Abstract
We previously reported that the peroxisome proliferator-activated receptor γ (PPARγ) agonist rosiglitazone (RSG) improved hippocampus-dependent cognition in the Alzheimer's disease (AD) mouse model, Tg2576. RSG had no effect on wild-type littermate cognitive performance. Since extracellular signal-regulated protein kinase mitogen-activated protein kinase (ERK MAPK) is required for many forms of learning and memory that are affected in AD, and since both PPARγ and ERK MAPK are key mediators of insulin signaling, the current study tested the hypothesis that RSG-mediated cognitive improvement induces a hippocampal PPARγ pattern of gene and protein expression that converges with the ERK MAPK signaling axis in Tg2576 AD mice. In the hippocampal PPARγ transcriptome, we found significant overlap between peroxisome proliferator response element-containing PPARγ target genes and ERK-regulated, cAMP response element-containing target genes. Within the Tg2576 dentate gyrus proteome, RSG induced proteins with structural, energy, biosynthesis and plasticity functions. Several of these proteins are known to be important for cognitive function and are also regulated by ERK MAPK. In addition, we found the RSG-mediated augmentation of PPARγ and ERK2 activity during Tg2576 cognitive enhancement was reversed when hippocampal PPARγ was pharmacologically antagonized, revealing a coordinate relationship between PPARγ transcriptional competency and phosphorylated ERK that is reciprocally affected in response to chronic activation, compared with acute inhibition, of PPARγ. We conclude that the hippocampal transcriptome and proteome induced by cognitive enhancement with RSG harnesses a dysregulated ERK MAPK signal transduction pathway to overcome AD-like cognitive deficits in Tg2576 mice. Thus, PPARγ represents a signaling system that is not crucial for normal cognition yet can intercede to restore neural networks compromised by AD.
Introduction
The nuclear receptor peroxisome proliferator-activated receptor γ (PPARγ) is a well established therapeutic target in type 2 diabetes since its transcriptional activity leads to improved insulin sensitivity in the periphery. Clinical studies suggest that PPARγ agonists such as rosiglitazone (RSG) improve cognitive function in Alzheimer's disease (AD) patients and in several rodent models of the disease (Watson et al., 2005; Pedersen et al., 2006; Risner et al., 2006; Escribano et al., 2009, 2010; Rodriguez-Rivera et al., 2011). However, the mechanism by which PPARγ agonists achieve these CNS effects is unclear.
Some mechanistic insight is provided by recent work demonstrating reciprocal PPARγ and extracellular signal-regulated protein kinase mitogen-activated protein kinase (ERK MAPK) activity in several neurological disorders and cancer suggesting a potential action for PPARγ in amelioration of memory deficits in AD (Camp and Tafuri, 1997; Kim et al., 2003; Papageorgiou et al., 2007; Schroeter et al., 2007; Rosa et al., 2008; Zhang et al., 2011). In support of this, chronic elevated β-amyloid (Aβ) leads to dysregulation of hippocampal ERK MAPK in vitro and in vivo (Dineley et al., 2001a; Bell et al., 2004; Swatton et al., 2004), while PPARγ agonism ameliorates cognitive deficits in vivo and can prevent Aβ-induced deficits in hippocampal plasticity in vitro (Costello et al., 2005; Rodriguez-Rivera et al., 2011). Likewise, both ERK and PPARγ are dysregulated in AD brain and certain PPARγ polymorphisms are associated with increased risk for the disease (Kitamura et al., 1999; Scacchi et al., 2007).
To investigate the molecular mechanism underlying PPARγ agonism with RSG on AD-like cognitive function, we used an extensively characterized AD mouse model, Tg2576, that expresses a transgene encoding the human amyloid precursor protein containing a mutation that causes AD in humans (Hsiao et al., 1996). Importantly, Tg2576 mice exhibit age-dependent cognitive decline as measured in several behavioral paradigms but most notably in those requiring proper hippocampal ERK MAPK function that are also impaired in humans with AD (Atkins et al., 1998; Dineley et al., 2001b; Dineley et al., 2001a,b, 2002; Hamann et al., 2002; Hoefer et al., 2008).
Therefore, the current study tested whether regulation of hippocampal PPARγ coincided with ERK MAPK signaling following RSG-mediated cognitive improvement. In the hippocampal PPARγ transcriptome of the Tg2576 AD animal model, we found significant overlap between peroxisome proliferator response element (PPRE)-containing PPARγ target genes and cAMP response element (CRE)-containing ERK MAPK [cAMP response element-binding protein (CREB)] target genes. Using quantitative mass spectrometry and bioinformatics on the dentate gyrus, we identified many proteins related to synaptic plasticity and memory formation that were induced concomitant with RSG-mediated cognitive rescue and activation of PPARγ and ERK2, actions reversed when hippocampal PPARγ was pharmacologically antagonized to reverse RSG-mediated cognitive improvement. We conclude that the hippocampal transcriptome and proteome induced by cognitive enhancement with RSG harnesses a dysregulated ERK MAPK signal transduction pathway to overcome AD-like cognitive deficits in Tg2576 mice. Thus, PPARγ represents a signaling system that is not crucial for normal cognition yet can intercede to restore neural networks compromised by AD.
