|Home | About | Journals | Submit | Contact Us | Français|
The unique vulnerability of the olfactory system to Alzheimer’s disease (AD) provides a quintessential translational tool for understanding mechanisms of synaptic dysfunction and pathological progression in the disease. Using the Tg2576 mouse model of β-amyloidosis, we show aberrant, hyperactive olfactory network activity begins early in life, prior to detectable behavioral impairments or comparable hippocampal dysfunction and at a time when Aβ deposition is restricted to the olfactory bulb (OB). Hyperactive odor-evoked activity in the piriform cortex (PCX) and increased OB-PCX functional connectivity emerged at a time coinciding with olfactory behavior impairments. This hyperactive activity persisted until later-life when the network converted to a hyporesponsive state. This conversion was Aβ-dependent, as liver-x-receptor agonist treatment to promote Aβ degradation, rescued the hyporesponsive state and olfactory behavior. These data lend evidence to a novel working model of olfactory dysfunction in AD and, complimentary to other recent works, suggest that disease-relevant network dysfunction is highly dynamic and region specific, yet with lasting effects on cognition and behavior.
Alzheimer’s disease (AD) is a progressive brain disease wherein patients suffer from sensory, cognitive, and motor loss (Bishop et al., 2010). Olfactory system function is especially vulnerable to AD. Deficits in odor detection, identification, and recognition occur early in the disease (Mesholam et al., 1998; Murphy, 1999) as opposed to effects on other modalities, such as vision (Gilbert and Murphy, 2004) or cognition (Bacon et al., 1998). These findings are consistent with the concept that AD preferentially impacts specific vulnerable neural networks in the brain (Arnold et al., 1991; Buckner et al., 2009; Seeley et al., 2009). Molecular and network-level mechanisms for this vulnerability remain unresolved yet likely hold major promise in understanding disease progression.
Multiple pathogenic features accumulate throughout the course of AD. The amyloid-β (Aβ) peptide is a particularly potent modulator of synaptic transmission (Kamenetz et al., 2003; Abramov et al., 2009) and thus synchronous neural oscillations (Palop and Mucke, 2010). Neural oscillations are critical for a range of local circuit and regional network functions including sensory coding, sensorimotor integration, synaptic plasticity necessary for information storage, and information transfer between brain regions (Freeman, 1975; Varela et al., 2001; Buzsaki, 2006). In agreement with this, numerous neurological disorders, including AD, exhibit atypical neural synchrony and oscillations (Jeong, 2004; Uhlhaas and Singer, 2006).
Olfactory perception relies upon the hierarchal processing of odor information throughout a distributed network. In this system, the spatiotemporal pattern of odor-evoked activity in the olfactory bulb (OB) is transferred by means of temporally structured spike trains and coinciding synchronous oscillations into highly plastic olfactory cortical areas including the piriform cortex (PCX). From the PCX, odor information travels into the entorhinal cortex and hippocampus for consolidation and future retrieval. Proper information processing within the OB and PCX is critical for the spread of information into higher-order areas and disruption of either structure alone impairs perception (Nusser et al., 2001; Wilson, 2001; Doucette et al., 2007) in manners analogous to that observed in AD (Murphy, 1999). AD pathogenic factors, including Aβ aggregation, are found within the olfactory epithelium, OB, and PCX in both humans with AD and in rodent models (Price et al., 1991; Arnold et al., 2010; Wesson et al., 2010). Interestingly, in the AD rodent models, Aβ first deposits in the OB well before deposition in the PCX, entorhinal cortex, or hippocampus (Wesson et al., 2010). Additionally, the magnitude of olfactory behavioral dysfunction strongly correlates with the level of Aβ deposition (Wesson et al., 2010).
The above findings, along with the ability for Aβ to disrupt synaptic transmission, posits the apparent vulnerability of the olfactory system to Aβ as a major factor underlying olfactory impairments in AD. The impact of Aβ on olfactory network function however remains unresolved. This is important even beyond the scope of understanding olfactory impairments since, through the vast interconnectedness of this network, atypical olfactory activity may induce wide-spread changes in brain function, pathogen spread, and other disruptions in cell function throughout the brain which may exacerbate disease progression (Kapogiannis and Mattson, 2011).
Mice bred and maintained within the Nathan S. Kline Institute for Psychiatric Research animal facility were used. Tg2576 (APP) mice were generated previously (Hsiao et al., 1996) by overexpressing the 695-amino acid isoform of human APP containing the K670N-M671L mutation. Age-matched non-transgenic (WT) littermates on B6SJLF1/J background were used as controls.
