Transposable Elements and Psychiatric Disorders

Abstract

Transposable Elements (TEs) or transposons are low-complexity elements (e.g., LINEs, SINEs, SVAs, and HERVs) that make up to two-thirds of the human genome. There is mounting evidence that TEs play an essential role in genomic architecture and regulation related to both normal function and disease states. Recently, the identification of active TEs in several different human brain regions suggests that TEs play a role in normal brain development and adult physiology and quite possibly in psychiatric disorders. TEs have been implicated in hemophilia, neurofibromatosis, and cancer. With the advent of next-generation whole-genome sequencing approaches, our understanding of the relationship between TEs and psychiatric disorders will greatly improve. We will review the biology of TEs and early evidence for TE involvement in psychiatric disorders.

INTRODUCTION

Transposable Elements (TEs) or transposons are low-complexity elements (e.g., LINEs, SINEs, SVAs, and HERVs) that make up to between half [Lander et al., 2001; Kuhn et al., 2007] and two-thirds [de Koning et al., 2011] of the human genome. Barbara McClintock discovered TEs in 1948 by observing color variegation in maize kernels and leaves. The observation that some grains were multi-colored was not in accordance with Mendel’s basic principles and lead to the identification of genes that have the ability to move from one location to another within the genome, also referred to as “jumping genes.” She called this phenomenon “transposition,” from which the name transposon is derived. The color variegation in grains suggested that transposons might also act as “controlling elements” and exert regulatory control over genes in their proximity. The insertion of a transposon in a pigment-producing gene can block pigment production in some cells, but insertion adjacent to a pigment-producing gene can also ultimately inhibit the pigment production. There is a growing understanding that TEs, and more generally the entire retroviral/retrotransposon gene machinery, play an essential role in normal, as well as disease-related, regulation of the human genome [Muotri et al., 2007; Goodier and Kazazian, 2008; Huang et al., 2010]. TEs are a significant part of the non-coding regions of our genome, whose architectural and functional complexity is the focus of the ENCODE project [Dunham et al., 2012; Harrow et al., 2012; Maurano et al., 2012]. They are also particularly relevant in shaping the non-coding to coding ratio pattern, which is mostly evident in brain-related genes [Taft et al., 2007]. Therefore, it is important to consider the role that TEs, as a key component of the regulatory non-coding DNA, may play in neuropsychiatric brain disorders.

Genome-wide association studies (GWAS) have produced a growing number of replicated Single Nucleotide Polymorphisms (SNP) associations in complex psychiatric diseases, including schizophrenia, autism, and bipolar disorder. Collectively, however, these common risk SNPs explain only a small fraction of the total heritability [Visscher et al., 2012]. Copy Number Variations (CNVs), and other rare de novo or almost essentially private mutations, represent another proportion of risk for these disorders [Coe et al., 2012; Malhotra and Sebat, 2012]. In most cases SNPs, Insertions/Deletions (Indels) and CNVs associated with complex psychiatric traits are mostly found in non-gene/non-coding regions and map predominantly to regulatory elements [Ripke et al., 2013], which include both TEs themselves and TE-derived elements, like long intergenic non-coding RNAs (lincRNAs) [Lin et al., 2011; Lee, 2012; Qureshi and Mehler, 2012; Ng et al., 2013].

The first report of a disease-associated TE insertion was in 1988 when Kazazian and others [1988] showed that mobilization of a LINE1 into the factor VIII blood clotting gene was responsible for hemophilia A in a young patient. The list of humans diseases caused by LINE1 insertions (retrotransposons insertional mutagenesis in germ cells) has expanded to include almost 100 single locus disorders (see the comprehensive review by Hancks and Kazazian, 2012). A number of studies in cancer and other diseases using next-generation sequencing (NGS) techniques combined with procedures enriching for repetitive sequences to detect germline polymorphic TE insertions [Witherspoon et al., 2010; Ewing and Kazazian, 2011], are beginning to characterize the landscape of mobile elements in disease. These studies are expanding into complex neuropsychiatric and other complex phenotypes.

We will review the main principles of TE activity, and then present our current understanding of how TEs may contribute to psychiatric disorders. First, we describe the TE classes and the mechanisms by which they mobilize within genomes. Then, we review what is known about TEs as genetic variants, their relevance in genome dynamics and their contribution to complex Neuropsychiatric diseases.

THE TRANSPOSABLE ELEMENTS

Transposable Elements (TEs) are classified into two major classes: DNA transposons and Retrotransposons (Fig. 1). DNA transposons account for ~3–5% of the genome and are apparently not mobile in humans, and no human disease has been reported to arise as a result of their activity to date [Solyom and Kazazian, 2012], although they contributed a fundamental evolutionary role in shaping the human genome [Feschotte and Pritham, 2007].

Fig. 1.

