The aim to discover abnormally expressed genes in MeCP2-deficient mice is two-fold: to reveal direct targets of MeCP2 modulation and to identify components of neuronal pathways and circuits that are deregulated in affected brain regions. The ultimate goals are to understand the pathophysiologic processes that lead to morbidity in RTT and to design rational therapies to prevent or revert the functional abnormalities. To reduce microarray data noise caused by cellular heterogeneity of tissues, we focused on the cerebellum, a part of the brain not previously studied. We compared expression profiles in two different MeCP2-deficient mouse strains that differ on the molecular level, the Mecp2tm1.1Bird (B-allele) and Mecp2tm1.1Jae (J-allele). Furthermore, we compared expression profiles at three postnatal ages that span different developmental stages in the cerebellum. At two weeks, granule cells are still mitotic and migrating to their final position in the internal granule layer. At four and eight weeks, granule cells are post-mitotic and have begun forming synapses with neurons in the cerebellum and other brain regions.
Consistent with previous studies on forebrain and hippocampus samples by Tudor and colleagues [23
], we did not observe major expression changes of a large number of genes in the cerebellum. MeCP2-deficiency does not significantly alter the cerebellar transcriptome on a global scale. This is remarkable given that the vast majority of cells in the cerebellum are granule cells that had shown morphological abnormalities in J-allele mice [30
]. Any significant overexpression of genes specifically in granule cells should have been detectable, as it would not be averaged out by cell type heterogeneity. Our results should caution investigators who engage in cell-type specific profiling studies. Also, no large (> 3 fold) changes were detected for any single gene. The changes in cerebellum transcript levels tended to be small (1.5 – 2 fold), but increases greatly outnumbered decreases. Overall, a number of genes with increased transcript levels greater than expected by chance were observed only in B-allele mice at 8 wk. There are several possible explanations for these findings.
First, the mutations are different. While B-allele mice are null at the Mecp2
locus, the J-allele mutation consists of an in-frame deletion of exon 3 that produces a stable mutant transcript. The putative mutant protein, which lacks the MBD domain, but has an intact nuclear localization signal, transcription repression domain and C-terminal region, may have maintained certain functions. Some evidence has been reported for a truncated MeCP2 protein in the J allele mice. Luikenhuis et al. [54
], saw diffuse nuclear staining with a MeCP2 antibody and speculated that this probably arose from the detection of the truncated protein. This interpretation is in line with the original report of the J-allele mice [30
], where Western blot showed a smaller band that was interpreted to represent the truncated protein, even though the wild-type samples had a protein of apparently the same size, but at a lower level of intensity. The possibility cannot be excluded, however, that the diffuse nuclear staining was caused by a cross-reacting protein that does not localize to DAPI-positive foci. Using tissue microarrays to study the mosaic composition of Mecp2
-expressing and non-expressing cells in female mice, Braunschweig et al. [55
], reported a background immunoreactivity with a C-terminal antibody in male J-allele mice that was apparently not present in mice with the B-allele. These results would argue in favor of the presence of an internally deleted mutant protein present in the J-allele brain, although the intensity was rather low.
Second, the mice were on different strain backgrounds: B-allele mice on C57BL/6 and J-allele mice on a mixed background, mostly BALB/cJ. Our hierarchical clustering data clearly demonstrated a strong effect of strain background (Figure ). Random mixed strain background noise should be overcome, to some extent, by our comparing only wild-type and mutant sib-pairs and by including a sufficiently large number of sib-pairs (16 for the B-allele and 13 for the J allele).
Third, 80% of the B-allele mice at 8 wk were symptomatic, while fewer of the J-allele mice were showing symptoms. When the disease process in the brain is further advanced, the expression changes may be caused by pathologic events far downstream of direct MeCP2 action.
