|Home | About | Journals | Submit | Contact Us | Français|
While cerebrospinal fluid (CSF) biomarkers are of use in the prediction and diagnosis of Alzheimer’s disease our understanding of the background effects of age and the ApoE genotype is limited. Seventy-eight community-based normal volunteers (mean age 60 ± 10 years, range 36–86) were examined to determine the relationships between CSF measures of total tau (T-tau), hyperphosphorylated tau (P-tau 231), amyloid beta (Aβ42/Aβ40 ratio), and isoprostane (IP) with age and ApoE genotype. The results showed that age by ε4 genotype interactions were found for P-tau231 (β = 1.82; p < 0.05) and IP (β = 1.6; p < 0.05). T-tau CSF concentration increased with age. The increasing CSF concentrations of P-tau and IP in ε4 carriers suggest that early tauopathy and oxidative stress may be related to the increased risk for AD. The data also suggest that T-tau changes are more age dependent than Aβ changes. The evidence that P-tau231 and IP are the earliest markers for the neuronal damage related to AD awaits longitudinal study. © 2007 Elsevier Inc. All rights reserved.
Age is major risk factor for AD (Bachman et al., 1993). Neuropathology (Bennett et al., 2006) and imaging studies [for review (Glodzik-Sobanska et al., 2005)] have shown that Alzheimer pathology begins years before clinical presentation. Both neurofibrillary tangles (NFT) and amyloid plaques are present in the brains of clinically normal subjects (Knopman et al., 2003), as are volume (Rusinek et al., 2003) and metabolic losses (de Leon et al., 2001) in limbic regions in aging subjects that predict cognitive decline.
Several CSF assays have been developed to estimate in vivo tau and amyloid pathology. These have proven useful in the diagnosis of AD and in the prediction of AD from the MCI stage (Blennow et al., 2006), and for prediction of decline among normal to MCI (Fagan et al., 2007). The total CSF tau (T-tau) concentration mirrors neuronal and axonal loss (Mori et al., 1995). However, the increased CSF T-tau level is not specific for AD as levels are also elevated in other neurodegenerative diseases (Arai et al., 1997; Mollenhauer et al., 2005). Overall, CSF T-tau reflects both the normal metabolism of tau, and the non-specific release of tau following neuronal damage, whereas P-tau231 reflects abnormal tau metabolism that is specific for tangles and for AD (Mitchell and Brindle, 2003). Several isoforms for hyperphosphory-lated tau (P-tau) exist, all reflecting NFT pathology (Blennow, 2004). Meta-analysis indicated a specificity (AD vs. control) of 94% for P-tau and 82% for T-tau; sensitivity was 70% and 83%, respectively. (Blennow and Vanmechelen, 2003).
The decrease in CSF Aβ42 concentration is a reliable finding in AD samples as compared to controls, with both sensitivity and specificity fluctuating around 85% (Blennow and Vanmechelen, 2003). Recent studies suggested that the Aβ42/Aβ40 ratio instead of Aβ42 increases the overall correct classification (AD patients and controls) (Lewczuk et al., 2004a,b). The changes in Aβ42 may result not only from AD but also can be a consequence of total Aβ peptides concentrations (Aβ load), The load is best estimated by Aβ40 concentration, therefore the use of Aβ42/Aβ40 is recommended (Wiltfang et al., 2007).
CSF isoprostane, a marker of oxidative stress and inflammation has been also shown to be increased in AD (Pratico et al., 2002) and to predict AD at the MCI stage (Brys et al., 2007; de Leon et al., 2007).
The apolipoprotein E (ApoE) polymorphism is the most widely accepted genetic factor increasing the risk for sporadic AD (Farrer et al., 1997). The presence of an ApoE ε4 allele is associated with a reduced age of clinical onset of AD, an increased extent of AD pathology in post mortem studies (Ohm et al., 1999) and altered CSF T-tau and Aβ42 biomarkers levels in AD patients (Galasko et al., 1998).
