Objective To investigate whether cerebrospinal fluid (CSF) ferritin (reporting brain iron) is associated with longitudinal changes in CSF β-amyloid (Aβ) and tau.
Methods Mixed-effects models of CSF Aβ1-42 and tau were constructed using data from 296 participants who had baseline measurement of CSF ferritin and annual measurement of CSF tau and Aβ1-42 for up to 5 years.
Results In subjects with biomarker-confirmed Alzheimer’s pathology, high CSF ferritin (>6.2 ng/mL) was associated with accelerated depreciation of CSF Aβ1-42 (reporting increased plaque formation; p=0.0001). CSF ferritin was neither associated with changes in CSF tau in the same subjects, nor longitudinal changes in CSF tau or Aβ1-42 in subjects with low baseline pathology. In simulation modelling of the natural history of Aβ deposition, which we estimated to occur over 31.4 years, we predicted that it would take 12.6 years to reach the pathology threshold value of CSF Aβ from healthy normal levels, and this interval is not affected by CSF ferritin. CSF ferritin influences the fall in CSF Aβ over the next phase, where high CSF ferritin accelerated the transition from threshold preclinical Aβ levels to the average level of Alzheimer’s subjects from 18.8 to 10.8 years.
Conclusions Iron might facilitate Aβ deposition in Alzheimer’s and accelerate the disease process.
- alzheimer’s disease
- iron deposition
Statistics from Altmetric.com
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.
In the natural history of Alzheimer’s disease (AD), β-amyloid (Aβ) accumulation, detected either by Aβ-positron emission tomography (PET) imaging or measuring falling cerebrospinal fluid (CSF) Aβ1-42 levels, progresses in a prodromal period lasting decades decade-long prodromal period.1–4 Risk factors that underlie the considerable variability of amyloid accumulation rate are uncertain. Major genetic factors, such as familial AD mutations, or the ε4 isoform of APOE, cause amyloid to commence accumulation earlier, but they have little impact on Aβ accumulation rate.2 4
Cortical iron elevation is also a pathological feature of AD. We recently showed that elevated CSF ferritin, a reporter of brain iron load, was longitudinally associated with worse cognitive performance (controlling for inflammation and blood leakage).5 We also used quantitative susceptibility mapping, an MRI modality sensitive to brain iron levels, to demonstrate that elevated brain iron was associated with amyloid burden determined by PET and longitudinal cognitive decline.6 Here we investigate whether the CSF ferritin predicts longitudinal changes in AD pathology by investigating changes in CSF Aβ1-42 and tau over a 5-year period.
Data were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu; 30 August 2016). ADNI was launched in 2003 as a public–private partnership, led by Principal Investigator Michael W Weiner. The primary goal of ADNI has been to test whether serial MRI, PET, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of mild cognitive impairment and early AD.
The study received prior approval by institutional review boards of the study sites. Informed consent was obtained from each subject. The ADNI study protocols and patient inclusion criteria have been reported previously.3 5 7 CSF levels of ApoE and ferritin were measured with the RBM multiplex platform,5 and yearly levels of CSF Aβ1-42 and tau were measured for up to 5 years with the multiplex xMAP Luminex platform.3 5 7
The cohort was stratified according to the absence or presence of pathology using a previously determined threshold7 in the tau/Aβ1-42 ratio (AD >0.39 units). We chose the tau/Aβ1-42 index to stratify subjects in the separate longitudinal modelling of Aβ1-42 levels and tau levels in preference to stratification using equivalent threshold values in Aβ1-42 (192 pg/mL) and tau (93 pg/mL) that could have been applied for each biomarker, so as to ensure group membership was identical for each mixed-effects model. For the hypothetical model, we stratified subjects according to the CSF Aβ1-42 threshold (192 pg/mL) since we modelled only Aβ1-42. Thresholding for ferritin was calculated by including incremental increases in ferritin cut-off score in linear mixed models of CSF tau and Aβ1-42 levels.
Models were performed in R (V.3.2.4) and tested for normal distribution of the residuals and absence of multicollinearity.
