Diurnal Patterns of Soluble Amyloid Precursor Protein Metabolites in the Human Central Nervous System

The amyloid-β (Aβ) protein is diurnally regulated in both the cerebrospinal fluid and blood in healthy adults; circadian amplitudes decrease with aging and the presence of cerebral Aβ deposits. The cause of the Aβ diurnal pattern is poorly understood. One hypothesis is that the Amyloid Precursor Protein (APP) is diurnally regulated, leading to APP product diurnal patterns. APP in the central nervous system is processed either via the β-pathway (amyloidogenic), generating soluble APP-β (sAPPβ) and Aβ, or the α-pathway (non-amyloidogenic), releasing soluble APP-α (sAPPα). To elucidate the potential contributions of APP to the Aβ diurnal pattern and the balance of the α- and β- pathways in APP processing, we measured APP proteolytic products over 36 hours in human cerebrospinal fluid from cognitively normal and Alzheimer's disease participants. We found diurnal patterns in sAPPα, sAPPβ, Aβ40, and Aβ42, which diminish with increased age, that support the hypothesis that APP is diurnally regulated in the human central nervous system and thus results in Aβ diurnal patterns. We also found that the four APP metabolites were positively correlated in all participants without cerebral Aβ deposits. This positive correlation suggests that the α- and β- APP pathways are non-competitive under normal physiologic conditions where APP availability may be the limiting factor that determines sAPPα and sAPPβ production. However, in participants with cerebral Aβ deposits, there was no correlation of Aβ to sAPP metabolites, suggesting that normal physiologic regulation of cerebrospinal fluid Aβ is impaired in the presence of amyloidosis. Lastly, we found that the ratio of sAPPβ to sAPPα was significantly higher in participants with cerebral Aβ deposits versus those without deposits. Therefore, the sAPPβ to sAPPα ratio may be a useful biomarker for cerebral amyloidosis.


Introduction
Alzheimer's disease (AD) is the most common neurodegenerative disorder, affecting an estimated 30 million people worldwide [1]. Although the pathophysiology of this disease is incompletely understood, the study of brain and cerebrospinal fluid (CSF) proteins, such as amyloid-b (Ab) and tau, have provided insight into AD molecular pathophysiology [2][3][4][5][6]. The study of Ab production, transport, and clearance is important for insight into normal brain protein handling and also for the pathophysiology of AD.
The first studies of Ab concentrations over time indicated that CSF concentrations were sinusoidal over 24 hours in younger healthy participants [7] and suggested a possible circadian pattern. Subsequent studies in humans and animal models [8] demonstrated Ab concentrations in the brain could be regulated by sleep/wake cycles and orexin. We reported that Ab exhibits a diurnal pattern in both CSF [9] and blood [10] in healthy adults. The diurnal patterns, as determined by circadian amplitude, decreased with aging and amyloidosis. The immediate mechanism for diurnal regulation of Ab has not been previously described, and possible causes for the Ab diurnal pattern include, but are not limited to, diurnal regulation of Amyloid Precursor Protein (APP) transcription, translation, or transport, or diurnal regulation affecting the two secretases (b-secretase or c-secretase) that cleave APP to produce Ab. In this study, we evaluated the temporal relationship of Ab with other proteolytic products of APP to inform about the cause of Ab diurnal patterns in the CNS of healthy young and elderly humans, as well as those with amyloid pathology.
Amyloid precursor protein is a single-pass transmembrane protein processed through at least two pathways in the CNS: the b-(amyloidogenic) pathway and the a-(non-amyloidogenic) pathway [11]. This protein is cleaved in the amyloidogenic pathway by b-secretase releasing a soluble extracellular fragment called soluble APPb (sAPPb) [12][13][14]. The APP endodomain, Cterminal fragment 99 (CTF99), which remains in the transmembrane, is subsequently cleaved by c-secretase, resulting in the generation of Ab and the APP Intra-Cellular Domain (AICD). The non-amyloidogenic processing of APP occurs when asecretase cleaves APP, producing soluble APPa (sAPPa). The endodomain of APP (CTF83) may then be cleaved by c-secretase, resulting in the release of a fragment, p3. The formation of Ab is precluded by a-secretase cleavage.
To further elucidate the potential contributions of APP to the Ab diurnal pattern and the balance of the aand bpathways in APP processing, we measured APP proteolytic products sAPPb, sAPPa, Ab 40 , and Ab 42 over 36 hours in CSF from cognitively normal young and elderly participants, as well as in CSF from participants with AD.

Ethics statement
All human studies were approved by the Washington University Human Studies Committee and the General Clinical Research Center (GCRC) Advisory Committee. Written, informed consent was obtained from all participants prior to their enrollment in this study.

