Animals were bred and maintained in the Core Unit of Biomedical Research, Division of Laboratory Animal Science and Genetics, Medical University of Vienna. The animals lived in a separate experimental room 1 week before and throughout the experiment. Rats were housed individually in standard Makrolon cages filled with autoclaved woodchips (temperature: 22 ± 2°C; humidity: 55 ± 5%; 12 h artificial light/12 h dark cycle: light on at 7:00 am). The study was carried out in accordance the recommendation of the European Communities Council Directive of 24 November 1986 (86/609/EEC) evaluated by the ethics committee of the Medical University of Vienna, Vienna, Austria. The protocol was approved by Federal Ministry of Education, Science and Culture, Austria.

The hole-board consisted of a 1 m × 1 m board made of black plastic surrounded by translucent plexiglass walls. Each side of the wall bears proximal spatial cues. Distal cues were provided by room structures visible outside the board. Four out of sixteen regularly arranged holes (diameter and depth 7 cm) were baited (dustless precision pellets, 45 mg, Bioserv^®). The pattern of baited holes remained the same during the entire test. In order to avoid olfactory orientation pellets were also present in an area below the board. Rats were familiarized to the experimenter through 10 min handling sessions per day for 4 days prior to the experiment, followed by 2 days of habituation to the hole-board during which the animals were allowed to explore the maze for 15 min each day with access to food pellets. The weight of each rat was gradually reduced by food restriction to reach 85% of its free-feeding body weight before the start of the experiment. Tap water was given ad libitum. Thereafter the rats were trained for 3 days (five trials on day 1, four trials on day 2 and a retention trial at day 3). Every trial lasted for 120 s or until all four pellets were eaten. The apparatus was cleaned with 0.1% incidin between trials in order to remove the rat odor cues. The interval between two successive trials for an individual was 20 min. A camera mounted on the room ceiling recorded the performance of the rats in the maze. The visits of holes and removals of pellets were noted for each trial. In order to compare rats with similar levels of motivation, rats with less than 40 hole-visits in total over the 10 trials were excluded from the analysis. Reference memory errors were noted as the number of visits to the unbaited holes. The Reference Memory Index (RMI) was calculated using the formula: (first + revisits of baited holes)/total visits of all holes. All behavioral training/testing was performed during the light phase of the light–dark cycle. Rats were decapitated, the brains were rapidly removed and dissected on a Para Cooler (RWW Medizintechnik, Hallerndorf, Germany) at 4°C to obtain DG. The tissue was immediately stored at -80°C until further analysis.

All homogenization and centrifugation steps were carried out on ice and at 4°C, respectively. Brain tissues were homogenized in an ice-cold homogenization buffer [10 mM HEPES, pH 7.5, 300 mM sucrose, protease inhibitors (Roche)] using a Dounce homogenizer; the homogenate was centrifuged at 1000 × g for 10 min to remove cell debris and nuclei and the supernatant was collected. The pellet was resuspended again in the homogenization buffer and centrifuged at 1000 × g for 10 min. The pooled supernatants were then centrifuged at 100000 × g for 60 min to enrich membranes. The resulting membrane pellets were washed in 10 mM HEPES, pH 7.5, protease inhibitors (Roche) and solubilized in 50 mM TEAB buffer (Sigma–Aldrich), 7 M urea, 2 M thiourea, 4% CHAPS, 100 mM DTT and protease inhibitors (Roche). The protein concentration was determined by Pierce^TM 660 nm Protein Assay (Thermo Scientific).

