Loss of serum response factor in mature neurons in the dentate gyrus alters the morphology of dendritic spines and hippocampus-dependent behavioral tasks

Conditional Srf KO mice (mutant mice; Srf^{f/fCaMKCreERT2}) were generated as previously described (Kuzniewska et al. 2016). As a control, CreERT2-negative littermates were used (Srf^f/f; control mice; wild type [WT]). The mice were bred on a C57BL/6J genetic background. Adult male and female Srf^f/f and Srf^{f/fCaMKCreERT2} mice were intraperitoneally injected with 1 mg tamoxifen (TAM; cat. #T5648, Sigma) twice daily for 10 days, resulting in the translocation of Cre-recombinase to the nucleus. Experiments were performed at least 6 weeks after the TAM injection. The animals were bred in the Animal House of the Nencki Institute of Experimental Biology. The mice were housed in individual cages with a 12 h/12 h light/dark cycle, constant temperature (23–25 °C), and food and water available ad libitum. Both male and female mice were used for the experiments, except when marked otherwise. All of the experiments were performed by experimenters who were blind to mouse genotype. All of the work was conducted in accordance with the European Community Council Directive (86/609/EEC) and Animal Protection Act of Poland (Directive 2010/63/EU). The procedures were approved by the 1st Local Ethics Committee in Warsaw, Poland (permission no. 389/2012, 678/2015, and 144/2016).

Total RNA was isolated from the mouse hippocampus using the RNeasy Mini Kit (cat. #74104; Qiagen) as described by the manufacturer. DNA contamination was removed by digestion with DNase I (cat. #1023460, Qiagen). The RNA concentration was calculated from absorbance at 260 nm, and the 260/280 nm absorbance ratio determined the purity of RNA. RNA was reverse transcribed with SuperScript III or IV Reverse Transcriptase (cat. #18080-044, #18090050, Invitrogen) according to the manufacturer’s instructions. Quantitative real-time polymerase chain reaction (PCR) was performed using Fast TaqMan Master Mix (cat. #44456, Applied Biosystems) with an Applied Biosystems 7900HT Fast Real-Time PCR System using TaqMan probes (ThermoFisher). Fold changes in expression were determined using the ∆∆C_T relative quantification method. The values were normalized to relative amounts of GAPDH.

The animals were perfused with 4% paraformaldehyde (PFA). Brains were then fixed in 4% PFA overnight at 4 °C and cut into 40-μm coronal slices. For SRF staining, brain slices were stained with anti-SRF antibody (1:500; cat. #sc-13029; Santa Cruz Biotechnology) overnight at 4 °C followed by avidin/biotin complex (cat. #PK-6100, Vector) and visualized using SIGMAFAST™ DAB (cat. #D0426, Sigma). For Nissl staining, the brain sections were stained with 0.1% cresyl violet solution and 3% acetic acid for 5 min, washed, dehydrated, cleared in xylene, and coverslipped.

Twenty micrograms of protein extracts were run on polyacrylamide gels (cat. #4569033, BioRad). The standard procedure of Western blot was performed using anti-Ser 3-phospho cofilin 1 and cofilin 1 (cat. #3313, #5175, Cell Signaling), and β-actin (cat. #A1978, Sigma) antibodies. To ensure equal total proteins level, blots were reprobed with tubulin (cat. #T9026, Sigma) antibodies. For detection, the chemiluminescent method was used. To quantify individual bands, a scan of photographic films was analyzed by densitometry with GeneTools Software (Syngene).