Multiple, separate cohorts of mice were used in this study. In all experiments we kept animal numbers to the minimum required for sufficient statistical test power. First, one group of 3-4mo old (7 WT [4 male, 3 female] and 6 APP [3 male, 3 female]]) and 6-7mo old (5 WT [2 male, 3 female] and 5 APP [2 male and 3 female]]) mice were used for behavior (Figure 1A), OB and PCX single-site electrophysiology experiments (Figures 1&2) and Aβ quantification (data not shown). Second, a group of 3-4mo old (6 WT [4 male, 2 female] and 6 APP [5 male, 1 female]) mice were used for hippocampal CA1 electrophysiology (Figure 1G&H). A third group of 3-4mo old (4 WT [2 male, 2 female] and 4 APP [3 male, 1 female]]) and 6-7mo old (5 WT [2 male, 3 female] and 5 APP [2 male and 3 female]]) mice were used for simultaneous OB and PCX recordings (Figure 3). A fourth cohort of mice at 14-16mo of age (8 WT [3 male, 5 female] and 9 APP [5 male, 4 female]) were used for behavior, electrophysiology, GW3965 treatment (see below) and Aβ quantification (Figures 4&5). Finally, a fifth cohort of 14-16mo old mice (6 WT [6 male] and 9 APP [8 male, 1 female]) were used for GW3965 withdrawal behavior experiments.
Mice were genotyped by PCR analysis of tail DNA using standard methods. Confirmatory PCR of tail DNA was performed following experiments to ensure correct classification of groups. All experiments were conducted in accordance with the guidelines of the National Institutes of Health and were approved by the Nathan S. Kline Institute’s Institutional Animal Care and Use Committee.
As we described previously (Wesson et al., 2010), odors (2-heptanone, isoamyl acetate, (+)-limonene, ethyl valerate; Sigma Aldrich, St. Louis, MO) were diluted 1×10−3 in mineral oil and applied to a cotton-applicator stick. Odors were delivered over 4 trials, 20sec each, separated by 30sec intervals in the previously listed order. Snout-oriented sniffing was recorded by a single observer blind to genotypes. Mice were tested in a counter-balanced order. For GW3965 experiments (Figure 5), the same set of odors described above was used to assess olfactory behavior before and after drug or vehicle treatment.
Within 2 weeks following completion of olfactory testing, mice were anesthetized with urethane (1g/kg, i.p.) and positioned in a stereotaxic apparatus for in vivo local field potential (LFP) recordings. The stereotaxic frame was outfitted with a water-filled heating pad to maintain core body temperature (38°C). Mice were supplied local anesthetic (1% xylocaine, s.c.) to the cranial surface and the skin later removed exposing the dorsal skull. Small (~1.5mm diameter) ipsilateral holes were drilled over the anterior piriform cortex (PCX), the olfactory bulb (OB), lateral olfactory tract (LOT), and (in some cases) the hippocampus CA1 region according to (Paxinos and Franklin, 2000). A Vaseline well was created around the skull surface and filled with warm (38°C) 0.9% NaCl. For local recordings, a tungsten stimulating electrode was lowered onto the LOT to aid in PCX localization under physiologic control by LOT stimulation. For data collection, a recording electrode (A-M Systems Inc, Cat # 576000, 0.01”) was lowered first into the PCX (the anterior PCX region) and then after PCX data collection, the electrode was lowered into the mesial aspect of the OB for OB data acquisition. This order was selected to minimize the effects of experience-dependent changes in odor-evoked activity in the PCX (Wilson et al., 2006). For OB-PCX simultaneous recordings, a bipolar stainless steel recording electrode bundle (A-M Systems Inc., Cat # 7915, 0.007”) was lowered into the PCX and another into the mesial aspect of the OB for OB data acquisition. Electrode locations were verified with post-mortem histology using nuclear staining (DAPI) of slide-mounted 40μm coronal sections (as described in Histology). Recording electrode potentials along with stimulus presentation events were acquired using Spike2 software (Cambridge Electronic Design Ltd., Cambridge, England). Recordings across groups were performed in a pseudo-random order.
