Deficiency of Fhl2 leads to delayed neuronal cell migration and premature astrocyte differentiation

The adaptor protein four and a half LIM domains protein 2 (Fhl2) is capable of mediating protein–protein interactions. It is expressed in several organs including heart, skeletal muscle, bone, placenta, liver and lung, as well as in cancer cells, including squamous carcinoma, glioblastoma, chronic myelogenous leukemia, prostate, and breast cancers (Johannessen et al., 2006). As a scaffolding protein, Fhl2 is able to fulfil different functions in the cell, as it (i) is involved in cytoskeletal organization, (ii) modulates several signaling cascades and, (iii) works as a transcriptional co-activator/co-repressor in the nucleus (Johannessen et al., 2006).

By direct binding to β- and α-integrin subunits, Fhl2 is recruited to focal adhesion sites regulating their structure and function (Park et al., 2008; Samson et al., 2004). Loss of Fhl2 in mesenchymal stem cells causes altered bundling of focal adhesion structures and impaired assembly of secreted extracellular matrix (ECM) proteins, including fibronectin and laminin (Park et al., 2008). Furthermore, Fhl2-deficient mesenchymal stem cells expose an altered cytoskeletal organization, impaired actin-myosin contraction and delayed migration on different matrix proteins (Park et al., 2008; Wixler et al., 2007). These defects in Fhl2-deficient mice lead to hindrance of acute skin wound healing (Wixler et al., 2007). Besides, Fhl2 is able to interact with a plethora of transcription factors and transcription co-factors including androgen receptor (AR), CREB, WT1, AP1, IRF3, JUN, FOS, β-catenin and many others (Du et al., 2002; Fimia et al., 1999; Labalette et al., 2004; Martin et al., 2002, 2007; Morlon and Sassone-Corsi, 2003; Muller et al., 2000; Nordhoff et al., 2012). After being transported into the nucleus, Fhl2 acts as either an activator or an inhibitor of transcription. Fhl2-deficient mice develop osteoporosis (Bai et al., 2005; Govoni et al., 2006; Günther et al., 2005), show delayed skin wound healing (Wixler et al., 2007), and increased chronic inflammatory psoriasis and arthritis during stimulation of TNF stimulation (Wixler et al., 2015; Leite Dantas et al., 2017) due to dysregulated function of several signaling cascades. In bone tissue, Fhl2 contributes to maintain the bone resorption and/or formation equilibrium by cytoskeletal organization of osteoblasts and/or osteoclasts. Osteoclastogenesis and bone resorption are accelerated in Fhl2^–/– mice (Bai et al., 2005; Günther et al., 2005).

So far, only very sparse information about the functional impact of Fhl2 on brain activity in general, and on neural cell–cell and cell–matrix interaction in particular is available. Fhl2 has been speculated to be involved in some pathological alterations in brain, including epilepsies and neurodegenerative disease (Tanahashi and Tabira, 2000). Fhl2 interacts physically with presenilin-2 (PS-2, officially known as PSEN2) but not with mutant PS-2 comprising the Met to Val point mutation M238V, which is associated with familial Alzheimer disease (Tanahashi and Tabira, 2000). PS-2 is also involved in activation of Notch signaling. Studies of adult neurogenesis emphasize that the equilibrium between neuronal and glial cell differentiations is tightly regulated by several signaling cascades, including those of Notch and Wnt (Andersson et al., 2011; Lie et al., 2005). In this study, we investigated the effect of Fhl2 deficiency on neural progenitor migration and the equilibrium between neuronal and glial cell differentiation in Fhl2^–/– mice.

We show here that Fhl2 deficiency lead to disturbed arrangement and delayed migration of neuroblasts, as well as to the development of neural stem cells with insufficient ECM. We also show that when Fhl2 is lacking, early postnatal neural stem cells (NSCs) easily lose their characteristic features and undergo a premature astroglial differentiation when cultivated in vitro. Furthermore, in brain of Fhl2-deficient mice (hereafter referred to as Fhl2-deficient brains) we observed disturbance of the equilibrium between neuronal and glial cell differentiation, resulting in a gliosis-like accumulation of glial fibrillary acidic protein (GFAP)-positive (GFAP+) astrocytes that increased substantially with age.

To define the anatomical regions of the brain that express Fhl2, we took advantage of the Fhl2^−/− mouse that expresses the LacZ reporter gene instead of exon 2 of the native Fhl2 gene. X-gal staining of brains from 4-week-old Fhl2^−/− mice (hereafter referred to as 4-week-old Fhl2^−/− mouse brains; Fig. 1A) showed that β-galactosidase expression is widely distributed in a whole brain. To gain more insight into Fhl2 expression during brain development, we analyzed β-galactosidase staining at embryonic day 14.5 (E14.5) and identified it in most of the cells (Fig. 1B) throughout all serial sections, suggesting that Fhl2 is already expressed in most immature cells of whole brain areas. Also, at postnatal day 1, Fhl2 was expressed in neural stem cells, precursor cells of the hippocampal subgranular zone (SGZ), and cells of the subventricular zone (SVZ) (Fig. 1C,D). In the hippocampus, Fhl2/β-gal staining was detected in radial glial-like GFAP+ cells of the subgranular zone (Fig. 1C,b,c, Fig. S1B, arrows), whereas in the SVZ it was expressed by B1 astrocyte-like cells of the subventricular layer (Fig. 1D,b,c, arrows). At higher magnification, most Fhl2-expressing cells in the SGZ and SVZ were identified as neuronal migration protein doublecortin (DCX)-positive (DCX+) neuronal precursor cells (Fig. 1C,d; D,d, arrows). Among mature and differentiated cells, RBFOX3-positive (hereafter referred to as NeuN+) mature neurons (Fig. 1C,f, arrows) and PECAM-positive (PECAM+) endothelial cells (Fig. 1C,e; D,e, arrows) stained strongly for Fhl2/β-gal in the SGZ and SVZ. To the contrary, GFAP+ cells were seldom positively stained, but few Fhl2/β-gal positive cells could be identified in glial cell-dominant region of cerebellum (Fig. 1A, Fig. 1D,f, arrows).

