Part 1 of 2. Click here for Part 2.

When axons in the brain’s white matter start to fall apart—be it from insult or injury—they release the protein neurofilament light chain. Thus dumped, NfL finds its way into the cerebrospinal fluid and into the blood. Rising CSF NfL signals neuron loss in acute and chronic conditions, including stroke, traumatic brain injury, Alzheimer’s, multiple sclerosis, and others. Now, the advent of a supersensitive assay for blood NfL has led to an explosion in studies evaluating the protein as a biomarker for neurodegeneration. New data spilling out at the Alzheimer's Association International Conference, held July 20–26 in Chicago, offer a glimpse at the progress. The data is so convergent that consensus is already forming that blood NfL can help predict a person’s disease course. Besides helping to stage participants in clinical trials, NfL might serve both as a dynamic marker of neuronal injury and as a tool to measure treatment effects in trials.

  • When axons fall apart, NfL gets released, leaks into blood.
  • Sensitive assays for blood NfL show its rise with age, AD.
  • A longitudinal study correlated the rate of NfL increase with changes in imaging and cognition, predicted disease progression.

Scientists already knew that NfL was consistently higher in the CSF of people with AD than controls (see AlzBiomarker meta-analysis). It tracks with prognosis, whereby more elevated concentrations presage faster disease progression (Nov 2015 news). 

Now, scientists are realizing that as goes CSF, so goes blood. And that opens new doors. For years, reliably measuring NfL in blood, where concentrations run 50 times lower than CSF, was impossible. Then Henrik Zetterberg, University of Gothenburg, Sweden, developed an ultrasensitive NfL immunoassay on the Quanterix Simoa platform (Apr 2016 conference newsKuhle et al., 2016). His and other groups’ work (Bacioglu et al., 2016) established that blood NfL concentrations mirror those in CSF. In a cross-sectional study looking at plasma NfL in 570 ADNI participants, blood NfL, just like CSF, appeared to faithfully report disease severity and prognosis (Mar 2017 news). 

At AAIC, a large, longitudinal study addressed the next question, that is, how blood NfL changes over time. Stephanie Schultz, Washington University, St. Louis, presented data on serum NfL in more than 400 people in the Dominantly Inherited Alzheimer’s Network. DIAN offers a unique opportunity for biomarker investigation, because mutation carriers are destined to get AD, and researchers know approximately when they will develop symptoms, based on their specific genetic lesions and family history. Mathias Jucker’s group at the German Center for Neurodegenerative Disease in Tubingen, Germany, measured NfL using the Simoa assay, and Schultz and her colleagues did the statistical analysis in the cohort, who have been followed for years with comprehensive phenotyping, including repeated MRI and PET imaging, and CSF and blood sampling. All had given more than one blood sample; half had given more than two, collected every one or two years. The mostly short stretches of within-person longitudinal follow-up added up to span 30 years of Alzheimer’s progression from presymptomatic to advanced dementia.

First, Schultz and colleagues ran a cross-sectional analysis. It substantiated earlier findings from a much smaller ADAD cohort in the U.K. (Nov 2017 news). In both cohorts, mutation carriers had on average higher CSF and blood NfL concentrations than noncarriers, regardless of their cognitive status. This ongoing ADAD research is also measuring serum NfL as a function of time to symptom onset. Schultz and colleagues are finding that the difference emerges five to 10 years before the estimated age of onset (EYO), the same as when NfL was measured in CSF, Jucker wrote to Alzforum.

For most biomarkers being studied, scientists are realizing that longitudinal measures of a given marker’s rate of change are more sensitive than just the baseline concentration at predicting cognitive decline (e.g., Dec 2017 conference news). Schultz and colleagues found the same for NfL. In the longitudinal part of the study, the rate of change of NfL held steady over time in noncarriers, but appeared to quicken in presymptomatic carriers approximately 16 years before onset of symptoms, or about 10 years earlier than estimates from the cross-sectional analysis. The rate of change appeared to peak at the transition from normal to dementia, then plateaued.

When compared with imaging biomarkers, the rate of change of serum NfL was associated most strongly with cortical thinning, and less so with declines in FDG-PET or Aβ accumulation. This suggests the increase in blood NfL reflects ongoing neurodegeneration. Consistent with that, the rate of NfL change also predicted cognitive decline—in a subset of 39 people, speedier NfL accumulation tracked with greater decline in MMSE and logical memory in the following one to two years.

Blood NfL increased fastest in those transitioning from cognitively normal to symptomatic, and this rate was more sensitive to predicting a person’s progression than their absolute NfL levels. These findings highlight the value of longitudinal data, Schultz said. The next logical step, she said, will be to relate NfL to measures of white matter and to synapse integrity.

“Our work suggests that NfL in blood is going to be useful as a strong and clear signal of neurodegeneration,” said co-author Brian Gordon of Washington University. DIAN is adding plasma NfL as a core biomarker, so it will be a standard measure moving forward, he added.