Transposable element classification and molecular structure. LTR-transposons or human endogenous retroviruses (HERVs) are classified in 30–50 distinct families (Gifford and Tristem, 2003) and structurally resemble exogenous retroviruses with two Long Terminal Repeat sequences flanking the coding sequences of the functional polyproteins capsid (gag), protease, polymerase and reverse transcriptase (pol) and envelope (env) that, however, lack the envelope component that is required to exit the cell. Non-LTR-retrotransposons include long interspersed elements (LINEs, present in 20.4% of the human genome), short interspersed elements (SINEs, 10.2%), and SVAs (SINE-R/VNTR/Alu-like, 2.6%). L1 elements require an intact 5′ untranslated region (UTR) that functions as an internal promoter and the production of full-length L1 mRNA that contains functional ORF1 (open reading frame) and ORF2 proteins, encoding for a reverse transcriptase [Moran et al., 1996]. Alus consist of a tRNA-related region that represents the internal promoter stretch, followed by a tRNA unrelated region and a LINE-related region, which is used by the SINE element to bind LINE-encoded proteins to complete LINE-1-mediated retrotransposition (Kajikawa and Okada, 2002). Alus derive their name from the presence of a single recognition site for the restriction enzyme AluI. SINEs are the second most abundant class of mobile elements with more than 1.6 million sites, of which 1.1 million are Alus interspersed in the genome between genes and within the introns. SVAs consist of four domains including a CT-rich repeat at the 5′ end, commonly referred to as CT-hexamer, an Alu-like sequence, so-called for homology with two antisense Alu-like fragments, a GC-variable number tandem repeats (VNTR), whose length determines variation of the full-length SVA elements, a sequence derived from the envelope gene (env) and a SINE-R, an LTR of an extinct HERV-K10 [Hancks et al., 2012]. In the DNA transposons a protein called transposase is bound by terminal inverted repeats (TIRs) that flank the TE, excises the TE out of the donor position and re-integrates it into the genome. In this schematic representation, the genomes of different TEs are not shown to scale.

Retrotransposons may be classified into two groups depending on the presence or absence of Long Terminal Repeats (LTRs). Retrotransposons are the only active TEs in humans. An active TE has the potential to move within the genome to new locations where it inserts in the original DNA sequence.

LTR-transposons or human endogenous retroviruses (HERVs) are 10 Kb long (but truncated elements are relatively common) and make up about 8% of the human genome. HERVs are assumed to be remnants from ancient germ line viral infections, integrated through reverse transcriptase as provirus in the germ line chromosomal DNA of progenitors of Homo sapiens, and passed to subsequent generations in a Mendelian fashion. HERVs are especially active prenatally and may impact on neurodevelopment. Evidence supports that the long terminal repeat (LTR) regions of many HERVs contain binding sites for transcription factors. These promoters could either activate HERVs or other genes located down-stream of the LTR [Boeke and Stoye, 1997; Forsman et al., 2005; Bannert and Kurth, 2006; Yi et al., 2006; Dickerson et al., 2008; Rowe and Trono, 2011]. HERVs are characterized by a very limited mobile activity [Gifford and Tristem, 2003], although some elements may have retained their ability to move through a retroviral-like mechanism [Turner et al., 2001].

The majority of TEs are represented by non-LTR-retrotransposons that include long interspersed nucleotide elements (LINEs, ~20.4% of our genome), short interspersed nucleotide elements (SINEs, ~10.2%), and SVAs (SINE-R/VNTR/Alu-like, ~2.6%). There are three major families of LINEs, L1, L2, and L3, but only the L1 family conserved the ability to transpose in the contemporary human genome [Moran et al., 1996] using an RNA intermediate by a copy-and-paste mechanism, a process called direct retrotransposition (RT). Not all L1s are able to retrotranspose, since most are truncated and transpositionally inactive, and thus we can distinguish between retrotransposition-competent and non-competent LINEs. RT competent LINEs are directly or indirectly responsible for new TE insertions. The presence or absence of the insertion of repetitive elements of recent acquisition is considered a polymorphism (sometimes called “dimorphism”) in humans [Batzer and Deininger, 1991; Batzer et al., 1995; Boeke and Stoye, 1997; Forsman et al., 2005; Christensen, 2010; Witherspoon et al., 2010; Lin et al., 2011; Qureshi and Mehler, 2012] and is referred to as Retrotransposon Insertion Polymorphimsm (RIP) [Ewing and Kazazian, 2011].

In 2006, a database of RIPs (dbRIP) was introduced to provide a comprehensive compilation of human variations derived from retrotransposon insertions [Wang et al., 2006]. L1s undergo active transposition in both germline and somatic cells to much greater degree than previously thought, significantly contributing also to the structural variation landscape in human genome [Lupski, 2010]. Iskow and colleagues reported that 45% of these L1 alleles had MAFs of 5% in a sample of 76 genomes [Iskow et al., 2010]. Ewing and Kazazian proposed that the number of dimorphic L1 elements in the human population with gene frequencies >5% were between 3,000 and 10,000. Most of the insertions identified in these studies were detected by one group only, opening the concept of “personal” retrotransposons insertions. Non-reference RIPs refer to those insertions that are not in the reference genome [Ewing and Kazazian, 2011] and should not be considered synonymous of “de novo” insertions: unless a new insertion is proved to be present in an individual but not in his/her parents, an insertion cannot be defined “de novo.” However, some reports define new insertions as “de novo” even without considering the status of the parental transmission.

SINEs originally derived from small cellular RNAs [Kramerov and Vassetzky, 2011; Schmitz, 2012]. The predominant types in humans are Alus that are about 300 bp long. SINEs do not code for proteins and mobilize using the retrotransposition machinery of a partner LINE [Kajikawa and Okada, 2002]. Alus seem to be the preferred sequences in the breakpoint regions for intra-chromosomal homologous recombination leading to deletions and to CNVs [Uddin et al., 2006].

SVA elements are structurally complex human-specific non-coding RNAs derived from different genomic repeats with size from 600 to 4,000 bp [Ostertag et al., 2003; Wang et al., 2005; Xing et al., 2006; Hancks et al., 2012]. About 3,000 SVAs are present in humans and are quite active and highly polymorphic.

