Human Molecular Genetics Advance Access originally published online on January 30, 2006
Human Molecular Genetics 2006 15(6):853-869; doi:10.1093/hmg/ddl004
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Genomic and functional profiling of duplicated chromosome 15 cell lines reveal regulatory alterations in UBE3A-associated ubiquitin–proteasome pathway processes
1Department of Pathology and 2Department of Biochemistry and Molecular Medicine, University of California, Davis School of Medicine, Sacramento, CA 95817, USA, 3Department of Human Genetics, University of California–Los Angeles, Los Angeles, CA, USA and 4Center for Pediatric Research, Nemours Biomedical Research, Alfred I. duPont Hospital for Children, Wilmington, DE, USA
* To whom correspondence should be addressed at: MIND Institute Wet Lab, 2805 50th St., No. 2420, Sacramento, CA 95817, USA. Tel: +1 9167030362; Fax: +1 9167030367; Email: jpgregg{at}ucdavis.edu
Received November 21, 2005; Accepted January 25, 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Autism is a complex neurodevelopmental disorder having both genetic and epigenetic etiological elements. Isodicentric chromosome 15 (Idic15), characterized by duplications of the multi-disorder critical region of 15q11–q14, is a relatively common cytogenetic event. When the duplication involves maternally derived content, this abnormality is strongly correlated with autism disorder. However, the mechanistic links between Idic15 and autism are ill-defined. To gain insight into the potential role of these duplications, we performed a comprehensive, genomics-based characterization of an in vitro model system consisting of lymphoblast cell lines derived from individuals with both autism and Idic15. Array-based comparative genomic hybridization using commercial single nucleotide polymorphism arrays was conducted and found to be capable of sub-classifying Idic15 samples by virtue of the lengths of the duplicated chromosomal region. In further analysis, whole-genome expression profiling revealed that 112 transcripts were significantly dysregulated in samples harboring duplications. Paramount among changing genes was ubiquitin protein ligase E3A (UBE3A; 15q11–q13), which was found to be nearly 1.5–2.0-fold up-regulated in duplicated samples at both the RNA and protein levels. Other key findings from gene expression analysis included two down-regulated genes, APP and SUMO1, with well-characterized roles in the process of apoptosis. We further demonstrate in this lymphoblast model that the gene-dosage directed increases in UBE3A levels can lead to dysregulation of the process of ubiquitination in response to genotoxic insult. This study provides insight into the direct and indirect effects of copy number gains in chromosome 15 and provides a framework for the study of these effects in neuronal systems.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Childhood autism (MIM 209850 [OMIM] ) is characterized as a pervasive neurodevelopmental disorder with onset before 30–36 months of age, resulting in impairment in social interaction and communication and manifestation of repetitive stereotypic behavior. The estimated prevalence of this disorder is one to two per 1000 births (1
). The involvement of genetic factors for autism has been well established in family and twin studies. Additionally, there is an emerging evidence that the presence of chromosomal abnormalities may play a significant role in the etiology of autism. This is based on the high frequency of large chromosomal anomalies identified in patients with autism spectrum disorders (ASD), which range from 1.7 to 4.8% depending on the population screened and the method of detection applied (2,3). These have included unbalanced translocations, inversions, rings and interstitial deletions and duplications. Although most of these rearrangements are unique to a single patient or family, a number of recurrent chromosome abnormalities have been reported in unrelated patients, including duplication of 15q, deletions of 18q, Xp, 2q and sex chromosome aneuploidies: 47,XYY and 45,X (4,5).
In these studies, chromosomal abnormalities have been identified by using traditional cytogenetics and fluorescence in situ hybridization (FISH), which allow detection of whole-chromosome losses or gains or large chromosomal deletions and duplications. Over the last several years, however, smaller chromosomal aberrations in the range of 5 kb to 10 Mb, which have traditionally been undetectable with classic cytogenetics, have been detected in a number of diseases. New genomic technologies have been developed, which can scan the genome at a very high resolution (6) in order to analyze DNA copy number. Large insert clones that contain mapped genomic segments have been utilized for copy number alteration (CNA) detection either by directly spotting onto glass slides or by amplifying these regions and then spotting them (6). Similarly, oligonucleotides representing specific spacing and regions in the genome have been developed and utilized for copy number analysis (7). Recently, single nucleotide polymorphism (SNP) arrays, initially utilized for loss of heterozygosity in cancer (8–10), have been applied for use in copy number analysis (11).
In this study, we applied the SNP array-based comparative genomic hybridization (CGH) approach to a known duplication of the chromosome 15q11–q13 region that is associated with autism. Duplications of chromosome 15q11–q13 account for 
1–3% of autism and are the second most prevalent genomic aberration associated with autism next to fragile X syndrome. These duplications range from 8 to 12 Mb and, when associated with autism, are derived from the maternal chromosome (12,13). These duplications can occur either through generation of a supernumerary isodicentric chromosome 15 (idic15) or as an interstitial duplication. With proof of principle that the SNP array approach is robust in accurately identifying and determining the breakpoints of this region, this technology could be extended to analyze a large cohort of children with autism.
The importance of this genomic approach in autism is that there is a significant co-morbidity with mental retardation (MR). Recently, Lugtenberg et al. (14) applied array-based CGH to a cohort of mentally retarded patients who were considered to be cytogenetically normal based on classic cytogenetics and FISH. These studies demonstrated that 10% of these MR cases carried small chromosomal abnormalities ranging from 100 kb to 12 Mb. This suggests that in autism there may be a large number of undetected chromosomal aberrations.
