Formin 2 links neuropsychiatric phenotypes at young age to an increased risk for dementia.

Age-associated memory decline is due to variable combinations of genetic and environmental risk factors. How these risk factors interact to drive disease onset is currently unknown. Here we begin to elucidate the mechanisms by which post-traumatic stress disorder (PTSD) at a young age contributes to an increased risk to develop dementia at old age. We show that the actin nucleator Formin 2 (Fmn2) is deregulated in PTSD and in Alzheimer's disease (AD) patients. Young mice lacking the Fmn2 gene exhibit PTSD-like phenotypes and corresponding impairments of synaptic plasticity, while the consolidation of new memories is unaffected. However, Fmn2 mutant mice develop accelerated age-associated memory decline that is further increased in the presence of additional risk factors and is mechanistically linked to a loss of transcriptional homeostasis. In conclusion, our data present a new approach to explore the connection between AD risk factors across life span and provide mechanistic insight to the processes by which neuropsychiatric diseases at a young age affect the risk for developing dementia.

Introduction

Alzheimer's disease (AD) is the most common neurodegenerative disorder causing a huge emotional and economical burden to our societies. The vast majority of AD cases are sporadic and arise on the background of variable genetic and environmental risk factors. There is substantial evidence that pathological changes occur years before the onset of clinical symptoms (Bateman et al, 2012). To elucidate how the different risk factors contribute to disease onset is therefore of utmost importance. The starting point of this study is epidemiological data indicating that individuals suffering from neuropsychiatric diseases such post‐traumatic stress disorder (PTSD) at a young age have an increased risk of developing AD in old age (Yaffe et al, 2010; Burri et al, 2013; Weiner et al, 2013; Stilling et al, 2014a). PTSD develops in response to a traumatic event when normal psychological defense mechanisms fail. Individuals that suffer from PTSD are deficient in learning that a stimulus associated with an adverse event no longer presents a threat, a process that requires cognitive flexibility and extinction learning (Goswami et al, 2013). Memory extinction can be analyzed in rodents using the well‐established fear conditioning paradigm. This paradigm represents a specific form of learning that underlies the reduction of previously acquired fear memories (Myers & Davis, 2002; Lattal et al, 2006; Sananbenesi et al, 2007; Fischer & Tsai, 2008; Radulovic & Tronson, 2010). Although impaired fear extinction in rodents does not recapitulate the complex phenotypes observed in humans suffering from PTSD, it is suitable to study the mechanisms that underlie increased susceptibility to PTSD. Thus, we reasoned that a promising strategy to elucidate the molecular mechanisms that link PTSD at a young age to an increased risk for age‐associated memory decline and AD would be to screen for animal models that exhibit (i) impaired fear extinction at young age, while consolidation of new memories is still intact, but (ii) develop accelerated memory impairment in the presence of additional AD risk factors. By this, we identified the Formin 2 (Fmn2) gene. FMN2 is best known for its role in regulating actin dynamics (Schuh, 2011) and was previously detected in a screen for genes that are deregulated in the aging mouse hippocampus (Peleg et al, 2010). In addition, FMN2 has been linked to synapse formation and deletion mutations of the Fmn2 gene are associated with intellectual disability, pointing to a role for Fmn2 in memory function in mice and humans (Peleg et al, 2010; Almuqbil et al, 2013; Law et al, 2014). Here we show that loss of Fmn2 affects plasticity at the mossy fiber‐CA3 synapse and causes impaired fear extinction in young mice. We furthermore show that Fmn2 expression is decreased in PTSD patients and in post‐mortem brain samples from AD patients. Loss of Fmn2 accelerates age‐associated memory decline, which is further accelerated in the presence of amyloid pathology and is accompanied by deregulation of hippocampal gene expression. While the mechanisms that link reduced FMN2 levels to aberrant gene expression are likely to be multifactorial, we provide evidence that FMN2‐dependent synaptic actin dynamics signal via the ERK1/2 pathway to drive Elk1 and SP1‐dependent gene expression. Of note, memory impairment in all employed models is ameliorated after oral administration of the HDAC inhibitor Vorinostat. In sum, our data represent a new approach to explore the cross‐talk between AD risk factors across life span and provide evidence that loss of transcriptional plasticity is a key event down‐stream of AD risk factor exposure.

Results

Loss of FMN2 leads to impaired fear extinction in young mice

We started our analysis by employing animal models for age‐associated memory decline and tested whether extinction phenotypes precede the impairment of memory consolidation. We decided to test APPPS1‐21 and 5xFAD mice, two well‐established mouse models for amyloid deposition and age‐associated memory impairment that are linked to familiar Alzheimer's disease (Radde et al, 2006; Ohno et al, 2007; Govindarajan et al, 2013). In addition, we employed Fmn2 knockout (Fmn2 −/−) mice. Although the role of FMN2 in the adult brain is not well understood, FMN2 has recently been linked to age‐related memory impairment in mice (Peleg et al, 2010) and is associated with cognitive dysfunction in humans (Almuqbil et al, 2013; Law et al, 2014). Mice from all experimental groups were subjected to contextual fear conditioning training at 3 month of age. Freezing behavior, indicative of associative memory consolidation, was similar in all mutant mice when compared to the corresponding control groups during a memory test performed 24 h later (Fig 1A). To test extinction of fear memory, animals were re‐exposed to the conditioning context on consecutive days in the absence of the foot shock, a procedure that induces extinction learning and leads to the reduction in the aversive freezing response (Sananbenesi et al, 2007; Tronson et al, 2012). Fear extinction was similar in APPPS1‐21 (Fig 1B) and 5xFAD mice (Fig 1C) when compared to the corresponding control groups. In contrast, fear extinction learning was impaired in Fmn2 −/− mice (Fig 1D). Next we confirmed that all three lines of mutant mice exhibit impaired consolidation of new memories already at 8 months of age (Fig 1E), while wild‐type mice were impaired only at 16 months of age (Fig 1F). These data indicate that Fmn2 −/− mice develop deficits in fear extinction that precede impairment of memory consolidation. However, the possibility remained that the employed fear conditioning protocol does not allow for the detection of very mild memory impairments. Therefore, we also subjected Fmn2 −/− mice to a milder fear conditioning protocol (Kerimoglu et al, 2013). In line with our initial observation, Fmn2 −/− mice did not exhibit memory impairment at 3 months of age (see Appendix Fig S1). In line with these data, 3‐month‐old Fmn2 −/− mice were able to acquire spatial reference memory in the Morris water maze test similar to wild‐type mice (Fig EV1A and B). Reversal learning was, however, impaired (Fig EV1C and D), which supports the view that loss of Fmn2 in young mice does not affect the consolidation of new memories but is required for changing responses to existing memories. Taking together, these data support the view that the analysis of FMN2 would be a good starting point to investigate the mechanisms by which neuropsychiatric diseases at young age contribute to an increased risk for age‐associative memory decline. Before continuing our analysis, however, we sought to provide further evidence for this hypothesis. To this end, we tested Fmn2 expression in blood samples from PTSD patients via qPCR. We observed a significant reduction in Fmn2 expression in PTSD patients when compared to age‐matched control individuals (Zieker et al, 2007; Fig 1G). Although care has to be taken when interpreting data obtained from blood in the context of brain diseases, there is increasing evidence that adverse life events can induce long‐lasting changes in the expression of specific genes and that such changes in gene expression are detected in various cell types such as cells obtained from blood or saliva (Smith et al, 2015). Thus, the analysis of gene expression, for example, in blood, is viewed as a suitable approach to identify biomarker and surrogate marker for brain diseases (Rao et al, 2013; Ciobanu et al, 2016; Schmitt et al, 2016). The fact that Fmn2 levels are altered in blood samples from PTSD patients therefore indicates that exposure to PTSD‐inducing events may similarly altered Fmn2 levels in the brain. Since we do not have access to suitable post‐mortem tissue from PTSD patients, this hypothesis remains to be tested. However, PTSD and other neuropsychiatric diseases have been linked to altered glucocorticoid signaling (Du & Pang, 2015; Kim et al, 2015). It is therefore interesting that Fmn2 expression was decreased in human neuronal progenitor cells subjected to acute or chronic dexamethasone treatment, a synthetic glucocorticoid that can have detrimental effects on neuronal function and cognitive abilities (Crochemore et al, 2005; Tongjaroenbuangam et al, 2013; Feng et al, 2015; Lanshakov et al, 2016; Fig 1H). We also observed decreased Fmn2 expression in post‐mortem human brain samples (hippocampus) from AD patients when compared to age‐matched control individuals (Fig 1I).

