Ataxia telangiectasia (AT) is a rare autosomal recessive disease caused by mutations in the ataxia telangiectasia-mutated gene (ATM). AT carriers with one mutant ATM allele are usually not severely affected although they carry an increased risk of developing cancer. There has not been an easy and reliable diagnostic method to identify AT carriers. Cell cycle checkpoint functions upon ionizing radiation (IR)-induced DNA damage and gene expression signatures were analyzed in the current study to test for differential responses in human lymphoblastoid cell lines with different ATM genotypes. While both dose- and time-dependent G1 and G2 checkpoint functions were highly attenuated in ATM−/− cell lines, these functions were preserved in ATM+/− cell lines equivalent to ATM+/+ cell lines. However, gene expression signatures at both baseline (consisting of 203 probes) and post-IR treatment (consisting of 126 probes) were able to distinguish ATM+/− cell lines from ATM+/+ and ATM−/− cell lines. Gene ontology (GO) and pathway analysis of the genes in the baseline signature indicate that ATM function-related categories, DNA metabolism, cell cycle, cell death control, and the p53 signaling pathway, were overrepresented. The same analyses of the genes in the IR-responsive signature revealed that biological categories including response to DNA damage stimulus, p53 signaling, and cell cycle pathways were overrepresented, which again confirmed involvement of ATM functions. The results indicate that AT carriers who have unaffected G1 and G2 checkpoint functions can be distinguished from normal individuals and AT patients by expression signatures of genes related to ATM functions.
- ataxia telangiectasia-mutated
- ionizing radiation
- gene expression signature
inactivating mutations in the ataxia telangiectasia-mutated (ATM) gene on human chromosome 11q23 lead to ionizing radiation (IR) hypersensitivity, defects in DNA damage checkpoint functions, and chromosomal instability (1, 40, 44). Although ataxia telangiectasia (AT) is a rare disease with a frequency of ∼1/40,000, the frequency of heterozygous carriers is much higher. The prevalence of heterozygous mutations in the ATM gene is ∼1% in the general population (50). Epidemiological studies have shown that AT carriers have an increased predisposition for developing cancers over the general population, especially for developing breast cancer in women (2, 5, 38, 48, 49, 52). Although ATM heterozygosity does not affect initiation of chronic lymphocytic leukemia, it influences rapid disease progression through loss of the remaining ATM allele (16, 39). Animal and cellular studies have shown radiosensitivity, G2-irradiation chromosomal hypersensitivity, and defective cell cycle checkpoints in ATM heterozygotes (6, 28, 41, 43, 56, 58). However, clinical symptoms observed in AT patients are usually not seen in heterozygous ATM mutation carriers, and, to date, there is no reliable clinical test for AT carriers.
ATM is a large gene (62 exons spanning ∼150 kb), and >400 mutations have been documented in AT patients falling into three categories: truncating mutations, missense mutations, and mutations leading to partial expression of mutant protein that lacks kinase activity (12, 17, 42, 54). ATM sequencing, the current gold standard for its genotyping, is labor intensive and expensive, costing up to several thousand dollars per sample with a turnaround time of at least 2 wk (http://preventiongenetics.com/clinical-dna-testing/test/ataxia-telangiectasia-syndrome-via-the-atm-gene/1042/). The diversity and broad distribution of mutations without hot spots in AT patients greatly limit the utility of direct mutation screening as a diagnostic tool or method for carrier identification (12). Other methods have been sought in recent years for diagnosis of AT carriers to replace the laborious and costly sequencing method. The mRNA levels of ATM are normal in almost all AT carriers, and protein levels, although decreased in most AT patients, are decreased in only a small portion of ATM mutation carriers (7). Therefore, determination of levels of ATM mRNA and protein cannot reliably identify ATM mutation carriers. Determination of DNA damage-dependent phosphorylation of ATM and SMC1 by flow cytometry (21, 35), a protein truncation test combined with enzyme-linked immunosorbent assay (PTT-ELISA) (14), optimized RT-PCR plus direct sequencing (34), denaturing high-performance liquid chromatography (33), and gene expression profiling (47, 55) has been developed for this purpose. RNA expression profiling as a method for rapid quantitative assessment of hundreds of transcripts is being implemented in hospital laboratories for diagnosis, prognosis, monitoring, and predicting efficacy of therapy (51). ATM-related downstream gene expression profiling may be an useful biomarker for AT carrier detection at much lower expense and shorter turnaround time (<1/3 the cost and in 1 wk) compared with the sequencing method. In the current study, gene expression signatures were extracted from human lymphoblastoid cell lines obtained from individuals with wild-type ATM, AT patients, and ATM mutation carriers. The extracted expression signatures with or without gamma irradiation (IR) successfully separated ATM+/− cell lines from ATM+/+ and ATM−/− cell lines despite the ATM+/− cells showing normal cell cycle checkpoint functions upon DNA damage induced by IR.
