Synaptic localisation of SRF coactivators, MKL1 and MKL2, and their role in dendritic spine morphology

The megakaryoblastic leukaemia (MKL) family are serum response factor (SRF) coactivators, which are highly expressed in the brain. Accordingly, MKL plays important roles in dendritic morphology, neuronal migration, and brain development. Further, nucleotide substitutions in the MKL1 and MKL2 genes are found in patients with schizophrenia and autism spectrum disorder, respectively. Thus, studies on the precise synaptic localisation and function of MKL in neurons are warranted. In this study, we generated and tested new antibodies that specifically recognise endogenously expressed MKL1 and MKL2 proteins in neurons. Using these reagents, we biochemically and immunocytochemically show that MKL1 and MKL2 are localised at synapses. Furthermore, shRNA experiments revealed that postsynaptic deletion of MKL1 or MKL2 reduced the percentage of mushroom- or stubby-type spines in cultured neurons. Taken together, our findings suggest that MKL1 and MKL2 are present at synapses and involved in dendritic spine maturation. This study may, at least in part, contribute to better understanding of the molecular mechanisms underlying MKL-mediated synaptic plasticity and neurological disorders.

ScieNtific REPORtS | (2018) 8:727 | DOI: 10.1038/s41598-017-18905-7 is highly expressed in the brain [17][18][19] and double knock-out of MKL1/MKL2 in the brain displays phenotypical similarity to SRF conditional knock-out mice 20 . These studies suggest that in addition to SRF, its cofactors (specifically MKL family members), play important brain roles. Recent observations suggest that mutations or single nucleotide polymorphisms (SNPs) of MKL1 and MKL2 genes are found in patients with schizophrenia 21,22 and autism spectrum disorder (ASD) 23,24 , respectively. Therefore, elucidating the molecular mechanism by which MKL family members control neuronal function may shed light on understanding brain function and disorders 25 . Although MKL regulates dendritic morphology at the cellular level 17,18,26 , subcellular localisation of MKL1 and MKL2, and their role in spine maturation remain unknown. In this study, we generated anti-MKL1-and MKL2-specific antibodies that are suitable for western blotting and immunostaining. Using these newly characterised antibodies, we show synaptic localisation of MKL1 and MKL2 proteins in cultured neurons and a synaptosomal fraction. Furthermore, we show that MKL1 and MKL2 control dendritic spine morphology by RNA interference-mediated knock-down in cultured neurons. Neuronal knock-down of MKL1 and MKL2 reduced the percentage of mushroom-or stubby-shaped spines, suggesting novel roles of MKL in spine maturation.

Evaluation of new antibodies by western blotting and immunostaining.
A previous study demonstrated that the N-terminal region of MKL1 shows high homology to MKL2 13 . Therefore, to obtain MKL1 or MKL2 antibodies that specifically recognise MKL1 and MKL2, we chose the C-terminal regions of MKL1 and MKL2 as antigens (which show low homology [37.2%] to each other), and include transcriptional activation domains. First, we evaluated specificity of these new MKL1 and MKL2 antibodies. Cell lysates were prepared from NIH3T3 cells expressing FLAG-empty vector (FLAG-empty), FLAG-MKL1, HA-empty vector (HA-empty), or HA-MKL2, and subjected to western blotting (Fig. 1a). MKL1 antibody strongly reacted with samples expressing FLAG-MKL1 (Fig. 1a, lane 2) and appeared to detect endogenous MKL1 (Fig. 1a, lanes 1, 3, and 4). Whereas, MKL2 antibody strongly reacted with the sample expressing HA-MKL2 (Fig. 1a, lane 4) and appeared to recognise endogenous MKL2 (Fig. 1a, lanes 1, 2, and 3). Even though HA-MKL2 was expressed in the sample (Fig. 1a, lane 4), the intensity of the band detected by MKL1 was the same as the empty vector bands (Fig. 1a, lanes 1 and 3). Similarly, even though FLAG-MKL1 was expressed in the sample (Fig. 1a, lane 2), the intensity of the band detected by MKL2 was the same as the empty vector bands (Fig. 1a, lanes 1 and 3). These results suggest that MKL1 antibody does not cross-react with MKL2, and MKL2 antibody does not cross-react with MKL1. We also determined if these antibodies recognise rat MKL1 and MKL2 (Fig. 1b)