Figure 1

Impaired fear extinction precedes memory decline in Fmn2 mutant mice

Data information: Error bars indicated SEM.

Freezing behavior tested 24 h after training was similar in 3‐month‐old APPPS1‐21, 5xFAD, and Fmn2 −/− mice when compared to corresponding control groups (n = 10/group).

Fear extinction in 3‐month‐old APPPS1 mice is similar to a non‐transgenic control group (n = 9/group).

Fear extinction in 3‐month‐old 5xFAD mice is similar to a non‐transgenic control group (n = 9/group).

Extinction learning in the contextual fear conditioning paradigm is impaired in Fmn2 −/− mice (n = 9/group; ***P < 0.001, F = 21.77, two‐way RM ANOVA).

Freezing behavior was tested 24 h after training in 8‐month‐old APPPS1‐21, 5xFAD, and Fmn2 −/− mice. A significant impairment in associative memory was observed when comparing APPPS1 mice (n = 10) to a corresponding control group (n = 8; **P < 0.01, t‐test), when comparing 5xFAD mice (n = 10) to a corresponding control group (n = 8; **P = 0.01, t‐test), and when comparing Fmn2 −/− mice (n = 10) to corresponding Fmn2 +/+ wild‐type mice (n = 10; ***P < 0.0001, t‐test).

Freezing was analyzed in 3‐ (n = 11), 12‐ (n = 9), and 16 (n = 8)‐month‐old male wild‐type mice. A significant impairment (***P < 0.0001, t‐test) was observed in 16‐month‐old animals when compared to either 3‐ or 12‐month‐old animals.

Fmn2 expression measured in blood samples via qPCR was reduced in PTSD patients (n = 9) when compared to age‐matched controls (n = 7; *P = 0.0361, t‐test).

Fmn2 mRNA expression is decreased after dexamethasone treatments in human hippocampal progenitor cells (HPCs) during neuronal proliferation (3 days) and differentiation (10 days) (***P adj < 3.27 × 10−6, one‐way ANOVA followed by Bonferroni correction).

qPCR revealed that Fmn2 expression is decreased (*P = 0.037, t‐test) in AD patients (n = 9) when compared to age‐matched controls (n = 8).

Figure EV1

Hippocampal FMN2 is required for reversal learning in the water maze paradigm

Data information: T, target quadrant. Error bars indicate SEM.

Mice were trained in the Morris water maze test. Three‐month‐old Fmn2 −/− and wild‐type mice show similar escape latencies throughout training (n = 11/group).

A probe test performed after 7 days of water maze training revealed no significant difference amongst groups. Both groups showed a significant preference for the target quadrant (n = 11/group; *P < 0.05, t‐test). The color key below the x‐axis refers to the 4 quadrants used for the schematic water maze pool shown in panel (A).

Left panel: Three‐month‐old mice were subjected to water maze training until they could reliably find the hidden platform on day 7 of training. Right panel: Afterward, mice were subjected to reversal learning. Escape latencies during reversal learning were severely impaired in Fmn2 −/− mice when compared to the control group (control, n = 15, Fmn2 −/−, n = 13; ***P < 0.001, F = 5.377, two‐way RM ANOVA).

A probe test was performed after 5 days of reversal training. While wild‐type mice showed a preference for the new target quadrant, target preference in Fmn2 −/− mice was significantly impaired (control, n = 15, Fmn2 −/−, n = 13; *P < 0.001 versus Fmn2 +/+ control group, t‐test).

Encouraged by these observations, we decided to study the role of FMN2 in the adult brain in greater detail. Fmn2 was highly expressed in mouse and human hippocampal neurons (Fig 2A). In line with these findings, robust FMN2 protein levels were observed in the mouse and human hippocampus (Fig 2B). The specificity of the FMN2 antibody employed for immunoblotting was confirmed using corresponding hippocampal tissue from Fmn2 −/− mice as negative control (Fig 2C). Since we could not identify a suitable antibody to reliably detect FMN2 via immunostaining, we generated Fmn2‐EGFP knock‐in mice. These mice are healthy and show no obvious phenotype (see Appendix Fig S1B and C). The analysis of FMN2 protein localization in FMN2‐EGFP knock‐in mice confirmed high levels of FMN2 protein in the hippocampus and especially in the mossy fiber pathway (Fig 2D), suggesting that FMN2 can localize to the pre‐synaptic compartment. In line with this, we detected robust levels of FMN2 in hippocampal synaptosomes (Fig 2E). Furthermore, the pre‐synaptic marker protein synaptoporin co‐localized with FMN2 within the stratum lucidum of the hippocampal CA3 region (Fig 2E).

Figure 2

FMN2 is highly expressed in mossy fibers and is required for synaptic depotentiation

Data information: For all experiments, parasagittal brain slices from 14‐ to 21‐day‐old mice were used. Cere, cerebellum; Hip, hippocampus, mf; mossy fiber, str. luc.; stratum lucidum, cc, corpus callosum. Error bars indicate SEM.

Fmn2 expression was analyzed via qPCR in brain regions from 3‐month‐old mice (left panel, n = 5) and in mouse primary hippocampal neurons, astrocytes (n = 5) and human post‐mortem hippocampus (right panel; n = 3/group).

Representative immunoblot images showing FMN2 protein levels in the mouse hippocampus (Hip M), cerebellum (Cere M), and human hippocampus (Hip Hu).

