The transcriptional response after oxidative stress is defective ... - Nature

Oncogene (2003) 22, 1135–1149

& 2003 Nature Publishing Group All rights reserved 0950-9232/03 $25.00 www.nature.com/onc

The transcriptional response after oxidative stress is defective in Cockayne syndrome group B cells Kasper J Kyng1,3,4, Alfred May1,4, Robert M Brosh Jr1, Wen-Hsing Cheng1, Catheryne Chen1, Kevin G Becker2 and Vilhelm A Bohr*,1 1 Laboratory of Molecular Gerontology, National Institute on Aging, 5600 Nathan Shock Drive, National Institutes of Health, Baltimore, MD 21224, USA; 2DNA Array Unit, National Institute on Aging, National Institutes of Health, Baltimore, MD 21224, USA

Cockayne syndrome (CS) is a human hereditary disease belonging to the group of segmental progerias, and the clinical phenotype is characterized by postnatal growth failure, neurological dysfunction, cachetic dwarfism, photosensitivity, sensorineural hearing loss, and retinal degradation. CS-B cells are defective in transcription-coupled DNA repair, base excision repair, transcription, and chromatin structural organization. Using array analysis, we have examined the expression profile in CS complementation group B (CS-B) fibroblasts after exposure to oxidative stress (H2O2) before and after complete complementation with the CSB gene. The following isogenic cell lines were compared: CS-B cells (CS-B null), CS-B cells complemented with wild-type CSB (CS-B wt), and a stably transformed cell line with a point mutation in the ATPase domain of CSB (CS-B ATPase mutant). In the wt rescued cells, we detected significant induction (>two-fold) of 112 genes out of the 6912 analysed. The patterns suggested an induction or upregulation of genes involved in several DNA metabolic processes including DNA repair, transcription, and signal transduction. In both CS-B mutant cell lines, we found a general deficiency in transcription after oxidative stress, suggesting that the CSB protein influenced the regulation of transcription of certain genes. Of the 6912 genes, 122 were differentially regulated by more than two-fold. Evidently, the ATPase function of CSB is biologically important as the deficiencies seen in the ATPase mutant cells are very similar to those observed in the CS-B-null cells. Some major defects are in the transcription of genes involved in DNA repair, signal transduction, and ribosomal functions. Oncogene (2003) 22, 1135–1149. doi:10.1038/sj.onc.1206187 Keywords: microarray; Cockayne syndrome; transcription; DNA repair; DNA damage; aging

*Correspondence: VA Bohr; E-mail: [email protected] 3 Current address: Department of Molecular and Structural Biology, Danish Center for Molecular Gerontology, University of Aarhus, DK-8000 Aarhus C, Denmark 4 Contributed equally to this work This article is a ‘United States Government Work’ paper as defined by the US Copyright Act. Received 2 July 2002; revised 18 October 2002; accepted 23 October 2002

Introduction An increasing body of evidence suggests that the accumulation of DNA damage leads to genomic instability, which can contribute to aging and ageassociated diseases such as cancer. Cells have evolved coordinately regulated mechanisms to facilitate DNA repair and cell survival, including induction and repression of genes involved in DNA repair, transactivation, signal transduction, and proteins normally associated with tissue injury and inflammation (Friedberg et al., 1995). Since transcriptional regulation changes with age (Weindruch et al., 2002), we hypothesized that this change contributes to the decline in DNA repair and the accumulation of mutations seen with aging. Recent studies of stress-induced expression patterns (Zou et al., 2000) show that these partially overlap with ageassociated changes, supporting a role of stress-induced DNA damage in the causation of aging and aging syndromes. Cockayne syndrome (CS) is a human hereditary disease belonging to the group of segmental progerias, and the clinical phenotype is characterized by postnatal growth failure, neurological dysfunction, cachetic dwarfism, photosensitivity, sensorineural hearing loss, and retinal degradation (Nance and Berry, 1992). Complementation studies demonstrate that there are at least two genes involved in CS, designated CSA and CSB (Friedberg, 1996). The CSA and CSB genes have been cloned and their products characterized biochemically. Whereas the CSA gene product belongs to the ‘WD repeat’ family of structural and regulatory proteins that lack enzymatic activity (Henning et al., 1995), the CSB protein is a DNA-dependent ATPase and chromatin remodelling factor (Citterio et al., 2000) that belongs to the SWI/SNF family of proteins. Cockayne syndrome complementation group B (CSB) cells are deficient in transcription-coupled repair (TCR), a subpathway of nucleotide excision repair (NER) (Balajee and Bohr, 2000) and the CSB protein is thought to function as a transcription elongation factor (Lee et al., 2001). It appears that the CSB protein plays a pivotal role in the DNA damage response, functioning in several pathways. On a cellular level, CS shows increased sensitivity to a number of

Cockayne syndrome group B pathways KJ Kyng et al

1136

DNA-damaging agents including UV radiation, ionizing radiation, and hydrogen peroxide (H2O2) (Balajee and Bohr, 2000). Recent studies from this laboratory demonstrate that CS-B cells are defective in the incision of 8-oxoguanine (8-oxoG) in DNA, a lesion that is removed by base excision repair (BER) (Tuo et al., 2001). The biological function of CSB in these different pathways may be mediated by distinct functional domains of the protein. We have stably transfected isogenic human CS-B fibroblast cell lines with the wildtype CSB gene (CS-B wt), a CSB ATPase motif II point mutant (CS-B ATPase mutant), or the expression vector alone (CS-B null) (Selzer et al., 2002). The transfected wt and mutant CSB alleles have been tested for their ability to function in DNA repair and for their effects on gene expression. While the absence of the CSB gene resulted in reduced incision of 8-oxoG lesions by cell extracts, mutation of the ATPase motif II abolished the ability of the CSB protein to complement the UVsensitive phenotypes of survival, RNA synthesis recovery, and apoptosis, but not the incision of 8-oxoG lesions (Selzer et al., 2002). These results suggest that the CSB protein functions by different mechanisms dependent upon the specific DNA repair pathway involved. Although CS-B cells may have a lower basal transcription rate, we have previously found that in the absence of exogenous stress treatment, CS-B null and wild-type (CS-B wt) cells express a large number of genes in a similar pattern (Selzer et al., 2002). To investigate whether CSB is involved in the transcriptional regulation after oxidative DNA damage, we used cDNA microarrays to characterize the transcriptional response to H2O2 in the three transfected cell lines. The high throughput mapping of gene induction profiles by cDNA microarrays enabled investigation of the global genome stress response, something that is not possible using traditional methods. The complementation of the CS-B cells with the wt CSB gene can be considered to represent a wt cell line, because we have shown that there is complete complementation of cellular phenotypes relating to DNA repair (Selzer et al., 2002). Thus, our studies here include the analysis of the transcriptional DNA damage response patterns in wt cells as well as in CS-B null cells. Expression changes associated with both stress response and aging depend on the genetic background, and variations between individual cell lines and different species are considerable (Lee et al., 1999, 2000; Cao et al., 2001; Kayo et al., 2001). The experimental design used here is unique in that the use of isogenic cell lines allowed us to characterize CSBdependent transcription in a uniform genetic background. We present herein results of a study based on microarray analysis to examine the effect of genotoxic stress on gene expression in human cells of a progeriod disorder. This is the first study on global genome expression in CS-B cells and we report a set of 122 genes (1.8% of those studied) whose transcription patterns in response to H2O2 are altered in CS-B cells. Of the 6912 genes screened, we identified 112 genes (1.6%) responsive to H2O2 in the wt CS-B cell line. The H2O2 response has Oncogene

not previously been characterized on a global genome scale in human wt cells, and a number of the identified genes in this study were not previously known to be induced by oxidative damage. Our results indicate that the consequences of the defect in CS are extensive and affect not only stress response and transcription regulation but also cell cycle checkpoints and central areas of signal transduction.

Results Validation of cDNA microarray data and statistical analysis Biological replicate experiments were performed for all experimental conditions and isogenic cell lines that had been isolated from individual colonies. For each point, two separate RNA purifications were performed and each RNA sample was hybridized to two different arrays (array replicate) yielding a total of four repeats per data point. The experimental procedure is outlined in Figure 1. The variation was addressed by calculating the coefficient of variation (CV), the standard deviation (s.d.) divided by the average, and scatter plot analysis. The average CV for this study was 0.16 between duplicate array hybridizations from the same RNA sample and 0.24 for all four replicates encompassing the variation resulting from duplicate biological experiments as well as from duplicate hybridizations. The microarray scans constituting the raw data were inspected manually and arrays with local nonuniform background were excluded from the analysis. An average of 3.2 replicates were used for the array analysis. Scatter plots corresponding to array replicates with a CV of 0.16 and biological replicates with a CV of 0.24 are shown in Figure 2a,b. Gene expression studies can be complicated by false positives if they rely on ratiobased criteria alone and do not include significance testing (Miller et al., 2001). Thus, in our analysis, we calculated statistically significant fold-changes with accompanying P-values based on the CVs for the gene expression values in this experiment. Tissue Culture H2O2 exposure for 15 minutes

Total RNA Isolation at Zero, 15 minutes, 6 and 24 hours DNase Treatment, Test and Quantify

Label via PCR Repeat using same RNA (Array replicate)

Quantify Arrays

Repeat entire biological experiment (Biological replicate)

Figure 1 Flowchart of the experimental procedure. Biological replicates and array replicates were each repeated twice


1137

Figure 2 Different representations verifying the reproducibility of the arrays. Panels a and b are scatter plots comparing either data from two separate hybridizations within the same biological experiment (a) or data from two separate experiments (b). (c) demonstrates the correlation between the expression ratios of the arrays and Northern data

To verify the cDNA microarray data, Northern analysis was performed to measure expression levels of six genes. As shown in Figure 2c, the ratios measured by Northern analysis confirmed those measured by microarrays, correlating for both under- and overexpressed genes. The Northern blots, together with the statistics on variability, validated the reproducibility of our data.