THE IMPACT OF TEs ON THE HUMAN GENOME AND THE CENTRAL NERVOUS SYSTEM (CNS)

TEs Introduce Genetic Variability in the Human Genome

Xing and colleagues estimated the Alu, L1, and SVA retrotransposition rates to be 1/21 births (~4%), 1/212 births (~0.4%), and 1/916 births (~0.1%), respectively [Xing et al., 2009]. Ewing and Kazazian estimated that the rate of L1 retrotransposition in humans is between 1/270 and 1/95 births (between 0.3% and 1%, respectively) [Ewing and Kazazian, 2011]. Studies on inter-generational transmission of insertions are still limited [Stewart et al., 2011]. The molecular mechanisms by which TEs affect the cell have been extensively reviewed: see, for example [Chen et al., 2005; Goodier and Kazazian, 2008; Cordaux and Batzer, 2009; Beck et al., 2011; Hancks and Kazazian, 2012; Kaer and Speek, 2013], as well as, the current literature from the ENCODE project, like [Dunham et al., 2012; Kavanagh et al., 2013]. The landscape of somatic retrotransposition is proving to be much more complex [Poduri et al., 2013]. We provide a graphical summary of the main effects of TEs on human genome architecture and expression in Figures 2 and 3, respectively.

Fig. 2.

Effects of TE on human genome architecture. TEs can affect genomic integrity through new insertions and post-insertional rearrangements resulting in deleterious mutations or aberrant expression. (1) The primary way by which TEs affect the genome is simply by retrotransposing to a new genomic location introducing an insertion mutation, like shown in [Kazazian et al., 1988; Kazazian 2000; Stewart et al., 2011; Burns and Boeke, 2012; Solyom and Kazazian, 2012], which may be coupled with a deletion (2) at the corresponding insertion site in the original DNA sequence. (3) When transposing, TEs can create new exons or new genes by carrying along non-retrotransposon DNA sequences [Xing et al., 2006]. In single-gene diseases, new TE insertions mostly affect coding exons, but this may not be the major effect of retrotransposition in complex traits. (4) When a retrotransposon duplicates, the 3′ flanking sequence or the 5′ flanking sequence (green boxes) may also duplicate (known as 3′ or 5′ transduction, respectively), leading to the retrotransposition of the non-TE 3′ flanking sequence (left) or the 5′ flanking sequence (right) along with the retrotransposons. (5) L1 non-allelic homologous and non-homologous recombination events can result in significant duplications or deletions of genomic DNA sequence [Gilbert et al., 2002; Symer et al., 2002; Cordaux et al., 2006; Mine et al., 2006; Takasu et al., 2007; Han et al., 2008] or (6) create microsatellites from homopolymeric genomic tracts. Moreover, Alus and LINEs are key factors in generating Copy Number Variants (for more details on this mechanism see the section of the article “The impact of TEs on the Human Genome and the Central Nervous System (CNS)” [Mills et al., 2006, 2011; Maiti et al., 2011; Coe et al., 2012; Malhotra and Sebat, 2012; Gokcumen et al., 2013]). Schematic representations have been inspired by images in Cordaux and Batzer [2009].

Fig. 3.

Effects of TEs on human genome expression. TE insertions can have a great impact on gene expression [Cowley and Oakey, 2013], generating or elongating new exons (“exonization” [Schmitz and Brosius, 2011]), but also with other mechanisms, like RNA truncation by premature polyadenylation [Perepelitsa-Belancio and Deininger, 2003], L1 splicing [Belancio et al., 2006], gene-breaking [Wheelan et al., 2005], sequence transduction [Xing et al., 2006]. (1) About 500,000 TE-derived Transcription Start Sites (TSSs) have been mapped to the human genome and as many have been identified although not all have been properly mapped within the genome [Faulkner et al., 2009]. Lack of validation for many TSSs suggests a possible procedural noise [Zhao et al., 2011] or a more complicated pattern of the L1 promoter, which exists in the 900 bp 5′ untranslated region (UTR). Recent data convincingly show a complex organization of the L1 5′ UTR, where one internal sequence within the L1 5′ UTR initially recruits the transcription initiation complex, which is then positioned onto the “minimal promoter” (i.e., within the first 100 bp of the L1 5′ UTR) or onto “alternative” TSSs within the 5′ UTR or in the 3′ UTR of ORF1 [Alexandrova et al., 2012]. Both in vitro and genomic analyses suggest that this internal regulatory complex also overlaps substantially with the region known to act as an antisense promoter (ASP) for genes present on the opposite strand. One possible consequence of this complex functional architecture of the L1 5′ UTR is that LINEs can modify the architecture of the transcriptome not only by new insertions, but also long after the integration process is completed and also through truncated LINEs, frequently in a tissue- or disease-specific way [Matlik et al., 2006; Kines and Belancio, 2012]. Further to the complexity, genes can be simultaneously transcribed by the native promoter and the L1 ASP or the L1 ASP activity can be regulated independently [Matlik et al., 2006]. (2) Retrotransposons carry transcription factor biding sites and a transcription factor can up- or down-regulate the expression of neighboring genes. (3) A methylated TE can initiate the formation of heterochromatin at a gene close to the retrotransposition event and rework the expression of the gene. (4) LINE elements can be responsible for transcriptional interference by intron retention, exonization or polyadenylation (Kaer and Speek, 2013). Alus can modify gene expression by RNA editing, interfering with RNA splicing, generating active Transcription Factor binding sites, generating miRNAs and miRNA target sites other than by the same mechanisms (like, for example, polyadenylation or exonization) described for LINEs [Lander et al., 2001; Jurka, 2004; Oei et al., 2004; Borchert et al., 2006; Piriyapongsa et al., 2007; Cordaux and Batzer, 2009; Kines and Belancio, 2012; Ichiyanagi, 2013; Spengler et al., 2013]. Schematic representations have been inspired by images in Cordaux and Batzer [2009].