MATERIALS AND METHODS
Cell Lines and Culture
EBV-transformed lymphoblastoid cell lines were obtained from the National Institute of General Medical Sciences (NIGMS) Human Genetic Mutant Cell Repository (Camden, NJ), including six normal individuals (ATM+/+: AG14778, AG14832, AG15014, AG15033, GM03714, GM03299), six AT patients (ATM−/−: GM13873, GM13989, GM03332, GM02782, GM09582, and GM03189), and four AT carriers (ATM+/−: GM02781, GM03188, GM03334, GM09579). Cells were cultured in RPMI supplemented with 15% FBS (Life Technologies, Grand Island, NY). All cell lines obtained from the NIGMS were maintained at 37°C in a humidified atmosphere of 5% CO2 and were routinely tested and shown to be free of mycoplasma contamination by a commercial assay (Gen-Probe, San Diego, CA).
Cells were exposed to IR in their culture medium, with a 137Cs source (Gammacell 40; Atomic Energy of Canada, Ottawa, Canada) at a dose rate of 0.84 Gy/min. Sham-treated controls were subjected to the same movements in and out of incubators as irradiated cells.
Cell Cycle Checkpoint Assays
Logarithmically growing cells were seeded at a density of 2.5 × 105/ml in T25 flasks. Forty-eight hours after seeding cells were irradiated at doses of 0.17, 0.5, 1.5 or 4.5 Gy IR for checkpoint function determination. For quantitative analysis of the G1 checkpoint, 10 μM 5′-bromo-2′-deoxyuridine (BrdU; Sigma Chemical, St. Louis, MO) was added to the culture medium and allowed to incubate for 2 h, at various times after irradiation to label synthesizing DNA. Cells were harvested, washed with phosphate-buffered saline (PBS), and fixed with 67% ethanol in PBS. Cells were stained with fluorescein isothiocyanate (FITC)-conjugated anti-BrdU antibody (BD Biosciences, San Jose, CA) and propidium iodide (PI, Sigma), then S phase cells were enumerated by flow cytometry as previously described (27, 29, 31). For quantitative analysis of the G2 checkpoint, mitotic cells were quantified by flow cytometry. Briefly, cells were harvested at various times after irradiation, washed with PBS, fixed in 1% formaldehyde in PBS for 30 min, and then fixed in 67% ethanol in PBS. Cells were incubated in 0.5 μg/100 μl anti-phospho-histone H3 antibody (Upstate Biotechnology, Lake Placid, NJ) for 2 h and then stained with FITC-conjugated anti-rabbit antibody (Santa Cruz Biotechnology, Santa Cruz, CA) and PI (25, 30, 57). Flow cytometric analyses to enumerate mitotic cells were done using a FACScan flow cytometer (BD, San Jose, CA) and Summit software (Dako Colorado, Fort Collins, CO).