With few exeptions the effects of age and ApoE genotype on CSF levels of AD biomarkers in normal individuals have not been addressed (Peskind et al., 2006; Sunderland et al., 2004). This paper examines these relationships in cross-section for five commonly used biomarkers.
Seventy eight healthy subjects (mean age 60.4 ± 10.4 years, range 36–86) were recruited to participate in longitudinal research studies at the Center for Brain Health (CBH) of the New York University School of Medicine NIH-NIA funded Alzheimer’s Disease Center. Volunteers were recruited through advertisement about the research program, or were the spouses and caregivers of impaired patients being evaluated at the center. Education ranged from 10 to 20 years with a mean of 16.5 ± 2 years. The study was approved by the NYU Institutional Review Board. All the subjects signed informed consent form allowing for lumbar puncture and the analysis of CSF.
All subjects received complete clinical assessments (medical, neurological, psychiatric, radiological and laboratory examinations). It was shown that cholesterol reducing medications (e.g. statins) (De Caterina et al., 2002) and antioxidant agents (i.e. vitamin E) (Sutherland et al., 2007) can affect IP levels in urine and plasma. As knowledge on what drugs may effect biomarker levels in the CSF is limited, specific information was gathered concerning the use of medications. Participants received a comprehensive blood work-up (including lipid profile, blood count, homocysteine levels) and MRI was used to exclude possible confounding brain pathology (e.g. tumor, infarction etc.). Individuals with medical conditions or history of significant conditions that might affect brain integrity (e.g. stroke, head trauma, other neurode-generative disease, or depression) were excluded.
All participants received a Global Deterioration Scale (GDS) rating of 1 or 2 (Reisberg et al., 1993). Both GDS 1and 2 represent normal functioning, where GDS = 1 indicates no subjective memory complaint, and GDS = 2 indicates awareness of memory change over the lifespan, in the absence of objective evidence of memory or functional problems on clinical interview. A semi-structured interview based on the brief cognitive rating scale (BCRS) (Reisberg et al., 1983) was used to query the subject and in most cases, also an informant for the subject’s recent memory performance and global functioning. In addition to the clinical evaluation, in every case, a neuropsychological test battery was administered but was not used to determine normal functioning. Rather cognitive tests were used for descriptive purposes. The measures include subtests of the Guild Memory Scale (Gilbert et al., 1968) assessing delayed and immediate recall of orally presented paragraphs (initial: PARI, and delayed: PARD); verbal paired associates (initial: PRDI, and delayed: PRDD); visual/ verbal paired associates with numbers (DESN). Subtests of the Wechsler Intelligence Scale Revised (WAIS-R) (Wechsler, 1981) were used to assess working memory (digits forward: (WAISDIG-F, and backward: WAISDIG-B), and attention (digit symbol substitution subtest: DSST).
After 12 h fasting, at 11a.m., 15 ml of clear CSF was collected into three polypropylene tubes using a fine 25G LP needle guided by fluoroscopy. All CSF samples were kept on ice for a maximum of 1 h until being centrifuged for 10 min at 1500 g at 4 °C. Samples were aliquoted to 0.25 ml and were stored in polypropylene tubes at −80 °C. All samples were blindly analyzed in batches. The subject distributions for genotype and age did not differ between the batches. For all the analytes the intra-observer and inter-observer variabilities were calculated by the collaborating labs.
CSF T-tau measurements were determined using the commercially available INNOTEST hTAU Antigen kit from Innogenetics (Hansson et al., 2006). The detection limit is 60 pg/ml for T-tau (INNOTEST hTau) and the coefficients of variability are 5.5% (intra-assay) and 11.6% (inter-assay).
A sandwich ELISA assay (Applied NeuroSolutions Vernon Hills, IL, USA) was used to detect tau phosphorylated at threonine 231 in CSF. In this assay, tau is captured with two backbone-directed antibodies, tau-1 and CP-27. The captured tau is then detected by CP9, which is specific for P-tau231 (Kohnken et al., 2000). The detection limits for this assay was 9 pg/ml. The coefficients of variation ranged from 6.0 to 10.3% (intra-assay) and 11.6 to 14.4% (inter-assay).