Representative immunoblot images showing FMN2 levels in the hippocampus of wild‐type and Fmn2 −/− mice.

Representative low (upper panel, scale bar 500 μm) and high magnification images (lower panels, scale bars 100 μm) showing FMN2 expression (green) in FMN‐EGFP knock‐in mice.

Left panel: Representative images showing the co‐localization of FMN2 (green) and synaptoporin in the mossy fiber pathway in low and high magnification. Right panel: Immunoblot analysis of hippocampal synaptosomes (Syn) compared to cell homogenate (H) showing that FMN2 is present in the pre‐synaptic compartment. Synaptophysin was used to confirm enrichment of synaptosomes. Scale bar: 20 μm.

LTP was similar in Fmn2 −/− (n = 12) and wild‐type control mice (n = 8).

The absolute level of LTD was similar amongst Fmn2 −/− and wild‐type mice (n = 10).

Depotentiation was significantly impaired in Fmn2 −/− mice (n = 12) when compared to control (n = 11; *P < 0.05, t‐test).

These data suggest that the susceptibility to develop PTSD‐like phenotypes observed in 3‐month‐old Fmn2 −/− mice involves compromised plasticity at the hippocampal mossy fiber‐CA3 synapse. Thus, we analyzed the electrophysiological properties of mossy fiber synapses in Fmn2 −/− mice at 3 months of age. We found that long‐term potentiation (LTP) and long‐term depression (LTD) were unaffected (Fig 2F and G). However, Fmn2 −/− mice exhibited impaired depotentiation, a well‐established phenomenon by which mossy fiber‐CA3 LTP can be reversed using long trains of low‐frequency stimulation that has been discussed as a molecular correlate of extinction processes (Hong et al, 2009; Kim et al, 2009; Fig 2H).

Although gross brain morphology, motor coordination, explorative behavior, and basal anxiety were similar when comparing 3‐month‐old Fmn2 −/− mice and control littermates (Fig EV2), our Fmn2 −/− mice constitutively lack FMN2. Thus, the possibility remained that the observed fear extinction phenotype could in part be due to subtle development abnormalities and in addition may not be specific to hippocampal function. To test for this possibility, we employed an RNAi approach and found that RNAi‐mediated decrease in FMN2 in the hippocampus impairs fear extinction, but has no effect on the consolidation of new memories. To this end, we first confirmed that intra‐hippocampal injection of an siRNA against Fmn2 reduces Fmn2 mRNA and protein levels when compared to the scramble control group (Fig 3A–C). Next we subjected mice to contextual fear conditioning and injected siRNA or scrambled control RNA for 3 days (every 12 h) before animals were subjected to fear extinction (Fig 3D). Injections were continued during extinction trials 1‐3. Extinction was significantly impaired when comparing the siRNA versus the scramble control RNA‐injected mice (Fig 3E). When animals were allowed to rest for 6 days after extinction trial 4 and did not receive further injections, normal extinction behavior was observed when mice were subsequently subjected to further extinction training. These data indicate that siRNA‐mediated impairment of extinction is not due to unspecific effects of the injection procedure (Fig 3D and E). Finally, we show that siRNA‐mediated knockdown of Fmn2 does not affect the acquisition of contextual fear memories (Fig 3F).

Figure EV2

FMN2 is expendable for locomotor function and basal anxiety

Data information: Error bars indicate SEM.

Upper panel: Representative images showing NeuN immunoreactivity in the hippocampus of 8‐month‐old Fmn2 −/− and control mice. Scale bar: 200 μm. Middle panel: Silver staining showing that gross brain morphology is normal in 8‐month‐old Fmn2 −/− mice. Lower panel: Brain weight is similar in 8‐month‐old Fmn2 −/− and control mice (n = 5/group). CA1, hippocampal subregion CA1; CA3, hippocampal subregion CA3.

Fmn2 −/− mice show normal explorative behavior as assayed in the open‐field test. Upper panel shows a representative image depicting the path in the open field. Lower panel shows the activity during the 5‐min open‐field exposure (n = 10/group).

Basal anxiety as measured by the time spent in the four corners (C1–C4) versus the center of the open‐field arena. No significant difference was found amongst groups (n = 10/group).

Fmn2 −/− mice show normal motor behavior in the rotarod test (n = 10/group).

Figure 3

Hippocampal FMN2 is essential for extinction learning and cognitive flexibility

Data information: Error bars indicate SEM.

Experimental design. Anti‐Fmn2 or scrambled siRNA oligomers were injected via microcannulae into the hippocampus of 3‐month‐old wild‐type mice every 12 h for 3 days.

Fmn2 mRNA levels were determined in the hippocampus of mice injected with either scrambled or Fmn2 siRNA via qPCR (n = 4/group). Fmn2 levels are reduced in mice injected with Fmn2 siRNA when compared to the scrambled control group (*P < 0.05, t‐test).

Similarly, hippocampal FMN2 protein levels were decreased in mice injected with Fmn2 siRNA when compared to the scramble control group (*P < 0.05, t‐test, n = 4/group).

Experimental design. Microcannulae were implanted into the hippocampus of wild‐type mice. All mice were subsequently subjected to contextual fear conditioning training. Fmn2 siRNA or scrambled control RNA was injected into the dentate gyrus 24 h later and then every 12 h for 3 days. On subsequent days, mice were subjected to extinction trial (E)1, E2, E3, and E4 (panel B). The injection protocol was continued until E3.

Fear extinction is significantly impaired (P < 0.001, F = 15.55; *P < 0.05, ***P < 0.001 for post hoc analysis) in 3‐month‐old mice that received intra‐hippocampal injections of Fmn2 siRNA when compared to the scrambled control group (n = 9/group). When siRNA injection was terminated, all mice exhibited normal extinction behavior.

Left panel: Experimental design. Right panel: Injection of Fmn2 siRNA before fear conditioning training does not affect the acquisition of fear memories (n = 10/group).

In conclusion, these data indicate that loss of FMN2 leads to impaired fear extinction which is accompanied by mild deficits in neuronal plasticity that mimic the behavioral alterations.