Although the global level of gene expression is not significantly altered, that does not preclude the possibility that some of the genes may have changed expression levels that are biologically relevant. Therefore, we focused on whether differentially expressed genes showed a similar change in other brain regions and in a different set of mice (biological validation), which is a more relevant evaluation than "statistical" validation. We chose four genes, Irak1, Fxyd1, Reln, and Gtl2, whose expression levels were increased in the most consistent fashion. By ChIP analyses, we show that MeCP2 binds to the promoters of Fxyd1 and Reln and to a differentially methylated region of the imprinted gene Gtl2 in normal mouse brain, suggesting that these genes may be direct targets of MeCP2.
The interleukin-1 receptor-associated kinase 1 gene revealed the largest significant expression increase of any gene, in cerebellum (2.0-fold on microarray and 3.5-fold on qRT-PCR assays) and in forebrain (1.9-fold), but only in B-allele mice. Irak1 is located ~ 3 kb downstream of Mecp2 in the same transcriptional orientation (UCSC Genome browser, Feb 2006 assembly). This suggests that a negative regulator of Irak1 could have been lost with the B-allele deletion, or that the altered local chromatin conformation leads to increased Irak1 expression. The less likely possibility that MeCP2 directly regulates the expression of Irak1 would assume that the J-allele has maintained a negative regulatory function, since Irak1 expression is not increased in J-allele mutants. This allele-specific difference raises the concern that the increased Irak1 expression may be responsible for some dysregulated genes in the B-allele mice that are not dysregulated in the J-allele mice. To identify bona fide MeCP2 targets we, therefore, focused on genes whose promoters bind MeCP2 in ChIP assays. Nevertheless, any studies done with the B-allele mice, at any level, from behavior to histology, should be considered in the light of possible downstream effects of Irak1 overexpression.
IRAK1 belongs to the ser/thr protein kinase family, pelle
subfamily. The protein is comprised of a death domain in the N-terminus, a central serine/threonine kinase domain, and a C-terminal serine/threonine rich region. Once phosphorylated, IRAK1 recruits the adapter protein PELI1 and causes activation of Toll-like-receptor-mediated intracellular signaling pathways. Three different splice variants in humans and mice differ in their relative abundance in brain, and differential splicing of IRAK1 may correlate with the aging process [56
]. Our qRT-PCR assay amplified all three splice forms. Given the postulated role of MeCP2 in alternative splicing [11
], it might be of interest to evaluate the relative abundance of the different Irak1
splice forms in Mecp2
-mutant B-allele brain.
In wild-type mice, Fxyd1
expression increases drastically from 2 wk to 8 wk in the cerebellum (our microarray data). In Mecp2
mutant mice, microarray and qRT-PCR studies revealed a significant increase of Fxyd1
transcripts at the 8 wk time point in B-allele mutants when compared to wild-type littermates. ChIP analyses confirmed the binding of MeCP2 to the Fxyd1
promoter. We, thus, have identified Fxyd1
as a validated target of MeCP2 binding. In contrast to a recent report by Deng et al. [53
], however, Fxyd1
transcripts were not increased in the mutant forebrain where expression levels are normally very low. Attempting to replicate the reported studies in humans, we have used qRT-PCR to measure FYXD1
expression in frontal cortex from male and female individuals with MECP2
mutations and unaffected controls. We found no evidence for increased expression in the mutant samples. Therefore, our data in humans and mice do not support the hypothesis that MeCP2 is responsible for keeping FYXD1
expression low in the majority of cells of forebrain. A small increase of expression in a neuronal subtype could have gone undetected, but the large increases reported by Deng et al. [53
] in RTT frontal cortex are not confirmed by our results.
The protein product, FXYD1, also called phospholemman (PLM), is a small molecule with a single transmembrane domain and a member of the FXYD family of small ion transport regulators. FXYD proteins regulate Na-K-ATPase activity in a tissue- and isoform-specific way. Each FXYD gene is expressed in one or more specific tissues [57
]. FXYD1 is a major membrane phosphoprotein in heart and muscle. In brain, FXYD1 is predominantly found in the cerebellum, specifically in the molecular layer, in Purkinje neurons, and in axons traversing the granule cell layer, as well as in the choroid plexus [58
]. The protein forms a helical structure and inserts into lipid membranes in a trans
-bilayer fashion [59
]. FXYD1 tetramers form ion channels selective for K+, Cl-, and taurine that are physically associated with the Na-K-ATPase [60
]. Recent studies revealed that FXYD1 also inhibits the cardiac Na+/Ca2+ exchanger (NCX1) [61
]. Overexpression of FXYD1 in adult cardiac myocytes acutely alters contractility as a function of extracellular Ca+ concentration, and Fxyd1
knock-out mice have cardiac hypertrophy [62
]. Alteration of Fxyd1
expression in the heart of Mecp2
-deficient mice and humans could conceivably contribute to cardiac dysfunction and sudden death.