Beyond identifying these aberrant regions, the ability to determine their importance and association with disease has remained elusive. In this study, we took a systems biology approach to analyze aberrations within a commonly duplicated region on the proximal long arm of chromosome 15. Expression microarrays were utilized to examine expression of genes in the duplicated region as well as potential down-stream genes using samples from patients with supernumerary isodicentric chromosomes. Analysis of differentially expressed genes demonstrated an overabundance of genes in pathways associated with ubiquitination and apoptosis, suggesting that these pathways may be altered in Idic15 subjects. Of prominence was UBE3A, located within the 15q11–q13 contig area, which was found to be overexpressed at both the transcriptional and protein levels. Furthermore, functional data presented here suggest that this overexpression of UBE3A is associated with dysregulated ubiquitination triggered by genotoxic stress and altered apoptosis in patients with Idic15. Therefore, we demonstrate a systems biology approach to analyze copy number changes, demonstrating a link between DNA copy, transcriptional activity and protein production, as well as identifying functional repercussions.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Affymetrix mapping arrays identify Idic15 amplifications in samples previously characterized by CGH
To identify chromosomal amplifications using commercially available SNP arrays and a nascent array-based CGH approach, we profiled 10 samples with known chromosome 15 duplications and nine control cell lines with Affymetrix arrays. The samples with duplications have been previously found to have variable duplications of chromosome 15q11–q13, a region spanning
10 Mb. The Affymetrix mapping 10K array includes 32 SNPs in this region separated by a median distance of 191 kb. To evaluate whether the SNP arrays would be able to differentiate between control and duplicated samples, we hybridized genomic DNA from nine samples from typically developing children to the arrays and analyzed using the Chromosomal Copy Number Tool (CCNT) for regions of significant consecutive copy change, which we defined to be above a –log10(meta P-value) of 5. None of the samples from the typically developing children showed meta P-value scores above 5 for the target region (Fig. 1A). Conversely, when the same region was examined in the samples with known duplications, chromosomal copy number increase for the 15q11–q13 region was identified in all samples. The meta P-values for these samples were found to be at the theoretical maximum of 20 (P-value=1x10–20) for the entire length of the duplication, a region containing 22–31 SNPs (Fig. 1B). Additionally, the data from the CCNT discriminated the characteristic breakpoints of each sample, making it possible to classify them by the length of their duplications (Fig. 2). In summary, these data support the utilization of the SNP arrays and CCNT as a valid approach for the identification of copy number changes in chromosomal regions.
|
|
Chromosomal copy number analysis using Affymetrix SNP arrays segregate duplicated chromosome 15 samples by breakpoint position
Copy number data obtained from the 32 SNPs, located in the 15q11–q13 region, provided the ability to resolve the point at which chromosomal amplification terminates in each sample and is detailed in Figure 2. The amplification termination points correlate well with previous data obtained through traditional CGH and two-color array-based CGH analysis conducted by Wang et al. (15
). Four samples, 99-70, 99-30b, 02-4 and 02-07 were classified as [ptel:BP4:BP5:ptel] duplications and showed evidence of increased copy number extending to the physical position of 31.1 Mb on chromosome 15 (Probe set ID: SNP_A-1512326). These samples were easily distinguished from other samples that carried [ptel:BP3:BP3:ptel] duplications. Copy number analysis indicates that amplification extends for 2.6 Mb distal to the BP3 breakpoint spanning an additional two to five contiguous SNPs on the array. Array 01–19, containing a [ptel:BP3:BP3:ptel] duplication, showed a deviation in apparent dosage compared with other samples with cytogenetically similar duplications. Data from this study indicate that the centromeric portion of the duplication 21.4 Mb was over 1.9 Mb (one to two SNP probe sets) beyond the other arrays. Additionally, this array showed an extended duplication range when compared with other [ptel:BP3:BP3:ptel] samples. Subsequent analysis of this sample with another oligonucleotide-based array platform corroborated this result (data not shown).
Gene expression analysis identifies distinct expression patterns between duplicated chromosome 15 and normal samples
To explore the transcriptome variation imparted by duplication of chromosome 15, we performed microarray gene expression analysis on Affymetrix HG-U133 Plus 2.0 arrays with approximately 38 000 well-characterized genes. Analysis of the data was performed to identify differential gene expression patterns between the two groups and to evaluate the correlation of changes in copy number to those in expression.
An initial filtering of more than 54 000 probe sets was conducted using criteria that eliminated genes that had fewer than three observations of centered signal values above an absolute value of 0.7 for all arrays. This reduction enabled the elimination of a large portion of probe sets with invariant expression profiles. The remaining 5534 probe sets were subjected to univariate significance testing with Benjamini–Hochberg false discovery rate (FDR) correction. This stringent analysis found 112 probe sets below a defined P-value threshold of 0.05. This resultant gene list contained several probe sets mapping to four unique genes located in the 15q11–q13 region of the genome. Additionally, we found that these genes were all up-regulated in samples with genomic amplifications when compared with controls. These genes and their corresponding fold change increase were UBE3A (1.92-, 1.6- and 1.6-fold, three different probe sets), NIPA1 (1.99-fold), NIPA2 (1.58-fold) and HERC2 (1.89-fold). As this probe list contained only four unique known genes out of a total of 80 located within the duplicated area of chromosome 15, the majority of the changes in the Epstein–Barr virus (EBV)-transformed lymphocytes are not due to increased gene dosage of chromosome 15, but rather represent potential down-stream effects of this duplication. Interestingly, we also detected up-regulation of a poorly annotated probe set, 1559343_at, which maps to GenBank AF400500, a sequence identified as containing untranslated, alternatively spliced exons of the SNRPN transcript (16). A search of the reference database RNAdb (research.imb.uq.edu.au/rnadb/) identifies the accession number as containing an exon that is antisense to UBE3A. A detailed investigation of the genomic positioning of this Affymetrix probe set found it to be located on the sense strand within the 3'-UTR of the antisense transcriptional unit of UBE3A. These genes were further evaluated using hierarchical clustering, which provides insight into potential co-regulation patterns (Fig. 3). This list of genes categorized the controls and duplicated 15 into their respective classes with the hierarchical clustering.
|
Gene proximity analysis
On the basis of the finding that a subset of the statistically relevant genes identified via t-test was located within the 15q region, the data were further analyzed to evaluate whether the four differentially expressed genes located on chromosome 15 occurred over a statistically relevant stretch of the genome. To accomplish this, we conducted a gene proximity analysis using dChip software (17
). Focussing on the 112 genes identified in the univariate analysis, the software evaluated the statistical likelihood that these genes would occur over a given genome stretch by chance. At a P-value threshold of 2x10–5, we found only three significant regions containing more than two probe sets as determined by normalized rank-distance (outlined in blue in Fig. 4). One of the identified regions contained the four genes in the critical chromosome 15 region. The other regions were identified as conserved histone clusters on chromosome 6p21 and 1q21. Referring back to the copy number data, we found no evidence of chromosomal change for those regions, indicating that the observed regional gene expression change may be merely due to co-regulation of the clusters.
