Markers in cerebrospinal fluid provide clues to a person’s risk of Alzheimer’s disease, but such tests still fall short of a definitive diagnosis. Two recent papers propose new markers to improve the predictive power of CSF testing. Each presents a potential biomarker that tracks with cognition and sharpens diagnosis in preliminary studies. In the May 19 Nature Communications, researchers led by Ashley Bush at the University of Melbourne, Australia, report that high CSF ferritin, a protein that reflects the iron load in the brain, predicts earlier progression to dementia. Meanwhile, in the May 5 Translational Psychiatry, researchers led by Sergio Ferreira and Rogerio Panizzutti at the Federal University of Rio de Janeiro, Brazil, proposed the neuromodulator D-serine as a marker of Alzheimer’s. In a small study, patients had fivefold higher CSF levels of this amino acid than age-matched controls.

If the findings hold up in larger studies, both molecules would make attractive markers, as they are relatively easy to measure reliably in CSF, noted Henrik Zetterberg at the University of Gothenburg, Sweden, who was not involved in either study. Zetterberg was even more intrigued by the clues to disease mechanisms provided by the data, and said both findings should stimulate more basic research.

The Australian researchers have long been interested in the link between brain iron and Alzheimer’s. Some histological and MRI studies detect more iron in the brains of patients, while others report changes in its oxidation state (see Jellinger et al., 1990; Bartzokis et al., 2004Quintana et al., 2006). Since ferritin is the main carrier of iron in the body, elevated CSF ferritin might indicate a higher brain iron burden. Previous studies gave mixed results, however, with only some reporting high CSF ferritin in AD patients (see Kuiper et al., 1994; Craig-Schapiro et al., 2011Paterson et al., 2014). 

To resolve this, joint first authors Scott Ayton and Noel Faux analyzed data from the Alzheimer’s Disease Neuroimaging Initiative, which included ferritin as a CSF analyte. In agreement with previous negative findings, ferritin levels were similar between 91 cognitively healthy controls, 144 people clinically diagnosed with mild cognitive impairment, and 67 AD patients. However, the authors saw something interesting when they looked at cognition: The amount of ferritin correlated inversely with a person’s cognitive scores, regardless of their diagnosis. The authors had to use different cognitive tests to detect decline in AD patients versus those at earlier stages. For AD patients, people in the highest tertile of ferritin levels scored about three points worse on the ADAS-Cog13 than did those in the lowest. The magnitude of the cognitive difference was similar to that seen in people with a high CSF tau/Aβ42 ratio. Likewise, controls and MCI patients who had the highest levels of ferritin scored lower on a verbal learning test than their peers with less ferritin. Unlike tau/Aβ42, however, ferritin levels did not associate with a higher rate of cognitive decline over time. Instead, ferritin correlated with a set amount of cognitive loss at all stages of the disease.

The authors took this data to mean that high CSF ferritin predicted earlier progression to dementia. They calculated that an elevation of ferritin by one standard deviation over the mean of the baseline population corresponded to a 9½-month-earlier diagnosis of AD. As has been seen in other studies, high CSF tau/Aβ42 and low CSF ApoE independently predicted faster disease progression (see Toledo et al., 2014). Combining ferritin with these markers improved the sensitivity and specificity of an AD diagnosis, the authors claim.

“This is the first time that ferritin has been reported as being of diagnostic value in Alzheimer’s disease. It looks to be as valuable as the gold-standard CSF biomarkers,” Bush told Alzforum. He is attempting to replicate the results in a separate cohort of MCI and AD patients. He is also comparing ferritin levels across several neurodegenerative diseases to find out if the marker is specific for AD. Brain iron accumulation has also been reported in Parkinson’s disease.

Commentators found the implications of the data intriguing. “This well-executed study points to a widespread change in iron metabolism in Alzheimer’s disease that is associated with cognition,” said George Perry at the University of Texas, San Antonio. “It advances research in this area substantially.”

