Caffeine Suppresses Amyloid-β Levels in Plasma and Brain of Alzheimer's Disease Transgenic Mice

Journal of Alzheimer’s Disease 17 (2009) 681–697 681DOI 10.3233/JAD-2009-1071IOS Press

Caffeine Suppresses Amyloid-β Levels inPlasma and Brain of Alzheimer’s DiseaseTransgenic Mice

Chuanhai Caoa,b, John R. Cirritoc, Xiaoyang Lina,b, Lilly Wangb,d, Deborah K Vergesc,Alexander Dicksonb,d, Malgorzata Mamcarzb,d, Chi Zhanga,b, Takashi Morie, Gary W. Arendashb,d,∗,David M. Holtzmanc,f,g and Huntington Pottera,b,h

aThe Byrd Alzheimer’s Center & Research Institute, Tampa, FL, USAbFlorida Alzheimer’s Disease Research Center, University of South Florida, Tampa, FL, USAcDepartment of Neurology, Washington University School of Medicine, St. Louis, MO, USAdDepartment of Cell Biology, Microbiology, and Molecular Biology, University of South Florida, Tampa, FL, USAeDepartments of Medical Science and Pathology, Saitama Medical Center and Saitama Medical University,Kawagoe, Saitama, JapanfDepartment of Developmental Biology, Washington University School of Medicine, St. Louis, MO, USAgHope Center for Neurological Disorders, Washington University School of Medicine, St. Louis, MO, USAhSuncoast Gerontology and Alzheimer’s Center, University of South Florida College of Medicine, Tampa, FL, USA

Abstract. Recent epidemiologic studies suggest that caffeine may be protective against Alzheimer’s disease (AD). Supportiveof this premise, our previous studies have shown that moderate caffeine administration protects/restores cognitive functionand suppresses brain amyloid-β (Aβ) production in AD transgenic mice. In the present study, we report that acute caffeineadministration to both young adult and aged AD transgenic mice rapidly reduces Aβ levels in both brain interstitial fluid andplasma without affecting Aβ elimination. Long-term oral caffeine treatment to aged AD mice provided not only sustainedreductions in plasma Aβ, but also decreases in both soluble and deposited Aβ in hippocampus and cortex. Irrespective of caffeinetreatment, plasma Aβ levels did not correlate with brain Aβ levels or with cognitive performance in individual aged AD mice.Although higher plasma caffeine levels were strongly associated with lower plasma Aβ1−40 levels in aged AD mice, plasmacaffeine levels were also not linked to cognitive performance. Plasma caffeine and theophylline levels were tightly correlated,both being associated with reduced inflammatory cytokine levels in hippocampus. Our conclusion is two-fold: first, that bothplasma and brain Aβ levels are reduced by acute or chronic caffeine administration in several AD transgenic lines and ages,indicating a therapeutic value of caffeine against AD; and second, that plasma Aβ levels are not an accurate index of brain Aβlevels/deposition or cognitive performance in aged AD mice.

Keywords: Alzheimer’s disease, amyloid-β, brain interstitial fluid, caffeine, plasma, transgenic mice

INTRODUCTION

Synthetic anti-Alzheimer’s disease (AD) drugs cur-rently on the market have mild symptomatic benefits,

∗Corresponding author: Gary W. Arendash, Ph.D., Departmentof Cell Biology, Microbiology, & Molecular Biology, University ofSouth Florida, Tampa, FL 33620, USA. Tel.: +1 813 974 1584; Fax:+1 813 974 1614; E-mail: [email protected].

are not known to be disease-modifying, and can havesignificant undesirable side-effects. Thus, it would bemost desirable to identify an inherently safe and read-ily available “nutriceutic” compound (e.g., natural andnormally in the diet) that can provide therapeutic util-ity against AD. Caffeine is a methylxanthine, high-ly concentrated in coffee, and probably the world’smost widely consumed psychoactive substance [1].The well known ability of caffeine to increase alert-

ISSN 1387-2877/09/$17.00 2009 – IOS Press and the authors. All rights reserved

UNDER EMBARGO UNTIL JULY 6, 2009, 00:00 CET

682 C. Cao et al. / Caffeine Suppresses Amyloid-β Levels in Plasma and Brain of Alzheimer’s Disease Transgenic Mice

ness and arousal primarily involves antagonism of cen-tral nervous system adenosine receptors, while addi-tional mechanisms of caffeine action (e.g., phospho-diesterase inhibition, calcium mobilization) have beenproposed [2], although they occur only at high, unphys-iologic concentrations of caffeine (mM range). Oralcaffeine is rapidly and almost completely absorbed viathe gastrointestinal tract, with blood caffeine levelsquickly equilibrating with brain tissue levels due tocaffeine’s unhindered traversal of the blood-brain bar-rier [3,4]. Caffeine is primarily metabolized in theliver to theophylline and paraxanthine, both of whichare at least as physiologically active as caffeine [3].There are no apparent differences in metabolism of, orphysiological responses to, caffeine in elder individualscompared to young individuals; moreover, caffeine’spharmacokinetics are similar after oral or intravenousadministration in humans and animals [3].

Recent longitudinal studies spanning 4–10 years sug-gest that habitual caffeine/coffee intake protects againstcognitive impairment in aging humans [5,6]. More-over, an epidemiologic study evaluated caffeine intakeduring the 20 years preceding AD diagnosis and foundthat AD patients consumed markedly less caffeine dur-ing that period compared to age-matched individualswithout AD [7]. Collectively, these and other obser-vational human studies [8] suggest that habitual caf-feine/coffee intake may protect against memory impair-ment and AD during aging. However, because suchstudies are not controlled and cannot isolate the effectof caffeine from the myriad of other lifestyle choic-es humans make, we performed a highly controlled“protection-based” study in mice [9]. In that study,we found that oral caffeine administration to AD trans-genic (APPsw) mice from young adulthood into old-er age: 1) protected these mice from otherwise cer-tain cognitive impairment in older age; and 2) limitedtheir brain production of the peptide amyloid-β (Aβ).The moderate amount of caffeine intake given to theseAPPsw mice (human equivalency of 5 cups of coffeeper day) suppressed both β-secretase (BACE1) and γ-secretase/PS1 levels in hippocampus, indicating thatcaffeine can directly impact AD pathogenesis in theseAD mice. This study in AD mice is consistent with thehuman epidemiologic literature supporting an ability ofmoderate caffeine intake to reduce risk of AD [7]. Para-doxically, human intake of coffee/caffeine declines ap-preciably during aging in Western cultures, in part dueto caffeine intake restrictions often suggested by healthcare professions. Such restrictions would appear un-warranted, based on comprehensive literature searches

indicating that moderate caffeine intake has no adverseeffects on the cardiovascular system, bone status, cal-cium balance, or the incidence of cancer during ag-ing [4,10]. Indeed, a recent study involving 18–24 yearfollow-ups found that coffee intake (4–6 cups per day)was associated with reduced mortality, particularly dueto cardiovascular disease [11].

The potential for caffeine to treat established ADhas yet to be explored in human studies. As an ini-tial step for elucidating possible efficacy of caffeine tostabilize or reverse established AD, we have recent-ly completed a “treatment-based” study in aged APP-sw mice that already contained Aβ pathology (see ac-companying paper [12]). These aged mice were con-firmed to be cognitively impaired in working memo-ry prior to receiving several months of oral caffeinetreatment. When re-tested, aged APPsw mice receiv-ing caffeine treatment exhibited working memory thatwas not only substantially better than APPsw mice thatdid not receive caffeine, but comparable to normal non-transgenic mice [12]. Thus, even with pre-existingand substantial Aβ neuropathology, aged APPsw miceexhibited memory restoration with caffeine treatment,suggesting a therapeutic potential of caffeine in casesof established AD.

Substantial evidence suggests that the brain’s pro-duction and aggregation of Aβ peptide represent keyevents underlying AD pathogenesis. As depicted inFig. 1A, newly produced Aβ enters a dynamic equilib-rium between soluble and deposited Aβ in the brain,with continual transport of soluble Aβ out of the brainand into plasma. In view of our findings that caffeinedecreases Aβ production in APPsw mice through sup-pression of both BACE1 and γ-secretase/PS1 [9], wehypothesize that resultant lower brain levels of solu-ble Aβ will at least acutely result in lower plasma Aβlevels (Fig. 1B). This hypothesis is supported by thefinding that Aβ is rapidly produced and cleared fromthe brain [13]. In the present study, we determine theeffect of both acute and chronic caffeine administrationon plasma and brain Aβ levels in AD mice. In addition,we explore possible relationships between: 1) plasmaand brain pools of Aβ; 2) plasma caffeine/theophyllineconcentrations and plasma/brain Aβ levels; and 3) plas-ma Aβ/caffeine levels and cognitive performance. Pos-sible associations between Aβ, caffeine, and cytokinesare also considered. Both young adult and aged ADmice were utilized in these studies to investigate theeffect of caffeine at both early (pre-Aβ deposition) andlate (robust Aβ deposition) stages of the disease inmice.


