Neurons adapt their molecular equipment during development and activity-induced plasticity processes through transcriptional regulation (Cohen and Greenberg 2008), including the mostly presynaptic cell-adhesion molecules neurexins (Rozic-Kotliroff and Zisapel 2007; Iijima et al. 2011; Rozic et al. 2013). All three neurexin genes (Nrxn1-3) are presumably transcribed from two independent promoters that lead to the generation of extracellularly longer α-Nrxn and shorter β-Nrxn variants (Missler et al. 2011). In addition to these six major variants, extensive usage of five alternative splice sites in α-Nrxn, and two in β-Nrxn, may lead to more than 3000 isoforms (Rowen et al. 2002; Tabuchi and Südhof 2002), which were shown to be differentially distributed at the mRNA level (Ullrich et al. 1995). Nrxn play important roles in synaptic transmission and maturation of contacts (Missler et al. 2003; Graf et al. 2004), and some functions require trans-synaptic complex formation with neuroligin (Nlgn) (Ichtchenko et al. 1995; Boucard et al. 2005).
Studies in Nrxn knockout (KO) mice demonstrated decreased Ca2 +-dependent synaptic release at excitatory and inhibitory synapses (Missler et al. 2003; Kattenstroth et al. 2004), and Nlgn KO mice similarly revealed an impaired synaptic transmission (Varoqueaux et al. 2006; Poulopoulos et al. 2009). Interactions of Nrxn with Nlgn, but also with other extracellular binding partners as LRRTM2, cerebellin, or dystroglycan (Missler et al. 2011), depend strongly on the molecular variants involved, and a code has been proposed for their binding affinities (Boucard et al. 2005; Chih et al. 2006; Koehnke et al. 2010). In addition, some Nlgn variants are localized to subtypes of synapses, with Nlgn1 restricted to excitatory (Song et al. 1999), Nlgn2 and Nlgn4 to inhibitory synapses (Varoqueaux et al. 2004; Hoon et al. 2011), while Nlgn3 occurs at both types (Budreck and Scheiffele 2007). Such regulation of localization and usage can be functionally relevant because Nlgn2 variants, for example, associated with inhibitory synapses, preferentially bind to α-Nrxn and this molecule, in turn, induces GABAergic post-synaptic differentiation (Graf et al. 2004; Chih et al. 2006).
Despite the wealth of information on their physiological roles and their link to autism-spectrum disorders and schizophrenia (reviewed in Südhof 2008), very little is known about the transcriptional regulation of Nrxn and Nlgn genes. Here, we observe that Nrxn/Nlgn expression is altered in brains of mice lacking the methyl-CpG-binding protein 2 (MeCP2), a transcriptional regulator in brain (Chahrour et al. 2008) and the cause of human Rett syndrome (Amir et al. 1999). Our strategy to start the analysis of Nrxn/Nlgn gene regulation in MeCP2 KO mice was based on the following considerations: (i) MeCP2 regulates gene expression dependent on DNA methylation (Shahbazian et al. 2002a; Kriaucionis and Bird 2003; Chang et al. 2006; Yasui et al. 2007), and Nrxn/Nlgn genes contain CpG-rich sequences; (ii) Rett syndrome is an autism-spectrum disorders (Chahrour and Zoghbi 2007), and MeCP2 KO mice that recapitulate symptoms of patients (Guy et al. 2001; Shahbazian et al. 2002a) suffer from impaired synaptic transmission (Chao et al. 2007, 2010; Medrihan et al. 2008), similar to the Nrxn/Nlgn mouse models. In this study, we identify regulatory sequences in Nrxn and Nlgn genes, some but not all containing CpG-islands, which are able to change reporter gene expression in heterologous and neuronal cell lines. In addition, we characterize novel transcription start sites and genomic fragments in Nrxn1 and Nlgn2 that depend on a distinct methylation frequency for their transcriptional activity. Thus, our study provides the first experimental data on the regulation of Nrxn and Nlgn expression at the transcriptional level.
