Psychiatric disorders such as bipolar disorder (BIP), schizophrenia (SCZ), and major depressive disorder (MDD) are classified based on neuropsychiatric symptoms (NPS), and these diseases remain to be the most prevalent among others. Patients suffering from NPS have been documented to experience symptoms that are representative of these disorders such as mood swings, hallucinations, and depression. The prevalence of BIP, SCZ, and MDD is considerable worldwide and also in the United States, with its 12-month prevalence amounting to 2.6, 1.1, and 6.7%, respectively (Regier et al., 1993; Kessler et al., 2005). Although the etiology of NPS is not yet clear, it is accepted that NPS are caused by alterations within the brain, namely changes in gene expression.

NPS refer to non-cognitive symptoms of apathy, anxiety, agitation, aggression, depression, hallucinations, and delusions, unlike cognitive symptoms of Alzheimer’s disease (AD) such as loss of memory and reduced executive functions. NPS represent a high proportion of the global burden of disease, primarily due to the lack of effective preventive and therapeutic methods for treating NPS (WHO, 2001). Interestingly, AD is significantly similar in NPS with BIP, SCZ, and MDD. AD patients have reported frequent experience of NPS, which has been found to have 80–90% prevalence across patients (Steinberg et al., 2004; Nowrangi et al., 2015). AD patients with NPS also show an increase in the rate of hospitalization and mortality while its presence cuts life expectancy by 10–20 years (Sweet et al., 2003; WHO, 2016).

Similarly, AD animals have also been found to elicit NPS in a number of transgenic models. Well-characterized AD transgenic models such as the hAPP/PS1 mouse has been observed to exhibit aggressive behavior, while TgCRND8 mice carrying Swedish and Indiana APP mutations were reported to demonstrate increased auditory response and reduced prepulse inhibition responses (McCool et al., 2003; Pugh et al., 2007). Notably, previous research has indicated that mice with forebrain specific presenilin 1 and 2 knockout (PSEN dKO) may possess a variety of NPS ranging from depression, apathy, and aggressive behavior – which mirrors many of the NPS present in AD patients (Yan et al., 2013).

In order to discover the molecular mechanisms underlying AD and psychiatric disorders, high throughput techniques providing comprehensive results of the transcriptome – such as microarrays – are utilized to identify genes responsible for pathogenesis (Mirnics et al., 2000; Hakak et al., 2001; Kang et al., 2007). Currently, most microarray studies screen for genes that significantly change compared to controls, and analyze the interaction of individual genes in pathways and networks to indirectly reveal the phenotype of these genes (Li and Chen, 2014). Similarly, previous microarray studies using AD models have selected genes and subsequently performed pathway and network analysis to explain the manifestation of cognitive deficits in AD. However, in our study, we sought to identify significantly altered genes within the gene ontology (GO) gene pools associated with psychiatric disorders (BIP, SCZ, and MDD), and to explain whether NPS in AD may have resulted from the changes in these genes.

To obtain the GO gene pools of psychiatric disorders (BIP, SCZ, and MDD), we cited a genomic-wide association study (GWAS) that was based on single nucleotide polymorphisms (SNP) (Network and Pathway Analysis Subgroup of Psychiatric Genomics Consortium, 2015). The study suggested 20 pathways relevant to psychiatric disorders such as BIP, SCZ, and MDD. As we believed that GO terms from SNP studies could represent the symptoms of BIP, SCZ, and MDD, we used 14 GO terms that were enriched with BIP, SCZ, and MDD susceptibility alleles. We observed significant expression changes of genes in the cortex and hippocampus of PSEN dKO mice, brain regions where neurodegeneration occurs in AD. These genetic alterations were not only related to cognitive decline in AD, but also associated with NPS observed in AD.

This is the first study to attempt the analysis of gene expressions for NPS in a mouse model of AD. In the present study, we profiled mRNA expressions in the cortex and the hippocampus in comparison to the wild type group using GO terms related with BIP, SCZ, and MDD. By exploring the patterns of gene expressions, we found significant differences in BIP, SCZ, and MDD-related genes in AD model. In addition, we found that the genes in the hippocampus showed more dramatic changes than in the cortex. Furthermore, we discovered an interesting point. The SCZ-related genes were the most noticeable genes among the genes of the three disorders and the altered SCZ-related genes were mainly involved in synaptic function. Accordingly, this study may help to clarify underlying mechanisms of NPS in AD.