Another member of the FXYD family, FXYD4/CHIF (corticosteroid hormone induced factor) was initially isolated as a glucocorticoid response gene [63
]. Interestingly, two putative MeCP2 target genes, Sgk1
are also glucocorticoid-inducible and were found to be upregulated in 8 wk old B-allele mice studied previously [25
]. Furthermore, in the Mecp2308
mouse model of MeCP2 deficiency, circulating corticosteroid levels were increased in response to stress, and the gene for corticotropin-releasing hormone (Crh
) was overexpressed in brain regions, leading to the conclusion that the mutant mice have an enhanced corticosteroid response to stress via increased CRH signaling [64
]. Future studies should evaluate whether FXYD1
, like FXYD4
, is also corticosteroid inducible.
Our identification of Reln
as a primary target of MeCP2 is supported by the following observations: Reln
transcript levels were increased in microarray and qRT-PCR experiments comparing cerebella from a total of 12 wild-type and 9 mutant mice. And the expression change was seen at the earliest time point (2 wk) when all mice were asymptomatic. No expression change was found in hippocampus. ChIP assay on whole brain documented that MeCP2 binds to the Reln
promoter in normal mice. The proximal promoter of Reln
is highly CpG-rich and methylated in whole brain extracts [65
]. Multiple hypermethylation sites at the Reln
promoter can be induced by injection of L-methionine. And in a methionine-induced model of schizophrenia, MeCP2 was shown to bind to the Reln
Reelin is a large extracellular protein that is produced by discrete populations of cells in the brain. It acts through the extracellular milieu on neighboring target cells. Until recently, the primary function of reelin was thought to involve regulation of neuronal migration during fetal brain development, by reelin providing an architectonic signal for the guidance of migrating neurons [68
]. Reelin-expressing cells include the Cajal-Retzius (CR) cells in the cortical marginal zone, that are the first neurons to express Mecp2
in development, and cerebellar granule cells. MeCP2-deficient mice have a thinner cortex and more densely packed neurons that could result from migratory defects.
expression continues in the adult, in particular in a subset of GABAergic neurons of the forebrain, as well as in the olfactory bulb [70
]. Reelin can modulate synaptic plasticity and enhance long-term potentiation (LTP) in adult hippocampal cultures [71
]. The signaling pathway that reelin acts on involves two high-affinity binding receptors, the very low density lipoprotein receptor (VLDLR) and the apolipoprotein E receptor 2 (ApoER2) [72
]. Binding of reelin to VLDLR and ApoER2 triggers phosphorylation of the cytoplasmic adaptor protein disabled 1 (Dab1) by Src
family kinases, in particular Fyn
. This series of reactions is required for cortical layer formation and for hippocampal dendrite development [73
]. In addition, cyclin-dependent kinase 5-dependent signals are required for the function of reelin not only in neuronal migration but also in synaptic transmission [74
]. The reelin signaling pathway is also involved in modulating learning and behavior [75
]. Reelin can regulate NMDA-type glutamate receptor activity and potentiate calcium influx through NMDA receptors in neuronal cultures [75
]. Sinagra et al. [76
] documented a reelin-controlled change in subunit composition of NMDA glutamate receptors during maturation. Reelin signaling through ApoE receptors plays an essential role in synaptic plasticity and function in the adult brain [77
Interestingly, BDNF, brain-derived neurotrophic factor, a primary target of MeCP2 [78
] regulates reelin production in cortical neurons. Bdnf
-/- mice have elevated reelin levels in Cajal-Retzius cells. Overexpression of BDNF produces brain abnormalities similar to the phenotype of reeler (rl)
mutant mice. This phenotype is preserved in slice cultures of hippocampus [80
]. Furthermore, in in vitro
co-culture systems, exogenous reelin caused dispersal of chain-migrating interneuron precursors in the olfactory bulb, suggesting that overexpression of reelin may affect neuronal migration [81
We hypothesize that Reln overexpression is responsible for part of the phenotype of MeCP2-deficient mice, and that this phenotype is caused by abnormal neuronal migration in the developing brain of MeCP2-deficient mice and by dysregulation of reelin signaling pathways, thus disturbing synaptic function, in postnatal brain. To test this hypothesis, we have initiated genetic interaction studies. We aim to reduce the level of Reln expression in the Mecp2 mutant background by crossing rl+/- heterozygous males to Mecp2+/- female mice. We have preliminary evidence that the onset of morbid phenotypes is delayed and lifespan prolonged in some of the Mecp2Y/- ;Rln+/- double-mutant mice, when compared to their Mecp2Y/- ;Rln+/+ littermates.
