Reversal of Fragile X Phenotypes by Manipulation of AβPP/Aβ Levels in Fmr1KO Mice

Fragile X syndrome (FXS) is the most common form of inherited intellectual disability and the leading known genetic cause of autism. Fragile X mental retardation protein (FMRP), which is absent or expressed at substantially reduced levels in FXS, binds to and controls the postsynaptic translation of amyloid β-protein precursor (AβPP) mRNA. Cleavage of AβPP can produce β-amyloid (Aβ), a 39–43 amino acid peptide mis-expressed in Alzheimer's disease (AD) and Down syndrome (DS). Aβ is over-expressed in the brain of Fmr1KO mice, suggesting a pathogenic role in FXS. To determine if genetic reduction of AβPP/Aβ rescues characteristic FXS phenotypes, we assessed audiogenic seizures (AGS), anxiety, the ratio of mature versus immature dendritic spines and metabotropic glutamate receptor (mGluR)-mediated long-term depression (LTD) in Fmr1KO mice after removal of one App allele. All of these phenotypes were partially or completely reverted to normal. Plasma Aβ1–42 was significantly reduced in full-mutation FXS males compared to age-matched controls while cortical and hippocampal levels were somewhat increased, suggesting that Aβ is sequestered in the brain. Evolving therapies directed at reducing Aβ in AD may be applicable to FXS and Aβ may serve as a plasma-based biomarker to facilitate disease diagnosis or assess therapeutic efficacy.


Introduction
FXS is an X chromosome-linked disorder characterized by highly variable intellectual disability, autistic-like behavior and seizures [1]. In the vast majority of cases, FXS results from a .200 trinucleotide (CGG) repeat expansion in the 59-UTR of the FMR1 gene [2] leading to transcriptional silencing and loss or reduction of expression of FMRP [3]. FMRP is a multi-functional mRNA binding protein involved in the dendritic transport, localization and translational regulation of several hundred mRNA ligands [4][5][6][7][8][9]. In the absence of FMRP, dendritic spine morphology and function are abnormal [10]. Thus, FXS is likely caused by the inappropriate, post-synaptic expression of one or more FMRP mRNA targets. We have previously demonstrated that postsynaptic translation of App mRNA is regulated by FMRP through a mGluR 5 -dependent pathway. In the absence of FMRP, excess AbPP and its catabolites Ab  and Ab  , accumulate in the brains of middle-aged Fmr1 KO mice [11].
There is very limited data regarding the roles of AbPP mRNA, protein and catabolites in persons with FXS. One group showed elevated App mRNA in the cerebral cortex, hippocampus and cerebellar cortex in Fmr1 KO mice [12], but we have not observed differences in cortical synaptoneurosomes [11]. Increased Ab levels would predict an increased incidence of AD pathology in aged FXS individuals, which has not been observed in neuropathological analyses of a very small number of specimens [13][14][15]. Likewise, there are no reports of an increased incidence of AD in FXS, but it is difficult to assess age-related dementia in the mentally retarded and very few elderly individuals with FXS have been studied in significant numbers. Recent data suggests that AbPP or its proteolytic derivatives may be aberrantly expressed in children with severe autism [16,17], which is extremely prevalent in FXS (67% of males and 23% of females) [18] and DS (7%) [19]. Thus, the increased production or altered processing of AbPP may contribute to the intellectual disabilities observed in all of these disorders.
Herein, we demonstrate rescue of several Fmr1 KO phenotypes in model mice by genetic modulation of AbPP/Ab levels. Treatment of primary neurons with Ab 1-42 rapidly triggered extracellularregulated kinase (ERK) signaling and altered the translation of multiple FMRP target mRNAs including App. Conversely, anti-Ab antibody decreased dendritic AbPP levels. These data suggest that Ab modulates its own production through a positive feedback loop. Finally, we show that plasma Ab 1-42 is significantly reduced while cerebral Ab 1-42 is likely increased in full-mutation FXS males. These data suggest evolving therapies directed at reducing Ab in AD may be applicable to FXS and plasma Ab 1-42 may be a biomarker for disease severity and drug efficacy in FXS.