The data underscore a potentially toxic effect of brain iron, Zetterberg said. He noted that when blood leaks into brain tissue, it causes a rare disease known as superficial CNS siderosis. Iron accumulates in the cortex, and tau becomes hyperphosphorylated. Left untreated, the condition leads to neurodegeneration and dementia. The disorder is seven times more common in AD patients than in the general population (see Feb 2014 news). “Iron poisoning of the brain may injure neurons and reduce function,” Zetterberg said. In the ADNI data, high baseline ferritin levels correlated with greater atrophy of the hippocampus over the next six years, supporting a role in neurodegeneration.

Pieter Jelle Visser at VU University Medical Center in Amsterdam agreed. “Given the relatively strong correlation of high ferritin with increased CSF tau and the weak association with increased CSF Aβ42, it may be possible that ferritin reflects neurodegeneration rather than an event associated with the onset of the disease. The current evidence seems insufficient to recommend studies on lowering brain iron, but rather emphasizes the need for further cross-sectional and longitudinal studies to define the role of iron in AD,” he wrote to Alzforum (see full comment below). 

The study raises numerous questions. What causes brain iron to rise? Bush and colleagues claim that APP and tau help export iron from neurons, suggesting that dysfunction of these molecules could allow the metal to build up in AD (see Feb 2012 news). Zetterberg suggested that microbleeds, which often occur in blood vessels peppered with amyloid, might help kick off iron accumulation. One surprise in the data was that CSF ApoE levels correlated strongly but inversely with ferritin levels. The effect was particularly strong in people who carried an ApoE4 allele; they had lower CSF ApoE and higher CSF ferritin than their peers. This hints at a role for ApoE in iron metabolism.

The Brazilian study took a different tack in the race for biomarkers, focusing on D-serine, an enantiomer of the common L-amino acid. Serine racemase converts the L to the D form and is expressed by neurons, astrocytes, and microglia. D-serine is abundant in mammalian brains. It acts as a co-agonist with glutamate, binding to the glycine site of NMDA receptors to help activate these channels, which are essential for learning and memory (for review, see Wolosker et al., 2008). In Alzheimer’s, extrasynaptic NMDA receptors may become overactive, leading to synapse damage and neuron death. The approved AD drug memantine blocks these channels. In animal models, D-serine contributes to excitotoxicity at NMDA receptors (see Dec 2011 conference newsMustafa et al., 2010). 

To investigate D-serine in Alzheimer’s pathology, joint first authors Caroline Madeira and Mychael Lourenco examined postmortem brains from 17 AD patients and 12 controls. They found a twofold elevation of D-serine in the patients’ hippocampuses and parietal cortices, but not in their occipital cortices, a region relatively spared by AD.

The authors wondered if the excess would show up in CSF, as suggested by one previous study (see Fisher et al., 1998). In a pilot study on 21 AD patients, the authors saw fivefold higher levels of D-serine than in 10 age-matched controls, and about twofold higher levels than in nine patients with hydrocephalus and another nine with depression. High D-serine correlated with poorer performance on the Mini-Mental State Exam and a worse Clinical Dementia Rating, regardless of a person’s diagnosis, suggesting an effect on cognition.

Moreover, D-serine analysis improved AD diagnosis in this small study. In the Brazilian cohort, the standard CSF biomarker tau/Aβ42 detected Alzheimer’s with 81 percent sensitivity and 94 percent specificity, and adding D-serine into the equation improved sensitivity to 96 percent and specificity to 100 percent. To validate the findings, Ferreira collected CSF from another 160 AD patients in Brazil, and will compare their levels of D-serine and other neurotransmitters to those of age-matched controls. Zetterberg suggested that it would be valuable to measure D-serine in other neurodegenerative diseases such as Parkinson’s as well, to determine if this marker is specific for AD.

What might explain such extreme levels of D-serine? To investigate, the authors used cell cultures. Adding synthetic Aβ42 oligomers to hippocampal neuron cultures for 24 hours pumped up levels of D-serine, as well as mRNA and protein levels of serine racemase, the enzyme that produces D-serine. Transgenic APPPS1 mice also displayed high levels of D-serine and serine racemase in the hippocampus compared to littermate controls. The results indicate that Aβ oligomers somehow turn up expression of serine racemase. Possibly, this is a compensatory response, Ferreira speculated. NMDA receptor levels fall early in AD, which may cause microglia to boost serine levels in an attempt to restore signaling. “The data support this notion that NMDA receptor signaling becomes deregulated in the AD brain,” Ferreira said.