Odors were presented to anesthetized mice using an air-dilution olfactometer at 1l/min flow using medical grade nitrogen. Stimuli included five monomolecular odorants (1,7-octadiene, 4-methyl-3-penten-2-one, ethyl propionate, heptanal and isoamyl acetate; Sigma Aldrich, St. Louis, MO). All odorants were pure in their liquid state except for 1,7-octadiene and 4-methyl-3-penten-2-one which were diluted 1:1 in mineral oil (all roughly 100ppm concentration). Odors were presented 2sec each, in a cross-habituation paradigm similar to that used for the odor cross-habituation behavioral test (see Behavior). Responses were recorded from both the PCX and OB using the same two odors. The paradigm for odor presentation was (‘stimulus (inter-trial interval)’): odor A (2min) → odor B (2min) → odor A (30sec) → odor A (30sec) → odor A (30sec) → odor A (30sec) → odor B. Thus for each brain region per mouse the first presentation of odor A & B were considered ‘novel odors’ (as used for analysis of basic odor-evoked activity, e.g., Fig 3) and presentations 2-5 for odor A were used to study cellular adaptation (Fig 4). Odor onset was triggered off of the animal’s respiration using a piezo-electric foil placed under the animal’s chest and a window discriminator to detect peaks of respiration (World Precision Instruments, Sarasota, FL).
A subset of 14-16mo old mice were gavaged with either the Liver-X receptor agonist GW3965 (GW, GlaxoSmithKline) [0.033g/kg / mouse/day]; n= 7 WT and 6 APP) or vehicle (veh, dimethyl dulfoxide; n= 6 WT and 5 APP) (Zelcer et al., 2007; Jiang et al., 2008). Several mice died early during treatment, likely due to complications from gavage. Final mouse numbers by study completion were, n= 3 WT and 4 APP (GW) and n= 5 WT and 5 APP (veh). Total volume gavaged was 0.1ml/10g body weight. The average bodyweight across all mice (33.29 ± 0.26g) did not significantly differ across groups (p>0.05, 2-way ANOVA), nor did any one group significantly change in bodyweight over the course of treatment (p>0.05, repeated measures ANOVA). An additional cohort of 14-16mo old mice were treated with GW (n= 6 WT and 4 APP) or veh (n= 4 APP) (same protocol as above) and then allowed to rest untreated for 1 week prior to behavior testing to examine effects of GW withdrawal.
Tissue collection was performed following urethane anesthesia (1g/kg, IP). Mice were decapitated and brains rapidly removed over ice. Hemi-brains were placed in 10% formalin for fixation. A subset of these brains were coronally sectioned (40μm) on a microtome for Aβ analysis and electrode placement verification.
Not all brains were examined for placement of electrodes given a combination of 1) the ease of location [e.g., OB] for electrode guidance and 2) the need for tissue for Aβ analysis versus electrode verification). However, visual inspection of electrode tip locations in Aβ-stained tissue confirmed that electrode placement was accurate. In other tissue, placement analysis occurred following cover-slipping with Vectashield hardmount with DAPI (4′,6-diamidino-2-phenylindole; Vector Laboratories, Inc., Burlingame, CA) for nuclear counter-stain. In all cases electrode tips were within targeted regions (PCX (anterior PCX layers i, ii, or iii), OB, or hippocampus CA1).
Floating sections containing the OB or PCX were stained with either Thioflavin-S or 4G8 (anti-Aβ). For Thioflavin-S staining, floating sections were left standing in 1% filtered Thioflavin-S solution for 10min and then rinsed in ddH2O (x3) for 1min. The remaining sections were left floating in tris-buffered saline (TBS) for 4G8 (anti-Aβ) immunohistochemistry (Mi et al., 2007). Sections were washed (3X @ 5min) in TBS after which they were treated with 85% formic acid for 5min to enhance amyloid staining. Sections were then washed in blocking buffer (0.05 M Tris-HCl pH 7.6, .9% NaCl, 0.25% Triton-X 100, 20% normal goat serum & 0.2% bovine serum albumin) 3X (10min each) before incubating for 12hrs in 4G8 primary antibody at 4°C (Signet Labs, Dedham, MA; 1:200 in blocking buffer). Sections were rinsed (3X @ 5min) in blocking buffer before incubating for 2hrs at room temperature in Alexa488 secondary antibody (Invitrogen, Carlsbad, CA). Finally sections were rinsed in TBS (3X @ 5min), placed on slides and cover-slipped.