Protein samples were digested overnight with trypsin (Promega) using filter-aided sample preparation (FASP) (Wisniewski et al., 2009) with 50 μg of protein per one reaction. The tryptic peptides were desalted using reversed-phase C18 stage tips (Rappsilber et al., 2007) and reconstituted in 50 μL of 50 mM TEAB (Sigma–Aldrich, Hamburg, Germany). The samples (50 μg in 50 μL) were subsequently differentially labeled by TMT10plex^TM Isobaric Label Reagent Set (Thermo Scientific) according to the manufacturer’s instructions. Using an actual peptide concentration determined by Pierce^TM Quantitative Colorimetric Peptide Assay (Thermo Scientific) the samples were pooled by mixing an equal amount of labeled peptides from individual samples. The pooled sample was separated by reversed-phase liquid chromatography (RP-LC) at high pH as previously described (Gilar et al., 2005) resulting in 25 fractions. The peptide fractions were analyzed by liquid chromatography-tandem mass spectrometry (LC–MS/MS) with two technical replicates per sample using the LC separation procedure described previously (Smidak et al., 2016). MS analysis was performed by the Thermo Scientific^TM Q Exactive^TM Plus Orbitrap mass spectrometer (Thermo Scientific) in positive ion mode with full-scan MS in the range of m/z 375–1400 at the resolution of 35,000 (at m/z 200). MS/MS scans were acquired at the resolution of 70,000 (m/z 200) through HCD fragmentation of 15 most intense ions at 32% normalized collision energy with a fixed mass of 100 m/z. Dynamic exclusion was enabled for 30 s and for unassigned, +1 and ≥ +8 charges.

MS raw data were searched against UniProtKB Rattus norvegicus complete proteome database (UP000002494, 31569 sequences, downloaded on December 2nd, 2016) using MASCOT 2.4 (MatrixScience) through Proteome Discoverer 2.1 platform (Thermo Scientific). The search criteria were as follows: trypsin with a maximum of two missing cleavage sites; fixed modification: carbamidomethylation (C); variable modification: oxidation (M); search mode: MS/MS ion search with decoy database search included; peptide mass tolerance ± 10 ppm; MS/MS mass tolerance ± 0.02 Da; and protein and peptide FDR 1%. Only proteins identified by two distinct peptides and one unique peptide were considered for the final dataset. The Reporter Ions Quantifier node in the Proteome Discoverer software was used to calculate the corresponding protein abundances based on the intensities of individual TMT reporter ion channels (126.13–131.14 m/z) corrected for reporter ions isotopic distribution. The protein abundances were normalized by the total intensity of each reporter in the PSMs (Peptide-Spectral-Matches) population. Differences in protein levels are expressed as log2-transformed normalized protein abundance ratio of aging versus young animal group evaluated by two-sided t-test using Perseus statistical package (version 1.3.0.4). p-Values < 0.05 with permutation-based FDR threshold 0.05 were used as cut-off for significance. Additional filtering for significantly changed proteins was applied based on log2 (protein ratio) threshold > |0.3|. Significantly changed proteins were analyzed for enrichment of GO annotations using ClueGO-Cytoscape platform (Bindea et al., 2009; Huntley et al., 2015) using GOA database 30.08.2017 and Benjamini–Hochberg correction (adjusted p-Values < 0.01). The lists of most representative pathways and biological functions/diseases, and protein interaction networks were generated by Ingenuity Pathway Analysis (IPA) software (Qiagen) using right-tailed Fisher’s exact test (p < 0.05). Functionally related protein clusters within the differential dataset were visualized by STRING v10.5¹ and MCL algorithm with confidence threshold 0.5 and inflation parameter 3.

The protein samples were separated by electrophoresis in 10% SDS-PAGE gel and transferred to PVDF membranes (GE Healthcare). The western blot procedure was performed as described previously (Shanmugasundaram et al., 2017) with minor modifications. Briefly, the membrane was blocked for 1 h in 5% non-fat milk, 1x TBS, 0.05% Tween-20 and probed with the corresponding primary and secondary antibodies. For detection of FMRP protein the membrane was incubated in 3% non-fat milk, 1x TBS, 0.1% Tween-20 with 1:1000 polyclonal mouse anti-FMRP antibody (MAB2160, Millipore) overnight at 4°C followed by incubation for 1 h at room temperature in 1:10000 HRP-conjugated anti-mouse secondary antibody (ab6728, Abcam). Actin as a loading control was detected by incubation in 3% non-fat dry milk, 1x TBS, 0.1% Tween-20 with 1:2000 anti-β-Actin antibody (4967S, Cell Signalling) overnight at 4°C followed by incubation for 1 h at room temperature in 1:10000 rabbit polyclonal antibody. The signals were quantified using ImageJ (NIH) software based on calculation of peak area and the intensity values for FMRP bands were normalized across the samples by actin control.

