|Home | About | Journals | Submit | Contact Us | Français|
Alzheimer’s disease (AD) affects millions of persons worldwide. Earlier detection and/or diagnosis of AD would permit earlier intervention, which conceivably could delay progression of this dementing disorder. In order to accomplish this goal, reliable and specific biomarkers are needed. Biomarkers are multidimensional and have the potential to aid in various facets of AD such as diagnostic prediction, assessment of disease stage, discrimination from normally cognitive controls as well as other forms of dementia, and therapeutic efficacy of AD drugs. To date, biomarker research has focused on plasma and cerebrospinal fluid (CSF), two bodily fluids believed to contain the richest source of biomarkers for AD. CSF is the fluid surrounding the central nervous system (CNS), and is the most indicative obtainable fluid of brain pathology. Blood plasma contains proteins that affect brain processes from the periphery, as well as proteins/peptides exported from the brain; this fluid would be ideal for biomarker discovery due to the ease and non-invasive process of sample collection. However, it seems reasonable that biomarker discovery will result in combinations of CSF, plasma, and other fluids such as urine, to serve the aforementioned purposes. This review focuses on proteins and peptides identified from CSF, plasma, and urine that may serve as biomarkers in AD.
Alzheimer’s disease (AD) is a neurodegenerative disorder that currently affects five million Americans. AD is the most common form of dementia among the elderly, as approximately 50% of adults aged 85 years develop this condition. Diagnosis of AD is based on National Institute of Neurological and Communicative Disorders and Alzheimer’s Disease and Related Disorders (NINCDS-ADRDA) Reagan criteria, neuropsychiatric testing, neuroimaging such as PET scanning/MRI, and an exclusion of other forms of dementia. AD is inherently difficult to diagnose, especially in very early stages when memory impairment is mild and may go unrecognized. Despite intense investigation into the mechanisms of neurodegeneration and possible causes of AD, currently no cure or definitive underlying cause exists. Therapeutics such as memantine, an NMDA receptor antagonist, and Aricept®, an acetlycholinesterase inhibitor, slow the rate of degeneration and are among the most widely prescribed medications to treat AD. These drugs are effective only briefly in the course of this disease and are usually administered at a time when the patient has exhibited a noticeable decline in cognition, and neurodegeneration has progressed to an irreversible state. Therefore, the need for a biomarker or set of biomarkers to diagnose AD earlier, distinguish AD from other forms of dementia, and monitor therapeutic efficacy is of critical need to aid and better protect the elderly from this devastating dementing disorder.
A biomarker is an abnormal signal from a bodily fluid or tissue that can provide distinguishable pathological information for a patient. According to the 1998 Consensus Report of the Working Group on Molecular and Biochemical Markers of Alzheimer Disease , ideal biomarkers for AD should be: 1) reflective or indicative of AD pathology; 2) reliable; 3) easy to perform/analyze; and 4) relatively inexpensive. Because biomarkers have the potential to assist in the diagnosis, delineate the disease stage, and monitor drug efficacy in AD, the notion of a single biomarker serving all of these functions is somewhat unrealistic. Rather, a combination of markers for diagnosing and managing AD and pre-clinical stages of AD, such as amnestic mild cognitive impairment (MCI), are continually being investigated. For AD and other neurodegenerative diseases, blood plasma and cerebrospinal fluid (CSF) have evolved as prime targets for biomarker investigation. This review focuses mainly on proteins and peptides in plasma and CSF as candidate biomarkers for AD, and briefly discusses advances in potential urinary biomarkers for AD.
Plasma is the liquid portion of blood that suspends cells such as red blood cells, white blood cells, and platelets. Plasma is pale-yellow in color, rather viscous, and can be isolated from whole blood by centrifugation at low speeds in the presence of an anticoagulant. Serum, which is oftentimes confused with plasma, is plasma with clotting factors, such as fibrinogen, removed. Serum is isolated by allowing blood to clot (therefore, no anticoagulant necessary) prior to centrifugation. During the coagulation process for serum preparation, the concentrations of some proteins in the sample are altered, and also, protein fragments are released by platelets and other cells . The plasma proteome, consequently, is altered and the opportunity for artifactual results is increased. Also, high fibrinogen levels have been shown to be associated with cognitive decline in patients with MCI  and increased risk of dementia . With this protein removed from the sample in serum preparation, a potential biomarker for AD may be overlooked. Therefore, depending on the investigation, plasma may be preferable for proteomic analysis/biomarker research instead of serum.