Gtl2 (gene trap locus 2)
By microarray analysis, we identified Gtl2
, called MEG3
(Maternally-expressed gene 3) in humans, as a significantly increased transcript in mutant cerebellum samples. qRT-PCR confirmed increased expression, although not significant due to large biological variation in different litters, in cerebellum and forebrain. Gtl2
encodes a maternally expressed imprinted non-translated RNA of unknown function. Gtl2
is expressed in most tissues, but most highly in brain. It lies within a cluster of imprinted genes on chromosome 12 that is conserved on distal 14q in human. Alternatively-spliced Gtl2
transcripts extend to include intron-encoded C/D box snoRNAs [82
] and microRNAs [83
]. Thus, Gtl2
might function as a host gene for these small RNAs. In our qRT-PCR assays, the increase in Gtl2
expression in cerebellum was less striking, but Gtl2
expression was consistently increased in the forebrain of Mecp2
Two previous studies have identified Gtl2
as a possible target of MeCP2. An expression microarray study on whole brain of one B-allele mouse revealed a 1.9-fold increase in Gtl2
]. Independently, by using a variant of differential display technology and qRT-PCR confirmation, Kriaucionis and colleagues [27
] detected increased Gtl2
expression in whole brains of B-allele mice. Since the increase was significant only at late symptomatic stages, the authors suggested that it may represent a secondary consequence of the disease pathology. In contrast, our data show a similar increase in Gtl2
expression at 2 wk and at 8 wk, albeit detected by two different cDNA clones on the array. Gtl2
has a CpG-rich promoter region that is unmethylated on the maternal allele and becomes hypermethylated on the paternal allele after fertilization [85
]. By ChIP assay, we found that MeCP2 binds in vivo
to this region with binding sites extending from 5' of exon 1 into intron 1, raising the possibility that it may play a role in silencing of the paternal allele. Alternatively, MeCP2 may bind to methylated CpG sites on the active maternal chromosome and thus regulate Gtl2
repression in certain brain regions or cell types. Future studies of allele-specific expression levels will address these possibilities.
Considering other previously reported candidate MeCP2 target genes that are represented on our microarrays: Only Snrp70
, reported to be increased by qRT-PCR in whole brain from late symptomatic B-allele mice (1.44 fold, p = 0.04)[27
], showed increased expression (1.5-fold) in cerebellum of 8 wk B-allele mice in our study. None of the following genes had significant expression changes in cerebellum: Uqcrc1
]. This is not surprising given that some of these genes may be primarily expressed in other brain regions and/or only in certain neuronal sub-types. Or transcript levels may be variable, when expression is modulated by activity state as for Bdnf
]. Our experience with biological validation and hierarchical clustering of independent samples taught us about the mitigating effects of different sibship identities. We need to consider that the different mothers are Mecp2
+/- heterozygotes, having the genotype of females with RTT. Although not obviously symptomatic yet, their variable X-inactivation status could influence expression profiles in their offspring.