Using ELISA, CSF Aβ levels of A(β40 and Aβ42 were measured using the monoclonal antibody 6E10 (specific to an epitope present on Ab-16) and rabbit antisera to Aβ1–40 and Aβ1–42, respectively, in a double antibody sandwich ELISA (Metha et al., 2000). The detection limit for Aβ1–40 and Aβ1–42 was 10pg/mL. The coefficients of variability ranged from 8–14% (intra-assay) and 10–18% (inter-assay). The amyloid beta assays were not N-terminus specific. The epitopes used were x-40 and x-42. Consistent with recent reports, we used the ratio of these two species of Aβ–Aβ42/Aβ40 (Lewczuk et al., 2004a,b).
CSF samples were spiked with a fixed amount of internal standard ((d4)8,12-iso-iPF2a-VI) and extracted on a C18 cartridge column. The eluate was purified by thin layer chromatography and finally assayed by negative ion chemical ionization gas chromatography/mass spectrometry (Pratico et al., 2000). The coefficient of variation ranged 4–7% (intraassay) and 4.5–6.5% (inter-assay).
Genotyping was conducted using methods previously published by Main et al. (1991). Study subjects were classified as ApoE ε4 positive if they had one or two ε4 alleles and otherwise ε4 negative.
To compare clinical measures, demographic characteristics, and medication histories between ApoE genotype groups we used χ2 and t-test. Cognitive variables were studied using MANOVA (general linear models). To examine the relationships between biomarker concentrations and age and ApoE genotype we used linear regression models. We determined the statistical power for the reported CSF models. Altogether for different analytes it ranged from around 80–98%. When appropriate the biomarker data was square root or log transformed to a normal distribution and analyses were repeated to insure that deviation from normality did not affect the results. For regression models, the standardized β are reported. To examine the relationships between biomarkers the Pearson-product correlation coefficient was used. The proportions of patients taking different classes of drugs in ApoE groups were compared with χ2 test. Lipid profile (cholesterol, LDL, HDL, triglycerides) and blood pressure (diastolic and systolic) were examined in two separate MANOVA models,with ApoE group as the grouping factor.
All reported p values were declared statistically significant when p ≤ 0.05. The analyses were performed.
There were no differences between ApoE ε4 negative (n = 48) and ApoE ε4 positive (n = 30) groups for age, education, gender or GDS score. The ApoE groups also did not differ in their neuropsychological test performances (F = 1.6, df = 9.65, p > 0.05, see Table 1). The percentages of subjects using different types of drugs in two ApoE groups did not differ (see Table 2).
Only BMI showed genotype related effects with lower BMI in the ApoE ε4 positive group (t = 3.2, df = 1.72, p < 0.01). For lipid profile only a trend was found (MANOVA, F = 2.3, df = 65, p = 0.07) and further analysis revealed that is was due to higher triglyceride levels in ApoE ε4 negative group (F = 8.9, p < 0.05).
Significant age (β = 0.50, p < 0.001) and genotype (β = 0.26, p < 0.01) effects were observed for T-tau concentrations (for the whole model: F = 15.5, df=2, p < 0.001).
Higher T-tau was associated with older age and the ApoE ε4 positive group (means: ε4 positive = 344.4 ± 152.8; ε4 negative = 277.1 ± 135.6). No age by genotype interaction effect was observed (see Fig. 1 A). Fig. 1A shows that the regression lines for T-tau by age as a function of ApoE group are parallel. The results remained unchanged after square root transformation of the T-tau concentrations.