The ease with which blood, and therefore plasma, can be obtained from a human makes this fluid ideal for biomarker investigation. Human plasma is estimated to contain thousands of proteins and peptides that span a dynamic range in concentration of >1015 . Although an extremely rich source for disease-related biomarkers, the concentrations of the most abundant proteins, the “classical proteins,” make plasma a difficult sample to analyze. Albumin, the most abundant protein in plasma (albumin comprises greater than half of the total plasma protein concentration), is capable of masking lower abundance proteins that may be important as biomarkers. Depletion methods for high abundance proteins have spawned a plethora of commercially available dyes, spin cartridges, and high performance liquid chromatography (HPLC) columns that are capable of depleting approximately 20 of the most highly abundant proteins in plasma. Thus, depletion of high abundance proteins will increase the accessible dynamic range (Fig 1) and uncover lower abundance proteins that would otherwise be masked in a non-depleted sample (for example, see the 2D electrophoretogram in Fig 2).
In the last decade, numerous possible candidate biomarkers for AD have been reported. A considerable number of AD studies of plasma have exploited differences in the expression of individual proteins or peptides in AD patients compared to healthy controls. Some of these proteins have direct effects on brain/cognitive processes or are localized in brain, and therefore fulfill the requirement that an ideal biomarker be reflective of the pathology it represents. Also, because of the multidimensionality of a biomarker, peptides or proteins identified as having a characteristic difference in AD plasma may be useful in other ways. For example, proteins indicative of pathological conditions of AD may lead to new therapeutic strategies.
Numerous studies have identified plasma proteins whose expression levels in AD patients differ from controls. One proteomic study found alpha-2-macroglobulin (α2M) and complement factor H (CFH) to have increased expression in AD plasma compared to controls, and the increase for CFH was observed only for AD when compared with other forms of dementia . Both α2M and CFH have been shown to be present in senile plaques [7, 8], a hallmark of AD pathology. α2M is associated with damage to the blood brain barrier (BBB) , which may be evident in AD . Also, α2M has been proposed as a genetic polymorphism in AD [11, 12]
Alpha-1-antitrypsin (A1AT), a serine protease inhibitor, was found to be increased in AD plasma [13, 14] and serum  compared to controls and oxidized in its precursor form in AD . A1AT is localized in both senile plaques and neurofibrillary tangles (NFT) . The physiological role of A1AT is to suppress overexpressed proteases during inflammation; oxidation of this protein can lead to an alteration of its activity, and as a result, contribute to inflammatory processes observed in conditions such as AD .
Alpha-1-antichymotrypsin (A1ACT) has been reproducibly found to be increased in plasma/serum of AD patients compared to controls [19–25], although some groups report otherwise [26, 27]. A1ACT is a serine protease inhibitor synthesized by the liver. Elevated levels of this protein in the periphery may indicate the presence of systemic inflammation . Plasma and CSF A1ACT levels have been shown to correlate with disease severity . A1ACT is localized to senile plaques observed in AD brain . Interestingly, A1ACT differentially interacts with Aβ(1–40) and Aβ(1–42), which could reflect the greater neurotoxicity of the latter [31, 32]. A1ACT also induces tau phosphorylation in neurons, suggesting that this protein could contribute to hyperphosphorylation and subsequent development of tangles .
Apolipoprotein A1 (APOA1), the main component of high-density lipoproteins (HDL), has been reported to be lower in AD serum compared to controls [34, 35]. The main function of APOA1 is to bind to the ATP binding cassette transporter A1 (ABCA1) to promote cholesterol export from cells and tissues into blood. Because ABCA1 is also present at the BBB, it has been suggested that lower APOA1 levels in serum may contribute to the pathogenesis of AD, due to the involvement of cholesterol in APP processing . Moreover, though not proven yet, statins are proposed as a promising potential therapy to Aβ, which conceivably could be related to APOA1.