Results
Genetic Modulation of AbPP Levels in Fmr1 KO Mice Reduces AGS Compared to WT, Fmr1 KO mice exhibit AGS [20], altered anxiety [21][22][23] and dendritic spine dysmorphogenesis [24][25][26], features shared with patients with FXS [1,27,28]. In order to establish if AbPP or Ab directly contributed to FXS pathogenesis, we modulated AbPP and Ab expression in Fmr1 KO mice. Thus, we generated Fmr1 KO /App HET and Fmr1 KO /App KO mice all in a pure C57BL/6 background to evaluate the effects of genetic modulation of AbPP and Ab expression. Western blot analyses confirmed that AbPP levels were reduced by 50% in Fmr1 KO /App HET mice ( Figure 1A). At 21 days of age, animals were evaluated for AGS. Wild running (WR) and seizures decreased 32% and 54%, respectively, in Fmr1 KO /App HET mice ( Figure 1B). These data suggest that seizures are enhanced when AbPP is absent or overexpressed and that AbPP or one of its metabolites significantly contributes to the AGS phenotype seen in Fmr1 KO mice.

Fmr1 KO /App HET Mice Lack FXS Behavioral, Dendritic Spine and mGluR-LTD Phenotypes
Hyperactivity, social anxiety and autistic-like behaviors are characteristic features of FXS [1]. We first assessed marble burying in these mice as a measure of repetitive behavior [32]. Fmr1 KO male mice buried significantly fewer marbles than WT (p = 0.04) (Figure 2A), which was rescued in male Fmr1 KO /App HET mice (statistically different compared to Fmr1 KO , p = 0.03; not different from WT, p = 0.95). Thus, repetitive digging is a normal mouse behavior that can be rescued in Fmr1 KO male mice by genetically reducing AbPP/Ab levels.
Next, we assessed hyperactivity ( Figure 2B) and anxiety ( Figure 2C, D) in the open field test. Fmr1 KO exhibit increased center time (reflecting a loss of anxiety) but equivalent locomotion in the open field compared with WT mice [21,22]. App HET and App KO mice exhibited significantly more anxiety/thigmotaxis than WT mice ( Figure 2C). The increased thigmotaxis in the App HET is partially due to decreased locomotion compared with WT mice ( Figure 2B); however, total distance traveled was equivalent between App KO and App HET indicating that decreased AbPP /Ab levels correlate with increased anxiety. Fmr1 KO /App HET mice exhibited equivalent center time as WT mice and significantly less than Fmr1 KO mice indicating that thigmotaxis was rescued. Total distance traveled in the open field was not statistically different between Fmr1 KO /App HET , WT and/or Fmr1 KO mice indicating equivalent locomotion.
Pathological examination of brains from FXS patients has shown an increased density of long and tortuous dendritic spines suggesting a delay in spine maturation [15,33]. We assessed dendritic spine length in primary cultured neurons and found a statistically significant 1.6-fold increase in protrusion length in the Fmr1 KO cells compared to WT ( Figure 3A, B) in agreement with the literature [24][25][26]. A 4 hr treatment with mGluR 5 antagonists (MPEP or fenobam) rescued the spine/filopodia ratio in Fmr1 KO mice to WT levels [26]. Likewise, we observed that both dendritic spine length and the percentage of filopodia in primary cultured Fmr1 KO neurons were reverted to WT levels within 15 min of MPEP treatment ( Figure 3B). Average protrusion length was reduced by 11% in the Fmr1 KO /App HET cells (statistically significant p,0.03 compared to Fmr1 KO ), and the percentage of filopodia (immature spines) versus mature spines was completely reverted to WT levels. Similarly, minocycline treatment of Fmr1 KO neurons has been reported to rescue the percentage of mature spines to WT levels albeit without a significant change in dendritic protrusion length [24].
Fmr1 KO mice exhibit enhanced hippocampal mGluR-long term depression (LTD), which requires rapid protein synthesis [34,35]. We assessed hippocampal mGluR-LTD by field recordings in 3month-old male WT, Fmr1 KO , App HET and Fmr1 KO /App HET mice. mGluR-dependent synaptic depression was enhanced in the CA1 in Fmr1 KO slices, equivalent in WT and App HET and reduced in Fmr1 KO /App HET (statistically significant p,0.0002 comparing Fmr1 KO /App HET and Fmr1 KO ) ( Figure 4). The time course of CA1 fEPSP slopes after DHPG treatment reveals a significant difference between Fmr1 KO animals relative to wild type, App HET and Fmr1 KO /App HET mice. Input/output relationships and paired-pulse facilitation (ppf) were analyzed to assess the basal synaptic properties of the different groups ( Figure S1), and did not show significant changes between experimental groups indicating unaltered synaptic transmission in the Fmr1 KO hippocampus.