Global transcriptional response to H2O2 in a wt rescued cell line Thus far little is known about the normal global genome response to H2O2 in human cells. To compile a list of H2O2 responsive genes, we compared CS-B wt before and after oxidative cellular stress. (CS-B wt are the CS-B cells complemented with wt CSB.) Based on the variability of the data set we determined the extent of change that would reach statistical significance. We have

previously demonstrated that without genotoxic stress, expression patterns are not changed in CS-B cells (Selzer et al., 2002). Therefore, predamage mRNA levels were only measured a single time in this study. Analysing the probability of false positives in a single measurement, it was observed that when the variation corresponded to a CV of 0.24, one out of 1000 genes would randomly display a five-fold difference (P ¼ 0.001). Of the 6912 genes analysed, 136 were upregulated by more than fivefold after H2O2 damage. To add significance to the analysis by including a data point with four replicates, only those of the 136 genes for which the maximal expression value was elevated at least two-fold compared to a time point other than untreated were included. Together, the two criteria are very conservative and it limited the screening to 112 of the 6912 genes (1.6%), all of which were upregulated after cellular damage. While a number of genes were downregulated after cellular damage, none of these passed the significance criteria. The 112 genes were selected for further analysis, and are listed in Table 1 (ESTs not included in the table). Oncogene


1138 Table 1 Overexpressed genes in the wt rescued cell line after H2O2 exposure (ESTs not listed) C l ust er 1 Genbank AA236164

M axi mum g ene exp r essi o n chang e af t er H2 O2 exp o sur e r el at i ve t o unt r eat ed Gene Name Genbank Gene Name + 7.2 Cat hepsin S AA497002 + 5.3 M elanoma cell adhesion molecule

AA598950

+ 6.6

Cat hepsin B

AA450062

+ 5.3 Prostat e diff erentiation f act or (M IC-1)

AA504943 R71440

+ 6.4 + 6.3

Cryst allin, alpha B HsP47 (colligen)

H77855 AA019996

+ 5.2 G prot ein-coupled recept or 48 + 5.0 Prostaglandin E receptor 4

H15111

+ 6.0

Uracil-DNA glycosylase

AA496838

+ 60

T71757

+ 6.0

Heme oxygenase (decycling) 1

H23421

+ 5.8 Ribosomal prot ein L7a

AA070226 AA446453 R55075 T67029 AA454146 AA291398 AA608568

+ 5.6 + 5.3 + 5.2 + 6.7 + 6.6 + 5.9 + 5.1

R97066 AA464568 R01340 N54244 AA421296 AA278759 AA454563

+ 5.3 + 5.3 + 8.8 + 6.9 + 6.6 + 6.5 + 6.4

AA443546

+ 6.6

AA489219 AA490462 W73792 AA029889

+ 5.9 + 5.8 + 5.5 + 5.3

AA127116

+ 5.1

Selenoprotein P, plasma, 1 Pref oldin 5 M pV17 transgene, murine homolog Ornithine decarboxylase antizyme 2 Cyclin H Chromosome condensat ion 1 Cyclin A2 V-rel avian ret iculoendotheliosis viral oncogene homolog DUTP pyrophosphatase AE-binding prot ein 1 ATP-dependent RNA helicase (ROK1) Transferrin recept or (p90, CD71) Het erogeneous nuclear ribonucleoprot ein A1

T62100

+ 5.6

AA026102

+ 5.1

Transcript ion fact or 3

AA442092

+ 5.6

AA443982

+ 6.8

Prot ein phosphatase 1, cat alyt ic subunit , alpha isof orm

AA488072

+ 5.4 Cardiac ankyrin repeat prot ein

H60549

+ 6.6

CD59 antigen p18-20

N47972

+ 5.3

AA039851

+ 6.2

AA281667

+ 5.8

Prot ein tyrosine phosphat ase type IVA, member 3 Prot ein kinase inhibit or alpha

W47073

+ 5.6

Solut e carrier family 20 (phosphate) Transforming growt h f actor, bet a receptor III CD81ant igen (t arget of ant iprolif erat ive antibody 1)

R02609

+ 6.3

AA496810 AA446108 AA456931 AA143649

+ 6.1 + 6.1 + 5.,9 + 5.6

R11698 AA488715 R01669

Ribosomal prot ein L5

Transglutaminase 2 Proteasome 26S subunit , ATPase, 4 CGI-76 prot ein BANP homolog, SM AR1 homolog CD68 antigen Proteoglycan 1, secret ory granule CD63 ant igen (melanoma 1 antigen) Protein disulf ide isomerase-related protein Protein kinase C subst rate 80K-H Endoglin Cyt ochrome c oxidase subunit VIc HLA class II region expressed gene Complement component 4-binding protein, alpha Cat enin (cadherin-associated prot ein), beta 1

RAB36, member RAS oncogene family ubiquinol-cytochrome c reduct ase + 5.3 hinge prot ein + 5.2 ATPase, H+ t ransporting, lysosomal Protein disulf ide isomerase-related + 5.1 protein

AA432030

+ 5.1 Int erf eron, alpha-inducible prot ein

R06307

+ 5.0

H 62473

+ 5.5

AA486653

+ 5.4

C l ust er 2 Genbank AA490946

M axi mum g ene exp r essi o n chang e af t er H2 O2 exp o sur e r el at i ve t o unt r eat ed Gene Name Genbank Gene Name +10.5 Heat shock protein, DNAJ-like 2 AA460480 + 9.7 Creatine kinase, mitochondrial 2

N25897

+ 7.4

BCL2-associat ed at hanogene 4

AA233549

+ 7.9 D component of complement (adipsin)

AA481758

+ 9.8

DNAJ (Hsp40) homolog, subf amily B, member 1

AA180912

+ 9.5 Tubulin, alpha 1 (t estis specific)

H20743

+ 8.7

Cell division cycle 34

AA100696

+ 8.6

AA284408

+ 6.8

Cut (Drosophila)-like 1

AA486393

+ 8.5 Int erleukin 10 recept or, beta

R73545

+ 8.2

Flot illin 2

R33030

+ 8.5 Glucose regulated prot ein, 58kD

AA464034

+ 9.6

Ribosomal protein L21

AA599177

+ 8.0 Cyst at in C

W15277

+ 9.3

Ribosomal protein L31

AA062814

+ 9.6 Sialylt ransf erase

AA487637

+ 8.1

Transporter 1, ATP-binding cassett e, sub-f amily B

H48420

+ 7.9 Prothymosin, alpha

N54914

+ 7.7 CREBBP/ EP300 inhibitory protein 1

C l ust er 3 Genbank AA147214 AA487700 R84242

HIV-1 inducer of short t ranscript s binding prot ein

Transmembrane 4 superf amily member 7

M axi mum g ene exp r essi o n chang e af t er H2 O2 exp o sur e r el at i ve t o unt r eat ed Gene Name + 13.2 GADD45 + 11.8 Cyclin D1 + 11.8 Ret iculocalbin 1

DNA damage, st ress response, apopt osis Replicat ion, cell cycle Transcript ion, RNA met abollism

Signal t ransduct ion Prot ein met abolism Ot her

a The 112 genes (all of which were upregulated) that passed our two criteria of (1) expressing a five-fold difference after H2O2 exposure compared to untreated cells and (2) the maximum expression was two-fold different from a time point other than the untreated. The three clusters were generated using GeneSpring software based on their expression patterns (see Figure 3 for details)

Three clusters of temporal H2O2 induction profiles Gene clustering by temporal expression pattern is a powerful method to organize array data identifying genes that are coexpressed and participate in related pathways (Eisen et al., 1998). The coregulation of Oncogene

groups of genes provides information about the relation between previously known and newly identified genes. Using the GeneSpring expression analysis software (Silicon Genetics, CA, USA), we applied K-means clustering using a distance correlation to classify the 112 genes that passed our stringent filtering criteria


Normalized intensity (fold scale)

1139

Cluster 1

Cluster 2

Cluster 3

20

20

20

15

15

15

10

10

10

5

5

5

0

//

undam 0min 15min

//

6h

0

24h

//

undam 0min 15min

6h

//

0 24h

//

undam 0min 15min

6h

//

24h

Time Figure 3 k-Means clustering of the 112 genes that were significantly induced after H2O2 damage in the CS-B wt cell line. The genes were clustered into three predefined groups (GeneSpring software). The parameters used for k-means clustering were as follows: correlation type, distance; number of clusters. 3; maximum iterations, 100. Table 1 contains the genes found in each cluster. The genes in each cluster were not homogeneous with respect to function

based on their temporal expression profiles. The genes were clustered into three groups (Figure 3, Table 1). As a control, we classified the same genes using hierarchical clustering and the results are comparable (data not shown). The three groups were heterogeneous with respect to the function of genes. This was expected based on other array studies showing a similar variety of genes to be coexpressed (Zhao et al., 2000).