Although these and other proteins may reflect pathological processes observed in AD and have been identified as having characteristic differences in diseased plasma compared to controls, these differences have yet to achieve the diagnostic power, sensitivity, and reproducibility necessary for widespread use in a clinical setting. Reproducibility issues may arise in part due to different analytical methodologies between laboratories, including choice of anticoagulant and depletion strategy and storage among other considerations that may affect the plasma proteome (Kim and Kim 2007). Therefore, standardization using a valid universal protocol for sample handling and analytical technique is essential before assessing which proteins are, in fact, reproducible biomarkers of AD .
Amyloid-beta (Aβ) peptide is the central component of senile plaques present in AD brain and exists primarily as 40- and 42-amino acid peptides. Aβ can be detected via an enzyme linked immunosorbent assay (ELISA). Plasma Aβ levels have been shown to correlate with age ; however, reports of Aβ differences in plasma on the basis of AD diagnosis have found no significant differences . One possibility for the broad overlap in plasma Aβ measurements may lie in the analytical difficulties associated with this fluid. For example, Aβ is capable of binding to plasma proteins such as albumin ; in fact, one study showed that 95% of Aβ present in the blood is bound to plasma proteins, and therefore undetected by ELISA, probably due to epitope masking . ELISAs have been developed to detect total Aβ, irrespective of being bound or unbound; however, some ELISAs only measure “free Aβ,” which could possibly explain some group-to-group variation .
A recent study in 2007 examined intra- versus inter-person differences in plasma and serum Aβ levels . Results of this study show lower intra-person than inter-person variability in plasma/serum levels of Aβ. This suggests that, for measuring peripheral Aβ as a potential marker, comparative studies of the same subjects before and after diagnosis of MCI/AD may be more advantageous than one comparing diseased subjects to normal controls, which would serve to decrease between-group variability . Nevertheless, levels of Aβ have not yielded conclusive results in plasma to serve as a biomarker for differentiating between AD and healthy controls.
CSF is the “shock absorbing” fluid that is produced by the choroid plexus and bathes the space between the arachnoid and pia mater. Approximately 500 mL of CSF is produced by the brain daily, which is the same as its absorption rate into plasma . The main functions of CSF are to cushion the brain and spinal cord, and shuttle waste products from the central nervous system into the blood. CSF is obtained via lumbar puncture. Although lumbar puncture is invasive and potentially painful for the patient, CSF is probably the most informative obtainable fluid for neurodegenerative disease prognosis. For example, CSF has more physical contact with brain than blood, as it is not separated from brain by the tightly regulated BBB. As a result, proteins or peptides that may be directly reflective of brain specific AD pathology would be most likely to diffuse into CSF than any other bodily fluid.
Changes in brain chemistry may be reflected in CSF, since as noted above, CSF is in direct contact with the central nervous system . Analysis of CSF has produced some of the most reproducible biomarkers for AD, such as decreased Aβ42, increased total tau (t-tau), and increased phosphorylated tau (p-tau) [44–46]. Tau protein is involved in tubulin polymerization and microtubule stabilization, which helps maintain the cytoskeleton of neurons [47, 48]. Tau protein activity is regulated by phosphorylation and dephosphorylation by various kinases and phosphatases; disruption of these events can lead to hyperphosphorylation of tau, and assembly into neurofibrillary tangles (NFTs), one of the pathologic hallmarks of AD .
CSF t-tau is suggested to be a marker of neuronal death  due to the fact that tau is an intracellular protein, and can diffuse into CSF as a result of lysis of the neuron . Also, CSF levels of t-tau are highest in conditions known to exhibit severe neuronal damage, such as Creutzfeldt-Jakob disease (CJD) . Accordingly, investigators have reported increased CSF levels of t-tau in AD . Also, measurements using ELISAs targeting six different phosphorylated epitopes of tau have reported increases of p-tau in AD CSF compared to controls . Interestingly, levels of p-tau are not elevated in CJD, despite such a dramatic increase in levels of t-tau; this conceivably may be explained by the absence of NFTs in brains of most CJD patients , which are composed of hyperphosphorylated tau. CSF p-tau levels have been suggested to be indicative of the phosphorylation state of tau, and possibly reflect tangle burden in brain . In fact, levels of p-tau231 in CSF have been shown to correlate with NFTs and hyperphosphorylated tau load in several neocortical areas, namely frontal, temporal, and parietal lobes ; however, there are recent reports of no correlation between p-tau181 and neurofibrillary pathology [56, 57]. Therefore, the role that CSF tau plays in the pathology of AD requires more attention. Nevertheless, increased CSF t-tau and p-tau have been shown to be reproducible markers in differentiating AD from controls, and have potential for differentiating AD from other dementia forms as well (see below).