Loss of FMN2 leads to accelerated age‐associated memory impairment

Next, we decided to explore the mechanisms by which FMN2‐mediated phenotypes at a young age increase risk for late life dementia. We first confirmed our observation that 8‐month‐old Fmn2 −/− mice exhibit impaired memory formation (see Fig 1E) in an additional memory test, namely the hippocampus‐dependent Morris water maze paradigm. There was no difference in the escape latency when comparing 3‐ versus 8‐month Fmn2 +/+ mice (Fig 4A). However, 8‐month‐old Fmn2 −/− mice were significantly impaired when compared to age‐matched wild‐type mice (Fig 4B). This finding was confirmed in a probe test performed after 10 days of training. While 3‐ and 8‐month‐old wild‐type mice showed a significant preference for the target quadrant, 8‐month‐old Fmn2 −/− mice were significantly impaired (Fig 4B). These data confirm that loss of Fmn2 accelerates age‐associative memory impairment. Next we decided to investigate the impact of FMN2 on cognitive decline in the presence of another risk factor for AD, namely amyloid deposition. To this end, we employed APPPS1‐21 mice. We observed that 3‐month‐old APPPS1‐21 mice subjected to the open‐field test and the contextual fear conditioning paradigm showed explorative (Fig 4C) and freezing behavior that was similar to wild‐type mice, suggesting that at 3 months of age APPPS1‐21 mice show no defect in associative memory formation (Fig 4D). Thus, we reasoned that the analysis in 3‐month‐old mice would enable us to detect synergistic effects on memory impairment in Fmn2 −/− and APPPS1‐21 mice. To this end, we crossed Fmn2 −/− mice with APPPS1‐21 (APP) mice and tested memory function in animals at 3 months of age. As expected, brain weight was not affected (see Appendix Fig S2). Moreover, explorative behavior in the open‐field test did not differ significantly amongst groups (Fig 4C). When subjected to the fear conditioning paradigm, neither 3‐month‐old Fmn2 −/− nor APPPS1‐21 mice showed impaired freezing behavior in comparison with control littermates (Fig 4D). Freezing behavior was, however, significantly impaired in 3‐month‐old Fmn2 −/−_APPPS‐21 mice (Fig 4D). Next we assayed spatial memory in the Morris water maze paradigm. All groups were able to acquire spatial memory throughout the 8 days of training (Fig 4E). When subjected to the probe test, all groups except Fmn2 −/−_APPPS1‐21 mice showed a significant preference for the target quadrant (Fig 4F). In conclusion, these data show that reduced levels of FMN2 further accelerate memory decline in a mouse model for aging and in a model for amyloid deposition.

Figure 4

FMN2 is linked to age‐associated memory decline

Data information: Error bars indicate SEM.

Escape latency in 3‐month‐old Fmn2 +/+ mice (n = 12) compared to 8‐month‐old Fmn2 +/+ (n = 20) and Fmn2 −/− mice (n = 20). Escape latency is significantly impaired in 8‐month‐old Fmn2 −/− mice (***P = 0.001, F = 24.1 versus age‐matched Fmn2 +/+, two‐way RM ANOVA).

A probe test revealed that the time spent in the target quadrant significantly differed amongst groups (P = 0.0004, F = 9.4; one‐way ANOVA) and that 8‐month‐old Fmn2 −/− mice (n = 20) spent significantly reduced time in the target quadrant when compared directly to age‐matched (*P = 0.0143, n = 20, post hoc analysis) or 3‐month‐old Fmn2 +/+ mice (***P < 0.0001, n = 12, post hoc analysis).

Left panel: The total distance traveled during a 5‐min open‐field exposure did not differ significantly amongst groups (n = 8/group). Right panel: The time spent in the center versus the corners during a 5‐min open‐field exposure was similar amongst groups (n = 8/group).

Contextual freezing in 3‐month‐old Fmn2 +/+ (n = 14), APPPS1‐21 (n = 13), APPPS1‐21_Fmn2 −/− (n = 14), and Fmn2 −/− mice (n = 14). Freezing behavior differed amongst groups (P = 0.0005, F = 6.947, one‐way ANOVA) and was significantly impaired in APPPS1‐21_Fmn2 −/− mice when compared to the other groups (***P < 0.0001, post hoc analysis).

Escape latency in the water maze test in 3‐month‐old Fmn2 +/+ (n = 15), APPPS1‐21 (n = 15), APPPS1‐21_Fmn2 −/− (n = 15), and Fmn2 −/− mice (n = 15). Comparing the learning curves amongst all groups revealed no significant difference.

During the probe test, there was a significant difference amongst groups in the time spent in the target quadrant (P = 0.04, F = 3.22, one‐way ANOVA). Except for APPPS1‐21_Fmn2 −/− mice (n = 15), all groups showed a significant preference for the target quadrant (*P < 0.05, post hoc analysis). A direct comparison between Fmn2 −/− (n = 15) and APPPS1‐21_Fmn2 −/− mice (n = 15) revealed a significant reduction in the latter group (*P = 0.017).

Loss of FMN2 accelerates age‐ and amyloid‐induced deregulation of gene expression

We have previously hypothesized that the various AD risk factors eventually cause aberrant gene expression, thereby contributing to the loss of homeostasis and memory decline (Sananbenesi & Fischer, 2009; Fischer, 2014). To test whether this hypothesis could help to explain how loss of Fmn2 at young age contributes to age‐associative memory decline, we employed the hippocampal DG region for RNA‐sequencing. First we compared gene expression levels in 3‐month‐old Fmn2 −/− mice and age‐matched control littermates. In addition to the Fmn2 gene, we detected only 26 differentially expressed sequences (Fig 5A; Dataset EV1). Nevertheless, GO‐term and functional pathway analysis identified “oxidative phosphorylation”—and especially subunits of the NADH dehydrogenase and ATPase—to be increased (Fig 5B and C). When we compared the gene expression in cognitively impaired 8‐month‐old Fmn2 −/− mice to age‐matched control littermates, we detected 461 differentially expressed genes, of which the majority were down‐regulated (Fig 5A). GO‐term and pathway analysis revealed that the genes deregulated were linked to ribosome function, RNA splicing, Alzheimer's disease, and oxidative phosphorylation (Fig 5D and F). While genes linked to oxidative phosphorylation were increased in 3‐month‐old mice, genes of the same complexes and especially genes encoding subunits of the NADH dehydrogenase, ATPase, and Cytochrome C reductase were all down‐regulated in 8‐month‐old mice (Fig 5E, Dataset EV1). These data suggest that the susceptibility to develop PTSD‐like phenotypes observed in 3‐month‐old Fmn2 −/− mice is not accompanied by massive changes in gene expression, while memory impairment in 8‐month‐old Fmn2 −/− mice correlates with substantial deregulation of transcriptome plasticity. The fact that the spliceosome appeared to be deregulated in 8‐month‐old Fmn2 −/− mice is in line with recent data suggesting that during aging, synaptic plasticity genes are deregulated at the level of differential exon usage (Stilling et al, 2014b; Benito et al, 2015). Thus, we also analyzed differential exon usage. In 3‐month‐old Fmn2 −/− mice, we detected 46 genes that showed altered differential exon usage when compared to 3‐month‐old control littermates (Fig 5G, Dataset EV1). No significant pathways were identified. The same comparison was performed in 8‐month‐old Fmn2 −/− and age‐matched control mice and revealed 286 differentially expressed exons (Fig 5G). Further GO‐term and pathways analysis showed that processes affected by differential splicing in 8‐month‐old Fmn2 −/− mice are linked to synapse function (Fig 5H). In line with the gene expression data, we also observed that processes linked to energy metabolism and oxidative phosphorylation were deregulated (Fig 5H). In conclusion, these data indicate that the PTSD‐like phenotypes observed in Fmn2 −/− mice are not linked to substantial changes in gene expression, while aberrant gene expression is accelerated in the context of aging as an additional AD risk factor. To further substantiate this finding, we also analyzed gene expression in the DG of 3‐month‐old Fmn2 −/−_APPPS‐21. Comparing the gene expression in 3‐month‐old APPPS1‐21 and age‐matched wild‐type control mice revealed 50 differentially expressed genes, while 268 differentially expressed genes were detected comparing 3‐month‐old APPPS1‐21 mice to age‐matched Fmn2 −/−_APPPS1‐21 mice (Fig 5I). The genes affected in 3‐month‐old APPPS1‐21 mice were mainly linked to pathways representing inflammatory processes (Fig 5J). The top pathways affected in Fmn2 −/−_APPPS1‐21 mice when compared to APPPS1‐21 mice were mainly linked to metabolic processes, suggesting that loss of Fmn2 does not accelerate Aβ‐induced inflammation but is associated with deregulation of general cellular processes.

