The Fragile X syndrome (FXS) as the most common monogenetic cause of cognitive impairment and autism indicates how tightly the dysregulation of synapse development is linked to cognitive deficits. Symptoms of FXS include excessive adherence to patterns that point to compromised hippocampal network formation. Surprisingly, one of the most complex hippocampal synapses connecting the dentate gyrus (DG) to CA3 pyramidal neurons has not been analyzed in FXS yet. Intriguingly, we found altered synaptic function between DG and CA3 in a mouse model of FXS (fmr1 knockout [KO]) demonstrated by increased mossy fiber-dependent miniature excitatory postsynaptic current (mEPSC) frequency at CA3 pyramidal neurons together with increased connectivity between granule cells and CA3 neurons. This phenotype is accompanied by increased activity of fmr1 KO animals in the marble burying task, detecting repetitive and obsessive compulsive behavior. Spine apparatus development and insertion of AMPA receptors is enhanced at postsynaptic thorny excrescences (TEs) in fmr1 KO mice. We report age-dependent alterations in TE morphology and in the underlying actin dynamics possibly linked to a dysregulation in profilin1 expression. TEs form detonator synapses guiding CA3 network activity. Thus, alterations described here are likely to contribute substantially to the impairment in hippocampal function and therefore to the pathogenesis of FXS.
The Fragile X syndrome (FXS) is the leading monogenetic cause for cognitive impairment and autism (Crawford 2001; Jacquemont et al. 2007; Bagni et al. 2012). Patients with FXS show attention deficits and autism spectrum disorder (ASD) symptoms like stereotypic behavior and excessive adherence to patterns (repetitive and stereotyped patterns of behavior, interests, activities as well as adherence to specific nonfunctional routines or rituals) (Cohen et al. 1988; Merenstein et al. 1996; Tsiouris and Brown 2004; Garber et al. 2008; Hernandez et al. 2009). One of the most important brain regions involved in pattern separation/completion and memory formation is the hippocampus (Guzowski et al. 2004; Sahay et al. 2011; Yassa and Stark 2011). In this respect, the mossy fiber pathway, connecting granule cells of the dentate gyrus (DG) to the CA3 region, is especially important for gating information transfer in the hippocampus and considered as central for pattern separation and memory formation (Leutgeb et al. 2007; McHugh et al. 2007; Bakker et al. 2008; Nakashiba et al. 2012).
FXS is caused by translational silencing of the fmr1 gene, which results in partial or full absence of the respective protein FMRP (Verkerk et al. 1991). This leads to an abnormal development of dendritic spines, which form the postsynaptic compartment in highly plastic brain regions, such as the neocortex and hippocampus (Comery et al. 1997; Braun and Segal 2000; Irwin et al. 2001; Nimchinsky et al. 2001; Galvez and Greenough 2005; McKinney et al. 2005). Spine abnormalities in the FXS have been studied extensively as they provide an intriguing example how aberrations in spine number, shape, and maturity may be correlated with behavioral and cognitive deficits (Rudelli et al. 1985; Wisniewski et al. 1985, 1991; Hinton et al. 1991; Comery et al. 1997; Irwin et al. 2000, 2001; Greenough et al. 2001; Nimchinsky et al. 2001; McKinney et al. 2005; Cruz-Martin et al. 2010; Bhakar et al. 2012; Portera-Cailliau 2012; He and Portera-Cailliau 2013).
In the hippocampus, abnormal spine numbers and shapes were described mainly for the CA1 region correlated to impairments in synaptic plasticity like a reduction in long-term potentiation (LTP) and an exaggerated mGluR-dependent LTD (Huber et al. 2002; Bear et al. 2004; Volk et al. 2007; Zhang et al. 2009; Cruz-Martin et al. 2012; Portera-Cailliau 2012; Padmashri et al. 2013). Yet, the mossy fiber synapse, as one of the central connections for hippocampal network function, has not been studied in the FXS mouse model before. These synapses comprise large mossy fiber terminals (LMTs) of the granule cells connecting to thorny excrescences (TEs) on proximal dendrites of CA3 pyramidal neurons in the “stratum lucidum.” Each TE consists of multiple postsynaptic densities, rendering this synapse especially strong (Amaral and Dent 1981; Chierzi et al. 2012). Together with their location close to the soma of CA3 neurons, this enables these synapses to drive the firing of CA3 neurons just by activation of a few TEs (Siegel et al. 1992). Accordingly, the mossy fiber synapse can be described as a “detonator” or “teacher” directing the storage of information in the CA3 network (Urban et al. 2001). This allows single granule cells to precisely time the activity of CA3 pyramidal neurons and thereby provide the necessary depolarization needed for Hebbian plasticity at the associational/commissural inputs (McNaughton and Morris 1987; Henze et al. 2000). The stimuli responsive nature of regular dendritic spines is driven by Hebbian synaptic plasticity, however, homeostatic synaptic plasticity is needed in addition to restrain neuronal activity within physiological boundaries (Turrigiano 2008). Recently, a key role of the mossy fiber synapse in mediating homeostatic synaptic plasticity was reported and shown to be the main locus of homeostatic synaptic plasticity in mature hippocampal neurons (Lee et al. 2013). Behavioral phenotypes appearing in ASDs, as hyperactivity and hypersensitivity, point toward an impaired homeostatic synaptic plasticity which therefore renders the mossy fiber synapse a key structure for investigating the cellular basis of ASD.