Here we found altered expression of neuropsychiatric disorder associated genes in AD model. As we believed that GO terms by SNP studies could represent the symptoms of BIP, SCZ, and MDD, we interpreted the changed expressions of these genes in AD as probable causes of BIP-like, SCZ-like, and MDD-like NPS in AD such as mood swings, hallucinations, and depression. Also, it is notable that the genes in the AD animal model are being investigated by using the results of human studies. Therefore, transcriptome analysis of this AD mouse model based on human studies allows for the application of results to NPS in AD patients in the future.

The cortex and hippocampus are brain regions show neurodegeneration in both AD patients and AD animal models (Saura et al., 2004). The AD animal model, PSEN dKO mice, showed earlier progression of neurodegeneration in the cortex than in the hippocampus, and gene expression changes at later stages were mainly observed in the hippocampus than in the cortex (Wines-Samuelson et al., 2010; Watanabe et al., 2014; Ham et al., 2017). Consistent with these previous studies, in this study, marked alterations in expressions of BIP, SCZ, and MDD related genes in the hippocampus were observed compared to the cortex due to old age of AD mice (7 and 18 month aged mice). An important point to consider is that it is easy to conceive that these brain regions are suitable for screening genes associated with memory loss and cognitive symptoms than NPS in AD. However, the cortex and hippocampus are brain regions associated with cognitive dysfunction and hallucination of BIP, SCZ, and MDD as well as memory and cognition functions of AD (Frodl et al., 2006; Amad et al., 2014). Therefore, the present investigation performed within the regions of cortex and hippocampus was adequate for NPS in AD. However, in order to confer higher accuracy in terms of regional association for NPS in AD, further studies will be needed within the regions of caudate putamen, nucleus accumbens, amygdala, and hypothalamus, as well as the cortex and hippocampus (Sun et al., 2013).

Schizophrenic symptoms such as hallucination and delusion occur considerably in AD to such an extent that new diagnosis, “AD with psychosis” (AD+P) has been formed (Murray et al., 2014). Those with AD+P account for 30–50% among all AD patients (Murray et al., 2014). On the other hand, depression is more common in AD than schizophrenic symptoms (about 50%) (Garcez et al., 2015). However, our results showed the lowest relationship between AD and depression. Interestingly, the most dominant changes in gene expression were observed in schizophrenia-related genes within both brain regions. It is important to note that schizophrenic symptoms are manifest usually in the later stages of AD whereas depressive symptoms appear especially during the early and middle stages of AD^1,2. Consequently, due to using old aged AD mice in this study, SCZ-related genes than MDD-related genes in the AD model were differentially expressed compared to normal mice.

The SCZ associated genes were Hist1h1c, Pcdhb16, Arc, Cnn3, and Stx3. Among these genes, the protocadherin beta 16 (Pcdhb16) is a member of the protocadherin beta gene cluster. The protocadherin gene family encode proteins with an ectodomain comprising cadherin repeats. These protocadherin proteins have a single transmembrane domain and a distinct cytoplasmic domain, and can be further subdivided into clustered and non-clustered protocadherins. In the clustered protocadherins, there are alpha-, beta-, and gamma-protocadherin families while non-clustered protocadherins consists of individual protocadherins (1, 7, 9, 10, 11X, 11Y, 12, 17, 18, 19, and 20) (Gul et al., 2017). Studies have reported alpha-, and gamma- clustered protocadherins proteolysis by presenilins (Buchanan et al., 2010), whereas the non-clustered protocadherin-X, Y have been investigated as candidate genes for schizophrenia (Giouzeli et al., 2004; Kalmady and Venkatasubramanian, 2009). However, no study to date have been performed regarding the protocadherin beta 16 (Pcdhb16) as a schizophrenia-associated gene. The preexisting evidence of the role of protocadherin beta 16 pertains to associations with synaptic connectivity and neuronal networks, and is still under investigation (Junghans et al., 2008).