TE activity in germline and somatic tissues is highly regulated and normally repressed through a variety of mechanisms, including epigenetic [Slotkin and Martienssen, 2007] and post-transcriptional processes [Nishikura, 2006; Goodier et al., 2012]. Repression of expression through methylation of CpG sites is a key mechanism since non-LTR-retrotransposons contain one-third of all human CpG sites [Cordaux et al., 2006]. The many host silencing mechanisms, although largely unexplored, appear to be particularly complex [Bogerd et al., 2006; Tanay et al., 2007; Aravin and Bourc’his, 2008; Edwards et al., 2010; Garcia-Perez et al., 2010; McCue and Slotkin, 2012; Schmidt et al., 2012; Ahn et al., 2013; Horn et al., 2013; Ward et al., 2013].

Despite the sophisticated machinery used by the cell to regulate TE activity, some TEs escape repression and generate new insertions in germ cells during early embryonic development, as well as, in somatic tissues later in life [Baillie et al., 2011; Kazazian, 2011; Lee et al., 2012]. While the majority of the 970,000 L1 that are referenced in the current (hg19) annotation of the human genome, are retrotranspositionally inactive, there are about 100–170 retrotransposition-competent LINEs in an average diploid human genome. These active LINEs belong to one L1 family, the L1 Human Specific (L1HS), and are responsible for the bulk of ongoing retrotransposition in the human population [Brouha et al., 2003; Beck et al., 2010]. These active elements retrotranspose to new sites altering the architecture of the human genome as summarized in Figure 2, and can also create CNVs and other structural variation of the genome [Lupski, 2010] in both germline and somatic cells [Mills et al., 2006, 2011; Maiti et al., 2011; Coe et al., 2012; Malhotra and Sebat, 2012; Gokcumen et al., 2013]. These effects have been implicated in Autism and Developmental Disorders [Sebat et al., 2007; Coe et al., 2012; Michaelson et al., 2012; O’Roak et al., 2012] and suggested in Schizophrenia and other neuropsychiatric disorders [Consortium, 2008; Walsh et al., 2008; Maiti et al., 2011; Guffanti et al., 2013].

Next-generation sequencing studies and specially modified sequencing protocols have investigated polymorphic insertions segregating in different populations [Beck et al., 2010; Ewing and Kazazian, 2010; Huang et al., 2010; Iskow et al., 2010; Stewart et al., 2011]. Many L1HS elements are highly different across populations, for example many L1HS are specific to African populations, making them excellent markers for studies of population-specific diseases (Ewing and Kazazian, 2011). HERVs and Alus also show population-specific insertions in healthy humans from different ethnic origins with Alus representing the most abundant and active mobile element in the human genome [Mirsattari et al., 2001; Witherspoon et al., 2013]. TEs can differ in the specific locus position of individual “new” insertions, which are probably more frequent than previously thought. However, TEs do not transpose randomly within the genome: LINEs tend to re-insert into GC-poor regions, and Alus (and HERVs) into GC-rich regions [Pavlicek et al., 2001]. Together these findings provide a highly dynamic portrait of the genome where individuals differ not only with respect to presence or absence of various L1 insertions, but also to the relative position of the insertion [Lupski, 2010].

Somatic TE Retrotransposition in the Brain

Recent data suggest that TEs might contribute to neuronal plasticity [Steel et al., 1998; Muotri et al., 2005; Muotri and Gage, 2006; Coufal et al., 2009; Muotri et al., 2010; Singer et al., 2010; Baillie et al., 2011; Hunter and McEwen, 2013]. Unlike other organs, the brain has a great diversity of cell types and current findings suggest that TE retrotransposition might contribute to some of this diversity [Muotri and Gage, 2006]. The identification of expression of L1s during differentiation of rat hippocampal neural stem cells inspired the investigation of TE insertions in human neuronal progenitor cells. Initially, L1 elements were thought to mobilize only during early development [Ergun et al., 2004]. Then, Muotri and colleagues provided evidence that human engineered L1s could retro-transpose in transgenic adult rat hippocampus progenitor cells in vitro and in the mouse brain in vivo [Muotri et al., 2005]. This finding suggested that L1 mobilization might not only occur during neurogenesis, but also during later neuronal development, leading to individual somatic mosaicism [Lupski 2013; McConnell et al., 2013; Poduri et al., 2013]. In other words, neurons originating from the same progenitor cells could have different genetic makeup due to the newly inserted genomic sequence, a process that seems to be more evident in the brain than in other human organs. New L1 retrotransposition events detected in the hippocampus and several regions of the adult human brain showed significantly increased rates compared to the degree of endogenous L1 mobilization in heart or liver from the same individual [Coufal et al., 2009]. However, while the high-throughput sequencing study by Baillie et al. [2011] to detect L1, Alu and SVA germline and somatic insertions suggested an intense retrotransposons activity in protein–coding genes expressed in the brain using post-mortem tissues, other experiments conducted in single cells showed low frequency of unique somatic insertions per neuron [Evrony et al., 2012]. Moreover, no other data supporting Alu’s somatic retrotransposition is available. Whether these contrasting findings are the consequence of different experimental conditions, for example, tissue versus single cell sequencing, or still unknown mechanisms is still under scrutiny [Reilly et al., 2013].

