biomarkers to date, and it tells us some important things we didn’t know,” said Henrik Zetterberg, University of Gothenburg, Sweden.

The results emphasize the shortcomings of extrapolating biomarker curves from cross-sectional data, said Fagan. “We can’t be naïve about it anymore by saying you can estimate longitudinal change by looking at between-person levels of biomarkers in cross-sectional studies,” she told Alzforum.

In cross-sectional studies of CSF biomarkers, total tau (t-tau) and phospho-tau (p-tau) start to rise long before signs of dementia appear and continue to trend upward through to end-stage disease. However, in previous work, Fagan’s team analyzed longitudinal samples from a small number of cases of dominantly inherited AD and were surprised to find that p-tau levels in a given person actually dropped after that person had started developing symptoms (Mar 2014 news). The same study reported a similarly timed decrease in visinin-like protein-1, a calcium-sensing protein released from injured neurons. The WashU group had identified VILIP-1 as a potential CSF biomarker several years earlier (Mar 2012 news). 

In the new study, first author Courtney Sutphen followed up this initial finding in DIAN by running a similar analysis in a larger group of people with late-onset sporadic AD from the Alzheimer’s Disease Neuroimaging Initiative. Sutphen identified 148 ADNI participants who had between two and seven CSF samples available for analysis, taken in a one- to seven-year span. At baseline, 56 of them were classified as cognitively normal, 76 had mild cognitive impairment, and 16 AD. Based on a CSF Aβ42 threshold of 192 pg/ml or less, 21 of the cognitively normal group, 58 with MCI, and all of the AD group were deemed amyloid-positive.

Sutphen measured VILIP-1, neurogranin, and SNAP-25 by ELISA, using antibodies developed in collaboration with Jack Ladenson at WashU. Like VILIP-1, both CSF neurogranin and SNAP-25 concentrations rise in people with AD, and are thought to signal neuronal injury or death. Rising neurogranin is associated with brain atrophy and reduced glucose uptake (Sep 2015 newsJan 2015 news; Sep 2015 news). All the ELISAs showed good reproducibility, Sutphen said, with variances of 7 percent or less on repeat testing. She also analyzed Aβ42, t-tau and p-tau181 using Elecsys automated assays, which have been developed for those markers and are more reproducible than ELISAs.

At baseline, levels of all five injury markers were higher in the amyloid-positive MCI and AD groups than in cognitively normal people. Average concentrations of t-tau and p-tau in the AD group were twice those in cognitively normal, amyloid-negative controls, with the other three markers showing slightly less of an increase.

Spaghetti Time. Serial measures of CSF neurogranin show no change in cognitively normal people (left), or most people with MCI (middle), but a decline in AD (right). [Courtesy of Sutphen et al., Alzheimer’s & Dementia.]

When Sutphen looked at within-person changes, t-tau and p-tau showed consistent increases in the amyloid-positive CN and MCI participants, while p-tau fell significantly in those with AD. Their average change, a loss of 1.65 pg/ml/year, added up to a 3.9 percent yearly decrease in p-tau. Sutphen saw similar reductions in the other neuronal damage markers. After baseline, people with AD averaged 3.4, 2.5, and 6.9 percent annual declines in VILIP-1, SNAP-25, and neurogranin levels, respectively.

None of the novel markers showed a significant movement up or down over time in the CN or MCI groups.

The inflammatory marker YKL-40 told a different story. It appeared to go up with age, but was highly variable both at baseline and over time. Its fluctuations appeared unrelated to amyloid status. Fagan speculated that YKL-40 reflects inflammation that is not specific to AD.

None of these biomarkers tracked with cognitive decline, hippocampal shrinkage, or cortical thinning, all of which accelerated as disease worsened.

Other investigators have noted declines in p-tau in late-onset AD patients (Aug 2017 conference news; Toledo et al., 2013; Seppälä et al., 2011). However, no one knows why tau, and now three other markers of neuronal injury, decline as disease advances. Fagan rules out the possibility that the proteins were simply diluted in a larger volume of CSF that results from brain shrinkage and ventricle enlargement with advancing disease. She said she still saw reductions after controlling for ventricular volume. 

Alternatively, the drop in CSF concentrations could signal the slowing or ending of an acute neuronal die-off. Zetterberg said he’d like to see measures of neurofilament in these same samples, because as a generic marker of cell death it might shed light on how CSF tau and the other markers relate to neuronal loss.

The results highlight how little is known about the origins of CSF tau and other markers, and they left Fagan wondering what markers scientist might look for to track response to therapies. “It’s logical to assume that if you slow down neuronal injury, that would be accompanied by a reduction in tau,” Fagan said. “But if tau is already declining, would a positive therapeutic outcome mean the decrease would become less steep, or maybe tau would rise, like it did earlier in the disease process?”

Zetterberg sees a practical take-home from the study. “In clinical trials, the active arm and the placebo group will have to be carefully matched by disease stage. Minor differences in how far patients have deteriorated in the neurodegenerative process might translate into major differences in biomarker trajectories, and that could be misinterpreted as a treatment effect,” he warned.