Ab 1-42 Alters Dendritic Expression of FMRP Targets
To begin to understand how excessive production and/or processing of AbPP mediates cell signaling events, we examined dendritic expression of select FMRP targets after treating WT primary neurons with soluble Ab 1-42 . There was a .2-fold increase in the expression of AbPP, no change in Arc, 40% increase in Map1B, 85% increase in RhoB and 50% decrease in PSD95 ( Figure 5A). The Ab 1-42 -mediated increase in dendritic AbPP could be blocked with MPEP or anisomycin indicating that it was mGluR 5 -and translation-dependent ( Figure 5B). Conversely, reduction of Ab in the cell culture media by treatment with anti-Ab antibody through a transwell reduced dendritic AbPP expression by 31% ( Figure 5C). We utilized transwells to avoid direct contact between the antibody and the cells as direct application of anti-Ab to the culture media dramatically increased AbPP expression presumably due to cell signaling events initiated by anti-Ab binding to cell surface receptors (data not shown). MPEP and Ab 1-42 altered phosphorylated levels of ERK ( Figure  S2), suggesting that the mitogen activated protein kinase pathway is as an intermediate in Ab-mediated AbPP translation.

Ab 1-42 Levels are Abnormal in Humans with FXS
We are unaware of any published data evaluating AbPP or its products in the blood of adult FXS patients. We found that plasma AbPP/AbPPa and Ab 1-40 levels were comparable in adult FXS patients and controls ( Figure 6A, B). These results are distinct from those previously observed in children [16] and suggest that AbPP expression and processing decrease with age. Indeed, AbPPa levels are higher in children 7 years and younger than those 10 years and older [16]. While plasma Ab 1-40 levels were unchanged between FXS and controls, Ab 1-42 was significantly lower in the FXS group (2.1-fold decrease, p,0.004) ( Figure 6B). As seen in other amyloidogenic diseases, the Ab 1-42 /Ab 1-40 ratio (1.4:1) in blood plasma was substantially reduced compared to controls (3.4:1) suggesting decreased clearance from the brain. There was no statistically significant increases in cell-associated AbPP, secreted AbPPa nor APP 695/751/770 mRNA in peripheral blood mononu-clear cells (PBMC) samples from FXS and control donors ( Figure 6A and data not shown).
We next assessed AbPP/AbPPa, Ab 1-40 and Ab 1-42 in hippocampal and neocortical control and FXS autopsy brain tissue. The data represents analyses of four FXS and three control brains. There is a strong trend toward increased soluble Ab 1-40 in lysates of FXS brain samples ( Figure 6D) while AbPP/AbPPa levels are reciprocally decreased ( Figure 6C) in both the neocortex and hippocampus. Despite the small sample size due to the unavailability of tissue, the 1.7-fold decrease in AbPP/AbPPa in neocortex approaches statistical significance. These data suggest that the brain may act as a sink for Ab and that lower blood plasma levels may indicate increased brain deposition. These results are similar to those in Fmr1 KO mice, which exhibited elevated Ab in the brain [11]. Murine Ab levels in blood plasma were below the ELISA detection limit and could not be determined.
These data strongly support the hypothesis that modest overexpression of AbPP and/or Ab, in the context of the Fmr1 KO , is necessary for many of the pathological phenotypes including AGS, anxiety, dendritic dysmorphogenesis and mGluR-LTD observed in the mice and that these effects are likely mediated by mGluR 5 signaling. Furthermore, these data suggest a positive feedback loop whereby extracellular Ab stimulates dendritic translation of AbPP   through a mGluR 5 signaling pathway providing more target for amyloidogenic processing and the generation of additional Ab.