Two hundred and ninety-six participants had baseline measurements of CSF ferritin and repeated measurement of CSF tau and Aβ1-42 annually for up to 5 years (table 1A). In a mixed-effects model of CSF Aβ1-42 in subjects with high baseline tau/Aβ1-42, CSF ferritin predicted decreasing Aβ1-42 levels when included as a dichotomous variable, with a median cut-off of 6.7 ng/mL (p=0.005), but not when included as a continuous variable (p=0.597). To determine the optimal threshold of CSF ferritin to predict change in Aβ1-42 levels, we repeated the mixed-effects model using incremental increases of CSF ferritin cut-off level (figure 1A). We found that 6.2 ng/mL was the best performing threshold, and strongly predicted decreasing levels of CSF Aβ1-42 (reporting brain Aβ deposition) in the mixed-effects model, which was controlled for age, sex, APOE ε4, CSF ApoE, CSF tau and diagnosis (β (SE)=−2.85 (0.7), p=0.0001; figure 1B, table 1B). CSF ferritin was not predictive of change of CSF Aβ1-42 in subjects with low baseline tau/Aβ1-42 levels (whether included as a continuous variable or for any threshold), nor was CSF ferritin predictive of change in similar models of CSF tau, and regardless of baseline CSF tau/Aβ1-42 level (table 1C).
We used the beta coefficients of the association between ferritin and annual change in Aβ1-42 derived from these mixed-effects models (table 1B) to calculate the impact of high CSF ferritin on the natural history of AD (figure 1B). We calculated that the time required for Aβ1-42 levels to fall from average levels of cognitively normal subjects with no amyloid pathology (233 pg/mL), to the threshold cut-off determined for brain Aβ positivity (192 pg/mL7), was 12.6 years (95% CI 10 to 15) and was not affected by CSF ferritin. Our model indicates that people with low CSF ferritin levels (<6.2 ng/mL) would follow a similar trajectory of Aβ1-42 decline when tau/Aβ1-42 levels are above the threshold for positive Alzheimer’s pathology, and it would take a further 18.8 (95% CI 17 to 20) years for CSF Aβ1-42 levels to decrease from the lower limit of normal (192 pg/mL) to the average value for clinical AD (132.6 pg/mL). However, subjects with high CSF ferritin would arrive at the average CSF Aβ1-42 value for clinical AD 8 years sooner than subjects with low CSF ferritin, because it would take only 10.8 (95% CI 9 to 12) years for CSF Aβ1-42 levels to decrease from the pathological threshold value to the average value for AD.
The calculation of the 30-year time frame of Aβ deposition is consistent with prior work,1–3 8 with the caveat that we model a linear change in CSF Aβ levels over age, whereas prior studies model a sigmoidal profile. The sigmoidal profiles previously reported were produced by models where baseline levels of either CSF Aβ levels or PET-determined amyloid load predicted the rate of change of the respective biomarkers.1 3 In our analysis, we performed mixed-effects modelling of non-transformed Aβ1-42 values at each collection point (table 1B), which is unsuitable for predicting rate changes as a function of baseline levels of Aβ1-42 (because baseline Aβ1-42 levels are also included in the outcome measure of the mixed-effects model, and therefore cannot also be included as a predictive variable). Therefore, we also performed modelling of change in Aβ1-42 values over time and included baseline Aβ1-42 as a predictive variable. In this model, CSF ferritin was again predictive of the linear rate of Aβ1-42 decline in subjects with high baseline tau/Aβ1-42 levels (p=0.001), similar to the model presented in table 1B. But in the model of Aβ1-42 change, we found that baseline Aβ1-42 levels exhibited high multicollinearity (variance inflation factor 28.6), which causes unreliable estimates of regression coefficients, so we could not use the regression coefficients of baseline Aβ1-42 levels in our hypothetical model. We therefore only estimated the change in CSF Aβ1-42 levels up until the average value of CSF Aβ1-42 in AD, where other studies of this epoch of the natural history have also observed a linear change in amyloid burden (whether determined by either CSF Aβ1-42 levels or PET), after which the rate of change begins to plateau.1 3 8
We provide evidence that brain iron elevation, as reported by CSF ferritin, might accelerate Aβ deposition in people with biomarker-determined AD pathology. Our findings accord with recent cross-sectional studies showing the association between iron elevation (measured with MRI) and amyloid deposition (measured with PET) in vivo.6 9 Brain iron has been implicated in plaque pathogenesis since iron is enriched in Aβ plaque,10 11 iron increases the aggregation of Aβ in vitro12 13 and promotes deposition of Aβ in murine models,14 15 but the present study is the first to show a relationship between iron and Aβ deposition in a longitudinal clinical study. While we have previously shown that ferritin is not elevated in people with high CSF tau/Aβ1-42 (ref 5) and therefore not an underlying cause of the disease, our data suggest that iron may be a major moderator of disease since we have shown that those people with higher ferritin values have accelerated symptom progression5 16 and pathology accumulation (current results) than those with low ferritin.