|
Gene ontology comparison analysis
The observation that the majority of significantly differentially expressed genes may be a product of down-stream effects of increased gene dosage on chromosome 15 suggests that the observed changes may reflect disruption of specific pathways. In an attempt to uncover similarities among these genes, expression data and significance analysis results for the 5534 probe sets were analyzed using the ErmineJ Gene Score Resampling application. The software analyzed gene significance scores between the Idic15 and control groups and identified 15 significant (P-value<0.001) gene ontology (GO) biological process terms, shown in Table 1. It is noteworthy to mention that several functional categories associated with macromolecular catabolic processes were identified, including the ubiquitin-dependent protein catabolism category. Not surprisingly, many of the genes identified by t-test are members of these enriched functional categories, most notably UBE3A and HERC2, both mapping within the duplicated region of chromosome 15, as well as SUMO1, a ubiquitin-like molecule located on chromosome 2.
|
TaqMan confirmation
Quantitative TaqMan RT-PCR was utilized to verify that the genes identified by the array analysis were differentially expressed. Four genes were selected for this confirmation based upon being implicated in playing vital roles in cellular homeostasis, such as in the regulation of ubiquitin-mediated proteolysis and/or apoptosis. All of the genes analyzed by TaqMan demonstrated significant differential expression and showed similar directional change when compared with microarray data (Table 2).
|
Idic15-bearing cells display altered regulation of ubiquitination in response to genotoxic stress
The findings of genomic duplication of the UBE3A locus and its approximate 1.5–2-fold overexpression (log expression difference range of 0.684–0.9870) suggest that cells harboring the Idic15 chromosome possess a deregulated ubiquitin-mediated proteasome pathway (UPP). In turn, this could lead to altered levels of proteins that are degraded via UBE3A-mediated ubiquitination. Interestingly, UBE3A exhibits complex regulation of expression in that, although many cell types (e.g. fibroblasts, lymphoblasts) exhibit biallelic expression, UBE3A is transcribed from the maternal allele in restricted subsets of cells, such as the Purkinje cells and hippocampal neurons (18
). The basis for specific repression of paternal gene suppression is thought to be due to the expression of an antisense UBE3A transcript driven by the Prader–Willi syndrome-imprinting center of the paternal allele (19).
As a first step in investigating this, we wanted to confirm that the increase in UBE3A message resulted in a similar increase in protein levels in the cells carrying the chromosome 15 duplications. Immunoblot analysis of UBE3A expression in a subset of our unaffected and Idic15 cell lines revealed a 1.6–1.9-fold up-regulation of UBE3A protein relative to the controls (Fig. 5A). On the basis of this, we next wanted to test the hypothesis that this would result in higher levels of ubiquitination of its target substrates. Because p53 is a prominent UBE3A substrate (20,21), we used it as a surrogate marker of UBE3A ubiquitin–protein ligase activity. However, steady-state p53 protein levels were found to be uniform among the panel of cell lines when normalized to ß-actin expression (Fig. 5A). To investigate the possibility that stress-responsive modulation of p53 levels was altered, cells were treated with doxorubicin, a DNA-damaging chemotherapeutic agent known to mediate cell cycle inhibition and apoptosis in a p53-dependent manner (22). Exposure of both control and Idic15 cell lines to doxorubicin resulted in a marked elevation of p53 protein levels (Fig. 5B), but a significant difference in the magnitude of the increase between the two groups of cells was not observed. Therefore, we were curious as to whether there were differences in basal or stress-induced levels of ubiquitination. To better visualize mono- and poly-ubiquitinated forms of p53, cells were pre-treated with the proteasome inhibitor MG-132 to prevent degradation of ubiquitinated proteins (Fig. 5B). As expected, inhibition of the proteasome led to stabilization and increased levels of the native p53 protein (Fig. 5B, middle panel, MG-132 lanes). Ubiquitinated p53 species were visualized by increasing the exposure time of the same blot (Fig. 5B, upper panel). In control cell lines, the levels of ubiquitination, particularly of poly-ubiquitinated species, declined markedly in response to doxorubicin treatment. In contrast, this trend was not universally obvious in the duplicated 15 cell lines; in one case, p53 ubiquitination increased substantially by doxorubicin. In summary, these results demonstrate that UBE3A expression is indeed overexpressed in the cell lines carrying the chromosome 15 duplications and raise the possibility that p53 ubiquitination triggered by genotoxic stress is dysregulated.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
We performed whole-genome, array-based CGH analysis of a panel of samples derived from children with autism harboring known duplications of chromosome 15. Expanding on the considerable amount of data previously published on characterization of chromosome 15q11–q13 duplications using traditional CGH approaches, as well as dual-channel array-CGH data (15
), we verified the efficacy of Affymetrix technology's ability to accurately detect the dosage and the applicability of this approach for the characterization of genomic duplications (Fig. 1A). In our study of 10 samples with known duplications, we found convincing evidence of duplication in the critical 15q11–13 region in all samples analyzed (Fig. 2). Moreover, the significance data generated by the CCNT software for these samples were particularly compelling with meta P-values (the metric that expresses the likelihood of consecutive dosage variations) reaching the theoretical maximum of 20 in –log10 space. Conversely, data derived from our panel of cell lines from typically developing individuals showed no instances of significant copy number or meta P-value changes, indicating that this array-based CGH approach has a high level of accuracy and can easily discriminate between both groups with good resolution.
Evaluation of mapping 10K-based array-CGH applicability
Chromosomal gain and loss analysis using mapping 10K SNP arrays was conducted in this directed study of duplications on chromosome 15. With a mean inter-SNP distance of 244 kb (10), this approach enabled CNA discovery at a reasonably high level of resolution not typically available using traditional methods of cytogenetics. Although the length of duplication studied was in a range detectable by other approaches, we are confident, based on successful mapping of duplication breakpoints, that it would be possible to detect much smaller CNAs using the strategies employed in this study. The mapping 10K array and CCNT algorithm together provided sufficient resolution for the detection of duplication endpoints and the variation in length characteristic in the two sub-classes of duplicated samples included in this study. For the distal breakpoints, the SNP data correlate with the data described by Wang et al. (15) using a BAC array approach. Although the nucleotide position reported here is different from that by Wang et al., the mapping of these BACs has been updated in Santa Cruz database and correlates with the data reported for the SNP arrays. In contrast, the start point of duplicated regions may have been underestimated in some samples due to the paucity of probes in the most proximal region of 15q. However, in the case of sample 01-19, the evidence of a more distal start point was reproduced using another oligonuceotide high-resolution array approach (data not shown). These effects may be less important when conducting similar experiments with higher density SNP arrays, such as the 100K and 500K sets. It is also assumed that the ability to resolve increasingly smaller duplications will be possible with the implementation of SNP arrays of increased density, namely the mapping 100K and 500K sets.