A control section (containing the cortex and hippocampus of a 16mo old Tg2576 mouse) was concurrently stained as a positive control (used within all staining wells for GW3965 experiment)(Fig S5). Optical scans of brain sections were collected within 7 days following staining by use of a Zeiss Axioscope microscope (model 200M) and a Zeiss digital camera (Carl Zeiss, Inc., Thornwood, NY). Sections were excited at 488nm and captured with a FITC filter (Fluorescein-5-isothiocyanate).
Human Aβ levels were determined by solid phase sandwich ELISAs to the 1-40 and 1-42 Aβ sequence as previously described (Schmidt et al., 2005) using manufactured ELISA kits (Invitrogen). Ten percent (weight / vol.) homogenates were prepared from flash frozen hemibrains (n≥4 mice/group) lacking the OBs and cerebellum in a homogenization buffer containing 250mM sucrose, 20mM Tris base, 1mM EDTA, 1mM EGTA, and protease inhibitors. DEA (diethylamine) extracted supernatant was used for soluble Aβ quantification, and the remaining DEA pellet sonicated with formic acid (FA) for measures of plaque-associated Aβ. Both soluble and plaque-associated fractions were plated in duplicate along with WT mouse control homogenates and kit-provided Aβ1-40 and Aβ1-42 standards.
For analysis of olfactory behavior data (as described previously (Wesson et al., 2010)), all raw investigatory values (sec) were pooled within animals and organized according to odor presentation (trial) number. As a measure of odor habituation, the raw investigatory values were normalized to the maximum investigatory duration per animal for each odor (max during trials 1–4). These normalized data were selected for analysis due to subtle group differences in trial # 1 odor investigation behavior. The maximum investigation duration was assigned a value of “1” and the lesser investigation times a proportion of 1. These normalized investigation durations of subjects for each odor were analyzed. In some cases (Figure 5B), ‘% habituation’ values were created on trial #4 data as a function of the maximum investigation time (“1” [only data from trial #4 was used in the analysis]).
Histological analysis of Aβ levels was performed in NIH ImageJ (http://rsbweb.nih.gov/ij). Fluorescence levels of anti-Aβ 4G8 and Thioflavin-S were thresholded and regions of interest (ROIs) determined with the guidance of the DAPI counter stain and standard anatomical coordinates (Paxinos and Franklin, 2000). Aβ deposition area (%) was quantified within the OB and PCX. Individual layers in both structures were manually outlined as described previously (Wesson et al., 2010). Aβ deposition (% area) was defined as the cumulative area of fluorescent pixels above threshold within each ROI. At least 3 coronal brain sections (range 3–4) containing each ROI per mouse were used for analysis. Percent area values were in some cases analyzed using one-way ANOVAs for independent groups followed by post-hoc group comparisons using Fisher’s PLSD. Percent area values for each ROI within each section were treated as independent measures for analysis.
Analysis of electrophysiological data was performed in Spike2 (CED, Inc) using Spike2 scripts. In most cases, raw data were processed with an off-line Fast-Fourier transform (FFT) analysis to allow classification of data within components of different LFP frequencies (0.49s Hanning window, 2.035Hz resolution). A power-spectrum was extracted containing the LFP power from 0-100Hz. For odor-evoked data, we normalized the LFP power spectrum before to that during odor presentation to calculate the odor-evoked response magnitude. OB-PCX simultaneous recordings were analyzed by first subtracting the signals of each bipolar electrode to create one OB and one PCX channel which would more closely reflect localized events. From this we then performed waveform correlations or analyzed olfactory LFP coherence (Chabaud et al., 1999) using the COHER openware script freely available at (http://www.ced.co.uk) as described previously (Wilson and Yan, 2010). For the COHER performance, the OB was set as the reference signal. All values are reported as mean ± standard error of the mean (SEM) unless otherwise stated.
We previously reported that in mice overexpressing human mutations of the amyloid precursor protein (APP) (Tg2576) (Hsiao et al., 1996), Aβ first deposits in the OB well before deposition in the PCX, entorhinal cortex, or hippocampus (Wesson et al., 2010). Additionally, the magnitude of olfactory behavioral dysfunction, which emerges early in life, strongly correlates with the level of Aβ deposition (Wesson et al., 2010). Here we employed neural synchrony measures along with olfactory behavioral and pathological assays to examine the influence of progressive β-amyloidosis on olfactory network activity and function.