Thus, our findings indicate that Fhl2 is widely expressed in cells of the embryonic brain but, in early postnatal brain, its expression is mostly restricted to neuroprogenitors and neuronal cells. Flow cytometry analysis of single cell suspensions from different brain parts of 4-week-old Fhl2^−/− mice for expression of Fhl2/β-gal supported this suggestion (Fig. S1A). Noticeably, the principal distribution of Fhl2 protein in the stem/precursor cells of the SGZ and SVZ of young (3-week-old) or adult (6-month-old) mice (Fig. S1B,C) does not differ from that of newborn animals (Fig. 1C,D). In the hippocampus, the Fhl2/β-gal staining was detected in radial glial-like GFAP+ cells and DCX+ neuronal precursors of the subgranular zone in 1-day-old (Fig. 1C,b–d, arrows), 3-week-old and 6-month-old mice (Fig. S1B, arrows). Also, mature neurons were positively stained for Fhl2/β-gal in hippocampi of 6-month-old mice, however, GFAP+ cells were rarely Fhl2/β-gal-positive even in the cerebral cortex of 3-week-old (Fig. S1C) as well as neonatal mice.

To investigate whether Fhl2 expression in the SVZ as a stem cell niche is also maintained in young mice, whole-mount immunostainings of lateral walls of the SVZ from 3-week-old wild-type (WT) animals were performed as described by Mirzadeh and colleagues (2008). In serial confocal z-stack images of the superficial ventricular wall, most of the (α-tubulin/β-catenin+) ependymal cells stained positively for Fhl2 (Fig. S1D, a-c,f). In the same layer (Fig. S1D,d,e), B1 astrocyte-like cells that represent the core of a pinwheel-like architecture of the superficial ventricular zone (shown as a schematic drawing in Fig. S1D, left panel) were also strongly Fhl2-positive (Fig. S1D, d,e arrows). This indicates that B1 astrocyte-like cells in the SVZ are Fhl2-positive. In the deeper layer of the SVZ, PECAM/Fhl2-positive microvessels (Fig. S1D, h,i) communicate with end-feet (Fig. 1D, g, arrows) of B1 cells in the upper layer (Fig. S1D, d) detected in z-stack confocal images.

Taken together, in the postnatal brain, Fhl2 is mainly expressed in NSCs, and in immature and mature neuronal lineage cells of the hippocampus and SVZ.

To study whether Fhl2 expression influences brain development, brain sections of WT and Fhl2^−/− mice were analyzed immunohistochemically at embryonic days 14.5 and 18.5 (E14.5 and E18.5, respectively) and postnatal day (P1) of mouse development. The whole forebrain of E14.5 old Fhl2^−/− embryos showed much less DCX protein signal than the appropriate forebrains of WT mice (Fig. 2A). By contrast, glial precursor marker A2B5+ cells were also easily detected in Fhl2^−/− E18.5 forebrain and their staining intensity there was much higher than in WT forebrains. In Fhl2^−/− mice astroglial preference of differentiation was also identified in 1-day-old cerebral cortex and hippocampus (Fig. 2C,D), and expression of Gfap mRNA was increased in E18.5 cerebral cortex (Fig. 2E). Furthermore, sagittal sections of 3-week-old brains revealed a more abundant presence of GFAP+ cells in Fhl2^−/− than in WT mice, which were widely detected throughout the whole brain including cerebral cortex, hippocampus, lateral ventricle and olfactory bulb areas (Fig. 2F). Also in whole-mount SVZ tissue, an abundant accumulation of GFAP+ cells was seen throughout whole lateral walls of Fhl2^−/− SVZ in 3-week-old brains (Fig. 2G), which was consistent with Gfap mRNA expression shown in Fig. 2H. The amount of Gfap mRNA in the Fhl2^−/− cerebral cortex from 4-, 6- and 9-month old mice was continuously higher than in WT brains and the difference increased with the age of the mice (Fig. 2H). By contrast, expression of DCX, a marker of neuronal precursor cells, rather decreased with age in the cerebral cortex (Fig. 2H). In good agreement with qRT-PCR data, the age-dependent increase of GFAP within cerebral cortex and hippocampus tissues of Fhl2^−/− mice was much higher than in WT mice (Fig. 2I).

Analysis of hippocampus, corpus callosum and cerebral cortex of 6-months-old mice revealed a dense astroglial distribution in Fhl2^−/− brains, judged by extensively high GFAP staining (Fig. S2B,C). Collectively, these findings suggest that a gliosis-like GFAP+ cell deposition accumulates in Fhl2^−/− mouse brains with age. Of note, not only a much more intensive GFAP staining was observed in Fhl2^−/− compared with WT cerebral cortex but also a higher number of Olig2- and ALDH1A1-positive cells was determined there (Fig. S2A), demonstrating an abnormal accumulation of glial cells in Fhl2^−/− cerebral cortex. Taken together, our in vivo data indicate that, in the absence of Fhl2, adult NSCs are persistently pushed towards astrocyte differentiation. This resulted in a gliosis-like accumulation of astrocytes in the cerebral cortex, hippocampus and SVZ, and the aberrant development was exaggerated in older Fhl2^−/− mice.

The fact that Fhl2 is expressed in NSCs and mature neurons suggests that it plays a role in the maintenance of stem cells and/or in their differentiation into neurons. To explore this possibility, we isolated neurospheres (Fig. 3B) from postnatal 1-day-old brains of Fhl2^+/+ and Fhl2^−/− mice, and cultivated them in vitro (Fig. 3A). Surprisingly, Fhl2-deficient neuroprogenitors spontaneously underwent premature astroglial differentiation in neurosphere culture. For neurospheres cultured for 7 days, more than half of the Fhl2-deficient cells showed an intense intracellular GFAP signal (Fig. 3C, 59%) in flow cytometry. Comparison of the gene expression profiles of 7-day-suspension cultures (Fig. 3B) showed that Fhl2-deficient neurospheres expressed significantly more Gfap mRNA. Gene expression of the stem cell marker CD133 (officially known as PROM1), the neuronal precursor marker Dcx and the Notch signaling-associated genes Jagged1 (Jag1) and Notch1 (Notch1) in Fhl2-deficient neurosperes was downregulated compared with that in WT neurospheres. Furthermore, as early as day 1 after plating on Matrigel-coated culture dishes, the vast majority of Fhl2-deficient cells sprouting out of neurospheres were intensively GFAP-positive when analyzed by immunofluorescence, whereas in WT cells, only a weak GFAP signal was detectable (Fig. 3C). In the absence of Fhl2, a misbalanced differentiation of NSCs also occurred in older brains. The levels of Gfap transcripts in neurospheres derived from 3-week-old Fhl2^−/− hippocampus and the SVZ were always higher in than in WT mice (Fig. 3D). The premature astroglial shift in Fhl2^−/− cells was further confirmed by analysis of Gfap mRNA transcripts and GFAP protein levels. They were always increased in Fhl2^−/− cells as compared with WT cells, both in neurospheres and adherent undifferentiated (Undiff.) cells (Fig. 3E,F). Furthermore, levels of Gfap mRNA increased during cultivation of neurospheres in vitro, whereas in WT cells it did not change significantly, and was virtually similar in neurosphere suspension and adherent cultures (Fig. 3E).