L1 mediated retrotransposition has been studied more extensively but other TEs, such as HERVs and Alus, may also play a role in the evolution and development of the brain, independently from their ability to retrotranspose. For example, Sasaki and colleagues demonstrated that a SINE acted as an enhancer of the FGF8 (fibroblast growth factor 8) gene expression in two regions of the developing forebrain, namely the diencephalon and the hypothalamus [Sasaki et al., 2008]. New somatic insertions may end up generating considerable transcriptional diversity in the Central Nervous System suggesting that neurons are not static and that TEs can play an active role in determining neural plasticity [Singer et al., 2010].

TEs in Neuropsychiatric Conditions

TEs have been associated with psychiatric disorders, although their specific mechanism of action is not definitively known (Table I). HERVs have been the most studied TEs in neuropsychiatric disorders thus far. In 2000, Yolken and collegues followed by Karlsson and colleagues in 2001 identified retroviral sequences in the brain and CSFs of individuals with schizophrenia and not in controls [Yolken et al., 2000; Karlsson et al., 2001]. Karlsson and colleagues reported, for the first time abnormal expression of HERV genes in the CSF of the same subjects for whom they identified retroviral sequences, providing evidence for their relative potential functional impact. Since then, abnormal expression of HERV-W and -K families has been found in blood [Karlsson et al., 2004; Perron et al., 2008; Yao et al., 2008; Huang et al., 2011; Perron et al., 2012b], cerebrospinal fluid (CSF) [Karlsson et al., 2001], and post-mortem brain samples [Frank et al., 2005; Weis et al., 2007; Christensen, 2010] from patients diagnosed with schizophrenia, both at onset and in later stages of the disease.

TABLE I.