Whole-cell patch-clamp recordings were carried out in voltage-clamp mode. Mice (male and female) were anesthetized with isoflurane and decapitated. Brain slices (coronal, 250 μm thick) prepared using Leica VT 1200S vibratome in ice-cold NMDG solution (135 mM NMDG, 1.2 mM KH₂PO₄, 1 mM KCl, 1.5 mM MgCl₂·6H₂O, 0.5 mM CaCl₂·2H₂O, 20 mM choline bicarbonate, 10 mM d-glucose, saturated with carbogen—95% O₂, 5% CO₂) were then transferred to the artificial cerebrospinal fluid (ACSF) solution (119 mM NaCl, 2.5 mM KCl, 1.3 mM MgCl₂·6H₂O, 1 mM NaH₂PO₄, 26 mM NaHCO₃, 20 mM d-glucose, 2.5 mM CaCl₂·2H₂O, saturated with carbogen), incubated for 12 min at 31 °C and then at least 1 h at room temperature. During recordings, slices were held in a recording chamber perfused with carbogenated ACSF solution supplemented with 100 μM picrotoxin and 0.5 μM tetrodotoxin and heated up to 31 °C. Granule cells of the upper blade of dentate gyrus were identified visually. Borosilicate glass capillaries (4–7 MΩ resistance) used for patch-clamp recordings were filled with Cs-based internal solution (130 mM Cs-gluconate, 20 mM HEPES, 0.4 mM EGTA, 3 mM TEA-Cl, 4 mM Na₂ATP, 0.3 mM NaGTP, 4 mM QX-314Cl; osmolarity: 285–290 mOsm, pH = 7.0–7.1). Data were acquired using custom algorithms in Igor Pro (Wavemetrics) with an NPI amplifier and digitized at 10 kHz with an ITC-18 InstruTECH/HEKA. Recorded currents were filtered at 2 kHz. Series and input resistances were monitored throughout the experiment. To measure miniature excitatory postsynaptic currents (mEPSCs) 10- to 20-min-long voltage-clamp recordings were collected. Miniature events were analyzed using the MiniAnalysis software (Synaptosoft). mEPSCs amplitude detection threshold was set to 7 pA. All mini events automatically detected by the software were verified visually by the experimenter.

Serum response factor KO and WT mice (male and female n = 8) were anesthetized and transcardially perfused with phosphate-buffered saline (PBS) and with 1.5% paraformaldehyde (PFA) at room temperature. The brains were placed in 1.5% PFA for 20 min for postfix and then transferred to ice-cold PBS for at least 20 min. Next, these brains were cut on a vibratome into 130-µm-thick slices and placed in PBS at room temperature for 1 h. Gene gun technique was used to label sections with tungsten particles (Bio-Rad, Hercules, CA, USA) coated with lipophilic dye DiI (1,1ʹ-dioctadecyl-3,3,3ʹ,3ʹ-tetramethyl indocarbocyanine perchlorate; d-3911, cat. #D282, ThermoFisher). The slices were incubated in PBS at the room temperature for about 3 h and next in 1.5% PFA in 4 °C overnight, which enables diffusion of the dye into neuronal processes and allows to visualize dendritic spines. Z-stacks of confocal images of the seven dendrites per animal from the middle molecular layer of the upper blade of dorsal DG were acquired with 561-nm laser line with Zeiss LSM780 confocal system. Neurons of immature morphology (very sparse dendritic spines) were excluded from the analysis. The semiautomatic SpineMagick! software was used to measure and analyze dendritic spines by obtaining the virtual skeletons (Ruszczycki et al. 2012). We used a scale-free parameter (the length/width ratio) which reflect the spine shape (Michaluk et al. 2011). Length/width ratio was calculated and plotted using a logarithmic scale. The density of spines was calculated as a number of spines per 1 μm of dendrite length. Spines’ shapes were divided into clusters and then sorted into two groups: “long” and “mushroom and stubby” spines using custom scripts (Jasinska et al. 2016).

Open field The apparatus had a wooden floor (50 cm × 50 cm) surrounded by walls (34 cm hight). Animals’ behavior was monitored by a video camera placed above the center of the apparatus. Adult females (WT n = 12, SRF KO n = 12, about 6 months old or older) were put individually in one corner of the open field facing the wall and were allowed to explore freely for 15 min. The floor of the apparatus was cleaned with 5% ethanol after each session. Data were analyzed using EthoVision 3.1 System (Noldus Information Technology), total distance traveled was counted.