Functional groups involved in the H2O2 damage response Reactive oxygen species (ROS) activate various signal transduction pathways resulting in diverse cellular responses (Dalton et al., 1999). To elucidate the involvement of subgroups of genes, the H2O2 responsive genes from each cluster were classified into different functional groups (Table 1). Signal transduction genes were the largest single group of genes induced and belonged primarily to cluster 1 of genes that were only moderately induced (5–6-fold) after H2O2 treatment. In addition, genes involved in transcription/RNA metabolism and translation/protein metabolism as well as DNA damage response and apoptosis genes were affected. Genes encoding ribosomal proteins L21 and L31 were in cluster 2, and ribosomal proteins L7a and L5 in cluster 1. TCF3, an immunoglobulin transcription factor, was induced, indicating that elements of the damage response mimicked an inflammatory response. AEbinding protein 1 (AEBP1) is a transcriptional repressor that activates mitogen-activated protein kinase (MAPK) by complexing and protecting it (Kim et al., 2001), and has not previously been implicated in the DNA-damage response. The Cyclin D1 gene, which regulates G1/S transition, belongs to cluster 3, the most prominently induced group. CDC34, Chromosome condensation 1, Cyclin A2, and Cyclin H were all induced. GADD45, which inhibits entry of cells into S phase, was upregulated 13.8-fold after H2O2 treatment indicating that control of cell cycle progression after H2O2

treatment requires the coordinated regulation of several genes. Stress response and DNA repair genes were also induced after H2O2 treatment. GADD45 was upregulated and two HSP70 partners were in cluster 2. Interestingly, the BER gene encoding uracil DNA glycosylase (UDG) was upregulated over six-fold after H2O2. To our knowledge, there is no previous report of the inducibility of this gene. Our observation is in agreement with a recent finding that expression of the rat MutY DNA glycosylase homolog is strongly elevated in response to oxidative damage (Lee et al., 2002). To further investigate the regulation of BER genes, we searched the complete array gene 1ist for the following genes: APE, HAP, XRCC1, FEN1, polymerase b, PCNA, OGG1, and ligase 1. XRCC1 was upregulated two-fold at 0 min and PCNA 1.6-fold at 0 min (Suppl. Table 1, http://www.grc.nia.nih.gov/ branches/rrb/dna/dnapubs.htm). No other BER genes were present on the arrays. To further validate that stress induction occurred in the cells after H2O2 treatment, genes known to respond to oxidative stress were examined. A recognized indicator of cellular redox state is the heme catabolic enzyme, heme-oxygenase-1 (HO-1) (Keyse et al., 1990), which was originally observed in primary human cultured skin fibroblasts (Tyrrell and Basu-Modak, 1994). HO-1 induction occurs in most oxidative stressed mammalian cells (Keyse and Tyrrell, 1987). The heme oxygenase gene is increased in CS-B wt (Table 1) after H2O2 treatment. Some of the genes upregulated in human fibroblast by H2O2, which were also detected by the microarray method, are protein kinase B (Akt) (Huang et al., 2001) and MAPK (Guyton et al., 1996). Ornithine decarboxylase is increased in human fibroblasts after paraquat treatment (Kuo et al., 1995). In our system, ornithine decarboxylase mRNA was increased in wt cells (over six-fold). These genes have been shown to contribute to cell survival after oxidative stress through the activation of kinases. Oncogene


1140

Transcriptional defects in CS-B after H2O2 exposure Having investigated the response to H2O2 in the CS-B wt cells, we next examined the responses in CS-B mutant cell lines. Using isogenic cell lines, microarray analysis was used to screen the effects of a single genetic alteration on the expression of thousands of genes in response to oxidative stress. We compared the H2O2induced expression in the wt rescued cells with that of two cell lines of identical genetic background except for the CSB locus: the CBS-null cell line and the CS-B ATPase mutant where a site-specific mutation was introduced in motif II of the CSB gene (Figure 4). We have previously shown that steady-state expression of the genes on the array between these cell lines is not significantly different prior to DNA damage (Selzer et al., 2002); therefore, we normalized undamaged mRNA levels to one for all the cell lines. A main finding of this study is the extent to which overall transcriptional regulation after oxidative damage is altered in CS-B cells (Figure 5). In summary, the general temporal pattern of gene expression in the CS-B wt rescued cells can be interpreted as a transcriptional response to DNA damage followed by a return to near predamage mRNA levels (Figure 5a). The overall induction of expression following oxidative stress is impaired in CS-B cells and in the CS-B null cells; in particular, there is a delay in the transcriptional response. In the CS-B null cells, induction of expression at 0 min reaches approximately half the level of that in the wt rescued cell line, and the peak is delayed from 15 min to 6 h after damage (Figure 5b). The CS-B ATPase mutant cells appear to be defective in a way similar to the CS-B null cells, though notably there is no peak of 6 h after damage (Figure 5c). The overlapping expression patterns of the two CS-B mutant cell lines add statistical significance to our comparison with the wt rescued cell line. To identify individual genes whose expression levels differed significantly between different cell lines we applied a two-fold difference criteria. In this data set (CV ¼ 0.24), 11.2% of the genes would randomly display a twofold difference if no replicates had been done. However, with an average of 3.2 replicates, this frequency was reduced to a very conservative 0.1% (P ¼ 0.001). Thus, based on the variability in our data and the number of replicates, it was expected that 6.9 genes out of the 6912 assayed genes pass the significance criteria owing to random variability. Out of the 6912

genes assayed, 122 (1.8%) showed expression changes greater than two-fold when CS-B null or CS-B ATPase mutant cells were compared to wt rescued cells (Table 2). This group includes 23 (21%) of the 112 genes identified as H2O2 responsive in the wt rescued cell line (Figure 7). The remaining 99 genes were not identified in the wt response; however after oxidative damage, the mRNA levels displayed significantly different patterns of change when CS-B cell lines were compared with the wt rescued cell line. Relative to the number of genes (112) involved in the wt response, the differential regulation of as many as 122 genes in the CS-B cell lines indicates that CSB is required in the general processes controlling the efficiency of the transcriptional response to oxidative DNA damage likely via its ATPase domain. Key CSB-dependent genes identified by clustering of expression profiles The 122 genes showing differences in gene induction profiles between CS-B wt and the CSB mutant cell lines were classified into three groups by k-means clustering of their temporal expression profiles (Figure 6, Table 2). Hierarchical clustering was performed as a control and classified the genes in a similar way (data not shown). Genes of cluster 1 in Table 2 are those that show greater induction in both CS-B null and ATPase mutant cells than in the CS-B wt cells. Genes in cluster 2 are relatively less induced in both mutant cell lines compared to the wt. Those genes, in CS-B null cells, show a delay in induction. Genes in cluster 3 show a much greater difference between wt and CS-B null cells. The three distinct expression profiles of CSB-dependent genes revealed by the clustering process suggest that CSB affects different regulatory pathways, some of which are more critical to the oxidative damage response than others. Genes in cluster 3 are the ones whose transcription following oxidative damage is most clearly affected in CS. The 23 genes identified in the Venn diagram in Figure 7 pass both the criteria of significant upregulation in the wt rescued cells and differential expression in the CSB mutants relative to CS-B wt by a Xtwo-fold difference. Table 3 is a list of the 17 (out of 23) genes that have assigned names. The 17 genes include all nine genes in cluster 3 and eight genes from cluster 2 in Figure 6. The regulation of these genes by CSB is likely to be particularly important in the

CSB gene

NH2

ACIDIC

CSB ATPase mutant:

I IA

II III

638

IV V VI

COOH

DWHYVILDQGH648

Figure 4 Map of the CSB protein. The ATPase motif II amino-acid sequence with the point mutation is listed Oncogene