Copy number and significance values generated by the whole-genome sampling analysis (WGSA) algorithm are derived from a reference data set containing over 110 normal samples and quantitative dosage–response experiments. The copy number data generated from our typically developing children control samples showed minimal alteration compared with our experiment group, indicating that the control pool accurately approximates a representative control sample. The contrast observed between the two groups possibly indicates that a minimum number of control samples are required to be tested in parallel with experimental groups, thereby increasing the overall feasibility of screening large numbers of samples at time.
Gene expression analysis
To determine whether the effect of increased gene dosage on 15q11.2–13.3 would eventuate in distinctive gene expression patterns for the duplicated 15 group versus the control group, we analyzed representative samples for differential gene expression. Results from an FDR-corrected t-test found 112 probe sets to have P-values below the targeted significance level. Examination of the genes within this culling revealed four unique genes located within the critical region, all of which were up-regulated in our duplicated 15 group. A gene proximity analysis found this group of genes to be one of only three stretches of the genome to be significantly dysregulated at P=2x10–5 or lower (Fig. 4). This finding, coupled with the fact that all of the genes in this region were up-regulated, leads us to suggest a regional gene-dosage effect resulting in increased transcriptional activity. In addition, these data suggest that a minority of the genes in this region (5%) are directly affected by copy number, potentially suggesting that a majority of the genes are regulated through alternative mechanisms. It is important to note that expression profiling was conducted using cultured lymphoblasts, not neuronal cells, and therefore may be subject to unique tissue-specific regulation. Additionally, there are several imprinted and neuro-related genes located in the duplicated 15q region, which are known to undergo alternative regulation. This may explain why we were unable to find evidence for increased expression levels for many of the traditional positional candidates located on chromosome 15q11.2–13.3, including GABRB3 (23), GABRG3 (24) and ATP10C (25) for autism. Although the use of blood genomics to serve as proxy for neurological systems in the identification of neurological disorders is not well documented, it is assumed and many of the same genes or pathways may be implicated. In a study of a pair of twins discordant with respect to bipolar disorder, gene expression profiling identified down-regulation of expression of genes related to endoplasmic reticulum stress response in both affected twins. Subsequent analysis of DNA from the identified genes and pathways found a polymorphism (–116C
G) in the promoter region of XBP1 (26). This study, along with our data, suggests that gene expression of peripheral blood can be used to identify genes or pathways that have importance in CNS phenotypes.
Notably, we found UBE3A, a ubiquitin E3 ligase that is deficient in patients with Angelman syndrome (MIM 105830
[OMIM]
), to be the most significantly (P=0.00057) differentially expressed gene overall, showing an approximate 2-fold increase in the duplicated group. UBE3A has been found to exhibit exclusive maternal expression in neurons (27) and preferential biallelic expression in lymphocytes (28). In certain regions of the brain, expression regulation of UBE3A has been found to be regulated through an antisense transcript, first proposed by Rougeulle et al. (29), in part through the transcriptional repressor MeCP2 (29,30). Further study of the effect of MeCP2 on the expression of UBE3A has demonstrated that the former protein helps to facilitate a closed chromatin structure to repress the expression of an antisense sequence (UBE3A-ATS) on the maternal chromosome (31). In typically functioning neurons, this mechanism leads to the silencing of the paternal allele, leaving only monoallelic expression of the maternal allele. Only trace levels of the this antisense transcript were detected in blood in a study conducted by Runte et al. (32). From the data in this experiment, it seems likely that UBE3A expression responds to dosage effects due to genomic copy number gain. However, the differential expression seen in the probe set 1559343_at may indicate that UBE3A-ATS levels are also elevated in duplicated 15 samples relative to controls; however, it seems to have had neglible effect on UBE3A signal attenuation.
Copy number-dependent overexpression of UBE3A has been predicted to have down-stream effects on the stability of its protein targets (33), as well as potentially causing anomalous neurological development (28). Recently, Jiang et al. (34) have proposed a model describing a modest role for UBE3A in autism, emphasizing that a gene subject to ubiquitination of UBE3A may be directly implicated in this disorder. In addition to validation of our microarray findings by qRT–PCR, a comparable up-regulation of UBE3A protein levels was verified (Fig. 5). The most notable interaction involving this E3 ubiquitin ligase is with p53, in which the latter protein is targeted for ubiquitin-mediated degradation (20). On the basis of this interaction, it has been assumed that cells harboring duplications of UBE3A could potentially exhibit decreased apoptotic potential (28) or a propensity to accumulate DNA lesions. To discern potential aberrant regulation of the UPP in the duplicated 15 samples, we monitored the p53 response to treatment with the chemotherapeutic agent doxorubicin to induce genotoxic stress (i.e. stabilization of topoisomerase II complexes). Ubiquitinated species of p53 were stabilized by co-treatment with the proteasome inhibitor MG-132. Although both steady-state and doxorubicin-induced levels of p53 were similar in both sets of cell lines, increased levels of poly-ubiquitinated p53 in response to doxorubicin were detected in a cell line overexpressing UBE3A. This suggests that the integrity of regulation of p53 levels and function may not be conserved in the cell lines carrying chromosome 15q duplications. In contrast, decreased p53 ubiquitination in the control cell lines under the same conditions is consistent with the enhanced stability of p53 required to mediate its normal functions of transcriptional regulation, growth arrest and/or apoptosis. This could impact cellular homeostasis when present in the appropriate setting (tissue, biological process, stimulus, etc.). Taken together, we believe our findings, which point to a potential link between genomic copy number and up-regulation of UBE3A, add to the growing literature implicating this gene's involvement in autism.
In light of UBE3A overexpression and potential deregulation of the ubiquitin–proteasome pathway, it is noteworthy that expression of small ubiquitin-like modifier 1 (SUMO-1) is down-regulated in the Idic15 cell lines. Indeed, deregulated sumoylation has been implicated in the pathogenesis of neurological diseases (35,36). Similar to ubiquitination, sumoylation is a reversible, post-translational modification occurring at lysine residues of specific target proteins and is mediated by the cooperative actions of the SUMO conjugating (E2) enzyme Ubc9 and SUMO ligase (E3) at specific consensus motifs (37). This brings up an important point regarding our observations. Despite having a statistically significant change, SUMO-1 expression exhibited a down-regulation of 1.61-fold (microarray gene expression profiling, Table 3) and a modest 1.19-fold (qRT–PCR, Fig. 3). If this difference is preserved through translation, this would correlate with a 60 or 20% reduction in SUMO-1 protein levels, respectively. Whether such a change would have an influence on cellular function is worth consideration, but this could conceivably impact the levels of sumoylation as a consequence of decreasing the level of ‘SUMO-charged’ Ubc9 that interacts with its substrates. Depending upon the protein, a modest alteration in the levels of its sumoylated species (which in many cases are at low levels in vivo) could produce an aberration in the process it participates in. Additionally, it is quite possible that the two SUMO paralogs (i.e. SUMO-2/3) cannot serve as substitutes for SUMO-1 in vivo because it has been demonstrated that the three SUMO paralogs are functionally distinct based upon their dynamics of utilization and localization during the cell cycle (38).
|
Rather than marking proteins for degradation, SUMO modification functions to regulate cellular localization, transcriptional activity, genome maintenance and signal transduction (37
). In some cases, SUMO promotes protein stability through competition with ubiquitin at binding sites. Accordingly, protein sumoylation influences a variety of cellular processes, including cell cycle progression and apoptosis.