Here, we set out to characterize the mossy fiber synapse in a mouse model of FXS (fmr1 knockout [KO] mice). This connection has not been thoroughly analyzed in the context of FXS before. Interestingly, while spines have mostly been described to be delayed in maturation in the FXS mouse model, we found a novel role of FMRP in restricting synapse development as TEs were found to be premature during development compared with wild-type (WT) cells in both size and function. This might indicate that this pathway also displays a critical period where FMRP is of special relevance for the formation of synaptic connections between DG and CA3. The functional and structural alterations of the DG-CA3 connection described here can provide cellular mechanisms for some of the key symptoms of FXS as hyperactivity and repetitive behaviors and point to an important involvement of disrupted information processing in CA3 neurons. Therefore, the phenotypes described are of high importance for an understanding of the pathogenesis of FXS and ASD in particular and of hippocampal function in general.
Materials and Methods
All experiments performed were authorized by the animal welfare representative of the TU Braunschweig and the LAVES (Oldenburg, Germany, Az. §4 (02.05) TSchB TU BS). Animals were kept in standard condition cages, exposed to a 12-h dark/light cycle. Fmr1 KO mice in a FVB129 background (The Dutch-Belgium Fragile X Consortium 1994), Jackson laboratory (Strain number 004624) were used. Mice were genotyped according to distributor instructions using the following primer sequences: mutant allele: TGT GAT AGA ACT AGT GAG ACG TG, WT allele: TGT GAT AGA ATA TGC AGC ATG TGA, common CTT CTG GCA CCT CCA GCT T.
Organotypic hippocampal slice cultures (OHC) of hemizygous males and heterozygous females were prepared as previously described (Stoppini et al. 1991; Michaelsen-Preusse et al. 2014) and cultivated with BME medium containing 25% HBSS, 25% equine serum, 1 mM GlutamaX (Invitrogen) and 0.5% glucose. Seventy-two hours after slice preparation, antimitotic drugs were added to the medium and removed after additional 24 h by a 100% medium exchange. Once a week 50% of the medium was exchanged.
Organtoypic slices days of in vitro (DIV) 13–15 were incubated for 20 min at 32 °C in the recording chamber, perfused with artificial cerebrospinal fluid (aCSF) saturated with 95%O2/5% CO2 (all in mM: 125 NaCl, 2.5 KCl, 1 MgCl2, 2 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, 25 glucose, 0.0005 TTX, 0.01 bicuculline, 0.01 APV) via a peristaltic pump at a flow rate of 1 mL/min. Whole-cell recordings at CA3 pyramidal neurons were performed using 5–7 MΩ pipettes filled with intracellular solution (all in mM potassium gluconate 126, HEPES 10, KCl 4, Mg-ATP 4, Na phosphocreatine 10, Na-GTP 0.3, pH 7.3 set with KOH, 290–295 mOsM). To abolish the influence of mossy fiber activity, 4 µM DCG-IV was applied. Miniature miniature excitatory postsynaptic currents (mEPSCs) were recorded at −60 mV via voltage clamp recordings. A Multiclamp 700B amplifier (Molecular Devices) and a Digidata 1440 A digitizer (Molecular Devices) were used to obtain recordings. mEPSCs were recorded for 2 min in a 5-min interval over a period of 30 min (under control conditions and with DCG-IV application). Data were analyzed by using Mini Analysis (Synaposoft), with an amplitude threshold of 7 pA.
Analysis of In Vivo Morphology
Male mice were transcardially perfused with 4% paraformaldehyde (PFA) and 2% sucrose in 0.1 M PB (phosphate buffer: NaH2PO4·2H2O 0.04 mM; Na2 HPO4·2H2O 0.17 mM). The brain was then removed and the hemispheres separated, followed by a postfixation in the perfusion solution for 30 min at 4 °C. For 400 µm whole brain slices, hemispheres were washed with phosphate-buffered saline (PBS: KCl 2.7 mM; KH2PO4 1.5 mM; NaCl 137 mM; Na2HPO4 10.4 mM) and embedded in 2% Agar in 0.1 M PB. Cutting was performed using a vibratome (VT1000S, Leica). Subsequently, the slices were again treated with 4% PFA for 1–2 h. Slices were biolistically shot with DiI (Rauskolb et al. 2010). The slices were kept in PBS at room temperature (RT) overnight.
Mice were sacrificed by cervical dislocation and decapitated. Hippocampi were removed and fixed in 4% PFA in 0.1 M PB overnight. Hippocampi were transferred into 30% sucrose in 0.1 M PB until they were dehydrated. For coronal 30 µm hippocampal slices, a microtome (Frigomobil, Jung) was used. Hippocampal microtome slices were incubated with a blocking solution (10% goat serum, 0.2% Triton X-100 in PBS) for 1 h at RT. The tissue was incubated with the primary antibody (1:1000 synaptoporin, SySy, 1:500 synaptopodin, Sigma) in PBS containing 1% goat serum overnight at 4 °C. The secondary antibody (1:500 Cy2, Cy3 goat, Dianova) was diluted in PBS and incubated for 2 h at RT. In the last 10 min, DAPI (1:1000 diluted in PBS) was added. Organotypic slice cultures (OHC) were stained against synaptoporin or synaptopodin after a 2-h fixation in 4% PFA in 0.1 mM PB and washing with PBS. The OHCs were treated with 1% Triton-X in TBS (20 mM Tris/HCl, 0.14 M NaCl) overnight at 4 °C followed by blocking solution (20% goat serum in TBS) for 1 h. OHCs were incubated in the primary antibody solution (1% goat serum in PBS, antibody dilution 1:500) overnight at 4 °C. Secondary antibody was applied for (anti-rabbit Cy2 1:500, 1% goat serum in TBS) 3 h at RT. OHCs transfected with SEP-GluR1 (Addgene) were fixed and blocked for 1 h at RT (20% goat serum in PBS). To increase the signal, anti-GFP antibody (1:500 rabbit, Millipore) was used as described above while PBS was used instead of TBS. OHCs transfected with wheat germ agglutinin (WGA) were additionally labeled with an anti-Lectin antibody (1:500 rabbit, Sigma).