C. Cao et al. / Caffeine Suppresses Amyloid-β Levels in Plasma and Brain of Alzheimer’s Disease Transgenic Mice 683

Fig. 1. Diagrams depicting brain Aβ production/clearance, the suppressive actions of caffeine on Aβ production, and resultant effects on brainand plasma Aβ levels. (A) Unmodulated: Aβ is primarily produced in neurons, secreted into the brain extracellular space in soluble form,then enters a dynamic equilibrium between soluble and deposited (insoluble) Aβ. Continual transport of soluble Aβ occurs into plasma. (B)Caffeine: Short-Term: Caffeine suppression of both β- and γ-secretase activities reduces Aβ production, resulting in lower soluble Aβ in brainand plasma. The equilibrium between soluble and deposited Aβ is not impacted by this short-term reduction in brain soluble Aβ levels. (C)Caffeine: Long-Term: Continued caffeine suppression of Aβ production and resultant lower levels of brain soluble Aβ induce a flux of deposited(insoluble) Aβ to the soluble form, which is cleared from brain into plasma via soluble Aβ transport. Plasma Aβ levels may be reduced ornot changed, depending on degree of caffeine-induced suppression of Aβ production. In aged APPsw mice given chronic caffeine treatment,their lower brain Aβ levels/deposition results in reversal of cognitive dysfunction. (Colours are visible in the electronic version of the article atwww.iospress.nl.)

Based on encouraging results from our complet-ed caffeine administration studies in AD transgenicmice [9,12], clinical trials in aged individuals are cur-rently being conducted to investigate the effects of caf-feine administration on plasma Aβ levels. As such, amajor aim of the present study was to attain advanceinsight into potential results from those clinical trialsthrough similar studies in AD transgenic mice.

MATERIAL AND METHODS

Animals

In these studies, all mice (with two exceptions) had amixed background of 56.25% C57, 12.5% B6, 18.75%SJL, and 12.5% Swiss-Webster and were derived froma cross between heterozygous mice carrying the mu-tant APPK670N, M671L gene (APPsw) with heterozy-gous PS1 (Tg line 6.2) mice. This resulted in offspringconsisting of mutant APPsw, PS1, APPsw+PS1, andnon-transgenic (NT) genotypes. The two exceptionswere: 1) APPsw+PS1 mice used in the 7-day gav-age treatment study, which had a B6C3 backgroundand were obtained from the Jackson Laboratory (BarHarbor, MA); and 2) Tg2576 mice used in the in vivo

microdialysis study, which had a C57/BL6/SJL back-ground and were a gift from Karen Hsiao Ashe (Uni-versity of Minnesota; Hsaio et al. 1996). All micewere maintained on a 12-hour dark and 12-hour lightcycle with ad libitum access to rodent chow and wa-ter/caffeinated water. All animal procedures were per-formed in AAALAC-certified facilities under protocolsapproved by Institutional Animal Care and Use Com-mittees at University of South Florida and the JA HaleyVA Hospital.

General protocol

A spectrum of studies involving both acute andchronic caffeine administration was performed inAPPsw/Tg2576 and APPsw+PS1 transgenic lines,with plasma, neurochemical, and/or behavioral mea-sures collected. It should be underscored that thesetransgenic lines have measurable levels of soluble Aβin both brain and plasma in young adulthood, with theirbrain levels of Aβ increasing appreciably during ag-ing. This results in Aβ plaque formation beginningaround 10–11 months of age for APPsw/Tg2576 miceand by 6 months of age for APPsw+PS1 mice. Thus,the 3–4 month-old APPsw/Tg2576 mice used in thesestudies had no Aβ deposition, while the 14 month and



older APPsw mice exhibited age-dependent Aβ depo-sition. All APPsw+PS1 mice utilized in these studieswere at least 15 months of age and therefore had robustAβ deposition. The general protocol for each study isindicated below:

Acute caffeine cffects on plasma Aβ levelsAcute (single treatment) administration of caffeine

or saline vehicle was given by intraperitoneal (i.p.) in-jection or gavage to the following groups of mice: 3–4month-old APPsw mice (i.p.), 14 month-old APPswmice (i.p. or gavage), and 14 month-old APPsw+PS1mice (gavage). A pre-treatment blood sample (0.15ml) was taken by sub-mandibular vein puncture 3–4days before treatment. At 3–4 hours following caffeine(1.5 mg/0.2 ml; Sigma, St. Louis, MO) or vehicle ad-ministration, another blood sample was taken. In theseacute studies, each group of caffeine- or vehicle-treatedanimals consisted of 5–7 Tg mice. For all acute caf-feine treatment studies, an equivalent volume of 0.9%saline (0.2 ml) was given immediately following anyblood sample taken.

Long-term caffeine effects on plasma Aβ levelsLonger term caffeine administration was given by

gavage to aged APPsw+PS1 mice. At 3–4 days fol-lowing a pre-treatment blood sample, 15–20 month-old APPsw+PS1 mice (n = 6) were started on twice-daily caffeine treatment (1.5 mg/0.2 ml each) via gav-age for 7 consecutive days. A blood sample was tak-en on the final day of caffeine treatment, as well as 9days thereafter. A second group of four 20 month-oldAPPsw+PS1 mice were pre-treatment bled, and thengiven two caffeine treatments (1.5 mg/0.2 ml each) viagavage every 4th day for up to two months. In this sec-ond long-term study, blood samples were taken every8th day during treatment, always on the day followinga treatment. As was the case for acute studies, volumereplacement with 0.9% saline occurred immediatelyfollowing each blood sample.

In vivo microdialysis: Acute caffeine effects oninterstitial fluid (ISF) levels of Aβ in hippocampus

In vivo microdialysis was used to assess brain ISFAβx−40 in the hippocampus of awake, freely movingTg2576 mice. Microdialysis was performed similarto previously described methods [14]. This techniquesamples soluble molecules within the extracellular flu-id that are smaller than 38-kilodaltons, the molecularweight cut off of the probe membrane. Aβ capable ofentering the probe has been dubbed “exchangeable Aβ

or eAβ” [14]. The pool of eAβ is in dynamic equi-librium with the total pool of ISF Aβ. During micro-dialysis, mice were housed in a constant light condi-tion and remained awake with freedom of movementand ad lib food and water during microdialysis. Mi-crodialysis perfusion buffer was artificial cerebrospinalfluid (CSF) containing 0.15% bovine serum albuminthat was filtered through a 0.1 µm membrane. Flowrate was a constant 1.0 µl/minute, which recovers 23.4± 1.7% (mean ± SEM) of exchangeable Aβ withinthe brain ISF of Tg2576 mice. Samples were collectedevery 30–60 minutes with a refrigerated fraction col-lector into polypropylene tubes and assessed for Aβ bysandwich ELISA at the completion of each experiment,as described by Cirrito et al. [15]. Briefly, Aβx−40

was assessed using an Aβ40-specific mouse monoclon-al antibody, mHJ2, as a coating antibody and a bi-otinylated central domain antibody, mHJ5.1, as the de-tecting antibody, followed by streptavidin-poly-HRP-40 (Fitzgerald Industries, Concord, MA). All ELISAassays were developed using Super Slow ELISA TMB(Sigma) and absorbance read on a Bio-Tek FL-600 platereader (Winooski, Vermont) at 650 nm. Basal levels ofISF Aβ were defined as the mean concentration of Aβ

over 5–6 hours preceding drug treatment. The meanin vivo concentration of basal ISF eAβ was 3.9 ± 0.78ng/ml (n = 6). For each animal, all Aβ levels werenormalized to the basal Aβ concentration. Once basalISF Aβ levels were established, Tg2576 mice were ad-ministered caffeine i.p. and ISF Aβ levels were sam-pled every 30 minutes until the end of the study. Micewere studied at 3 months of age, which is prior to Aβdeposition in this mouse model.