Study design
Participants were recruited from the general public or through Washington University's Charles F. and Joanne Knight Alzheimer's Disease Research Center (Knight ADRC). All participants were in good general health. These participants were divided into three groups by age and brain amyloid status: 1) an Amyloid+ group of participants greater than 60 years of age and with probable amyloid plaques in the brain. Amyloid plaque status was determined by positron emission tomography using Pittsburgh compound B (PET PiB) or determined by an Ab 42 CSF mean concentration less than 350 pg/mL; 2) an Amyloid2 age-matched group with no probable amyloid plaques in the brain as measured by PET PiB or determined by an Ab 42 CSF mean concentration greater than 350 pg/mL; 3) a Young Normal Control (YNC) group (18-50 years of age) that are likely PET PiB- [15]. PiB binds to fibrillar amyloid plaques in the brain [16]. A mean cortical binding potential (MCBP) was calculated for each participant to determine PET PiB (Amyloid) ''+'' or ''2'' status [15]. To measure the MCBP, binding potentials of PiB were averaged from specific brain regions: prefrontal cortex, precuneus, lateral temporal cortex, and gyrus rectus. MCBP scores of 0.18 or greater were designated as amyloid plaque positive (Amyloid+), while those less than 0.18 were designated as amyloid plaque negative (Amyloid2) [15]. Some participants did not have reported MCBP values, and, in those cases, a surrogate marker of amyloid deposition was used to assign the participant group. This surrogate marker was a low CSF Ab 42 concentration which has been shown to be inversely correlated with PET PiB measurements [17]. A CSF Ab 42 concentration was considered low (and the participant classified as Amyloid+) if it was detected as less than 350 pg/mL from an Ab 42 ELISA that used 21F12 (anti-Ab 42 ) as the capture antibody and biotinylated 3D6 antibody (directed against Ab 1-5 ) as the detection antibody.

Demographics of study participants
A total of 49 participants (both men and women) were assessed in at least one part of this study. Specific sample size in each group varied depending on the experiment, and sample size for each group when diurnal patterns were observed is listed in the cosinor analyses section of the Methods. For the part of this study where APP metabolites were measured in a single CSF time point, there were 15 participants in the YNC group, 15 in Amyloid2, and 18 in Amyloid+.
The mean (SD) age for each participant group when all 49 participants were taken into account: YNC = 37.11 (68.71) years; Amyloid2 = 69.6 (64.5) years; and Amyloid+: 76.3 (67.5) years. Clinical Dementia Rating (CDR) at study onset was available for all participants. Of the Amyloid2 participants, 33.3% had a CDR score greater than zero (exhibited cognitive deficits). Of the Amyloid+ participants, 29.4% had a CDR score equal to zero. All YNC subjects were free from any cognitive deficits.

Sample collection and storage
Sample collection and handling were done as previously described [18]. Briefly, for all participants an intrathecal lumbar catheter was placed between the L3 and L4 interspace or the L4 and L5 interspace between 7:30 A.M. and 9:00 A.M. Collection of CSF began between 8:00 A.M. and 9:30 A.M. Every hour for 36 hours, 6 mL of CSF and 12 mL of plasma were withdrawn. Aliquots of CSF (1 mL) were immediately frozen at 280uC in Axygen maximum-recovery polypropylene tubes.

Sample and standard handling
Aliquots (1 mL) from even hours with two freeze-thaw cycles were measured by sAPPa and sAPPb ELISA. The effect of two freeze-thaw cycles was determined to not significantly change sAPPa and sAPPb concentrations. Before plating, CSF samples were diluted in phosphate buffered saline-0.05% Tween20 (PBS-T) 75-to 150-fold for sAPPa, and 10-to 25-fold for sAPPb. Recombinant standards from E.coli were used for both sAPPa (Sigma-Aldrich; St. Louis, MO) and sAPPb (Sigma-Aldrich; St. Louis, MO). The concentration of the standards ranged from 1.6-75 ng/mL for sAPPa and 2.7-125 ng/mL for sAPPb. Single freeze-thaw CSF aliquots from both odd and even hours were thawed on ice for the Ab 40 and Ab 42 ELISAs. They were diluted in a final buffer consisting of 2 mg/mL BSA (bovine serum albumin (Sigma-Aldrich; St. Louis, MO))-PBS-T, 3 M Tris, 10% Azide, 16 protease inhibitor cocktail. Each CSF and standard sample was assessed in triplicate.