First, we confirmed olfactory behavioral impairments prior to recordings by testing 3-4 and 6-7mo old wild type (WT) and APP mice in the odor habituation test (Wesson et al., 2010) (n>12/age, >5/genotype). In this test, the time mice spend sniffing odors is measured across successive odor presentations as a gross behavioral index of olfactory function. Odor habituation was similar between WT and APP mice at 3-4mo of age (p>0.05, 2-tailed t-test within each odor trial). However, 6-7mo APP mice displayed statistically less habituation to repeated odor presentation at trials 2 (F(1,49)=9.502; p=0.0034), 3 (F(1,49)=12.21; p=0.001), and 4 (F(1,49)=13.978; p=0.0005) in comparison to WTs (Figure 1A). Given the importance of the PCX and OB to odor habituation (Wilson and Linster, 2008), these results suggest that odor information processing is disrupted in these regions of APP mice and highlight the potential benefits of examining network activity at these two ages (3-4 and 6-7mo), since they uniquely represent time-points wherein olfactory behavior is respectively either functional or dysfunctional.
Next we recorded 10min of spontaneous OB and PCX LFP (local field potential) activity from the same mice used for behavior (Figure 1A) and used Fast-Fourier transforms (FFT) to allow analysis of different LFP frequency components (Figures 1 B&C). Olfactory LFP activity can be organized into theta (1-10Hz), beta (10-35Hz), and gamma (40-70Hz) frequency bands. Olfactory theta activity mostly reflects large synchronous volleys of afferent input from the nasal epithelium (Adrian, 1951). Gamma oscillations reflect local circuit interactions within both the OB and PCX (Rall and Shepherd, 1968; Neville and Haberly, 2003). Beta oscillations are believed to reflect more large-scale network activity (Kopell et al., 2000), for example between the OB and PCX (Neville and Haberly, 2003; Kay and Beshel, 2010). Each of these bands reflect distinct circuit activities contributing unique aspects to odor perception and odor guided behavior (Kay et al., 2009). Importantly, all recordings were performed under anesthesia to control for changes in behavioral state which may exert top-down influences on olfactory function.
At just 3-4mo of age, APP mice display abnormal spontaneous OB and PCX LFP activity, with prominent increases in both beta and gamma band power (Figure 1B, right). This finding is fairly consistent across APP animals, with perhaps greater expression among females (Figures 1C & E). Increases in spontaneous beta and gamma band power were observed within both the OB and PCX in 3-4mo old APP mice compared to age-matched WT (p<0.001, 2-tailed t-tests) (Figures 1D & F). No differences in spontaneous theta power were observed between genotypes at this age. Similar recordings performed from the CA1 of the hippocampus from a separate cohort of 3-4mo mice (n=6/genotype) revealed a near complete preservation of hippocampal LFP activity at this age (p>0.05, within theta, beta, and gamma frequency bands) (Figure 1G). Thus the aberrant spontaneous activity in the OB and PCX is selectively localized.
Surprisingly, the enhanced spontaneous activity in the OB and PCX of 3-4mo APP mice was transient (Figures 1D & F). In 6-7mo old mice, only within PCX gamma band activity was there a significant increase in APP mice (p<0.01). Instead, APP beta and gamma band power significantly decreased from 3-4 to 6-7mo within both the OB (p<0.05, beta, and p<0.001, gamma) and PCX (p<0.001, beta & gamma). Theta band power showed a progressive trend across age within both genotypes; however, no significant differences were observed (p>0.05). To determine if these changes affect the power of specific sub-frequency bands we performed an analysis of peak LFP frequencies within each band. No group differences within either peak theta, beta, or gamma band activity within the OB or PCX were found (p>0.05, data not shown).
One advantage the olfactory system brings to the study of AD is that network function can be probed during the processing of discrete stimuli (odors), thus allowing conclusions regarding behaviorally-relevant information processing. We recorded odor-evoked LFP activity (Figure 2A) in the PCX and then later in the OB (2 odors/region, same mice as in Figure 1). All groups showed significant odor-evoked responses (relative changes) compared to the 2 sec pre-odor (p<0.05, 2-tailed t-tests within frequency bands and groups), however, as detailed below, with significant differences between groups (Figure 2B).
As shown in Figure 2B, heightened odor-evoked beta and gamma band response magnitudes were found in the PCX of 6-7, but not at 3-4mo APP mice. In contrast, while odor-evoked gamma activity in the OB was not affected in any age group of APP mice, odor-evoked beta (in 3-4 and 6-7mo) and theta (in 6-7mo) activity were significantly suppressed in the OB. No group effects were observed within peak LFP bands of odor-evoked recordings (p>0.05, data not shown). Thus, the display of aberrant local circuit and network function at 6-7mo in the PCX corresponds with the emergence of impaired olfactory behavior at this time.