Examples of Conditions With Neuropsychiatric Features Associated With TEs

Disorder	TE type	Affected gene(s)	TE feature	Tissue	Sample size	Reference
Schizophrenia	HERV(-W), HERV(-K)		Upregulation of HERV-W and HERV-K transcription in the cortex of post-mortem brains of SCZ and BD vs. HC	Post-mortem brain	4 SCZ, 4 BD, 6 HC	Yolken et al. [2000]
Schizophrenia	HERVs (-W)		(i) Over-representation of HERV (-W) family in CSF of SCZ vs. HC; (ii) up-regulation of differential expression of HERV(-W) in frontal cortex brain regions in SCZ vs. HC	Cerebrospinal fluid and post-mortem brain	55 SCZ and 52 HC	Karlsson et al. [2001]
Schizophrenia	HERV(-W)		Presence of particles containing HERV-W RNA in the plasma of patients with recent-onset schizophrenia (subgroup of the sample described in Karlsson et al., 2001)	Blood plasma	54 SCZ, 46 HC	Karlsson et al. [2004]
Schizophrenia	HERV(-K)		(i) Over-expression of HERV-K in SCZ 8c BD vs. HC	Post-mortem brain	215 brain tissue samples derived from 39 SCZ, 38 BD, 39 HC	Frank et al. [2005]
Schizophrenia	HERV(-K)		The HERV-K115 insertion appeared to be more frequent in patients with younger onset of schizophrenia than those with later onset.	DNA	178 SCZ, 181 HC	Lin et al. [2011]
Schizophrenia	HERV(-W)		(i) reduced expression of HERV-W gag in neurons and astroglial cells in SCZ, BP and MDD vs. HC	Post-mortem brain (anterior cingulate cortex, hippocampus)	15 SCZ, 15 HC	Weis et al. [2007]
Schizophrenia	HERV(-K)	SLAM	Two-polymorphisms haplotype in the envelope region of Herv K-18, located in the CD48 signaling lymphocyte activating [SLAM] gene on chr1, is highly associated with type 2 diabetes in a population of 229 individuals with SCZ vs. 136 HC	DNA	229 SCZ, 136 HC	Dickerson et al. [2008]
Schizophrenia	HERV(-W)		(i) Significantly higher antigenemia of HERV-W ENV and GAG proteins in individuals with SCZ vs. HC; (ii) significant correlation of antigenemia with C-reactive protein (CRP)	Serum	49 SCZ, ~49 HC	Perron et al. [2008]
Schizophrenia	HERV(-W)	PTD015 (or C11orf67)	(i) Higher prevalence of HERV-W gag transcripts in SCZ vs. HC; (ii) intronic elements on the noncoding strand of PTD015 (11q13.5) appear to be transcribed at a higher rate in the patients during the transition from susceptibility to manifestation of symptoms.	PBMC	30 SCZ, 26 HC	Yao et al. [2008]
Schizophrenia	HERV(-W)	BDNF, NTRK2, DRD3, CREB	(i) Higher prevalence of HERV-W env transcripts in SCZ vs. HC; (ii) over-expression of HERV-W upregulate BDNF, NTRK2, DRD3 and increase CREB protein	Blood plasma	118 SCZ, 106 HC	Huang et al. [2011]
Schizophrenia	HERV(-W)		(i)Significantly higher HERV-W env transcription in BD and in SZ vs. HC (ii)The corresponding DNA copy number was paradoxically lower in the genome of patients with BD or SZ than in HC.	PBMC	91 BD, 45 SZ, 73 HC	Perron et al. [2012a,b]
Schizophrenia (& Parkinson–s)	Retrotransposition activity		Effect of somatic retrotransposition on brain metabolism, biosynthesis of 250 metabolites, including dopamine, serotonin and glutamate	in silico	Not Available	Abrusan [2012]
Bipolar disorder	HERV(-W), HERV(-K)		Upregulation of HERV-W and HERV-K transcription in the cortex of post-mortem brains of SCZ and BD vs. HC	Post-mortem brain	4 SCZ, 4 BD, 6 HC	Yolken et al. [2000]
Bipolar disorder	HERV(-K)		(i) Over-expression of HERV-K in SCZ & BD vs. HC	Post-mortem brain	215 brain tissue samples derived from 39 SCZ, 38 BD, 39 HC	Frank et al. [2005]
Bipolar disorder	HERV(-W)		(i)Significantly higher HERV-W env transcription in BD and in SZ vs. HC (ii)The corresponding DNA copy number was paradoxically lower in the genome of patients with BD or SZ than in HC.	PBMC	91 BD, 45 SZ, 73 HC	Perron et al. [2012b]
Bipolar disorder	HERV(-W)		(i) Reduced expression of HERV-W gag in neurons and astroglial cells in SCZ, BD and MDD vs. HC	Post-mortem brain (anterior cingulate cortex, hippocampus)	15 SCZ, 15 MDD, 15 HC	Weis et al. [2007]
Major depression	HERV(-W)		(i) Reduced expression of HERV-W gag in neurons and astroglial cells in SCZ, BD and MDD vs. HC	Post-mortem brain (anterior cingulate cortex, hippocampus)	15 SCZ, 15 MDD, 15 HC	Weis et al. [2007]
Post-traumatic stress disorder	LINE-1 and Alu		LINEs were hypermethylated in controls post- versus predeployment, hypomethylated in cases versus controls postdeployment. Alus wer hypermethylated for cases versus controls predeployment	Serum, pre- and post-deployment, DNA methylation assessed via pyrosequencing	75 post-deployment PTSD and 75 no postdeployment PTSD (controls)	Rusiecki et al. [2012]
Post-traumatic stress disorder	LINE-1		Identification of a module of stress regulated highly co-expressed transcripts with no annotations mapping to genomic locations corresponding to LINE-1 and up-regulated in the basolateral amygdala of rats	Genome-wide gene expression profile in amygdala	Stress-enhanced fear learning (SEFL) animal model on 56 adult male rats	Ponomarev et al. [2010]
Alcohol dependence	LINE, SINE, LTR		Several module of co-expressed SINE and LTR up-regulated in the central and basolateral amygdala and superior frontal cortex of the alcoholic brain; upregulation of the unique probe corresponding to a currently “active” LINE-1 family, L1HS; regions corresponding to upregulated probes show increased hypomethylation in alcoholics	Post-mortem brain from central (CNA) and basolateral nucleus [BLA] of amygdala, as well as the superior frontal cortex (CTX)	17 alcoholics and 15 matched control (gene expression); 6 alcoholics and 6 matched control (H3K4 histone methylation); 5 alcoholics and 5 matched controls (Methyl-Histone H3K4 ChIP)	Muotri et al. [2009]
Autism	HERV(-H, -W,- E, -K)		(i) HERV-H over-expression in ASD vs. HC; (ii) HERV-W down-expression in ASD vs. HC	PBMC	28 ASD, 28 HC	Balestrieri et al. [2012]
Attention deficit hyperactivity disorder (ADHD)	HERV(-H, -K, -W)		HERV-H over-expression in patients with ADHD vs. HC; no differences in the expression levels of HERV-K and W.	PBMC	30 ADHD, 30 HC	Balestrieri et al. [2013]
Rett syndrome	LINE-1	MECP2	MeCP2 mutations increase susceptibility for L1 retrotransposition in RTT patients	In vitro cell culture of neuronal progenitor cells derived from human induced pluripotent cells	Not available	Muotri et al. [2010]
Ataxia telangiectasia (neurodegenerative disorders)	LINE-1	ATM	(i) Increase in retrotransposition activity in L1s cells with lack of ATM; (ii) increased copy number of L1HS in ataxia telangiectasia patients vs. HC	L1RP-EGFP transgenic mouse, human brain post-mortem tissue, in vitro cell culture	Not available	Coufal et al. [2011]
Infantile encephalopathy (neurodegenerative disorder)	LINE-1	lincRNA SLC7A2	Point mutation in intron of SLC7A2 in a primate-specific retrotransposon sequence and related increase of neuronal apoptosis in in vitro KO for this transcript	DNA, post-mortem brain tissue, in vitro cell culture	9 families, 15 patients and 17 unaffected siblings	Cartault et al. [2012]
Frontotemporal lobar degeneration (FTLD) (neurodegenerative disorder)	LINE, SINE and LTR	TDP-43 (RNA-binding protein)	(i) TDP-43 broadly targets TE-derived transcripts, including many SINE, LINE and LTR classes as well as some DNA elements; (ii) the association between TDP-43 and TE-derived RNA targets is reduced in FTLD patients relative to healthy subjects; (iii) over-expression of TE derived transcripts in each of two different mouse models with TDP-43 dysfunction.	In vivo crosslinking-immunoprecipitation sequencing (CLIP-seq) datasets of human healthy and FTLD post-mortem brain, RNA immunoprecipitation sequencing (RIP-seq) data of rat cortical neuron cells, mouse CLIP-seq and mRNA-seq datasets	Not available	Li et al. [2012]
Neurodevelopment	LINE-1, Alus, SVAs	CAMTA1, DRD3, SLCBA5, SLC6A6 SLCBA9	High-throughput de novo insertions of LINEs, Alus and SVA wit over-representation in protein-coding genes implicated in neurogenesis and synaptic function pathways, in introns (LINEs), exons (Alus)	Post-mortem brain	3 HC	Baillie et al. [2011]
DiGeorge syndrome (mental retardation)	Alu		Alu-mediated chromosomal rearrangements leading to microdeletions on 22q11	DNA	Not available	Babcock et al. [2003]