Materials and Methods
Animals.
Animals were bred in The University of Texas Medical Branch animal care facility by mating heterozygous Tg2576 males with C57BL6/SJL (F1) females (Jackson Laboratory). The University of Texas Medical Branch operates in compliance with the United States Department of Agriculture Animal Welfare Act, the Guide for the Care and Use of Laboratory Animals, and Institutional Animal Care and Use Committee-approved protocols.
Mice were housed, n ≤ 5 per cage, with food and water ad libitum. All animal manipulations were conducted during the lights-on phase (0700–1900 h). Male and female 8 months old (8MO) Tg2576 and wild-type (WT) littermates were fed control or 30 mg/kg RSG diet (Bio-Serv) for 30 d, as previously described (Rodriguez-Rivera et al., 2011). Animals were killed by decapitation and the brain was rapidly removed from the skull for hippocampus dissection.
Intracerebroventricular injection.
Using a modified free-hand method (Clark et al., 1968), mice were anesthetized (isoflurane, 1–4%) and, with aseptic technique, the skull was exposed with a small incision along the midline. Hemostatic forceps held the needle 1 mm anterior and 1 mm lateral of the bregma. GW9662 (32.5 pmol, 3 μl) or vehicle (dimethyl sulfoxide) were delivered by an electronic programmable microinfuser (Harvard Apparatus) at 3 μl/min and the needle left in place for 1 min postinjection. This dose was based on previous reports of intracerebroventricular (ICV) injection of GW9662 to antagonize PPARγ function in the CNS (Maeda et al., 2007; Zhang et al., 2008).
Fear conditioning.
Two-pair fear conditioning (FC) training and FC hippocampus-dependent contextual testing was performed on awake and alert subjects 4 h after ICV injection. Eight to 12 mice per group (male and female) were trained in the FC chamber following our standard FC protocol, as described previously (Dineley et al., 2002). Twenty-four hours later, mice were returned to the training chamber for testing in the hippocampus-dependent contextual FC paradigm. Cued FC was not included in this study since Tg2576 are not deficient in the hippocampus-independent cued FC task and RSG treatment has no effect on WT or Tg2576 performance in this task (Dineley et al., 2002; Rodriguez-Rivera et al., 2011). Following testing, mice were decapitated and the hippocampus and cortex were immediately dissected, frozen on dry ice, and stored at −80°C.
Shock threshold.
Approximately 9 animals per group were subjected to shock threshold test to assess shock sensitivity, as described previously (Dineley et al., 2002). Briefly, a sequence of single foot shocks was delivered to animals placed on the same electrified grid used for fear conditioning. Initially, a 0.1 mV shock was delivered for 1 s, and the animals' behavior was evaluated for flinching, jumping, and vocalization. At 30 s intervals the shock intensity was increase by 0.1 mV up to 0.7 mV and then returned to 0 mV in 0.1 mV increments at 30 s intervals. Threshold to vocalization, flinching, and then jumping was quantified for each animal by averaging the shock intensity at which each animal manifested a behavioral response to the foot shock.
Nuclear extraction.
Following the manufacturer's instructions (Active Motif), extracts were prepared from the hippocampi of individual animals. Immunoblot analyses for subcellular fraction markers determined >95% purity (data not shown).
DNA binding assays.
Eight micrograms of nuclear extract was assayed for PPARγ binding to the PPRE with TransAM ELISA kit (Active Motif) according to the manufacturer's instructions. Data are reported as mean ± SEM normalized to WT signal.
Antibodies.
Phospho-Ser84 PPARγ (1:500; MAB3632) and PPARγ (1:200; 07-466) were obtained from Millipore. ERK (1:1000; 9102) and phospho-Thr202/Tyr204 ERK (1:1000; 9101) were obtained from Cell Signaling Technology. β-Actin (1:5000; A5441) was obtained from Sigma. Lamin A/C (1:100; SC-20681) was obtained from Santa Cruz Biotechnology. HRP-conjugated anti-mouse IgG (1:50,000; NA931V) and anti-rabbit IgG (1:100,000; NA934V) were obtained from GE Healthcare.
Quantitative immunoblot.
Using our previously described method (Dineley et al., 2001b), 10–40 μg (DC Protein Assay, Bio-Rad) of nuclear or cytosolic hippocampal extract from individual animals was resolved by SDS-PAGE, transferred to PVDF membrane (Immobilon, Millipore), then probed with the appropriate primary and secondary antibodies. Protein bands were detected by chemiluminescence (Advance ECL, GE Healthcare) and film exposures in the linear range for the antigen-antibody combination were developed with a Kodak imager (Kodak). Band densities were measured with ImageJ (NIH) and normalized to control level. Normalized control values were determined for each immunoblot by averaging control values, dividing each control and test sample density by the average of the control set, and then determining the average and SEM for control and test samples for n = 6–10 animals/group. All blots were sequentially probed for PPARγ phosphorylated on Ser84, PPARγ, ERK phosphorylated on Thr202/Tyr204, ERK, then lamin or actin for normalization.