Species-typical behaviors All species-typical behaviors tests were performed on the same group of animals (WT n = 10, SRF KO n = 7; adult female and male, about 6 months old or older), starting from overnight nesting test and then marble burying and digging.

Nesting Mice were housed individually in their home cages with standard bedding for at least 1 week. After that time paper towel (divided into six pieces 9 cm × 9 cm each) was placed in the middle of each cage and left overnight to assess the nest building ability of the mice. The following scoring system was used: (1) paper towel was mostly untouched (> 90%) and was left in the middle of the cage; (2) flat pad-shaped nest, mostly moved to a corner; (3) more complex nest with biting the towel but not gathered in one place; (4) nest with shredded paper to form a cup; (5) perfect nest, paper towel torn up to form a crater, where walls are higher than mouse body height. Nest building is spontaneous home cage behavior of mice. It indicates healthy functioning and well-being of an animal (Jirkof 2014; Moretti et al. 2005). It is also interpreted by some authors as social behavior (Crawley 2012).

Digging Clean cage was prepared with the flat surface of wooden, 5-cm-deep bedding. Mice were placed individually in the experimental cage. The duration of the test was 3 min and latency to start digging, a number of digging bouts and total duration of digging were recorded and analyzed. Coordinated movements of force and/or hindlimbs which displace the wooden bedding defined the digging. The new bedding was used for all tested animals. Marble burying measures spontaneous digging behavior, which is typical for mice and is dependent on hippocampal function (Deacon 2006; Deacon and Rawlins 2005).

To compare the distributions, Shapiro–Wilk normality test was performed. To test the differences between two groups, unpaired t test or Mann–Whitney test (nonparametric) was used. When required repeated-measures ANOVA was performed, followed by Bonferroni’s multiple comparisons post hoc test. To compare the cumulative distributions of mEPSCs amplitudes and frequencies, the Kolmogorov–Smirnov test was used. The number of animals and neurons which were used for analysis is provided in figures legend. Data on the graphs are expressed as cumulative probability or means ± standard errors of the means (SEM). The difference between the experimental groups was considered as significant when p < 0.05. Results were analyzed in GraphPad Prism software.

The genome-wide analysis of basal gene expression in animals with the adult deletion of neuronal SRF showed changes in mRNA that were restricted to genes that encode the actin cytoskeleton, such as Actb and Actg1 (Kuzniewska et al. 2016; Losing et al. 2017; Parkitna et al. 2010). To confirm the downregulation of actin expression in the hippocampus in SRF KO animals, we used inducible conditional Srf KO mice to ablate Srf exclusively in adult, excitatory forebrain neurons (Erdmann et al. 2007; Kuzniewska et al. 2016). Actb, Actg1, and Acta1 mRNA levels were analyzed in the hippocampus in mutant and control littermates using quantitative real-time PCR. We found decreases in the levels of mRNA of β-actin and non-muscle γ-actin in the hippocampus in SRF KO animals, but not in smooth muscle α-actin (Fig. 1a).

To assess the consequences of SRF depletion in adult neurons, we focused on the morphology of the hippocampus. Using Nissl staining, we confirmed the lack of gross neuroanatomical defects in SRF KO animals with low SRF expression in the hippocampus (Fig. 1c) (Kuzniewska et al. 2016). We also analyzed the size of the hippocampal formation in adult WT and SRF KO animals. No differences were found in the total weight of the hippocampus (Fig. 1b; Mann–Whitney test, p > 0.05).