ISF Aβ half-life was determined similar to Cirritoet al. [14]. Microdialysis probes were inserted as de-scribed previously. Basal levels of ISF Aβ were estab-lished for five hours, followed by i.p. administrationof 30 mg/kg caffeine or vehicle (PBS). Three hoursafter treatment, the γ-secretase inhibitor Compound E(Alexis Biochemicals, San Diego, CA; 100 nM) wasadded to the microdialysis perfusion buffer to rapidlyinhibit Aβ production near the probe. The IC50 forthis compound to inhibit γ-secretase activity in vitro is0.3 nM. Microdialysis samples were collected every 30minutes for an additional four hours and then assessedfor Aβx−40 by sandwich ELISA. The half-life of Aβ

was calculated based on the slope of decline in ISFAβ levels [14] beginning one hour after the onset ofCompound E administration.



Long-term caffeine effects in behaviorally-tested miceAt 18–19 months of age, APPsw and NT littermate

controls were pre-tested in the radial arm water maze(RAWM) task of working memory according to ourestablished protocol [9,12,16]. Following confirma-tion that APPsw mice were cognitively-impaired, theywere divided into two groups, with half of Tg micestarted on caffeine administered in their drinking wa-ter (0.3 mg/ml, providing a daily dose of ≈ 1.5 mgcaffeine/mouse) and the other half remaining on stan-dard tap water, as detailed previously [9]. At 4–5weeks into caffeine treatment, all mice were re-testedin the RAWM, with both last block and overall errorsbeing analyzed for working memory Trials 4 and 5.Following completion of behavioral testing at 20–21months of age, mice were deeply anesthetized withsodium pentobarbital, a blood sample was then takenfor neurochemical analysis, followed by transcardialperfusion with 100 ml of 0.9% saline. Postmortembrains were immediately removed and bisected sagital-ly. The hippocampus and cerebral cortex was dissectedfrom the right hemisphere and processed for solubleAβ1−40 and Aβ1−42 determinations by ELISA, as wellas for cytokine levels. Briefly, 30 mg brain sampleswere homogenized in 400 µl RIPA buffer [100 mMTris [pH8.0], 150 mM NaCl, 0.5% DOC, 1% NP-40,0.2% SDS, and 1 tablet proteinase inhibitor per 100ml (S8820, Sigma)], and sonicated for 20 seconds onice. Samples were then centrifuged for 30 minutes at27,000 g at 4◦C, and supernatants were transferred intonew screw cap tube. The supernatants obtained fromthis protocol were then stored at −80◦C for determina-tion of soluble Aβ levels using ELISA kits (KHB3482for 40, KHB3442 for 42, Invitrogen, Carlsbad, CA).Standard and samples were mixed with detection an-tibody and loaded on the antibody pre-coated plate asthe designated wells. HRP-conjugated antibody wasadded after wash, and substrates were added for col-orimetric reaction, and then stopped with sulfuric acid.Optical density was obtained and concentrations werecalculated according a standard curve. Plasma Aβ1−40

and Aβ1−42 levels were determined with the same pro-tocol and using the same ELISA kits. Homogenatesfrom the right hippocampus and cerebral cortex werealso analyzed for cytokine levels (see next section).

The left hemisphere was histologically processedfor analysis of total Aβ deposition, as previously de-scribed [17]. Briefly, equally-spaced 5-µm sections atthe hippocampal level were immunostained for totalAβ deposition using a biothinylated human Aβ mon-oclonal antibody (clone 4G8; Covance Res. Products,

Everyville, CA). Quantitative image analysis was thenperformed according to Mori et al. [18] and Aβ bur-den was determined as a percentage of immunolabeledarea (positive pixels) relative to the full area captured(total pixels). Quantitative image analysis was donebased on previous methods with modifications [18,19].Images were acquired using an Olympus BX60 mi-croscope with an attached digital camera system (DP-70, Olympus, Tokyo, Japan). The digital image wasthen routed into a Windows PC for quantitative anal-ysis using SimplePCI software (Compix Inc., ImagingSystems, Cranberry Township, PA). Images of five 5-µm sections (150 µm apart) through the hippocampuswere captured from each animal, and a threshold op-tical density was obtained that discriminated stainingfrom background. Each region of interest was manual-ly edited to eliminate artifacts. For Aβ burden analy-sis, data are reported as percentage of immunolabeledarea captured (positive pixels) relative to the full areacaptured (total pixels). Each analysis was done by asingle examiner blinded to sample identities.

Plasma neurochemical analysis

Plasma from blood samples was analyzed for levelsof Aβ1−40, caffeine, theophylline (a caffeine metabo-lite), and cytokines. Aβ1-40 levels were determinedaccording to the aforementioned methodology involv-ing brain tissues. Plasma caffeine and theophyllineconcentration were measured with ELISA Kits fromNeogen (Lansing, MI, USA) by following the manu-facture protocol. In brief, the enzyme conjugate solu-tion was prepared by diluting the 180X enzyme conju-gate stock 1 to 180 in the EIA buffer provided. Caf-feine (or theophylline) standard was then diluted withEIA buffer at two fold dilutions from 200 ng/ml to 0.39ng/ml. Then 20 µl standard of each dilution was addedinto the coated plate. Plasma samples were then dilutedwith EIA buffer, with 20 µl of this dilution added intothe coated plate. Both standard and samples were runin duplicate in the plate. Positive and negative controlsof 20 µl were loaded to each plate. Then 180 µl of di-luted drug-enzyme conjugate was added into each welland mixed by gently shaking the plate. Plates werecovered with plastic film and incubated at room tem-perature for 45 minutes. During the incubation, a 10xwash buffer was diluted to 1X with DI water and mixedthoroughly. Once incubation was completed, the liquidwas dumped from the wells. Plates were then taped ona clean lint-free towel to remove any remaining liquidin the wells. Then each well was washed with 300 µl of



diluted wash buffer 3 times. After completing the lastwash step, the bottom of the wells was wiped with alint-free towel to remove any liquid on the outside of thewells. Then 150 µl of the K-Blue Substrate was addedto each well with a multi-channel pipette. The plate wasthen mixed by gently shaking, followed by incubationat room temperature for 5 to 20 minutes. To stop theenzyme reaction, 50 µl of red stop solution was addedto each well and gently mixed. The absorbance wasthen measured with plate reader (Synergy HT, Biotek,VT) at a wavelength of 650 nm. The absorbance wasconverted into concentration using Gen5 software.

Cytokine expression profiles were detected using theBio-Plex kits (Bio-Rad, Richmond, CA, catalogue #171F11181). Samples and standards were prepared us-ing company protocols with the initial concentration ofstandards ranging from 32 ng/ml to 1.95 pg/ml. Plas-ma samples were prepared for analysis by diluting 1volume of the serum sample with 3 volumes of the Bio-Plex mouse sample diluents. Details of this procedurewere performed by followed the protocol provided bythe manufacture. Finally, the plates were read. Eachcytokine level was calculated based on its own stan-dard curve. Brain cytokine and chemokine levels weredetected with the same method.

Statistical analyses

Group comparisons involving levels of soluble/de-posited Aβ, caffeine, theophylline, and cytokines wereperformed using ANOVA. For determination of pre-treatment versus post-treatment differences in plasmaAβ level, paired t-tests were employed. All other sta-tistical analyses are as indicated in the text. In order totest if relationships were present between plasma, neu-rochemical, and behavioral measures, correlation anal-ysis was performed using the Systat analytical softwarepackage.

RESULTS

Acute caffeine administration reduces plasma Aβ level

We have previously found that Aβ1−40 and Aβ1−42

generation is decreased in N2a neuronal cell culturesfollowing 6 hours of caffeine treatment [9]. More-over, these reductions in generation of both Aβ1−40

and Aβ1−42 peptides were dose-dependent, occurringat caffeine concentrations (� 10 µM), which are typi-cally present in plasma following coffee consumption

in humans. Given the rapid delivery of caffeine to allbody organs including the brain, we therefore hypothe-sized that acute caffeine administration would quicklysuppress brain Aβ production in AD transgenic mice,resulting in a significant reduction in plasma Aβ levelswithin hours. In an initial study in which 3–4 month-old APPsw mice (pre-plaque) were given a single i.p.injection of caffeine (1.5 mg), plasma Aβ1−40 levelswere significantly reduced by 41% at 3 hours post-treatment compared to pre-treatment levels (Fig. 2A);vehicle injection failed to affect plasma Aβ levels inother littermate APPsw mice. Caffeine-induced reduc-tions in plasma Aβ1−40 levels were also observed inaged 14 month-old APPsw mice (plaque-bearing) at 3hours following either a single i.p. or gavage treatmentwith the same dose of caffeine (Fig. 2B). Moreover,plasma caffeine concentrations in these same 14 month-old APPsw mice were negatively correlated with plas-ma Aβ1−40 levels; higher plasma caffeine levels wereassociated with lower plasma Aβ levels in individualanimals (Fig. 2C). Even aged 19-month APPsw+PS1mice (plaque-bearing) exhibited significant reductionsin plasma Aβ1−40 levels at 3 hours following gavagetreatment with caffeine (Fig. 2D).