RNA extraction and PPARγ1 and PPARγ2 reverse transcriptase-PCR.
Hippocampi from WT mice were dissected out and stored in RNAlater RNA protection solution (Ambion, catalog #AM7024) for further analysis. Total RNA was isolated from the tissue using RNAqueous-Micro Kit (Ambion) following the manufacturer's instructions. RNA sample quality and quantity were analyzed using Agilent 2100 Bioanalyzer and Nanodrop ND1000, respectively. One microgram of total RNA was synthesized into cDNA using Transcriptor High Fidelity cDNA Synthesis Kit (Roche Applied Science) according to the manufacturer's instructions and subjected to PCR with primers (Sigma-Genosys) specific for PPARγ1 and PPARγ2 transcripts. PCR (25 cycles) was performed (PerkinElmer PE2400) under the following conditions: 94°C, 30 s; 58°C, 20 s; 72°C, 20 s. PCR products were analyzed in 2% agarose gels in Tris-acetate-EDTA buffer with base pair marker.
Quantitative PCR.
Individual hippocampi were collected from 4 animals (male and female) of each group (WT untreated, untreated Tg2576, RSG-treated Tg2576) and suspended in 20-fold excess (w/v) TRIzol (Invitrogen). The tissue was homogenized in a 1 ml Dounce homogenizer on ice and RNA extracted according to the manufacturer's instructions. Quality control assessment of total RNA was performed on an Agilent 2100 Bioanalyzer (Agilent Technologies) as well as A260/A280 and A260/A230 nm ratio analyses using NanoDrop technology (Thermo Scientific). cDNA was synthesized from 5 μg of hippocampal mRNA using Superscript III (Invitrogen) according to the manufacturer's instructions. Individual animal mRNA was quantified for a custom array of predominantly PPRE-containing PPARγ genes on 1 μl of cDNA using a Roche LightCycler 480 and LightCycler 480 SYBR Green I Master reagent (Roche Applied Science) in the University of Texas Medical Branch Molecular Genomics Core Facility. All oligos (Table 1) were purchased from Integrated DNA Technologies ΔCT values were calculated by subtracting the average CT of three housekeeping genes (GAPDH, Rpl19, and Bpol) from each gene of interest and the ΔΔCT method (Applied Biosystems) was used to calculate fold-change values between treatment groups. −ΔCT values are shown (Figs. 1B, 2D) to indicate increased number of mRNA transcripts.
Quantitative mass spectrometry.
Stable isotope labeling was used to quantify differential protein expression as previously described (Sadygov et al., 2010; Starkey et al., 2010). Briefly, the dentate gyrus from 10 mice each of Tg2576 fed control or RSG diet were homogenized in TRIzol and the protein pellet resuspended in guanidine. Following reduction and alkylation, proteins were digested with trypsin and peptides desalted with SepPack C18 cartridges. Dried peptides were then treated with immobilized trypsin (Applied Biosystems) in normal water (H216O) or heavy water (H218O) for trypsin-mediated exchange of oxygen atoms from water onto the C terminus of peptides. Desalted peptides were then pooled to prepare a mixture of 16O-labeled peptides from control-fed mice and 18O-labeled peptides from RSG-fed mice. To reduce the sample complexity and increase the depth of analysis into the proteome, the peptide mixture was resolved into 60 fractions using strong cation exchange chromatography.
Two-dimensional liquid chromatography-tandem mass spectrometry.
Each SCX fraction was injected onto a C18 peptide trap (Agilent), desalted, and eluted peptides separated on a reversed phase nano-HPLC column with a linear gradient over 120 min at 200 nl/min. Liquid chromatography-tandem mass spectrometry (LC-MS/MS) experiments were performed with a LTQ linear ion trap mass spectrometer (ThermoFinnigan) equipped with a nanospray source. The mass spectrometer was coupled online to a ProteomX nano-HPLC system (ThermoFinnigan). The mass spectrometer was operated in the data-dependent triple-play mode. In this mode, the three most intense ions in each MS survey scan were automatically selected for moderate resolution zoom scans which were followed by MS/MS. Each of the peptide mixtures was repetitively analyzed by nano-HPLC-MS/MS three times. The acquired MS/MS spectra were searched with SEQUEST algorithm performed on the Bioworks 3.2 platform (ThermoFinnigan) using conservative filtering criteria of Sp ≥300, ΔCn ≥0.12, and Xcorr of 1.9, 2.0 and 3.0 for data from a singly, doubly or triply charged precursor ions, respectively.
The zoom scan data were used to calculate the relative abundance ratios of 18O-labeled peptide/16O-unlabeled peptide pairs using MassXplorer (Sadygov et al., 2010). Peptides with charge >3, false discovery rate >3%, 18O/16O ratios <0.1 or >10, and reversed sequences were removed from further analysis. Calculated peptide ratios were log2 transformed and mean centered before statistical analysis. Significance was determined by assessed using the Wilcoxon rank-sum test with Benjamini-Hochberg false discovery rate correction for multiple testing comparisons as indicated (Benjamini and Hochberg, 1995).