Discussion
FXS is the most common form of inherited mental retardation and autism. It is caused by the loss of FMRP, an mRNA binding protein, which localizes to dendrites and regulates protein synthesis. Considerable effort has focused on characterizing the FMRP mRNA ligands and signaling pathways, particularly mGluR 5 [36], that contribute to FXS phenotypes. We have demonstrated that App mRNA is an FMRP mRNA ligand whose translation is regulated through mGluR 5 [11].
Maximal AbPP expression coincides with the critical period of sensory development in rodents (postnatal weeks 2-6) [43]. FMRP is also developmentally regulated in the neonatal brain where it peaks at the end of the first postnatal week and declines thereafter [44]. AbPP is processed by a-, band/or c-secretases to produce soluble N-terminal domains of AbPP (sAbPPa and sAbPPb), Ab and C-terminal fragments. Ab, which is over-expressed in AD and DS [45], is strongly implicated in impaired synaptic function and synapse loss observed early in the development of AD [46,47], and we propose that their dysregulated production in FXS contributes to disease pathology. From conception, DS individuals overexpress APP mRNA, AbPP and extracellular amyloid [48,49]. Like FXS, DS patients show mental retardation, craniofacial abnormalities and dendritic dysmorphogenesis [1,50]. Thus, increased AbPP and/or AbPP proteolytic products could provide a common effector at the molecular level for the neuroanatomic and behavioral phenotypes observed in all of these disorders [51]. We found that the peripheral concentration of Ab 1-42 and the Ab 1-42 /Ab 1-40 ratio were significantly decreased in full-mutation FXS males compared to control donors. A reduced Ab 1-42 /Ab 1-40 ratio is an independent risk factor for AD irrespective of the total Ab load [52][53][54][55][56][57][58]. Plasma Ab 1-42 levels are increased in patients with mild cognitive impairment, but drop to control levels by the time of AD diagnosis [59]. In DS, elevated plasma Ab 1-42 is associated with earlier onset of AD [60] and the Ab 1-42 /Ab 1-40 blood plasma ratio is lower than in controls [61]. Thus, our finding of a reduced Ab 1-42 /Ab 1-40 ratio in FXS patients compared to control plasma is consistent with other amyloidogenic diseases.
The effectiveness of drug therapy in FXS is currently assessed exclusively by behavioral testing. Perhaps most importantly, the availability of a plasma biomarker for FXS may permit the monitoring of drug therapy as well as predict disease progression. Other studies have demonstrated reduced cAMP production in platelets [62] and delayed early-phase phosphorylation of ERK in lymphocytes [63] from FXS patients. Our data demonstrates that Ab 1-42 is significantly lower in FXS plasma than control subjects, and in conjunction with two studies demonstrating elevated sAbPPa in autistic children [16,17], suggests that both AbPP and Ab are viable biomarkers for FXS. We did not observe increased sAbPPa in adult FXS plasma suggesting that AbPP expression and processing decrease with age consistent with prior studies [16]. Previously reported control plasma levels of Ab range from 130-208 pg/mL for Ab 1-40 and 15 pg/mL-85.7 ng/mL for Ab 1-42 [53,58,64,65]. Thus, there is wide range of reported plasma Ab 1-40 and Ab 1-42 levels in the literature and our studies fall within that range. A critical question remains regarding if these catabolites are associated with disease severity or progression. There is a paucity of human FXS brain tissue available for analyses. Our preliminary studies indicate a trend for elevated Ab in FXS brain, which agrees with data in Fmr1 KO mice [11].
FXS patients exhibit hyper-reactivity to visual, olfactory, tactile and auditory stimulation [1,[66][67][68][69]. This hypersensitivity phenotype is manifested as AGS in Fmr1 KO mice [20]. If Ab contributes to AGS, then AD and DS mice would also be susceptible to seizures. Consistent with this, Tg2576, FRAXAD and DS mice, which all over-express hAbPP with the Swedish familial mutation and/or mouse AbPP, exhibit AGS [70]. Fmr1 KO /App HET male mice exhibited a statistically significant decrease in AGS but not in WR whereas MPEP significantly reduced WR in Fmr1 KO mice. These data suggest that the reduction of AbPP/Ab in Fmr1 KO mice is not particularly effective at reducing the