Long-term oral caffeine treatment provides asustained reduction in plasma Aβ levels

Our prior work has demonstrated that hippocampallevels of both soluble and insoluble Aβ are reducedby greater than 30% following 51/2 months of oral caf-feine treatment (≈1.5 mg/day) to young adult APP-sw mice [9]. In the same study, we also found thissame treatment to be effective in reducing insolublehippocampal Aβ levels in aged 17 month-old APPswmice following 18 days of treatment. We therefore hy-pothesized that long-term oral caffeine treatment wouldresult in a sustained suppression of plasma Aβ levels.In fact, aged 15–20 month-old APPsw mice (plaque-bearing) that were given oral treatment with 1.5 mgcaffeine twice daily for one week did exhibit signifi-cant reductions in plasma levels of both Aβ1−40 andAβ1−42 compared to pre-treatment values (Fig. 2E). At9 days following cessation of this caffeine treatment,however, plasma levels of both Aβ1−40 and Aβ1−42

had returned to their pre-treatment levels. In a secondlong-term treatment study, 20 month-old APPsw+PS1mice (plaque-bearing) were given oral caffeine treat-ment (1.5 mg) twice daily every 4th day over a 2month period. Periodic analysis of plasma Aβ1−40 andAβ1−42 levels during this treatment period revealed not



Fig. 2. Caffeine treatment induces a rapid and sustained decrease in plasma Aβ levels. (A,B,D) A single i.p. or gavage treatment with caffeinesignificantly reduces plasma Aβ levels in 3–4 month-old (A) and 14 month-old (B) APPsw mice; as well as in 19 month-old APP+PS1 mice(D). In all three studies, saline vehicle treatment had no effect. (C) Plasma caffeine levels in 14M old APPsw mice are inversely correlatedwith plasma Aβ1−40 levels. (E) Oral caffeine treatment for one week to 15–20 month-old APPsw mice results in decreased plasma levels ofboth Aβ1−40 and Aβ1−42 immediately following treatment, with a rebound in plasma Aβ levels occurring by 9 days after cessation of caffeinetreatment. Each group in A, B, D, and E consisted of 5–10 mice. (F) As exemplified by two aged 20 month-old APP+PS1 mice, oral caffeineadministration every 4th day over a 2-month period induces a sustained and continual decrease in plasma levels of both Aβ1−40 and Aβ1−42

over the treatment period. Post-treatment versus pre- or delayed post-treatment Aβ levels were evaluated by paired t-tests. ∗p < 0.05–0.01versus pre-treatment levels; ∗∗p < 0.05 versus both pre-treatment and 9-day post-treatment levels. (Colours are visible in the electronic versionof the article at www.iospress.nl.)



Fig. 3. Caffeine lowers interstitial fluid (ISF) Aβ levels. (A) Brain ISF Aβx −40 levels were measured by in vivo microdialysis. Prior to treatmentwith caffeine, ISF Aβ levels fluctuated very little. A low dose (5 mg/kg) and a higher dose (30 mg/kg) of caffeine caused ISF Aβ levels todecrease significantly compared to basal levels. (B) Following a 5 mg/kg dose of caffeine i.p., ISF Aβx −40 levels decreased by 19% compared tobasal levels (∗p < 0.05), while a 30 mg/kg administration decreased Aβ levels by 33% (∗∗p < 0.01, n = 6 per group). The mean concentrationof ISF Aβ represents an average of hours 2–3 after each dose of caffeine when Aβ levels had stabilized. (C) Tg2576 were pre-treated with vehicleor 30 mg/kg caffeine for 3 hours (n = 6 per group), followed by administration of 100 nM Compound E, a γ-secretase inhibitor, via reversemicrodialysis. In both groups, ISF Aβ levels decreased rapidly when APP cleavage was blocked. (D) The elimination half-life of ISF Aβx −40

was comparable in vehicle-treated and caffeine-treated mice, suggesting that caffeine does not affect ISF Aβ elimination, but likely impacts Aβproduction instead. (Colours are visible in the electronic version of the article at www.iospress.nl.)

only a sustained reduction in both Aβ peptides, but al-so a continuing decrease in their levels throughout the2-month treatment. Given that the half-life of caffeinein rodents is only 0.7 to 1.2 hours [3], these later ob-servations suggest that the effects of caffeine on Aβsuppression greatly exceed its own half-life.

Caffeine administration lowers brain interstitial fluidlevels of Aβ in vivo

Aβ is primarily produced in neurons and secretedinto the brain extracellular space where it is normally

soluble within ISF. To determine if caffeine adminis-tration acutely affects brain ISF Aβ levels, 3 month-oldTg2576 mice (pre-Aβ plaque) were treated with sev-eral doses of caffeine during in vivo microdialysis tomeasure ISF Aβ levels. Microdialysis probes were im-planted into the hippocampus, permitting us to sampleISF Aβ levels every 30 minutes for up to 24 hours inawake, behaving animals [14,15]. Basal ISF Aβ lev-els were determined over 6 hours in each mouse fol-lowed by i.p. administration of caffeine at 5 mg/kg(∼0.15 mg/mouse), then at 30 mg/kg (∼1 mg/mouse)



Fig. 4. Long-term caffeine administration to aged, cognitively-impaired APPsw mice reduces soluble Aβ levels in both cortex and hippocampus,while also decreasing deposited (insoluble) Aβ in hippocampus. Caffeine was administered to 18–19 month-old APPsw mice for two months intheir drinking water. Both brain Aβ1−40 and Aβ1−42 were decreased in caffeine-treated Tg mice. Although plasma Aβ levels were unaffectedfor all Tg/Caff mice inclusively, see Fig. 7B. Immunohistochemical Aβ deposition in the hippocampus was reduced by 40% in caffeine treatmentmice. ∗p < 0.05; ∗∗p < 0.001. Each group consisted of 5–8 mice. (Colours are visible in the electronic version of the article at www.iospress.nl.)

3 hours later (Fig. 3A). Caffeine significantly loweredISF Aβ levels by 19% and 32% at the low and highdoses, respectively, as compared to the basal ISF Aβlevels in each mouse (Fig. 3B). There was also a trendthat Aβ levels were reduced to a greater extent by the30 mg/kg dose of caffeine than by the 5 mg/kg dose(p = 0.1).

Caffeine can have many affects in a living animalwhich could potentially alter Aβ production or Aβelimination, thus lowering ISF Aβ levels [3,20]. Assuch, we determined the elimination half-life of ISF Aβin mice treated with vehicle or 30 mg/kg caffeine. BasalISF Aβ levels were measured in Tg2576 mice, followedby i.p. administration of 30 mg/kg caffeine or vehicle(Fig. 3C). As expected, caffeine reduced ISF Aβ lev-els by 35% compared to no change in vehicle-treatedmice. Three hours later, the microdialysis perfusionbuffer was switched to contain a potent γ-secretase in-hibitor, Compound E, to rapidly block Aβ production.In vehicle-treated mice, the elimination half-life of Aβwas 1.5 hours (Fig. 3D), which is similar to previous re-ports of the ISF Aβ half-life in this mouse model [21].Importantly, the half-life of ISF Aβ in caffeine-treatedmice (1.3 hours) was not significantly different fromvehicle-treated mice. This suggests that caffeine doesnot alter ISF Aβ elimination, but likely impacts someaspect of Aβ production instead.