One caveat of the study was the short follow-up times available, which averaged from 2.5 to four years. To get a fuller picture, Fagan plans to measure these biomarkers as people progress from asymptomatic through early symptomatic or MCI to end-stage AD, a process that takes 20 years or more. “That’s the only way to find out if we really do have elevations early on that then come down,” Fagan said. That’s one goal of WashU’s Adult Children Study, where CSF draws start at age 45.

Longitudinal studies might be easier if markers could be measured in blood. That’s the way forward, wrote Harald Hampel of the Sorbonne University in Paris in a comment (below). His group reported a longitudinal decline in CSF p-tau131 in patients with AD (Hampel et al., 2001). For future studies, a shift to blood-based biomarker panels seems the rational course, he wrote.—Pat McCaffrey


  1. The interesting aspect of this study is that the longitudinal decrease in biomarkers observed in the familial DIAN cohort is now replicated in the ADNI cohort of LOAD patients.

    Such longitudinal data are highly needed for clinical trials, and the observed variation in clinical, MRI, and biomarker outcome measures (e.g., high in YKL-40) and decreases indeed are very relevant when evaluating treatment effects. However, a trial will normally include placebo, so any effect should be different in the treatment group compared to controls. But now we are more aware that biomarker levels do not only increase but can also decrease as part of the natural course.

    There is a gap to fill next, because in ADNI participants are relatively old and in DIAN they are quite young. There are clear age relationships of many biomarkers, which increase due to normal aging. So, it would be interesting if similar findings can be obtained in early onset, sporadic AD patients.

    It is interesting that levels of tau, but not the other markers, increase over time. This may suggest that some different biology is involved. Alternatively, the stronger dynamics of tau might be due to having assays of higher quality (lower variation) or to the protein being under stronger influence of altered CSF flow than the other biomarkers.

    Assuming that the different biomarker changes are due to biology, this data suggests that axonal damage does not precede tissue destruction but that markers of axonal damage are released due to tissue injury. Moreover, there is a need for earlier biomarkers that change as early as does Aβ42.

    The five-year change for the majority of markers is clear, but given the slope of the spaghetti plots, inter-individual variation will be high. So, the ideal biomarker that shows a change in every single patient is not yet discovered.

    Moreover, the cause of change could indeed be a loss of tissue mass or ventricular enlargement, i.e., a general physiological parameter, rather than pathology-specific alteration. The true dynamic change downstream of neuroprotection has yet to be proven in positive trials.

  2. Longitudinal biomarker studies are very important to better understand the molecular changes in AD. However, considering the protracted disease course of AD (spanning two to three decades), we probably need intra-individual measurements over at least eight to 15 years to be able to make firm conclusions. There have been a few such studies (e.g. Stomrud et al., 2015), but they have been quite small.

    I fully agree with the authors that the current study warrants the need of further long-term studies on CSF biomarkers that might reflect ongoing neurodegeneration (total tau, Ng, SNAP25) before such biomarkers are used as outcome measures in trials evaluating the efficacy of new treatments. We need to better understand which factors are associated with increases or decreases in the levels of these biomarkers over time, including the effects of changes in neuronal activity, neuronal degeneration, as well as changes in the dynamics of the flow of ISF and CSF during different stages of the disease. That said, the combinations of the CSF Aβ42/40 ratio together with t-tau or phospho-tau have been very thoroughly validated for their ability to accurately identify cases with preclinical or prodromal AD, and these biomarkers can be used when selecting individuals for treatment trials.


    . Longitudinal cerebrospinal fluid biomarker measurements in preclinical sporadic Alzheimer's disease: A prospective 9-year study. Alzheimers Dement (Amst). 2015 Dec;1(4):403-11. Epub 2015 Oct 9 PubMed.

  3. To investigate longitudinal dynamics of pathophysiological biomarkers across all AD stages is of paramount importance for the understanding of disease initiation, progression, and biomarker-guided, targeted drug development. Ideally, we should obtain comparative paired CSF and plasma/blood dynamics over time in a large-scale observational study of healthy elderly individuals until a proportion of them reaches clinical endpoints, such as mild cognitive impairment or AD dementia. This is currently been done by the Alzheimer Precision Medicine Initiative (APMI) through the APMI cohort program (APMI-CP) using the large-scale observational mono-center INSIGHT-preAD cohort in Paris (Hampel et al., 2016; Hampel et al., 2017; Hampel et al., 2018). In addition to longitudinal (more than seven years) fluid biomarker analysis, the APMI-CP includes complementary, comprehensive modalities, such as structural, functional, and metabolic imaging and neurodynamics (including EEG and ERP). Previously, we applied up to seven serial CSF measurements in patients with AD demonstrating that p-tau231 concentrations decreased linearly with time (Hampel et al., 2001). 