induction of AGS, but does retard progression to clonic-tonic seizures. Fmr1 KO mice also exhibit enhanced mGluR-LTD [34], which is lost in Fmr1 KO / App HET mice. The Fmr1 KO reduces mGluR-LTD in an App HET background, but enhances mGluR-LTD in an App WT background. The large difference in the maximal depression of synaptic transmission between Fmr1 KO /App HET and Fmr1 KO in opposite directions from WT mice, which exhibit equivalent mGluR-LTD as App HET , suggests that FMRP and AbPP/Ab play important and synergistic roles in modulating mGluR-LTD. Thus, the overexpression of AbPP or an AbPP catabolite lowers seizure threshold and enhances mGluR-LTD, and approaches to attain normal synaptic levels of these proteins could prove therapeutic.
MPEP is a potent and highly selective noncompetitive antagonist of mGluR 5 [71,72] that reduces AGS, anxiety phenotypes and dendritic spine protrusion morphology in Fmr1 KO [22,26]. We have previously demonstrated that mGluR 5 blockade inhibits translation of AbPP in synaptoneurosomes [11] and herein demonstrate reversion of several FXS phenotypes by genetic manipulation to reduce AbPP/Ab. Minocycline, a secondgeneration tetracycline compound, reverts several FXS phenotypes [24], possibly by altered Ab fibril formation [73] or Abinduced neuronal death and glial activation [74]. In aggregate, these data suggest that mGluR 5 antagonists and minocycline therapies converge on a similar signaling pathway resulting in decreased Ab levels/activity as obligatory for the rescue of FXS phenotypes.
How Ab mediates synaptic dysfunction remains unclear. Ab promotes AD-like cytoskeletal abnormalities and can promote intracellular accumulation of sAbPP in primary cultured neurons [75]. Soluble oligomers of Ab increase LTD in WT hippocampal slices and inhibit long-term potentiation [76][77][78][79][80]. A similar enhancement of mGluR 5 -mediated LTD occurs in the hippocampus of Fmr1 KO mice [34], and MPEP prevents the block in long-term potentiation [81]. Ab causes membrane depolarization and calcium influx, activates mGluR 1 [82] and functions as an extracellular scaffold for mGluR 5 [83]. In vitro, Ab 1-42 altered the expression of important dendritic proteins regulated by FMRP including AbPP, Map1B and PSD95 with known roles in synaptogenesis and/or dendritic spine morphology [37,38,44,84]. A previous report utilizing neuronal hybrid cells demonstrated that Ab 1-40 increases AbPP levels also supporting the existence of an Ab-driven positive feedback loop [85]. Lower levels of PSD95 are observed in Tg2576 neurons, which constitutively overexpress Ab and ADDLs [84]. In our hands, levels of the immediate-early gene Arc were not changed during a 1 hr stimulation with low molecular weight oligomers of Ab 1-42 ; however, others have observed that higher molecular weight Ab oligomers (10-100 kDa) bind in a punctate pattern to the surface of neurons, colocalize with PSD95 and upregulate Arc [47]. We can block the Abinduced increase in AbPP levels with either MPEP or anisomycin suggesting that an mGluR 5 -and translation-dependent pathway is involved. The varied expression of known FMRP targets in response to Ab suggests that dendritic translation can be modulated through FMRP-dependent and independent pathways. Ab 1-42 rapidly increases phosphorylated ERK levels suggesting that ERK mediates downstream signaling.
In conclusion, our work demonstrates that AbPP translation is regulated through an mGluR 5 /FMRP-mediated pathway. Excessive signaling through mGluR 5 in the absence of FMRP leads to increased AbPP production and processing and we have observed elevated AbPP and Ab levels in Fmr1 KO mice [11]. Genetic reduction of AbPP levels in Fmr1 KO mice has reverted or partially rescued FXS seizure, behavioral, dendritic spine and mGluR-LTD phenotypes. Furthermore, treatment of primary neurons with Ab 1-42 increased while anti-Ab reduced dendritic AbPP expression suggesting that an Ab-driven positive feedback loop drives synthesis/processing of AbPP through a mGluR 5 signaling pathway. Our results have potential implications for the treatment of FXS as plasma AbPP/Ab can be readily screened as biomarkers to evaluate potential therapies including mGluR 5 antagonists as well as secretase inhibitors and anti-Ab, which are currently undergoing testing for the treatment AD.