Arc has been mainly investigated for its roles in activity-driven synaptic plasticity and long-term memory consolidation (Rudinskiy et al., 2012). However, several literatures have examined the relevance between the Arc gene and schizophrenia (Manago and Papaleo, 2017). Particularly, schizophrenic mouse models showed reduced Arc expression (Takao et al., 2013; Takagi et al., 2015). In addition to alterations in gene expression, Arc knockout mice showed schizophrenic behaviors such as social impairment, impaired pre-pulse inhibition (which is a major characteristic observed in schizophrenia patients), and memory deficits (Manago et al., 2016). Furthermore, other possible roles of Arc has been suggested such as modulation of glutamatergic and dopaminergic systems which are key disrupted neurotransmitter systems in schizophrenia (Manago and Papaleo, 2017). Therefore, in the scope of the present study, altered Arc expression implicates that synaptic dysfunction via abnormal expression level of Arc may be involved in schizophrenic symptoms as well as memory loss and impaired cognitive function in AD.

Calponin-3 (Cnn3) is an actin-interacting protein and modulates postsynaptic currents and the size of dendritic spines (Rami et al., 2006). In a previous microarray study that was performed within the dorsolateral prefrontal cortex of schizophrenia patients, increased expression level of Cnn3 was reported (Glatt et al., 2005). However, the roles of calponin-3 in schizophrenia have not been elucidated. Therefore, differentially expressed Cnn3 in this study suggests the need for further investigation of the role of calponin-3 in schizophrenic symptoms of AD. This study suggests that the Pcdhb16, Arc, and Cnn3 may be associated with synaptic functions on the basis of their GO terms. Together, considering the present study along with the aforementioned literature, abnormal expression levels of these genes in the cortex and hippocampus of AD point towards and provides evidence for synaptic dysfunction. This synaptic dysfunction in the cortex and hippocampus of AD may implicate schizophrenic symptoms of AD as well as memory loss and cognitive deficits.

As an important point to note, Arc and Paxbp1 were expressed in opposite directions at 7 and 18 months. Expressions of Arc and Paxbp1 at 7 month were increased while these expressions at 18 month were decreased (Figure 5). This complex pattern of expression may have originated from the characteristics of these genes in the pathological conditions of AD. According to current literature, Arc expression in AD patients and mice showed discrepancies in results (Kerrigan and Randall, 2013). Furthermore, the expression level of Arc in the same AD mice changed depending on their ages (Kerrigan and Randall, 2013). One recent study with APP/PS1 AD mice provided a plausible explanation to these discrepancies (Rudinskiy et al., 2012). Amyloid plaques may elicit locally aberrant increase of Arc expression in active neurons. However, this study also observed a significant decrease of Arc expression in active neurons near amyloid plaques of mice. Due to unestablished role of Arc in AD, the Arc expression pattern in this study mice needs to be elucidated with definite experimental designs as a part of a future study.

PAX binding protein 1 (PAXBP1) is an adaptor protein linking the transcription factor PAX3 and PAX7 to the H3K4 histone methyltransferase complex. Although DNA hypo-methylation in vitro and higher levels of DNA methylation in vivo as epigenetic alterations in AD have been reported (Sanchez-Mut and Graff, 2015), roles of target genes of PAX3 and PAX7 such as Cdc20 and Id3 have not been researched in neither AD nor MDD (Diao et al., 2012). Therefore, we can restrictively interpret the opposite expression of PAXBP1 as potential roles in AD pathology based on epigenetic mechanisms with the need of further elucidation on the role of PAXBP1 in AD and MDD.

Until recently, NPS in AD have received less attention than memory loss and cognitive function deficits in AD. This big data analysis of genes associated to NPS in the AD animal model reveals the need for additional studies of NPS in AD. In the future, studies elucidating the underlying causal molecular mechanisms of NPS in both AD patients and AD animal models are needed.

We used PSEN dKO mice with partial deletion of PSEN1 in the forebrain and PSEN2 null. The wild type and PSEN dKO mice were fed ad libitum, and 2–4 animals were housed in a cage under the following cage conditions: constant humidity (45–55%), temperature (21–23°C), and 12 h light-dark cycle in a laboratory breeding room at the Korea Institute of Science and Technology (KIST). All procedures regarding the use and the handling of animals were approved by the Institutional Animal Care and Use Committee (IACUC) of KIST. All procedures were performed in accordance to the institutional guidelines and regulations on animal care and use from the IACUC.

H-II and Y-PT designed the study. H-II carried out the RNA isolation and microarray, and reviewed and commented on the manuscript. SH and TK analyzed, interpreted data, and drafted the manuscript. HH and YK discussed and edited the manuscript. All authors read and approved the final manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.