1141

c

b

a

11

1

// 24h

// 0min 15min 6h

//

// Undam

Undam

// 0min 15min 6h

Undam


21

24h

//

0min 15min 6h 24h

Time

Figure 5 Overview of the transcriptional response over 24 h to 250 mm H2O2 damage of the three cell lines; CS-B wt (a), CS-B null (b), CS-B ATPase mutant (c). The figure is an overlay of 6912 lines, each representing a gene. There is a delay in the mutants to return to predamage mRNA levels and both have approximately half the level of response that the wt cell produces. The mutants have similar expression patterns as the wt

response to cellular stress induced by H2O2. With respect to function, the CSB key regulated genes are heterogeneous and individual genes are discussed in other sections. Functional grouping of CSB-dependent genes To illuminate the cellular functions affected in the absence of the CSB protein within each cluster in Table 2, the genes were organized according to function. A set of genes belonging to the same functional categories as those in Table 1 are activated in CS-B null cells (cluster 1) indicating that the cells without CSB activated distinct pathways compared to wt rescued cells. Notably, seven genes implicated in carcinogenesis were upregulated in the CS-B null cells but not in CS-B wt cells (cluster 1). Clusters 2 and 3 contain the genes with reduced expression after H2O2 in CS-B null cells compared to the wt cells. With respect to function, these groups are dominated by genes involved in the stress response, transcription, translation, and signal transduction. It is clear from the results that the coordinated regulation of transcription following oxidative damage extends beyond proteins directly involved in the processing of DNA damage. However, CSB status did affect the expression of known heat-shock proteins and other stress response genes including UDG (Figure 8a), suggesting that CSB may play a role in the upstream events of the response to oxidative stress. UDG is a major protein in BER, where it functions as a glycosylase in the incision step of the repair process

(Krokan et al., 1997) and there has been no previous report indicating that it is induced after oxidative stress. Control of cell cycle progression is essential for successful DNA repair and the defective induction of CDC 34 and cyclin D1 (cluster 3), and cell cycle-related genes in cluster 2 confirm that core aspects of the stress response may be dependent on CSB status. In CS-B null cells we further observed a differential expression of genes encoding ribosomal protein suggesting a compromised translation from mRNA to protein (Figure 8b, c). Ribosomal proteins L21 and L31 belong to cluster 3, the group of genes displaying the most prominent lack of induction in CS, a finding that prompted further analysis of the microarray data with respect to genes encoding ribosomal proteins. Investigation of genes that did not pass the initial two-fold difference criteria revealed that a number of ribosomal proteins were regulated after H2O2 exposure (Table 4). The array analysis of the CS-B-transfected cell lines demonstrated that oxidative stress results in the modulation of gene expression for ribosomal proteins of both the large and small subunits, suggesting that ribosome structure and function, as well as protein synthesis in the mitochondria, may be modulated by CSB status in the cell. In other studies, evidence indicates that ribosomal protein expression is regulated during aberrant cellular growth and cancer in human cells (Wang et al., 2000). The role of CSB in the upregulation of ribosomal protein expression may be important to cellular homeostasis during the stress response to oxidative damage by impacting the translational machinery.

Discussion The use of isogenic CS-B stably transfected cell lines has enabled us to limit differences in genetic backgrounds in our analysis of the human CS-B cells. From the microarray analysis, we concluded that the transcriptional response to oxidative stress was compromised in the absence of CSB protein. The observation that differential expression in CS-B cell lines includes more than 100 genes and particularly several transcription factors and DNA-processing proteins suggests that CSB protein plays an upstream role in the transcriptional regulation following oxidative damage. We propose that the mechanisms underlying the repair deficiency and premature aging phenotype of CS include the observed downregulation of transcription of genes involved in protein turn-over, cell cycle regulation, and altered signal transduction in CS-B cells after oxidative DNA damage. In the wt rescued cell line, only upregulated genes passed the significance criteria. From a recent review of oxidative stress and gene regulation (Allen and Tresini, 2000), it is evident that upregulation of genes is more prevalent than repression. Applying less stringent criteria would have allowed identification of transcripts repressed by oxidative stress, but at the cost of Oncogene


1142 Table 2 Genes expressing differences in the CS-B null or ATPase mutant cells when compared to the wt cell line (ESTs not listed) C lust er 1

M aximum d if f er ence in g ene exp r essi o n af t er H2 O 2 exp o sur e r elalt i ve t o w t r escued cell l ine

Genbank

null

ATPase Gene Name

Genbank

null

AA461231

2,1

1,4

Translin

N62179

1,5

R17654 AA459213 W72621 AA459263 AA057436

2,5 2,4 2,6 2,3 2,3

1,5 2,3 1,6 2,0 1,1

KIAA0376 protein Cyclin A2 RING1 and YY1 binding protein BCL2-related protein A1 Regulatory factor X-assoc. prot ein

T73556 AA460319 W19461 AA486332 AA488609

1,4 3,2 2,5 2,3 2,3

AA292583

2,2

1,6

TAF2H

AA457097

2,0

H46553 AA486221 AA019459 AA521243

2,1 2,0 2,9 2,6

1,4 1,3 1,6 1,4

Transcription factor 8 Poly(A)-binding protein Prot ein tyrosine kinase 9 M itochondrial ribosomal protein L19

R32952 H29407 T98612 AA455523

2,0 1,7 2,7 2,2

T55931

2,2

1,3

Solute carrier family 20 (phosphate)

AA457118

2,1

R52852 N20798 AA428749

2,2 2,2 2,1

1,8 1,9 2,2

G prot ein-coupled receptor 37 V-kit Hardy-Zuckerman 4 feline Prot ein phosphat ase 1, regulat ory

AA481585 AA495746 W15263

2,1 2,1 2,1

R15708

2,1

1,9

Insulin-like 4 (placenta)

AA258396

2,0

AA464470 N50073

2,1 2,1

2,0 1,2

Valyl-t RNA synthet ase 2 Butyrate-induced transcript 1

AA456156 R76281

2,0 2,0

H80359

2,0

1,7

PIP5K2B (kinase)

AA449037

1,0

AA126356 AA485677

1,5 1,1

2,2 2,4

Calnexin Thyroid hormone receptor interactor 6

ATPase Gene Name M ethylmalonate-semialdehyde 2,1 dehydrogenase 2,0 Fatty-acid-Coenzyme A ligase 2,1 LIM D1 1,1 AIP-1( abl-interactor 12) 1,2 PTOV1 1,3 Nucleoporin 88kD V-akt murine thymoma viral oncogene 1,7 homolog 2 2,1 S100 calcium-binding prot ein P 2,0 LIV-1 protein, estrogen regulated 3,9 Collagen, type III, alpha 1 1,7 CD39-like 2 Translocase of out er mitochondrial 2,7 membrane 34 (yeast) homolog 1,1 Probe hTg737 1,7 Phosphatidylinosit ol glycan 1,3 Glycoprot ein M 6B Pleckstrin homology-like domain, 1,0 family A, member 1 1,8 Uroporphyrinogen III synthase 1,6 Solute carrier family 31 (copper) Capping protein (act in filament) 2,0 muscle Z-line, alpha 1

C lust er 2


Genbank AA487452 H15111 H09997

null -2,2 -2,2 -2,2

ATPase -1,5 -2,5 -2,1

AA446839

-2,1

-2,2

H93393

-2,0

-1,7

AA608557

-1,3

-2,1

AA504943

-1,1

Gene Name DNA fragmentation factor Uracil-DNA glycosylase M atrix metalloproteinase 16 BCL2/ adenovirus E1B 19kD-interacting prot ein 3 Fusion, derived from t(12;16) malignant liposarcoma Damage-specific DNA binding prot ein 1 (127kD)

Genbank AA480880 AA019316 H77855

null ATPase Gene Name -2,9 -1,4 Butyrate(EGF) response factor 2 -2,7 -1,2 Chloride channel 4 -2,2 -1,1 G protein-coupled receptor 48

AA453275

-1,4

-2,4

Integral membrane prot ein 2B

W23931

-1,3

-2,1

SH2-B homolog

H23421

-3,1

-4,2

Ribosomal protein L7a

-2,1

Crystallin, alpha B

AA455267

-2,5

-1,4

Vacuolar protein sorting 29

-3,7

Chromosome condensation 1

AA070997

-2,3

-2,7

Proteasome subunit, beta type, 6 Ribosomal protein S5 Ubiquinol-cytochrome c reductase hinge prot ein 3-hydroxybutyrate dehydrogenase Pyrroline-5-carboxylate synthetase Quinone oxidoreduct ase homolog Cytochrome c oxidase subunit Via M HC class I region ORF KIAA0300 protein Guanylate binding protein 1, Cytochrome P450, 51 GAS2-related on chromosome 22 Profilin 1

R93875

no value -2,5

AA490263

-2,3

T67029 W45690 AA026102 AA112660 AA410435 AA029041 T68202 AA291995 R36006 R96220

-2,2 -2,0 -2,5 -2,4 -2,0 -1,5 -1,4 -1,2 -3,2 -2,9

C lust er 3


Genbank

null

AA291398

AA481758 AA236164 H20743 AA487700 AA284408

-1,5 no value -1,8 -2,2 -2,3 -2,2 -2,2 -2,0 -2,1 -2,1 -2,2 -1,8

Nucleosome assembly prot ein 1-like

AA456616

-1,5

-2,0

NIM A-related kinase 3

R11698

-2,7

-2,8

Ornithine decarboxylase antizyme 2 M itogen-activated protein kinase 1 Transcription factor 3 Forkhead box F1 ATRX (RAD54 homolog) Seven in absentia homolog 2 RNA binding motif protein 10 Cleavage stim. factor, 3' pre-RNA Thyroid receptor interacting prot. 15 Guanine nucleotide binding protein