Data were then analyzed through the use of the extensively curated Ingenuity Pathways Analysis (Ingenuity Systems) with a significance cutoff of p ≤ 0.05 and ≥20% change in protein expression. Functional Analysis using Gene Ontology classifiers identified the biological functions that were most significant to the dataset. Right-tailed Fisher's exact test was used to calculate a p-value determining the probability that each biological function assigned to that dataset is due to chance alone. Network Analysis generates a graphical representation of the molecular relationships between molecules. Molecules are represented as nodes, and the biological relationship between two nodes is represented as a line. All lines are supported by at least one reference from the literature, from a textbook, or from canonical information stored in the Ingenuity Knowledge Base. Nodes are displayed with various shapes that represent the functional class of the gene product.
Total β-amyloid quantification.
Cortex from 18 Tg2576 and 18 Tg2576 RSG-treated (male and female) was homogenized in 8× (volume by wet weight) 5 m guanidine HCl, 50 mm Tris HCl, pH 8.0. Signal Select colorimetric sandwich ELISA (BioSource) for either human Aβ1–40 or Aβ1–42 was used in comparison to a standard curve.
Statistics.
Statistical analyses were conducted with ANOVA followed by either Bonferroni's or Dunnett's post hoc comparison. Where appropriate, Student's t test was used for pairwise comparison. Significance was set to p < 0.05.
Results
Initially we evaluated whether oral RSG treatment increased PPARγ activity in the CNS by measuring hippocampal PPARγ binding to its PPRE. Nuclear extracts prepared from the hippocampus of Tg2576 and WT littermates showed that RSG treatment resulted in a statistically significant (∼30%) increase in PPARγ DNA binding in both Tg2576 and WT groups (Fig. 1A), confirming that oral RSG is blood–brain barrier permeable (Strum et al., 2007; Festuccia et al., 2008; Diano et al., 2011; Lu et al., 2011; Ryan et al., 2011) and increases steady-state DNA binding. We were unable to affect DNA binding with the PPARγ antagonist GW9662 (data not shown).
Consistent with the prevailing concept that PPARγ binding to PPREs is necessary yet insufficient for regulating target gene expression, we assessed the hippocampal PPARγ transcriptome using quantitative PCR on hippocampal mRNA isolated from mice treated with or without RSG. Expression analysis from a custom array of 45 genes chosen for enrichment in PPREs, demonstrated that 34 were downregulated in untreated Tg2576 compared with WT and 32 of those were induced by RSG treatment in Tg2576 (Table 1). For example, the PPRE-containing apolipoprotein O gene (APO-O) was decreased in untreated Tg2576 compared with WT, and RSG treatment reversed this (Fig. 1B). As such, untreated Tg2576 mice exhibited a −1.97-fold-change in APO-O mRNA transcripts compared with WT, and RSG induced a +10.82-fold increase in this mRNA transcript in Tg2576.
We next probed the hippocampal PPARγ proteome with quantitative mass spectrometry using the stable isotope 18O-/16O-water and LC-MS/MS method (Sadygov et al., 2010; Starkey et al., 2010). This method of differentially labeling and quantifying dentate gyrus proteins from untreated and RSG-treated Tg2576 revealed that PPARγ agonism significantly upregulated 147 proteins and downregulated 67 proteins related to energy, synaptic structure, plasticity, biosynthesis, and transport (Fig. 1C). For example, this approach determined that the PPARγ target gene, APO-O, exhibited 2.9-fold increased protein in RSG-treated Tg2576 compared with untreated Tg2576 (Benjamini-Hochberg rank sum p = 0.0015) and the ERK phosphatase PP2A was downregulated by 16% of untreated Tg2576 (Benjamini-Hochberg p = 2.54 × 10−6).
To evaluate potential functional relationships between the Tg2576 hippocampal proteins whose expression was augmented by RSG treatment, we performed bioinformatics analysis on proteins involved in synaptic plasticity. ERK MAPK emerged as a central node following Ingenuity Pathways Analysis. PPARγ itself was a target regulator of ERK MEK (mitogen-activated protein kinase kinase) in addition to glutamate decarboxylase, GSK3-α, α-synuclein, metabotropic glutamate receptor 5, and glutamate receptor 2 (Fig. 1D).
The mouse PPARγ gene gives rise to two mRNAs (PPARγ1 and PPARγ2) that differ only at their 5′ ends (Fig. 1E). The mouse PPARγ2 mRNA encodes an additional 30 aa N-terminal to the first ATG codon of PPARγ1 (Zhu et al., 1995). Our immunoblot analysis of mouse hippocampus from WT or Tg mice treated with any intervention had only revealed a single band at ∼67 kDa. In an attempt to determine which of the two isoforms was detected by immunoblot, we performed PCR on WT mouse hippocampus using primer pairs that would selectively produce amplicons either only within the PPARγ2-specific exon 1′ (primer set 1) or within exon 2 (primer set 2) that is common to both PPARγ1 and PPARγ2 (Zhu et al., 1995). This illustrated that both mRNA forms were expressed in the hippocampus (Fig. 1F, top). However, quantitative PCR indicated that the ratio of PPARγ1 to PPARγ2 was >7 (Fig. 1F, bottom). Therefore, immunoblots most likely detected PPARγ1 protein. This was further confirmed by using a PPARγ2-specific antibody (Santa Cruz Biotechnology) to probe mouse hippocampal extracts which failed to produce a signal (data not shown).