Figure 5

Memory decline in 8‐month‐old Fmn2 −/− mice correlates with deregulated transcriptome plasticity

Data information: Error bars indicate SEM. GO:CC, gene‐ontology domain “cellular compartment”; GO:MF, gene‐ontology domain “molecular function”; GO:BP, gene‐ontology domain “biological process”; pathway, “KEGG pathways”.

Volcano plot showing genes differentially expressed in the dentate gyrus of either 3 (red)‐ or 8‐month (black)‐old Fmn2 −/− mice compared to corresponding wild‐type control group (FDR < 0.05, log2 fold change < ± 0.25). The upper panel depicts the number of altered genes.

Top GO domains and pathways affected in 3‐month‐old Fmn2 −/− mice when compared to wild‐type control animals.

Expression of selected genes affecting the oxidative phosphorylation pathway that are increased in 3‐month‐old Fmn2 −/− mice (*P < 0.05; t‐test, n = 5/group).

Top GO domains and pathways deregulated in 8‐month‐old Fmn2 −/− mice when compared to an age‐matched control animals.

Expression of selected genes of the oxidative phosphorylation pathway that were decreased in 8‐month‐old Fmn2 −/− mice (*P < 0.05; t‐test, n = 3–5/group).

Cytoscape‐generated network of genes deregulated in 8‐month‐old Fmn2 −/− mice.

Volcano plot showing differential exon usage in the dentate gyrus of either 3 (red)‐ or 8‐month (black)‐old Fmn2 −/− mice when compared to the corresponding wild‐type control group (FDR < 0.05, log2 fold change < ± 0.2). The upper panel depicts the number of altered genes.

Selected GO domains and pathways affected by differential exons usage in 8‐month‐old Fmn2 −/− mice when compared to a wild‐type control group.

Volcano plot showing genes differentially expressed in 3‐month‐old APPPS1‐21 mice compared to a wild‐type control group (green) and APPPS1‐21_Fmn2 −/− mice compared to wild‐type control mice (purple; FDR < 0.05, log2 fold change < ± 0.5). The upper panel depicts the number of altered genes.

Functional pathways affected when comparing APPPS1‐21 (green bars) or APPPS1‐21_Fmn2 −/− mice (purple bars) to wild‐type mice.

In sum, these findings show that loss of FMN2 in young mice leaves the hippocampal transcriptome rather unaffected. However, chronically reduced FMN2 levels accelerate deregulation of hippocampal transcriptome plasticity in response to aging or amyloid deposition.

Since our data indicate that FMN2 is a synaptic protein, we hypothesize that chronically low levels of FMN2 will eventually affect pathways linked to synapse‐to‐nucleus signaling (Jordan & Kreutz, 2009). FMN2 has been implicated with the regulation of actin dynamics, and it is interesting to note that changes in actin polymerization were found to signal to the nucleus and thereby induce gene expression changes (Olson & Nordheim, 2010). In line with this, we observed that hippocampal actin dynamics were altered in Fmn2 −/− mice (Fig EV3A). When we analyzed the promoter regions of genes deregulated in 8‐month‐old Fmn2 −/− mice for the presence of any consensus sequences, we observed a significant enrichment for the motifs “CCCGCCCC” and “CCGGAAGC” which represents the ETS family of transcription factors (e.g., ELK) and the transcription factor specificity protein 1 (SP1), respectively (Fig EV3B). This is interesting, since SP1 and ELK have been linked to memory function and Alzheimer's disease (Sananbenesi et al, 2002; Sung et al, 2013; Szatmari et al, 2013; Citron et al, 2015; Wei et al, 2016) and are—amongst other kinases—activated via ERK1/2 (Bonello & Khachigian, 2004; Salim et al, 2007; Besnard et al, 2011; Kim et al, 2012). ERK1/2 is highly expressed in the mossy fiber pathway (Hu et al, 2004; Provenzano et al, 2014); has been linked to memory function, fear extinction, and Alzheimer's disease (Sweatt, 2004; Fischer et al, 2007; Kim & Choi, 2015); interacts with the actin cytoskeleton; and translocates to the nucleus upon stimulation (Wang & Hatton, 2007; Berti & Seger, 2017). It is thus tempting to hypothesize that chronically reduced FMN2 levels cause subtle change to the actin cytoskeleton that will eventually alter ERK1/2‐SP1/ELK signaling leading to aberrant gene expression. In line with this hypothesis, we observed that FMN2 and ERK1/2 interact and colocalize at the mossy fiber synapse (Fig EV3C–F) and that levels of active ERK1/2, SP1, and ELK1 are reduced in 8‐month‐old Fmn2 −/− mice (Fig EV3D). Although these data suggest one possible link between loss of FMN2 function and aberrant gene expression (Fig EV3G), certainly also other processes contribute to this phenotype. Irrespective of the multifactorial processes that link decreased FMN2 levels to aberrant gene expression, it is interesting to note that drugs aiming to state physiological gene expression such as inhibitors of histone‐deacetylases (HDAC) were shown to improve learning behavior in mouse models for age‐associated memory decline and amyloid deposition (Guan et al, 2009; Kilgore et al, 2010; Peleg et al, 2010; Govindarajan et al, 2011; Benito et al, 2015). Therefore, we wondered if HDAC inhibitors may also help to improve memory function in 8‐month‐old Fmn2 −/− mice and in 3‐month‐old Fmn2 −/−_APPPS‐21 mice.

Figure EV3

FMN2 affects actin dynamics and ERK1/2‐linked gene expression programs

Data information: Error bars indicate SEM.

FMN2 has been implicated with the regulation of actin dynamics and the actin cytoskeleton is essential for pre‐synaptic plasticity and synaptic vesicle dynamics (Hatada et al, 2000; Owe et al, 2009). As such, we hypothesized that alterations to actin‐related processes may contribute to the observed phenotypes. The levels of G and Factin were measured in the dentate gyrus isolated from Fmn2 +/+ and Fmn2 −/− mice (n = 4) using the “F‐Actin/G‐actin In Vivo Assay” from Cytoskeleton, INC. The upper panel shows a representative image of the Western blot performed with the anti‐actin antibody provided with the kit. The lower panel shows the quantification of the F/G ratio. There was a significant reduction in F/G‐actin ratio in Fmn2 −/− mice (*P = 0.0268, t‐test).