Of note, most cells in the core of WT neurospheres stained strongly for Fhl2 and only weak for GFAP, whereas those of Fhl2-positive neurospheres did not colocalize with GFAP (Fig. S3A). This implies that undifferentiated NSCs strongly express Fhl2 but, once differentiated toward astroglial lineage, Fhl2 expression is rapidly lost. Prolonged cultivation of already attached neuroprogenitors in differentiation medium supplemented with 1 µM retinoic acid but without FGF and PDGF led a high number of Fhl2^−/− cells mature to astrocyte-like cells (Fig. 3G,a) but not to a neuronal lineage as expected (Fig. 3G,c). Immunostaining of these cells for oligodendrocyte (Olig2) or another astrocyte marker (ALDH1A1) revealed a high number of Fhl2^−/− cells that stained positive for both markers (Fig. 3G,b,d). Analysis of the differentiation profile of neuroprogenitors cultured for 5 days in differentiation medium clearly showed that Fhl2^−/− cells are prone to glial rather than neuronal differentiation (Fig. 3H). These data confirmed that GFAP+ Fhl2^−/− cells, indeed, undergo glial differentiation rather than being maintained as GFAP+ neuroprogenitors in an Fhl2^−/− neurosphere-derived pool. Serial counts of neurospheres in suspension cultures cultivated for over seven passages (Fig. S3B), showed that Fhl2^−/− neurosphere formation was less efficient than WT neurosphere culture. Not only was the number of Fhl2^−/− neurospheres over the whole cultivation period decreased, but they also were smaller (Fig. S3B, inset). These data indicate that the self-renewal capacity of many Fhl2^−/− neurospheres is rapidly lost in suspension culture, resulting in premature differentiation.

As Fhl2-deficiency showed impaired assembly of secreted ECM proteins, such as laminins and fibronectin, and severe disturbance of migration of mesenchymal stem cells on ECM proteins (Park et al., 2008; Wixler et al., 2007), we attempted to verify whether Fhl2 has also a role in neuronal migration. Interestingly, in brain sections of E14.5 mice, Fhl2-deficient neuroblasts showed perturbed migration towards the cortical plate. Whereas WT forebrain cells at this embryonal stage were located at the cortical plate border (Fig. 4A, dotted lines), Fhl2-deficient cells showed a slower rate of migration. These cells did not properly reach at the cortical plate and were scattered around the cortical plate (Fig. 4A, arrows). Considering that Fhl2-deficient forebrains showed insufficient expression of laminin a2 (LAMA2) and laminin a5 (LAMA5), and impaired assembly of the ECM (Fig. 4B,C), these factors might be in part responsible for disturbed neuroblast migration. Indeed, insufficient expression of laminin was consistently observed in hippocampi of P1 (Fig. S4A) and cerebral cortex 6-month-old (Fig. S4D) Fhl2^−/− mice, as well as in Fhl2-deficient in vitro neurosphere cultures (Fig. 4E). Insufficient expression of laminins in the forebrain of Fhl2-deficient mice was observed at the same time as decreased PS-2 signals in E14.5 old forebrain (Fig. 4D) and in in vitro neurosphere cultures from brain of P1 mice (Fig. 4E).

Considering that Fhl2 binds to PS-2 and that PS-2-deficient mice show perturbed neuronal migration (Tanahashi and Tabira, 2000; Hartmann et al., 1999; Louvi et al., 2004), it is conceivable that insufficient PS-2 expression at the Fhl2^−/− background leads to migration abnormalities. Indeed, in vitro migration of Fhl2-deficient NSCs was severely delayed (Fig. 4F, Movies 1 and 2). Remarkably, the migration defect of these cells was simply rescued by retroviral introduction of a human FHL2 cDNA (Fig. 4F, lower panel).

In addition to migration deficiency of embryonal neuroblasts, abnormal migration and/or abnormal neuronal cell orientation and alignment was also revealed in postnatal cerebral cortex and the hippocampus. In the cerebral cortex of 3-week-old Fhl2-deficient mice, DCX+ neuronal precursors distributed without a clear localization zone, whereas in the cerebral cortex of WT these cells was well demarcated (Fig. 4G). This abnormal cell distribution was also found in the hippocampus of 3-week-old Fhl2-deficient mice (Fig. 4H). DCX+ cells were less in number and randomly orientated in the Fhl2-deficient dentate gyrus of the hippocampus compared to well-aligned DCX+ cells in the subgranular zone of hippocampus in WT mice. Also in the cerebellum (Fig. 4I) as well as the lateral ventricle (Fig. S4B) of 3-week-old Fhl2-deficient mice, the DCX+ organization was missing. Taken together, Fhl2 plays an important role in migration of neuronal cell and their alignment, and these processes are perturbed in Fhl2-deficient mice.

To verify whether deficiency of Fhl2 is responsible for preferential astroglial differentiation in Fhl2^−/− neurospheres, protein expression of Fhl2 in Fhl2-deficient progenitors was restored by retroviral gene transfer of Myc-tagged human FHL2 (Fig. 5). Of note, FHL2 expression also restored the altered gene expression pattern in Fhl2-deficient progenitors, when analyzed 12 days after FHL2 transduction (Fig. 5B). Not only were the gene expression levels of astrocyte, neuronal precursor and mature neuron gene markers (Gfap, Dcx and NeuN, respectively) reverted, the reduced expression of genes belonging to the Notch signaling pathway was also dramatically enhanced in Fhl2^−/− cells after overexpression of recombinant FHL2. Because Notch is known to promote proliferative signaling during neurogenesis, but is inhibited itself during neuronal differentiation, we hypothesized that altered Notch signaling is responsible for the premature astroglial differentiation of Fhl2-deficient neuroprogenitors.