Adaptation to odor is a normal feature of olfactory circuit function which is important to odor perception (Dalton and Wysocki, 1996). Since 6-7mo APP mice show deficient odor habituation behavior, we hypothesized that cortical odor adaptation may be compromised in these animals. To test this we compared the percentage of adaptation in odor-evoked response magnitudes (‘% adaptation’) across 4 trials of the same odor (n=4 odors/animal). 6-7mo APP mice showed enhanced odor-evoked response adaptation specifically within the PCX, but not OB (Figure 2C). This was significant within the beta frequency band (p<0.05, 6-7mo WT v. APP). No other within-age effects were found. Thus, one factor which may contribute to the modified behavioral odor habituation is atypical cortical adaptation.
Information transfer between the OB and PCX is thought to be critically dependent on beta oscillations (Neville and Haberly, 2003; Kay and Beshel, 2010). Disruption of inter-regional communication may provide additional mechanisms for network dysfunction in AD (Buckner et al., 2009). Therefore, we performed simultaneous OB and PCX LFP recordings with bipolar electrodes in an additional cohort of mice (Figure 3A) to provide a test of this hypothesis.
First, we recorded 10min of spontaneous OB-PCX activity to examine resting-state OB-PCX dynamics. While waveform correlation analysis did not detect an effect of age or genotype (data not shown), spectral coherence measures found heightened OB-PCX coherence within beta and gamma band frequencies in 6-7mo APP mice compared to age-matched WTs (Figure 3B)(p<0.05). Similar to that found during spontaneous state, waveform correlation analysis did not detect an effect of age or genotype on odor-evoked data (data not shown). However, spectral coherence of odor-evoked activity within the beta (p<0.05), and also low gamma band (p<0.01), was greatly enhanced within 6-7mo APP mice in comparison to age-matched WTs (Figure 3C). Thus, an additional contributing mechanism to olfactory behavioral dysfunction in APP mice may be impaired, perhaps hyperactive, OB-PCX functional connectivity.
Does hyperactivity in the PCX of 6-7mo APP mice become progressively greater with age? We examined PCX odor-evoked activity in 14-16mo APP and WT mice (same methods used for 3-4 and 6-7mo mice). Strikingly, in 14-16mo APPs we found significantly hypoactive odor responsivity (p<0.05 in beta & gamma separately, WT v. APP) (Figure 4A). Spontaneous PCX activity was not affected in APP mice of this age (p>0.05, 2-tailed t-tests in each band, data not shown). Thus, whereas WT PCX activity is relatively consistent throughout age (3-16mo), odor-evoked activity in APP mice shows a transient dramatic increase which later decreases to hypoactive levels by 14-16mo of age (Figure 4B) (theta p<0.05, beta p<0.001, 6-7mo v. 14-16mo APP, n≥4 mice/group).
To more directly test whether Aβ plays a role in olfactory network dysfunction we employed acute pharmacotherapy with the liver-X receptor (LXR) agonist, GW3965 (GW, (Zelcer et al., 2007; Jiang et al., 2008)) (see Materials & Methods), to promote Aβ proteolytic degradation. Prior to treatment, WT and APP mice were randomly divided into two groups (Figure 5A) for repeated measures testing, one for vehicle (veh) treatment and the other for GW (WT+veh [n=5], APP+veh [n=5], WT+GW [n=3], APP+GW [n=4]) and screened for ‘baseline’ odor habituation behavior (Figure 5B). Both groups of APP mice performed significantly worse than WT controls, showing deficient odor habituation (as measured with ‘% habituation’) (p<0.01 in both groups). Following testing, mice were orally administered either GW or vehicle for 2 weeks. Mice were then again tested for odor habituation behavior. Odor habituation was significantly improved in APP+GW mice versus baseline (p<0.01, 2-tailed t-test) and versus the APP+veh group (p<0.01, 2-tailed t-test)(Figure 5B) reflecting functional improvement of a PCX-dependent olfactory behavior with GW treatment. In an additional cohort of mice we explored whether GW treatment had long-lasting effects after withdrawal by performing baseline odor habituation testing, then administering GW or veh as described above, and then allowing mice to rest untreated for 1 week (see Materials & Methods). Following 1 week of withdrawal, APP+GW mice returned to APP+veh mouse levels (p>0.05, 2-tailed t-test) (data not shown) and were significantly different than WT+GW mice (p<0.05, 2-tailed t-test), suggesting no long-lasting effect of the GW treatment after its withdrawal.