The findings agree with a previous study from Steve Barger at the University of Arkansas for Medical Sciences in Little Rock. He reported that Aβ stimulates cultured microglia to make more serine racemase, and the resulting boost in D-serine amplifies toxicity (see Wu et al., 2004). “Together, the data suggest that serine racemase and its product provide a mechanistic link between neuroinflammation and neurodegeneration,” Barger wrote to Alzforum (see full comment below).—Madolyn Bowman Rogers


  1. Excitatory amino acids (EAA) have a long and complex history in AD research. Considerable evidence exists for EAA dysregulation in AD pathogenesis, including depletion of the so-called “glycineB” component of the NMDA receptor (Procter et al., 1989). Nevertheless, such evidence must be resolved in the context of a generalized hypoactivity in the AD brain, as well as the limited efficacy of memantine, an NMDA-receptor antagonist. These caveats are compounded by the fact that the direct impact of Aβ on excitatory synapses is predominantly inhibitory. The latter could belie a very different effect that is mediated indirectly through Aβ’s activation of glia. Microglia exposed to Aβ quickly begin to release substantial levels of glutamate via a chain of biochemical events that begins with the oxidative burst and ultimately involves the system-Xc glutamate/cystine antiporter (Barger et al., 2007). In addition to this early response, a somewhat slower induction of the gene for serine racemase can result in an order-of-magnitude elevation in production of dextrorotatory serine (Wu et al., 2004), now known to be the most relevant ligand for “glycineB” sites in the forebrain. The D-serine thus produced contributes significantly to the neurotoxicity that activated microglia exhibit in coculture with neurons, and serine racemase-knockout mice are less severely impacted by intracerebral injections of Aβ (Inoue et al., 2008). The serine racemase promoter was shown to be responsive to Jun-kinase activation in microglia via the AP-1 transcription factor (Wu et al., 2004). Together, the data suggest that serine racemase and its product provide a mechanistic link between neuroinflammation and neurodegeneration.

    Some years ago, a talented graduate student working in my laboratory, Shengzhou Wu, documented an elevated expression of serine racemase in AD temporal-lobe cortex (Wu et al., 2004). Madeira et al. have now reported that measurements of D-serine in the CSF can serve as a useful additional discriminator in AD diagnosis. Alone, [D-Ser]CSF proved more sensitive than the Aβ:Tau ratio; when coupled with the latter index, [D-Ser]CSF boosted specificity to essentially 100 percent. This report also makes several other useful contributions to our understanding of D-serine, including evidence of regional specificity of [D-Ser] in AD and rather impressive quantitative correlations between [D-Ser]CSF and AD staging (by MMSE and CDR). As the authors point out, this temporal and regional specificity may explain previous reports that failed to find a correspondence between AD and D-serine levels (Chouinard et al., 1993; Nagata et al., 1995). Serine racemase could be induced only transiently, elevating D-serine in a wave across the pathogenic progression from hippocampus to temporal neocortex and beyond; sampling of specific regions at the wrong time might miss the hypothetical wave. 

    This sort of conditional elevation of D-serine could also speak to the broader issues of excitatory amino acids and neuroinflammation in AD. Madeira et al. are careful to point out that D-serine is not wholly a bad thing. It is a deficiency in D-serine, not a glut, that appears to play a role in psychosis (Labrie et al., 2009); one must consider the possibility that elevation of this molecule after microglial activation represents part of a compensatory response. We would do well to remember that EAA stimulation can result in diminished electrophysiological activity. The classic work of John Olney, for instance, documented a role for such inhibition in the untoward effects of excitotoxins after their activation of GABAergic neurons (Olney et al., 1997). Likewise, acute elevations of glutamate, such as occur in ischemia or traumatic brain injury, can create a reactive downregulation of ionotropic glutamate receptors; remediation at later time points is effected not by glutamate-receptor antagonists but by agonists. One such agonist showing utility in preclinical models of acute excitotoxicity is D-cycloserine, a pharmacological cousin of D-serine. Another unexpected result came from an ALS mouse model, where elevation of D-serine hastened disease onset but also delayed progression (Thompson et al., 2012). This may reflect a delayed or secondary stage of disease that results not from hyperexcitation but from hyperinhibition (Schutz, 2005). Therefore, it is worth considering that serine racemase induction and other aspects of glial activation exist to supplant the trophic neurochemical activity otherwise reduced by Aβ. 