Long-term oral caffeine treatment to aged,cognitively-impaired APPsw mice reduces brainsoluble and deposited Aβ

To determine the effects of chronic caffeine treat-ment in cognitively-impaired AD mice, caffeine(∼1.5 mg/day) was orally administered in drinking wa-ter to 18–19 month-old APPsw mice (plaque-bearing)that were confirmed to be impaired in the RAWM taskof working memory prior to treatment. At 4–5 weeksinto caffeine treatment, impaired APPsw mice that hadbeen given caffeine (Tg/Caff) exhibited substantial-ly better RAWM working memory performance com-pared to the continuing impairment of control APPswmice [12]. After euthanizia at 20–21 months of age (2months into caffeine treatment), soluble Aβ1−42 levelsin both cortex and hippocampus of Tg/Caff mice weresignificantly reduced by 51% and 59%, respectively,compared to Tg controls (Fig. 4A). Cortical and hip-pocampal Aβ1−40 levels were also reduced by chroniccaffeine treatment, although the decrease in hippocam-pal Aβ1−40 did not reach statistical significance (p =0.09). Compared to Tg controls, plasma Aβ1−40 lev-els were not significantly decreased in Tg/Caff mice.However, when Tg/Caff mice were divided into twosub-groups based on higher versus lower plasma caf-feine levels, a significant decrease in plasma Aβ1−40



Fig. 5. Plasma Aβ levels do not correlate with brain Aβ levels in aged APPsw mice. (A,B) Strong correlations are present between levels ofAβ1−40 and Aβ1−42 for both soluble and insoluble Aβ in hippocampus, irrespective of caffeine treatment. By contrast, no correlations existedbetween hippocampus (or cortex) and plasma for soluble Aβ levels (C) or insoluble Aβ levels (D) irrespective of caffeine treatment. (Coloursare visible in the electronic version of the article at www.iospress.nl.)

was evident in higher plasma caffeine mice (see sectionbelow). Finally, and in the same behaviorally-testedaged Tg mice, chronic caffeine treatment resulted in aremarkable 40% reduction in hippocampal Aβ depo-sition compared to Tg controls (Fig. 4B). Brain levelsof insoluble Aβ (as measured by ELISA) were alsoreduced in caffeine-treated Tg mice, as exemplified bythe 29% and 33% decreases in Aβ1−40 and Aβ1−42,respectively, seen in posterior cortex (data not shown).

Plasma Aβ levels do not correlate with brain Aβlevels or cognitive performance in aged APPsw mice

To explore the relationship between brain and plas-ma pool of Aβ, as well as the association between var-ious Aβ pools and cognitive performance, we next per-formed correlation analysis between these measures inthe same aged APPsw mice whose brain and plasmaAβ measures were presented in Fig. 4. These 20–21month-old APPsw mice exhibited strong correlations

between Aβ1−40 and Aβ1−42 levels in hippocampusand cortex, and for both soluble and insoluble Aβ pools(Fig. 5A and B). These correlations were present irre-spective of whether all Tg mice, or only Tg controlswere included. In sharp contrast, plasma Aβ1−40 wasnot correlated with any of the 8 brain Aβ measuresevaluated (e.g., hippocampus or cortex, soluble or in-soluble, Aβ1−40 or Aβ1−40). Figure 5C and D showstwo representative plots involving hippocampal solu-ble/insoluble Aβ1−40 versus plasma Aβ1−40. The lackof correlations between brain and plasma Aβ measureswas present irrespective of whether all Tg mice, or onlyTg controls were included. Finally, there were no cor-relations between plasma Aβ1−40 levels and four mea-sures of cognitive performance in the RAWM task ofworking memory; two of these correlations are shownin Fig. 6 (A and B). However, as we have consistentlyshown in prior studies [22–25], Aβ levels in cerebralcortex, entorhinal cortex, and hippocampus were close-ly linked to RAWM performance; higher brain Aβ lev-



Fig. 6. Plasma Aβ levels do not correlate with cognitive performance in aged APPsw mice. (A,B) No correlations were present between plasmaAβ levels and RAWM working memory. (C,D). By contrast, strong correlations were evident between brain Aβ levels/deposition and cognitiveimpairment, as exemplified by correlations involving soluble Aβ levels in posterior cortex (C) and Aβ deposition in entorhinal cortex (D). Eachsymbol represents an individual Tg mouse. All correlations involve all Tg mice collectively (Tg controls and Tg/Caff), although identical resultswere observed when only Tg controls were included in the analysis. (Colours are visible in the electronic version of the article at www.iospress.nl.)

els were strongly correlated with poorer working mem-ory performance (Fig. 6C and D). Thus, plasma lev-els of Aβ are not an accurate index of: 1) soluble orinsoluble brain Aβ; or 2) cognitive performance.

The relationship between plasma caffeine levels inaged APPsw mice and their plasma/brain Aβ levelsand cognitive performance

Plasma taken at euthanasia from the 6 behaviorally-tested 20–21 month-old APPsw mice that have beengiven oral caffeine treatment for two months was an-alyzed for concentrations of caffeine and theophylline(a major biologically-active metabolite of caffeine). Astrong inverse correlation was evident between plasmacaffeine concentration in those mice and their plasmalevels of Aβ1−40 (Fig. 7A), with higher plasma caffeinelevels associated with lower plasma Aβ levels. Be-cause there was a significant range in plasma caffeine

concentrations among these animals, the three APP-sw mice with the highest plasma caffeine levels [mean= 9331 ng/ml] were compared to the three mice withthe lowest levels for plasma Aβ levels [mean = 769ng/ml]. Mice with higher plasma caffeine levels hadsignificantly reduced plasma Aβ1−40 levels comparedto those with lower plasma caffeine levels (Fig. 7B).However, plasma caffeine levels in these APPsw micewere not correlated with their Aβ1−40 or Aβ1−42 lev-els in either hippocampus or cerebral cortex (data notshown). Moreover, as was the case for plasma Aβ lev-els (Fig. 6A and B), there were no correlations betweenplasma caffeine levels and RAWM working memoryperformance in aged APPsw mice (Fig. 7C and D).Plasma caffeine levels were, however, strongly corre-lated with plasma levels of theophylline (Fig. 7E). Im-portantly for human relevance, the 6 aged APPsw micechronically treated with caffeine had a mean plasmacaffeine concentration of 26 ± 11.5 µM, which is com-



Fig. 7. Effects of caffeine and theophylline on plasma, brain, and cognitive measures. In aged 20–21 month-old APPsw mice that had beentreated for 2 months with caffeine and cognitively evaluated (n = 6), blood and brain tissues taken at euthanasia were analyzed. (A) High plasmacaffeine levels correlated with lower plasma Aβ levels. (B) Mice with higher plasma caffeine levels had lower plasma Aβ1−40 levels and lowerhippocampal cytokine levels than those mice with lower caffeine levels. ∗p < 0.05; ∗∗p < 0.01. (C,D) As exemplified by these 2 correlationgraphs, plasma caffeine levels in individual mice were not correlated with their radial arm water maze (RAWM) working memory performance.(E) Plasma caffeine and plasma theophylline levels were strongly correlated. (F) High plasma theophylline levels correlated with lower plasmaIFN-γ levels. (Colours are visible in the electronic version of the article at www.iospress.nl.)

parable to plasma caffeine levels in humans followingseveral cups of coffee [3].

The association of Aβ and caffeine levels in plasmawith cytokines in plasma/brain

Aβ and caffeine have been reported to have pro- andanti-inflammatory actions, respectively [20,26]. There-fore, we investigated whether plasma levels of these twocompounds impacted plasma or brain cytokine levelsin aged APPsw mice (plaque-bearing) following two

months of oral caffeine treatment. Plasma Aβ1−40 lev-els were not correlated with plasma cytokine levels forall Tg mice collectively, irrespectively of caffeine treat-ment (data not shown). Additionally, plasma caffeinelevels in caffeine-treated Tg mice were not correlatedwith plasma cytokine levels (data not shown). Indeed,there were largely no differences in “plasma” cytokinelevels between non-transgenic, Tg, and Tg/Caff groupsin this study (data not shown). Although this was alsothe case for “brain” cytokine levels, closer inspectionrevealed potential effects of caffeine on levels of key



inflammatory cytokines. Specifically among caffeine-treated Tg mice, those mice with higher caffeine lev-els had significantly reduced levels of hippocampal cy-tokines compared to mice with lower caffeine levels(Fig. 7B). These results suggest an anti-inflammatoryaction of caffeine in a key brain area for cognitive func-tion. The caffeine metabolite “theophylline” may al-so be contributory to the anti-inflammatory effect ofcaffeine, as exemplified by the lower levels of plas-ma interferon (IFN)-γ in APPsw mice having higherplasma theophylline levels (Fig. 7F). Indeed, plasmalevels of theophylline in caffeine-treated APPsw mice:1) were correlated with lower cytokine levels in bothhippocampus and cortex; and 2) showed the same anti-inflammatory profile as caffeine (Fig. 7B) when ana-lyzed in terms of high vs. low theophylline levels (datanot shown).