Ethics Statement
Adequate measures were taken to minimize pain or discomfort to the mice, and all husbandry, seizure and euthanasia procedures were performed in accordance with NIH and an approved University of Wisconsin-Madison animal care protocol administered through their Research Animal Resources Center (approval #G00468). Males with FXS and age-matched controls were recruited from the FXS Clinic at Rush University Medical Center (RUMC) in Chicago, IL. The study was approved by the RUMC Institutional Review Board and all donors or their legal guardians signed the appropriate consent forms for study participation.

Mouse Husbandry
WT and Fmr1 KO mice (C57BL/6 background) were bred and housed as previously described) [11]. Fmr1 KO females [86] were crossed with App KO males [87] (Jackson Laboratories #004133, C57BL/6 background) to generate Fmr1 HET /App HET females and Fmr1 KO /App HET males that were crossed to generate Fmr1 KO /App KO mice. For the behavioral testing, littermate controls were generated by crossing Fmr1 KO /App HET females with Fmr1 KO / App HET males to generate Fmr1 KO /App WT , Fmr1 KO /App HET and

Audiogenic Seizures
All mouse strains were tested at postnatal day 21, the peak of AGS sensitivity in C57BL/6. The experimental apparatus consisted of a clear, Plexiglas box (130L680W670H) with the sound source located inside the box (LOUD KEY TM jogger's alarm). Mice were weighed and then placed individually into the center of the chamber and exposed to a siren that generated noise at 118 dB for 5 min. Loud, acoustic stimulation causes WR within 20-30 sec followed by erratic leaping, clonic convulsions and tonic hind limb extension by 40-50 sec followed by respiratory arrest and death [32]. The percentage of mice exhibiting WR, AGS and death were scored versus gender and genotype and assessed for statistical significance by the Fisher exact test. The Fmr1 KO /App HET mice used in the AGS studies were offspring generated by crossing Fmr1 KO females with Fmr1 KO /App KO males to avoid effects due to maternal genotype in comparing Fmr1 KO and Fmr1 KO /App HET . MPEP was a kind gift from FRAXA Research Foundation (Newburyport, MA) and was dissolved at 1 mg/mL in DPBS before I.P. injection at 30 mg/kg body weight 30 min prior to AGS testing at age P21.

Marble Burying
Mice were acclimated to the behavioral testing room for at least 15 min prior to transfer to a clean cage containing corn cob bedding and 20 black marbles arranged in a rectangular 465 grid over 2/3 of the cage. The mouse was placed into the cage at the end that did not contain marbles and allowed to explore the new cage with the marbles for 30 min after which the mouse was returned to its homecage and the number of visible marbles (more than half not buried) counted. Background white noise in the room was set to 70 dB. There is variability in this assay with some laboratories observing that Fmr1 KO mice bury more marbles than WT (R. Paylor, personal communication); however, differences in behavioral outcomes can vary between facilities due to genetic background or altered environmental factors such as housing conditions and diet.

Open Field
Mice were acclimated to the behavioral testing room for at least 30 min before placement into the center of a clear, Plexiglas chamber measuring 14.50L614.50W6100H. Mouse movement was monitored in the chamber for 15 minutes with LimeLight2 software interfaced with an overhead camera. The open field arena was arranged into a 16 square grid in the camera window with 4 squares in the center and 12 squares around the perimeter. Time spent in the center 4 squares (measure of anxiety) and total distance traveled (measure of locomotion) were compared between genotypes.

DiI Labeling and Analyses of Dendritic Spines
Primary mouse neurons were prepared from embryonic (age E15-17) dissected brains from timed pregnant WT and Fmr1 KO female mice as previously described) [11]. To generate Fmr1 KO / App HET cells, Fmr1 KO females were mated with Fmr1 KO /App KO males. Cells were cultured for 15 days on poly(D)-lysine coated glass coverslips inside of 12-well tissue culture dishes, fixed with 4% paraformaldehyde and stained with lipophilic DiI dye (Life Technologies Corporation, Carlsbad, CA, USA). For the staining, the wells were aspirated and sprinkled with DiI crystals and a small amount of DBPS was added to the edge of the wells to prevent dehydration of the cells. Cells were stained for 10 min, copiously washed with DPBS to remove all crystals and fixed to slides with ProLong Gold Antifade (Life Technologies Corporation, Carlsbad, CA, USA). Slides were allowed to dry for at least 3 days to allow complete migration of the DiI into dendritic spines. Dendritic spines were imaged on a Zeiss Axioplan 2 Imaging Photomicroscope equipped with a MBF Biosciences automated XYZ stage and MicroFire A/R camera. Images were taken using the 1006 objective (Zeiss FLUAR 1006/1.30 oil) and Zeiss Immersol TM 518F oil at ambient temperature. Spine length was quantitated with StereoInvestigator v9 software. Contours were drawn around the protrusions and the feret max (length) and feret min (widest width) of the contours were calculated. A minimum of 2-6 coverslips were analyzed per neuronal cell prep and images of neurons were taken from multiple areas of those coverslips. Data is representative of multiple batches of neuronal cells. A minimum of 746 spines were quantitated per genotype. The feret width was divided by feret max and protrusions having a ratio less than 0.5 were classified as filopodia and those with a ratio greater than or equal to 0.5 were classified as spines. For MPEP treatments, 2.5 mM MPEP was added to the neurons for the indicated times followed by washing the cells with DPBS and fixation and staining as previously described. A minimum of 263 protrusions were quantitated per MPEP treatment.