Transfection and Dye Electroporation
Organotypic slice cultures were either transfected at DIV11 via the Helios gene gun method (Biorad) or by single cell DNA electroporation. Electroporation of DNA constructs was performed at 5 V, 1 mA and a 220-Hz train at RT in HBSS (Invitrogen). For analyzing neuronal morphology, either a farnesylated eGFP (Clonetec) or mApple (Shaner et al. 2008; Michaelsen et al. 2010a, 2010b) was used. For profilin1 (PFN1) overexpression, cells were electroporated at DIV7 (together with mApple as well as control cells with mApple only). Fluorescence recovery after photobleaching (FRAP) experiments were performed using eGFP-actin. To visualize AMPA receptor clusters and insertion, SEP-GluR1 (addgene, Kopec et al. 2006) was electroporated, as well as truncated WGA (Yoshihara et al. 1999) for anterograde and retrograde tracing. For analysis of TE morphology, OHCs were electroporated with AlexaFlour594 (Abcam) with the following settings: 10 V, 10 ms (single pulses) and quickly fixed in 4% PFA in 0.1 M phosphate buffer.
Imaging and Image Analysis
For the analysis of dendritic complexity, an Axioplan 2 microscope equipped with an Apotome module (Zeiss) was used with a z-stack size of 1 µm and a ×20 objective (0.8 NA), Neurolucida software (Microbrightfield) was used to determine dendrite morphology. Images of the anti-synaptoporin staining for the analysis of mossy fiber projections were acquired with a ×10 objective (0.3 NA) at the same setup. Unless otherwise specified for all other imaging experiments, a confocal laserscanning microscope (Fluoview1000 setup, Olympus) was used, equipped with the following objectives: ×20 (NA 0.5), ×40 (NA 1.3), and ×60 (NA 1.0). For time-lapse imaging, organotypic slice cultures were incubated 20 min at 32 °C in an imaging chamber (RC-22, Warner Instruments), perfused with carbogenated aCSF (all in mM: 125 NaCl, 2.5 KCl, 1 MgCl2, 2 CaCl2, 1.25 NaH2PO4, 26 NaHCO3, 25 glucose) via a peristaltic pump at a flow rate of 1 mL/min. For detailed analysis of neuronal structure, z-stack (0.35 µm step size) images were taken. Image analysis was performed via ImageJ. A spine in the stratum lucidum on the proximal apical dendrites of CA3 pyramidal neurons was classified as a TE when the area exceeded at least three times the size of the average spine area in the “stratum radiatum.” TE area was measured taking the outline of a TE in z level images at their maximum expansion.
Fluorescence Recovery After Photobleaching
FRAP experiments on single TEs were performed as described in detail previously (Michaelsen-Preusse et al. 2014, 2016). In eGFP-actin expressing neurons, a single TE was bleached using the 405-nm laser line with a Power of 2.3–3 mW (ca. 30%) for 30 ms. A SIM scanner allowed parallel imaging of GFP emission at 488 nm. A time-lapse series was acquired with an interval of 2 s for a total time of 120 s. At this time point, a plateau in the fluorescence recovery was reached. The analysis was performed using ImageJ. The background corrected mean intensity values for each time point were calculated per spine and plotted against the time. The fluorescence intensity of each time point was normalized to the average value derived from five prebleaching images (relative fluorescence). Nonlinear curve fitting was performed in GraphPad Prism where the net recovery after photobleaching is provided by the following equation: Y = Y0 + (Plateau − Y0) * (1 − exp(−K * x)) where Y0 is the Y value when time is 0 directly after the bleaching impulse, Plateau is the Y value at infinite times, expressed as a fraction of the fluorescence before bleaching and was used to determine the dynamic actin pool (F-actin dynamic). The stable pool (F-actin stable) is the fraction of fluorescence, which does not recover within the imaging period of 2 min calculated as 1 − (F-actin dynamic), K is the rate constant, Tau is the time constant, expressed in seconds, it is computed as the reciprocal of K. From this equation, the actin turnover rate was calculated as the time at 50% recovery of prebleaching fluorescence levels.
Marble burying behavior experiments were performed to observe species stereotypic behavior in an age- and hippocampus-dependent manner (Deacon and Rawlins 2005). Therefore, standard mice cages were filled up to 5 cm with bedding. The bedding was flatted and compressed slightly. Twelve marbles (with same color and size) were placed in a standardized scheme. After 60 min, the number of marbles buried was counted.
EM Imaging and Analysis
For the electron microscopic study of the mossy fiber boutons (MFBs), slice cultures from WT animals and fmr1 KO mice were fixed in 1% PFA and 1% glutaraldehyde in 0.1 M phosphate buffer, pH 7.4. Following washing in phosphate buffer, the slice cultures were osmicated (1% OsO4, 30 min), washed in 0.1 M phosphate buffer and dehydrated in increasing concentrations of ethanol (block staining of the sections in 1% uranyl acetate in 70% ethanol for 1 h). The sections were then soaked in propylene oxide and finally embedded in Epon (Sigma) and hardened at 60 °C. The CA3 region was trimmed and thin sections were cut using an Ultracut (Leica) and diamond knifes. Thin sections were mounted on copper grids, contrasted in 2% uranyl acetate and lead citrate and viewed in the electron microscope at a magnification of ×5000. In thin sections from six cultures per condition, the areas of boutons and TEs of mossy fiber synapses were measured (at least 10 MFBs per culture using SIS software). Then, the ratio of TE area/bouton area was calculated.
Data Presentation and Statistical Analysis
When not described differently, data are shown as mean ± SEM. An αlevel of P < 0.05 was chosen to reject the null hypothesis. Two different groups were compared by an unpaired Student's T-test (Microsoft Excel), while the comparison of three or more experimental groups was performed via one-way ANOVA in GraphPad Prism.