Nuclear translocation of MKL in NIH3T3 cells and cortical neurons.
Rho signalling promotes nuclear translocation of MKL1 16,27 . These findings encouraged us to detect nuclear translocation of MKL1 and MKL2 using our antibodies. Thus, we examined whether nuclear translocation of MKL1 and MKL2 was induced by constitutively active (ca) mDia, which is a downstream effector of Rho (Fig. 2, Supplementary Fig. S3). In NIH3T3 cells expressing camDia, the cellular distribution of MKL1 was 56%, 11% and 33% in the nucleus, cytoplasm, or both, respectively ( Supplementary Fig. S3a). Conversely, in NIH3T3 cells expressing constitutively inactive (cia)mDia, a negative control, the distribution of MKL1 was 5%, 65%, and 30% in the nucleus, cytoplasm, or both, respectively ( Supplementary Fig. S3b). The distribution of MKL2 in NIH3T3 cells expressing camDia was 50%, 12%, and 38% in the nucleus, cytoplasm, or both, respectively ( Supplementary Fig. S3c). In NIH3T3 cells expressing ciamDia, 24%, 30%, and 46% of MKL2 was distributed in the nucleus, cytoplasm, or both, respectively ( Supplementary Fig. S3d). Therefore, camDia effectively induced nuclear translocation of MKL1 and MKL2 in NIH3T3 cells. In cortical neurons expressing camDia, the distribution of MKL1 was 2%, 64%, and 34% in the nucleus, cytoplasm, or both, respectively ( Supplementary Fig. S3e). In cortical neurons expressing ciamDia 1%, 88%, and 11% of MKL1 was distributed in the nucleus, cytoplasm, or both, respectively ( Supplementary Fig. S3f). The distribution of MKL2 in cortical neurons expressing camDia was 8%, 43%, and 49% in the nucleus, cytoplasm, or both, respectively ( Supplementary Fig. S3g). In cortical neurons expressing ciamDia, 7%, 61%, 32% of MKL2 was found in the nucleus, cytoplasm, or both, respectively ( Supplementary Fig. S3h). Thus, the percentages of MKL1 and MKL2, localised in the nucleus only, were very small, and did not alter much even though camDia was expressed in cortical neurons. However, the levels of MKL1 and MKL2 in both the nucleus and cytoplasm were increased, and in turn, the levels in the cytoplasm were decreased in the presence of camDia. These findings indicate that the MKL1 or MKL2 antibodies are suitable for nucleocytoplasmic staining in NIH3T3 cells and cortical neurons.
Biochemical detection of MKL1 and MKL2 in subcellular fractions of rat brain. Next, we investigated subcellular localisation of MKL1 and MKL2 protein in brain extracts and cortical neurons. Initially, we performed biochemical detection of MKL1 and MKL2 protein by subcellular fractionation of rat brain. A strong signal was obtained from precipitates of postsynaptic density (PSD) fractions (PPT) using anti-PSD-95 antibody, a postsynaptic marker (Fig. 3, PSD-95). In addition, a strong signal was obtained from the crude synaptic vesicle (CSV) fraction using anti-synaptophysin antibody, a presynaptic marker (Fig. 3, synaptophysin). These results indicate successful subcellular fractionation. Both MKL1 and MKL2 were enriched in PSD (PPT), synaptic membrane 3 (SM3), and CSV fractions ( morphology was visualised by expressing green fluorescent protein (GFP) followed by immunostaining with GFP antibody and MKL1 or MKL2 antibody (Fig. 4a). We found MKL1 and MKL2 protein enriched in cortical neuronal cell bodies, with punctate staining for both MKL1 and MKL2 protein observed in dendrites. Puncta appeared to be localised within nuclei. Higher magnified images of dendrites revealed that some puncta were localised at GFP-positive dendritic spines (Fig. 4b, arrowheads) where PSD-95 and MKL1 or MKL2 signals merged (Fig. 4c, arrowheads), supporting postsynaptic localisation of MKL1 or MKL2. Furthermore, synaptophysin and MKL1 or MKL2 signals also merged (Fig. 4d, arrowheads), demonstrating pre-synaptic localisation of MKL1 or MKL2 protein as well. However, not all puncta stained with anti-MKL1 or anti-MKL2 antibodies colocalised with puncta stained with anti-PSD-95 or anti-synaptophysin antibodies.   irregular in left panel), they tended to increase in cells expressing MKL1 or MKL2 shRNA. In contrast, MKL1 and MKL2 shRNA expression had no effect on dendritic spine density (Fig. 5c, right panel). To confirm the role of MKL in regulation of spine morphology, hippocampal cultures were used (Fig. 6). Expression of shMKL1 and shMKL2 (whose sequences are different from those used in Fig. 5) significantly decreased the percentage of dendritic spines with mushroom-or stubby-shaped structures, and increased the percentage of long, filopodia spines ( Fig. 6b and Supplementary Fig. S4a). Consistent with the results in cortical neurons, shMKL1 and shMKL2 did not affect spine density ( Fig. 6c and Supplementary Fig. S4b). Taken together, these results indicate that MKL1 and MKL2 may play a key role in promoting dendritic spine maturation.

