The ability of ATRX to remodel chromatin 16 suggests that ATRX can regulate gene expression. To identify possible gene targets of the ATRX protein in the developing mouse brain, we used the previously described Atrx^Foxg1Cre mice that lack ATRX in the forebrain 19. In this model system, Atrx deletion is achieved by crossing Atrx^loxP "floxed" mice to mice that express cyclization recombinase (Cre) under the control of the forebrain-specific forkhead box G1 (Foxg1) promoter 20. We performed microarray analysis to compare the expression profiles of the Atrx^Foxg1Cre and control telencephalon at embryonic day 13.5 (E13.5) (n = 3 pairs) using an Affymetrix mouse genome expression array representing approximately 39,000 transcripts 21. Only probe sets showing a significant difference (p < 0.05) were included in all subsequent studies. By setting a threshold of 1.5 fold change we identified 202 disregulated probesets, and at a threshold of 2 fold change we identified only 22 altered probe sets. Approximately two-thirds of the probe sets demonstrating altered expression were upregulated (Additional file 1A, B).

We next compared gene expression patterns in control and Atrx-null forebrain tissue at postnatal day 0.5 (P0.5) (n = 4 pairs). At a threshold of 1.5 fold change, we identified 304 probe sets and at a threshold of 2 fold change, we identified 57 probe sets showing altered transcript levels. When we compared the microarray results at E13.5 and P0.5 we identified 14 common probe sets that were upregulated and 13 that were downregulated more than 1.5 fold, and one increased and three decreased more than 2 fold (Additional file 1A).

We used GeneSpring to identify significantly overrepresented Gene Ontology (GO) categories in the Atrx-null mouse forebrain. Several statistically and biologically significant categories of upregulated genes were related to the immune response. This could be an indirect response to the increased apoptosis that characterizes the Atrx-null forebrain at E13.5 in the developing cortex and to a lesser extent at P0.5 in the hippocampus 19. In particular, categories and genes involved in phagocytotic clearing of apoptotic cells, such as complement activation 22, were enriched at both E13.5 and P0.5. Several genes involved in cell adhesion processes were upregulated at P0.5 and, consistent with the abnormal forebrain development described in the Atrx-null forebrain19, genes involved in neurogenesis and nervous system development were downregulated at both timepoints (Additional file 2).

Five of the most downregulated transcripts identified in the microarray analysis were unidentified cDNA clones (Affymetrix IDs 1436320_at, 1448057_at, 1443755_at, 1429730_at and 1453066_at; Accessions [GenBank:W45978], [GenBank:BI202412], [GenBank:BE457721], [GenBank:AK007409] and [GenBank:BI320076] respectively). To further investigate these probe sets, their NCBI nucleotide sequences were used for a Basic Local Alignment Search Tool nucleotide (BLASTn) search of the nr database. The expressed sequence tag (EST) [GenBank:W45978] has similarity to Mus musculus Dhrsxy ([GenBank:NM_001033326], score = 120, E value 5e-24). The EST [GenBank:BI202412] displayed similarity to several unidentified mouse cDNA clones. Interestingly, a BLAST-like Alignment Tool (BLAT) search of this clone showed similarity to intron 1 of mouse Dhrsxy and it could represent an unknown splice variant of Dhrsxy. The EST [GenBank:BE457721] is annotated as similar to human Arse and a BLASTn search revealed high similarity to Rattus norvegicus Arse ([GenBank:NM_001047885], score 197, E value 6e-28). BLASTn of [GenBank:AK007409] showed high similarity to Asmtl in cow ([GenBank:BT02626], score = 248, E value = 6e-62) as well as dog, human, the putative rat Asmtl, and numerous other species. The EST [GenBank:BI320076] displayed no significant hits to any sequences by either BLASTn or BLAT. Interestingly, while Dhrsxy, Arse and Asmtl do not display an obvious connection, they do share a common link in that they are all pseudoautosomal genes in eutherians. In addition, the microarray data showed decreased expression of Cd99, Shox2 and Csf2ra, that also lie within the pseudoautosomal region in most eutherians. Therefore, while GO analysis identified a subset of downregulated genes involved in brain development at both timepoints, a more in depth analysis of downregulated targets revealed that many are orthologs of PAR1 genes residing on the tip of the X and Y chromosomes in most placental mammals. Overall, our transcriptional screen identified six of these genes, constituting approximately half of all PAR1 orthologs discovered in the mouse genome so far. The more intriguing aspect of this finding is that in the mouse, these genes no longer reside within the PAR1 region but have translocated to autosomes (Figure 1). It also identified two potential novel PAR1 orthologs–Arse and Asmtl–not previously identified in the mouse genome. At E13.5, these genes represent 6 of the 15 most downregulated transcripts identified by microarray analysis. Strikingly, they constitute 4 of the top 5 most downregulated genes in the microarray performed on P0.5 forebrain tissue (Arse and Shox2 were not significantly decreased in the microarray at P0.5) (Table 1, Additional file 1B). These results suggest that ATRX normally participates in the transcriptional activation of these genes during both the proliferative (E13.5) and more differentiative (P0.5) stages of forebrain development.