Significant age (β = 0.48, p < 0.001) and genotype (β = 0.25, p < 0.05) effects were observed for P-tau231 concentrations (for the whole model: F = 14.0, df = 2, p < 0.001). Higher P-tau231 was associated with older age and the ApoE ε4 positive group (means: ε4 positive = 13.7 ± 12.6; e4 negative = 9.4 ± 7.5). The age by genotype interaction was significant (β = 1.82; p < 0.05, (the change for the interaction step: F = 10.7, df= 1, p < 0.05). The results remained similar after the square root transformation of P-tau231 (for interaction β = 1.20; p = 0.05, the change in the interaction step: F =4.1, df= 1, p < 0.05). Overall, the results indicated that the P-tau231 concentration increased with age in both genotype groups, and that this increase was greater in the ApoE ε4 positive group (Fig. 1B).
Significant age (β = 0.21, p < 0.01) and genotype (β = 0.32, p < 0.01) effects were observed for IP (for the whole model: F = 6.1, df = 2, p < 0.01). Higher IP was associated with older age and with the ε4 positive groups (means: ε4 positive = 35.1 ± 16.2; ε4 negative = 27.4 ±7.9). The age by genotype interaction was significant (β = 1.60; p < 0.05, the change for the interaction step: F = 7.2, df = 1.73 p < 0.05). However with inclusion of the age by genotype interaction, age alone was no longer significant in the model. These results remained unchanged after the log transformation of the IP levels (interaction β = 1.40; p < 0.05, the change for the interaction step was F = 4.0, df = 1, p < 0.05). We repeated the regression analysis in the two genotype groups separately. For the ApoE ε4 positive group there was an age effect (β = 0.43, p < 0.01), which was not found in the ApoE ε4 negative group (β = 0.03, p = 0.80). Overall, the results indicated that the CSF isoprostane level did not change with age in the ApoE ε4 negative group, while it increased linearly in the ApoE ε4 positive group (Fig. 1C).
The Aβ42/Aβ40 ratio decreased with age, but only a statistical trend was observed (β = −0.21, p = 0.06) (for the whole model: F = 2.6, df = 2, p = 0.08). The ApoE genotype effect and age interaction were not significant (Fig. 1D). Mean Aβ42/Aβ40 values for the ApoE ε4 positive and negative groups were: 11.5 ±3.0 and 12.6 ±3.9, respectively). The results remained unchanged after square root transformation.
No significant age, genotype or interaction effects were found for A β40 or A β42.
For the ε4 positive and negative groups the values for Aβ40 were 8552.90±2420.04 and 8634.23±3418.51, respectively. The values for Aβ42 were 980.40±406.04 and 1109.2±566.0 for ε4 positive and negative groups, respectively.
No gender effects for any of the biomarkers were found.
To our knowledge, this is the first report describing normal lifespan changes in multiple CSF biomarkers used for Alzheimer’s disease. We examined, in normal subjects, the relationship between age and ApoE genotype for five CSF AD biomarkers. The results showed age by genotype interactions, for some analytes and for others only age effects. No CSF biomarker marker showed effect restricted to genotype. Specifically, both P-tau231 and IP showed age by ApoE genotype interactions in the direction consistent with the known risk that carriers have for AD. Separate age and genotype effects were found for T-tau, no effects were found for the Aβ ratio or the individual Aβ peptide fragments studied. Because other work suggests that P-tau231 and IP are excellent univariate discriminators of AD (Brys et al., 2007), these biological data highlight the importance of increasing age and the ApoE ε4 genotype in modifying the risk for AD in normal subjects.