For each psychiatric disorders (Schizophrenia, Bipolar Disorder, Major Depression, etc.) results are shown from published investigations, presenting the specific TE element involved and its features, the genes that are affected by the TE when this information is known, the tissue that has been investigated and a summary description of the sample that made it possible to detect a putative role of a given TE.

Perron and colleagues found that a higher level of HERV-W expression in schizophrenia was associated with a lower copy-number of HERV-W elements [Perron et al., 2012a]. Similar findings on HERV’s transcriptional activity were also reported for other neuropsychiatric disorders such as bipolar disorder [Frank et al., 2005; Weis et al., 2007; Perron et al., 2012b], major depression [Weis et al., 2007], autism [Balestrieri et al., 2012], and ADHD [Balestrieri et al., 2013]. While most of these studies focused on the overall detection of qualitative and quantitative profiles of HERVs transcriptional activity, two reports specifically addressed the effect of abnormal transcriptional profiles on gene activity. First, Yao and colleagues reported higher transcriptional activity of intronic elements on the non-coding strand of PTD015 (11q13.5) in schizophrenia patients during the transition from prodromal to symptomatic phases as a consequence of higher prevalence of HERV-W gag transcripts in schizophrenia [Yao et al., 2008]. Second, Huang and colleagues found upregulation of the schizophrenia relevant genes BDNF, NTRK2, DRD3 and increased CREB protein levels induced by the over-expression of HERV-W transcripts in schizophrenic patients [Huang et al., 2011]. The authors proposed a model in which infectious agents could trigger the expression of HERV proteins, which in turn would increase the promoter activation of genes such as those mentioned above, and whose expression would ultimately contribute to the pathogenesis of schizophrenia. These findings do not define whether over-expression of HERVs is a result of their reactivation due to disease or whether it is a causal event leading to disease.

The neuropsychiatric-related research on LINEs has mainly focused on the identification of active retrotranspositions in early and later brain development, potentially revealing signatures of expression. In their high-throughput study of somatic retrotransposition of the brain, Baillie and colleagues found that protein-coding loci are disproportionally affected by TEs, with over-representation of L1s in introns and Alus in exons [Baillie et al., 2011]. In particular, somatic L1 insertions affected genes detected in neuroblastoma and glioma (i.e., CAMTA1), dopamine receptors (DRD3), and several neurotransmitter transporters such as SLC6A5, -A6 and -A9. Overall, the gene ontologies that seemed to be enriched in the network of genes affected by new insertions were neurogenesis and synaptic function. Recently, a computational analysis proposed that the reduction of the expression of these genes, as a result of TE insertions, has the potential to influence brain metabolism and the biosynthesis of several neurotransmitters such as dopamine, serotonin and glutamate theoretically resulting in relatively larger inter-individual variability [Abrusan, 2012], but see also [Chen et al., 2006] and Figures 2 and 3 for added discussion.

Muotri and colleagues investigated Rett Syndrome subjects and found possible connections between disease-related genetic mutations and neuronal L1 retrotransposition [Muotri et al., 2010], showing that L1 retrotransposition was increased in the absence of methyl CpG binding protein (MeCP2) efficiency in both rodents knockdown animal models and in patients with Rett Syndrome mutations. These findings provide an example of a functional model for TEs in the brain and the influence of tissue-specific regulation/repression through epigenetic mechanisms. Similarly, Coufal and colleagues provided evidence of increased somatic L1 retrotransposition as a consequence of the ATM gene deficiency. The ATM gene is implicated in the pathway of DNA damage repair. Mutations in the gene have been linked to Ataxia telangiectasia, a syndrome characterized by progressive neurodegeneration. These findings partially conflict with previous results showing that DNA host repair mechanisms may rather lead to truncated insertions [Suzuki et al., 2009] or are even in complete disagreement with other data indicating that ATM is essential for L1 retrotransposition [Gasior et al., 2006].