Logarithmically growing cells were harvested for microarray analysis or were treated with 1.5 Gy IR and harvested 6 h later, and controls were harvested at the same time after sham treatment. Total RNA was isolated using Qiagen RNeasy kit (Qiagen Sciences, Germantown, MD). The quality of all RNA samples was confirmed using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). Microarray analysis was performed using a 22k human 1A array (Agilent) through a contract with Cogenics (Research Triangle Park, NC). Hybridization of sample RNA against global reference RNA (Stratagene, La Jolla, CA) was done twice with dye swap. For details, refer to our previous publication (60). We processed 24 arrays with four ATM+/+, four ATM+/−, and four ATM−/− cell lines for baseline gene expression analysis, and 64 arrays were processed in separate experiments with six ATM+/+, four ATM+/−, and six ATM−/− cell lines for analysis of IR-induced changes in gene expression.
Microarray Data Analysis
The complete transcript data set used in our study has been deposited at the Gene Expression Omnibus database under accession number GSE45850. The extraction of gene expression patterns that were associated with ATM genotype was carried out by a method known as “extracting gene expression patterns and identifying coexpressed genes” (EPIG) (9). Three parameters, the Pearson correlation coefficient within a specific pattern, the magnitude of change in gene expression (signal), and the signal-to-noise ratio (SNR), were employed for selection of significant genes (9, 60). The expression patterns were generated as described briefly here. The expression of each gene was represented as a log2 ratio (sample/reference). For baseline pattern extraction, the average of gene expression from ATM+/+ cell lines was aligned to zero, and gene expression from ATM+/− cell lines and ATM−/− cell lines was adjusted accordingly to observe absolute changes in gene expression from ATM+/− and ATM−/− cell lines. Correlation coefficient was calculated for each gene against all other genes in all cell lines, and genes highly correlated were clustered to get a set of discrete gene expression patterns, and then SNR, magnitude, and correlation coefficient were used for significant gene selection in each pattern. For IR-treated pattern extraction, the average of gene expression from controls in each genotype group was aligned to zero, and gene expression after IR treatment was adjusted accordingly to observe absolute IR-induced changes in gene expression in different genotype groups. Correlation coefficient was calculated for each gene against all other genes in all IR-treated cell lines to get the expression patterns, and significant genes in each pattern were selected in the same way as described above. Two-dimensional principal component analysis (PCA) and hierarchical cluster analysis (HCA) were performed using Partek software (Partek, St. Louis, MO). HCA genotypes were grouped by Euclidean distance and average linkage. Fuzzy ARTMAP was used for prediction of ATM genotype (19, 22, 32). Gene ontology (GO) and pathway analysis using DAVID (version 6.7) was done as previously described (23).
Cell Cycle Checkpoint Functions in Response to IR-induced DNA Damage
The G1 arrest induced by IR was quantified by measuring the incorporation of BrdU 6–8 h after 0.5, 1.5, and 4.5 Gy of IR or sham treatment in six ATM+/+, six ATM−/−, and four ATM+/− lymphoblast cell lines (Fig. 1A). ATM+/+ and ATM+/− cells displayed very similar dose-dependent reductions in the fraction of BrdU-labeled S phase cells in the first half of S phase (2–3N DNA content) because of ATM- and p53-dependent G1 arrest (29). At the dose of 0.5 Gy, the reduction of cells in the first half of S was ∼50% relative to the sham-treated control. Treatment with 1.5 Gy IR seemed to have saturated the response with a reduction of 70%, close to that caused by 4.5 Gy in both ATM+/+ and ATM+/− cell lines. The G1 arrest was markedly attenuated in ATM−/− cell lines compared with ATM+/+ and ATM+/− cell lines, with reductions of S phase fractions of 10, 25, and 35%, respectively, for 0.5, 1.5, and 4.5 Gy IR treatments (Fig. 1A, left). While ATM−/− cell lines displayed a significant defect in IR-induced G1 arrest, the ATM+/− cell lines were not significantly different from ATM+/+ cell lines.