Generally, CSF levels of total Aβ were not found to significantly differ between AD and control patients [58–60]. As mentioned previously, Aβ42 levels in CSF are largely decreased in patients with AD . It was hypothesized that a reduction in CSF Aβ42 is a consequence of deposition of this peptide into plaques found in AD brain, with less diffusing into CSF . Strozyk et al. (2003) reported that low levels of Aβ42 correlated with high neuritic plaque load in both neocortex and hippocampus. Reductions in CSF Aβ levels that correlate with Aβ deposition in the brain have also been observed in the Tg2576 mouse model of AD . However, this issue is still controversial, as low levels of Aβ42 have also been found in dementias without senile plaques , and Engelborghs et al. (2007) showed no association between plaque burden and CSF Aβ42 levels in 50 autopsy-confirmed AD cases.
Combinations of some CSF markers have also been successful at discriminating AD from control cases. One study reported abnormally high CSF p-tau/Aβ42 ratio has high sensitivity and specificity for differentiating AD from normal controls [i.e., 86%/97% sensitivity/specificity], and from subjects with other non-AD dementias [i.e., 80%/73% sensitivity/specificity] . Also, several studies have investigated the usefulness of the Aβ42/t-tau CSF ratio in AD, which has been found to discriminate AD from controls with high sensitivity but variable specificity .
Proteomic studies in CSF have revealed proteins that show expression/oxidation differences in AD patients versus controls. For example, fibrinogen gamma A-chain precursor was shown to be increased in both AD and MCI patients compared to age matched controls, and expression was dependent on the severity of dementia . Decreases in the concentrations of λ-chain, β-trace, and transthyretin, as well as an increase in the carbonylation status of λ-chain in the CSF of patients with probable AD have been reported . Although differences in these and other individual proteins have been exploited in CSF in AD [for review, see ], greater attention has been given to tau and Aβ as potential biomarkers. However, Finehout et al. (2007) recently reported a panel of CSF proteins using 2D gel electrophoresis-based proteomics that distinguished between AD and non-AD with sensitivities and specificities of 94%; this panel consists of proteins involved in Aβ transport, inflammatory response, proteolytic inhibition, and in neuronal membrane, thus fulfilling criteria for a conclusive AD biomarker. A larger scale study will be needed to confirm the clinical usefulness of this panel in AD diagnosis.
AD pathology probably occurs 20–30 years prior to clinical onset . Patients diagnosed with amnestic MCI have a memory complaint despite no observable signs of dementia. A significant percentage of patients with memory complaints that do not meet clinical criteria of dementia are diagnosed as having MCI . However, MCI diagnosis does not imply AD is 100% imminent, as other forms of dementia, such as vascular dementia (VD), can be preceded by MCI . Also, some patients diagnosed with MCI revert back to a normal cognitive state. A set of biomarkers capable of predicting dementia prior to onset would have tremendous potential to improve the quality of life of patients, namely, earlier treatment administration that would serve to slow or possibly prevent dementia-related symptoms. Experimentally, studies designed to determine protein/peptide alterations for disease prediction and/or staging require long-term monitoring of subjects from a cognitively normal or slightly impaired state throughout the development of dementia.