Without experimental evidence of the critical SUMO substrates in the model system used in this study, we can only speculate upon the significance of SUMO-1 down-regulation and assume this would lead to a generalized reduction in the modified level of its targets. This could have varied effects on the regulation of apoptosis signaling. For instance, decreased SUMO-1 levels could sensitize cells to death receptor-mediated apoptosis (39) and heighten p53 levels and responses via decreased mdm2 levels resulting from its diminished sumoylation and consequent increased self-ubiquitination and degradation (40).
Alternatively, survival would be favored by reduced p53 transcriptional activity as a result of decreased K386 sumoylation (41) and reduced relocalization of p53 to nuclear bodies (42), as well as by enhanced NF-
B signaling resulting from decreased competition for ubiquitination of IBß at K21 (43). It is also possible that the reduced levels of SUMO-1 provide a partial basis for the expression signature of the Idic15 cells as a result of altered activity (i.e. de-repression in the absence of sumoylation) of one or more transcription factors (44). Finally, the fact that sumoylation figures prominently in regulating DNA replication, recombination and repair processes suggests that our findings of overexpression of several nucleosome core histone genes might be related to dysregulation of one or more of these pathways as a result of SUMO-1 down-regulation.
Further examination of the data found that amyloid precursor protein (APP) was significantly down-regulated in Idic15 samples, nearly 10-fold with a corrected P-value of 0.01178. Aside from its implication in Alzheimer's (MIM no. 104300
[OMIM]
) disease and Down syndrome (no. 190685), expression of APP has been found to play a pivotal role in the induction of neuronal apoptosis via activation of a ‘ubiquitin-like’ neddylation pathway, as reviewed by Chen (45). It has been found that disruption of this pathway hinders cellular transition out of S-phase (46), which, in neurons, may lead to an impaired ability to undergo apoptosis. Additionally, it has been found that APP knockout mice exhibit reactive gliosis and impaired motor function (47), although follow-up knockout studies were unable to confirm the essential function of APP in neurological development (48). Dramatic reductions of mRNA levels of APP may provide further support for a pro-survival tendency in these cells.
Interestingly, ATP10A, which lies immediately distal to UBE3A, showed no evidence of expression dysregulation in the cell lines, despite having been previously found to show preferential maternal expression (49). On the basis of its maternal expression, ATP10A has been suggested to be involved in chromosome 15-associated autism (50). The lack of concordant expression patterns between UBE3A and ATP10A could indicate that the latter gene is subject to a specific regulation beyond that of the SNRPN-IC and is therefore less sensitive to increased maternal copy number in lymphoblasts. Additionally, expression data from this experiment seem to be in agreement with linkage studies interrogating the ATP10A region that did not find statistical evidence of the involvement of this gene in autism (25).
The pattern of anti-apoptotic elements seen in this studymay provide a possible mechanism for chromosome 15-related autism cases. It has been previously postulated that increased gray and white matter seen at particular development stages in autism cases could be due to excessive sparing of neurons or axon formations (51). If the findings of this study using a lymphoblast model were confirmed in a neuronal model, a broad study of apoptosis regulation in early development would be well merited. A focussed interaction network of the findings of this study (Fig. 6) reveals how the observed transcriptional and protein variations may impact cell stability.
|
Overall, this study demonstrates that an SNP array approach can identify relatively small chromosomal gains. Further, this segmental duplication of proximal chromosome 15q leads to an increased expression of genes directly in the contig, with evidence for dysregulation of numerous down-stream genes scattered throughout the genome. The specific overexpression of UBE3A protein may be a primary mediator of the down-stream effects, as the data presented here indicate augmented ubiquitination triggered by genotoxic stress in coincidence with UBE3A overexpression. In turn, this might translate into an inappropriate reduction in apoptosis due to decreased sensitivity to apoptotic stimuli and tolerance to DNA damage. Taken together, these may represent contributing factors to the etiology of ASD.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
Cell lines
Human EBV-transformed lymphocytes from patients with duplication of chromosome 15 were obtained after informed parental consent with the use of protocols approved by the institutional review boards of the University of California, Los Angeles, Alfred I. duPont Hospital for Children, and University of California, Davis. These cell lines have been verified to have duplications in the 15q11–q13 region by Southern blot, FISH and genomic array CGH and have been described previously (15
). Cytogenetically normal cell lines derived from non-affected siblings in families with autism were obtained from similar EBV-derived cell lines from the autism genetic resource exchange (AGRE). This internationally recognized central repository and family registry for autism was created by Cure Autism Now (CAN) and the Human Biological Data Interchange for autism studies. The family recruitment and diagnostic assessments for the cytogenetically normal cell lines were based on the procedures established by CAN and are fully described by Geschwind et al. (52). A complete description of each sample used in this study is available through AGRE at http://www.agre.org. The selection criteria for control cell lines used in this study required that the individual was categorized as non-affected; for the purposes of this paper, these samples will be termed as typically developing or control samples. The control cell lines used in this study include HI 1276, HI 1715, HI 0083, HI 1871, HI 1556, HI 1085, HI 1688, HI 1539 and HI 0238.
Cell culture
Cell lines were cultured in RPMI-1640, 15% fetal bovine serum, 50 U/ml penicillin, 50 µg/µl streptomycin and 2 mM L-glutamine in a 5% CO2-humidified tissue culture incubator. For DNA harvest, a total of 2x106 cells were seeded out in a T-75 flask in 10 ml of medium and harvested after 2 days of growth. For RNA, a total of 2x106 cells were seeded out in a T-75 flask in 10 ml of medium and harvested when cells reached a density of 8x105–1.5x106 cells/ml. For each line, RNA and protein were extracted during the logarithmic growth phase, in order to minimize experimental variability that might lead to differences in gene and protein expression. This procedure ensures that RNA and protein are collected in a reproducible manner and can be compared between cell lines. Previous studies have shown that this procedure produces RNA transcript profiles that are highly correlative between replicate cell culture experiments (57).