We performed Western blot assay to confirm actin downregulation in DG neurons. We observed a ~ 50% drop in β-actin protein expression levels in the DG in SRF-deficient neurons (Fig. 1d, e). In non-neuronal cells, the downregulation of β-actin induces cofilin 1 phosphorylation at Ser3, leading to its inhibition (Liu et al. 2007). Cofilin 1 is an actin-binding protein that regulates actin filament dynamics, stimulating the severance of actin filaments. To test whether the low level of β-actin and γ-actin in adult neurons correlates with changes in the actin-binding proteins, we examined the expression of cofilin 1 in the DG. Western blot analysis revealed that cofilin 1 phosphorylation was upregulated in SRF KO DG extracts (Fig. 1f) and coincided with a lower level of β-actin (Fig. 1d).

Actin dynamics play a crucial role in neurons, especially in the regulation of dendritic spine morphology (Hotulainen and Hoogenraad 2010; Hotulainen et al. 2009). Therefore, we investigated whether a decrease in β-actin protein expression and increase in cofilin 1 phosphorylation affect dendritic spine morphology. We evaluated the effects of SRF deficiency on the density and morphology of dendritic spines of GCs in the molecular layer of the dorsal DG (Fig. 2a). The analysis of spine density showed no significant changes in the number of spines of DG neurons (upper blade; Fig. 2b; t test, p > 0.05). To assess whether SRF depletion modified dendritic spine morphology, we measured the length and width of the spines to evaluate spine shape (Michaluk et al. 2011). The cumulative probability graph shows a significant increase in length-to-width ratio in DG neurons from SRF KO animals (Fig. 2c; Kolmogorov–Smirnov test, p < 0.001) which testify change towards longer and thinner dendritic spines. To further investigate the way in which SRF depletion affects dendritic spine shape, we clustered spines into two morphological categories: long spines vs. mushroom and stubby spines (Fig. 2d) (Jasinska et al. 2016). Serum response factor KO mice exhibited a significant increase (~ 6 percentage points) in the frequency of long spines in the DG and a significant decrease in the population of mushroom and stubby spines (χ² test, p < 0.0001) compared with WT animals.

The morphology of dendritic spines under many conditions is highly correlated with changes in synaptic strength (Engert and Bonhoeffer 1999; Kasai et al. 2003; Yuste and Bonhoeffer 2001). To test whether SRF deletion affects basal glutamatergic transmission in adult neurons, α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor-mediated miniature excitatory postsynaptic currents (mEPSCs) were measured (Fig. 3). The comparison of cumulative probability of cellular responses between WT mice and SRF KO mice demonstrated a significant decrease in mEPSC amplitude (Fig. 3b; Kolmogorov–Smirnov test, p < 0.001) and frequency (Fig. 3d; Kolmogorov–Smirnov test, p = 0.01).

To determine whether changes in the morphology and function of DG neurons in SRF KO animals correlate with behavioral deficits, we compared locomotor activity in WT and SRF KO animals in the open field test. Serum response factor KO animals were significantly more active than WT animals during a 15-min exploration session (Fig. 4a). Differences between genotypes were apparent from the first minute of exploration until the end of the test session (Fig. 4b; repeated-measures analysis of variance [ANOVA], genotype: F_1,11 = 25.63, p = 0.0004, time: F_4,44 = 21.28, p < 0.0001, genotype × time: F_4,44 = 0.1032, p = 0.9808, ns).

Next, we evaluated the role of SRF using simple, hippocampus-dependent tasks, including digging, marble burying, and nest building (Deacon 2006; Deacon et al. 2002). In the digging paradigm, SRF KO animals performed fewer digging bouts (Fig. 5a; Mann–Whitney test, p < 0.05). Serum response factor KO animals also spent less time digging than control animals (Fig. 5b; Mann–Whitney test, p < 0.01). No difference in the latency to start digging was found between KO and WT animals (Fig. 5c; Mann–Whitney test, p > 0.05). In the marble burying test, SRF KO animals exhibited deficits, in which they buried fewer marbles than WT littermate controls (Fig. 5d, e; Mann–Whitney test, p < 0.01). Finally, SRF KO mice exhibited impairments in nest building compared with WT littermates (Fig. 5f, g; Mann–Whitney test, p < 0.01).