sAPPa ELISA protocol
For the sAPPa ELISAs, 96-well Nunc MaxiSorp flat bottom ELISA plates (eBiosciences, Inc.; San Diego, CA) were coated with 100 mL per well of 5 mg/mL of 8E5 (a monoclonal antibody raised to a bacterially expressed fusion protein corresponding to human APP 444-592 of the APP 770 transcript [19], courtesy of Eli Lilly). Plates were incubated for 24 hours on a shaker at 4uC, and then blocked with 3% dry milk in PBS-T for 1 hour 20 minutes at 37uC. To avoid plate position effects, samples were randomly assigned to a well on the plate. Secondary (detection) antibody (50 mL of 1:10,000 6E10 [20], a monoclonal antibody reactive to Ab 1-16 , otherwise known as APP 672-687 (in the APP 770 transcript), and having the epitope at Ab 3-8 , or APP 674-679 ) (Signet Covance; Dedham, MA) was added to each well. Samples and secondary antibody were incubated on a shaker at 4uC for 24 hours. Plates were washed 5 times with PBS-T and then Streptavidin Poly-HRP20 (Fitzgerald Industries International; Acton, MA), diluted at 1:15,000 in 1% BSA-PBS-T, was added to each well at 100 mL/ well. Plates were incubated in the dark for 1 hour at 37uC on a shaker. Plates were then washed 5 times with PBS-T and 5 times with PBS. The plates were developed as described for the sAPPb ELISA below.
To test the specificity of the sAPPa assay, we ran a titration curve of sAPPa and sAPPb protein standards on the same ELISA. The results demonstrated that this assay was specific for sAPPa and there was no detectable cross-reactivity with sAPPb, as even the highest sAPPb standard (300 ng/mL) did not produce an OD value above zero ( Figure S1). The diluted CSF OD values fell within a linear range of the sAPPa standard curve sAPPb ELISA protocol For the sAPPb ELISA, 96-well Nunc MaxiSorp flat bottom ELISA plates (eBiosciences, Inc.; San Diego, CA) were coated with 100 mL per well of 10 mg/mL of the monoclonal antibody, 8E5. Plates were incubated for 24 hours on a shaker at 4uC and subsequently blocked with 3% dry milk in PBS-T for 1 hour 20 minutes at 37uC. Samples were randomly assigned a plate well position and incubated for 24 hours on a shaker at 4uC. They were then washed 5 times with PBS-T. An antibody against the neo-epitope of sAPPb (APP 670/671 of the APP 770 transcript) (courtesy of Eli Lilly) was used as the secondary (detection) antibody at a volume of 50 mL and a concentration of 0.5 mg/mL, diluted in PBS-T pre-warmed to 37uC. The sAPPb detection antibody was added to each well and incubated at 37uC for 90 minutes. Plates were washed 10 times with PBS-T, and 100 mL Streptavidin Poly-HRP40 (Fitzgerald Industries International; Acton, MA), diluted at 1:20,000 in 1% BSA-PBS-T, was added to each well. Plates were incubated in the dark for 1 hour at 25uC on a shaker and washed 5 times with PBS-T and 5 times with PBS. For the sAPPa and sAPPb ELISAs, 100 mL/well of ELISA TMB Super Slow (Sigma-Aldrich; St. Louis, MO), pre-warmed to 25uC, was then added to each well. Optical density (OD) was measured at 650 nm using a Biotek Synergy 2 plate reader after 5-30 minutes.
We tested the specificity of the sAPPb assay by running a titration curve of the sAPPb and sAPPa protein standards on the same ELISA. The results demonstrated that this assay was specific for sAPPb and that cross-reactivity with sAPPa was negligible. The OD value for the sAPPb standard of 8.5 ng/mL was approximately the same as that for the sAPPa standard of 300 ng/ mL ( Figure S2). This indicated that this ELISA was approximately 35-fold more selective for sAPPb than for sAPPa. The diluted CSF OD values fell within a linear range of the sAPPb standard curve and well above the highest sAPPa standard's (300 ng/mL) OD value. Given that in biological samples sAPPa and sAPPb were nearly equal in molar concentrations, this minimal cross-reactivity of sAPPa in the sAPPb ELISA was negligible. Thus, we concluded that any fluctuations we observed in sAPPb levels using this ELISA were attributed solely to sAPPb, and not to sAPPa.

Ab 40 and Ab 42 ELISA protocols
Corning 96-well half area clear flat bottom polystyrene high bind ELISA plates (Corning Life Sciences, Tewksbury, MA) were coated with 1.25 mg/mL HJ7.4 (Ab 37-42 ) or 2.5 mg/mL HJ2 (Ab [33][34][35][36][37][38][39][40] ) in PBS plus 20% glycerol (PBS-G), then incubated 1 hour at 25uC followed by overnight incubation at 4uC. The next day the plates were blocked with 2% BSA-PBS-T for 90 minutes at 4uC. Samples were randomly assigned a well on the plate. Diluted CSF samples and standards were pipetted at a volume of 50 mL per well onto freshly washed plates. The samples were loaded in triplicate and incubated overnight at 4uC. After incubation and washing, the plates were incubated for 90 minutes at 25uC with 0.2 mg/mL HJ5.1-Biotin (Ab 13-28 ) in 1% BSA-PBS-T-G. Plates were then washed three times with 190 mL PBS-T, followed by incubation in Streptavidin Poly-HRP40 (Fitzgerald Industries International; Acton, MA), diluted at 1:12,000 in 1% BSA-PBS-T-G, for 90 minutes at 25uC. Plates were subsequently washed three times with 190 mL PBS-T. They were then incubated with 50 mL/well of Slow ELISA TMB (pre-warmed to 25uC) for 5-30 minutes. Optical density (OD) was read at 650 nm using a Biotek Synergy 2 plate reader.