We analyzed the genes deregulated in 8‐month‐old Fmn2 −/− mice for consensus motifs and target transcription factors at promoters of candidate genes using MEME‐ChIP platform (Bailey et al, 2009; Machanick & Bailey, 2011) with default parameters. Targets with E‐value above 0.05 were considered significant. Promoter regions were defined as ± 2 kb from TSS. We observed a significant enrichment of two motifs; “CCCGCCCC” which is a motif common for SP transcription factor binding sites and “CCGGAAGC” which is a target of ETS/ELK family of transcription factors.

Representative images showing co‐localization of ERK1/2 and FMN2 within the hippocampal mossy fiber pathway. Scale bars: 500 μm for low magnification images and 100 μm for high magnification images. mf, mossy fiber; str. luc, stratum lucidum.

Immunoblot analysis showing decreased levels of active ERK1/2, SP1, and ELK1 that was measured using phospho‐specific antibodies in the hippocampal dentate gyrus. Right panel shows quantification (n = 5/group; *P < 0.05; t‐test).

Immunoblot analysis of hippocampal synaptosomes (Syn) compared to cell homogenate (H), showing that FMN2 and pERK1/2 are present in the pre‐synaptic compartment, while pELK1 is absent. Synaptophysin was used to confirm enrichment of synaptosomes (n = 3).

Lysates from hippocampal synaptosomes were used for immunoprecipitation using an antibody detecting pERK1/2 (IP). IgG and homogenate (H) were used as negative and positive controls (n = 3). FMN2 and pERK1/2 but not pELK1 were detected in pERK1/2 immunoprecipitates (IP). These data indicate that FMN2 and pERK1/2 interact at synapses. As expected from our findings showing that pELK1 is not detected at synapses, we could not detect interaction of synaptic pERK1/2 and pELK1.

A model suggesting how synaptic FMN2 may impact on gene expression. FMN2 and ERK1/2 interact at mossy fiber synapses (green lines indicate actin cytoskeleton). By mechanisms that remain to be determined FMN2 contributes to the activation of ERK1/2 that subsequently activates ELK and SP1 transcription factors which translocate to the nucleus and initiate gene expression. Of course we cannot exclude that FMN2 also affects gene expression by other mechanisms. In the absence of FMN2 (as in Fmn2 −/− mice), the described pathway is less efficient which eventually contributes to altered gene expression. The fact that reduced ERK1/2 signaling is observed in the aging brain and in APP mice may explain why loss of FMN2 accelerates age‐associated changes in gene expression.

Vorinostat rescues memory impairment in aged Fmn2 −/− and Fmn2 −/−_APP mice

Vorinostat (SAHA) is an FDA‐approved HDAC inhibitor with beneficial effect on memory function (Kilgore et al, 2010; Benito et al, 2015). Thus, we decided to test its effect in aged Fmn2 −/− mice and in Fmn2 −/−_APPPS1‐21 mice. First, we treated 7‐month‐old Fmn2 −/− mice orally with Vorinostat (50 mg/kg) or placebo for 4 weeks. Explorative behavior measured in the open‐field test was similar amongst groups (Fig 6A). Basal anxiety analyzed by comparing the center activity in the open field (Fig 6B) and by subjecting animals to the elevated plus maze test was also similar amongst groups (Fig 6C). Next we analyzed hippocampus‐dependent associative memory in the contextual fear conditioning paradigm. Freezing behavior during the memory test was significantly increased in Vorinostat‐treated Fmn2 −/− mice (Fig 6D). We also analyzed hippocampus‐dependent spatial learning in the water maze paradigm. When compared to the placebo group, Fmn2 −/− mice treated with Vorinostat showed a significantly improved escape latency, indicative of improved spatial memory formation (Fig 6E). Swim speed was similar amongst groups (Fig 6E). During the probe test, only Vorinostat‐treated Fmn2 −/− mice showed a significant preference for the target quadrant (Fig 6F). We also tested the expression of selected genes linked to oxidative phosphorylation that were down‐regulated in the DG of 8‐month‐old Fmn2 −/− mice when compared to wild‐type control littermates (see Fig 5). All selected genes were increased in the Vorinostat group when compared to placebo‐treated Fmn2 −/− mice (Fig 6G). HDAC inhibitors can affect gene expression by changing histone‐acetylation. Reduced acetylation of histone 4 lysine 12 (H4K12ac) has been previously linked to age‐associated memory decline (Guan et al, 2009; Peleg et al, 2010; Benito et al, 2015). In line with the gene expression data, we observed reduced H4K12ac in aged Fmn2 −/− mice, which was, however, increased in mice treated with Vorinostat (Fig EV4). To further substantiate these findings, we performed a similar experiment in Fmn2 −/−_APPPS1‐21 mice. Two‐month‐old Fmn2 −/−_APPPS1‐21 mice were treated orally with either Vorinostat or placebo. Placebo‐treated APP mice served as an additional control. Behavior in the open‐field and elevated plus maze test did not differ amongst groups. However, similar to the findings in aged Fmn2 −/− mice, we observed that Vorinostat treatment was able to ameliorate memory deficits, H4K12ac and gene expression in Fmn2 −/−_APPPS1‐21 (Figs EV4 and EV5).

Figure 6

Cognitive decline in 8‐month‐old Fmn2 −/− mice is rescued by the HDAC inhibitor Vorinostat

Data information: Error bars indicate SEM.

The total distance traveled during a 5‐min open‐field exposure was similar amongst groups (Fmn2 −/− placebo, n = 9; Fmn2 −/− Vorinostat, n = 12).

The time spent in the center and the corners during a 5‐min open‐field exposure was similar amongst groups (Fmn2 −/− placebo, n = 9; Fmn2 −/− Vorinostat, n = 12).

The time spent in the open (o) and closed arms (c) or the center region (Ce) of an elevated plus maze was similar amongst groups (Fmn2 −/− placebo, n = 7; Fmn2 −/− Vorinostat, n = 10).

Freezing behavior during the memory test was significantly increased in the Fmn2 −/− Vorinostat group when compared to the placebo group (*P = 0.04, t‐test; Fmn2 −/− placebo, n = 8; Fmn2 −/− Vorinostat, n = 9).

Left panel: The escape latency in the water maze test was significantly improved in the Fmn2 −/− Vorinostat group when compared to the placebo group (***P < 0.0001, F = 4.86; two‐way RM ANOVA, n = 10/group). Right panel: Swim speed was not altered amongst groups.

Mice of the Fmn2 −/− Vorinostat group spent significantly more time in the target quadrant (T) compared to the other quadrants (OQ). Please note the values for OQ are presented as average of all three other quadrants (***P < 0.0001; post hoc analysis, n = 10/group).