Increased Mossy Fiber-Dependent Input to CA3 Pyramidal Neurons Accompanied by Altered AMPA Receptor Insertion and Spine Apparatus Formation in fmr KO
In fmr1 KO mice, synapse formation and maturation have been found to be compromised in the neocortex and hippocampus. However, the DG-CA3 synapse, one of the most powerful synapses in the CNS, has not been studied in this context yet. Therefore, we set out to characterize both synaptic function and morphology of this connection in detail. We compared WT mice with fmr1 KO littermates carrying a deletion of the exon V within the fmr1 gene, which mimics the total loss-of-function of the FMRP protein found in patients with the full mutation (The-Dutch-Belgian-Fragile-X-Consortium 1994).
Excitatory input to CA3 pyramidal neurons is in parts comprised of large MFBs of DG granule cells, which are especially powerful (Henze et al. 2002). As a first approach, we tested the function of this synaptic connection. Therefore, we recorded mEPSCs from CA3 pyramidal neurons (whole-cell mode, Vm = −70 mV, 0.5 µM TTX) in OHC (DIV13–15). As a first step, we recorded mEPSCs under baseline conditions in fmr1 WT and KO cells over 30 min with a duration of 2 min in 5 min intervals. To prove that the observed effects are due to the mossy fiber input only, we applied in a second step DCG-IV suppressing mossy fiber release function (Nicoll and Schmitz 2005). We found mEPSC frequency to be significantly increased by 66% in fmr1 KO CA3 pyramidal neurons in comparison with fmr1 WT cells while the amplitude was not altered (Fig. 1A–C). The application of 4 µM DCG-IV for at least 15 min reduced the frequency of both fmr1 WT and KO significantly to the same level, while the amplitude was not affected (Fig. 1A–C). Thus, the effect we observed upon the deletion of FMRP is due to altered mossy fiber-dependent input but not commissural inputs.
As a next step, we addressed the question what molecular mechanisms might lead to this altered synaptic function. For this purpose, CA3 hippocampal neurons in slice cultures were electroporated with mApple and SEP-GluR1 (Addgene). The expression of SEP-GluR1 is an established tool to study the insertion of GluR1-containing AMPA receptors at synapses (Nicoll and Schmitz 2005; Kopec et al. 2006, 2007; Makino and Malinow, 2009; Korte and Schmitz 2016). Here, the AMPA receptor subunit GluR1 is fused to superecliptic pHluorin (SEP), which allows for an analysis specifically of surface receptors. Seventy-two hours after transfection, cultures (DIV14) were either transferred into normal medium (baseline) or into medium with increased potassium levels (60 mM KCl, synaptic activation medium) for 20 min and afterwards transferred back to the former medium for 60 min. Interestingly, already under baseline conditions the SEP-GluR1 signal was stronger in fmr1 KO TEs as compared with WT neurons, indicating an increased content of surface AMPA receptors (Fig. 1D,E). Synaptic activation via KCl led to a significant increase in the SEP-GluR1 signal both in WT neurons (46%) and in fmr1 KO cells (91%); however, this increase was more pronounced in the absence of FMRP (Fig. 1E). The results thus indicate that TEs in fmr1 KO cells have an increased number of surface AMPA receptors, pointing to an increase in synaptic strength. Furthermore, the insertion of AMPARs upon increased synaptic activity is enhanced.
These findings indicate that there might indeed be an increased state of maturity for TEs of fmr1 KO neurons. Therefore, as a next step, we performed an anti-synaptopodin staining. Synaptopodin is an actin-binding protein localized to the spine apparatus, which is reported to be involved in synaptic plasticity and is mainly found in mature spines (Mundel et al. 1997; Deller et al. 2000; Vlachos et al. 2009; Zhang et al. 2013; Korkotian et al. 2014). CA3 pyramidal neurons of both genotypes expressing mApple were analyzed at DIV7 and DIV14 (Fig. 1F). The anti-synaptopodin staining revealed already at the early time point a significantly increased number of clusters per TE in fmr1 KO cells. This phenotype was even more pronounced at DIV14 (Fig. 1F,G). In line with these findings, the proportion of TEs without synaptopodin clusters was significantly higher in WT cultures both at DIV7 and DIV14 in comparison with fmr1 KO neurons (Fig. 1F,H).
Taken together, the deletion of FMRP strongly affected synaptic function due to altered TE maturation. Here, the question arises whether the strong functional alterations are in addition accompanied by alterations in morphology.
Both TE Synapses and Collateral/Commissural Synapses Are Affected by the Loss of FMRP
TEs are powerful synapses that are highly involved in modulating network activity in the CA3 region of the hippocampus. The majority of TEs are located on proximal apical dendrites of CA3 pyramidal neurons. We used single cell electroporation of Alexa594 in organotypic slice cultures of different ages to investigate TE morphology. Within the CA3 area, different subregions have been described (Lorente de Nó 1934; Li et al. 1994; Kesner 2013), therefore, we divided the CA3 region into an area distal to the DG, which corresponds to CA3a/b and an area proximal to the DG corresponding to CA3c (Lorente de Nó 1934; Li et al. 1994; Kesner 2013). The cultures were prepared at P5 and at the earliest time point investigated (DIV7) no differences between fmr1 KO and WT neurons were found (Fig. 2A–E). The deletion of FMRP had no effect on TE density at DIV14 per se (Fig. 2B). However, at this time point, we were able to detect an increase in total TE number, especially at CA3a/b neurons (Fig. 2C). We also found an increased distance of the most distal TE from the cell body, indicating a widening of the mossy fiber projection in slice cultures as well (Fig. 2D). This phenotype can also be visualized by an anti-Synaptoporin staining. Synaptoporin is highly enriched in LMTs (Grabs et al. 1994; Grosse et al. 1998); and therefore, this staining allows for a detailed quantification of both mossy fiber band width and area. The staining revealed an increase in area, length, and width of the mossy fiber band (Fig. 2F–I). At DIV14, we also found the area of TEs to be significantly larger compared with control cells (Fig. 2E). Two weeks later at DIV28 both the maximum distance to the soma and TE area were indistinguishable from WT cells (Fig. 2D,E) whereas total TE density was still increased compared with control cells accompanied with an increase in TE number (Fig. 2B,C).