Discussion
In this study, we initially generated and evaluated MKL1-or MKL2-specific antibodies. Our first biochemical analysis by western blotting showed that MKL1 and MKL2 antibodies do not cross-react with MKL2 and MKL1, respectively, and specifically recognise MKL1 and MKL2 (Fig. 1a). Using cell lysates from rat cortical neurons, we found that these antibodies react with rat MKL1 and MKL2 (Fig. 1b). Antibody absorption tests with antigen-conjugated Sepharose beads and immunocytochemical knock-down experiments with shMKL1 and shMKL2 revealed that immunostained signals were derived from endogenous mouse and rat MKL1 and MKL2 (Supplementary Figs S1and S2).
Our previous biochemical analyses with commercially available antibodies indicated that MKL is localised in both the cytoplasm and nucleus of cortical neurons 18 . Consistent with this, we confirmed nucleocytoplasmic staining of NIH3T3 cells and cortical neurons with our MKL1 and MKL2 antibodies (Figs 2 and 4a). Furthermore, our antibodies stain nuclear MKL1 and MKL2 due to excessive activation of Rho signalling by constitutively active mDia in NIH3T3 cells (Fig. 2a). The quantitative data suggest that MKL1 and MKL2 efficiently translocate into the nucleus in NIH3T3 cells expressing camDia (Supplementary Fig. S3). In contrast, in cortical neurons, the levels of MKL1 and MKL2, which were located in the nucleus only, were not altered much by camDia ( Supplementary Fig. S3). Rather, the levels in both the nucleus and cytoplasm were up-regulated, and the levels in the cytoplasm were down-regulated ( Fig. 2c and Supplementary Fig. S3). Previously, we detected nuclear MKL in cells expressing camDia using another antibody that recognised the N-terminal of MKL 27 . Although our antibodies distinctly recognise the C-termini of MKL1 and MKL2, we also successfully detected alteration of cellular localisation of MKL induced by camDia. MKL in NIH3T3 cells tends to more readily translocate than MKL in cortical neurons. Efficient nuclear translocation of MKL1 and MKL2 might be dependent upon cell type. Taken together, these results indicate that our antibodies are suitable for studying cellular localisation and expression of MKL1 and MKL2.
Next, we performed western blotting by subcellular fractionated proteins (Fig. 3) and immunostaining (Fig. 4). Western blotting shows faint MKL1 and MKL2 bands in cytoplasmic fractions compared with those in CSV and PSD fractions. Contradictorily, cytoplasmic staining was predominant, rather than synaptic staining (Fig. 4). There is one possibility to explain this inconsistency. The CSV and PSD fractions include not only CSV and PSD but also a variety of membrane components. With immunostaining, there are a lot of puncta in cell bodies of cortical neurons. Collectively, punctate cell body signals may imply that MKL1 and MKL2 translocate from the cytoplasm to synapses or dendrites or from synapses or dendrites to the cytoplasm via vesicular (membrane) transport. Thus, membrane vesicles containing MKL1 and MKL2 in cell bodies might have been concentrated  We immunocytochemically confirmed that some MKL1 and MKL2 puncta are localised postsynaptically although not all MKL1 and MKL2 puncta are colocalised with PSD-95-positive puncta (Fig. 4). Considering our biochemical analysis with subcellular fractions (Fig. 3), postsynaptic localisation of MKL1 and MKL2 is, at least in part, observed. These findings suggest that MKL1 and MKL2 play important roles in synaptic transmission and function at the PSD. To investigate this possibility, we determined if MKL1 and MKL2 are involved in modification of dendritic spine morphology (Fig. 5). ShMKL1 and shMKL2 reduced the percentage of mushroomor stubby-shaped structures, suggesting that MKL may be involved in spine maturation. Consistent with the results obtained in cortical neurons, hippocampal neurons expressing shMKL1 or -2 reduced the percentage of mushroom-and stubby-shaped spines (m and s) and did not affect dendritic spine density (Fig. 6b and c). Furthermore, shMKL1 or -2 expression increased long-and filopodia-shaped spines (l and f), supporting our finding that MKL1 and MKL2 influence spine maturation (Fig. 6b). To exclude the possibility of shRNA off-target effects, we used a separate set of shMKL1 and shMKL2 for morphological analyses. As shown in Figs 5,6 and Supplementary Figure S4, we obtained the same results with shRNAs targeting diverse sequences, providing convincing evidence that our results are due to MKL1-or MKL2-specific knock-down and not off-target effects. Taken together, MKL1 and MKL2 may act at the PSD to promote dendritic spine maturation. Since MKL1 and MKL2 contain actin-binding motifs, they can bind to G-actin 16 . In contrast, MKL1 can bind to filamin A as an F-actin binding protein 28 . Moreover, F-actin polymerisation is crucial for dendritic spine maturation 29 . Therefore, MKL1 and MKL2 may act as regulators for actin-based alteration of dendritic spine morphology at the PSD. Since MKL1 and MKL2 are SRF transcriptional cofactors, it is possible that MKL1 and MKL2 promote dendritic spine maturation at the transcriptional level. Indeed, microarray analysis of conditional MKL1 and MKL2 knock-out in mouse brain demonstrates regulation of a subset of genes 20 . Some of them (including actin, gelsolin, and pctaire1) are related to cytoskeletal regulation 20 . So far, we cannot exclude the possibility that MKL1 and MKL2 promote spine maturation via cytoskeletal gene regulation.
As well as the presence of MKL in the PSD fraction, MKL was detected in CSV and SM fractions (Fig. 3). To clarify this significance, future studies may be required.
Aberrant dendritic spine morphology is associated with neurological disorders including schizophrenia and ASD 30 . Corroboratively, SNPs of the mkl1 gene are found in patients with schizophrenia 21,22 . In addition, mkl2 gene mutations or SNPs are candidates for ASD 23,24 . Therefore, these mutations may disturb MKL1 and MKL2 function at synapses. In this study, using our newly-generated antibodies, we show that MKL1 and MKL2 are localised at synapses and may contribute to spine maturation. Our results provide improved understanding of the regulatory mechanism mediated by MKL family members in brain function and disorders.