DNA isolation
For each sample, a total of 4x106 cells were collected for DNA isolation using the QIAamp DNA Blood Mini Kit (Qiagen, Valencia, CA, USA). A cell pellet was obtained after centrifugation. The cell pellet was re-suspended in PBS and then treated with Proteinase K with lysis buffer added subsequently to ensure efficient lysis and digestion of proteins. Following these steps, ethanol was incorporated in the reaction volume to allow the binding of the DNA to the silica gel membrane of the QIAamp spin column after centrifugation. The pellet was then washed several times in order to purify the DNA before the elution of the DNA in buffer AE. The purity and quality of the DNA were analyzed by fiberoptic spectrophotometry using the Nanodrop ND-1000 and by gel electrophoresis on a 1% agarose gel. The genomic DNA isolated and utilized for the genomic analysis had an A260/A280 absorbance ratio of 1.8; gel electrophoresis revealed a major band at 10–20 kb. The DNA samples were then processed and hybridized to GeneChip® Human 10K Mapping arrays (Affymetrix Inc., Santa Clara, CA, USA).
RNA isolation
Total RNA was isolated immediately upon cell harvest, as described by the TRIzol Reagent (Invitrogen, Carlsbad, CA, USA) RNA isolation procedure, and further purified using the RNeasy kit (Qiagen, Chatsworth, CA, USA). RNA quality was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Foster City, CA, USA) and quantified by fiberoptic spectrophotometry using the Nanodrop ND-1000 (Nanodrop Inc., Wilmington, DE, USA). RNA yielding both an A260/A280 absorbance ratio greater than 2.0 and a 28s/18s rRNA ratio equal to or exceeding 1.8 was utilized. The RNA samples were then processed and hybridized to Human Genome GeneChip human U133 2.0 Plus arrays (Affymetrix Inc.).
SNP mapping array processing
The Affymetrix Mapping 10K GeneChip is a single array that enables genotyping of approximately 11 500 SNPs with the use of a single PCR primer. This mapping array provides substantial coverage with an average genomic distance between SNPs of only 210 kb. Starting with 250 ng of genomic DNA, each sample was digested with an XbaI restriction enzyme. A process-negative control was included at the beginning of the assay to assess the presence of contamination. The digestion step generated single-stranded overhangs specific to the restriction enzyme, which was then targeted by a specific adaptor in a ligation step. PCR was then performed with a single primer that recognized the adaptor-ligated fragments of genomic DNA. The PCR reaction is optimized to amplify fragments from 250 to 1000 bp size. This size range and the absence of contamination were checked by gel electrophoresis. PCR products were purified using the MinElute 96 UF PCR Purification kit (Qiagen) and quantified by fiberoptic spectrophotometry using the Nanodrop ND-1000. The samples underwent fragmentation to reduce the average size to <180 bp, which was verified by gel electrophoresis. They were then labeled with a proprietary biotin-ddNTP and hybridized for 16 h on the array. The arrays were washed and stained on a Fluidics Station 450 and were scanned on a GeneChip Scanner 3000. Each SNP is interrogated by 10 probe quartets of 25mers, where each probe quartet is comprised a perfect match and a mismatch probe for each allele.
Gene expression array processing
Purified total RNA (10 µg) was used for cDNA synthesis followed by in vitro transcription to incorporate biotin labels and subsequent hybridization to Human Genome U133 Plus 2.0 (Affymetrix) according to manufacturer's protocol. The arrays were washed and stained on a Fluidics Station 450 and were scanned on a GeneChip Scanner 3000. The HG-U133 Plus 2.0 array represents 38 500 well-characterized genes.
Microarray mapping and copy number analysis
Data obtained from the SNP mapping arrays were processed and analyzed by the GeneChip DNA Analysis Software (GDAS) v3.0. GDAS, using a model-based algorithm, and generated SNP allele calls and intensity measurements for all the arrays analyzed. These data were subsequently processed by the Affymetrix CCNT to generate copy number estimations and significance calculations. The CCNT uses an algorithm, WGSA, to extract intensities from the array image file and to estimate the genomic copy number of each SNP. Calculation of the genomic copy was enabled using a variety of dosage-controlled samples and was calibrated for strong dosage–response in the 1X–5X copy number range (10). Additionally, the CCNT performed a significance calculation by comparing each SNP's calculated copy number to a sample pool of 110 normal individuals to determine the significance of each probe set's deviation from the normalized data set. Finally, a meta-analysis of the data was conducted, which assigned greater significance to consecutive SNPs that show copy number gains or losses in the same direction. The data generated by the CCNT and GDAS were exported from the software and imported into a custom Microsoft Access database for filtering and data visualization.
Gene expression statistical analysis and data visualization
Data analysis of all gene expression arrays was conducted using ArrayAssist 3.3 software (Stratagene, La Jolla, CA, USA). Probe-level analysis, normalization and summarization were performed by ArrayAssist's implementation of the GC-RMA algorithm (53). GC-RMA provides increased accuracy over the RMA algorithm by extending the model-based approach through the use of thermodynamic informatics to include the use of mismatch probes to generate final signal values. The arrays were classified into two biological groups: a control group derived from typically developing children from the AGRE cell lines (n=8) and a group of samples from cell lines with known chromosome 15 duplications (n=7). Signal values for all arrays were centered relative to the mean value of the normal cell line group and log2 transformed. Significance analysis was conducted using a stringent unpaired t-test with Benjamini–Hochberg FDR correction and filtered for genes showing significant differential gene expression above a log difference of 0.5 and below an FDR-adjusted P-value of 0.05.
The GC-RMA signal values were exported from ArrayAssist and imported into the dChip software v1.3 (17) for data visualization. Supervised, two-way, centroid-linkage hierarchical clustering analyses were performed using the genes identified by t-test. The gene branches of the hierarchical clustering tree were labeled using green to represent genes up-regulated in duplicated chromosome 15 samples and red to signify genes down-regulated in the same group when compared with normal.
dChip's ‘Genome View’ tool was used to visualize the genes from the FDR-corrected t-test to evaluate whether the identified changes occurred over a statistically significant stretch of the genome. Briefly, the software ranks all genes by physical position and then tests the likelihood that the proximity of the genes in the subset could have occurred merely by chance. The genes in the list were plotted by chromosome and labeled with the corresponding green or red label from the clustering analysis. A significance calculation was used to assess the likelihood of multiple instances of genes at close, relative proximity as described in the dChip manual. Those stretches of the genome passing a significance threshold of P<0.001 were labeled with a blue outline by the software.