IR-induced G2 arrest was quantified by measuring mitosis-specific phospho-histone H3 immunostaining 2 h after IR or sham treatment. IR at the dose of 0.17 Gy caused a similarly moderate reduction (∼10%) in the fraction of mitotic cells because of G2 arrest in all cell lines; IR at the dose of 0.5 Gy caused 84 and 73% reductions in ATM+/+ and ATM+/− cell lines, but only 36% reduction in ATM−/− cell lines; IR at the dose of 1.5 Gy caused ∼95% reduction in both ATM+/+ and ATM+/− cell lines but only 70% in ATM−/− cell lines (Fig. 1A, right). Similar to the G1 checkpoint response to IR, ATM−/− lines displayed a significant defect in G2 arrest relative to ATM+/+, but ATM+/− cell lines preserved normal G2 checkpoint function.
Dynamic changes in DNA content, DNA synthesis, and mitosis were monitored 2–24 h after 1.5 Gy IR (Fig. 1B). Similar to the results observed in the checkpoint analyses, ATM+/+ and ATM+/− cell lines showed very similar responses to IR over the broader time-course. In ATM+/+ and ATM+/− cell lines, the percentage of cells with 2N DNA content that did not incorporate BrdU (G1 population) showed a 10–15% decline from 2 to 24 h post-IR compared with sham-treated controls. S phase cells with 2–4N DNA content and labeled with BrdU declined by 30–35% at 6 h and by 60–65% from 12 to 24 h after IR treatment compared with sham-treated controls. There was a rapid postirradiation accumulation of 4N cells with no BrdU incorporation (predominantly G2) that peaked at 12 h (5 times the control) and maintained the same level in ATM+/− cells or recovered moderately in ATM+/+ cell lines at 24 h. Mitosis was severely inhibited (>95%) 2 h after IR and then recovered to 50% of control levels at 6 and 12 h and to 65 and 53% of control levels at 24 h in ATM+/+ and ATM+/− cells, respectively. The changes in ATM−/− cell lines were quite different from ATM+/+ and ATM+/− cell lines after IR irradiation. Generally ATM−/− cell lines showed marked reduction in G1 phase cells from 6 to 24 h, decreased inhibition of S phase (6–12 h) and mitotic cells (2–12 h), and reduced accumulation in G2 phase cells. However, by 24 h postirradiation, the inhibitions of S phase and mitotic cells in ATM−/− cell lines were very close to those in ATM+/+ and ATM+/− cell lines, suggesting delayed G1 and G2 checkpoint functions, which might be signaled by the ATR pathway (1, 24, 44).
Differential Gene Expression Profiles in ATM+/+, ATM+/−, and ATM−/− Cells
Previous work showed that p53-dependent radiosensitivity in cancer cell lines could be predicted by both basal gene expression and IR-induced changes in gene expression (3), and we showed that basal gene expression could be used to predict p53-dependent G1 checkpoint function in melanoma cell lines (8). Therefore, gene expression profiles at baseline level and post-IR treatment in the current study were analyzed to identify expression signatures that may help distinguish ATM genotypes.
EPIG was used for analysis (9). The criteria for significant gene selection were SNR > 3, signal magnitude > 0.5, and Pearson correlation coefficient > 0.64.
Gene expression signatures at baseline level.
A subset of the lymphoblastoid cell lines, four ATM+/+, four ATM+/−, and four ATM−/−, was utilized to acquire gene expression profiles at the baseline level when in logarithmic growth. ATM+/− cell lines that were used in the current study were obtained from adults, while ATM−/− cell lines from AT patients were from children. ATM+/+ cell lines were from normal children that were age-matched to AT patients. Consequently, there was a concern about age-related bias in analysis of gene expression profiles. To try to eliminate the effect of age on gene expression, gene expression profiles were compared between children (ATM+/+ and ATM−/− grouped together) and adults (ATM+/−). Age-related genes identified by EPIG (data not shown) were removed from the analysis. After removal of age-related genes, EPIG identified 203 probes that were placed in nine expression patterns (Fig. 2, A and B; Supplement 1).1
Among the nine expression patterns identified for this gene set, basal expression levels generally displayed large variation in individual cell lines within the same genotype in each expression pattern (idiosyncratic changes). Consequently, it was not possible to distinguish ATM genotypes by using the expression profile from a single pattern. However, when the 203 probes from all nine expression patterns were subjected to PCA and HCA, cell lines with different ATM genotypes were clearly separated from each other (Fig. 2, C and D). Interestingly, HCA results show that the expression signature of the 203 probes from ATM+/− cell lines was closer to that from ATM−/− cell lines than from ATM+/+ cell lines (Fig. 2D). Prediction of ATM genotypes using the expression signature consisting of the 203 probes in Fuzzy ARTMAP leave-one-out cross validation was 100% correct (19, 22, 32).