- Top of page
- Materials and methods
- Conflict of interest
- Supporting Information
We investigated promoter-like genomic sequences in Nrxn and Nlgn genes that are able to regulate transcriptional activity. Nrxn and Nlgn are pre- and post-synaptic molecules that affect important properties of synapses throughout the brain (Craig and Kang 2007; Missler et al. 2011). Nrxn and Nlgn represent candidates for transcriptional regulation because they are expressed in various isoforms and undergo extensive alternative splicing, leading to differential distribution and localization of distinct variants in subpopulations of neurons (Püschel and Betz 1995; Ullrich et al. 1995; Varoqueaux et al. 2004; Hoon et al. 2011) or even synapses (Song et al. 1999; Graf et al. 2004; Poulopoulos et al. 2009). Recent work has shown that expression levels, and possibly distribution, of specific alternatively spliced Nrxn variants can be adapted by synaptic activity (Iijima et al. 2011) and depolarization of neurons (Rozic-Kotliroff and Zisapel 2007). Such adaptation appears meaningful because Nrxn variants that are alternatively spliced at splice site #4 display different binding properties to their trans-synaptic partners Nlgn, LRRTM2, and cerebellin (Boucard et al. 2005; de Wit et al. 2009; Koehnke et al. 2010; Joo et al. 2011). In addition, splice variant expression may influence synaptic plasticity (Iijima et al. 2011). Similarly, regulated distribution of specific isoforms is also relevant for Nlgn because deletion of Nlgn2 demonstrated that it specifically affects a subset of inhibitory interneurons with distinct physiological properties in the neocortex (Gibson et al. 2009). While significant insight into the mechanisms of alternative splicing of Nrxn was provided by identifying the involved RNA-binding proteins Sam68 (Iijima et al. 2011), T-STAR (Ehrmann et al. 2013), and the heterogeneous nuclear ribonucleoprotein hnRNP L as the binding partner (Rozic et al. 2013), little progress was made in identifying regulatory genomic sequences in Nrxn or Nlgn genes since their initial description (Rowen et al. 2002; Tabuchi and Südhof 2002). Our investigation has now revealed three novel aspects of this process:
First, we observed that Nrxn1-3 and Nlgn1-3 expression levels are almost simultaneously altered in brains of mice lacking MeCP2 (Figs 1, 2). MeCP2 is a transcriptional regulator specific to the nervous system that can function as enhancer or repressor depending on the context (Adkins and Georgel 2011). Altered expression does not imply that MeCP2 directly regulates Nrxn or Nlgn. However, we provide experimental evidence that seem to favor this possibility at least for some instances because binding of MeCP2 to genomic DNA is frequently mediated by methylated CpG-dinucleotides clustered in CpG-islands at transcriptionally active regions (Chen et al. 2003; Chahrour et al. 2008; Skene et al. 2010): (i) We demonstrate that CpG-islands occur in some Nrxn and Nlgn promoter-like regions and can affect transcriptional activity in luciferase reporter gene assays (Figs 3, 6). This method is commonly used to quantify and compare transcriptional activity (Ikeda et al. 2002; Tsuritani et al. 2007; Chahrour et al. 2008; Suter et al. 2011), albeit additional effects because of translation rates cannot be excluded. The reporter gene assay was performed in HEK293 cells and in neuronal-like PC12 cells because Nrxn/Nlgn are neuron-specific and differentiation of PC12 with NGF leads to increased expression of MeCP2 (Impey et al. 2004; Chahrour et al. 2008; Mullenbrock et al. 2009). In addition, the strong transcriptional activation of CpG-rich sequences in the 3′-region of Nrxn2 (Fig. 3) is consistent with the role of 3′-flanking promoters in alternative splicing (Kornblihtt 2005). (ii) Nrxn1 and Nlgn2 promoter analyses demonstrate proof-of-concept that MeCP2 binds to these regions depending on the methylation frequency of CpG-islands (Figs 5, 8) and in vitro-methylation of exon1 CpG sequences increases transcriptional activity (Fig. 8). Despite these results, it has to be emphasized that Nrxn and Nlgn, even if some of the regulation is direct, are only two of hundreds of target genes of MeCP2 (Jordan et al. 2007; Chahrour et al. 2008; Ben-Shachar et al. 2009) and that the mode of action of MeCP2 includes different mechanisms (Adkins and Georgel 2011). Likewise, our data do not indicate that Nrxn or Nlgn gene regulation is solely responsible for the phenotype of MeCP2 KO mice (Medrihan et al. 2008; Chao et al. 2010) or even accounts for key pathophysiological mechanism in Rett patients.