The expression data were further analyzed to detect significantly dysregulated GO categories using ErmineJ software (54) (http://microarray.cu-genome.org/ermineJ/). The analysis conducted a gene score resampling that provides a robust statistical approach to evaluate GO biological function gene classes. Unfiltered results of FDR-corrected t-test described above were imported by ErmineJ. The software then determined whether any Biological Process categories were found to be significantly represented using a 100 000 iteration re-sampling technique. Categories with raw P>0.001 were exported and ranked by Benjamini–Hochberg FDR-corrected P-value.
Quantitative RT–PCR confirmation of array results
RT–PCR was performed for four genes (UBE3A, APP, HERC2 and SUMO1) identified as differentially expressed. These genes were selected on the basis of their differential expression and their relevance to our proposed dysregulated pathways.
Total RNA, remaining from probe synthesis, was used in conjunction with SS II reverse transcriptase and random hexamers (Invitrogen) to generate first strand cDNA. For the PCR step, 10 ng of total RNA converted to cDNA was used in 20 µl reactions with TaqMan® Universal PCR Master Mix (Applied Biosystems, Foster City, CA, USA) and the corresponding sequence-specific primers/probes assay mix. Each reaction was performed in triplicate according to the default PCR thermal cycling conditions described in the product protocol. Fluorescence detection was measured utilizing an ABI Prism 7900 platform (Applied Biosystems). After an initial step at 50°C for 5 min, there was an activation step at 95°C for 10 min. Thermal cycling proceeded with 40 cycles at 95°C for 15 s and 1 min at 60°C. Normalization to 18s RNA was performed to account for variability in the initial concentration and quality of total RNA and in the conversion efficiency of the reverse transcription reaction. 18s rRNA was used rather than beta-actin, GAPDH, or other possible internal controls because it was shown to be the most stable RNA species from cell lines. The endpoint, CT, identified by the ABI software as the PCR cycle number at which the signal crosses a specified threshold value was calculated for each gene. Transcript abundance is directly related to the cycle at which this threshold is crossed. Fold difference was calculated for each gene between comparison groups utilizing the comparative CT method, as outlined in the ABI Sequence Detector User Bulletin No. 2 (http://docs.appliedbiosystems.com/pebiodocs/04303859.pdf/), and normalized to an 18S rRNA endogenous reference. Significance analysis was performed by computing P-values of mean RQ values from the duplication group versus control group, using an unpaired t-test on the RQ values.
Growth curves
Cells were seeded at a concentration of 2x105 cells/ml into triplicate wells of 24-well plates. Aliquots were removed at each time point for determination of cell numbers using standard hemocytometer counting.
Reagents
Mouse monoclonal antibodies used in these experiments included anti-p53 (clone DO-1, IgG2A; BD Pharmingen, San Diego, CA, USA), anti-UBE3A (clone 13, IgG1; BD Transduction Laboratories, San Diego, CA, USA) and anti-ß-actin (clone AC-15; IgG1, ascites fluid; Sigma Chemical Company, St Louis, MO, USA). MG-132 and doxorubicin were purchased from Calbiochem (San Diego, CA, USA). Stock solutions of each were made in DMSO and stored at –20°C.
Immunoblot analysis
Immunoblot analysis was performed using standard protocols as previously described (55,56). Nuclear-free cell lysates were made in NP-40 lysis buffer (25 mM Tris–Cl, pH 7.5, 150 mM NaCl, 1% NP-40, 10% glycerol) supplemented with a cocktail of proteases inhibitors (1 µg/ml of aprotinin, leupeptin and pepstatin A) and phosphatase inhibitors (1 mM each of sodium fluoride and sodium orthovanadate). Nuclei and cellular debris were removed by centrifugation at 16 000g for 15 min at 4°C. Protein concentration of the 16K supernatant was then quantified with the bicinchoninic acid protein assay reagent (PIERCE, Rockford, IL, USA). Proteins (20 µg) were separated by denaturing SDS–PAGE and transferred to Immobilon-P PVDF membranes (Millipore) with semi-dry transfer (Bio-Rad, Hercules, CA, USA). Membranes were blocked for 1 h with 5% non-fat dry milk or bovine serum albumin in Tris-buffered saline (TBS) containing 0.05% Tween-20 (TBST). Primary antibodies were diluted in blocking buffer according to manufacturers' protocols and subsequently incubated with the blots for 2 h at room temperature or overnight at 4°C. The membranes were washed three times with TBST and incubated with a 1:2500 dilution of the appropriate horseradish peroxidase-conjugated anti-IgG in blocking buffer for 1 h at room temperature. After washing, the blots were developed with ECL Plus western blotting reagents (Amersham Biosciences, Piscataway, NJ, USA) and exposed to BioMax X-ray film (Kodak). The films were subsequently scanned and densitometry performed on digital TIFF images using TotalLab 2.0 software (Nonlinear Dynamics, BioSystematica, UK). Relative p53 expression was determined and normalized to ß-actin signals.
|
ACKNOWLEDGEMENTS |
|---|
The authors would like to extend their gratitude to the IsoDicentric 15 Exchange, Advocacy and Support (IDEAS) group, and to all participating patient families for their roles in enabling this study. We would also like to thank Barb Malone and Dawn Milliken for their excellent technical assistance. This work was supported by R01-HD-37874 (N.C.S.), U19-HD35470 (N.C.S.) from the Collaborative Programs for Excellence in Autism Research, P20RR020173 (N.C.S.), MIND Institute Biomarkers Initiative (J.P.G.), MIND Institute Genomics Core (J.P.G.) and grant number PO1-ES-11269 (J.P.G.) from the NIEHS and US EPA. We gratefully acknowledge the resources provided by the Autism Genetic Resource Exchange (AGRE) Consortium and the participating AGRE families. The Autism Genetic Resource Exchange is a program of Cure Autism Now and is supported, in part, by grant MH64547 from the National Institute of Mental Health to Daniel H. Geschwind (PI).