As cleavage of the Notch1 intracellular domain (NICD) is an essential event for canonical Notch signaling, we first analyzed whether the amount of NICD protein changed in Fhl2-deficient neuroprogenitor cells. Fhl2-deficient cells isolated from P1 mouse brains and cerebral cortex tissue extracts from 4-week-old Fhl2^−/− mice showed much less NICD protein than equivalent WT samples (Fig. 5C). Of note, the difference between WT and Fhl2-deficient cells could be negated by infection of Fhl2^−/− cells with FHL2-containing retroviruses (Fig. 5C). Thus, our data suggest that the presence of Fhl2 in NSCs is not only crucial for their maintenance but also for unfolding of their differentiation programs.

We further investigated whether the amount of NICD protein was changed in the brain tissue of Fhl2^−/− mice by analyzing cerebral cortex lysates using western blotting with an anti-NICD antibody, detecting only active NICD cleaved at valine at position 1744. Indeed, 4-week-old Fhl2^−/− cerebral cortex tissue extracts showed much less NICD protein than WT samples. Interestingly, they also showed diminished amounts of active α-secretase (ADAM17), which is needed for cleavage of the Notch receptor (Andersson et al., 2011; Brou et al., 2000) (Fig. 6A).

As Fhl2 coordinates transcription of many genes as a cofactor (Johannessen et al., 2006), we next attempted to investigate whether it also regulates the transcriptional activity of the transcription factor and Notch-target gene Hes1. HEK293 cells were transiently transfected with the luciferase reporter gene construct pGL3/Hes1-luc, containing the Hes1 promoter, either alone or together with NICD and FHL2 transgenes. Interestingly, when FHL2 was coexpressed with NICD, the reporter gene activity was repressed and the repression efficiency was clearly FHL2-dependent (Fig. 6B). Further, an immunoprecipitation assay from 4-week-old WT cerebral cortex tissue extract showed that Fhl2 indeed physically interacts with NICD (Fig. 6C). To underscore the significance of Fhl2 in Hes1 gene regulation, the FHL2 transgene was co-transfected with pGL3/Hes1-luc plasmid into WT and Fhl2^−/− neuroprogenitor cells, and activation of the reporter gene was studied. Western blot analysis of transfected cells showed that considerable amounts of recombinant FHL2 were achieved in both WT and Fhl2^−/− neuroprogenitors (Fig. 6D). Luciferase assays showed that the NICD-mediated activation of the Hes1 promoter-driven luciferase reporter was strongly repressed by Fhl2 in a dose-dependent manner, both in both WT and Fhl2^−/− neuroprogenitors (Fig. 6E). Furthermore, addition of the Fhl2 transgene reverted the diminished neuronal differentiation of both WT and Fhl2^−/− neuroprogenitors that have been induced in these cells by overexpression of NICD (Fig. S4C). Considering that Fhl2 directly binds NICD (Fig. 6C), these reporter gene data indicate that Fhl2 regulates the NICD-mediated transcription activity of Hes1 as a co-repressor.

Finally, to analyze whether Fhl2, as a binding partner of NICD, also regulates the NICD-driven transcription of Gfap, we expressed the luciferase reporter gene construct pGL3/GFAP-luc containing the Gfap gene promoter in WT and Fhl2^−/− neuroprogenitors, either alone or together with NICD and FHL2 transgenes. Activity of the Gfap promoter was, indeed, dramatically repressed in a dose-dependent manner by Fhl in both WT and Fhl2-knockout cells (Fig. 6F).

Together, our data suggest that Fhl2 modulates, directly or indirectly, the transcription of Hes1 and GFAP genes, which are responsible for the differentiation of NSCs into neuronal/astroglial lineages. Previous studies reported that overexpression of Hes1 induces preferential glial differentiation of glial restricted precursor cells (Wu et al., 2003), whereas suppression of Hes1 expression initiates differentiation of neural stem cells into neurons (Kabos et al., 2002). In our study, the fact that Fhl2 negatively regulates the NICD-mediated activity of both Hes1 and Gfap promoters indicates that the absence of Fhl2 protein in Fhl2^−/− neuroprogenitors may lead to Notch-mediated activation of Hes1 and GFAP transcription, resulting in tenacious development of these cells towards astrocytes.

To our knowledge, our study demonstrates for the first time that knockout of Fhl2 leads to phenotypic changes in the brain. Our results show that Fhl2 is expressed almost in all brain areas during embryonal development, but that expression is restricted to occur postnatal in particular cell populations. Remarkable is that, during embryonic development and early postnatal periods, immature neuroblasts and postnatal neural stem cells/progenitors in hippocampus and subventricular zones strongly express Fhl2. Another particular feature of Fhl2 expression is its prevalence in neuronal progenitors and in cells of the neuronal lineage. The majority of immature and mature neurons express Fhl2, including those that are DCX+ and NeuN+. However, among the GFAP+ glial cells, Fhl2-positive cells were rare. These Fhl2 expression pattern suggests that Fhl2 is involved in maintenance and decision regarding lineage differentiation fate of NSCs. Fhl2-deficient NSCs in vitro showed a low self-renewal activity of neurospheres and a premature differentiation of GFAP+ glial lineage concomitantly with a decreased neuronal differentiation ratio. Also, compared with WT brain regions, Fhl2-deficient brain regions revealed a decreased DCX+ cell distribution in vivo, with disturbed orientation in hippocampus and cerebral cortex, both in embryonal and postnatal periods. By contrast, astrocyte lineage cells, including A2B5+ or GFAP+ cells, were abundantly present in the brain both during embryonal and postnatal periods of development, and their presence continued to increase with age. These data indicate that NSCs/progenitor cells require Fhl2 to maintain their stem cell pool and that Fhl2 plays a role in keeping an equilibrium axis on differentiation regulation.