At AAIC, Philip Weston, University College London, showed longitudinal data on his center’s ADAD cohort. It agreed closely with Schultz’s findings. Among the 15 mutation noncarriers and 47 carriers Weston follows, 41 thus far have given blood repeatedly, for a total of 118 time points. As published before, baseline NfL was higher in symptomatic people, 22.6 pg/ml in the AD group versus 12.6 in asymptomatic carriers and 10.34 in noncarriers. More strikingly, though, Weston detected a dramatic escalation of the annual rate of change of NfL from 0.51 and 0.67 pg/ml/year in noncarriers and asymptomatic carriers, respectively, to 3.52 pg/ml/year in the five symptomatic people who had given multiple samples.

When he modeled the longitudinal changes versus time of onset, blood NfL concentration appeared to diverge between carriers and noncarriers around a decade prior to onset of clinical symptoms, and rose progressively after. Raquel Sanchez-Valle, Barcelona University Hospital, presented additional data in her Spanish ADAD cohort that supports the idea that serum NfL is a good proxy for tracking disease onset and progression in this group.

Weston had too little data to say what happened in the years after onset, but he did notice a difference with type of mutation: Presenilin mutation carriers had 34.5 percent higher NfL than APP mutation carriers. Weston emphasized that in this small sample, he sees great inter-individual variability in the rate of change, which may limit NfL’s usefulness in tracking individuals. Other factors can affect NfL; Its concentration creeps up with age, with vascular disease, and with neurodegenerative processes besides AD. “That’s something we need to think about more,” he said.

Like ADAD, like LOAD?
How these results in ADAD will apply to sporadic AD remains unclear. “I don't know of any studies that directly compared NfL in sporadic and familial AD. This is extremely important in knowing the wider applicability of these markers,” Weston said. The ADNI NfL study reported no difference in blood NfL between asymptomatic amyloid-positive and -negative people, but it was not clear how close those people might be to developing symptoms. “Perhaps they were too early in the disease process,” Weston said. Gordon told Alzforum he hopes to measure plasma NfL in WashU’s sporadic AD cohort.

Scientists at the Mayo Clinic in Rochester, Minnesota, are starting to widen the lens by analyzing NfL in the Mayo Clinic Study of Aging, a large community–based sample of people not selected for family history or risk of AD. At AAIC, Silke Kern of the Mayo and the University of Gothenburg, Sweden, reported MCSA NfL data from CSF, not blood. Kern showed longitudinal results in 648 participants. Their baseline NfL levels turned out to predict progression from normal to MCI better than their concentrations of CSF Aβ42, p-tau, T-tau, or the synaptic marker neurogranin.

Did NfL behave the same in blood? Michelle Mielke, also of the Mayo Clinic, reported on NfL concentrations in plasma from 79 people (64 cognitively normal, 15 with MCI) from the same study. She sent blood, collected at baseline and then again 15 or 30 months later, to be analyzed by Simoa in Kaj Blennow and Zetterberg’s group in Gothenburg. Baseline NfL did not correlate with baseline hippocampal volume, cortical thickness, amyloid PET, FDG-PET, or cognitive measures. However, higher baseline NfL did match with greater decline over time in the structural measures, FDG-PET and global cognition. A faster increase in NfL over the follow-up period signaled a steeper decline in global cognition. The effects were independent of amyloid levels.

The results mean plasma NfL could work as a short-term prognostic marker of ongoing neurodegeneration. The Mayo data, too, suggest that changes in plasma NfL parallel changes in cognition; however, Mielke needs larger samples and longer follow-up to say for sure. Like Weston, Mielke stressed the need to define what else affects plasma NfL in the general population.

In the proposed A/T/N framework for biomarker-based diagnosis and staging of AD, positive biomarkers for cerebral amyloid (A) and neurofibrillary tangles (T) define AD, while the neurodegeneration (N) category encompasses nonspecific measures of neuronal dysfunction and death. As proposed, this N category currently contains FDG-PET, brain atrophy, and CSF total tau (Apr 2018 news). “CSF NfL is a risk factor for cognitive decline, whether people are Aβ-positive or not. It’s not specific for AD, but might be a better marker for neurodegeneration in the A/T/N scheme,” said Kern. The results had many at the meeting asking if blood NfL might offer an even better measure of N.

In toto, while the older CSF-based NfL measurements are more established, every lab will now be trying the blood markers, Gordon told Alzforum. A combination of blood Aβ, which is specific for AD, and NfL, which is not, could revolutionize AD clinical trials and the ability to monitor people over time, he said. “Because blood NfL is a nonspecific marker, it should have high utility for other diseases, too,” he said.—Pat McCaffrey


  1. It was very interesting to see the range of studies around immunodetection of neurofilament proteins in blood in the context of Alzheimer’s disease, made possible by sensitive new SIMOA techniques developed at the University of Gothenburg. While this is often interpreted as a relatively nonspecific change of Alzheimer’s disease, and indeed, neurofilament pathology occurs in several other neurological conditions (Vickers et al., 2009), there is an existing evidence base that indicates that this approach may be useful for specific staging of important pathological changes in Alzheimer’s disease. 