T67058 AA143509 AA668595 AA482243 T58146 AA405458 AA486849 AA477893 R35292 AA521431

-2,2 -1,5 -1,3 -1,2 -2,1 -2,0 -1,8 -1,4 -1,2 -1,2

-2,2 -2,0 -2,1 -2,1 -2,0 -2,1 -2,0 -2,3 -2,1 -2,6

ATPase Gene Name DnaJ (Hsp40) homolog, subfamily B, -3,6 member 1 -1,1 -2,2 Cathepsin S -3,5 -3,4 Cell division cycle 34 -1,4 -2,8 Cyclin D1 -1,4 -2,6 Cut (Drosophila)-like 1 -1,5

DNA damage, stress response, apoptosis Replicat ion, cell cycle Transcription, RNA metabollism Signal transduction

Genbank

null

R73545

-1,3

ATPase Gene Name -2,9

Flotillin 2

AA460480 W15277 AA464034

-1,5 -3,8 -3,4

-2,8 -4,1 -3,4

Creatine kinase, mitochondrial 2 Ribosomal protein L31 Ribosomal protein L21

Protein met abolism Energy metabolism Cancer related Other

a Clusters were generated by GeneSpring using k-means clustering (see Figure 6 for details). Genes were clustered according to (1) being upregulated compared to the wt, (2) upregulated in the wt, upregulated with a delay in the CS-B null mutant, and less upregulated in the ATPase mutant, and (3) upregulated more in the wt than in either of the mutants

introducing more false positives. The lesser magnitude of repression is illustrated in a study of protein expression after H2O2 exposure in Saccharamyces Oncogene

cerevisiae (Godon et al., 1998), where 115 induced and 52 repressed proteins were identified. However, of the 44 analysed repressed proteins 14 (32%) were repressed by


1143

Cluster 3 10

9

8

7

CS-B wt CSB null CSB ATPase mutant

6

5

4

4

3

3

2

2

2

1

1

1

//

//

//

0min 15min 6h 24h

Undam

//

0min 15min 6h 24h

Undam

Cluster 1

Undam


Cluster 2

//

//

0min 15min 6h 24h

Time Figure 6 Average expression profiles of k-means generated clusters after H2O2 damage comparing the wt and the mutant cell lines. Cluster 1 is the genes that were upregulated in the mutants compared to the wt, cluster 2 are those genes that were upregulated in all three cell lines but delayed at the 6 h time point in the null mutant, and cluster 3 are the genes upregulated in the wt compared to the mutants. The parameters used for k-means clustering were as follows: correlation type, distance; number of clusters, 3; maximum iterations, 100. Table 2 contains the gene list from each cluster

A

89

C 23

B

99

Figure 7 Venn diagram representing the number of genes overexpressed (Table 1) in the CS-B (wt) cell line (A-genes in Table 1) after H2O2 damage or deficient in the two mutant cell lines (CS-B null and CS-B ATPase mutant) or (B-genes in Table 2) compared to the wt after damage (Table 2). Panel (C) is the overlap between the two. This number includes the ESTs as well as the genes in (C). See Table 3 for the list of genes (not including the ESTs)

less than two-fold and only 15 (34%) by more than three-fold. In contrast to the consistent expression profiles seen in the wt rescued cell line after oxidative stress, CS-B null and CS-B ATPase mutant cell lines displayed a pattern of reduced and delayed induction of a specific subset of genes. Of the analysed genes, 122 (1.8%) were

significantly differentially expressed in the CS-B mutant cell lines when compared to the wt rescued cell line. Genes involved in transcriptional regulation were differentially regulated in CS-B null cell lines and it is likely that their regulation occurs upstream in a network involving other genes identified in this study. One implication of this is that determining the relations between the expression of individual genes will be possible only after the accumulation of expression data from multiple array studies examining different conditions. Expression of the gene encoding the stress response protein Hsp40 (DnaJ) that interacts with and stimulates Hsp70 was less induced in the CS-B cells than in wt. Hsp40 is a member of the heat-shock protein family that includes many proteins involved in regulating folding of other proteins, thereby serving to protect proteins from various forms of cellular stress including heat shock and oxidative stress (Feder and Hofmann, 1999). A number of heat-shock proteins and chaperones display changes in their mRNA levels after oxidative stress. Recently, it was reported that mammalian SWI/SNF complexes, Oncogene


1144 Table 3

List of genes in group C from Figure 7 (ESTs not listed)

M aximum difference in gene expression aft er H2O2 exposure relaltive t o wt rescued cell line Genbank

null

H15111 -2,2 AA481758 -1,5 AA504943 -1,1 AA236164 -1,1 H20743 -3,5 T67029 -2,2 AA487700 -1,4 AA291398 no value AA026102 -2,5 AA284408 -1,4 H77855 -2,2 R73545 -1,3 W15277 -3,8 AA464034 -3,4 H23421 -3,1 R11698 -2,7 AA460480 -1,5

ATPase -2,5 -3,6 -2,1 -2,2 -3,4 -1,8 -2,8 -3,7 -2,3 -2,6 -1,1 -2,9 -4,1 -3,4 -4,2 -2,8 -2,8

Gene Name Uracil-DNA glycosylase DnaJ (Hsp40) homolog, subf amily B, Crystallin, alpha B Cathepsin S Cell division cycle 34 Ornithine decarboxylase antizyme 2 Cyclin D1 Chromosome condensation 1 Transcription factor 3 Cut (Drosophila)-like 1 G protein-coupled receptor 48 Flot illin 2 Ribosomal protein L31 Ribosomal protein L21 Ribosomal protein L7a Ubiquinol-cytochrome c reduct ase Creatine kinase, mitochondrial 2

DNA damage, stress response, apopt osis Replication, cell cycle Transcription, RNA met abollism Signal transduction Protein metabolism Energy metabolism

a Genes that have the same profile as cluster 3 (Figure 6) and were both upregulated in the wt cells and were differentially expressed by the mutants compared to the wt by a twofold or greater margin. These genes appear to be important to the cellular response to H2O2 exposure

related to CSB by amino-acid sequence homology, contribute to activation of the hsp70 gene in response to certain forms of cellular stress (Finkel and Holbrook, 2000). HSF2, the most conserved heat-shock transcription factor, has been shown to be induced when the ubiquitin–proteasome pathway is impaired (Mathew et al., 1998). The array data from our study demonstrated 9.8-fold Hsp40 overexpression in a wt rescued cell line after exposure to H2O2. The significantly reduced induction of the Hsp40 gene in both CS-B null (1.5-fold less than wt rescued) and CS-B ATPase mutant (3.6-fold less than wt rescued) suggests that a functional CSB protein is required for the appropriate transcriptional upregulation of Hsp40. The evidence presented here suggests that the inability to elicit the appropriate transcriptional response in the CS-B defective cell lines contributes to the compromised cellular response to oxidative damage exposure, culminating in cellular dysfunction and pathogenesis. This notion is supported by the finding that elevating the level of the Hsp40-associated Hsp70 results in increased survival of stressed cells and animals, whereas inhibiting Hsp70 induction reduces survival after stress treatment (Morimoto and Santoro, 1998). H2O2 has been shown to prevent ubiquitination and degradation of the epidermal growth factor (EGF) receptor (De Wit et al., 2001). This is of interest because of the proteasome class of genes that is differently Oncogene

regulated in CBS null, ATPase mutant, and wt. One of the roles of the proteasome degradation pathway is to eliminate damaged or oxidized proteins in cells (Sitte et al., 1998). Therefore, it is expected that proteasome genes may be affected after oxidative stress and in aging (Sitte et al., 2000). Furthermore, it has been reported that ubiquitination of UV-induced Pol II is deficient in CS-B fibroblasts (Bregman et al., 1996), demonstrating that CSB may affect the ubiquitin– proteasome pathway. The proteasome subunit PSMB6 mRNA level was less induced in mutant (2.7-fold) and vector (2.3-fold) CS-B cells compared to wt cells. Similarly, genes encoding other proteins involved in the ubiquitin proteasome pathway were induced to a lesser extent (between 1.8 and two fold less), for example, ubiquitin protein ligase E3A, proteasome subunits alpha type 1, 3, and beta unit 10, ubiquitin-conjugating enzyme E2D 3, and ubiquitin-specific protease 13 (not in table). The accumulation of unwanted proteins may be detrimental to the proper functions of the cell, thus contributing to the damage-sensitive phenotype of CS. In our search for relevant genes whose expression is dependent upon CSB, certain genes were of particular interest. UDG is one of the most abundant BER enzymes, and is responsible for the removal of uracil in DNA (Krokan et al., 2001). We find that UDG mRNA is induced after H2O2 in the wt cells, suggesting