Here, we have investigated (i) a delayed neuroblast migration, (ii) an early exhaustion of stem/progenitor pool and (iii) an abnormal glial cell accumulation throughout the brain in Fhl2-deficient mice, all of which point towards disturbed adult neurogenesis. Adult neurogenesis in mammals mostly occurs in restricted stem cell niches, i.e. in the dentate gyrus of the hippocampus (the SGZ) and in the SVZ (Lie et al., 2004; Ming and Song, 2011; Morrison and Spradling, 2008). Our data confirmed that postnatal stem cells in the SGZ and SVZ strongly express Fhl2. In the SGZ and SVZ of 3-week-old Fhl2-deficient mice, DCX+ cells rapidly diminished and showed abrupt cell alignment and orientation. Instead, excessive glial cell accumulation was observed surrounding the SGZ and SVZ. Regarding the decreased speed of migration of neuroprogenitors shown in our in vitro and in vivo experiments, more than one factor seems to be involved. On the one hand, insufficient ECM expression, including laminin a2 and a5, might be responsible in part for the in vitro and in vivo delayed migration of Fhl2-deficient stem/progenitors. Previous studies have also shown the role of laminins in supporting migration of NSCs and neuronal progenitor cells (Perris, 1997; Franco and Müller, 2011; Barros et al., 2011). On the other hand, the decreased expression of presnilin-2 (PS-2) might delay cell migration, which was consistently observed in the brain of Fhl2^−/− mice and in Fhl2-deficient NSCs in vitro. Previous studies in which mutant embryos homozygous for presenilin-1 (PS-1, officially known as PSEN1) had been used, found that cortical dysplasia is strongly linked to defective migration of neurons (Hartmann et al., 1999; Louvi et al., 2004). Loss of PS-1 function perturbs both radial and tangential migration in the cerebral cortex, and several tangential migratory pathways in the brainstem. Of note, restoration of Fhl2 expression in Fhl2-deficient neuroprogenitors also restored both the expression of PS-2 and their migration capacity. The effect of PS-2-deficiency on neuroblast migration has not been reported so far. Considering that Fhl2 binds to PS-2 (Tanahashi and Tabira, 2000), it would be interesting to further investigate how Fhl2 may be involved in the regulation of PS-1/2 expression to verify the mechanism not only for the migratory disturbance but also for Fhl2-deficient stem cell fate decisions mentioned above.

In vertebrates, Notch signaling via Hes1 – in which Hes1 functions as a transcriptional repressor of proneuronal genes – has been described to maintain the NSC pool and to inhibit neuronal differentiation. Previous studies in mice have demonstrated that, in the developing nervous system, conditional ablation of Notch1 (Lutolf et al., 2002) or lack of Hes1 activity (Ishibashi et al., 1995; Kabos et al., 2002; Nakamura et al., 2000) leads to the premature onset of neurogenesis. Our results demonstrated that both Notch1 receptor and its ligand Jagged1 were significantly reduced in Fhl2-deficient neurospheres. The proteolytically cleaved product of the Notch1 receptor, the NICD, was also downregulated both in Fhl2-deficient cells cultured in vitro and in brain tissue. This occurred together with reduced PS-2 expression and diminished ADAM17 activity. The fact that rescue of Fhl2-deficency in mouse NSCs by introduction of a human FHL2 transgene directly restored the transcription of Notch1, Jagged1 (Jag1) and Gfap as well as several other neuronal differentiation markers supports the suggestion that the Notch-mediated stage-specific astroglial differentiation is modulated by Fhl2. We initially assumed that reduced Notch signaling under Fhl2-deficiency leads to neuronal differentiation of NSCs. Surprisingly, further experiments showed that Fhl2-deficiency results in disturbed NSC maintenance, giving rise to premature gliogenesis both in vitro and in vivo. Several reports have demonstrated that downregulation of Notch signaling induces premature neuronal differentiation (Imayoshi et al., 2010; Hatakeyama et al., 2004; Breunig et al., 2007). However, our experiments showed that, on an Fhl2^−/− background, downregulation of Notch signaling induced premature astroglial differentiation. These seemingly contradictory results highlight the importance of the interaction between Fhl2 and NICD in order to modulate differentiation. A previous study reported that overexpression of Hes1 induces preferential differentiation of glial restricted precursor cells (Wu et al., 2003). With regard to our immunoprecipitation and reporter gene assay results, it is very likely that Fhl2 negatively regulates Hes1 and Gfap promotor activation in a dose-dependent manner, thereby acting as a co-repressor of the NICD. Further, the fact that NICD-guided activity of the GFAP promoter was also diminished by the FHL2 transgene may explain, at least partly, the premature astroglial differentiation of Fhl2^−/− neuroprogenitors. Recently, Notch signaling has been proposed as a stage-specific regulatory cue for the differentiation of neuronal and glial cells (Morrison et al., 2000; Gaiano and Fishell, 2002; Grandbarbe et al., 2003; Ramasamy and Lenka, 2010). It is well known that neurogenesis precedes glial differentiation in the developing CNS (Qian et al., 2000) and that, during the neurogenic period, the astrocyte differentiation program is actively suppressed (Ge et al., 2002). Although the Notch signaling pathway per se is intact in astrocytes, it is unable to stimulate transcription of Gfap, suggesting that Notch signaling is suppressed during neurogenesis (Ge et al., 2002; Tanigaki et al., 2001). Hence, it is plausible to speculate that Fhl2 is required for Notch signaling to inhibit gliogenesis, and to maintain the pool of NSCs during brain development and adult neurogenesis. This would imply that Fhl2 is involved in the NSC fate decision and explain why Fhl2 deficiency allows initiation of the astrocyte differentiation program, which is actively suppressed during the neurogenic period in the presence of Fhl2 (Fig. 7). Although we performed some Notch signaling-related assays in this study, a complete verification of mechanisms regulating Fhl2-deficient phenotype changes in brain is far beyond the aim and scope of the present study. In addition, the mechanism by which Fhl2 regulates transcription of Notch1 and Jag1 has to be further validated.