Recently, in an infantile encephalopathy study Cartault and others found a point mutation on chromosome 8p22 mapping in a long intergenic non-coding RNA (lincRNA) within the SLC7A2 gene, and whose sequence was derived from a primate-specific retrotransposon that is responsible for the alteration of the expression of the RNA gene associated with the disease [Cartault et al., 2012]. In particular, the authors found that the single point mutation segregating with the disease was mapping in the first intron of the gene SLC7A2 in a degenerated transposable element, whose sequence was composed of the 3′ end of a primate-specific L1, L1PA8, and was embedded in a SINE element, AluSz. The mutation did not seem to determine dysregulation of SLC7A2 expression or splicing, but looking at the transcriptional activity associated with the repetitive elements the authors identified a unique expressed sequence tag (EST), a short sub-sequence of a cDNA sequence used to identify gene transcripts, that included the mutation. The analysis of this EST lead to the identification of two transcripts collinear with SLC7A2 and expressed in several different brain structures. The transcripts were further characterized as lincRNAs and were named SLC7A2-IT1A and SLC7A2-IT1B. Since the expression of SLC7A2-IT1A/B was more than eightfold in patient brain tissue compared with control, the authors decided to investigate the function of wild-type SLC7A2-IT1A/B and the effects of the reduced levels observed in the disease by performing in vitro knockdown in neuronal cells. Knockdown lead to the observation of increased active Caspase-3 protein levels, which resulted in a significant increase in neuronal apoptosis. Further in silico analyses of the lincRNA sequence lead the authors to propose several hypotheses to explain the phenotypic deleterious effect in the cell induced by the mutation in the TE-derived lincRNA. First, the mutation could induce the recruitment of the signal recognition particle protein SRP54 influencing the post-transcriptional regulation of the gene [Sauer-Eriksson and Hainzl, 2003]. Second, the mutated sequence could be a privileged target of piRNA leading to silencing of the transcript [Khurana and Theurkauf, 2010]. The occurrence of the predicted mechanisms and the neuronal apoptosis observed in vitro remains to be tested in the disease. Nevertheless, the differential expression of the lincRNA harboring the mutation suggests the intriguing hypothesis that intronic mutations, usually presumed to be non-functional, might induce gain of putative functional sites ultimately able to alter expression in cells, in the presence of repetitive elements such as the AluSZ/L1PA8 fragment of SLC7A2-IT1A lincRNA.

The involvement of Alus has been suggested in a study of the DiGeorge Syndrome, a disorder characterized by mental retardation and greatly increased risk to develop schizophrenia [Babcock et al., 2003]. The authors found that Alus can mediate chromosomal rearrangements leading to the disease-causing microdeletion observed on chromosome 22q11. The potential relevance of the link between TE-mediated functional mechanisms and CNVs is highlighted by the greatly increased risk that individuals with 22q11 microdeletion syndrome have for developing schizophrenia, although clearly homologous and non-homologous recombinations involving Alus are not restricted to neurological disorders. A list of TEs-related findings in psychiatric disorders is shown in Table I.

CONCLUSIONS AND FUTURE DIRECTIONS

Evidence is slowly accumulating that TEs can contribute to the genetic component of vulnerability to psychiatric disorders. So far, findings have been mostly anecdotal. Few studies provide sufficient data on specific TE expression or new insertions, making it difficult to estimate the extent to which TE might represent an etiopathogenic component in psychiatric disorders. However, despite our limited knowledge of mutation rates and disease/population-distribution of TEs, the concept that TE copies rarely mobilize within germline cells and that their activity is silenced in somatic tissue is breaking down. Identification of active retrotransposition of LINEs, Alus and SVAs in several different human brain regions, generated the provocative hypothesis that active mobilization and differential expression of TEs play a role in normal brain development (and adult physiology), and possibly in psychiatric disorders.

The most relevant evidence of the potential role of TEs comes from the well-documented impact of TEs in generating Indels (insertions/deletions) and CNVs (Copy Number Variants). Indels and CNVs are well-established risk factor for psychiatric disorders, from autism [Sebat et al., 2007; Marshall et al., 2008], to Developmental Disorders [Coe et al., 2012], schizophrenia [Consortium, 2008; Walsh et al., 2008] and Alzheimer’s Disease [Heinzen et al., 2010; Swaminathan et al., 2011; Guffanti et al., 2013]. A possible hypothesis is that TEs act as “re-shufflers” of the architecture of each individual genome. Our ability to detect and then prove the relationship between TEs and CNVs as important risk factors in psychiatric disorders will greatly improve with the data generated by whole-genome sequencing approaches.

Much is yet to be studied, such as the inheritance pattern of new insertions. Study designs that focus on single genes or on characterization of single TE types are a limitation of current TE research. With next-generation whole-genome approaches, we expect that our understanding of the relationship between TEs and psychiatric disorders will greatly improve. Initially, we need to catalogue TEs in diverse populations to confirm whether there is a role for RIPs in common psychiatric diseases. Then, it will be necessary to explore how these RIPs impact the disease phenotype. Current genome-scale technologies offer a unique opportunity to integrate next-generation DNA and RNA sequencing data to explain the potential effects of RIPs on gene expression. The combination of different platforms investigating the methylation and chromatin remodeling patterns across the human genome will eventually lead to defining the pathogenicity of RIPs individual profiles. We believe that understanding the mechanisms that finely dys-regulate the transcriptome in psychiatric disorders has the potential to also inform the development of new therapeutic strategies and the design of new drugs.

Schizophrenia and autism are by far the most studied psychiatric disorders. Yet, the search for genetic factors affecting risk for these disorders has produced results that are still somewhat conflicting and clearly not definitive, despite the major efforts of GWAS and NGS. Recent data highlights the potential value of investigating the overlap between GWAS hits and known RIPs: while the GWAS-associated SNPs are usually interpreted as proxies for nearby genes of potential pathogenic relevance, the possibility exists that the TE variants themselves may ultimately be among the etiopathogenic events. It is clear that the complex machinery by which genes are regulated may be of pivotal importance in the etiology of psychiatric disorders. This creates a broader context than only considering the functional effects of polymorphisms within a given “causal” gene and urges to focus on gene networks or pathways as we explore these genetic risk factors. Finally, the factors that promote gene regulation can be both environmental and genetic. TEs offer the intriguing possibility to combine both elements.