For the analysis of FMRP expression eight young and eight aging rats were deeply anesthetized with isoflurane and transcardially perfused with 4% paraformaldehyde solution in 0.1 M sodium phosphate buffer, pH 7.4. Brains were gently removed and post-fixed in the same fixative solution for 20–24 h at 4°C. Each brain was rinsed well with phosphate buffered saline (PBS) to remove formaldehyde and incubated in 30% sucrose with 0.02% NaN₃ at 4°C. Immediately prior to cutting, brains were embedded in O.C.T. Tissue-Tek (Sakura Finetek) and frozen at -20°C in a cryo chamber. Tissue was sectioned at 20 μm with a cryostat CryoStar NX50 (Thermo Scientific). Coronal cryo sections (Bregma 3.30–3.80 mm) were blocked in 5% normal goat serum (NGS) (G9023, Sigma–Aldrich), 0.3% Triton X-100 (T8787, Sigma–Aldrich), 2% BSA (23208, Thermo Scientific) in 0.1 M PBS 7.4 for 2 h at 22–24°C and incubated with primary antibody (mouse anti-FMRP 1:500 MAB2016, Merck) for 48 h at 4°C in 0.1 M PBS, 1% NGS, 0.1% Triton X-100, 0.1% BSA with continuous stirring. After three washing steps in 0.1 M PBS, sections were incubated with the secondary antibody (1:1000 AlexaFluor488 anti-mouse 4408S, Cell Signaling Technology) for 2 h at 22–24°C in 2% BSA. In addition, to remove lipofuscin-like autoimmunofluorescence, all sections were incubated in autoimmunofluorescence eliminator reagent (2160, Merck) for 5 min and then underwent six washing steps for 1 min each in 70% ethanol. DAPI staining was used to assess the gross cell morphology (D1306, Thermo Scientific). All sections were mounted with DAKO fluorescence mounting medium (Agilent Technologies) and analyzed by fluorescence microscopy. Immunoreactivity signal of FMRP in cells, processes and neutropil in the DG were acquired on a Zeiss LSM780 confocal microscope (Zeiss, Jena, Germany) using 20x magnification. The image analysis software ImageJ (NIH) calculated the percentage of FMRP-positive area per section by thresholding FMRP immunoreactivity above background levels. The cell body layer of DG granule cells and hilus were separately outlined as regions of interest (ROI) according to the DAPI signal in each slice. Immunofluorescence imaging and data analysis were duplicated, evaluating two brain slices per rat in total.

Data obtained from western blot and immunofluorescence microscopy was analyzed using Microsoft Excel 2013 or Prism 5 (GraphPad Software) and the difference between animal groups was evaluated by Student’s t-test with the significance cut-off p < 0.05. The behavioral data was analyzed with SPSS Version 22 software (IBM^® SPSS^® Statistics) using general linear model – repeated measures ANOVA for the differences over the entire training, or separately for days 1 and 2, and univariate ANOVA for the retention trial. The significance cut-off was set at p < 0.05. The values are expressed as mean ± SEM.

Behavioral testing was performed to assess hippocampal-dependent memory performance in the subarea of interest. The cohort of 18 young and 104 aging Sprague-Dawley rats were evaluated for spatial reference memory using the hole-board test. A significant effect of the entire training on memory performance (p = 0.032) was observed indicating that learning took place. Furthermore, the data showed a significant effect of the training-age factor interaction (p = 0.001) when the entire training was compared. The significant group differences were observed on day 2 (p = 0.001) and at the retention test at day 3 (p = 0.011) with the group of young animals performing significantly better than aging group (Supplementary Figure S1).

A quantitative MS-based approach was employed to identify proteins with significantly changed levels in the aging rats. The experimental design included two – young and aging – animal groups used in the previous behavioral test, five biological replicates per group. The membrane-enriched protein samples from DG were analyzed by LC–MS as TMT10plex labeling experiments in two sets representing two technical replicates per one biological replicate, i.e., two MS runs of each peptide fraction. Quantitative data for each individual are expressed as normalized protein abundance values for the corresponding TMT channel. A total number of 5832 proteins was identified in both groups, of which 153 were showing significantly different levels [p < 0.05, log2 (protein ratio) > |0.3|] between the groups. The complete list of significantly changed proteins as well as all identified proteins in the study is provided as supporting information (Supplementary Table S1).