Increased levels of CSF p-tau and t-tau, as well as decreased Aβ42, have been reproducibly found to discriminate AD patients from controls. While similar trends are also present in MCI patients compared to controls , these individual differences have generally not been capable of predicting AD in longitudinal studies, although Ewers et al. (2007) report CSF p-tau231 as a good predictor of AD in MCI patients. On the other hand, combinations of biomarker levels (along with some others) have proven to be effective in increasing sensitivity and specificity of predictive AD diagnosis. A 1999 study by Andreasen revealed that high CSF concentrations of t-tau and/or low concentrations of Aβ42 identified 14 of 16 MCI subjects who later developed AD . Another investigation supported the diagnostic capability of these markers, reporting 90% sensitivity and specificity of predicting probable AD in MCI patients 18 months after the initial diagnosis . However, because MCI patients convert to AD at a rate of about 15% per year , it is essential to investigate the diagnostic capabilities of these markers over a longer follow-up period . A more recent investigation reported increased baseline concentrations of CSF t-tau (>350 ng/L) and lower Aβ42 (<530 ng/L) predicted which MCI patients would later (i.e., after 4–6 years) convert to AD with 95% sensitivity and 83% specificity . t-tau levels combined with the Aβ42/p-tau181 ratio showed a slightly higher specificity for predicting MCI to AD conversions . A study of similar sample size and follow-up time reports low Aβ42/40 ratios predict MCI to AD conversions ; the Aβ42/40 ratio has been suggested to improve reliability of dementia-related diagnosis compared to raw concentrations of Aβ peptides . Another report showed high CSF tau/Aβ42 ratios in control subjects corresponded to increased frequency of the known AD risk factor, the APOE4 allele, as well as increased risk for conversion from healthy to MCI after a 42-month follow-up period . Levels of p-tau and Aβ42 at baseline were also capable of predicting cognitive decline in healthy elderly adults, which may be indicative of early AD signs .
As mentioned previously, plasma Aβ levels increase as a function of age . While total levels of plasma Aβ may seem rather insignificant diagnostically in AD, low plasma Aβ42/40 ratios have been associated with increased risk of developing MCI or AD . A case-cohort study in 2006 found that high Aβ40 and low Aβ42 in plasma resulted in a greater than ten-fold increase in risk of dementia . Very recently however, Hansson et al. (2008) reported plasma Aβ levels did not predict MCI to AD conversions in two sets of cohorts, instead referring to CSF levels of Aβ42 as a much better predictor of incipient AD.
Recently, Ray and colleagues proposed that, because the brain controls the release and regulation of signaling proteins in the periphery, AD pathology would have a disease-specific phenotype detectable in plasma . These researchers investigated the relative concentrations of 120 signaling proteins in plasma using sandwich ELISAs and report 18 proteins that were able to predict MCI to AD conversions, as well as distinguish between AD and cognitively normal patients with approximately 90% accuracy. Proteins involved in immune response, apoptosis, hematopoeisis, and neuronal support were identified as having characteristic concentration differences in plasma of subjects with AD and in subjects with MCI who later developed AD . It will be interesting to see in the future if these protein changes in plasma, possibly combined with predictive markers in CSF and brain imaging techniques, will provide accurate enough results to allow for a definitive clinical assessment of AD prior to onset of dementia.
One of the main criteria physicians use to diagnose AD is an exclusion of other forms of dementia. Biomarkers capable of fulfilling this criterion would greatly aid in the management and, possibly, prevention of this disease. Unfortunately, some of the individual CSF biomarkers capable of differentiating AD from normal aging are not highly capable of differentiating AD from other forms of dementia. For instance, while CSF levels of Aβ42 have been reproducibly found to be decreased in AD patients relative to controls, a similar decrease has been reported in dementia with Lewy bodies (DLB) [81, 82], frontotemporal lobar dementia (FTLD) , and amyotrophic lateral sclerosis (ALS) . Similarly, increased levels of CSF t-tau are evident not only in AD but also in conditions with neuronal degeneration such as CJD  and stroke , although levels of t-tau are so markedly increased in CJD that they were able to differentiate CJD from AD patients . On the other hand, p-tau in CSF has had some success as a discriminatory marker for AD. High CSF levels of p-tau181 have been shown to differentiate patients with AD from normal pressure hydroencephalus  and DLB . CSF p-tau231 levels discriminated between frontotemporal dementia (FTD) and AD subjects with 88% sensitivity and 92% specificity . This marker was also capable of differentiating between AD and geriatric depression, conditions which are difficult to distinguish clinically due to overlapping symptoms . CSF P-tau231 was significantly higher in AD compared to FTD, DLB, VD, other neurological disorders, and controls, and discriminated AD from FTD with sensitivities and specificities greater than 90% .
p-tau combined with Aβ in CSF has yielded some promising results for differentiation of AD from other neurological diseases as well. Studies have shown an increase in the CSF p-tau/Aβ42 ratio as a specific AD marker, including one that reported high degrees of differentiation from normal controls, patients with other dementias and patients with neurological disorders without dementia . This ratio was shown to distinguish AD subjects from FTLD, alcohol dementia, and patients with major depression . Consistently, a low CSF Aβ42/p-tau ratio differentiated AD subjects from controls and VD subjects with specificities and sensitivities greater than or equal to 85% , thus fulfilling the criteria for a reproducible biomarker; however, an attempt to validate the findings of this study reported low specificity using this ratio, which may limit differentiation between VD and AD .