The AGRE Consortium
Dan Geschwind, MD, PhD, UCLA, Los Angeles, CA; Maja Bucan, PhD, University of Pennsylvania, Philadelphia, PA; W. Ted Brown, MD, PhD, FACMG, NYS Institute for Basic Research in Developmental Disabilities, Staten Island, NY; Rita M. Cantor, PhD, UCLA School of Medicine, Los Angeles, CA; John N. Constantino, MD, Washington University School of Medicine, St Louis, MO; T. Conrad Gilliam, PhD, University of Chicago, Chicago, IL; Martha Herbert, MD, PhD, Harvard Medical School, Boston, MA; Clara Lajonchere, PhD, Cure Autism Now, Los Angeles, CA; David H. Ledbetter, PhD, Emory University, Atlanta, GA; Christa Lese-Martin, PhD, Emory University, Atlanta, GA; Janet Miller, JD, PhD, Cure Autism Now, Los Angeles, CA; Stanley F. Nelson, MD, UCLA School of Medicine, Los Angeles, CA; Gerard D. Schellenberg, PhD, University of Washington, Seattle, WA; Carol A. Samango-Sprouse, EdD, George Washington University, Washington, DC; Sarah Spence, MD, PhD, UCLA, Los Angeles, CA; Matthew State, MD, PhD, Yale University, New Haven, CT; Rudolph E. Tanzi, PhD, Massachusetts General Hospital, Boston, MA.
Conflict of Interest statement. None declared.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONRESULTSDISCUSSIONMATERIALS AND METHODSREFERENCES |
|---|
-
Yeargin-Allsopp, M., Rice, C., Karapurkar, T., Doernberg, N., Boyle, C. and Murphy, C. (2003) Prevalence of autism in a US metropolitan area. JAMA, 289, 49–55.
[Abstract/Free Full Text] - Lauritsen, M., Mors, O., Mortensen, P.B. and Ewald, H. (1999) Infantile autism and associated autosomal chromosome abnormalities: a register-based study and a literature survey. J. Child Psychol. Psychiatry, 40, 335–345.[CrossRef][ISI][Medline]
- Wassink, T.H., Brzustowicz, L.M., Bartlett, C.W. and Szatmari, P. (2004) The search for autism disease genes. Ment. Retard. Dev. Disabil. Res. Rev., 10, 272–283.[CrossRef][ISI][Medline]
- Reddy, K.S. (2005) Cytogenetic abnormalities and fragile-x syndrome in autism spectrum disorder. BMC Med. Genet., 6, 3.[CrossRef][Medline]
- Gillberg, C. (1998) Chromosomal disorders and autism. J. Autism Dev. Disord., 28, 415–425.[CrossRef][ISI][Medline]
- Li, J., Jiang, T., Bejjani, B., Rajcan-Separovic, E. and Cai, W.W. (2003) High-resolution human genome scanning using whole-genome BAC arrays. Cold Spring Harb. Symp. Quant. Biol., 68, 323–329.[CrossRef][ISI][Medline]
- Auburn, R.P., Kreil, D.P., Meadows, L.A., Fischer, B., Matilla, S.S. and Russell, S. (2005) Robotic spotting of cDNA and oligonucleotide microarrays. Trends Biotechnol., 23, 374–379.[CrossRef][ISI][Medline]
-
Zhao, X., Li, C., Paez, J.G., Chin, K., Janne, P.A., Chen, T.H., Girard, L., Minna, J., Christiani, D., Leo, C. et al. (2004) An integrated view of copy number and allelic alterations in the cancer genome using single nucleotide polymorphism arrays. Cancer Res., 64, 3060–3071.
[Abstract/Free Full Text] -
Lin, M., Wei, L.J., Sellers, W.R., Lieberfarb, M., Wong, W.H. and Li, C. (2004) dChipSNP: significance curve and clustering of SNP-array-based loss-of-heterozygosity data. Bioinformatics, 20, 1233–1240.
[Abstract/Free Full Text] - Huang, J., Wei, W., Zhang, J., Liu, G., Bignell, G.R., Stratton, M.R., Futreal, P.A., Wooster, R., Jones, K.W. and Shapero, M.H. (2004) Whole genome DNA copy number changes identified by high density oligonucleotide arrays. Hum. Genomics, 1, 287–299.[Medline]
-
Rauch, A., Ruschendorf, F., Huang, J., Trautmann, U., Becker, C., Thiel, C., Jones, K.W., Reis, A. and Nurnberg, P. (2004) Molecular karyotyping using an SNP array for genomewide genotyping. J. Med. Genet., 41, 916–922.
[Abstract/Free Full Text] - Cook, E.H., Jr, Lindgren, V., Leventhal, B.L., Courchesne, R., Lincoln, A., Shulman, C., Lord, C. and Courchesne, E. (1997) Autism or atypical autism in maternally but not paternally derived proximal 15q duplication. Am. J. Hum. Genet., 60, 928–934.[ISI][Medline]
- Dawson, A.J., Mogk, R., Rothenmund, H. and Bridge, P.J. (2002) Paternal origin of a small, class I inv dup(15). Am. J. Med. Genet., 107, 334–336.[CrossRef][ISI][Medline]
- Lugtenberg, D., de Brouwer, A.P., Kleefstra, T., Oudakker, A.R., Frints, S.G., Schrander-Stumpel, C.T., Fryns, J.P., Jensen, L.R., Chelly, J., Moraine, C. et al. (2005) Chromosomal copy number changes in patients with non-syndromic X-linked mental retardation detected by array CGH. J. Med. Genet., jmg.2005.036178, in press.
- Wang, N.J., Liu, D., Parokonny, A.S. and Schanen, N.C. (2004) High-resolution molecular characterization of 15q11–q13 rearrangements by array comparative genomic hybridization (array CGH) with detection of gene dosage. Am. J. Hum. Genet., 75, 267–281.[CrossRef][ISI][Medline]
-
Runte, M., Huttenhofer, A., Gro, S., Kiefmann, M., Horsthemke, B. and Buiting, K. (2001) The IC-SNURF-SNRPN transcript serves as a host for multiple small nucleolar RNA species and as an antisense RNA for UBE3A. Hum. Mol. Genet., 10, 2687–2700.
[Abstract/Free Full Text] -
Li, C. and Wong, W.H. (2001) Model-based analysis of oligonucleotide arrays: expression index computation and outlier detection. Proc. Natl Acad. Sci. USA, 98, 31–36.
[Abstract/Free Full Text] - Albrecht, U., Sutcliffe, J.S., Cattanach, B.M., Beechey, C.V., Armstrong, D., Eichele, G. and Beaudet, A.L. (1997) Imprinted expression of the murine Angelman syndrome gene, UBE3A, in hippocampal and Purkinje neurons. Nat. Genet., 17, 75–78.[CrossRef][ISI][Medline]