CSF protein level quantification
Soluble APPa, sAPPb, Ab 40 , and Ab 42 concentration levels were quantified using the Biotek Gen5 software (version #1.08.4) based on the non-linear five parametric standard curves generated from the recombinant sAPPa, sAPPb, Ab 40 , and Ab 42 standards.
The OD values of the CSF samples fell within the linear range of the standard curve and were converted to concentration levels. The product of the concentration and the dilution factor was calculated in order to determine the final CSF concentration of each protein.
Total protein levels of each sample were measured by BCA assay (Thermo Fisher Scientific, Inc.; Rockford, IL) as previously reported [9]. The intra-sample coefficient of variation mean was 2% for duplicates.

Group-averaged cosinor analyses
Serial sAPPa and sAPPb concentrations were binned in two hour increments as samples were from every other hour. Serial Ab 40 and Ab 42 concentrations were left unbinned because hourly concentrations were measured. For each APP metabolite, each participant's hourly metabolite's concentration was normalized to that metabolite's mean concentration over 36 hours. The normalized value was calculated as a percentage of each participant's mean (1006value/mean). Hourly (Ab 40 and Ab 42 ) and bi-hourly (sAPPa and sAPPb) concentrations of each metabolite were averaged among all participants in each participant group to produce normalized mean 36 hour concentrations. Next, the linear concentration rise over time that was observed in each metabolite was subtracted out of the mean concentrations and a single cosinor fit was applied for each metabolite as described previously [9]. Briefly, a cosine transformation was applied to the time variable using 24 hours as the default circadian cycle, and Graphpad Prism version 5.01 for Windows (GraphPad Software; San Diego, CA) was used to estimate the parameters of the circadian rhythms for each metabolite. The amplitude (distance between the peak to the midline of the cosine wave) was determined for each participant group. For all cosinor analyses, the YNC group consisted of 13 participants. The Amyloid2 group included 19 participants for sAPPa and sAPPb cosinor analyses, and 15 participants for Ab 40 and Ab 42 cosinor analyses. The Amyloid+ group had 17 participants for sAPPa and sAPPb cosinor analyses, and 14 participants for Ab 40 and Ab 42 cosinor analyses.

Individual cosinor analyses
For each participant, sAPPa, sAPPb, Ab 40 , and Ab 42 levels over 36 hours were analyzed using a single cosinor analysis as described above. Mesor (midline of the metabolite oscillation), amplitude (distance between the peak and mesor), amplitude-to-mesor ratio, and acrophase (time at which the peak occurs) were calculated for each metabolite for each participant. Then, participant group means for each of the metabolites' cosinor parameters were determined. Group sample size for these analyses was the same as for the group-averaged cosinor analyses.

Statistical analyses
Analyses were performed using Microsoft Office Excel 2007 and GraphPad Prism version 5 for Windows (GraphPad Software, San Diego, California, USA). Student's t-tests and ANOVAs were used to determine whether there were differences in cosinor parameters between groups. 95% confidence intervals were reported. Correlations between APP metabolites were measured by calculating the correlation coefficient (Pearson r values reported). Soluble APPa, sAPPb, and sAPPb/a ratio were compared among groups using a student's t-test and ANOVA. 95% confidence intervals were reported.

Circadian patterns of APP metabolites
In order to determine APP processing over time within the same participant, temporal CSF samples from a particular participant were randomly assigned a well position on four sandwich ELISAs: specific for sAPPa, sAPPb, Ab 40 , or Ab 42 . This allowed for analysis of APP metabolite concentrations in the CSF over time.
To compare age and amyloid deposition effects on hourly dynamics of APP metabolites, the Young Normal Control (YNC) group was compared to the Amyloid2 and Amyloid+ groups.

sAPPa and sAPPb exhibit circadian patterns
Cerebrospinal fluid sAPPa and sAPPb hourly concentrations had significant fits to a 24 hour cosinor pattern in the YNC group. The average amplitude of the diurnal pattern for sAPPa was 2.9%61.3% (SEM) ( Figure 1A). For sAPPb, the average amplitude was 4.4%61.6% (SEM) ( Figure 1D).
Group-averaged sAPPa and sAPPb circadian amplitudes lower with older age When a 24 hour cosine curve was fit to the three groupaveraged sAPPa hourly concentrations, the YNC group exhibited an amplitude that significantly deviated from zero (2.9%) and was significantly greater than the Amyloid2 (0.9%) and Amyloid+ (2%) groups, which both did not deviate significantly from zero ( Figure 1A-C). A similar trend was observed when a cosine curve was fit to the three group-averaged sAPPb hourly concentrations ( Figure 1D-F). Amplitude of sAPPb for the YNC group was 4.4%, Amyloid2 was 1.2%, and Amyloid+ was 2%. Only the sAPPb amplitude of the YNC group significantly deviated from zero. Amplitude of Ab 40 for the YNC group was 0.9%, Amyloid2 was 3.2%, and Amyloid+ was 2.6% ( Figure 1G-I). Only the Ab 40 amplitude of the Amyloid2 group significantly deviated from zero. Amplitude of Ab 42 for the YNC group was 2.9%, Amyloid2 was 3.8%, and Amyloid+ was 0.4% ( Figure 1J-L). Only the Ab 42 amplitude of the YNC group significantly deviated from zero.