Although our results showed a regulatory effect of Fhl2 on both Hes1 and Gfap transcriptional activity, we cannot exclude the possibility that additional factors are also involved in enhanced astrogliogenesis of Fhl2^−/− progenitors. Recent reports have shown that the promoter of human JAG1 contains binding sites for the bipartite transcription factor β-catenin-LEF/TCF (Katoh and Katoh, 2006), and that JAG1 is a β-catenin target gene in adult epidermis (Estrach et al., 2006) and colorectal cancer cells (Rodilla et al., 2009). Moreover, it has recently been reported that β-catenin and NICD bind to each other (Kwon et al., 2011; Shimizu et al., 2008), forming a complex on the promoter of the Hes1 gene, allowing its expression (Shimizu et al., 2008). These data suggest a link between Wnt and Notch signaling. Interestingly, Fhl2 also binds β-catenin and, being transported into the nucleus, acts as a co-activator in cancer cells (Wei et al., 2003) or as a co-repressor in myogenic stem cells (Martin et al., 2002). The fact that Fhl2 can bind β-catenin (Martin et al., 2002; Wei et al., 2003) as well as NICD (this study), suggests that Fhl2 plays a modulatory role in Jag1 and/or Hes1 transcriptional activities in concert with β-catenin and the NICD nuclear protein complex. Whether Fhl2 is part of a link between Wnt and Notch signaling needs to be further investigated.

Taken together, Fhl2-deficiency interrupts the maintenance of adult NSCs and disturbs their balanced differentiation in stem cell niches, resulting in preferential glial differentiation and early exhaustion of the NSC pool (Fig. 7).

Since Fhl2 mRNA and protein are already abundantly expressed in the heart during early embryonic development (Chu et al., 2000b), the role Fhl2 in heart development and function has been investigated intensively. However, Fhl2-knockout mice have normal heart function, and the heart and blood vesicles develop normally (Chu et al., 2000a; Kong et al., 2001). Based on numerous reports, Fhl2 deficiency itself does not seem to directly provoke pathophysiological conditions or disease during development and postnatal growth (Chu et al., 2000b). Rather, Fhl2 plays an important role as an essential regulator of stem cell behavior during stress and or inflammation, and in wound-repair processes. In hematopoietic stem cells under regenerative stress (Hou et al., 2014), mesenchymal stem cells during skin wound healing (Wixler et al., 2007; Park et al., 2008), during epithelial–mesenchymal transition (Zhang et al., 2010; Cai et al., 2018), chronic inflammatory arthritis (Wixler et al., 2015) or liver regeneration (Dahan et al., 2013), Fhl2-deficient stem cells show diverse phenotypes. These include reduced self-renewal capacity and migration deficiency, and Fhl2-deficient tissue revealed delayed wound healing, or abnormal inflammatory signal processing and matrix organization during repair (this study; Wixler et al., 2007; Park et al., 2008; Leite Dantas et al., 2004).

The complex functional activity of Fhl2 seems to be derived from its structure. As a LIM domain-containing adaptor protein, it binds to numerous other proteins, including membrane receptors, cytosolic structural proteins, kinases and transcription factors. So far, ∼50 different proteins have been described that specifically interact with Fhl2 (Johannessen et al., 2006). Furthermore, Fhl2 is able to move between cellular membrane and nucleus, transferring signals and regulating gene transcription either as repressor or activator. The dualistic nature of Fhl2 most probably depends on the composition of a protein complex comprising Fhl2, which depends on cell-specific expression and cell-specific subcellular distribution of Fhl2 (Hou et al., 2014). Fhl2 may act as an oncogene or as a tumor suppressor. Its gene expression is often dysregulated in cancer cells, as it is either downregulated or overexpressed depending on the tumor type, i.e. in rhabdomyosarcoma, prostate cancer, ovarian cancer, melanoma, lung cancer, breast cancer and liver cancer (Johannessen et al., 2006; Hou et al., 2014). The complexity of Fhl2 functions makes it often difficult to collect clinical evidence of Fhl2-associated pathophysiology, although diseases linked to Fhl2 have been reported, including familial isolated dilated cardiomyopathy, hemophagocytic lymphohistiocytosis or rheumatoid arthritis (Friedrich et al., 2014; Cetica et al., 2016; Wixler et al., 2015). However, a clear clinical proof that perturbed Fhl2 expression and/or activity contributes to human disease is still lacking (Johannessen et al., 2006). Nevertheless, several studies, including this work, underscore the relevance to further investigate the functional role of Fhl2 in the brain and other organs. Particularly, the altered stem cell characteristics and prematurely exhausted stem cell pool in brain tissue of Fhl2-deficient adult mice (present study), as well as the potential association of Fhl2 with Alzheimer's disease (Tanahashi and Tabira, 2000), early-onset autosomal recessive Parkinson's disease (Xu et al., 2005), disturbed wound repair (Wixler et al., 2007; Park et al., 2008) and chronic inflammation (Leite Dantas et al., 2004), emphasize the pivotal role of Fhl2 in preventing pathogenic processes. Adoption of RNA interference to target Fhl2-regulating genes, in order to reduce or restore levels of Fhl2 in diseased cells, may be an option to improve pathogenic conditions in brain and other tissues.

The Fhl2^−/− mouse model was generated as described previously (Kong et al., 2001). In short, part of exon 2, including the translation initiation codon ATG, of the Fhl2 allele was replaced with a LacZ reporter gene and a neomycin resistance cassette. The LacZ gene product β-galactosidase was used for localization of the Fhl2 protein in brain tissues. The study was performed under institutional animal care approval (Kinder- und Jugendklinik TS-9/12, Friedrich-Alexander University, Erlangen, Germany). For X-gal staining, brains of 4-week-old Fhl2^−/− mice were cut mid-sagittally and fixed with 4% paraformaldehyde (PFA) solution overnight at 4°C. Afterwards, the probes were washed three times for 15 min with washing buffer (0.1 M phosphate buffer pH 7.3, supplemented with 2 mM MgCl₂) and further incubated with X-gal solution (1 mg/ml X-gal, Boehringer Mannheim, 5 mM potassium ferricyanide, 5 mM potassium ferricyanide in washing buffer) overnight at 4°C with gentle shaking. On the following day, the probes were washed again three times with PBS, for 15 min each time, and images were taken with an AxioCam digital camera and AxioVision software (Carl Zeiss MicroImaging Inc.). For fluorescein di-β-D-galactopyranoside (FDG, Life Technologies) staining, brains of 4-week-old Fhl2^−/− and WT mice were harvested and digested by dispase (1 U/ml, Sigma) for 20 min, resuspended in FACS-PBS (PBS containing 2% FCS and 0.02% NaN₃) and centrifuged. Then, 100 µl of single cell suspensions, kept at 37°C, were thoroughly but rapidly mixed with 100 µl of pre-warmed 2 mM FDG solution and incubated in a 37°C water bath for exactly 1 min. Then, 1.5 ml of ice-cold FACS-PBS was added, and the cell suspension was analyzed for FDG+ cells by flow cytometry. WT cerebral cortex was used as a negative control.