Taken together, our data point toward an increase in the number of TEs. This phenotype was in addition accompanied by an enlargement of TE area. Previous studies investigating spine morphology and the number of synapses formed by CA3 collaterals were inconclusive as some reported an immature spine phenotype comparable to the one described for CA1 neurons (Huber et al. 2002; Bear et al. 2004; Volk et al. 2007; Bilousova et al. 2009; Cruz-Martin et al. 2010, 2012; Portera-Cailliau 2012; He and Portera-Cailliau 2013; Padmashri et al. 2013), whereas others failed to detect differences between fmr1 KO cells and WT neurons (Levenga et al. 2011). When we investigated spine number and morphology in organotypic slice cultures at DIV14, the time point where also the TE phenotype was strongest, we found spine density unaltered (Fig. 2K), whereas spine length (Fig. 2J,L) and the proportion of thin spines were significantly increased in fmr1 KO cells compared with WT neurons (Fig. 2M). At this point, it is interesting to note that our data provide a first example that different subtypes of spines on CA3 neurons were affected differentially by the loss of FMRP, a fact that might have a strong impact on information processing in these cells. In order to conclude our analysis of CA3 pyramidal neurons, we analyzed the complexity of the dendritic tree using the Sholl analysis method (Sholl 1953). With this approach, we found a significant increase in the overall number of nodes and intersections (Fig. 2N–P), which was consistent with a significantly increased dendritic complexity at a distance between 160 and 200 µm from the soma. The presence of TEs on apical dendrites started directly at the soma and they spread up to 70 µm along the dendritic tree (Fig. 2D); therefore, the increased number of TEs in mutant cells was not a consequence of an increase in dendritic complexity as the number of intersections within the first 100 µm did not differ between fmr1 KO and WT cells (Fig. 2Q).
TE Motility and Actin Dynamics Are Altered in fmr1 KO Neurons
It has been shown previously that spatial learning and environmental enrichment increased the number and size of LMTs (Schöpke et al. 1991; Pleskacheva et al. 2000; Ramı́rez-Amaya et al. 2001; Galimberti et al. 2006; Holahan et al. 2006; Rekart et al. 2007; Gogolla et al. 2009; Ruediger et al. 2011) and that activity regulates morphological plasticity of the presynaptic compartment (Chierzi et al. 2012). Moreover, also TEs display structural plasticity (Frotscher et al. 1977; Zhao et al. 2012). Therefore, as a next step we investigated short-term changes in TE morphology.
TEs show a great range of structural variations (Wilke et al. 2013). Previous publications revealed TEs as structures with multiple synapses (Amaral and Dent 1981; Chierzi et al. 2012); hence, structural plasticity of the postsynaptic compartment might play an important functional role for the output of CA3 neurons. To analyze TE dynamics, we transfected CA3 pyramidal neurons at DIV11 OHC with mApple. For the analysis of TE motility under baseline conditions, stretches of proximal apical dendrites were imaged at 5 min intervals for 20 min in DIV14 cultures. Remarkably, the deletion of FMRP caused a significant decrease in the average change in TE area per 5 min interval (Fig. 3A,B).
Changes in spine shape are correlated to the dynamic actin cytoskeleton, which is highly enriched in dendritic spines (Fischer et al. 1998; Fukazawa et al. 2003; Honkura et al. 2008; Hotulainen et al. 2009; Bosch et al. 2014; Michaelsen-Preusse et al. 2016). We were therefore interested to know whether the decrease in structural plasticity could be directly linked to alterations in actin polymerization rates. FRAP experiments were used to monitor actin dynamics in single TEs of fmr1 KO neurons and WT cells, an approach that we previously applied to monitor actin dynamics at spines in the CA1 region (Michaelsen-Preusse et al. 2014, 2016). As TE size differed significantly between fmr1 KO and WT cells, we first addressed the question whether diffusion could be affected which would interfere with the read-out of actin dynamics using GFP-actin FRAP. Cytoplasmatic eGFP was expressed in single CA3 neurons and FRAP experiments showed no alteration in the recovery rate after the bleaching impulse. This indicates that there are no significant changes in diffusion rate (Fig. 3C,D).
In our previous work, we showed that FMRP interacts with the mRNA of PFN1, an important modulator of actin dynamics in neurons (Michaelsen et al. 2010a; Michaelsen-Preusse et al. 2016), and that PFN1 levels are decreased in the hippocampus of fmr1 KO mice. Indeed, FRAP experiments using single CA3 neurons expressing GFP-actin revealed a significant decrease in actin polymerization rates in TEs of fmr1 KO mice (Fig. 3E–G). While actin polymerization was slower in TEs as indicated by an increase in the turnover time (Fig. 3F), the stable pool of actin filaments that did not recover during the 2 min time window was not significantly altered in FMRP-depleted cells (Fig. 3G). Further in line with a proposed role of PFN1 downstream of FMRP (Michaelsen-Preusse et al. 2016), the significant alteration in actin turnover between fmr1 KO and WT cells was abolished in cells overexpressing PFN1 (Fig. 3E–G).