Another group of candidate CSF markers that has received some attention for differentiation of AD from other dementia forms is that of neurofilament proteins. Three isoforms of neurofilament proteins exist, light, medium, and heavy chain, which together help form the axonal skeleton. Similar to CSF t-tau, high levels of CSF neurofilament proteins also are thought to indicate neuronal death/axonal degradation [94, 95]. Neurofilament protein levels in CSF have been shown to be increased in AD, but also in VD, FTD  and ALS . De Jong et al. (2007) reported neurofilament protein levels were increased in FTLD compared with early AD (EAD), and when combined with p-tau181 and Aβ42 levels, increased diagnostic accuracy (i.e., sensitivity 86%, specificity 100%). Although somewhat promising, more insight is needed to assess the discriminatory capacity of neurofilament proteins in CSF.
No one protein/peptide biomarker or set of markers in plasma has yet had the success of differentiating AD from other neurodegenerative diseases in contrast to that in CSF. Increased plasma CD40, a member of the tumor necrosis factor receptor superfamily, and a concomitant decrease in TGF-β1 levels were found in AD patients, and this trend was not observed in PD or non-AD dementia patients . De Servi et al. (2002) also observed decreased TGF-β1 in AD plasma compared to controls, although Rodriguez-Rodriguez et al. (2007) reported no difference in TGF-B1 in serum between AD and controls. Serum brain-derived neurotrophic factor (BDNF) was significantly higher in AD compared to VD and controls . Levels of soluble receptor for advanced glycation end products (sRAGE) were significantly higher in plasma from AD patients compared to VD as well . However, more discriminatory studies are needed to validate which plasma markers are specific only for AD.
The effects of AD and MCI therapeutics on plasma and CSF biomarkers, particularly Aβ, are thought to provide useful information for monitoring drug efficacy. Aβ is central to the pathogenesis of AD, as this peptide associates into oligomers and causes structural and functional damage via oxidative stress mechanisms . Thus, many potential therapeutics are designed with the intention of reducing the Aβ load in AD brain. Because the brain is in contact with both CSF and plasma, changes in the Aβ content in brain may be reflected in one or both of these compartments following pharmacological administration.
Aβ is produced by sequential cleavages of the amyloid precursor protein (APP) by two enzymes, beta- and gamma- secretase. Beta-secretase cleaves APP first at the N terminus, leaving a C terminal fragment (CTF) spanning the inner membrane of the lipid bilayer. γ-Secretase subsequently cleaves this CTF, releasing Aβ. Several therapeutics have been designed to reduce the Aβ load act by inhibiting γ-secretase. The first in vivo study of γ-secretase inhibitors reported decreased brain Aβ in APP transgenic mice 3 hours after oral administration . Despite showing potential as an AD therapeutic, the use of γ-secretase inhibitors is questionable due to safety issues, as inhibition of γ-secretase may have adverse health effects related to Notch signaling ; however, some γ-secretase inhibitors are well tolerated by patients in clinical trials and have potential as treatment for AD.