Individual acrophases are not significantly different with age or amyloidosis
There is much inter-subject variability within groups for each metabolite's acrophase. Thus, any differences in time at peak/ trough among participant groups are not significantly different. Data are provided in Tables 1-4. In the case of all four metabolites, differences among average acrophase of participant groups never reached statistical significance (p.0.05). Differences among metabolites' group-averaged acrophases were not evaluated because when no significant cosinor fit is found (as in Figure 1B, C, E, F, G, I, K, L), the acrophase is not a valid parameter to compare groups.

No diurnal pattern exhibited in total protein levels of Amyloid2 and Amyloid+ groups
As a negative control for diurnal rhythms, we assayed total CSF protein over 36 hours using a micro BCA assay. Total protein data was only available for a subset of participants in each group. We measured that, on average, total protein concentrations were significantly lower in YNC as compared with the older participants (YNC = 797.2 mg/mL (n = 6), Amyloid2 = 895.1 mg/mL (n = 6), and Amyloid+ = 871.4 mg/mL (n = 5), ***p,0.0001). A cosinor fit was applied to the mean of each group's total protein level. A significant cosinor fit was found in the YNC group, with an amplitude 4.5% (95% CI: 26.1% to 22.9%). Cosinor fits for both older groups were insignificant because the amplitudes' 95% CIs crossed zero: Amyloid2 (95% CI: 21.4% to +8.6%) and Amyloid+ (95% CI: 28.4% to +1.4%) ( Figure S3). Acrophase was calculated only for the YNC (1.160.7 h), as the other groups did not exhibit a significant cosinor fit. Owing to high inter-subject variability within the YNC group and approximately only 46% of participants having BCA data for analysis, we cannot conclude that a significant cosinor fit in the YNC group would hold up with a full dataset.

sAPP and Ab positively correlated, except in amyloidosis
In order to determine the relationship of aand b-secretases on APP processing, correlations of sAPPa, sAPPb, Ab 40 , and Ab 42 were calculated in CSF from a single time-point at the onset of the study (between 7:30 A.M. and 9:00 A.M.) in the three participant groups: YNC, Amyloid2, and Amyloid+. Soluble APPa and sAPPb were positively correlated in all groups (YNC: r = 0.95, significantly highest in YNC when compared to Amyloid2 (*p = 0.04) and Amyloid+ (***p,0.0001). The Amyloid2 group also had a significantly higher Ab 42 amplitude than the Amyloid+ group (***p = 0.0008). H) The Ab 42 amplitude-to-mesor ratios did not differ significantly among groups. doi:10.1371/journal.pone.0089998.g002

sAPPb/sAPPa ratio is elevated in amyloidosis
In order to determine the effects of age and amyloidosis on the APP processing pathways, APP metabolites from a single CSF time-point at the onset of the study (between 8:00 A.M. and 10:00 A.M.) were compared among three participant groups: YNC, Amyloid2, and Amyloid+. The sAPPb to sAPPa ratio was 0.2660.01 (<1:3 ratio, n = 15) in YNC, and 0.2660.02 (<1:3 ratio, n = 15) in Amyloid2. However, the ratio increased to 0.3260.05 (<1:2 ratio, n = 10) for Amyloid+. The sAPPb/sAPPa ratio was significantly higher in Amyloid+ participants than in Amyloid2 (*p = 0.02) and YNC (**p = 0.002) ( Figure 4A). However, taken independently, mean sAPPa and sAPPb concentrations were not significantly different among groups, suggesting that the sAPPb/sAPPa ratio corrected for other variances which were not associated with amyloidosis ( Figure 4B-C).
In order to determine if there was a similar pattern in sAPPb/ sAPPa ratio differences among groups when measurements were taken over a full 36 hour time-course (versus at a single time-point: hour 0), for each participant sAPPb and sAPPa concentrations were individually averaged over 36 hours. Each participant's 36 hour averaged sAPPb concentration and their respective 36 hour averaged sAPPa concentration were then used to determine the mean sAPPb/sAPPa ratio. These mean ratios were then, in turn, averaged to determine a participant group average of the mean sAPPb/sAPPa ratio. The mean sAPPb to sAPPa ratio was 0.5960.04 (n = 15) in YNC, which was significantly higher (*p = 0.03) than either the Amyloid2 (n = 19) or the Amyloid+ (n = 17) ratio (both ratios were 0.4260.06) ( Figure S4A).
Additionally, each participant's sAPPb mesor and sAPPa mesor were used to determine individual mesor sAPPb/sAPPa ratios. The mesor sAPPb to sAPPa ratio was 0.5960.04 (n = 15) in YNC, which was significantly higher than the Amyloid2 and Amyloid+ mesor ratios. Mesor ratio means and error for the two older groups were identical to averaged ratios and errors ( Figure S4B).
The results from the mean sAPPb to sAPPa ratio and the mesor sAPPb to sAPPa ratio are almost identical because they represent nearly the same parameter. These results also contrast with the increased sAPPb to sAPPa ratio with amyloidosis when only the first CSF sample collected (hour 0) is analyzed. The mean concentrations and the mesor are calculated from runs on multiple ELISA plates over many months and may not be directly comparable, while the hour 0 samples were run on the same plate and can be directly compared. Thus, we conclude the increased sAPPb to sAPPa ratio in amyloidosis when measuring at hour 0 is most reliable as it avoids assay drift and also the modeling of the calculated mesor value.