So far we showed that in the absence of FMRP, TEs in cultivated neurons are larger and more numerous, their spread along the apical dendrite is increased and that TE motility is decreased together with decreased actin dynamics in single TEs. Next, we asked the question whether TEs would also be affected in vivo by the absence of FMRP.
TE Density Is Altered in a Region- and Age-Dependent Manner in fmr1 KO Mice In Vivo
To support our data by in vivo analysis, we investigated TE morphology and the mossy fiber projection in brain slices derived from mouse brains perfused with PFA. We used diolistics (Rauskolb et al. 2010) to label single CA3 cells with the lipophilic Dye DiI which due to its restriction to the membrane reveals even subtle morphological details (Fig. 4A). To investigate TE morphology, we chose the peak of TE development (P19) (Wilke et al. 2012, 2013) and 2 weeks thereafter (P30–40). We focused our analysis on the suprapyramidal mossy fiber band running in stratum lucidum. Here, we divided again the CA3 region into a CA3a/b and CA3c area (Fig. 4B–E), like described above. Interestingly, we found the density of TEs to be significantly increased only at CA3a/b neurons, whereas TE density of CA3c cells did not significantly differ from WT neurons at P19 (Fig. 4B), a phenotype comparable to the one investigated in cell culture. This differential phenotype was only transient as 2 weeks later (P30–40) the number of TEs was indistinguishable from WT neurons in both regions. Also the total number of TEs was transiently increased in CA3a/b neurons at P19 (Fig. 4C) in comparison with WT cells together with a larger spread along the apical dendrite. Thus, the maximal distance of the last TE to the neuronal soma was increased at this time point (Fig. 4D), whereas TE area was not altered (Fig. 4E). The observation of a widening of the distal mossy fiber projection (in CA3a/b) could be confirmed by an anti-synaptoporin staining (Fig. 4F–L). As expected from the TE analysis, a detailed quantification of the LMT projection revealed an elongation and widening especially in the most distal portion in fmr1 KO mice at P19 (Fig. 4F–I) which could, however, no longer be detected at P30–40 (Fig. 4F,J–L). From our in vivo analysis, it became obvious that the phenotype was in most parts comparable to the one found in cultured neurons. In order to obtain a complete overview of the mossy fiber pathway in the FXS mouse model, we also investigated the morphology and structural plasticity of the presynaptic partner, the LMTs.
LMT Morphology Is Altered in the FXS Mouse Model Accompanied by Increased CA3-DG Connectivity and Altered Compulsive Behavior
LMTs are complex structures as the MFB forms excitatory contacts with TEs, whereas filopodial extensions contact inhibitory interneurons (Blackstad and Kjaerheim 1961; Hamlyn 1962; Frotscher 1985; Acsády et al. 1998). Thereby, this connection allows feedforward excitation and feedforward inhibition of CA3 neurons (Ruediger et al. 2011). Single granule cells in DIV11 organotypic slice cultures were transfected with eGFP and the size of the core MFB as well as filopodia length and number were analyzed at DIV14 (Fig. 5A,B). In contrast to the increased size of TEs found in slice cultures derived from fmr1 KO animals, MFBs were significantly reduced in size (Fig. 5B). Filopodia length and number were not altered (Fig. 5B). Structural plasticity in the form of short-term changes in MFB area and filopodia length (referred to as motility) was investigated as described above in 5 min intervals for a total imaging period of 20 min. The motility of MFBs was, however, not significantly increased in fmr1 KO neurons, but filopodia motility was significantly reduced (Fig. 5C,D).
To analyze LMT morphology and the corresponding TEs in greater detail, electron microscopy (EM) was performed. Indeed, EM analysis supported the results obtained by confocal imaging. The TE/LMT ratio was significantly increased in fmr1 KO slices in comparison with WT neurons (Fig. 5E–G). The combination of a significantly increased TE area in the absence of FMRP (Fig. 5E,H) and at the same time a decreased LMT body area led to an especially pronounced effect (Fig. 5E,I).
As the number of TEs was increased both in vitro and in vivo, we were interested to learn whether this would be also reflected in the connectivity pattern between DG and the CA3 region. We used retrograde labeling via WGA to determine the number of granule neurons connected to a single CA3 cell in DIV14 organotypic slice cultures of both genotypes. A truncated form of WGA (Yoshihara et al. 1999) was expressed in a single CA3 neuron and an anti-lectin immunostaining was performed to detect WGA-positive cells in the granule cell layer (Fig. 5J–L). Interestingly, we indeed found an increased number of WGA-positive cells in the DG of fmr1 KO slices when compared with WT controls (Fig. 5J–L) indicating hyperconnectivity of the mossy fiber pathway.
Symptoms of FXS patients include stereotypic behavior and excessive adherence to patterns like repeated movements (Cohen et al. 1988; Hernandez et al. 2009), which can also be observed in the FXS mouse model (The-Dutch-Belgian-Fragile-X-Consortium 1994). In this respect, it is noteworthy that especially the mossy fiber pathway is known to be important for pattern completion (Leutgeb et al. 2007). As our phenotypes described here, suggested perturbation of DG to CA3 signaling in FXS, we used the marble burying task (Poling et al. 1981; Broekkamp et al. 1986; Handley 1991; Deacon 2006) to assess stereotypic behavior in our FXS mouse model (Fig. 5M,N). While the TE phenotype could be detected at P19 it was absent at the later time point investigated (P30–P40). Therefore, mice of both ages were studied. The results show that at the young age fmr1 KO animals buried significantly more marbles than their WT littermates whereas both genotypes were indistinguishable at P30–40 (Fig. 5N).