Expression of selected genes that were down‐regulated in 8‐month‐old Fmn2 −/− mice was significantly increased in the dentate gyrus from mice of the Fmn2 −/− Vorinostat group when compared to placebo (*P < 0.05; t‐test, n = 4/group).

Figure EV4

Impaired H4K12ac in aged Fmn2 −/− and Fmn2 −/−_APP mice is rescued by Vorinostat treatment

Data information: Error bars indicate SEM.

Quantification of ChIP‐qPCR analysis showing H4K12ac at the promoter regions of three selected genes that showed decreased expression in 8‐month‐old Fmn2 −/− mice (compared to wild‐type mice) and were increased after Vorinostat treatment. In line with the gene expression data, we observed decreased H4K12ac in 8‐month‐old Fmn2 −/− mice when compared Fmn2 +/+ animals (one‐way ANOVA revealed a significant group difference for Atp5d, P = 0.003; Ndufb7, P = 0.005, and Uqcrq, P = 0.001). Post hoc comparison revealed a significant difference when comparing H4K12ac levels in the placebo‐treated Fmn2 −/− versus Fmn2 −/− Vorinostat‐treated or wild‐type mice for all three genes (*P < 0.05; **P < 0.01; n = 5/group).

Quantification of ChIP analysis showing H4K12ac at the promoter regions of three selected genes that showed decreased expression in 3‐month‐old Fmn2 −/−_APP mice (compared to 3‐month‐old APP mice) and were increased after Vorinostat treatment. One‐way ANOVA revealed a significant group difference for Cacng6, P = 0.004; Cdh19, P = 0.004, and Htr5b, P = 0.017. Post hoc comparison revealed a significant difference when comparing H4K12ac levels in the placebo‐treated APP versus Fmn2 −/−_APP mice and placebo versus Vorinostat‐treated Fmn2−/−_APP mice for all three genes (*P < 0.05; **P < 0.01; n = 5/group).

Figure EV5

Cognitive decline in 3‐month‐old Fmn2 −/−_APPPS1‐21 mice is rescued by the HDAC inhibitor Vorinostat

Data information: Error bars indicate SEM.

The total distance traveled during a 5‐min open‐field exposure was similar amongst groups (n = 7/group).

The time spent in the center or the corners during a 5‐min open‐field exposure was similar amongst groups (n = 7/group).

The time spent in the open arms (o) and the closed arms (c) of an elevated plus maze was similar amongst groups (n = 9/group).

One‐way ANOVA revealed a significant difference in freezing behavior during the memory test (n = 12/group; F = 8.975, P = 0.0006). Post hoc analysis revealed increased freezing behavior when comparing the Fmn2 −/− _APP Vorinostat to the Fmn2 −/−_APP placebo group (*P = 0.139).

Left panel: The escape latency in the water maze test was similar amongst groups (n = 12/group). Right panel: Swim speed was not altered amongst groups.

Mice of the Fmn2 −/−_APP Vorinostat group spent significantly more time in the target quadrant (T) compared to the other quadrants (OQ). Please note that the values for OQ are presented as average of all three other quadrants (*P < 0.03, post hoc analysis, n = 12/group).

The expression of selected genes affected in Fmn2 −/−_APP mice was increased when comparing the Vorinostat group to the placebo group (*P < 0.05; t‐test, n = 4/group).

Discussion

Our study was inspired by the observation that young individuals suffering from psychiatric diseases such as PTSD have an increased risk to develop AD as they age (Yaffe et al, 2010; Burri et al, 2013; Weiner et al, 2013). We reasoned that one possible way to begin elucidating this phenomenon would be to select genes that have been implicated with age‐associative memory decline and to test whether these genes may also play a role in the development of PTSD‐like phenotypes, which we analyzed in mice via fear extinction as a commonly used and robust paradigm. Nevertheless, we like to reiterate that results from animal models of neuropsychiatric diseases have to be interpreted with care, and while impaired fear extinction in rodents may point to the mechanisms that underlie increased susceptibility for PTSD, it does not fully recapitulate the phenotypes observed in PTSD patients. We observed that deficits in fear extinction precede memory decline in Fmn2 −/− mice, and moreover, Fmn2 expression was decreased in PTSD and AD patients and decided to follow up on this novel observation. We found that Fmn2 is highly expressed in neurons of the mouse and human hippocampus and is especially enriched in the hippocampal mossy fiber pathway, where it is localized to pre‐synaptic terminals. These data are in line with previous in situ hybridization findings showing that in the adult mouse brain Fmn2 expression is highest in the hippocampal dentate gyrus (Schumacher et al, 2004). Nevertheless, it is likely that posttranslational mechanisms contribute to the enrichment of FMN2 at the mossy fibers. Interestingly, loss of Fmn2 had no effect on LTP or LTD at the mossy fiber‐CA3 synapse. Taking into account that LTP and LTD have been considered as molecular correlates of memory consolidation, these data are in agreement with the fact that young Fmn2 mutant mice have no deficit in the consolidation of new memories. However, depotentiation was impaired in Fmn2 −/− mice. It is tempting to link this observation to the behavioral phenotypes. Hence, at the synaptic and the behavioral level, mice that lack Fmn2 at a young age are able to induce plasticity‐related processes that are essential for information storage, but are impaired in modifying this information subsequently. These data are in line with previous studies, suggesting that depotentiation in the amygdala is required for fear extinction (Hong et al, 2009; Kim et al, 2009). The precise mechanisms by which loss of FMN2 affects depotentiation remain to be investigated. However, depotentiation has been linked to altered actin dynamics which in turn affects multiple synaptic processes such as vesicle trafficking (Galvez et al, 2016). We suggest that future research should test if such processes play a role in FMN2‐mediated fear extinction. However, impaired depotentiation is certainly not the only process responsible for impaired fear extinction in Fmn2 −/− mice.

A key observation of our study is that chronically reduced levels of Fmn2 accelerate age‐associative memory decline. The fact that in young Fmn2 −/− mice the formation of hippocampus‐dependent memories is not affected suggests that memory decline observed during aging is most likely an indirect consequence of the molecular changes triggered by chronically reduced FMN2 function. We hypothesize that such changes eventually lead to altered gene expression contributing to memory impairment. Indeed, we observed neglectable changes in hippocampal gene expression in 3‐month‐old Fmn2 −/− mice, while loss of Fmn2 dramatically increases the number of deregulated genes during aging or in response to amyloid pathology. These data support our idea that AD risk factors synergistically drive aberrant gene expression and thereby eventually contribute to dementia (Fig 7). Importantly, neither amyloid deposition nor FMN2 are likely to affect gene expression directly. While the precise mechanisms that couple FMN2‐mediated defects in hippocampal function to deregulation of gene expression are likely multifactorial, it is interesting to note that stimuli which promote rearrangement of the actin cytoskeleton were found to induce gene expression changes (Olson & Nordheim, 2010). In line with this, we observe that loss of FMN2 affects actin dynamics which confirms previous data from other cellular systems (Pfender et al, 2011; Schuh, 2011; Montaville et al, 2016; Sahasrabudhe et al, 2016) and is in agreement with the fact that blocking actin dynamics in the hippocampus impairs fear extinction (Fischer et al, 2004). Alteration of actin dynamics has been associated with ERK1/2‐mediated changes in gene expression (Wang & Hatton, 2007; Berti & Seger, 2017). Indeed, we found reduced ERK1/2 activity in Fmn2 −/− mice. The finding that ERK1/2 activates the transcription factors ELK1 and SP1 is in line with our data showing that binding sites for these factors are enriched amongst the genes deregulated in aged Fmn2 −/− mice (Fig EV3G). However, there are other processes linking synaptic plasticity to gene expression (Jordan & Kreutz, 2009) that likely also contribute to a loss of transcriptional homeostasis. The view that disturbed transcriptome plasticity causatively contributes to memory loss is supported by our observation that the HDAC inhibitor Vorinostat reinstates hippocampal memory formation in aged Fmn2 −/− mice and also in APPPS1‐21 mice that lack Fmn2.