1145


a

6 CS-B wt CSB null CS-B ATPase mutant

5 4 3 2 1 undam

0 min

//

15 min

6h

//

24 h

Time


b 10

CS-B wt CSB null CS-B ATPase mutant

c 10

8

8

6

6

4

4

2

2

//

undam 0 min 15 min

6h

//

24 h

undam 0 min

15 min

//6 h // 24 h

Figure 8 Average expression profile in the three cell lines of (a) the uracil DNA glycosylase (UDG) gene (Table 2) and two ribosomal genes: Panels (b) L31 and (c) L21 (Table 3). Note that the induction is significantly greater in the wt rescued cell line compared to the mutants suggesting a role for CSB in the induction. Both ribosomal genes encode components of the 60S subunit and L21 is part of the mammalian mitochondrial ribosome Table 4 List of ribosomal genes found on the array that were up or downregulated after H2O2 exposure Genbank

N76229 AA496838 W15277 AA464034 AA486919 H23421 AA456616 T55092 AA411107 AA127116 AA083577 AA453015 T69519

Gene name

Ribosomal protein L39 Ribosomal protein L5 Ribosomal protein L31 Ribosomal protein L21 Ribosomal protein L28 Ribosomal protein L7a Ribosomal protein S5 Small nuclear ribonucleoprotein Small nuclear ribonucleoprotein Heterogeneous nuclear ribonucleoprotein A1 Ribosomal protein L 19 Mitochondrial ribosomal protein U2 small nuclear ribonucleoprotein

Maximal overexpression after damage CS-B wt

CS-B null

+++ +++ +++ +++ ++ ++ ++ ++ ++ ++ + + +

++ ++ ++ ++ + + + + + + +

CS-B ATPase mutant ++ ++ + + + + + + + +

Approximate expression levels: +++: 6–10-fold overexpressed; ++: 3-5-fold overexpressed; +: three-fold overexpressed a Most were underexpressed in the mutants compared to the wt suggesting that the CSB may upregulate ribosomal protein gene expression after oxidative stress

that this form of repair is inducible, and we also find that CSB is involved in the regulation of UDG gene expression. To our knowledge, there has been no

previous report of induction of UDG after DNA damage, but there are observations that other BER enzymes can be induced (Hollenbach et al., 1999). Oncogene


1146

Cell cycle arrest plays an important role in regulating the DNA damage response. We found that cyclin D1 expression was induced significantly by H2O2 treatment (250 mm). However, this induction was significantly reduced in the CS-B null and ATPase domain mutant, suggesting a critical role of CSB in G1 phase cell cycle progression in response to DNA damage. Cyclin D1 is generally the rate-limiting step for its association with cyclin-dependent kinase-4 and -6 in G1 phase. The formation of this complex phosphorylates and inactivates the retinoblastoma protein (Rb), which, in turn, results in the expression of E2F-regulated genes that are required for G1 to S phase transition (Weinberg, 1995). In agreement with Weber et al. (1997) who showed that the induction of cyclin D1 required MAPK1’s activation, we found a correlation between MAPK1 and cyclin D1 expression. The induced expression of the cyclin D1 gene by H2O2 in the wt is much greater than that of MAPK1 (Table 1), but the trend was significantly attenuated by the CSB mutation. This result could suggest a role of CSB in maintaining cellular MAPK1 and cyclin D1 levels to precede the arrested cell cycle in response to DNA damage. A direct role of CSB in transcription has previously been suggested by in vitro and in vivo studies (Dianov et al., 1999; Balajee and Bohr, 2000). More recently, evidence supports a requirement for yeast RAD26, the homolog of the human CSB gene, in elongation by RNA polymerase II (Lee et al., 2001). Although RAD26 has a role in transcription elongation in the absence of DNA damage in yeast, it is possible that the human CSB protein may also impact transcription during a stress response following treatment with an oxidizing agent, such as H2O2. Another important observation in our study is that the transcription of a number of ribosomal genes appears to be regulated by CSB. It is possible that some of the clinical signs and symptoms of CS are a result of defects in translation. So far, there is no evidence for a role of CSB in rDNA transcription by Pol I, but only in Pol II-dependent transcription. In the light of our observations here, it would be important to revisit this question and examine ribosomal transcription in CS-B cells. A model to explain why a CSB deficiency results in locus-specific fragility of genes encoding highly structured RNAs was recently proposed (Yu et al., 2000). CSB could function as an elongation factor for RNA pol I or RNA pol II transcribed genes that encode highly structured RNAs. This is consistent with other evidence (Balajee et al., 1997; Selby and Sancar, 1997), which indicate that CSB serves as a transcription elongation factor. Regulation of the expression of genes encoding ribosomal proteins would directly modulate the protein translation machinery, ultimately providing a response mechanism to the stress. A decline in protein synthesis accompanies aging, and is likely to contribute to ageassociated disease. In general, there is a decrease in enzyme biosynthesis and protein turnover by the ubiquitin–proteasome pathway with aging. In old cells, Ly et al. (2000) demonstrated that genes involved in Oncogene

protein processing were decreased compared to young cells, such as the cyclin-selective ubiquitin carrier protein, 26S proteasome subunit p44.5, proteasome subunit p55, HC3, HC8, p42, and the 26S proteasome subunit. Our results concurred with this observation that genes involved in protein processing were less induced in the vector and mutant cells. A diminished signal transduction is a hallmark of senescence (Wheaton et al., 1996; Bootcov et al., 1997), and our findings indicate that the downstream effects of CSB are mediated by differential expression of signal transduction genes as well as by direct regulation of transcription and translation. Implication of the transforming growth factor-b (TGF-b) signalling pathway in DNA mismatch repair has been suggested (Parsons et al., 1995). The binding of TGF-b to its receptor on cell membranes is required for the TGF-b signalling pathway. We found that the mRNA levels of MIC-1, a member of the TGF-b superfamily, and type III TGF receptor b were both increased after oxidative stress. Recently, the MIC-1 promoter region has been shown to be a target for p53 protein and cause G 1 arrest (Tan et al., 2000) suggesting a role of MIC-1 in stress-induced cellular responses. The mRNA levels of MIC-1 and type III TGF receptor b were much less increased in the CS-B mutant cells, indicating that the activity of CSB protein might be involved in the H2O2-induced DNA mismatch repair. MAP kinase has been termed stress-activated protein kinase (SAPK) because of its activation by chemicals, heat, osmotic shock, ultraviolet light radiation, and inhibitors of protein synthesis (Kyriakis et al., 1994). In the stress response, JNK1/SAPK (MAP kinase) is activated after H2O2 treatment in human fibroblast, but not in CS-B cells (Dhar et al., 1996). MAPK8 is activated after H2O2 treatment in human fibroblast but not in CS-B cells (Dhar et al., 1996). In the wt rescued cell line, three genes whose products promote dephosphorylation were induced: protein phosphatase 1, protein tyrosine phosphatase IV, and protein kinase inhibitor alpha (Table 1). Protein tyrosine kinase 9 expression was increased in CS-B null cells simultaneously with protein phosphatase 1 inhibitor and other genes involved in signal transduction. Thus, while cellular homeostasis requires both positive and negative regulation of kinase activity, we found that in some cases CSB status affects the expression of genes whose products are needed for dephosphorylation and kinase inhibition, possibly halting normal cellular processes while DNA repair takes place. There has been considerable discussion about whether the complex clinical phenotype of CS is because of a primary defect in DNA repair or transcription. The CS defect in TCR of certain types of DNA damage such as UV-induced lesions does not explain the characteristic growth and neurological deficiencies associated with the disease. Our observations would argue that a major defect in this disease lies in the transcriptional response to oxidative stress. In conclusion, by analysing the dynamics of gene expression following cellular stress, we have identified a number of genes not previously known


1147

to be associated with CS or aging. The large numbers of genes affected provide new evidence that CSB protein plays a central role in DNA and RNA processing and the variety of genes affected correlate with the diverse cellular and clinical phenotypes of CS.