Hippocampal Slice Preparation and mGluR-LTD Field Recordings
WT, Fmr1 KO , App HET and Fmr1 KO /App HET mice were rapidly decapitated by cervical dislocation (3-month old males). Hippocampal slices were prepared and electrophysiology performed as previously described [88]. After a 50 min stable baseline, slices were bathed in 50 mM S-DHPG for 10 min to induce mGluR-LTD. A 2 min trace of an average of four 30 sec traces was recorded over 100 min post-drug application and the slope of the fEPSP was measured and graphed as a function of time.

Blood Collection
Males with FXS and age-matched controls were recruited from the FXS Clinic at Rush University Medical Center (RUMC) in Chicago, IL. All FXS subjects (ages 9-32 years old) were positive by DNA analyses for a fully methylated expansion mutation in the FMR1 gene. Controls (age 23-33) were normal volunteers working at RUMC and had no history of cognitive or mental health disorders. The age and medications of the donors are listed in Table S1. Blood was drawn from donors into lithium heparincoated blood collection tubes and spun at 1,500 rpm. The plasma supernatant was removed and frozen at 280uC. The anticoagulated blood was mailed by overnight delivery from RUMC to the University of Wisconsin-Madison where PBMC were isolated within 24 hr.
Assessment of AbPP/AbPPa, Ab 1-40 and Ab 1-42 by ELISA Plasma was thawed and clarified at 12,000 rpm for 10 min at 4uC prior to ELISA assays for AbPP/AbPPa, Ab 1-40 and Ab 1-42 per the manufacturer's instructions (BioSource #KHB0051, KHB3482, KHB3442) with the following modifications for the Ab assays: (1) the sample volume was doubled from 50 mL to 100 mL, (2) the incubation time was extended from 3 hr to overnight at 4uC, and (3) after the overnight incubation, the samples were removed from the antibody-coated wells prior to addition of the detection antibody. PBMC were isolated as previously described [91] and cultured for 24 hr prior to harvesting the cells and culture media for ELISA analyses. Hippocampus and neocortex (pre and post central gyri) samples were sectioned from left cerebral hemispheres of control and FXS brain autopsy tissue. Detergent-soluble lysates were prepared as previously described) [11] for analyses of AbPP and Ab 1-40 .

Statistical Analyses
One-way ANOVA was performed using GraphPad Prism version 5.0 d for Mac OS X (GraphPad Software, San Diego, CA) to compare the means of three or more unmatched groups for the behavioral and dendritic spine analyses. Student T-test analyses were used to quantitate statistical significance for the western blotting, marble burying, open field, dendritic spine protrusion length, immunofluoresence and ELISA data. Fisher exact tests were used to quantitate statistical significance for the AGS data. Two-way ANOVA with Bonferroni posthoc tests was used to quantitate statistical significance for the mGluR-LTD analyses.

Supporting Information
Figure S1 Assessment of hippocampal slice health. Hippocampal slices from WT (black circles), Fmr1 KO (white circles), App HET (black triangles) and Fmr1 KO /App HET (white triangles) mice exhibit equivalent synaptic transmission as determined by the input/output relationship (A) and ppf (B). i/o was also measured at the end of the recordings to assess slice health and showed similar i/o relationships to those shown in (A) (data not shown). (B) fEPSP size as percent of first stimulus versus interstimulus interval (msec). Data were analyzed by two-way ANOVA/Bonferroni multiple comparison analyses. There were no statistically significant differences in the i/o relationships. The WT ppfs were statistically different (p,0.05) from App HET (many time points), WT versus Fmr1 KO (50 msec time point); WT versus Fmr1 KO /App HET (first four time points). There were no other statistically significant differences in ppfs for the remaining datasets.