    . Glutamate release from activated microglia requires the oxidative burst and lipid peroxidation. J Neurochem. 2007 Jun;101(5):1205-13. Epub 2007 Mar 30 PubMed.

    . Presence of the N-methyl-D-aspartate-associated glycine receptor agonist, D-serine, in human temporal cortex: comparison of normal, Parkinson, and Alzheimer tissues. J Neurochem. 1993 Oct;61(4):1561-4. PubMed.

    . NMDA- and beta-amyloid1-42-induced neurotoxicity is attenuated in serine racemase knock-out mice. J Neurosci. 2008 Dec 31;28(53):14486-91. PubMed.

    . Serine racemase is associated with schizophrenia susceptibility in humans and in a mouse model. Hum Mol Genet. 2009 Sep 1;18(17):3227-43. Epub 2009 May 30 PubMed.

    . d-serine levels in Alzheimer's disease: implications for novel biomarker development. Transl Psychiatry. 2015 May 5;5:e561. PubMed.

    . Free D-serine concentration in normal and Alzheimer human brain. Brain Res Bull. 1995;38(2):181-3. PubMed.

    . Excitotoxic neurodegeneration in Alzheimer disease. New hypothesis and new therapeutic strategies. Arch Neurol. 1997 Oct;54(10):1234-40. PubMed.

    . Reduced glycine stimulation of [3H]MK-801 binding in Alzheimer's disease. J Neurochem. 1989 Sep;53(3):698-704. PubMed.

    . Imbalanced excitatory to inhibitory synaptic input precedes motor neuron degeneration in an animal model of amyotrophic lateral sclerosis. Neurobiol Dis. 2005 Oct;20(1):131-40. PubMed.

    . Paradoxical roles of serine racemase and d-serine in the G93A mSOD1 mouse model of amyotrophic lateral sclerosis. J Neurochem. 2012 Feb;120(4):598-610. PubMed.

    . Induction of serine racemase by inflammatory stimuli is dependent on AP-1. Ann N Y Acad Sci. 2004 Dec;1035:133-46. PubMed.

    . Induction of serine racemase expression and D-serine release from microglia by amyloid beta-peptide. J Neuroinflammation. 2004 Apr 20;1(1):2. PubMed.

    View all comments by Steve Barger
  2. Ayton et al. published interesting data on CSF ferritin as a biomarker for AD. Based on the cognitive data, the authors conclude that ferritin is a trait marker because it correlates cross-sectionally with cognitive scores, but not with change in cognition over time. Still, they also observed that in MCI, baseline ferritin concentrations predicted progression to AD-type dementia, which seems to contradict the first finding. One possible explanation for the discrepancy is that in the first analysis both control and MCI subjects were included. The authors added diagnosis as a factor but did not report that they tested the interaction between diagnosis and change over time; previous studies clearly showed that cognitive change is different between controls and MCI subjects. Hence, it may be possible that a lack of prediction in cognitive change was because the controls improved in cognition at follow-up while MCI subjects declined. Another surprising finding is that ferritin did not differ between diagnostic groups, though it was associated with cognitive scores and also with AD biomarkers that all differ between diagnostic groups. Given the relatively strong correlation of increased ferritin with increased CSF tau and the weak association with increased CSF Aβ42, it may be possible that increased ferritin reflects neurodegeneration rather than an event associated with the onset of the disease. The current evidence seems insufficient to recommend studies on lowering brain iron, but rather emphasizes the need for further cross-sectional and longitudinal studies to define the role of iron in AD.

    The paper of Madeira et al. convincingly shows that D-serine is increased in AD, although the studies, based on CSF, had a relatively small sample size. D-serine correlated with a ratio of CSF Aβ42 and total tau providing additional support for the involvement of D-serine in AD. It would be of interest to see whether D-serine more strongly correlates with CSF tau or CSF Aβ42, as this would give an indication of whether D-serine reflects neuronal injury or amyloid aggregation.