Materials and methods Cell lines and culture conditions Three previously described cell lines were derived from CS1AN.S3.G2, a CS complementation group B SV40-transformed human fibroblast cell line (Selzer et al., 2002). Briefly, CS1AN.S3.G2 was transfected with the mammalian expression vector pc3.1 (In vitrogen, San Diego, CA, USA) containing the wt human CSB gene (designated CS1AN/ pc3.1-CSBwt), the CSB wt gene with a point mutation in the ATPase/helicase motif II region (CS1AN/pc3.1-CSBE646Q) (Figure 4) or with the vector alone (CS1AN/pc3.1). Stable isogenic transfected cell lines isolated from individual clones were characterized for CSB protein expression and genetic function (Selzer et al., 2002). Cells were cultured in minimal essential media supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin, 1% l-glutamine, and Geneticin G418 (400 mg/ml) (all components from Life Technologies, Gaithersburg, MD, USA). Cells were treated with 250 mm H2O2 (Sigma, St Louis, MO, USA) for 15 min in serum-free medium and RNA was isolated before H2O2 exposure, and at specific time points after H2O2 exposure (0 min, 15 min, 6 h, and 24 h). Duplicate RNA samples were isolated from two separate populations of cells grown and treated identically but on different days. RNA was isolated with RNA STAT-60 (Tel-Test, Friendswood, TX, USA) and following the procedure from the Clontech Expression Array Manual (Palo Alto, CA, USA), DNase treated to remove any contaminating DNA. All samples were tested for DNA contamination with RT–PCR using a betaactin primer (Ambion, Austin, TX, USA). cDNA microarrays To demonstrate reproducibility and increase statistical significance, each of the two RNA samples per experimental condition were hybridized twice to different arrays producing four replicate data points per experimental condition. Probes were labelled as follows: 1 mg of oligo-dT (Research Genetics, Huntsville, AL, USA) was mixed with 5 mg of total RNA and this was brought to a final volume of 15 ml with DEPC-treated water. The mixture was heated to 701C for 10 min and then placed on ice. Then, 8 ml 5 first strand buffer, 4 ml 0.1 m DTT, 1 ml RNAseOUT (all three from Life Technologies), 4 ml 0.5 mm dNTP (minus dCTP), 5 ml [33P] dCTP (Amersham, Piscataway, NJ, USA), 16 ml of H2O, and 2 ml Superscript II reverse transcriptase (Life Technologies) were added. The reaction was incubated for 35 min at 421C followed by the addition of 2 ml of Superscript and another 35 min incu bation at 421C. RNA was removed with 5 ml 0.5 m EDTA and 10 ml 0.1 n NaOH followed by incubation at 651C for 30 min and addition of 25 ml of 1 m Tris-HCl (pH 8). The probes were then purified through Biospin P-30 columns (BioRad, Hercules, CA, USA) following the manufacturer’s instructions. cDNA arrays containing 6912 genes and ESTs printed on three nylon filter membranes (www.grc.nia.nih.gov/branches/ rrb/DNA/DNA.htm) were produced essentially as previously described (Tanaka et al., 2000). For prehybridization (501C,

4 h), 100 ml Human Cot 1 DNA (Life Technologies), 8 mg/ml poly A (Sigma, St Louis, MO, USA) and 10 mg/ml sheared salmon sperm (Sigma) was heat denatured at 951C for 5 min, then added to 4 ml of Microhyb (Research Genetics, Huntsville, AL, USA). The probes were denatured (951C, 5 min) before their addition to the arrays and then hybridized overnight. Membranes were washed three times at room temperature with 2 SSC/0.1% SDS and two times then with 1 SSC/0.1% SDS. The arrays were then exposed overnight on a phosphoimager screen and scanned on a Molecular Dynamics STORM phosphoimager (Amersham, Piscataway, NJ, USA). Verification by Northern analysis A volume of 10 mg of total RNA was electrophoresed on a 1% agarose, 1.11% formaldehyde gel and transferred to HybondXL membrane (Amersham) using a PosiBlot pressure blotter (Stratagene, La Jolla, CA, USA). The membrane was prehybridized with ExpressHyb hybridization solution (Clontech). Probes for the genes of interest were generated via PCR from either pT3T7-pac or pT7T3-D plasmid containing the same cDNA for the genes as found on the arrays. Hybridization took place overnight at 421C and membranes were washed two times with 2 SSC/0.1% SDS and two times with 0.2 SSC/0.1% SDS. Data analysis Spot intensity was quantitated using Image Quant analysis software (Amersham) and subsequently analysed with Excel (Microsoft) and GeneSpring. Array scans were inspected visually and arrays with local nonuniform background were excluded from the analysis. Nonspecific uniform background across entire arrays because of experimental variation was normalized using global normalization. The data value for each spot on each membrane was divided by the average intensity value of that membrane to obtain a normalized intensity value. The median of the four replicate data points per experimental condition was used in the subsequent analysis. A gene-to-gene normalization was applied in the analysis when calculating ratios between expression values before and after damage and between different cell lines. Ratio values were generated in Excel, and data were visualized and clustered using GeneSpring. GeneSpring’s k-means clustering algorithm divides genes into a user-defined number (k) of equal-sized groups. It then creates centroids (in expression space) at the average location of each group of genes. With each iteration, genes are reassigned to the group with the closest centroid. After all of the genes have been reassigned, the location of the centroids is recalculated and the process is repeated until the maximum number of iterations has been reached. The goal is to produce groups of genes with a high degree of similarity within each group and a low degree of similarity between groups. In a time series experiment, kmeans clustering can identify unique classes of genes that are upregulated or downregulated in a time-dependent manner. For each clustering analysis, we empirically determined the number of clusters giving the best visual separation of clusters.

Abbreviations: CS, cockayne syndrome; CSB, cockayne syndrome group B; CV, coefficient of variation; Wt, wild type CS-B wt, CS1AN/ pc3.1-CSBwt; CS-B null, CS1AN/pc3.1; CS-B ATPase mutant, CS1AN/pc3.1-CSBE646Q; TCR, transcription-coupled repair; Oncogene


1148 BER, base excision repair; UDG, uracil DNA glycosylase; NER, nucleotide excision repair.

Acknowledgements We thank Chris Cheadle for statistical help and discussions. Complete results for all genes on the arrays can be found at the

following website: http://www.grc.nia.nih.gov/branches/rrb/ dna/dnapubs.htm. This work was supported in part by the Danish Center for Molecular Gerontology, the Danish Medical Research Council (9902876), the Danish Cancer Society (9914403 9132/99 14403), Fonden til Lægevidenskabens Fremme and Direktr E Danielsens Fond.

References Allen RG and Tresini M. (2000). Oxidative stress and gene regulation. Free Radic. Biol. Med., 28, 463–499. Balajee AS and Bohr VA. (2000). Genomic heterogeneity of nucleotide excision repair. Gene, 250, 15–30. Balajee AS, May A, Dianov GL, Friedberg EC and Bohr VA. (1997). Reduced RNA polymerase II transcription in intact and permeabilized Cockayne syndrome group B cells. Proc. Natl. Acad. Sci. USA, 94, 4306–4311. Bootcov MR, Bauskin AR, Valenzuela SM, Moore AG, Bansal M, He XY, Zhang HP, Donnellan M, Mahler S, Pryor K, Walsh BJ, Nicholson RC, Fairlie WD, Por SB, Robbins JM and Breit SN. (1997). MIC-1, a novel macrophage inhibitory cytokine, is a divergent member of the TGF-beta superfamily. Proc. Natl. Acad. Sci. USA, 94, 11514–11519. Bregman DB, Halaban R, Van bool AJ, Henning KA, Friedberg EC and Warren SL. (1996). UV-induced ubiquitination of RNA polymerase II: a novel modification deficient in Cockayne syndrome cells. Proc. Natl. Acad Sci. USA, 93, 11586–11590. Cao SX, Dhahbi JM, Mote PL, Spindler SR. (2001). Genomic profiling of short- and long-term caloric restriction effects in the liver of aging mice. Proc. Natl. Acad. Sci. USA, 98, 10630–10635. Citterio E, Van Dem Boom, Schnitzler G, Kvanaar R, Bonte E, Kingston RE, Hoeijmakers JH and Vermeulen W. (2000). ATP-dependent chromatin remodeling by the cockayne syndrome B DNA repair-transcription-coupling factor. Mol. Cell. Biol., 20, 7643–7653. Dalton TP, Shertzer HG and Puga A. (1999). Regulation of gene expression by reactive oxygen. Annu. Rev. Pharmacol. Toxicol., 39, 67–101. De Wit R, Makkinje M, Boonstra J, Verkleij AJ and JA. (2001). Hydrogen peroxide reversibly inhibits epidermal growth factor (EGF) receptor internalization and coincident ubiquitination of the EGF receptor and Eps15. FASEB J., 15, 306–308. Dhar V, Adler V, Lehmann A and Ronai Z. (1996). Impaired jun-NH2-terminal kinase activation by ultraviolet irradiation in fibroblasts of patients with Cockayne syndrome complementation group B. Cell Growth Differ., 7, 841–846. Dianov G, Bischoff C, Sunesen M and Bohr VA. (1999). Repair of 8-oxoguanine in DNA is deficient in Cockayne syndrome group B cells. Nucleic Acids Res., 27, 1365–1368. Eisen MB, Spellman PT, Brown PO and Botstein D. (1998). Cluster analysis and display of genome-wide expression patterns. Proc. Natl. Acad. Sci. USA, 95, 14863–14868. Feder ME and Hofmann GE. (1999). Heat-shock proteins, molecular chaperones, and the stress response: evolutionary and ecological physiology. Annu. Rev. Physiol., 61, 243–282. Finkel T and Holbrook NJ. (2000). Oxidants, oxidative stress and the biology of ageing. Nature, 408, 239–247. Oncogene