Second, we identified a novel TSS in Nrxn1 that may be alternatively spliced in neuronal cells and could serve in the regulation of activity-dependent expression (Iijima et al. 2011; Rozic et al. 2013). The genomic sequence in this region is highly conserved (Figure S1) and the sequences preceding the first exon exerted a strong activation in our luciferase assay (Figs 4, 5). Interestingly, the transcriptional activity is abolished when fragments from the first intron or the second exon are included, similar to Nlgn2, in which the untranslated first exon also exerts a strong activation in the luciferase assay (Figs 7, 8). This arrangement fits the idea that MeCP2 has more activating influence on expression when bound to upstream, mostly untranslated sequences such as promoters and inhibits gene activity when bound to more downstream sequences (Chahrour et al. 2008). Recently, usage of such alternatively spliced TSS was confirmed for distribution of Fxyd1 in different brain regions, another target regulated by MeCP2 (Jordan et al. 2007; Banine et al. 2011). Thus, while our analysis of promoter-like sequences in Nrxn1 and Nlgn2 argues for an involvement of MeCP2 in both cases, the activity profiles may differ.
mRNA and expressed sequence tags (EST) databases suggest that for Nrxn1 the TSS in exon1 is used for abundant expression of Nrxn1α variants in brain, whereas transcripts starting with exon2 can be found in liver, a tissue with little or no Nrxn1 protein. This distribution is consistent with our finding that exon1 sequences, including only very few CpG, are activating and exon2 sequences, containing many methylated CpG-islands, are inhibiting translational activity (Figs 4, 5). This could indicate that in Nrxn1, the methylation of exon2 CpG-islands mediates the classical mode of transcriptional inhibition by MeCP2 complexed to Sin3/HDAC (Adkins and Georgel 2011). To increase expression, the transcription start would be switched to exon1 sequences that exert an increased activity (Figs 3, 4).
Although mRNA and EST databases suggest that expression of Nlgn2 in brain starts from exon1 but not from exon2, similar to Nrxn1, its actual regulation should be different because unlike Nrxn1 there are CpG-islands in both of the two-first Nlgn2 exons (Fig. 8). In support, we show that the two Nlgn2 CpG-rich regions contain different methylation frequency, and that they mediate opposite effects on transcriptional activity: exon2 sequences show little activation, whereas exon1 sequences are active (Fig. 8). Interestingly, we found that activity of exon1 sequences can be enhanced by additional methylation (Fig. 8), a challenging result because generally promoter methylation has been associated with gene silencing, for example, in immune cells (Brenet et al. 2011) but also in many studies on MeCP2 (Adkins and Georgel 2011). However, it was shown that MeCP2 binds to methylated CpG-islands more often when they are associated with promoter regions, and that this binding triggers activation of expression (Yasui et al. 2007; Chahrour et al. 2008), consistent with our data. Since electrical activity is known to influence the methylation pattern via chromatin remodeling (Martinowich et al. 2003), we hypothesize that availability of Nlgn2 for inhibitory synapses could be stimulated by increased neuronal/excitatory activity that leads to addition of methyl groups to exon1 CpGs, followed by association of MeCP2 and increased translational activation (Chahrour et al. 2008). Co-activation of other transcription factors such as Creb1 (Chahrour et al. 2008) could then generate the molecular machinery needed for newly formed or enlarged inhibitory contacts.
Third, we characterized a bidirectional head-to-head configuration in the promoter region of Nlgn2 and 18100RIK on opposite genomic strands which is highly conserved (Figure S2). Molecular dissection of the genomic sequences that activate transcription in these opposite directions revealed two crucial regions (Fig. 7). Since these regions are 218 bp apart, we believe that expression of Nlgn2 and 18100RIK can be simultaneously regulated as shown for similar promoter arrangements in mice (Kornblihtt 2005) and humans (Wei et al. 2011), and could be used for concomitant activation of both genes. Since the expression pattern and function of 18100RIK is completely unknown, and no homologs exist in the databases, it remains unclear at present if the 18100RIK mRNA is translated (Figure S3) or used as a regulatory RNA.
Work on neuronal promoters has revealed that transcriptional regulation in brain extends beyond the action of MeCP2. For example, it was shown that expression of synaptic vesicle proteins is down-regulated by RE1-silencing transcription factor (REST)-binding to methylated promoter sequences which lead to chromatin rearrangement by histone acetylation (Shahbazian et al. 2002b; Ekici et al. 2008). Along this line, a binding site for repressor element 1 (RE1/NRSE) has previously been identified in Nrxn3 (Bruce et al. 2004). The activity of REST may also influence the expression of Nlgn2 because REST leads to an activity-dependent developmental switch from excitatory to inhibitory synapses (Varoqueaux et al. 2004; Yeo et al. 2009). These results suggest that different regulatory pathways may intersect as expression patterns of Nrxn genes are activity dependent (Shapiro-Reznik et al. 2012) and expression of REST is adjusted by MeCP2 (Ballas et al. 2005). Thus, expression of Nrxn and Nlgn is likely not affected by MeCP2 alone but is influenced by a larger group of transcriptional regulators which tune the levels of these functionally important molecules during development, plasticity and maintenance of synapses. The quest for the mechanisms involved has just started.