T-tau levels increased with age and were higher in ApoE ε4 carriers. The finding of positive association between age and T-tau is in agreement with other previously published data (Blomberg et al., 2001; Briani et al., 2002; Itoh et al., 2001; Shoji et al., 1998; Sjogren et al., 2001). However, there are also contrary findings (Andreasen et al., 1998; Blomberg et al., 1996; Burkhard et al., 2004; Jensen et al., 1995; Lewczuk et al., 2004a,b). However, in some of the published negative reports, the samples were much smaller than in our study. T-tau reflects the non-specific neuronal and axonal turnover and damage due to any pathological condition (for review (Blennow and Vanmechelen, 2003)). The increase in T-tau with age is not unexpected as age is related to multiple sources of neuronal damage and pruning (for review (Uylings and de Brabander, 2002)). We observed that over the lifespan, the ApoE ε4 positive group had higher overall T-tau levels. This result is in agreement with the report by Golombowski et al. (1997), who described this ApoE ε4 effect in group of three healthy ApoE ε4 carriers and nine non-carriers. However, two other studies showed negative results (Blomberg et al., 2001; Sunderland et al., 2004). The current study examined risk factors for atherosclerosis in the sample. We hypothesized that the E4 positive carriers might be predisposed to vascular diseases which in turn could contribute to age-related brain damage and therefore to elevated T-tau levels. Our ApoE groups were similar with respect to the distribution of medical (atherosclerosis and medication related) factors. It remains a possibility that our careful screening, which removed from the sample those individuals who had clinical evidence for brain damage due to vascular causes differentially reduced the prevalence associated with the ApoE genotype and diminished the age and interaction effects.
The P-tau231 concentration increased with age in both ApoE groups; and the increase was significantly steeper in the ApoE ε4 positive group. There is not much information about age effects on P-tau231 levels in normal controls, but our results are at variance with three other published reports, which did not show age-dependent changes either in P-tau181 (Lewczuk et al., 2004a,b), P-tau199 (Itoh et al., 2001) or P-tau231 (Buerger et al., 2002). Greater age-related P-tau231 elevations in the ε4 positive group are consistent with the increased risk of this group for AD and for an earlier accumulation of AD pathology. Several reasons led us to use P-tau231. (1) Neuropathological data shows that P-tau231 is a marker for the earliest stages of the disease. Augustinack et al. (2002) showed that TG3 antibody (for thr231 epitope) was one of three, which stained predominantly pre-neurofibrillary tangles stage while staining with the antibody specific for thr175/181 was more prominent in the later stage of NFT evolution. This distinction is potentially of great value while assessing first, disease- related signal in normal population (as our study group). (2) Antemortem CSF P-tau231 correlates with post mortem NFT pathology in neocortical regions (Buerger et al., 2006) and P-tau181 does not (Buerger et al., 2007; Engelborghs et al., 2007). (3) Excellent diagnostic specificity data exist for P-tau231 (Hampel et al., 2004).
Interestingly, across the lifespan and in relationship to ApoE genotype, the P-tau231 data parallels the results observed for IP. IP levels also showed steep age-related slope in ε4 carriers. This may reflect the possible link between oxidative stress and tau phosphorylation and is consistent with the hypothesis of an early role for oxidative damage in AD (Zhu et al., 2003). Examination of the relationship between these two analytes showed they were associated (r = .48, p < .01). We are not aware of other studies addressing the effect of ApoE genotype on CSF isoprostanes levels in healthy subjects. There is nonetheless, evidence of their higher levels in plasma in healthy ε4 carriers (Dietrich et al., 2005). Oxidative damage to peptidyl prolyl-cis, trans-isomerase (Pin), an enzyme regulating phosphorylation-dephosphorylation of tau has been suggested as an initial event prompting tangle formation (Sultana et al., 2006). Oxidative stress results in the elevation of stress-activated kinases (SAPK) (Zhu et al., 2003), which in turn are crucial for the phosphorylation of tau protein. The correlations between IP and P-tau levels in our sample, speaks in favor for the hypothesis linking oxidative stress and tau phosphorylation. Our data, however, does not address the issue of temporal and causative relationships between biomarkers; to fully answer this question longitudinal and animal model studies are needed. Nevertheless, the association between P-tau231 and IP suggest the early diagnostic potential for these biomarkers.