Further analysis of the expression signature from the nine patterns was performed to understand the potential biological processes behind the variation in expression of these genes. When expression profiles in ATM−/− and ATM+/− cell lines were compared with those in ATM+/+ cell lines, a total of 88 genes from patterns 1–3 had reduced expression levels, while a total of 97 genes from patterns 4–7 had increased expression levels in ATM−/− and ATM+/− cell lines relative to ATM+/+ cell lines. Pattern 8 showed six genes having higher expression levels in ATM−/− cells and lower expression levels in ATM+/− cells relative to ATM+/+ cells. Twelve genes from pattern 9 had the lowest expression level in all ATM−/− cell lines and in one ATM+/− cell line, while four ATM+/+ and three ATM+/− cell lines showed similar expression levels.
GO and pathway analysis on the identified genes using DAVID indicated ATM function-related biological processes (Table 1). Genes in patterns 1–3 are related to cell cycle control and DNA metabolic processes. Genes in this group include CDCA5, E2F1, EGR1, FANCA, KIF21B, MCM10, MCM3, MCM4, RAD54L, and RFC5 (Table 1, Supplement 2). Genes in patterns 4–7 are related to regulation of cell death and p53 signaling pathway. Genes in this group include BCL2L14, BMF, CASP4, CASP5, DNAJB9, FAS, GADD45A, LAG3, MAP3K5, OPTN, SESN1, SESN2, and TNFSF9 (Table 1, Supplement 3). No significant biological process (BP) category and pathway are generated from genes in pattern 8 and pattern 9. Named genes in pattern 8 are GP9, KCNN2, and PSAP. Genes in pattern 9 include ATM, BCAR3, C3orf54, C6orf64, FGD6, FHOD1, IGH@, KCNK6, RGS20, and TNS4 (Table 1, Supplement 4).
Gene expression signatures post-IR treatment.
IR-induced changes in gene expression were analyzed in six ATM+/+, four ATM+/−, and six ATM−/− cell lines. Cells received a sham treatment or 1.5 Gy IR and gene expression were analyzed 6 h later using EPIG with the same criteria for significant gene selection as described previously. With basal expression levels (sham-treated controls) adjusted to zero (log2 scale), three stereotypic expression patterns and 126 significant genes that responded to IR were identified (Fig. 3, A and B; Supplement 5). In pattern 1, genes were highly induced in the ATM+/+ and ATM+/− cell lines and only moderately induced in the ATM−/− cell lines. GO and pathway analysis revealed that these genes were mainly related to p53-mediated apoptosis and cell cycle arrest upon IR-induced DNA damage, including many p53 target genes such as BAX, CDKN1A, DDB2, FDXR, GADD45A, LRDD, SES2, TP53AP1, and TP53I3 (Table 2, Supplement 6). In pattern 2, genes were markedly repressed in ATM+/+ and ATM+/− after IR but were not changed in ATM−/− cells, including many cell cycle- and cell proliferation-regulated genes and genes involved in negative regulation of metabolic process, such as CHEK1, E2F8, EXO1, FBXO5, MCM3, MCM6, MSH6, MYC, RAD54L, RFC2, RFC4, UHRF1, and UNG (Table 2, Supplement 7). In pattern 3, genes were only induced in ATM+/− cells and these genes mainly function in glycine, serine, and threonine metabolism pathway, including ADA, GAMT, ASS, APOA2, SILV, and NQO1 (Table 2, Supplement 8).