The FMRP was unambiguously identified in the proteomic samples as downregulated [log2 (protein abundance ratio) = -0.39, p = 0.0005] in the aging group (Supplementary Table S1) which was confirmed by western blotting (Figure 1A). The aging-dependent changes for FMRP in hippocampus were further analyzed by immunoflourescence microscopy with eight young and eight aging rats of the same behaviorally tested cohort. Consistently with general reduction in FMRP levels, there was a significant decrease (p = 0.002) in the total FMRP immunoreactivity per DG region in aging (7.86 ± 2.38) compared to young (12.85 ± 2.21) animals (Figure 1B). Furthermore, the comparison of individual DG subareas showed a significant decrease in FMRP labeling in aging rats in both, hilus (14.27 ± 3.03 vs. 8.72 ± 2.15, p = 0.002) and granular cell layer (21.55 ± 3.80 vs. 13.09 ± 6.08, p = 0.01).

To dissect the global proteomic changes identified in the current study functional enrichment analysis was applied to differential protein dataset. The biological processes and molecular functions that may be altered in aging animals were identified based on Gene Onthology (GO) enrichment (p < 0.01). The results for biological processes showed the overrepresentation of GO terms associated with RNA metabolism, regulation of RNA splicing and stability, membrane raft organization, cell junction assembly, regulation of endocytosis, and synaptic plasticity. The most significant molecular functions were related to RNA binding activity, glycosaminoglycan binding activity, structural constituents of extracellular matrix and cytoskeleton (Figure 2A). For more detailed annotation the data were submitted to IPA software (Qiagen) that infers activation states (z-score) of the representative pathways and biological functions (p < 0.05). The aging-dependent downregulation of EIF2 signaling (z-score = 2.646) followed by dysregulation in 14-3-3 mediated signaling, axonal guidance signaling and cell junction signaling were found as the most significant pathways. The analysis of biological functions showed coordinated expression changes leading to an increase of RNA splicing (z-score = 2.00), cell-to-cell contact (z-score = 2.05), a decrease in development of neurons (z-score = -2.13) and formation of microtubules (z-score = -2.14) in the aging group. The complete results obtained from GO enrichment and IPA analysis are provided as a supporting information (Supplementary Table S2).

The STRING network (STRING v10.5 database²) of significant proteins formed several clusters of closely related members associated with previously enriched processes. Most of the clusters were connected with each other reflecting the functional closeness of the identified machineries (Supplementary Figure S2). These include small groups of proteins involved in the formation of membrane structures and membrane vesicle trafficking (FLOT1, FLOT2, PASCIN2, BIN1, CLINT1), proteins associated with cell junctions (GJB2 GJB6, GJA1), and separate cluster of proteoglycans and proteoglycan-linked proteins (HAPLN1, VCAN, BCAN, ACAN, NCAN, TNC) linked to the group containing guanidine nucleotide binding (G) proteins (GNB4, GNG13). The enrichment of RNA metabolic activity in the significant proteomic dataset was represented by two big interconnected clusters associated with RNA processing and translation. Both clusters are linked to RNA-binding regulators involved in RNA transport and stability including FMRP, CAPRIN1, the members of Pumilio (PUM) and CUG-BP, Elav-like (CELF) family, and YTH N(6)-methyladenosine RNA-binding proteins (YTHDF).

The IPA network analysis on the individual RNA/DNA – protein or protein – protein interactions of FMRP with the other significant proteins connected FMRP to nuclear FMRP interactor – NUFIP2, RNA-binding regulator – PUM2, purine rich element binding protein – PURB, splicing factor – SFPQ, BCL2 binding anthanogene – BAG3 and tubulin alpha-4A chain – TUBAA. Additionally, the network showed close association to several signaling nodes of ERK1/2 and SRC family and to general regulation of gene expression through histones and RNA polymerase (Figure 2B).