DISCUSSION

In this report, we show that caffeine treatment to ADtransgenic mice lowers Aβ levels in plasma and brainISF within a few hours and can provide continued re-ductions in plasma Aβ through oral treatment periodsof 1-8 weeks. Even in aged, cognitively-impairedAPP-sw mice bearing pre-existing and substantial Aβ bur-dens, oral caffeine administration over several monthsreduced both soluble and deposited brain Aβ. Related-ly, this same caffeine treatment reverses the poor mem-ory performance of aged APPsw mice back to the levelof non-transgenic (normal) mice [12]. However, be-cause there were no correlations between hippocampaland plasma Aβ levels in individual mice, we concludethat plasma Aβ levels are not an accurate index of brainAβ levels in aged AD mice. Moreover, because therewere no correlations between plasma Aβ levels andcognitive performance in the same aged AD mice, weconclude that blood Aβ levels are not an accurate indexof cognitive performance. From an overall perspec-tive, then, plasma Aβ levels are not a viable biomarkerfor the cognitive and neurochemical characteristics ofaged AD mice. The inability of plasma Aβ1−40 levelsto acutely reflect decreased brain Aβ1−40 levels mayinvolve a consequent, acute decrease in Aβ1−40 trans-porter activity (below saturability levels) at the bloodbrain barrier [27]. Secondly, our collective results un-derscore that caffeine, its metabolites, and/or analogsshould be considered for prevention and treatment trialsin AD because caffeine can suppress Aβ production,but not its clearance from the brain.

The concentration of Aβ in brain tissue and CSFis some 50–100 times higher than in plasma [28]. Itis thus likely that a sizable amount of plasma Aβ1−40

originates from production in, and out of, the brain.Indeed, Aβ is rapidly produced and cleared from thebrain, with similar ISF clearance rates in AD mice andhumans [13,14]. Given the relatively short (<2 hours)half-life of ISF Aβ in AD mice [14], it is apparent thatnewly produced Aβ enters a dynamic equilibrium be-tween soluble and insoluble/deposited Aβ in the brain,with continual transport of soluble Aβ out of the brainand into plasma (Fig. 1, Unmodulated). In the presentstudy, a single treatment with caffeine rapidly reducedboth brain ISF and plasma levels of Aβ1−40 withinseveral hours, indicating a direct and immediate effectof caffeine on brain Aβ levels. Alternatively, becauseblood platelets appear to be the primary source of cir-culating APP and Aβ, it is possible that caffeine wasindependently lowering plasma Aβ by suppressing Aβproduction from blood platelets. The peripheral pro-duction of Aβ could certainly be one explanation for thelack of correlations between plasma and brain Aβ – orbetween plasma Aβ and any behavioral/neurochemicalmarkers evaluated in this study.

A single treatment with caffeine did not affect thehalf-life of ISF Aβ, demonstrating that caffeine hadaffected brain Aβ production rather than its elimi-nation. Underscoring this premise, our prior stud-ies [9,12] show that caffeine affects Aβ productionthrough suppression of both β-secretase (BACE1)and γ-secretase/PS1 expression. Whether or not thiscaffeine-induced suppression of Aβ production is di-rect or involves adenosine receptor blockade/mediationis currently unknown and a subject of current researchin our laboratories. Acutely, such decreased Aβ pro-duction would result in lower soluble Aβ in brain ISFand consequently lower plasma Aβ levels (Fig. 1, Caf-feine – Short-Term). This is exactly what we observedin AD mice that were young adults (no brain Aβ de-position), as well as aged AD mice (robust Aβ de-position). It is noteworthy that even caffeine treat-ment every fourth day was sufficient to provide a sus-tained reduction in plasma Aβ levels to aged AD miceover several months. Thus, the Aβ-reducing effects ofcaffeine exceed its own half-life, which does not pre-clude adenosine receptor-mediated effects of caffeine.Indeed, adenosine receptor activation has been shownto impact gene expression/signal transduction [3].

We recently reported that caffeine, when adminis-tered in drinking water from young adulthood througholder age, protected APPsw mice from otherwise cer-



tain impairment and reduced their brain Aβ levels [9].These results, suggestive that moderate daily intake ofcaffeine (the human equivalent of 5 cups of coffee dai-ly) could delay or reduce the risk of AD, underscore epi-demiologic studies reporting that caffeine is protectiveagainst both AD [7] and cognitive impairment associ-ated with normal aging [5,6,8]. In a new study [12], weshow that this same caffeine treatment paradigm, whengiven for two months to cognitively-impaired APPswmice, reverses their working memory impairment tonormal levels. We further report that this 2-month caf-feine administration regime, given to aged APPsw micewith pre-existing and substantial Aβ burdens, reducesboth soluble and deposited brain Aβ levels – in essence,reversing their brain Aβ neuropathology. Althoughthe mechanism(s) for caffeine-induced reversal of Aβdeposition requires further investigation, we hypothe-size that “chronic” caffeine suppresses Aβ productionlong-term, resulting in consistently lower brain levelsof soluble ISF Aβ (Fig. 1, Caffeine – Long-Term). Thiswould then induce a flux of insoluble Aβ out of thedeposited form and into the soluble form due to dy-namic equilibrium. Newly-solubilized Aβ would thenbe cleared into the plasma via both blood-brain-barriertransport and bulk fluid flow. The dynamic equilibri-um between soluble (ISF) and insoluble/deposited Aβin the brain is highlighted by our earlier work show-ing that young adult and aged APPsw mice had similarsteady-state levels of ISF Aβ, but rapid inhibition of Aβproduction resulted in a two-fold longer Aβ half-lifein the aged, Aβ deposit-bearing mice [14]. We believethis longer Aβ half-life is caused by solubilization ofa portion of the deposited Aβ pool due to the dynamicISF Aβ ↔ deposited Aβ equilibirum.

In contrast to caffeine’s clear ability to reduce brainAβ levels through chronic two-month treatment to agedTg mice, these same mice did not have significantly re-duced Aβ levels in their plasma. However, correlationanalysis revealed not only a strong inverse correlationbetween plasma caffeine levels and plasma Aβ levelsin these aged Tg mice, but also a differential effect ofhigh versus low plasma caffeine levels; aged Tg micewith high plasma caffeine levels had significantly re-duced plasma Aβ levels compared to those with lowplasma caffeine levels. Thus, whether or not a chronicreduction in plasma Aβ levels occurs in aged Tg micewould appear to be dependent on plasma caffeine lev-els. Our elucidation of the same inverse correlation be-tween plasma caffeine and plasma Aβ in young adultTg mice given a single caffeine treatment indicates thatthis relationship transcends: 1) length of caffeine treat-

ment; 2) age of AD mouse recipient; and 3) whether ornot Aβ plaques are present. Indeed, given the fact thatplasma caffeine concentrations in both age groups werewell within physiologic range, it would not be surpris-ing for a similar inverse relationship between plasmacaffeine and plasma Aβ to be present in humans. In thecontext that caffeine suppresses γ-secretase [9,12], thereduced plasma Aβ1−40 levels seen in our aged APPswmice with higher plasma caffeine levels is consistentwith the reported reduction of plasma Aβ levels in ADpatients following a similar 6-week treatment with aγ-secretase inhibitor [29].

For aged APPsw mice, strong correlations were evi-dent between Aβ1−40 and Aβ1−42 levels (both solubleand insoluble) in hippocampus and cortex irrespectiveof caffeine treatment. Moreover, we have found brainAβ levels from aged APPsw mice to be strongly corre-lated with cognitive impairment [12,23–25]. Thus, ourprior work has established an intimate and presumablycausative association between “brain” Aβ levels andcognitive impairment. In sharp contrast, results fromthe present study in the aged APPsw mice clearly showthat no such relationship exists between “plasma” Aβlevels and cognitive performance. Furthermore, plas-ma Aβ levels in these mice were not correlated withany of eight brain Aβ measures. We conclude that, atleast in aged AD transgenic mice, plasma Aβ levelsare not an accurate index of soluble or insoluble brainAβ levels, nor are they an accurate index of cognitiveperformance. This conclusion is consistent with a pre-vious study in old APP transgenic mice [30], as well asa recent report showing that plasma Aβ levels in ADpatients did not correlate with Aβ levels measured inneocortex following death [31] nor with brain Aβ bur-den as measured in vivo with amyloid imaging [32]. Aswith plasma Aβ, we also found no correlations betweenplasma caffeine concentrations and brain Aβ levels orcognitive performance. Thus, higher plasma caffeinelevels were associated with lower plasma Aβ levels,but were not reflective of brain Aβ levels or cognitiveperformance.

From the present studies involving several AD trans-genic lines, it is apparent that baseline plasma levelsof Aβ remain relatively stable during aging and irre-spective of APPsw versus APPsw+PS1 transgenicity(Figs 2 and 4). This finding suggests either close reg-ulation of plasma Aβ in AD transgenic mice or lim-ited/saturable Aβ transport into the blood. Regardingthe latter, we propose that during aging in AD mice,there is a limit to the amount of Aβ that can be trans-ported out of the brain and into plasma. After this lim-



ited transport capacity is reached, any additional Aβwould remain in the brain and “stored” in Aβ depositsthrough the dynamic Aβ equilibrium present with ISFAβ (Fig. 1).