A few studies have examined plasma and CSF levels of Aβ after administration of γ-secretase inhibitors. Administration of the γ-secretase inhibitor LY450139 resulted in a 40% decrease in human plasma Aβ40 [103, 104]. Larger doses of this compound intensified this effect, with a maximum 72.6% reduction in Aβ40 using a 140 mg dose  that lasted for 12 h before returning to baseline. Decreases in plasma, CSF, and brain Aβ40 have been observed in several rodent species after administration of various γ-secretase inhibitors [106–109]. Whether the decreases in plasma Aβ40 after administration of a γ-secretase inhibitor is reflective of reduced neuronal Aβ is debatable. Brain-derived Aβ is thought to infiltrate plasma via the actions of two transporters at the blood brain barrier (BBB), low-density lipoprotein-related protein 1 (LRP1)  and p-glycoprotein , as well as bulk flow from CSF/interstitial fluid to plasma (Figure 3). Peripheral Aβ can also be transported into the brain via the receptor for advanced glycation end products (RAGE) . While still controversial, recent studies have reported that plasma Aβ levels do not reflect brain Aβ levels . On the other hand, plasma Aβ levels have been reported to decline along with CSF Aβ levels which correlates with Aβ deposition in the brains of Tg2576 mice . Other cells, such as platelets, also produce Aβ, and it would seem reasonable that γ-secretase inhibitors would affect this source as well. However, Grimwood et al. (2005) report decreased plasma, CSF, and cortex Aβ40 in guinea pigs after γ-secretase inhibition by compound E. A marked decrease in plasma Aβ40 was seen also at lower doses of compound E, suggesting that one potential mechanism for cortical Aβ40 reduction may result in part from a primary reduction in plasma Aβ40 , which could lead to less being transported into brain. Also, El Mouedden et al. (2006) reported reductions of both Aβ40 and Aβ42 in rat plasma, CSF, and brain extract in a dose- and time- dependent manner with an orally administered γ-secretase inhibitor. Therefore, it seems feasible that brain Aβ alterations resulting from γ-secretase inhibitors could be detected in plasma or CSF, though more investigation is necessary to decisively prove this point.
Another therapeutic strategy undergoing clinical trials for the treatment of AD is immunization against Aβ. Lower levels of anti-Aβ antibodies in AD patients versus controls led to the suggestion that immunoglobulin therapy could be a viable treatment option [115, 116]. Immunization can be active or passive; active immunity is the production of antibodies via stimulation by an antigen, while in passive immunity, preraised antibodies are directly administered to the subject. Active immunization studies reportedly improved cognition and decreased cerebral amyloid deposits in transgenic animals [117–119], but clinical trials in humans were halted after the development of meningoencephalitis in a number of patients [120, 121].
While lower levels of Aβ antibodies were observed in AD patients in some reports [115, 122], other studies have contradictory results [123, 124]. In plasma, a steady-state equilibrium dependent on antigen and antibody concentrations exists . Based on this phenomenon and the numerous variable reports on plasma Aβ levels in AD, conflicting reports on the levels of antibodies against Aβ are not surprising. As with Aβ ELISAs, epitope masking by the antigen (Aβ in this case) may also contribute to the relative concentration variability of this molecule in plasma/sera of AD patients. In this regard, a recent report by Gustaw et al. (2008) details a method of dissociating Aβ-antibody interactions for measurement of total titers in plasma or serum, therefore circumventing this potential pitfall. After dissociation, significant increases in Aβ antibodies were observed in plasma of AD patients compared to controls, whereby the same analysis on non-dissociated samples revealed no significant differences . Similar investigations in plasma and CSF using this methodology may further validate or refute immunotherapy as a treatment option for AD. Nevertheless, immunotherapy has been reported to have positive effects for AD patients and is briefly reviewed here.
A small-scale investigation into the levels of Aβ in CSF and plasma after passive Aβ immunization in mild AD patients revealed increases in plasma Aβ40 and Aβ42, decreases in CSF Aβ40 and 42, and a 2.5-point increase in mini mental state exam (MMSE) score after 6 months of therapy . Cessation of therapy resulted in CSF Aβ40 and Aβ42 increasing to pre-treatment baseline levels and a decrease in MMSE scores after 3 months, and a decrease in CSF Aβ was observed again when treatment was resumed after the three-month discontinuation. Dodel and colleagues (2004) reported increased total serum Aβ and decreased total CSF Aβ , as well as a slight general increase in MMSE scores as a result of passive immunoglobulin therapy in mild to moderate AD subjects. However, safety is also an issue with passive immunotherapy, as cerebral hemorrhaging has been reported with this treatment method . This side effect may be minimized by removal of the Fc portion of the antibody against Aβ . Nevertheless, if this treatment passes clinical trials, CSF levels of Aβ may be a potential biomarker to monitor intravenous immunoglobulin efficiency.
Urine is another bodily fluid that has received attention for biomarker investigation. This medium is less invasive to the patient than plasma and CSF. While urine would be an ideal source for biomarkers, analysis of this fluid is difficult due to very low protein concentrations, high salt levels, and high overall variability . Also, many of the proteins that are excreted into urine are of low molecular weight, thus rendering some proteomic strategies analytically inappropriate . Some published proteomic studies on urinary biomarkers for AD have included non-protein species, such as isoprostanes . Nevertheless, neural thread protein (NTP) levels have been reproducibly identified as an AD biomarker in urine, such that specific NTP assays to urine have been developed as a diagnostic tool [130, 131]. A simple biochemical assay called AlzheimAlert® from Nymox Pharmaceuticals measures levels of this protein in urine and has reported sensitivity and specificity levels of >80% and 90% for AD detection, respectively .