FXS as one of the leading causes for autism and cognitive impairment is characterized by hyperactivity, stereotypic behavior, and repeated movements in patients (Cohen et al. 1988; Hernandez et al. 2009), a phenotype that is replicated in the fmr1 KO mouse model of FXS (The-Dutch-Belgian-Fragile-X-Consortium 1994). Earlier studies already tried to correlate these behavioral phenotypes to an apparent defect in synaptogenesis, which manifests itself as alteration in the general number of spines as well as in their morphology and stability (Comery et al. 1997; Irwin et al. 2000; Huber et al. 2002; Cruz-Martin et al. 2010; Thomas et al. 2011; Portera-Cailliau 2012; He and Portera-Cailliau 2013). However, reports about spine density and morphology are controversial. The changes described in the literature vary with respect to the brain region investigated (neocortex vs. hippocampus), the onset and maintenance of the spine phenotype (alterations in spine density and/or an immature spine profile) and the neuronal subpopulations affected. Some studies described the phenotype as being transient and primarily restricted to development with no alterations in mature neurons whereas others found changes in spine density or morphology in adult animals (Portera-Cailliau 2012; He and Portera-Cailliau 2013). Yet, a common denominator of the spine phenotypes described so far is their immaturity (Bakker et al. 1994; Nimchinsky et al. 2001; Galvez and Greenough 2005; McKinney et al. 2005; Irwin et al. 2001; Grossman et al. 2010).
The excessive adherence to patterns as repeated movements in patients also indicates impairments especially in hippocampal network formation. Characteristic symptoms such as deficits in visual short-term memory, visual-spatial abilities, processing of sequential information, and attention deficits as well point to impairments in hippocampus formation in FXS patients (Ojemann et al. 1988; Creutzfeldt et al. 1989; Cianchetti et al. 1991; Reiss and Freund 1992; Murray et al. 1993; Maes et al. 1994; Sobesky et al. 1994; della Rocchetta et al. 1995; Bouras et al. 1998; Cornish et al. 2004). Moreover, hippocampal volume is increased both in FXS and in ASD individuals and the severity of symptoms can be correlated to the magnitude of the hippocampal phenotype (Reiss et al. 1994; Kates et al. 1997; Sparks et al. 2002; Salmond et al. 2005; Endo et al. 2007).
In pyramidal neurons of the CA1 region, an overabundance of long and thin spines has been described before (Antar et al. 2006; Grossman et al. 2006; de Vrij et al. 2008; Bilousova et al. 2009; Grossman et al. 2010; Levenga et al. 2011; Swanger et al. 2011). Also for the other principal cell types (CA3 pyramidal neurons and granule cells of the DG), immature spine profiles were observed (Bilousova et al. 2009; Grossman et al. 2010; Levenga et al. 2011). Interestingly, so far a detailed morphological analysis was restricted only to the prototypical mushroom spines which build the postsynapse to the Schaffer collaterals or recurrent collaterals of CA3 neurons and to the perforant path fibers. Yet, one of the most important synaptic connections directly involved in pattern separation and completion in the hippocampus has not been studied in the FXS mouse model. The DG is highly involved in pattern separation whereas especially the mossy fiber connection to the CA3 region is involved in pattern completion (Leutgeb et al. 2007; Bakker et al. 2008). Both types of information processing are affected in FXS and are likely to contribute to the excessive adherence to patterns observed in patients (Cohen et al. 1988; Hernandez et al. 2009).
In the work presented here, we therefore focused on function and structure of one of the most powerful and morphologically complex synapses in the hippocampus, the synapse formed by the axons of granules cells with proximal dendrites of CA3 pyramidal neurons. In contrast to other synaptic alterations in FXS described above, we found the postsynaptic compartment—the TEs—intriguingly premature in fmr1 KO mice. TEs were increased in size especially during a critical time window (P19, DIV14 in organotypic slice cultures) and this was accompanied by a widening of the stratum lucidum in the fmr1 KO mice compared with WT littermates. At a later time point (DIV28, P30–40), this phenotype was abolished. It was reported previously that the infrapyramidal and intrapyramidal mossy fiber projection was even reduced in adult mice whereas the suprapyramidal projection was not altered (3–4 months of age) (Mineur et al. 2002). Besides the morphology of the TEs, the synaptic function is altered as well. We could show that the mossy fiber-dependent synaptic input, which is received by TEs is drastically increased. Most notably, also the number of surface GluR1-containing AMPA receptors in TEs was increased compared with WT cells already under baseline conditions. An elevation of activity via KCl stimulation led to an increase in the amount of GluR1 both in WT and fmr1 KO neurons; however, this increase was significantly stronger in fmr1 KO TEs pointing to dysregulated synaptic plasticity at the TE synapse in the absence of FMRP. Additional support for alterations in synaptic function came from our analysis of synaptopodin clusters, which indicated an increased number of TEs displaying already a spine apparatus at DIV7 in fmr1 KO neurons compared with control cells. At DIV14, the number of synaptopodin cluster was even further increased in TEs of fmr1 KO neurons whereas WT neurons only showed a spine apparatus in a significantly lower number of synapses. This was accompanied by a decrease in motility and slower actin dynamics in TEs, indicating increased morphological stability that could be rescued by an overexpression of PFN1. We previously reported that FMRP is able to bind PFN1 mRNA, which is an important actin-binding protein promoting actin polymerization (Michaelsen-Preusse et al. 2016). Additionally, the protein levels of PFN1 were found to be decreased in hippocampal tissue obtained from fmr1 KO mice compared with WT mice. Interestingly, the mRNA of the brain-specific isoform profilin2a (PFN2a) is not bound by FMRP (Michaelsen-Preusse et al. 2016). This is especially surprising since PFN1 and PFN2a display 65% amino acid sequence identity and high biochemical similarity (Gieselmann et al. 1995). In our recent study, we could reveal two distinct functions of both proteins in spines. While PFN1 is crucial for synaptogenesis, PFN2a is mandatory for adult spine plasticity. This is in line with results presented here where an overexpression of PFN1 could rescue the defect in actin dynamics in developing TEs. Therefore, the dysregulation of PFN1 in fmr1 KO mice might be indeed at least in large parts responsible for the documented age-dependent structural alterations of TEs.