Methods
Animals. Sprague-Dawley dams for primary cultured cortical neurons were purchased from Japan SLC Inc.
(Hamamatsu, Japan). Wistar P1 pups were used for primary cultured hippocampal neurons. All experiments were approved by the Animal Experimentation and Ethics Committee of Kitasato University School of Medicine. Experiments were performed in accordance with the guidelines of the Animal Care and Experimentation Committee of the University of Toyama, Sugitani Campus and Animal Care Welfare Committees at the University of Yamanashi, and according to guidelines of the First Warsaw Ethical Committee on animal research, with appropriate permission. Every effort was made to minimise the suffering of animals.
The following antibodies were used at the indicated dilutions: goat CF594-conjugated anti-guinea pig IgG Western blotting. Western blotting was performed as previously described 18 . Protein detection was performed using the enhanced chemiluminescence protocol (ECL Western blotting detection reagents, GE Healthcare; ImmunoStar LD, ImmunoStar Zeta, Wako Pure Chemical Industries, Ltd., Osaka, Japan).

Generation of antibodies.
Generation of antibodies against MKL1 or MKL2 was as follows. The C-terminal regions of mouse MKL1 (333 amino acid residues) and mouse MKL2 (333 amino acid residues) proteins were bacterially expressed as maltose-binding protein (MBP) fusion proteins and purified using amylose-resin (New England Biolabs, Ipswich, MA, USA). MBP-MKL1 or MBP-MKL2 fusion proteins were conjugated with Freund's adjuvant and subcutaneously injected into guinea pigs five times at 2-week intervals. After sera were incubated with MBP to eliminate immunoglobulins against MBP, specific antibodies were further affinity-purified using MBP-MKL1 or MBP-MKL2 fusion proteins coupled to CNBr-activated Sepharose 4B (GE Healthcare).
Immunostaining. Immunostaining methods of NIH3T3 cells and cortical cultures were performed as previously described 27,32,33 . Purification of synaptic proteins. Subcellular fractionation of 6-week-old male rat brain was performed as previously described 34 . Morphological analysis. To observe dendritic spine morphology of cortical neurons, an average of 40 neurons were selected and images containing two full-length dendrites collected. The number of spines was quantified and classified into three groups: mushroom/stubby (width of spine head is longer than spine length), thin/filopodia (spine length is longer than width of spine head), and irregular (branched spine or protrusion from spine head or much larger spine size). Simultaneously, spine density was counted. Images were obtained by microscopy (LSM700; Carl-Zeiss, Oberkochen, Germany). To observe dendritic spine morphology of hippocampal neurons, fluorescent images were captured with a Leica SP8 SMD confocal microscope (Wetzlar, Germany). For spine density and morphological analyses, only transfected neurons that were distinct from other transfected cells were evaluated with SpineMagick Software 35 . Spine clustering was performed using custom scripts 36 .