Friedberg EC. (1996). Cockayne syndrome – a primary defect in DNA repair, transcription, both or neither? BioEssays, 18, 731–738. Friedberg EC, Walker GC and Siede W. (1995). DNA Repair and Mutagenesis. ASM Press: Washington, DC. Godon C, Lagniel G, Lee J, Buhler JM, Kieffer S, Perrot M, Boucherie H, Toledano MB, Labarre J. (1998). The H2O2 stimulon in Sacharomyces cerevisiae. J. Biol. Chem., 273, 22480–22489. Guyton KZ, Liu Y, Gorospe M, Xu Q and Holbrook NJ. (1996). Activation of mitogen-activated protein kinase by H2O2. Role in cell survival following oxidant injury. J. Biol. Chem., 271, 4138–4142. Henning K, Li L, Legerski R, Iyer N, McDaniel L, Schultz R, Stefanini M, Lehmann A, Mayne L, Friedberg E. (1995). The Cockayne syndrome complementation group A gene encodes a WD-repeat protein which interacts with CSB protein and a subunit of the RNA pol II transcription factor IIH. Cell, 82, 555–566. Hollenbach S, Dhenaut A, Eckert I, Radicella JP and Epe B. (1999). Overexpression of Ogg1 in mammalian cells: effects on induced and spontaneous oxidative DNA damage and mutagenesis. Carcinogenesis, 20, 1863–1868. Huang C, Li J, Ding M, Leonard SS, Wang L, Castranova V, Vallyathan V and Shi X. (2001). UV induces phosphorylation of protein kinase B (Akt) at Ser-473 and Thr-308 in mouse epidermal Cl 41 cells through hydrogen peroxide. J. Biol. Chem., 276, 40234–40240. Kayo T, Allison DB, Weindruch R and Prolla TA. (2001). Influences of aging and caloric restriction on the transcriptional profile of skeletal muscle from rhesus monkeys. Proc. Natl. Acad. Sci. USA, 98, 5093–5098. Keyse SM, Applegate LA, Tromvoukis Y and Tyrrell RM. (1990). Oxidant stress leads to transcriptional activation of the human heme oxygenase gene in cultured skin fibroblasts. Mol. Cell. Biol., 10, 4967–4969. Keyse SM and Tyrrell RM. (1987). Both near ultraviolet radiation and the oxidizing agent hydrogen peroxide induce a 32-kDa stress protein in normal human skin fibroblasts. J. Biol. Chem., 262, 14821–14825. Kim SW, Muise AM, Lyons PJ and Ro HS. (2001). Regulation of adipogenesis by a transcriptional repressor that modulates MAPK activation. J. Biol. Chem., 276, 10199–10206. Krokan HE, Otterlei M, Nilsen H, Kavli B, Skorpen F, Andersen S, Skjelbred C, Akbari M, Aas PA and Slupphaug G. (2001). Properties and functions of human uracil-DNA glycosylase from the UNG gene. Prog. Nucleic Acid Res. Mol. Biol., 68, 365–386. Krokan HE, Standal R and Slupphaug G. (1997). DNA glycosylases in the base excision repair of DNA. Biochem. J., 325, 1–16. Kuo ML, Lee KC, Lin JK and Huang TS. (1995). Pronounced activation of protein kinase C, ornithine decarboxylase and


1149 c-jun proto-oncogene by paraquat-generated active oxygen species in WI-38 human lung cells. Biochim. Biophys. Acta, 1268, 229–236. Kyriakis JM, Banerjee P, Nikolakaki E, Dai T, Rubie EA, Ahmad MF, Avruch J and Woodgett JR. (1994). The stressactivated protein kinase subfamily of c-Jun kinases. Nature, 369, 156–160. Lee CK, Klopp RG, Weindruch R and Prolla TA. (1999). Gene expression profile of aging and its retardation by caloric restriction. Science, 285, 1390–1393. Lee CK, Weindruch R and Prolla TA. (2000). Gene-expression profile of the ageing brain in mice. Nat. Genet., 25, 294–297. Lee HM, Wang C, Hu Z, Greeley GH, Makalowski W, Hellmich HL and Englander EW. (2002). Hypoxia induces mitochondrial DNA damage and stimulates expression of a DNA repair enzyme, the Escherichia coli MutY DNA glycosylase homolog (MYH), in vivo, in the rat brain. J. Neurochem., 80, 928–937. Lee SK, Yu SL, Prakash L and Prakash S. (2001). Requirement for Yeast RAD26, a homolog of the human CSB gene, in elongation by RNA polymerase II. Mol. Cell. Biol., 21, 8651–8656. Ly DH, Lockhart DJ, Lerner RA and Schultz PG. (2000). Mitotic misregulation and human aging. Science, 287, 2486– 2492. Mathew A, Mathur SK and Morimoto RI. (1998). Heat shock response and protein degradation: regulation of HSF2 by the ubiquitin-proteasome pathway. Mol. Cell. Biol., 18, 5091–5098. Miller RA, Galecki A and Shmookler-Reis RJ. (2001). Interpretation, design, and analysis of gene array expression experiments. J. Gerontol. A Biol. Sci. Med. Sci., 56, B52– B57. Morimoto RI and Santoro MG. (1998). Stress-inducible responses and heat shock proteins: new pharmacologic targets for cytoprotection. Nat. Biotechnol., 16, 833–838. Nance M and Berry S. (1992). Cockayne syndrome: review of 140 cases. Am. J. Med. Genet., 42, 68–84. Parsons R, Myeroff LL, Liu B, Willson JK, Markowitz SD, Kinzler KW and Vogelstein B. (1995). Microsatellite instability and mutations of the transforming growth factor beta type II receptor gene in colorectal cancer. Cancer Res., 55, 5548–5550. Selby CP and Sancar A. (1997). Cockayne syndrome group B protein enhances elongation by RNa polymerase II. Proc. Natl. Acad. Sci. USA, 94, 11205–11209. Selzer RR, Nyaga S, Tuo J, May A, Muftuoglu M, Christiansen M, Citterio E, Brosh Jr RM and Bohr VA. (2002). Differential requirement for the ATPase domain of the Cockayne syndrome group B gene in the processing of UV-induced DNA damage and 8-oxoguanine lesions in human cells. Nucleic Acids Res., 30, 782–793. Sitte N, Merker K and Grune T. (1998). Proteasomedependent degradation of oxidized proteins in MRC-5 fibroblasts. FEBS Lett., 440, 399–402.

Sitte N, Merker K, von Zglinicki T, Grune T and Davies KJ. (2000). Protein oxidation and degradation during cellular senescence of human BJ fibroblasts: part I – effects of proliferative senescence. FASEB J., 14, 2495–2502. Tan M, Wang Y, Guan K and Sun Y. (2000). PTGF-beta, a type beta transforming growth factor (TGF-beta) superfamily member, is a p53 target gene that inhibits tumor cell growth via TGF-beta signaling pathway. Proc. Natl. Acad. Sci. USA, 97, 109–114. Tanaka TS, Jaradat SA, Lim MK, Kargul GJ, Wang X, Grahovac MJ, Pantano S, Sano Y, Piao Y, Nagaraja R, Doi H, Wood WH, III, Becker KG and Ko MS. (2000). Genome-wide expression profiling of mid-gestation placenta and embryo using a 15,000 mouse developmental cDNA microarray. Proc. Natl. Acad. Sci., USA, 97, 9127–9132. Tuo J, Muftuoglu M, Chen C, Jaruga P, Selzer RR, Brosh RM Jr, Rodriguez H, Dizdaroglu M and Bohr VA. (2001). The Cockayne Syndrome group B gene product is involved in general genome base excision repair of 8-hydroxyguanine in DNA. J. Biol. Chem., 276, 45772–45779. Tyrrell RM and Basu-Modak S. (1994). Transient enhancement of heme oxygenase 1 mRNA accumulation: a marker of oxidative stress to eukaryotic cells. Methods Enzymol., 234, 224–235. Wang Y, Cheong D, Chan S and Hooi SC. (2000). Ribosomal protein L7a gene is upregulated but not fused to the tyrosine kinase receptor as chimeric trk oncogene in human colorectal carcinoma. Int. J. Oncol., 16, 757–762. Weber JD, Raben DM, Phillips PJ and Baldassare JJ. (1997). Sustained activation of extracellular-signal-regulated kinase 1 (ERK1) is required for the continued expression of cyclin D1 in G1 phase. Biochem. J., 326 (Part 1), 61–68. Weinberg RA. (1995). The retinoblastoma protein and cell cycle control. Cell, 81, 323–330. Weindruch R, Kayo T, Lee CK and Prolla TA. (2002). Gene expression profiling of aging using DNA microarrays. Mech. Ageing Dev., 123, 177–193. Wheaton K, Atadja P and Riabowol K. (1996). Regulation of transcription factor activity during cellular aging. Biochem Cell Biol., 74, 523–534. Yu A, Fan HY, Liao D, Bailey AD and Weiner AM. (2000). Activation of p53 or loss of the Cockayne syndrome group B repair protein causes metaphase fragility of human U1, U2, and 5S genes. Mol. Cell, 5, 801–810. Zhao R, Gish K, Murphy M, Yin Y, Notterman D, Hoffman WH, Tom E, Mack DH and Levine AJ. (2000). Analysis of p53-regulated gene expression patterns using oligonucleotide arrays. Genes Dev., 14, 981–993. Zou S, Meadows S, Sharp L, Jan LY and Jan YN. (2000). Genome-wide study of aging and oxidative stress response in Drosophila melanogaster. Proc. Natl. Acad. Sci. USA, 97, 13726–13731.

Oncogene