We next determined whether there were differences between WT and Tg2576 hippocampal PPARγ, Ser84 phosphorylated PPARγ (pPPARγ), or subcellular distribution. Quantitative immunoblot analysis of hippocampal cytoplasmic fractions from sham-treated Tg2576, WT, and RSG-treated Tg2576 showed no significant differences in either total or pPPARγ (data not shown). However, Tg2576 hippocampal nuclear fractions contained significantly less PPARγ than WT (Fig. 2A). ERK MAPK phosphorylation of PPARγ at Ser84 is considered inhibitory by decreasing PPARγ transcriptional competency (Camp and Tafuri, 1997; Shao et al., 1998). Although nuclear pPPARγ is lower in untreated Tg2576 (Fig. 2B), nuclear PPARγ transcriptional competency in Tg2576 hippocampus is likely diminished since the ratio of phospho/total PPARγ indicates a net increase in the ERK MAPK phosphorylated, inhibited form of PPARγ (Fig. 2C).
PPARγ agonists have been shown to ameliorate several forms of cognitive deficits in Tg2576 and other AD mouse models (Pedersen et al., 2006; Escribano et al., 2009, 2010; Rodriguez-Rivera et al., 2011). We found that RSG cognitive improvement also ameliorated Tg2576 deficiencies in hippocampal nuclear PPARγ (Fig. 2A,B). These changes resulted in a ratio of phospho/total PPARγ statistically indistinguishable from WT (Fig. 2C). Finally, quantitative PCR analysis of hippocampal mRNA showed that PPARγ gene expression was reduced in Tg2576 compared with WT and normalized by RSG (Fig. 2D) with an 8.7-fold increase in PPARγ gene transcripts although PPARγ is not a PPRE-containing gene (Table 1), suggesting that RSG treatment has diverse effects on gene expression. This is further supported by our observation that several genes lacking identifiable PPREs were also induced by RSG treatment (Table 1). In summary, nuclear-PPARγ gene transcripts and protein are deficient in Tg2576 hippocampus and both are normalized with RSG treatment concomitant with reversal of hippocampus-dependent cognitive deficits.
Given the importance of ERK2 MAPK in hippocampus-dependent memory (Selcher et al., 2001), including contextual FC, we also evaluated RSG effects on hippocampal ERK2 protein, its phosphorylation (activation) status, and nuclear-cytosolic distribution. Quantitative immunoblot analysis of total-ERK2 in hippocampal nuclear and cytoplasmic fractions showed no significant differences between Tg2576 and WT animals (data not shown). Tg2576 RSG treatment, however, led to increased nuclear ERK2 activity, as noted by an increase in Thr202/Tyr204 phosphorylated ERK2 (pERK2) compared with untreated Tg2576 (Fig. 2E). No significant effects on cytosolic total or pERK2 cytoplasmic samples were found (data not shown). Thus, nuclear ERK2 activity in the hippocampus is enhanced during RSG rescue of hippocampus-dependent cognition in Tg2576 mice. Consistent with our previous observation that RSG has no effect on hippocampus-dependent cognition in WT littermates (Rodriguez-Rivera et al., 2011), RSG also had no effect on WT PPARγ or ERK (data not shown).
A recurring concern with thiazolidinediones (TZDs) is whether peripheral administration can actually affect the molecular target PPARγ in the CNS. Thus, to test whether CNS PPARγ mediates RSG cognitive improvement in 9MO Tg2576, we directly injected GW9662 (Leesnitzer et al., 2002) into the lateral ventricles of RSG-treated mice to block CNS PPARγ activity. Such ICV administration of GW9662 has been used to establish that CNS PPARγ mediates RSG effects in animal models of energy balance and feeding behavior (Diano et al., 2011; Ryan et al., 2011). The dose used was based on previous reports of ICV injection of GW9662 to antagonize PPARγ function in the CNS (Maeda et al., 2007; Zhang et al., 2008).
Tg2576 and WT mice were infused with either vehicle or GW9662 4 h before FC training (Fig. 3A). No significant difference in behavior was detected between the groups during training, indicating that ICV injection and PPARγ manipulations do not interfere with behavior during the acquisition phase of this associative learning paradigm (Fig. 3B). The contextual test for FC memory consolidation performed 24 h later, further demonstrated that RSG (or ICV injection of vehicle) does not affect WT performance and that RSG-treated Tg2576 now freeze to the same extent as WT in contrast to Tg2576 treated with vehicle alone (Fig. 3C). These results confirm that RSG treatment ameliorates cognitive deficits in 9MO Tg2576 (Rodriguez-Rivera et al., 2011) and that antagonism of CNS PPARγ in RSG-treated Tg2576 prevents consolidation of the hippocampus-dependent contextual FC memory (Fig. 3C). Neither RSG nor RSG+GW9662 affected WT performance, emphasizing that PPARγ activity is not critical to hippocampus-dependent learning and memory in non-diseased mice. Additional studies in WT mice also demonstrated that RSG treatment does not augment cued FC learning and that ICV-delivered GW9662 alone had no behavioral effect (data not shown). Furthermore, we detected no effect of genotype or treatment in an animals' tendency to flinch, vocalize, or jump to increasing shock intensities during a shock threshold test; indicating that 9MO Tg2576 exhibit equivalent sensory processing of the footshock in the FC paradigm and RSG treatment has no effect on this process in WT or Tg2576 mice (Fig. 3D). Together, these results suggest that RSG rescue of hippocampus-dependent cognitive deficits in Tg2576 AD mice is mediated by hippocampal PPARγ to compensate for a signal transduction system that is typically necessary for this form of learning.