For immunofluorescence, 6-μm tissue slides were cut from formalin-fixed, frozen or paraffin-embedded specimens, or from cells grown on Matrigel-coated microchamber dishes (Ibidi). Cells were fixed with 4% PFA in PBS at RT for 10 min. After fixation, they were permeabilized by incubation with 0.2% Triton X-100 in PBS for 10 min and washed with PBS. Primary antibodies used were: rabbit anti-Fhl2 antibody (1:200, Wixler et al., 2007), mouse anti-Fhl2 clone F4B2 (1:10, Wixler et al., 2007), rabbit anti-GFAP (1:300, Abcam), rabbit anti-ALDH1A1 (1:300, Abcam), rabbit anti-Olig2 (1:500, Millipore), rabbit anti-MAP2 (1:500, Abcam), rabbit anti-laminin a2 and anti-laminin a5 (1:200, both gifts from Dr Takako Sasaki), rabbit anti-Preselin2 (1:200, Abcam), mouse anti-β-galactosidase (1:200, Sigma), guinea pig anti-DCX (1:1000, Millipore), mouse anti-NeuN (1:300, Millipore), mouse anti-PECAM (1:200, Sigma), mouse anti-A2B5 (1:50, R&D systems) and mouse ßIII tubulin (tubulin beta-3 chain (TUBB3, 1:600, Sigma) diluted in antibody diluent (DAKO) in overnight incubation at 4°C. Probes were washed with PBS two times followed by incubation with fluorochrome-conjugated secondary antibodies. Staining in the absence of primary antibodies confirmed the specificity of the immunolabeling. For differentiation profile analysis, adherent WT and Fhl2^−/− neuroprogenitors on Matrigel-pre-coated dishes passaged twice were trypsinized, plated on Matrigel-pre-coated dishes and cultivated with differentiation medium for 5 days. Cells were immunostained with antibodies against MAP2, Olig2, GFAP or ALDH1A1. Positively stained cells of at least three independent experiments per each lineage-marker were counted and the percentage of each marker-positive cell was calculated in each experiment. For neuronal differentiation in response to transfection with NICD and FHL2-encoding plasmids, adherent WT and Fhl2^−/− neuroprogenitors on Matrigel-pre-coated dishes that had been passaged twice were trypsinized, plated on Matrigel-pre-coated dishes and transfected either alone or together with NICD and FHL2-encoding plasmids (pCDNA-FHL2) 1 µg per well. Cells were cultivated with differentiation medium for 5 days and immunostained for DCX. Fluorescence was monitored using an Axiophot microscope (Carl Zeiss MicroImaging Inc.) or confocal microscopy (Leica Confocal Laser Scanning Microscope, Model LSM 780 NLO). Note that some images, acquired by using the multiple tile scan tool at high resolution by confocal microscopy may show some discontinuity between tiles.

Brains of 3-week-old Fhl2^−/− or WT mice were extracted after cervical dislocation and the lateral ventricle was dissected from the caudal part of the telencephalon and hippocampus before the septum was removed. The following procedures of whole-mount preparation and immunostaining were performed as described (Mirzadeh et al., 2008). Briefly, the dissected lateral wall was fixed in 4% PFA/0.1% Triton-X overnight at 4°C. For immunostaining, specimens were incubated with rabbit anti-Fhl2 antibody (1:200), mouse anti-Fhl2 antibody clone F4B2 (Wixler et al., 2007), mouse anti-acetylated tubulin (1:1000, Sigma), rabbit anti-β-catenin (1:1000, Sigma), anti-PECAM (1:1000, Sigma) primary antibodies diluted in PBS with 0.5% Triton-X and 10% normal goat serum for 24 h at 4°C. On the following day, probes were washed with 0.1% Triton-X100/PBS three times for 20 min at room temperature, and incubated with secondary FITC- and Cy3-conjugated antibodies for 24 h at 4°C. The anti-mouse (1:400, Molecular Probes) and anti-rabbit (1:400, Jackson Immuno Research) secondary antibodies were diluted in 0.5% Triton-X/PBS plus 10% normal goat serum. After secondary antibodies were washed off, nuclear counter-staining was performed by incubating specimens with DAPI diluted in PBS for 30 min at room temperature. After washing with PBS, ventricular walls were released from the underlying parenchyma, dissected as 200–300 µm thick slices covered with mounting medium and a coverslip, and confocal images were taken (Leica Confocal Laser Scanning Microscope, Model LSM 780 NLO).

Isolation of neuroprogenitors from brain of newborn mice and neurosphere formation was performed as previously described (Brewer and Torricelli, 2007). Isolated cells from neonatal 1-day-old WT and Fhl2^−/− mouse brains were plated on Ultra-low attachment culture dishes (Corning Costar). Neurospheres were cultivated for 7 days as suspension under non-differentiation conditions (Neurobasal medium without retinoic acid (Invitrogen) but containing B27, 0.5 mM glutamine, 10 μg/ml gentamycin and supplemented with 10 ng/ml of mouse recombinant FGF, EGF and PDGF, Immunotools). Afterwards, neurospheres were plated into Matrigel (BD science)-coated culture dishes and cultivated as monolayers under non-differentiation condition for additional 7 days. Cells at confluence were trypsinized and plated on Matrigel-pre-coated dishes for cell expansion. To stimulate neuronal differentiation in monolayer cells, growth medium was replaced with differentiation medium [Neurobasal medium with 1 µM retinoic acid (Invitrogen) and B27, 2 ng/ml EGF, 0.5 mM glutamine and 10 μg/ml gentamycin] for 10 days.