The reports about lifespan changes in the CSF Aβ levels of normal aging subjects are controversial. In a recent study, Peskind et al. (2006) found an age effect for Aβ42 and Aβ42/Aβ40. The authors showed a decrease in CSF Aβ42 after the age of 60, the decrease in Aβ42/Aβ40 occurred somewhat earlier. Significantly more pronounced age-related Aβ reductions were observed in subjects with an ApoE ε4 allele. However, others have not reported age effects for the Aβ ratio (Shoji et al., 1998). In our sample only a statistical trend for lower Aβ42/40 ratio with increasing age was found and we did not find an ApoE effect nor an interaction contrary to other reports (Peskind et al., 2006; Sunderland et al., 2004). We believe the discrepancy might have come from different assays used. Moreover, the Fagan et al. (2000) findings support ours, as she found significantly elevated Aβ40/Aβ42 ratios (inverse of Aβ42/Aβ40) between ε4 groups only in the lipoprotein fraction of CSF, but not in the unfractionated CSF (as was ours). Additionally, the possibility exists CSF levels may not be exclusively brain derived, as the platelets are the biggest source of the soluble amyloid in the blood (Chen et al., 1995). Thus, the interpretation of level changes may be problematic.
As different ApoE isoforms differentially affect lipid metabolism (Xhignesse et al., 1991), atherosclerosis (Greenow et al., 2005) and subsequent lipid peroxidation, it is possible that this influence accounts for our observed increase in the IP of ε4 positive subjects. Similarly, drugs may influence the levels of some biomarkers (De Caterina et al., 2002; Sutherland et al., 2007). This later issues are not likely to be a major confounding factors in our study, as the ε4 groups did not differ in variables associated with atherosclerosis, lipid metabolism, or the use of lipid modifying, antioxidant or other drugs. Although we did find a difference in BMI values, in neither group the means were in the obesity range. Similarly, there was a trend toward more favorable lipid profile in ε4 positive individuals, but this was limited only to triglycerides. We do not think the slightly higher number of subjects taking statins among the ε4 positive individuals, could be an explanation. Statins are known to have only negligible influence on triglyceride levels. None of our subjects were taking fibrates. However, the ε4 positive group was a little (but not significantly) younger, this could influence the triglyceride levels. Nevertheless, our analysis may be also compromised by the lack of information pertaining the types and dosages of statins and antioxidants.
The presence of an ApoE ε4 allele modifies the age of onset of AD (Khachaturian et al., 2004). Cognitively intact ApoE ε4 positive individuals display brain glucose metabolic patterns similar to those observed in AD patients (Reiman et al., 1996) and more recently it was reported that E4 carriers with subjective memory complaints showed elevated IP and P-tau231 levels in association with reduced medial temporal lobe glucose metabolism (Mosconi, 2007). Our CSF findings are in agreement with PET and other data indicating biological differences between ApoE ε4 positive and ε4 negative subjects that point to an AD predisposition or possibly to already existing preclinical pathology. Overall, this suggests a mechanism for their greater risk for cognitive decline and an early age of AD onset. However the observation of symptom development predicted by lower metabolism or altered biomarkers remains unexamined. The ApoE influence on CSF biomarkers may be exerted on different pathways: Lipid metabolism, neurite outgrowth, tau phosphorylation, mitochondrial alteration (Mahley et al., 2006) or oxidative stress (Ramassamy et al., 2000). It remains unclear whether the CSF differences between ε4 carriers and non-carriers are a reflection of increased sub-clinical disease or whether there is a life long trait of biomarker differences found in association with the APOE genotype. This important question can only be answered with longitudinal studies with post-mortem or other validations, assessing the impact of both genotype and biomarkers levels on progression rates from normal to MCI and to AD.
In summary, the present study examines the cross sectional effect of age (through the adult life-span) and ApoE genotype for five CSF analytes. The data show that CSF P-tau231 and IP are particularly sensitive to the interaction of age and genotype and therefore may be useful early predictors of the pathology leading to clinical recognition of AD.
This study was supported by: NIH-NIA AG12101, AG08051, AG022374, NIH-NCRR MO1RR0096, the American Health Assistance Foundation, and the Alzheimer’s Association.
Dr. Zinkowski owns stocks and stock options in Applied NeuroSolutions, Inc. The other authors declare no conflicts of interest.