Figure 7

AD risk factors cause loss of gene expression control

Our central hypothesis is that the various AD risk factors eventually lead to aberrant gene expression and loss of transcriptional control. This is based on the assumption that proper gene expression is a core feature of cellular homeostasis and that changes occurring at various compartments of the cell (e.g., at the synapse) will eventually signal to the nucleus and cause gene expression changes. In the case of FMN2, we provide, for example, evidence that chronically low levels of FMN2 disturb synaptic actin dynamics and thereby affect ERK1/2‐dependent gene expression programs. Thus, even if an AD risk factor may play no direct role in gene expression, it may eventually contribute to a loss of transcriptional plasticity. We therefore suggest that targeting pathological gene expression could be a suitable therapeutic approach, especially for multifactorial diseases such as AD, where it is nearly impossible to determine all of the genetic and environmental factors that eventually contribute to clinical phenotypes.

In conclusion, our data provide insight to the molecular mechanisms by which neuropsychiatric diseases at a young age lead to an increased risk for dementia when individuals age. We suggest that our experimental approach should be applied to additional genes and environmental factors implicated with PTSD or other neuropsychiatric diseases. Our data also provide a more general insight of how the various AD risk factors (in our case, FMN2‐mediated PTSD‐like phenotypes, amyloid deposition, and aging) contribute to dementia. Targeting transcriptome plasticity would thus be a very suitable therapeutic approach that is independent of the precise knowledge to the upstream pathological events (Fig 7). In line with this, we show that the HDAC inhibitor Vorinostat reinstates hippocampal memory formation in aged Fmn2 −/− mice and also in APPPS1‐21 mice that lack Fmn2. Of course, we cannot exclude that the memory enhancing effect of Vorinostat also involves processes not directly related to gene expression. It is, however, interesting to note that HDAC inhibitors are also discussed as a novel therapeutic avenue to treat PTSD (Whittle & Singewald, 2014). In fact, Vorinostat not only improves memory function, but also facilitates extinction of fear memories (Whittle et al, 2013), a finding we could confirm in our paradigm (see Appendix Fig S3). Our data therefore indicate that it might be possible to develop therapeutic strategies for PTSD patients that at the same time lower the risk to develop Alzheimer's disease, a line of research that should also be taken into consideration for other neuropsychiatric diseases.

Materials and Methods

For detailed description of methods, see the Appendix Supplementary Methods.

Animals and human tissue

Male mice were housed under standard conditions with free access to food and water. All experiments were carried out in accordance with the animal protection law and were approved by the District Government of Germany. Fmn2 −/− and APPPS‐21 mice have been described before (Leader et al, 2002; Radde et al, 2006). Three‐, 12‐, and 16‐month‐old wild‐type mice were obtained from Janvier Labs. Post‐mortem hippocampal tissue from AD patients and controls was obtained with ethical approval from the Alzheimer's disease Research Center Brain bank at Massachusetts General Hospital, Boson, MA, and from Brigham & Women's Hospital Autopsy Service, Boston, MA, USA. Samples were matched for age and post‐mortem delay. AD patients Braak and Braak stage were 3–5. EDTA blood samples from PTSD patients and controls were from the Ramstein cohort described previously (Zieker et al, 2007). These individuals were exposed to the trauma of the air show catastrophe in Ramstein, Germany, 1988. We employed nine age‐ and gender‐matched controls (46.38 ± 4.363 years of age) and seven PTSD patients (52.29 ± 5.515 years of age). Patients were diagnosed according to DSM IV (SKID‐I; Wittchen et al, 1997). Symptom severity was measured using the German versions of the clinician‐administered PTSD scale [CAPS] (Schnyder & Moergeli, 2002; mean = 40.0, SD = 22.9; > 40 moderate PTSD), the Posttraumatic Stress Diagnostic Scale [PDS] (Foa et al, 1993; mean = 18.8, SD = 11.1; > 20 moderate to severe PTSD), the 22‐item Impact of Event Scale Revised [IES‐R] (Horowitz et al, 1979; mean = 48.0, SD = 18.7), and the Questionnaire on Dissociative Symptoms [FDS] (Freyberger et al, 1999; mean = 9.8, SD = 8.5; > 8.4 dissociative symptoms). To characterize possible comorbidity with a depressive disorder, the German version of the Beck Depression Inventory [BDI] (Hautzinger, 1991) was performed (mean = 14.3, SD = 9.1; > 18 relevant depression). None of the patients had been taking psychoactive drugs on a regular basis 3 months prior to the study. All experiments were approved by the local ethical committees, and tissue/blood samples were obtained upon informed consent.

Behavior analysis

Mice were subjected to a series of behavioral tests as described before (Kerimoglu et al, 2013).

Cannulation and siRNA injection

Intra‐hippocampal injections were performed as described previously (Bahari‐Javan et al, 2012).

Statistical analysis

Unless specifically mentioned otherwise, data were analyzed by unpaired Student's t‐test, two tailed t‐test, Bonferroni test for multiple comparisons, or one and two‐way and ANOVA (analysis of variance) when appropriate. Errors are displayed as standard error of mean (SEM). Unless otherwise stated, analysis was performed using GraphPad Prism.

Gene expression

Analysis of RNA‐sequencing has been described before (Stilling et al, 2014a). Data have been deposited to GEO database: GSE100070.

Author contributions

RCA‐B performed most of the experiments; PP, NR, and CM performed electrophysiological measurements; CK, EB, GJ, and SB performed gene expression analysis; MG and SB‐J helped with behavioral experiments related to siRNA injections; ID provided post‐mortem human brain tissue; AJ, AS, and PF provided blood samples from PTSD patients; PAZ, JCP, and EBB performed experiments in human neuronal progenitor cells; AF and FS designed and coordinated the experiments and wrote the manuscript. MD generated FMN2‐EGFP mice.

Conflict of interest

The authors declare that they have no conflict of interest.

Appendix

Click here for additional data file.

Expanded View Figures PDF

Click here for additional data file.

Dataset EV1

Click here for additional data file.

Review Process File

Click here for additional data file.