Discussion
We evaluated whether APP exhibited diurnal fluctuations similar to that of Ab, which would help inform why Ab demonstrates a diurnal pattern. We also determined normal aand b-processing of APP in the human CNS and assessed whether AD pathology is associated with alterations in APP processing. The regulation of APP by aand b-secretase over time, including potential dynamic changes of sAPPa and sAPPb within an individual, has not been previously evaluated, although Ab diurnal activity has been described in healthy, young human participants [7]. We recently demonstrated that both in CSF [9] and in plasma [10], the physiological Ab diurnal fluctuation described in young participants diminishes significantly with increasing age, but is not further decreased in amyloidosis. Further, previous studies in mice indicated that sleep regulation may play a critical role in the risk and development of AD [8], but more recent findings indicate that it may be Ab aggregation that disrupts both the sleep-wake cycle and Ab diurnal fluctuation [21]. For example, longitudinal studies have found a strong relationship between sleep circadian patterns, as well as sleep disordered breathing and risk of mild cognitive impairment and AD [22][23]. Therefore, we sought to determine the relationship between aand bprocessing pathways in individuals over time, and also determine if APP regulation contributes to Ab circadian patterns.
In the YNC group, we found that sAPPa, sAPPb, Ab 40 , and Ab 42 concentrations were dynamic over 36 hours, with diurnal patterns. The lowest concentrations were in the morning (approximately 9:00 A.M.), and the concentrations peaked in the evening, approximately12 hours later. This suggests that in the YNC group, dynamic changes in these protein levels were due to dynamic changes in APP availability, whether by its production (transcription or translation) or transport to the site of processing (i.e. axonal transport). Amyloid-b also demonstrated a diurnal pattern with a peak and trough approximately three hours after sAPPa and sAPPb. This suggests that APP diurnal availability likely plays a role in Ab diurnal patterns.
Diurnal patterns of sAPPa and sAPPb were diminished in the Amyloid2 group. Ab 42 did not show any significant diurnal pattern in the Amyloid2 group similarly to prior work from our laboratory [9]. However, whereas our present work did not show a diurnal pattern of Ab 40 in the Amyloid2 group, there was a slight, but significant diurnal pattern observed in [9]. Potential reasons for this discrepancy include different ELISA assays employed for the different studies. Both Ab ELISA assays from [9] used 3D6 as detection antibodies, and capture antibodies were 2G3 (anti-Ab 40 ) and 21F12 (anti-Ab 42 ). These are fairly common Ab antibodies, and those assays provided lower intra-sample CV of duplicates than the antibodies we used for Ab in this study. More noisy data may have contributed to slightly differing results. Further, although several of our participants in the two studies overlapped, many participants were not from the same dataset as [9]. Lastly, [9] had more variable sample size among groups (YNC = 20, Amyloid2 = 15. Amyloid+ = 11), whereas our groups were more balanced (YNC = 13; Amyloid2 = 15; Amyloid+ = 14). Taken together, these findings indicate that with age there is a loss of APP dynamics or availability, which results in the noted loss of not only sAPP, but also Ab diurnal patterns. It was recently reported that sleep facilitates Ab clearance [24], thus the physiological tightly-regulated diurnal patterns of Ab may diminish with age due to an increase in sleep fragmentation that is common in normal aging [25] or by a general increased Ab production due to wakefulness [8].
Lack of a diurnal pattern of sAPPa and sAPPb was exhibited to a similar extent in the Amyloid+ group as was seen in the Amyloid2 group. However, the diurnal patterns in Ab 40 and Ab 42 were even more significantly diminished in the Amyloid+ group than was seen in the Amyloid2 group. The further marked decrease in Ab 40 and Ab 42 diurnal patterns in the presence of amyloidosis does not correspond to any decrease in sAPP diurnal patterns. This disconnect may be an effect of downstream APP cleavage events and not due to APP dynamics or availability, which seems to be the case in general aging. Potentially the extent of c-secretase cleavage of APP, which is controlled by availability of the c-secretase components or the c-secretase level of activity, may play a role in diminishing the diurnal patterns of the two Ab species we measured. Also, the build-up of Ab plaques in the brains of those with amyloidosis may serve as a buffering system that decreases the dynamic nature of Ab that is observed in healthy, younger humans. Although, the Amyloid+ group has a lower Ab 42 amplitude than YNC or Amyloid2, this result is not intended to suggest that Ab 42 amplitude should be added as an Alzheimer's diagnostic test. Currently, other tests (a combination of CSF Ab 42 /tau, PIB PET, and FDG PET scanning) have good predictive outcomes for determining AD diagnosis. The potential minor additive diagnostic benefit of Ab 42 amplitude is questionable and would require a patient to be catheterized for 24 hours.