The ability of plasma Aβ levels to be predictive ofimpending AD cognitive impairment in humans is cur-rently unresolved, with some studies reporting that highplasma Aβ1−40 levels are indicative of increased ADrisk [33] and others showing no predictive ability ofplasma Aβ for progression to AD [34]. Results fromthe present study are consistent with the latter of theseassertions in that baseline plasma Aβ levels remainedessentially unchanged during aging in our inbred ADtransgenic mice (compare control Aβ levels in Fig. 2A,B, D, and F), yet these mice develop robust brain Aβpathology and become cognitively impaired as theyage. The question nonetheless remains: Are studiesinvolving the predictive value of plasma Aβ equivo-cal because the assays being utilized are inadequate,or is plasma Aβ simply just not a good biomarker forAD? Indeed, some laboratories are focusing on signal-ing proteins and intercellular communication factors,rather than plasma Aβ, as potential biomarkers for AD.A notable recent study employing this strategy eluci-dated a set of 18 signaling proteins in plasma that col-lectively were highly accurate in classifying subjects asaged normal or AD [35].

Our measurement of plasma caffeine and theo-phylline levels from aged APPsw mice given twomonths of caffeine treatment underscore the humanrelevance and physiologic significance of the caffeinetreatment in this long-term study. At the caffeine con-centration in drinking water that was employed, plasmacaffeine levels averaged 26 µM, equivalent to the plas-ma caffeine concentration expected in humans follow-ing intake of several cups of coffee [3]. This caffeineconcentration is very close to the 20 µM concentrationthat we found to be optimal in our N2a cell culturestudies for suppression of Aβ1−40 and Aβ1−42 pro-duction, although caffeine concentrations in the muchlower nM range were also effective [9]. Predictably,plasma theophylline levels were tightly correlated withplasma caffeine levels. As with plasma caffeine levels,however, no correlations where evident between plasmatheophylline and plasma Aβ or cognitive performance.As one of several active metabolites of caffeine, theo-phylline may be responsible for a significant portionof the cognitive, neuropathologic, and neurochemicalbenefits of caffeine. This premise is underscored by thefact that caffeine metabolites such as theophylline havea considerably longer half-life than caffeine. Indeed,

brain concentrations of theophylline in mice followinglong-term caffeine ingestion are usually higher thanthose of caffeine [36]. We are currently exploring thecontribution of such caffeine metabolites to the benefitscurrently being attributed to caffeine in our AD mice.

No correlations were found between plasma Aβ lev-els and plasma/brain cytokine levels, suggesting theirindependence of one another. Although plasma caf-feine levels were also not correlated with plasma cy-tokines, a clear reduction in hippocampal inflamma-tory cytokines was evident for caffeine-treated micewith higher plasma caffeine levels compared to thosewith lower caffeine levels. This was not only alsotrue for plasma theophylline levels, but higher plas-ma theophylline levels were strongly and consistent-ly correlated with lower inflammatory cytokine levelsin both hippocampus and cortex; these theophyllinefindings suggest a more profound interaction betweenthis caffeine metabolite and brain cytokines than caf-feine itself, perhaps due to the aforementioned higherbrain concentrations of theophylline [36]. The abil-ity of caffeine and theophylline to provide beneficialanti-inflammatory effects in the brain of AD mice isconsistent with a large body of data supportive of theiranti-inflammatory capacities [20] and, as such, couldrepresent a potent “non-amyloidogenic” mechanism ofcaffeine action that may contribute to its ability to pro-tect against/reverse cognitive impairment and synapticdysfunction. Other such beneficial mechanisms of caf-feine action in AD mice (and perhaps in human AD)include the antioxidant actions caffeine [37], its abili-ty to block disruptions of the blood-brain barrier [38],and its well-established antagonism of brain A1 and/orA2a adenosine receptors [1,2,39,40]. Indeed, adeno-sine receptor antagonism may be central to most or allof the beneficial actions of caffeine against AD and anycombination of “caffeinergic” actions may collectivelyprovide the cognitive benefits we have documented inAD mice.

Anti-AD drugs currently on the market (cholineste-rase inhibitors and NMDA receptor antagonists) haveminimal benefits and do not appear to address ADpathogenesis. Although caffeine is the most widelyconsumed psychoactive agent in the world, its intake inWestern countries decreases substantially during aging.Given that: 1) blood caffeine concentrations equilibratealmost instantaneously with brain caffeine levels [3]; 2)caffeine appears to suppress AD pathogenesis [9,12];3) caffeine has few, if any, deleterious side effects formost individuals [10]; and 4) caffeine is readily avail-able and inexpensive, therapeutic actions of this psy-



choactive agent against AD could provide immediatebenefits. Based on the robust protective [9] and treat-ment effects [12] of caffeine that we have observed inAD transgenic mice, Phase II clinical trials involvingacute caffeine administration are currently in progress.To our knowledge, no prior studies have investigatedthe effect of acute or chronic caffeine administrationon plasma biomarkers related to AD. It is importantto note that this report’s acute caffeine administrationstudies involved AD transgenic mice “naıve” to caf-feine, whereas clinical trials involve many occasionalor habitual caffeine users. Therefore, it is difficult topredict the impact of acute caffeine administration inthese ongoing human trials.

To summarize, the present report demonstrates thatcaffeine can acutely decrease brain and plasma Aβ lev-els, as well as reduce brain Aβ deposition and im-prove cognitive function following chronic administra-tion. As such, caffeine appears to have excellent po-tential as a safe and effective therapeutic against AD.We also provide evidence from multiple studies andmethodologies indicating that plasma Aβ is not an ac-curate index of brain Aβ levels/deposition or cognitiveperformance in aged AD mice.

ACKNOWLEDGMENTS

This research was supported by grants to G.W.A. andH.P. within the NIA-designated Florida Alzheimer’sDisease Research Center (P50AG025711), AG029524(J.R.C.), AG13956 (D.M.H.), Cure Alzheimer’s Fund(D.M.H. and J.R.C.), and funds from the ByrdAlzheimer’s Center & Research Institute.

REFERENCES

[1] Fisone G, Borgkvist A, Usiello A (2004) Caffeine as a psy-chomotor stimulant: mechanism of action. Cell Mol Life Sci61, 857-872.

[2] Daly J (2007) Caffeine analogs: biomedical impact. Cell MolLife Sci 64, 2153-2169.

[3] Fredholm B, Battig K, Holmen J, Nehlig A, Zvartau E (1999)Actions of caffeine in the brain with special reference to factorsthat contribute to its widespread use. Pharmacol Rev 51, 83-133.

[4] Higdon J, Frei B (2006) Coffee and health: a review of recenthuman research. Crit Rev Food Sci Nutr 46, 101-123.

[5] Ritchie K, Carriere I, de Mendonca A, Portet F, Dartigues J,Rouaud O, Barberger-Gateau P, Ancelin M (2007) The neuro-protective effects of caffeine: a prospective population study(the Three City Study). Neurology 69, 536-545.

[6] van Gelder B, Buijsse B, Tijhuis M, Kalmijn S, GiampaoliS, Nissinen A, Kromhout D (2007) Coffee consumption isinversely associated with cognitive decline in elderly Europeanmen: the FINE Study. Eur J Clin Nutr 61, 226-232.

[7] Maia L, de Mendonca A (2002) Does caffeine intake protectfrom Alzheimer’s disease? Eur J Neurol 9, 377-382.

[8] Rosso A, Lippa C, Mossey J (2008) Caffeine: Neuroprotec-tive Functions in Cognition and Alzheimer’s Disease. Am JAlzheimers Dis Other Demen 23, 417-422.

[9] Arendash G, Schleif W, Rezai-Zadeh K, Jackson E, ZachariaL, Cracchiolo J, Shippy D, Tan J (2006) Caffeine protectsAlzheimer’s mice against cognitive impairment and reducesbrain beta-amyloid production. Neuroscience 142, 941-952.

[10] Nawrot P, Jordan S, Eastwood J, Rotstein J, Hugenholtz A,Feeley M (2003) Effects of caffeine on human health. FoodAddit Contam 20, 1-30.

[11] Lopez-Garcia E, van Dam R, Li T, Rodriguez-Artalejo F, Hu F(2008) The relationship of coffee consumption with mortality.Ann Intern Med 148, 904-914.