  3. Following a suggestion posted by Pieter Jelle Visser, we have re-analyzed our data (Madeira et al., 2015) to examine individual correlations between CSF D-serine levels and Aβ42/tau/p-tau181 levels in AD and control individuals. The first conclusion from this re-analysis was that there is no correlation between CSF D-serine levels and p-tau181 levels (R2 = 0.0002; P = 0.46). The second conclusion was that there is indeed a negative correlation (R2 = 0.45; P < 0.0001) between CSF D-serine and Aβ42 levels, as might be expected if D-serine is up-regulated and released from neurons and/or glia in response to brain accumulation of Abeta oligomers and amyloid aggregation.

    Intriguingly, however, D-serine levels were positively correlated with total tau levels in the CSF (R2 = 0.37; P = 0.0008), which might indicate that, at least in part, elevation of D-serine in the CSF might be related to neurodegeneration. The overall correlation between the IATI index and D-serine levels that we reported thus appears to originate from both Aβ42 and total tau measures in CSF.

    We note, however, that the sample size used in this study was small. Results were robust when the IATI index was correlated with D-serine levels, but it is possible that the study was insufficiently powered to detect individual correlations between the various analytes in CSF. It will be interesting to further investigate the origin of the D-serine increase in CSF in a future study with a larger cohort of demented and control patients.


    . d-serine levels in Alzheimer's disease: implications for novel biomarker development. Transl Psychiatry. 2015 May 5;5:e561. PubMed.

Make a Comment

To make a comment you must login or register.


News Citations

  1. Alzheimer’s Disease Linked to Superficial Siderosis. What Does it Mean?
  2. Tau Joins APP in the Ironworks
  3. DC: Targeting Toxic Extrasynaptic Signaling in HD and AD

Research Models Citations

  1. APPswe/PSEN1dE9 (line 85)

Paper Citations

  1. . Brain iron and ferritin in Parkinson's and Alzheimer's diseases. J Neural Transm Park Dis Dement Sect. 1990;2(4):327-40. PubMed.
  2. . Brain ferritin iron as a risk factor for age at onset in neurodegenerative diseases. Ann N Y Acad Sci. 2004 Mar;1012:224-36. PubMed.
  3. . Study of the localization of iron, ferritin, and hemosiderin in Alzheimer's disease hippocampus by analytical microscopy at the subcellular level. J Struct Biol. 2006 Jan;153(1):42-54. PubMed.
  4. . Cerebrospinal fluid ferritin levels of patients with Parkinson's disease, Alzheimer's disease, and multiple system atrophy. J Neural Transm Park Dis Dement Sect. 1994;7(2):109-14. PubMed.
  5. . Multiplexed immunoassay panel identifies novel CSF biomarkers for Alzheimer's disease diagnosis and prognosis. PLoS One. 2011;6(4):e18850. PubMed.
  6. . Cerebrospinal fluid markers including trefoil factor 3 are associated with neurodegeneration in amyloid-positive individuals. Transl Psychiatry. 2014 Jul 29;4:e419. PubMed.
  7. . CSF Apo-E levels associate with cognitive decline and MRI changes. Acta Neuropathol. 2014 May;127(5):621-32. Epub 2014 Jan 3 PubMed.
  8. . D-amino acids in the brain: D-serine in neurotransmission and neurodegeneration. FEBS J. 2008 Jul;275(14):3514-26. PubMed.
  9. . Serine racemase deletion protects against cerebral ischemia and excitotoxicity. J Neurosci. 2010 Jan 27;30(4):1413-6. PubMed.
  10. . Free D- and L-amino acids in ventricular cerebrospinal fluid from Alzheimer and normal subjects. Amino Acids. 1998;15(3):263-9. PubMed.
  11. . Induction of serine racemase expression and D-serine release from microglia by amyloid beta-peptide. J Neuroinflammation. 2004 Apr 20;1(1):2. PubMed.

Further Reading

Primary Papers

  1. . Ferritin levels in the cerebrospinal fluid predict Alzheimer's disease outcomes and are regulated by APOE. Nat Commun. 2015 May 19;6:6760. PubMed.
  2. . d-serine levels in Alzheimer's disease: implications for novel biomarker development. Transl Psychiatry. 2015 May 5;5:e561. PubMed.