Further, sAPPa and sAPPb were positively correlated in all groups. Positive correlation of the aand b-secretase products suggests a non-competitive model of APP pathways: that the total APP availability drove changes in sAPPa and sAPPb. Soluble APPa and sAPPb were positively correlated with both Ab species in YNC and elderly controls. However, the correlation between the sAPP species and Ab 42 was lost with amyloidosis. Prior evidence in the human CNS shows a positive sAPPa to sAPPb correlation in individuals also suggesting non-competitive aand b-pathways [26][27][28][29] However, in vitro studies of secretase inhibitors or activators, or genetically decreasing BACE1 (a b-secretase protein) or ADAM10 (an a-secretase protein) [30][31][32][33][34][35][36] support the hypothesis that aand b-secretase pathways compete for the same APP pool due to inverse correlations during secretase inhibition (i.e. when processing through one pathway decreases, the processing of the alternative pathway increases). These studies suggest that there may be an inverse relationship between the aand bpathways in inhibitor studies, while our study shows that during physiologic APP processing in the human CNS, aand bprocessing are positively correlated. We found that the molar ratio of sAPPa to sAPPb was approximately 3:1 with a shift to 2:1 from ato b-processing in the setting of amyloid deposition. The differences in ratios among these groups were not age-related since there was no significant difference between YNC and Amyloid2 groups. Prior reports estimated a to b ratios of 10:1 [31,33], however, these in vitro estimates likely had lower b-secretase activity than is present in the CNS, since b-secretase is mostly found in the brain [12][13][14]. We further showed that on average sAPPb/sAPPa was significantly higher in Amyloid+ participants than in Amyloid2 participants and YNC; therefore, the ratio may be a useful indicator of Ab plaque deposition. This result further supports the hypothesis that sporadic AD may be the result of an upregulation of b-secretase processing of APP, with respect to a-secretase. Our results are consistent with recent findings of increased CSF sAPPb in the presence of decreased Ab 42 and increased tau [26][27]. However, some reports indicate increased sAPPa [28] while others show no difference [26][27], similarly to our findings. Recently, it was reported that neither sAPPa, nor sAPPb, measured from CSF by both ELISA and mass spectrometry, was altered in AD [32]. This parallels results of an ELISA study from a decade earlier that also showed no difference in sAPPa, nor in sAPPb, when healthy controls were compared to sporadic AD patients [37]. None of these groups, however, reported sAPP metabolite ratios. To summarize, amyloidosis, and not age, was associated with a constitutive change in ato bprocessing of APP among individuals.
In conclusion, in our study we report diurnal dynamics of APP metabolites diminished with age, and, only for Ab, were further attenuated with amyloidosis. These results may explain some possible confounding factors of other studies that have measured sAPPa, sAPPb, Ab 40 , and Ab 42 levels in CSF collected at a single time point from AD versus non-AD participants. This may clarify the discrepancy in results and the wide range of concentrations of APP metabolites presented by various groups. We also indicate that taking a ratio of sAPPb/sAPPa may correct for these inconsistencies. Further, we demonstrated that there is a positive correlation among soluble APP metabolites, which diminishes with amyloidosis. This dissociation is probably due to CSF Ab 42 levels in AD no longer being representative of APP processing due to the sequestering of Ab, particularly Ab 42 , in plaques.
Advantages of this study included that the samples were obtained from the human CNS in three different participant groups and total protein concentrations showed stability over time in the older groups. Fewer than half of the YNC had total protein data available, and this, along with high inter-subject variability, does not allow us to state conclusively whether a diurnal pattern of total protein does or does not exist in the whole YNC group. However, the similar diurnal patterns among APP metabolites seem to indicate that CSF APP dynamics are likely independent of CSF total protein levels. Nevertheless, we did not directly measure aand b-secretase activities or production rates of APP metabolites. Thus, our study does not answer the question of what causes APP to rise and fall in a diurnal pattern, although possibilities include transcription, translation, or transport. Future studies into APP processing pathways, including production rates of APP and aand b-secretases may be useful to inform about causes of APP dynamics.