For transduction of an exogenous human FHL2 transgene into Fhl2^−/− neurosphere-derived cells, 3-day-old undifferentiated monolayer cultured cells were retrovirally infected by overnight incubation in supernatants from virus-producer Phoenix cells and 10 µg/ml polybrene as described in (Samson et al., 2004). Next day, virus-containing supernatants were replaced by fresh stem cell medium and infected cells were incubated for two more days before selection for Zeocin resistance (750 µg/ml Zeocin, Invitrogen). On reaching confluence, FHL2-transgenic cells were split and further incubated in the presence of Zeocin according to the schedule shown in Fig. 5A.

For gap migration analysis, cell migration studies were performed essentially as previously described (Wixler et al., 2007). In brief, under non-differentiation condition 5×10⁴ undifferentiated neural progenitors expanded as monolayer culture from neurospheres were plated in 2-well silicone-inserts providing a 500 µm cell-free gap (Ibidi) in 48 well-plates pre-coated with Matrigel, except for a cell-free gap area. Previously, retrovirally infected cells with FHL2 transgene as mentioned above were used as rescued cells. Cells were fixed 24 h after insert removal for migration experiments. Bright-field microscopy gap migration images following immunostainings for laminin a5 were taken 24 h after migration start. Gap closure areas in each three different depicted areas out of each three independent experiments (n=9) were analyzed by using ImageJ software. WT and Fhl2^−/− cell migrations were also monitored by video microscopy from 6 h until 24 h after migration start (Movies 1 and 2).

For neurosphere assays, neurospheres were counted in each passage until passage 7. At first, non-passaged neurospheres after 7-day suspension culture of freshly isolated WT and Fhl2^−/− neuroprogenitors were dissociated. A roughly equal number of cells (∼3000/cm²) in each dish were cultivated in suspension culture for 2 days (passage 1) under non-differentiation conditions as mentioned above. At day 3, the neurospheres formed were counted. For subsequent passaging, neurospheres were collected by brief centrifugation (3 min, 300 g) and dissociated. The cells were then counted and roughly the equal number of cells (∼3000/cm²) in each dish was freshly plated and further cultivated (passage 2) under the same culture condition as described above.

Total RNA was extracted from cells of different differentiation stages (see Figure legends for details) with Trizol reagent (Life Technologies). Two independent samples – Fhl2^+/+ and Fhl2^−/− neural stem cells – were always used for each experiment. Reverse transcription of RNA (1 μg) was performed with oligo(dT) or random hexamer primers and M-MuLV reverse transcriptase (Promega) (Table S1). PCR reactions were done in triplicate for each cDNA probe using the QPCR Master Mix (Eurogentec) and the Real-time PCR system (Applied Biosystem, TaqMan 7500) according to the manufacturer's instructions.

Soluble proteins were extracted from cells or mouse brain specimens with lysis buffer (150 mM NaCl, 50 mM Tris pH 7.4, 1% NP-40, 0.25% deoxycholate and 1 mM EDTA, PMSF 1 mM, NaF 1 mM, Na3VO4 1 mM, 1 µM of aprotinin, leupeptin, pepstatin) for 10 min at 4°C. After centrifugation at 13,000 g for 20 min at 4°C, protein lysates were denatured at 90°C for 10 min, separated using 10% SDS-PAGE, and electroblotted onto a PVDF membrane (Roti-PVDF; Roth GmbH) according to standard protocols. After blocking in 5% nonfat dry milk/PBST for 2 h, membranes were incubated overnight at 4°C with a monoclonal anti-Fhl2 antibody and anti-NICD antibody detecting only active Notch1 intracellular domain (NICD) that had been cleaved at valine at position 1744 (V1744) (Cell signaling, dilution 1:1000). Membranes were than washed, incubated with horseradish peroxidase-conjugated secondary antibody (dilution 1:500; Cell signaling), and proteins detected by using enhanced chemiluminescence (ECL). Equal protein loading was ensured by probing blots with an anti–GAPDH antibody (dilution 1:20,000; Abcam). Co-immunoprecipitations were performed as described previously for Myc-tagged Fhl2 proteins (Martin et al., 2002) by using protein A/G PLUS-agarose beads (Santa Cruz).

Transfection of HEK293 cells or neurosphere-derived progenitors was performed with the K2 transfection system (Biontex, Martinsried, Germany) as recommended by the manufacturer. Cells grown in 6-well plates were transfected with reporter plasmid pGL3/Hes1-luc (0.5 µg for HEK293 cells and 1 µg for neural progenitors) either alone or with expression plasmids encoding the NICD or FHL2 (pCDNA-FHL2), with 0.5 µg per well for HEK293 cells and 1 µg per well for neural progenitors. NICD and FHL2 constructs have been described previously (Martin et al., 2002; Romer et al., 2003). When pGL3/GFAP-luc was used as a reporter plasmid for the transfection of WT or Fhl2^−/− neural progenitor cells, 0.4 µg per well was added. While the amount of single plasmid DNA varied depending on the experimental protocol, the total quantity of plasmid DNA per dish was kept constant by adding the appropriate amount of empty vector. Each transfection was carried out in triplicate. Luciferase activities were measured 24 h after transfection using the Luciferase Assay System (Promega) and relative light units were measured using Veritas luminometer (Promega, Turner Biosystems).

Fhl2^−/− and WT neurospheres cultured in suspension for 7 days were trypsinized for 5 min; after centrifugation for 5 min at 600 g, cells were resuspended in FACS-PBS. After fixation of the probe with 2% PFA for 15 min and permeabilization with 0.4% Tween 20 for 15 min, cells were washed twice with FACS-PBS. Then, cells were incubated with an anti-GFAP antibody (Abcam) for 30 min at 4°C, washed twice with FACS-PBS and incubated with FITC-conjugated secondary antibody (Dako Cytomation) for an additional 20 min. After washing, detection of GFAP+ cells was performed using a BD FACS DIVA flow cytometer (BD Biosciences).

Data are provided as the mean±s.e.m. Statistical analysis was performed by using the Student's t-test and GraphPad Prism software, with P values <0.05 being regarded as statistically significant.

We thank Ida Allabauer, Elisabeth Koppmann and Tanja Schied for expert technical assistance, and Ute Saunders for kindly providing the reporter plasmids pGL3/Hes1-luc, pGL3/Hes1ΔAB-luc and expression plasmids for NICD. The present work was in part performed in fulfillment of the requirements for the degree ‘Dr. med.’ of S.Y.K.