Since ERK MAPK is essential for hippocampus-dependent learning and memory in general, and contextual FC in particular, we hypothesized that PPARγ agonism in Tg2576 mice recruits the ERK MAPK pathway to overcome AD-like cognitive deficits in associative learning and memory. Therefore, we evaluated whether PPARγ antagonism with ICV GW9662 affected hippocampal PPARγ and ERK in RSG-treated Tg2576. We killed animals and collected hippocampi to evaluate GW9662 effects 4, 8, and 16 h following ICV infusions; if these animals had been FC trained, these time points would have correlated with 0, 4, and 12 h post-training. Quantitative immunoblot revealed that ICV injection of GW9662 had no significant effect on nuclear or cytosolic forms of total or pPPARγ at the 4 and 16 h time points compared with vehicle controls (Fig. 4A–D). However, 8 h after ICV infusion of GW9662 we observed decreased nuclear PPARγ (Fig. 4C) concomitant with increased cytoplasmic PPARγ (Fig. 4D). Further, cytoplasmic pPPARγ was also increased at 8 h (Fig. 4B). While total PPARγ decreased ∼30% in the nucleus at this time point, analysis of the phospho/total PPARγ ratio at the 8 h time point revealed no net change between the nuclear and cytosolic compartments (Student's two-tailed t test = 0.18, data not shown). These results are consistent with a model in which PPARγ phosphorylation at Ser84 might be instrumental in nuclear export or cytoplasmic retention. In summary, inhibition of CNS PPARγ with GW9662 in RSG-treated Tg2576 mice led to a net decrease in nuclear-PPARγ concomitant with an increase in total and pPPARγ in the cytoplasm suggesting that reversal of cognitive improvement through inhibition of PPARγ involves subcellular redistribution of the protein.
Since the maximal effect of GW9662 on nuclear PPARγ was achieved 8 h after ICV injection, we evaluated whether nuclear ERK2 activity was also affected at this time point. As might be expected, GW9662 antagonism of CNS PPARγ resulted in no change in total ERK2 but decreased nuclear ERK2 activation (one-way ANOVA: (F(2,14) = 6.01, F(2,15) = 0.42 (p < 0.05)) for total ERK and pERK, respectively). Because ERK activation and the ERK2 isoform has been shown to be necessary for FC consolidation (Atkins et al., 1998; Selcher et al., 2001), our findings that PPARγ antagonism both reverses RSG effects on FC performance and nuclear ERK activity supports our interpretation that cognitive improvement in Tg2576 with RSG treatment results from PPARγ effects on ERK2 MAPK activity in the hippocampus.
These RSG-mediated effects are consistent with the notion that RSG crosses the blood–brain barrier to activate CNS PPARγ (Willson et al., 1996; Strum et al., 2007; Festuccia et al., 2008; Diano et al., 2011; Lu et al., 2011; Ryan et al., 2011). Further, RSG increased both WT and Tg2576 hippocampal PPARγ DNA binding activity indicating that RSG effects in Tg2576 brain are not due to compromised BBB permeability. Finally, ICV administration of the specific PPARγ full antagonist GW9662 (Leesnitzer et al., 2002) reversed RSG cognitive improvement strongly implicates CNS PPARγ.
Last, we assessed whether cognitive improvement via PPARγ agonism correlates with altered Aβ accumulation. Total Aβ1–40 and Aβ1–42 were quantified by dissolving cortical tissue directly in guanidine-HCl to extract all forms of Aβ from untreated Tg2476 and RSG-treated Tg2576 that were ICV-injected with either vehicle or GW9662. Neither 1 month RSG treatment nor acute GW9662 PPARγ inhibition (8 h) significantly altered total Aβ1–40, or Aβ1–42 (Table 2). Therefore, neither RSG PPARγ agonism nor GW9662 PPARγ antagonism influenced Aβ accumulation in this animal model. Since we are focused on elucidating cognitive rescue mechanisms downstream of Aβ toxicity, we did not further characterize effects of RSG treatment on Aβ pathology although there are reports of Aβ mechanisms (Mandrekar-Colucci et al., 2012).
Discussion
We and others have previously shown that PPARγ agonists improve cognitive performance in mouse models of AD, mainly in tasks affected in human AD (Hamann et al., 2002; Pedersen et al., 2006; Hort et al., 2007; Hoefer et al., 2008; Escribano et al., 2010; Rodriguez-Rivera et al., 2011). It is also well established that hippocampal ERK MAPK is required for many of these forms of learning and memory (Sweatt, 2004). In these contexts, the current study addressed the convergence of the ERK MAPK and PPARγ signaling pathways in Tg2576 mice following cognitive improvement with RSG.