Following the second round of behavior during GW treatment, we performedPCX LFP recordings to explore whether PCX circuit function in APP+GW mice might be restored parallel to the observed recovery of olfactory behavior. Figure 5C shows averaged PCX LFP power spectrograms from APP and WT groups. Average theta, beta, and gamma band powers were calculated within each mouse similar to that performed above (Figures 3--4).4). As mentioned earlier (data not shown), no significant differences in spontaneous activity were found between groups at this age (APP+veh v. APP+GW mice showed a trend in spontaneous beta band activity (p=0.0502, 2-tailed t-test)). Like that shown in Figure 4, we found the PCX of 14-16mo APPs to be hyporesponsive to odors (Figure 5C). No significant effect of odor–presentation was observed in APP+veh mice (p>0.05 in either beta or gamma bands). In contrast, in both WT groups, a significant odor-evoked enhancement of LFP power within the beta band was observed (p<0.01, 2-tailed t-test of odor v. pre-odor, Figure 5C). Remarkably, GW treatment partially restored PCX odor-evoked activity in APP mice toward WT levels, particularly within the beta frequency band (p<0.01, APP+GW v. APP+veh)(Figure 5C).
In confirmation of the efficacy of the GW treatment, we observed GW-dependent reductions in Aβ deposition (Figure 5D & E) and in soluble and insoluble Aβ1-40 and Aβ1-42 as assayed with ELISA (Figure 5F). Together, these data demonstrate that GW-dependent reductions in Aβ (Figure 5D, E, & F), or perhaps modulation of other factors which co-occur with over-accumulation of Aβ in the OB and PCX, can restore olfactory behaviors and the underlying function of the network required to perform the behavior.
Many studies have examined the impact of Aβ on network activity and synaptic transmission (for review see (Selkoe, 2008; Palop and Mucke, 2010)). By probing neural circuit function in the olfactory system of APP transgenic mice we were able to examine the influence of low-levels of Aβ on behaviorally-relevant perturbations in network function throughout several processing centers and during Aβ progression. Our results demonstrate a non-linear emergence of network dysfunction in APP mice and support a novel working model for how the olfactory system becomes disrupted throughout the course of Alzheimer’s β-amyloidosis (Figure 6).
Abnormal network excitability is a prevalent feature of AD (Palop and Mucke, 2010; Santos et al., 2010). Aβ deposition is positively associated with hyperexcitability within the default-network of humans (Sperling et al., 2009) and in the neocortex and hippocampus of aged APP mice (Palop et al., 2007; Busche et al., 2008). In a recent model proposed by Palop and Mucke (Palop and Mucke, 2010), moderate levels of Aβ may enhance presynaptic activity and potentiate synaptic transmission. At high levels though, Aβ may have inverse effects including enhancing long-term depression (LTD) and loss of dendritic spines. Here, in the earliest parts of the olfactory system of APP transgenic mice, we revealed early-life hyperactivity which coincides with only modest levels of Aβ (that found solely in the OB glomerular layer (Wesson et al., 2010)). While in 3mo mice elevations in Aβ likely contribute (via enhancing synaptic transmission at low levels (Puzzo et al., 2008)) to network hyperexcitability which is maintained, if not in some cases enhanced in 6mo old mice, this hyperexcitability is not sufficient to disrupt behavior (at least to the extent detectable by the odor habituation assay). Later in life though (>14mo), the ‘hyperactive’ state subsides, and in effect renders the PCX hypoactive. Indeed, even brief periods of hyperactivity are thought sufficient to elicit a cascade of later-life system hypoactivity and dysfunction (Kapogiannis and Mattson, 2011). Thus, hyperexcitability in the olfactory network emerges at a time-point coinciding with behavioral dysfunction and later recedes leaving a hypoactive system. Importantly, at both stages (hyper- and hypoactive), behavioral dysfunction persists suggesting that even a transient change in local-circuit and/or network-level activity (that observed at 6mo of age) is sufficient to render the system incapable of properly responding to odor information, perhaps through disrupting plasticity among the ensembles required for normal odor coding (Barnes et al., 2008). Similar mechanisms may also underlie the observed spectral coherence results (i.e., Fig 3). Future use of more precise olfactory and behavioral assays may help link specific physiological dysfunction with specific behavioral abnormalities.