To apply IR-responsive gene expression signatures to distinguish ATM genotypes, IR-induced changes in gene expression were obtained by subtracting gene expression levels of sham-treated controls. The changes in the 126 IR-responsive genes were then subjected to PCA and HCA. While PCA produced good separation of all three ATM genotypes (Fig. 3C), HCA showed good separation of ATM−/− from ATM+/− and ATM+/+, but the separation of ATM+/− from ATM+/+ was not perfect (Fig. 3D), indicating cell cycle checkpoint functions in ATM+/− cell lines were well preserved. However, prediction of ATM genotypes using IR-responsive genes with Fuzzy ARTMAP analysis was still 100% correct.
ATM rapidly responds to DNA damage by initiating a series of signal transduction cascades associated with cell cycle checkpoints, DNA repair, and programmed cell death (1, 44). It is a central gene in the complex system of DNA damage response that maintains genome stability (10). AT patients who have mutations in both copies of ATM are characterized by neurodegeneration, immunodeficiency, sensitivity to IR, and high predisposition for malignancy (26, 45). However, there is usually no marked heterozygote phenotype in AT carriers, making their identification difficult. Epidemiology studies have confirmed the increased risk of breast cancer in relatives of AT patients (2, 4, 18, 52, 53). Analysis from a recent large study of 1,160 relatives of 169 AT patients estimated the overall relative risk of breast cancer in carriers to be 2.23, and for women under 50 yr of age relative risk was 4.9 compared with the general population (52). Given that up to 1% of the population might be AT carriers, even a relatively modest increase in breast cancer risk in carriers could equate to an appreciable population risk (2). Thus, a relatively easy and reliable method to identify AT carriers among the general population is needed.
A variety of assays have detected significant differences between the cells of AT patients and matched normal controls including radiation-induced clonogenic survival, apoptosis, chromosomal aberrations, and expression of biomarkers of DNA damage response (11, 13, 37, 46, 56). For most of these assays AT carriers display an intermediate response to IR that is significantly different from normal controls. Few of these laboratory assays are suitable for clinical laboratory testing because of time and expense, although a flow cytometric assay of IR-induced phospho-SMC1 shows promise for detecting AT carriers (35). As the cells of AT carriers display abnormal responses to IR, we reasoned that it should be possible to identify abnormalities in gene expression that characterize the carrier state.
Microarray analysis has been widely used to study ATM-dependent transcriptional regulation in response to DNA damage and oxidative stress (15, 20, 24, 59). However, only two studies have compared the three ATM genotypes by using gene expression signatures (47, 55). AT carriers were distinguished from noncarriers by gene expression profiles at baseline and post-IR treatment (55). Gene expression phenotypes were further categorized as recessive with the IR-response patterns being similar between AT carriers and noncarriers but significantly different from AT patients and dominant with the IR-induced expressions being similar between AT carriers and AT patients but different from noncarriers. The recessive gene expression phenotypes belonged to the ATM-p53 pathway, whereas the dominant gene expression phenotypes belonged to the ATM-AKT pathway (47). Results from our microarray analysis on ATM wild-type, heterozygous, and homozygous human lymphoblastoid cell lines support these previous findings in general. Gene expression signatures at either baseline or after IR treatment were able to distinguish the three ATM genotypes.