To validate the microarray results, we performed real-time quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) analysis of Dhrsxy, Cd99, Csf2ra, Shox2 and also the putative new orthologs of Asmtl and Arse in Atrx-null and control E13.5 and P0.5 forebrain (n = 3 at each time point). Since Arse and Asmtl have not yet been identified in the mouse, we sequenced the PCR products to ensure they corresponded to the transcripts identified on the microarray, and not to other contaminating sequences. The qRT-PCR results confirmed that five of the six genes exhibit decreased expression in the Atrx-null forebrain at E13.5, and that these genes remain downregulated at P0.5 (Figure 2A). In addition, analysis at P17 demonstrated decreased expression of ancestral PAR genes at this later time point as well (Figure 2A). One exception was Shox2 which exhibited highly variable expression differences between the Atrx-null and control tissue at E13.5, P0.5 and P17, ranging from a 170 fold decrease to a 90 fold increase (Figure 2B). Therefore, while the expression of Shox2 is clearly affected by the loss of ATRX protein, the outcome on expression levels appears to be highly variable and does not validate the consistent downregulation observed by microarray analysis.

Our discovery that the expression of several ancestral PAR1 genes is controlled by ATRX throughout the early developmental period of the mouse brain reveals an unexpected association between the levels of ATRX protein and the expression of these ancestral PAR1 genes.

In humans, a cluster of ARS genes are located approximately 115 kilobases centromeric to the PAR1 region on the X chromosome, but still possess the ability to escape XCI in females 23. Located outside the PAR1, these genes do not have an identical homolog on the Y chromosome but have pseudogenes, and in the evolutionary past it is believed that they were true pseudoautosomal genes with identical copies on both the X and Y chromosome 24.

A multiple alignment of amino acid sequences of [GenBank:BE45772] suggested that it is a fragment of the full length ARSE protein, aligning in the middle of the approximately 600 amino acid ARSE protein of multiple other species (Additional file 3). The putative mouse ARSE is 65% identical to rat and 47% identical to human.

Phylogenetic analysis demonstrated that the mouse ARSE sequence clusters with near certainty with the rat ARSE, however, this putative ARSE clustered within the ARSD proteins, not ARSE as expected (Figure 3). Therefore, we propose that we have identified a member of the PAR1 ARS family but at this time cannot determine the exact identity and will refer to this sequence as Arsd/e. We note that the long branch-length between the rodent ARS sequences and the remaining ARSD clade may be an artifact due to the short mouse sequence and its high similarity to the rat sequence, which has undergone seemingly accelerated evolutionary change.

Comparisons to available mouse Ars gene family members shows that [GenBank:BE457721] is more similar to Arse genes in rat than to other mouse arylsulfatase family members (Additional file 4), suggesting that we have identified Arse. This data, combined with our ability to specifically amplify this transcript from mouse brain cDNA and also from a commercially available E15 cDNA library (data not shown), indicates that we have likely identified the mouse homologue of a previously unidentified mouse Ars gene rather then a gene fragment from a known mouse family member.

To further confirm the identity of [GenBank:BE457721], we assessed the outcome of ATRX depletion on Arsd/e expression by RNA interference in the Neuro-2a cultured neuroblastoma cell line. Small interfering RNAs (siRNAs) were used to transiently deplete ATRX, as was done previously 25. Cells transfected with a non-specific siRNA or no siRNA ("Mock") were used as controls. At 72 hours following siRNA transfection, we monitored the effectiveness of ATRX depletion by indirect immunofluorescence using an ATRX-specific antibody (H300) and qRT-PCR analysis of Atrx expression levels using primers that simultaneously amplify both the full length isoform and the reported truncated isoform 26. In the siATRX-treated samples, approximately 95% of cells were negative for ATRX (Figure 4A) and Atrx transcript levels were depleted by approximately 5 fold (Figure 4B). We then used qRT-PCR to determine the outcome of ATRX silencing on the expression level of the Arsd/e. Similar to the results obtained in the Atrx-null forebrain, the expression of Arsd/e was decreased two fold (Figure 4B). These findings support that we have identified the mouse Arsd/e and confirm the regulation of this ancestral PAR gene by ATRX, and that this outcome on gene expression can be recapitulated in two different systems: in vivo in the ATRX-null developing forebrain and in vitro in ATRX-depleted cultured neuronal cells.