That CSF ferritin had no association with CSF Aβ1-42 levels in subjects without biomarker evidence of amyloid pathology could indicate that elevated brain iron might exaggerate the rate of amyloid accumulation only after the amyloid pathology is manifest. This would be consistent with genetic diseases causing brain iron elevation, such as neurodegeneration with brain iron accumulation, not usually being complicated by Aβ pathology.17 However, elevated brain iron levels in AD might facilitate further deposition of Aβ once Aβ pathology takes hold.
Limitations with the current study are that we used a surrogate marker of brain iron load, CSF ferritin. Serum ferritin is a routine clinical assay for body iron load; however, CSF ferritin is not established as a biomarker of brain iron load. The accumulative evidence supports that CSF ferritin reflects brain iron status; for example, CSF ferritin levels are decreased in a restless legs syndrome, a disorder of low brain iron,18 and increased in the brain iron-accumulating Parkinson’s disease, which is reversible by iron chelation therapy.19 But ferritin is also an acute phase protein, which increases in certain inflammatory diseases, and conceivably could be reporting inflammation instead of iron. However, we previously showed that CSF reporters of inflammation are only modestly associated with ferritin, and their association is abolished when other clinical variables are considered.5 Furthermore, CSF ferritin is not elevated in AD,5 suggesting that CSF ferritin levels are not responding to the inflammatory processes in this disease.
A further limitation is that we were not able to measure CSF albumin to determine whether the CSF values of ferritin were likely contaminated by ferritin from the blood. We however have previously shown in these subjects that CSF values of haemoglobin (also a reporter of blood contamination) were not related to the amount of ferritin in the CSF, and there was also only a modest association between plasma ferritin and CSF ferritin,5 so the ferritin values more likely reflect brain-derived ferritin.
Elevated brain iron levels in AD might facilitate the deposition of Aβ and accelerate the disease process.
SA and ID contributed equally.
Contributors SA: Scientific concept, wrote manuscript, data modelling. ID: Data modelling, preparation of figures and tables. AIB: Supervised analysis, edited manuscript, funded analysis.
Funding Data collection and sharing for this project was funded by ADNI (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). ADNI is funded by the National Institute on Aging, the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; BioClinica; Biogen Idec; Bristol-Myers Squibb Company; Eisai; Elan Pharmaceuticals; Eli Lilly and Company; F Hoffmann-La Roche and its affiliated company Genentech; GE Healthcare; Innogenetics, NV; IXICO; Janssen Alzheimer Immunotherapy Research & Development; Johnson & Johnson Pharmaceutical Research & Development; Medpace; Merck & Co; Meso Scale Diagnostics; NeuroRx Research; Novartis Pharmaceuticals Corporation; Pfizer; Piramal Imaging; Servier; Synarc; and Takeda Pharmaceutical Company. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organisation is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Disease Cooperative Study at the University of California, San Diego. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California. Analysis was supported by funds from the Australian Research Council, the Australian National Health & Medical Research Council (NHMRC), and the CRC for Mental Health (the Cooperative Research Centre (CRC) programme is an Australian Government Initiative). The Florey Institute of Neuroscience and Mental Health acknowledges support from the Victorian Government, in particular funding from the Operational Infrastructure Support Grant. No funder of this study had any role in the design and conduct of the study; collection, management, analysis or interpretation of the data; preparation, review or approval of the manuscript; and decision to submit the manuscript for publication.
Competing interests AIB is a shareholder in Prana Biotechnology, Cogstate, Eucalyptus, Mesoblast, Brighton Biotech, Nexvet, Grunbiotics and Collaborative Medicinal Development, and a paid consultant for Collaborative Medicinal Development. SA, ID and AIB have a filed a provisional patent that contains data from this manuscript. SA and AIB have received funding relevant to this study from the NHMRC, Alzheimer’s Association, Alzheimer’s Research UK, The Michael J Fox Foundation for Parkinson’s Research, and Weston Brain Institute.
Patient consent Obtained.
Ethics approval Multiple sites.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All raw data are available on the ADNI website.
Collaborators Data used in preparation of this article were obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database (adni.loni.usc.edu). As such, the investigators within the ADNI contributed to the design and implementation of ADNI and/or provided data but did not participate in analysis or writing of this report. A complete listing of ADNI investigators can be found at: http://adni.loni.usc.edu/wp1ontent/uploads/how_to_apply/ADNI_Acknowledgement_List.pdf