Initially, we evaluated hippocampal PPARγ in Tg2576 and WT littermates either untreated or treated with oral RSG for 1 month between 8MO and 9MO. RSG treatment of Tg2576 mice significantly enhanced hippocampal PPARγ DNA binding, mRNA, and protein. PPARγ phosphorylation at Ser84 has been shown to inhibit transcriptional competency (Camp and Tafuri, 1997). We found that the ratio of pPPARγ/total PPARγ in untreated Tg2576 hippocampus nuclear fractions was significantly elevated, indicative of net PPARγ inhibition, while RSG treatment normalized this ratio to WT level.
We discovered that concomitant with RSG cognitive enhancement, the hippocampal PPARγ transcriptome and proteome converge with the ERK MAPK cascade at several levels. First, the majority of PPRE-containing target genes induced by RSG treatment also contain CREs suggesting that some PPARγ target genes are also CREB target genes which themselves are highly regulated by ERK MAPK during memory consolidation (Guzowski and McGaugh, 1997; Ahi et al., 2004). Second, an unbiased proteomics and bioinformatics analysis of the dentate gyrus from untreated and RSG-treated Tg2576 found that ERK MAPK was a central, integrative node of the plasticity proteins augmented by RSG. Third, RSG-mediated changes in hippocampal PPARγ and ERK were reversed when RSG-treated Tg2576 memory consolidation was blocked by an irreversible, selective PPARγ full antagonist (GW9662). Thus, there is a coordinate relationship between PPARγ transcriptional competency and pERK that is reciprocally affected in response to chronic activation, compared with acute inhibition, of PPARγ. Finally, CREB-binding protein (CBP) was markedly induced during RSG cognitive enhancement. CBP can rescue learning and memory deficits in AD mouse models (Caccamo et al., 2010), is a nuclear coactivator of PPARγ (Bugge et al., 2009; Inoue et al., 2012) and CREB (Klein et al., 2005), and is thus well positioned to integrate the convergence of the PPARγ and ERK MAPK pathways.
From our data, we elaborate on one of many examples for convergent PPARγ and ERK pathway integration: RSG treatment impinged upon the protein sumoylation system. Protein sumoylation often leads to the functional inhibition of the target protein, e.g., MEK, the upstream kinase activator of ERK (Kubota et al., 2011). This post-translational inhibitory modification is reversibly regulated by the SENP family of SUMO proteases. A scenario can be considered in which increased Tg2576 hippocampal protein sumoylation (McMillan et al., 2011) leads to inhibition of MEK, thereby preventing proper ERK activation during memory consolidation. Elevated sumoylation could also account for the observed reduction in PPARγ transcriptional activity (Floyd and Stephens, 2012) as well as the PPARγ hippocampal coregulator PGC1-α (Rytinki and Palvimo, 2009) and the ERK target CBP (Kuo et al., 2005). RSG-mediated induction of SENP8 gene expression could conceivably contribute to disinhibition of the PPARγ transcriptome and the ERK MAPK cascade. Likewise, RSG induction of CBP, cyclin-dependent kinase 2, and nucleosomal assembly protein 1-like 1 would further contribute to PPARγ and ERK-dependent transcription by providing transcription coregulators and enhancing ERK nuclear translocation (Okada et al., 2011; Plotnikov et al., 2011). This hypothetical scenario built upon the observed PPARγ transcriptome supports our model that PPARγ agonism serves to integrate the ERK and PPARγ signaling pathways to facilitate hippocampal memory consolidation.
Analysis of the Tg2576 hippocampal proteome from untreated versus RSG-treated animals also supports the notion that PPARγ agonism serves to integrate the ERK and PPARγ signaling pathways. We found that RSG led to the upregulation of 147 proteins and downregulation of 67 proteins in Tg2576 dentate gyrus that can be functionally categorized into energy, biosynthesis, synaptic structure or plasticity; consistent with many of the proteins found affected in human AD hippocampus with similar approaches (Sultana et al., 2007; Di Domenico et al., 2011). Again, several of the identified proteins were related to the ERK MAPK cascade (e.g., GluR2, mGluR5, PKCγ) (Neary et al., 1999; Schroeter et al., 2007; Ménard and Quirion, 2012). If 9MO Tg2576 recapitulates a relevant and diagnosable stage of human AD, PPARγ agonism to selectively impinge upon the ERK MAPK cascade represents a disease modifying intervention for humans. Furthermore, given the adverse side effects attributed to RSG full agonism of PPARγ, it will be important to test alternative TZDs such as pioglitazone as well as next-generation PPARγ non-agonist and partial agonist ligands (Choi et al., 2010, 2011; Vidović et al., 2011).