[GenBank:AK007409] is the RIKEN cDNA 1810009N02 gene and contains a musculoaponeurotic fibrosarcoma (MAF) domain. A multiple sequence alignment of amino acid sequences was used to further determine the identity of [GenBank:AK007409] (Additional file 5). [GenBank:AK007409] aligns to the N terminus of ASMTL from multiple other species. The N terminal portion of ASMTL also contains a MAF domain. Human ASMTL was generated by a fusion of a duplicated acetylserotonin O-methyltransferase (ASMT) with the bacterial maf gene 27. While [GenBank:AK007409] contains a MAF domain, it lacks the ASMT domain. However, this is similar to the putative rat ASMTL (Accession [GenBank:NP_001099385]) which also lacks the ASMT domain. The putative mouse ASMTL is 54% identical to rat, and 51% identical to the human protein.

In contrast to ARSE, ASMTL has fewer discernable high-similarity full-length orthologs, and its evolution appears tied to the pseudoautosomal region 27. Therefore fewer sequences were available for analysis. Figure 5 shows the inferred phylogeny of the ASMTL family, with the primate branches collapsed for clarity. With fairly high bootstrap support, the tree mirrors the known branching of the placental mammals, marsupials, monotremes, birds, amphibians, and fish. The mouse sequence displays the only anomalous placement in the tree, clustering well outside the mammalian clade. Both the placement and the branch-length of the mouse sequence indicate that it is of considerably evolutionarily derived character compared to the putative ancestor, and it appears to have followed an evolutionary path quite different from its paralogs. The lack of the ASMT domain in the mouse sequence may also be responsible for the placement of the mouse sequence in the tree. As with ARSE, some of this divergence may be due to the availability of a partial mouse sequence, but the sequence remains quite unique, nonetheless.

Mice conditionally deficient for ATRX in the forebrain were generated by crossing Atrx^loxP females with heterozygous Foxg1Cre male mice, as previously described 19. Pregnant females were sacrificed at E13.5, embryos were recovered and yolk sac DNA was genotyped by PCR using the primers 17F, 18R and neo^ras described previously 19. For newborns (P0.5) and juveniles (P17), pups were sacrificed and tail DNA was used for genotyping as previously described 19.

Total forebrain RNA (10 μg) was isolated from three E13.5 and four P0.5 pairs of littermate-matched ATRX-null and control embryos using the RNeasy Mini kit (Qiagen). cRNA was generated and hybridized to an Affymetrix Mouse Genome 430 2.0 Array at the London Regional genomics Center (London, Canada). For the analysis at E13.5, RNA from two forebrains was pooled for each array. Probe signal intensities were generated using GCOS1.4 (Affymetrix Inc., Santa Clara, CA) using default values for the Statistical Expression algorithm parameters and a Target Signal of 150 for all probe sets and a Normalization Value of 1. Gene level data was generated using the RMA preprocessor in GeneSpring GX 7.3.1 (Agilent Technologies Inc., Palo Alto, CA). Data were then transformed (measurements less than 0.01 set to 0.01), normalized per chip to the 50^th percentile, and per gene to control samples. Probe sets representing Atrx transcripts were removed (10 sets). Remaining probe sets were filtered by fold change of either ≥1.5 or 2 between control and Atrx-null samples, and by confidence level of P < 0.05. Heatmaps were generated using the GeneSpring hierarchical clustering gene tree function. Significantly overrepresented GO categories were determined using GeneSpring: at E13.5 and P0.5, probe sets were filtered by 1.5 fold change, P < 0.05 and categorized as either up or downregulated. Where there were multiple probe sets for a gene, duplicates were removed. P < 0.001 was used as the significance cutoff.