The NTP associated with AD, called AD7c-NTP, is substantially increased in AD CSF relative to controls, and levels of this protein in CSF seem to indicate the severity of dementia in early disease stages [132, 133]. This protein is overexpressed at the gene level in brains of AD patients, accumulates in neurons with disease progression, co-localizes with senile plaques found in AD brain, and impairs mitochondrial function and causes apoptosis in vitro . AD7c-NTP is detected by ELISA, and is increased in urine of patients with early AD [131, 133]. The urinary concentration of this protein also increases with disease severity consistent with analysis of AD7c-NTP in CSF [131, 133]. While the discriminatory power of this protein in urine has yet to be established, levels in CSF in patients with MS, multi-infarct dementia, CJD, PD, diffuse LBD, and Pick’s disease were similar to controls [132, 133], which may be evident in urine but has yet to be reported. Effects on the levels of AD7c-NTP in urine after therapeutic intervention also remain to be investigated.
It is feasible that certain lifestyle factors could inadvertedly affect biomarker levels, specifically diet, medication, and exercise. From a dietary standpoint, antioxidant-rich diets would seemingly quench in vivo markers of oxidation in AD, namely isoprostanes as well as oxidized proteins serving as markers in either plasma or CSF. Cholesterol effects on brain Aβ would presumably affect the brain-derived contribution of this peptide to CSF, and possibly plasma as well. CSF levels of Aβ42 were decreased in guinea pig after treatment with simvastatin . However in patients without dementia, treatment with simvastatin and pravastatin for 14 weeks did not significantly alter CSF levels of either Aβ40 or 42 or t-tau . Other studies also confirm steady Aβ and t-tau levels with statins [136, 137], although some reports differ with respect to p-tau [135, 136]. NMDA receptor antagonists prevent against excitotoxicity, and therefore, cell death; while not yet proven, these agents may in fact lower t-tau CSF levels, proposed markers of neuronal death. If this were true, t-tau levels in CSF may be an indicator of in vivo efficacy of drugs such as memantine. However, Degerman Gunnarsson et al. (2007) found significantly reduced CSF p-tau with memantine treatment, whereas t-tau and Aβ were unaffected. Because the significance of p-tau levels in CSF is still debatable, the implication of this decrease in CSF remains to be investigated. Meanwhile, cholinergic agents were the subject of a 2002 investigation that probed CSF levels of Aβ, t-tau, and p-tau and no significant differences in these biomarkers were observed in patients after treatment with donepezil, rivastigmine, or galantimine .
Physical activity is believed to play a preventative role in AD. Several groups have shown that increased physical (and intellectual) activity decreases risk of dementia [140, 141], although some studies do not support this idea . Also, investigations using mouse models of AD have shown cognitive improvements and/or decreased brain Aβ in mice exposed to exercise [143–146]. It has been reported that environmental enrichment has cognitive, behavioral, and biochemical benefits [147, 148]. Decreases in brain Aβ could have downstream consequences on the levels of this peptide in other bodily fluids, however the extent of this effect is not known at this time.
Biomarkers have many advantages for managing and predicting AD. As mentioned previously, it is estimated that pathology in AD begins roughly 20–30 years prior to diagnosis; early diagnosis would shift the dynamic of treatment from slowing neurodegeneration to potential prevention, so that the patient would be more likely to die of natural causes and perhaps other factors, rather than AD. Markers sought in plasma or urine would be advantageous due to the ease with which the sample is obtained; however, to date, it appears that more reproducible markers have been found in CSF. On the whole, specificities and sensitivities of diagnosis seem to increase using combinations of CSF biomarkers as opposed to absolute values of a single marker. Moreover, it is reasonable to believe that the outcome of these studies will result in combinations of plasma, CSF, and/or urine biomarkers that will serve diagnostic, staging, and therapeutic purposes in AD.
This work was supported in part by N.I.H. grants to D.A.B. [AG-10836; AG-05119; AG-029839].
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.