[12] Arendash GW, Mori T, Cao C, Mamcarz M, Runfeldt M, Dick-son A, Rezai-Zadehe K, Tane J, Citron Ba, Lin X, EcheverriaV, Potter H (2009) Caffeine reverses cognitive impairment anddecreases brain amyloid-β levels in aged Alzheimer’s diseasemice. J Alzheimers Dis 17, 661-680.

[13] Bateman R, Munsell L, Morris J, Swarm R, Yarasheski K,Holtzman D (2006) Human amyloid-beta synthesis and clear-ance rates as measured in cerebrospinal fluid in vivo. Nat Med12, 856-861.

[14] Cirrito J, May P, O’Dell M, Taylor J, Parsadanian M, CramerJ, Audia J, Nissen J, Bales K, Paul S, DeMattos R, HoltzmanD (2003) In vivo assessment of brain interstitial fluid withmicrodialysis reveals plaque-associated changes in amyloid-beta metabolism and half-life. J Neurosci 23, 8844-8853.

[15] Cirrito J, Kang J, Lee J, Stewart F, Verges D, Silverio L, Bu G,Mennerick S, Holtzman D (2008) Endocytosis is required forsynaptic activity-dependent release of amyloid-beta in vivo.Neuron 58, 42-51.

[16] Ethell D, Shippy D, Cao C, Cracchiolo J, Runfeldt M, BlakeB, Arendash G (2006) Abeta-specific T-cells reverse cognitivedecline and synaptic loss in Alzheimer’s mice. Neurobiol Dis23, 351-361.

[17] Cracchiolo J, Mori T, Nazian S, Tan J, Potter H, ArendashG (2007) Enhanced cognitive activity–over and above socialor physical activity–is required to protect Alzheimer’s miceagainst cognitive impairment, reduce Abeta deposition, andincrease synaptic immunoreactivity. Neurobiol Learn Mem 88,277-294.

[18] Mori T, Town T, Tan J, Yada N, Horikoshi Y, Yamamoto J,Shimoda T, Kamanaka Y, Tateishi N, Asano T (2006) Arundicacid ameliorates cerebral amyloidosis and gliosis in Alzheimertransgenic mice. J Pharmacol Exp Ther 318, 571-578.

[19] Tan J, Town T, Crawford F, Mori T, DelleDonne A, CrescentiniR, Obregon D, Flavell R, Mullan M (2002) Role of CD40ligand in amyloidosis in transgenic Alzheimer’s mice. NatNeurosci 5, 1288-1293.

[20] Horrigan L, Kelly J, Connor T (2004) Caffeine sup-presses TNF-alpha production via activation of the cyclicAMP/protein kinase A pathway. Int Immunopharmacol 4,1409-1417.

[21] Cirrito J, Yamada K, Finn M, Sloviter S, Bales K, May P,Schoepp D, Paul S, Mennerick S, Holtzman D (2005) Synapticactivity regulates interstitial fluid amyloid-beta levels in vivo.Neuron 48, 913-922.



[22] Arendash G, Gordon M, Diamond D, Austin L, Hatcher J,Jantzen P, DiCarlo G, Wilcock D, Morgan D (2001) Behav-ioral assessment of Alzheimer’s transgenic mice followinglong-term Abeta vaccination: task specificity and correlationsbetween Abeta deposition and spatial memory. DNA Cell Biol20, 737-744.

[23] Arendash G, King D, Gordon M, Morgan D, Hatcher J, HopeC, Diamond D (2001) Progressive, age-related behavioral im-pairments in transgenic mice carrying both mutant amyloidprecursor protein and presenilin-1 transgenes. Brain Res 891,42-53.

[24] Leighty R, Nilsson L, Potter H, Costa D, Low M, Bales K, PaulS, Arendash G (2004) Use of multimetric statistical analysisto characterize and discriminate between the performance offour Alzheimer’s transgenic mouse lines differing in Abetadeposition. Behav Brain Res 153, 107-121.

[25] Arendash G, Jensen M, Salem N, Hussein N, Cracchiolo J,Dickson A, Leighty R, Potter H (2007) A diet high in omega-3fatty acids does not improve or protect cognitive performancein Alzheimer’s transgenic mice. Neuroscience 149, 286-302.

[26] Tuppo E, Arias H (2005) The role of inflammation inAlzheimer’s disease. Int J Biochem Cell Biol 37, 289-305.

[27] Kandimalla K, Curran G, Holasek S, Gilles E, Wengenack T,Poduslo J (2005) Pharmacokinetic analysis of the blood-brainbarrier transport of 125I-amyloid beta protein 40 in wild-typeand Alzheimer’s disease transgenic mice (APP,PS1) and itsimplications for amyloid plaque formation. J Pharmacol ExpTher 313, 1370-1378.

[28] Saido T (2003) Aβ Metabolism and Alzheimer’s Disease(Landes Bioscience).

[29] Siemers E, Quinn J, Kaye J, Farlow M, Porsteinsson A, Tar-iot P, Zoulnouni P, Galvin J, Holtzman D, Knopman D, Sat-terwhite J, Gonzales C, Dean R, May P (2006) Effects of agamma-secretase inhibitor in a randomized study of patientswith Alzheimer disease. Neurology 66, 602-604.

[30] DeMattos R, Bales K, Cummins D, Paul S, Holtzman D (2002)Brain to plasma amyloid-beta efflux: a measure of brain amy-loid burden in a mouse model of Alzheimer’s disease. Science295, 2264-2267.

[31] Freeman S, Raju S, Hyman B, Frosch M, Irizarry M (2007)Plasma Abeta levels do not reflect brain Abeta levels. J Neu-ropathol Exp Neurol 66, 264-271.

[32] Fagan A, Mintun M, Mach R, Lee S, Dence C, Shah A,LaRossa G, Spinner M, Klunk W, Mathis C, DeKosky S, Mor-ris J, Holtzman D (2006) Inverse relation between in vivo amy-loid imaging load and cerebrospinal fluid Abeta42 in humans.Ann Neurol 59, 512-519.

[33] van Oijen M, Hofman A, Soares H, Koudstaal P, Breteler M(2006) Plasma Abeta(1-40) and Abeta(1-42) and the risk ofdementia: a prospective case-cohort study. Lancet Neurol 5,655-660.

[34] Hansson O, Zetterberg H, Blennow K (2008) Evaluation ofplasma Abeta40 and Abeta42 as predictors of conversion toAlzheimer’s disease in patients with mild cognitive impair-ment. Neurobiol Aging, in press.

[35] Ray S, Britschgi M, Herbert C, Takeda-Uchimura Y, BoxerA, Blennow K, Friedman L, Galasko D, Jutel M, Karydas A,Kaye J, Leszek J, Miller B, Minthon L, Quinn J, RabinoviciG, Robinson W, Sabbagh M, So Y, Sparks D, Tabaton M,Tinklenberg J, Yesavage J, Tibshirani R, Wyss-Coray T (2007)Classification and prediction of clinical Alzheimer’s diagnosisbased on plasma signaling proteins. Nat Med 13, 1359-1362.

[36] Johansson B, Georgiev V, Kuosmanen T, Fredholm B (1996)Long-term treatment with some methylxanthines decreases thesusceptibility to bicuculline- and pentylenetetrazol-inducedseizures in mice. Relationship to c-fos expression and receptorbinding. Eur J Neurosci 8, 2447-2458.

[37] Azam S, Hadi N, Khan N, Hadi S (2003) Antioxidant andprooxidant properties of caffeine, theobromine and xanthine.Med Sci Monit 9, BR325-330.

[38] Chen X, Gawryluk J, Wagener J, Ghribi O, Geiger J (2008)Caffeine blocks disruption of blood brain barrier in a rabbitmodel of Alzheimer’s disease. J Neuroinflammation 5, 12.

[39] Dall’Igna O, Fett P, Gomes M, Souza D, Cunha R, Lara D(2007) Caffeine and adenosine A(2a) receptor antagonists pre-vent beta-amyloid (25-35)-induced cognitive deficits in mice.Exp Neurol 203, 241-245.

[40] Cunha G, Canas P, Melo C, Hockemeyer J, Muller C, OliveiraC, Cunha R (2008) Adenosine A2A receptor blockade preventsmemory dysfunction caused by beta-amyloid peptides but notby scopolamine or MK-801. Exp Neurol 210, 776-781.


Caffeine Suppresses Amyloid-β Levels in Plasma and Brain of Alzheimer's Disease Transgenic Mice

Documents

Transcript of Caffeine Suppresses Amyloid-β Levels in Plasma and Brain of Alzheimer's Disease Transgenic Mice