    For example, early studies by John Morrison, Michael Campbell, and Patrick Hof demonstrated that subsets of cortical pyramidal neurons particularly enriched in neurofilament proteins were susceptible to degeneration in Alzheimer’s disease (Morrison et al., 1987; Hof et al., 1990), and we have also shown that these specific neurons also form neurofibrillary tangles (Vickers et al., 1992). Furthermore, while β-amyloid deposition can occur in aging, it is the switch to increasing proportions of more fibrillar plaques that signals the clinical phase (Dickson et al., 1999). Such plaques are associated with damage to axons, involving substantial accumulation of neurofilaments in abnormal neurites (Vickers et al., 1996; Mitew et al., 2013). This axonopathy corresponds to the earliest substantial neuronal pathology of Alzheimer’s disease, prior to abnormal tau in dystrophic neurites, and is linked to subsequent transformation of the cytoskeleton that gives rise to neurofibrillary pathology (Vickers et al., 2016). Plaque-damaged axons with accumulated neurofilaments are also the most abundant neuronal change observed in multiple amyloid transgenic animal lines (Woodhouse et al., 2009; Mitew et al., 2013). 

    With respect to this pathological sequence, using new immunodetection technologies, it could be predicted that neurofilament protein levels would increase markedly during the stage in which fibrillar plaques develop in substantial numbers, which we might consider the most active phase of the disease, where the most substantial axonal pathology and neuronal cytoskeletal transformation occurs. Relatedly, once the majority of neurofilament-abundant neurons have degenerated, then peripheral neurofilament proteins may be expected to diminish.

    The great body of work already documented in the literature on neurofilament biomarkers, and in presentations at AAIC 2018, may open a window for detecting the most pathologically active phase of Alzheimer’s disease, and potentially monitoring the effectiveness of therapeutic strategies to protect axons and neuronal connections.


    . Neurochemical diversity of dystrophic neurites in the early and late stages of Alzheimer's disease. Exp Neurol. 1999 Mar;156(1):100-10. PubMed.

    . The morphological phenotype of beta-amyloid plaques and associated neuritic changes in Alzheimer's disease. Neuroscience. 2001;105(1):99-107. PubMed.

    . Quantitative analysis of a vulnerable subset of pyramidal neurons in Alzheimer's disease: I. Superior frontal and inferior temporal cortex. J Comp Neurol. 1990 Nov 1;301(1):44-54. PubMed.

    . Neurofibrillary pathology in transgenic mice overexpressing V717F beta-amyloid precursor protein. J Neuropathol Exp Neurol. 2001 Apr;60(4):357-68. PubMed.

    . Neurites containing the neurofilament-triplet proteins are selectively vulnerable to cytoskeletal pathology in Alzheimer's disease and transgenic mouse models. Front Neuroanat. 2013;7:30. PubMed.

    . A monoclonal antibody to non-phosphorylated neurofilament protein marks the vulnerable cortical neurons in Alzheimer's disease. Brain Res. 1987 Jul 28;416(2):331-6. PubMed.

    . Progressive transformation of the cytoskeleton associated with normal aging and Alzheimer's disease. Brain Res. 1992 Oct 30;594(2):273-8. PubMed.

    . Dystrophic neurite formation associated with age-related beta amyloid deposition in the neocortex: clues to the genesis of neurofibrillary pathology. Exp Neurol. 1996 Sep;141(1):1-11. PubMed.

    . Axonopathy and cytoskeletal disruption in degenerative diseases of the central nervous system. Brain Res Bull. 2009 Oct 28;80(4-5):217-23. PubMed.

    . Alterations in neurofilaments and the transformation of the cytoskeleton in axons may provide insight into the aberrant neuronal changes of Alzheimer's disease. Brain Res Bull. 2016 Sep;126(Pt 3):324-333. Epub 2016 Jul 27 PubMed.

    . Dystrophic neurites in TgCRND8 and Tg2576 mice mimic human pathological brain aging. Neurobiol Aging. 2009 Jun;30(6):864-74. Epub 2007 Oct 22 PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. For the Faint of Heart: Doc, Want to Check Their NfL?
  2. Neurofilament Light Chain: A Useful Marker for AD Progression?
  3. WANTED: Biomarkers for Drug Trials in Frontotemporal Dementia
  4. Blood Neurofilament Light a Promising Biomarker for Alzheimer’s?
  5. Serum NfL Detects Preclinical AD, Reflects Clinical Benefit
  6. At CTAD, Tau PET Emerges as Favored Outcome Biomarker for Trials
  7. New Definition of Alzheimer’s Hinges on Biology, Not Symptoms

Biomarker Meta Analysis Citations

Paper Citations

  1. . Comparison of three analytical platforms for quantification of the neurofilament light chain in blood samples: ELISA, electrochemiluminescence immunoassay and Simoa. Clin Chem Lab Med. 2016 Oct 1;54(10):1655-61. PubMed.
  2. . Neurofilament Light Chain in Blood and CSF as Marker of Disease Progression in Mouse Models and in Neurodegenerative Diseases. Neuron. 2016 Jul 6;91(1):56-66. Epub 2016 Jun 9 PubMed.

Further Reading

No Available Further Reading