Structurally and functionally premature TEs together with an increased connectivity of single CA3 pyramidal neurons to DG granule cells point toward hyperexcitability of CA3 pyramidal neurons since they are contacted by a larger amount of LMTs connecting to enlarged TEs with an increased number of synapses and therefore increased amount of inserted AMPA receptors. This hyperconnectivity is also represented by an increase in the frequency of mEPSCs but not the amplitude in fmr1 KO cells as we report here.
Taken together, we show that TE synapses are premature with a high abundance of spine apparatuses and increased surface GluR1 receptors. This is in contrast to the well-known spine immaturity of the CA1 region associated with decreased surface GluR1 levels and impaired LTP (Zhang et al. 2009; Cruz-Martin et al. 2010, 2012; Portera-Cailliau 2012; He and Portera-Cailliau 2013; Padmashri et al. 2013), and therefore reveals a new role of FMRP as a restricting factor for synapse development. In line with a previous study (Bilousova et al. 2009), we also found regular spines on CA3 neurons, likely contacting recurrent collaterals of CA3 pyramidal neurons, immature with a higher proportion of long and thin protrusions.
In contrast to most other studies describing spine abnormalities in FXS and in order to get a comprehensive view of the mossy fiber synapse, we investigated presynaptic structures in FXS as well. Interestingly, we found LMTs in fmr1 KO cultures to be significantly smaller in size compared with those of WT neurons. EM revealed that due to the size increase of TEs and the size reduction of LMTs the TE/LMT ratio was drastically increased in fmr1 KO neurons. Since FMRP is highly concentrated in FXS granules, self-contained functional packets, particularly observed in the LMTs of the mossy fiber pathway (Akins et al. 2012), our results indicate a direct effect of FMRP loss associated with dysregulation of protein synthesis at these structures. As previous investigators noticed an underdevelopment of TEs following input removal (Frotscher et al. 1977; Gaiarsa et al. 1992), we hypothesize that the presence of FMRP counteracts abnormal TE growth. These results emphasize that future experiments on spine morphology in FXS need to include the corresponding presynaptic structures to further elucidate the outcome for synaptic network development.
In addition to a detailed analysis of synaptic structure, we investigated behavior in fmr1 KO mice. The marble burying test is commonly used as a paradigm to detect repetitive and obsessive compulsive behavior, a phenotype highly dependent on hippocampal networks (Webster et al. 1981; Handley 1991; Deacon and Rawlins 2005). Here, we clearly show that young fmr1 KO animals display increased levels of repetitive behavior in the marble burying test. However, when the animals were tested at a later time point, when the premature TE phenotype had disappeared, the changes between fmr1 KO and WT littermates in marble burying were no longer observed.
Taken together, the immature spine profile of regular spines on CA3 neurons together with hyperconnectivity to the DG and premature TEs most likely have a significant effect on the output of CA3 pyramidal neurons, as these phenotypes might even enhance each other. Hence, information flow from the mossy fibers dominates over CA3 associative connections thereby likely resulting in detrimental effects for information processing. In this respect, it is noteworthy that TEs are highly plastic and can respond to different environmental inputs (Galimberti et al. 2006) as well as stress and take part in processes of learning and memory (Stewart et al. 2005). They have been described as independently tunable excitatory input in proximal apical CA3 neuron dendritic domains, which permit homeostatic adjustment without disturbing synaptic information storage at other synaptic contacts in the hippocampal network (Lee et al. 2013). The impairments in LMT-TE formation described here might therefore contribute to defects in homeostatic plasticity in the hippocampus in FXS and provide a possible explanation for some of the prominent symptoms as hyperactivity and increased anxiety in patients as well as higher risks for epilepsy and seizures (Gillberg et al. 1986; Hagerman et al. 1986; Musumeci et al. 1988; Musumeci et al. 1991; Musumeci et al. 1999; Sabaratnam et al. 2001).
In this study, we focused on the DG-CA3 connection and its potential relationship to ASD. Autistic patients show on the one hand excessive adherence to patterns (repetitive behavior) and on the other hand impairment in the detection of socially important patterns such as the recognition of facial expressions (Cohen et al. 1988; Merenstein et al. 1996; Tsiouris and Brown 2004; Garber et al. 2008). All these symptoms point to a prominent role of the DG and CA3 regions of the hippocampus as they are important for the encoding of new episodic memories (Squire 1992; Lisman 1999; Lisman et al. 2005; Leutgeb et al. 2007; Bakker et al. 2008). During this process, a crucial step is the amplification of the differences between new representations and already existing ones (pattern separation/pattern completion). It has been shown previously that LTP in the DG is reduced in fmr1 KO mice together with an impairment in adult neurogenesis (Eadie et al. 2009; Yun and Trommer 2011; Guo et al. 2012; Lazarov et al. 2012). Our data now further expand this view as we show that also the mossy fiber synapse is altered during development in the FXS mouse model, a phenotype that most likely contributes substantially to impairments in hippocampal network development and therefore to the pathogenesis of FXS and ASD.
K.M.P. and M.K. were supported by the Deutsche Forschungsgemeinschaft (KO 1674/8-1). M.F. was supported by the Deutsche Forschungsgemeinschaft (FR 620/14-1) and is a Research Professor for Neuroscience of the Hertie Foundation.
We wish to thank Diane Mundil for excellent technical assistance in the cell culture and Dagmar Drexler for her excellent assistance in the EM studies. Conflict of Interest: None declared.