The post-IR expression signature consisted of 126 significant genes. The ATM+/− and ATM+/+ cell lines had very similar changes in expression of genes playing important roles in DNA damage responses through p53 signaling, including induction of BAX, CDKN1A, DDB2, FDXR, GADD45A, TP53AP1, and TP53I3 (Fig. 3, A and B, pattern 1) and marked repression of CHEK1, E2F8, MCM3, MCM6, MSH6, MYC, RAD54L, RFC2, and RFC4 (Fig. 3, A and B, pattern 2) (1, 60), which were very different from the ATM−/− cell lines. This result was consistent with the observed G1/S and G2/M checkpoints that were normal in the ATM+/− cell lines and supported the theory of recessive expression phenotypes belonging to the ATM-p53 pathway (47). However, the dominant expression phenotype was not observed in our study. No significant expression pattern was shown to have similar gene expression post-IR treatment between AT carriers and AT patients but different from noncarriers. Two representative genes, CHEK1 and FBXO5, showing dominant expression phenotypes in the previous study, were observed as recessive expression phenotypes in our study (Fig. 2A, pattern 2). A unique expression pattern was identified in our study showing induction of genes functioning in glycine, serine, and threonine metabolism in ATM+/− cell lines but not in the ATM+/+ and ATM−/− cell lines. Our combined post-IR expression signature successfully separated the three ATM genotypes using PCA, and the result was confirmed by Fuzzy ARTMAP analysis. Although HCA was able to aid in directly observing differences in expression patterns, its power in separating ATM genotypes in this study is not as strong as PCA, which uses multidimensional parameters in calculation.
Basal levels of gene expression were also found to separate the three genotypes, and nine patterns of expression were identified. The gene expression signature at baseline consisted of 203 significant genes that function in several ATM-dependent pathways, including cell cycle control, p53 signaling, and apoptosis. Analysis of details in each expression pattern at baseline suggested that individual variation in genetic background, independent of ATM genotypes, might have a marked effect on gene expression, especially in ATM+/− cell lines. Two examples were in patterns 3 and 9. In pattern 3, gene expression levels in all four ATM+/+ cell lines and one ATM+/− cell line were the same, while the other three ATM+/− cell lines had the same expression levels as the four ATM−/− cell lines. In contrast, in pattern 9 the same ATM+/− cell line resembled the ATM−/− cell lines, while the other three ATM+/− cell lines resembled the ATM+/+ cell lines. Thus, genes from a single pattern were not sufficient to discriminate the ATM genotypes. However, when the genes from all nine expression patterns were considered in aggregate, the three ATM genotypes were well distinguished by both PCA and HCA.
An interesting observation in our study was that ATM transcriptional levels were significantly lower in ATM−/− cells than in ATM+/+ and ATM+/− cells, while no significant difference was observed between ATM+/+ and ATM+/− cells (Fig. 4A). This result is in contrast to a previous report that ATM mRNA levels were normal in almost all AT patients and carriers (55). Another interesting finding was that the gene encoding the immunoglobulin heavy chain locus (IGH@) in pattern 9 was highly repressed in AT cells with an average of 18-fold decrease in expression level (P < 0.001) compared with that in ATM+/+ cells. One of the four ATM+/− cell lines also showed a marked decrease in IGH@ expression, while the others displayed a normal expression level (Fig. 4B). The marked decrease in IGH@ expression in AT cells is consistent with studies showing that AT patients exhibit reduced expression of immunoglobulins alpha, gamma, and epsilon, reflecting a severe defect in class-switch recombination (36). In summary, gene expression signatures related to ATM-dependent signal transduction at either baseline or post-IR levels were able to distinguish three ATM genotypes and therefore may be useful in clinical screening for AT carriers.
This work was supported in part by Public Health Service Grants ES-10126 and ES-11391 and in part by the Intramural Research Program of the National Institutes of Health, National Institute of Environmental Health Sciences Grant Z01 ES102345-04.
No conflicts of interest, financial or otherwise, are declared by the author(s).
Author contributions: L.Z., J.C., and P.R.B. analyzed data; L.Z. and T.Z. interpreted results of experiments; L.Z. and T.Z. drafted manuscript; L.Z., D.A.S., C.L.I., J.C., P.R.B., R.S.P., W.K.K., and T.Z. approved final version of manuscript; D.A.S. and C.L.I. performed experiments; D.A.S. and P.R.B. prepared figures; D.A.S., C.L.I., J.C., P.R.B., R.S.P., W.K.K., and T.Z. edited and revised manuscript; W.K.K. and T.Z. conception and design of research.
We thank Yingchun Zhou for performing cell culture and tests for mycoplasma contamination and George Wu for helping with microarray data analysis and database management.
↵1 The online version of this article contains supplemental material.