Supporting Information
Figure S1 Specificity and selectivity of the sAPPa ELISA. Titration curves of sAPPa and sAPPb standards were run on the sAPPa ELISA assay. The OD values from the CSF samples fell well above baseline, and within the linear range of the sAPPa standard curve. This demonstrates that this assay is sensitive enough to measure sAPPa from the biological samples in this study. The sAPPb standard curve's OD values were zero, even at the highest concentration of 300 ng/mL, which indicates that sAPPb does not cross-react with the sAPPa assay. (TIFF) Figure S2 Specificity and selectivity of the sAPPb ELISA. Titration curves of sAPPa and sAPPb standards were run on the sAPPb ELISA assay. The OD values from the CSF samples fell well above baseline, and within the linear range of the sAPPb standard curve. This demonstrates that this assay is sensitive enough to measure sAPPb from the biological samples in this study. The optical density (OD) for the sAPPb standard of 8.5 ng/ mL was approximately the same as the OD value for the sAPPa standard at a concentration of 300 ng/mL. This indicates that this ELISA is approximately 35-fold more selective for sAPPb than for sAPPa. Thus, any cross-reactivity is negligible. (TIFF) Figure S3 No diurnal pattern in total CSF protein concentrations of Amyloid2 and Amyloid+ groups. Participants' total protein concentrations in CSF over 36 hours were determined by using a micro BCA assay. For each participant group, the mean total protein concentration for each hour was calculated and plotted. Cosinor fits were applied to each group's hourly mean total protein concentration. A significant cosinor fit was found in the YNC group (n = 6), with an amplitude 4.5% (95% CI: 26.1% to 22.9%). No significant diurnal patterns were apparent in the Amyloid2 group (n = 6; 95% CI: 21.4% to +8.6%) and the Amyloid+ group (n = 5; 95% CI: 28.4% to +1.4%). (TIFF) Figure S4 sAPPb/sAPPa ratios determined from 36 hour time-course. We measured the sAPPb/sAPPa ratio for each individual based on that participant's sAPPb and sAPPa concentrations over the 36 hour time-course. Individual ratios Figure 4. Separating participant groups by the sAPPb/sAPPa ratio. We compared sAPPb and sAPPa concentrations, as well as the sAPPb/ sAPPa ratio, among groups using the first CSF collection. Each participant's first CSF sample was drawn between 7:30 A.M. and 9:00A.M. sAPPb and sAPPa concentrations were measured using two separate metabolite-specific ELISAs. Student's t-tests were used and graphs show 95% Confidence Interval error bars. A) sAPPb/sAPPa ratio was higher with amyloid deposition (Amyloid+) as compared to healthy, older controls (Amyloid2) (*p = 0.02) or young healthy controls (YNC) (**p = 0.002). No significant difference was detected between the ratio of the YNC and Amyloid2 groups (p = 0.6). B) sAPPa concentrations were not significantly higher in Amyloid+ than in YNC (p = 1.0) or Amyloid2 (p = 0.5). No significant difference was detected between the sAPPa concentration of the YNC and Amyloid2 groups (p = 0.4). C) sAPPb concentrations were not significantly higher in Amyloid+ than in YNC (p = 0.09) nor Amyloid2 (p = 0.6). No significant difference was detected between sAPPb concentrations from the YNC and Amyloid2 groups (p = 0.3). doi:10.1371/journal.pone.0089998.g004 were calculated and averaged within participant groups. Student's t-test was used and graphs show 95% Confidence Interval error bars. A) Mean sAPPb/sAPPa ratio was calculated for each participant using that participant's 36 hour mean sAPPb concentration and 36 hour mean sAPPa concentration. Individual ratios were averaged in their respective participant groups. The groupaveraged mean sAPPb/sAPPa ratio is significantly higher in YNC than in Amyloid2 (*p = 0.03) or in Amyloid+ (*p = 0.03). No significant difference was detected between the group-averaged mean sAPPb/sAPPa ratio of the Amyloid2 and Amyloid+ groups (p = 0.92). B) Mesor sAPPb/sAPPa ratio was calculated for each participant using the sAPPb mesor value (determined from the cosinor fit of that participant's 36 hour sAPPb concentrations) and the sAPPa mesor value (determined from the cosinor fit of the 36 hour sAPPa concentrations). The mesor sAPPb/sAPPa ratio is significantly higher in YNC than in Amyloid2 (*p = 0.03) or in Amyloid+ (*p = 0.03). No significant difference was detected between the mesor sAPPb/sAPPa ratio of the Amyloid2 and Amyloid+ groups (p = 0.93). (TIFF)