GW9662 PPARγ antagonism in RSG-treated AD mice mimics the effect of ERK MAPK inhibitors on contextual FC in WT rodents (Atkins et al., 1998) further supporting the model that PPARγ can harnesses a dysregulated ERK MAPK pathway to overcome AD-like cognitive deficits in Tg2576 mice. At the biochemical level, GW9662 reversed the effects of RSG on nuclear PPARγ and ERK activity in Tg2576 hippocampus with a time course that suggests GW9662 interferes with FC consolidation through effects on ERK via PPARγ.
GW9662 also led to elevated cytoplasmic pPPARγ, indicating that GW9662 reversed RSG effects on nuclear PPARγ and promoted cytosolic redistribution of PPARγ. Since PPARγ Ser84 phosphorylation also promotes the rapid turnover of PPARγ through targeted ubiquitination, sumoylation, and proteosomal degradation (Genini and Catapano, 2006), this may account for the relatively rapid recovery (16 h) from GW9662. While our methodology cannot address PPARγ nuclear/cytosol shuttling or turnover, it can be said that GW9662 reversal of RSG cognitive improvement leads to reduced PPARγ nuclear localization and increased inhibitory phosphorylation accompanied by reduced nuclear ERK activity.
The ERK MAPK cascade has been shown to regulate PPARγ both through phosphorylation and nuclear/cytosol trafficking via interaction with MEK-ERK complexes which themselves shuttle in and out of the nucleus (Burgermeister et al., 2007; von Knethen et al., 2010). We found that RSG increased nuclear ERK activity concomitant with a decrease in ERK-mediated pPPARγ. This at first appears illogical but one possible consequence of RSG cognitive enhancement is concurrent effects on overall ERK activity as well as ERK substrate selectivity. We suggest that following RSG treatment, pERK performs many functions, some of which are in series and in parallel with PPARγ such that not all pERK directly affects PPARγ phosphorylation because some pERK is executing additional cognitive-enhancing functions. An alternative mechanism might be the upregulation of phosphatases that act upon PPARγ that lead to decreased pPPARγ. Our observation that serine/threonine protein phosphatase 1 (PP1) α and γ gene transcripts are upregulated in RSG-treated Tg2576 (Table 1) is consistent with this mechanism, although the PPARγ phosphatase has yet to be identified.
Although many examples of TZDs increasing pERK exist in the literature, the mechanism remains poorly defined (Gardner et al., 2003; Kim et al., 2003; Rosa et al., 2008). The following model attempts to integrate our data within a framework of potential relationships with the ERK MAPK cascade and ERK molecular mechanisms gleaned from the annotated literature. RSG cognitive enhancement may reflect a feed forward loop that begins with RSG-mediated PPARγ target gene induction, e.g., casein kinase II subunit II α (CK2α) (Table 1), which in turn stimulates ERK nuclear translocation (Plotnikov et al., 2011). We detected decreased PP2A by mass spectrometry similar to TZD (pioglitazone) effects during adipocyte differentiation (Altiok et al., 1997). Since PP2A specifically dephosphorylates and inactivates pERK (Alessi et al., 1995; Hu et al., 2009; Puustinen et al., 2009), decreased PP2A would be predicted to lead to a net increase in pERK as we found (Fig. 2E). These results suggest potential coordinate effects of decreased PP2A and increased CK2α on nuclear ERK activity. Furthermore, cross-regulatory feed forward loops have been extensively described in that some transcription factors induced by PPARγ also bind to the PPARγ gene promoter to increase its expression. Our finding of increased PPARγ transcripts and protein in RSG-treated Tg2576 support this notion. PPARγ, in turn, may then mediate the induction of other transcription factors and target genes that integrate the PPARγ transcriptome with the ERK MAPK cascade. One example of this comes from the C/EBP-PPARγ field (Wu et al., 1995, 1999; Lefterova et al., 2008).
Enhanced cognition in AD mice with RSG PPARγ agonism, coupled with our finding that neither PPARγ agonism nor antagonism affected WT performance, positions this nuclear receptor as a potential therapeutic target for the human disease. This idea is strengthened by the fact that PPARγ is dysregulated in AD brain and certain polymorphisms in the PPARγ gene are associated with increased risk for the disease (Kitamura et al., 1999; Scacchi et al., 2007). Furthermore, our discovery that the hippocampal PPARγ transcriptome and proteome converge with the ERK MAPK cascade at several levels, combined with the reciprocal effects of RSG and GW9662 on PPARγ and ERK activity and localization, suggest a multifaceted regulatory relationship warranting further investigation.
Footnotes
This work was supported by the National Institutes of Health under Grants F31 NS052928 to J.R.-R. and R01-AG031859 to K.T.D. and L.A.D. Additional funding was provided by the American Health Assistance Foundation, The Sealy Foundation for Biomedical Research, and a kind gift from J. and W. Mohn to K.T.D.; by the Emmett and Miriam McCoy Foundation to L.A.D.; and by the Cullen Trust for Health Care to the Mitchell Center. Behavioral testing was performed in The University of Texas Medical Branch Rodent In Vivo Assessment Core. Expert technical assistance was provided by Wei Song and Dr. Narayana Komaravelli.
The authors declare no competing financial interests.
- Correspondence should be addressed to Dr. Kelly T. Dineley, 301 University Boulevard, Galveston TX 77555-0616. ktdinele{at}utmb.edu