Total RNA was isolated using the RNeasy Mini kit (QIAGEN). First-strand cDNA was synthesized from 3 μg of total RNA using the SuperScript™ II Reverse Transcriptase kit (Invitrogen) with 25 mM dNTPs (GE Healthcare), 1 μL porcine RNAguard (GE Healthcare) and 3 μL random primers (GE Healthcare). PCR reactions were performed on a Chromo4 Continuous Fluorescence Detector in the presence of iQ™ SYBR Green Supermix and recorded using the Opticon Monitor 3 software (Bio-Rad Laboratories, Inc.). Samples were amplified as follows: 95°C for 10 seconds, annealed for 20 seconds, 72°C for 30 seconds (See Additional file 6 for primer sequences and annealing temperatures). After amplification a melting curve was generated, and samples were run on a 1.5% agarose gel (75 V for 1 h) to visualize amplicon purity. Standard curves were generated for each primer pair using three fold serial dilutions of control cDNA. Primer efficiency was calculated as E = [10^(-1/slope)-1]*100, where a desirable slope is -3.32 and r² > 0.99. Samples were normalized to β-actin expression and relative gene expression levels were calculated using GeneX software (Bio-Rad Laboratories, Inc.).

For Arsd/e and Asmtl, the PCR products were gel extracted using the QIAquick Gel Extraction Kit (QIAGEN) according to the manufacturer's instructions and sequenced at the DNA Sequencing Facility at Robarts Research Institute (London, Canada).

Probeset sequences were obtained from the Netaffx website http://www.affymetrix.com/analysis/netaffx and used for BLASTn searches http://www.ncbi.nlm.nih.gov/BLAST. For calculation of interspecies similarity, sequences were obtained from NCBI RefSeq http://www.ncbi.nlm.nih.gov/RefSeq or Ensemble http://www.ensembl.org where RefSeq sequences were not available, and pairwise comparisons made using Jalview 46.

For generation of trees and sequence alignments, human ARSE (SwissProt P51690, RefSeq NP_000038) and human ASMTL (SwissProt O95671, RefSeq NP_004183) were used as seeds and the GenBank NR database was searched for high-similarity, full-length orthologs and paralogs. Fifty-nine ARSE and twenty-two ASMTL sequences met or exceeded the similarity cutoff, with resultant species spanning the metazoa from anemone and urchin to a diverse set of vertebrates. Sequences were aligned using T-Coffee 5.56 47 using default parameters. Alignments were manually adjusted via inspection prior to further analysis. Approximate maximum-likelihood trees were built using PHYML 2.4.5 48 using the WAG model of protein evolution 49 and a seven-category Gamma-plus-invariant model of rate heterogeneity. All rate parameters were estimated from the data. One hundred bootstrap replicates were performed to assess support for the inferred tree topology. All trees are presented as midpoint-rooted phylograms. Since the given mouse sequences were quite short compared to the full protein length, two sets of trees were built for each family to assess if the mouse sequences were long enough to definitively support their taxonomic clustering. One set utilized a "trimmed" alignment where all alignment columns outside the mouse sequence domain were removed. The trees produced with this trimmed alignment were compared with the set of trees produced from the alignment of the mouse sequences to their respective full-length proteins. For both ARSD/E and ASMTL, very little difference was observed between full-length and trimmed-alignment trees. The trimmed alignments tended to exaggerate sequence divergence and modestly lower bootstrap support levels. Overall topology did not appear significantly different, however, and the text references the full-length sequence phylogeny exclusively.

Neuro-2a cells were grown at 37°C with 5% CO₂ in EMEM supplemented with 10% fetal bovine serum (Sigma-Aldrich). For siRNA treatment, 1.5 × 10⁴ cells were plated in a plastic six well dish (Corning Incorporated) on glass coverslips and allowed to grow to 15% confluency (approximately 24 hours). Cultures were transfected using Lipofectamine 2000 (Invitrogen) with 8 nM siATRX (Dharmacon), a non-specific control siRNA (Sigma-Aldrich), or with no siRNA ("Mock") according to the manufacturers' instructions (for siRNA sequences refer to 25). Total RNA was extracted from cells after 72 hours, cDNA was generated and qPCR analysis performed as described above. Alternatively, cells were processed for immunofluorescence staining as described below.

Neuro-2a cells were fixed using 3:1 methanol:ethanol, incubated for 1 h with the primary antibody (H300 anti-ATRX, 1:100 dilution; Santa Cruz) followed by the secondary antibody (goat-anti rabbit Alexa 594, 1:1500 dilution; Molecular Probes), then counterstained with 4',6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich) for 5 min. Coverslips were mounted with Vectashield (Vector Laboratories), Z-stack images were captured using a Leica DMI6000b inverted microscope and Openlab software (v5.0, Improvision) and processed using Volocity software (v4.0, Improvision); deconvolution was performed using iterative restoration set with a confidence limit of 95%.