We employed acute pharmacotherapy with the Liver-X receptor (LXR) agonist GW3965 to enhance the degradation of Aβ and thereby compare behavior and network activity of APP mice with high versus those with low levels of Aβ. It was recently shown that GW3965 promotes proteolytic degradation of soluble Aβ (Jiang et al., 2008). Similarly, we found that GW administration for 2 weeks reduced Aβ deposition in the OB and PCX, improved behavior, and recovered odor-evoked responsivity in the PCX of aged APP mice. Withdrawal from GW treatment for one week resulted in a return of behavioral impairment. These findings suggest that Aβ-dependent effects on synaptic dysfunction are transient, consistent with the ability to reverse these effects with γ-secretase inhibitors or anti-Aβ methods (e.g., (Janus et al., 2000; Kounnas et al., 2010)). The transient improvement may be associated with changes in cell excitability, synaptic function and/or reversible anatomical changes in synaptic or dendritic spine structure. In addition, while GW3965 lowered Aβ deposition within both the OB and PCX (and across other structures), alternative factors which are impacted by LXR treatment may have contributed to the restoration of behavior and olfactory network function. LXR activation drives macrophages into M2 or alternative activation states that are anti-inflammatory and facilitate phagocytosis (A-Gonzalez et al., 2009). This restoration of phagocytic competence with LXRs is due largely to the transrepression of cytokine expression (Koenigsknecht and Landreth, 2004). An ideal paradigm for future studies would be one wherein Aβ degradation and cytokine levels could be differentially regulated to assess impacts of both factors, independently, in behavior and network activity.
The present results greatly extend our understanding of AD-related olfactory impairments in several notable ways. First, we found that Aβ strongly impacts the activity and function of the OB and PCX. While PCX activity is clearly compromised in AD (e.g., (Li et al., 2010)), the OB’s role has remained elusive. Our results demonstrate that the OB may be a major culprit behind olfactory dysfunction in AD. Early-life aberrant activity in the OB activity may disrupt normal coding of odors as they spread into the PCX, thereby resulting in PCX disruption. Alternatively, given the relevance of synaptic activity in the postsynaptic upregulation of Aβ levels (Cirrito et al., 2005) and its spread between regions (Lazarov et al., 2002), these data suggest that early-life hyperactivity within the OB may effectively contribute to dysfunction in the PCX as observed here and elsewhere in humans (Li et al., 2010) by enhancing pathological accumulation of Aβ. This accumulation may ultimately elicit synaptic loss (Spires et al., 2005) and thus result in a hypo-responsive network as indeed we observed later in life. Second, our work shows that olfactory network dysfunction appears early in life, prior to the appearance of significant behavioral impairments. These findings suggest that more refined methods to examine olfactory network activity clinically may serve to identify prodromal AD populations. Additionally, the improvement of olfactory behavioral and network function by enhancing Aβ proteolysis suggests that olfactory loss may be reversible. Future clinical studies examining whether similar methods to reduce Aβ can prevent olfactory loss entirely or even just reduce it will be important in understanding the mechanisms of sensory loss in AD and other Aβ-related disorders.
The results of these studies provide novel evidence for a relationship between the onset of sensory dysfunction commonly reported in AD and early stage sensory network hyperactivity. These findings lend evidence to a model whereby early-life elevations in Aβ are associated with dramatic increases in high-frequency network oscillatory activity which, due to its own persistent activity, contributes to subsequent hyporesponsivity and dysfunction later in life (Figure 6). In light of these results it is interesting to consider that olfactory dysfunction may exacerbate AD through activity-dependent modulation of Aβ and possibly other pathogens. Future studies examining the role of other APP-related neurotoxic factors in combination with in vivo neural network activity measures will be valuable in establishing a complete understanding of the basis of neural dysfunction in AD.
We thank Julie Chapuis for helpful comments on an earlier version of this manuscript, Paige Cramer and Paul Mathews for advice on ELISAs, and Adrienn Varga-Wesson for assistance performing ELISAs. This work was supported by National Institutes of Health grants DC003906 to D.A.W., AG037693 to D.A.W, E.L., R.A.N., AG017617 to R.A.N., and AG030482 to G.E.L.