    Several relevant pathophysiological processes, i.e., synaptic dysfunction and loss of plasticity, neuroinflammation, and mitochondrial dysfunction, have been reported to underlie polygenic AD and should be investigated longitudinally using appropriate candidate fluid biomarkers. AD transgenic animal studies have previously shown a complex interplay among synaptic pathways, glial activity, and regulatory mechanisms of both Aβ and tau metabolism (Bachhuber et al., 2015). It is well established that these pathophysiological processes occur along the continuum of AD, beginning from asymptomatic phases, and directly contribute to neurodegeneration and successive cognitive decline.

    Fagan and colleagues have demonstrated in this study that synaptic dysfunction, neuroinflammation, and neuronal injury are tightly linked to AD pathophysiological “hallmarks”—i.e., Aβ and tau, both misfolding and widespread toxic accumulation. This is an important step toward exploring pathophysiological dynamics informative for clinical drug development programs.

    It is reported that the global and regional amount of brain amyloid plaques correlates well with neither the extent of neurodegeneration nor the degree of cognitive decline (Karran and Hardy, 2014). However, Aβ biomarkers do not seem to cover the entire temporal spectrum of polygenic AD pathophysiology.

    To date, a holistic and comprehensive understanding of the pathophysiological dynamics across the spectrum of individuals affected by polygenic AD is far from being reached, however, this study carried out by Fagan and colleagues provides relevant, novel incremental insights into the longitudinal CSF dynamics of the disease.

    From a pharmacological point of view, research should integrate and focus to better understand whether and how aberrant synaptic and inflammatory pathways synergistically interact with misfolding and accumulation of Aβ and tau. To move forward, a shift to, and integration of, evolving and available blood-based candidate biomarker panels seems rational (O’Bryant et al., 2017)

    The findings by Fagan and colleagues support the perspective of a biomarker-guided, multi-targeted therapeutic approach to AD under the Precision Medicine paradigm rather than “magic-bullet, one-size-fits-all” drug targeting of one pathophysiological process at unspecified time points along yet undefined AD stages.


    . PRECISION MEDICINE - The Golden Gate for Detection, Treatment and Prevention of Alzheimer's Disease. J Prev Alzheimers Dis. 2016 Dec;3(4):243-259. Epub 2016 Sep 6 PubMed.

    . A Precision Medicine Initiative for Alzheimer's disease: the road ahead to biomarker-guided integrative disease modeling. Climacteric. 2017 Apr;20(2):107-118. Epub 2017 Feb 9 PubMed.

    . Revolution of Alzheimer Precision Neurology. Passageway of Systems Biology and Neurophysiology. J Alzheimers Dis. 2018;64(s1):S47-S105. PubMed.

    . Tracking of Alzheimer's disease progression with cerebrospinal fluid tau protein phosphorylated at threonine 231. Ann Neurol. 2001 Apr;49(4):545-6. PubMed.

    . Inhibition of amyloid-β plaque formation by α-synuclein. Nat Med. 2015 Jul;21(7):802-7. Epub 2015 Jun 22 PubMed.

    . A critique of the drug discovery and phase 3 clinical programs targeting the amyloid hypothesis for Alzheimer disease. Ann Neurol. 2014 Aug;76(2):185-205. Epub 2014 Jul 2 PubMed.

    . Blood-based biomarkers in Alzheimer disease: Current state of the science and a novel collaborative paradigm for advancing from discovery to clinic. Alzheimers Dement. 2017 Jan;13(1):45-58. Epub 2016 Nov 18 PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. DIAN Longitudinal Data Surprises With Late Drop in Tau
  2. Research Brief: VILIP-1 a Potential CSF Marker for AD?
  3. Cerebrospinal Fluid Neurogranin Correlates with Markers of Neurodegeneration
  4. New Biomarkers? Synaptic Proteins in Spinal Fluid Predict Cognitive Decline
  5. Paper Alert: Longitudinal Data Support CSF Neurogranin as an Early Synaptic Marker of Alzheimer’s Disease
  6. Longitudinal Data Say: Nope, CSF Markers Do Not Track Progression

Paper Citations

  1. . Longitudinal change in CSF Tau and Aβ biomarkers for up to 48 months in ADNI. Acta Neuropathol. 2013 Jun 29; PubMed.
  2. . Longitudinal changes of CSF biomarkers in Alzheimer's disease. J Alzheimers Dis. 2011;25(4):583-94. PubMed.
  3. . Tracking of Alzheimer's disease progression with cerebrospinal fluid tau protein phosphorylated at threonine 231. Ann Neurol. 2001 Apr;49(4):545-6. PubMed.

External Citations

  1. Adult Children Study

Further Reading

No Available Further Reading

Primary Papers

  1. . Longitudinal decreases in multiple cerebrospinal fluid biomarkers of neuronal injury in symptomatic late onset Alzheimer's disease. Alzheimers Dement. 2018 Jul;14(7):869-879. Epub 2018 Mar 23 PubMed.