The Knick | Vedi tutte | Loosely Exactly Nicole

Distinct nuclear gene expression profiles in cells with mtDNA depletion and homoplasmic A3243G mutation

Distinct nuclear gene expression profiles in cells with mtDNA depletion and homoplasmic A3243G mutation

Mutation Research 578 (2005) 43–52 Distinct nuclear gene expression profiles in cells with mtDNA depletion and homoplasmic A3243G mutation Roshan S. ...

186KB Sizes 1 Downloads 10 Views

Mutation Research 578 (2005) 43–52

Distinct nuclear gene expression profiles in cells with mtDNA depletion and homoplasmic A3243G mutation Roshan S. Jahangir Tafrechi a , Peter J. Svensson b,c , George M.C. Janssen a , Karoly Szuhai a , J. Antonie Maassen a , Anton K. Raap a,∗ a

Department of Molecular Cell Biology, Leiden University Medical Center, P.O. Box 9503, 2300 RA Leiden, The Netherlands b Department of Toxicogenetics, Leiden University Medical Center, P.O. Box 9503, 2300 RA Leiden, The Netherlands c Department of Oncology, Radiology and Clinical Immunology, University Hospital, 75185 Uppsala, Sweden Received 8 November 2004; received in revised form 1 February 2005; accepted 18 February 2005 Available online 28 April 2005

Abstract The pathobiochemical pathways determining the wide variability in phenotypic expression of mitochondrial DNA (mtDNA) mutations are not well understood. Most pathogenic mtDNA mutations induce a general defect in mitochondrial respiration and thereby ATP synthesis. Yet phenotypic expression of the different mtDNA mutations shows large variations that are difficult to reconcile with ATP depletion as sole pathogenic factor, implying that additional mechanisms contribute to the phenotype. Here, we use DNA microarrays to identify changes in nuclear gene expression resulting from the presence of the A3243G diabetogenic mutation and from a depletion of mtDNA (␳0 cells). We find that cells respond mildly to these mitochondrial states with both general and specific changes in nuclear gene expression. This observation indicates that cells can sense the status of mtDNA. A number of genes show divergence in expression in ␳0 cells compared to cells with the A3243G mutation, such as genes involved in oxidative phosphorylation. As a common response in A3243G and ␳0 cells, mRNA levels for extracellular matrix genes are up-regulated, while the mRNA levels of genes involved in ubiquitin-mediated protein degradation and in ribosomal protein synthesis is down-regulated. This reduced expression is reflected at the level of cytosolic protein synthesis in both A3243G and ␳0 cells. Our finding that mitochondrial dysfunction caused by different mutations affects nuclear gene expression in partially distinct ways suggests that multiple pathways link mitochondrial function to nuclear gene expression and contribute to the development of the different phenotypes in mitochondrial disease. © 2005 Elsevier B.V. All rights reserved. Keywords: Mitochondria; mtDNA; Expression; Diabetes

Abbreviations: ES, enrichment score; GO, gene ontology; GSEA, gene set enrichment analysis; MIDD, maternally inherited diabetes and deafness; mtDNA, mitochondrial DNA; OXPHOS, oxidative phosporylation; PLA, probe level analysis ∗ Corresponding author. Tel.: +31 71 5276187; fax: +31 71 5276180. E-mail address: [email protected] (A.K. Raap). 0027-5107/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.mrfmmm.2005.02.002

44

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

1. Introduction Mitochondrial dysfunction caused by mutations in the mitochondrial genome is related to a variety of diseases such as type two diabetes mellitus (DM2), cancer and neuro-muscular diseases [1]. Mitochondria are involved in multiple cellular processes of which ATP production by oxidative phosphorylation (OXPHOS) is the most prominent one. However, mitochondria also accommodate other processes, such as the tricarboxylic acid cycle and fatty acid oxidation. In addition, mitochondrial function is linked to calcium, iron and ROS signalling and apoptotic pathways. The variation in clinical phenotype of mitochondrial diseases is difficult to explain by merely a reduced respiration rate [2]. Rather the consequences of additional mitochondrial dysfunction on retrograde signalling pathways may determine the distinct nature of the clinical manifestation [3]. A genome-wide differential gene transcription profile of normal cells and cells with dysfunctioning mitochondria is expected to give insight in the pathobiochemical pathways affected in mitochondrial disease [4,5]. In order to investigate how mitochondrial mutations affect the nuclear gene expression profile we created 143B cybrid cells with mitochondrial DNA being the only variable [6]. The first state of respiratory dysfunction is induced by an A–G conversion at location 3243 in the tRNAleu gene of the mitochondrial DNA. This mutation causes maternally inherited diabetes and deafness (MIDD) [7] in most carriers and associates also with the neuromuscular MELAS syndrome [8]. Another state of mitochondrial dysfunction is induced by a depletion of mtDNA (␳0 cells). Using cybrid cells with 100% wild-type mitochondrial DNA, cybrid cells with 100% A3243G mutant mitochondrial DNA of the same haplotype and ␳0 cells, we found both common mitochondrial-defect and MIDD-specific responses in nuclear gene expression.

2. Materials and methods 2.1. Cell culture, cell characteristics and GeneChip hybridisation The cybrid cells used in this report have been previously described [2,9]. In short, 143B osteosar-

coma cells were treated with ethidium bromide to create ␳0 cells devoid of mitochondrial DNA. Next, ␳0 cells were fused with enucleated cells from a MIDD patient, generating clones with different but stable heteroplasmy levels for the 3243 mutation. Two apparently homoplasmic mutants (VM48 and VM50) and two apparently homoplasmic wild-type (VW6 and VW7) cybrid clones were selected and used as biological replicates in this study. The cells were grown on Dulbecco’s modified eagle’s medium containing 4.5 mg/ml glucose and 110 ␮g/ml pyruvate (DMEM) supplemented with 50 ␮g/ml uridine and 10% fetal bovine serum. Heteroplasmy levels were monitored by use of PCR-RFLP and ApaI, which cleaves the mutated PCR product. The oxygen consumption of the cells was measured as described previously [6]. Mitochondrial copy numbers were determined by comparing the amount of mitochondrial DNA with the amount of ␤-globin DNA, in a SybrGreen real time PCR reaction with the primers described in Szuhai et al. [10] and using the ABi Prism 7700 spectrofluormetric thermal cycler (Applied Biosystems, USA). The same system was used for validation of the mRNA concentration data obtained by chip hybridisation analysis. Primer sets for 10 different genes were used, including 3 that were found to be differentially and consistently expressed: NADH dehydrogenase 1␤8 (5 ACGAACCTTACCCGGATGATG + 5 -CATGGATCTCTCTCATGCTGTGAG), ubiquitin-conjugating enzyme E2D3 (5 -ATCACAGTGGTCGCCTGCTT + 5 ATAGATCCGTGCAATCTCTGGC) and collagen VI␣2 (5 -CATCGATGACATGGAGGACGT + 5 -CAGCTCTGTTTGGCAGGGAA). The primers were all manufactured by Eurogentec, Belgium. Cytoplasmic protein synthesis rate was estimated from the incorporation of l-[4,5-3 H]-leucine into trichloroacetic acid precipitable material essentially as described [2]. In brief, series of cells at different densities in 6-well plates were washed with phosphatebuffered saline and incubated for 60 min at 37 ◦ C in 0.75 ml of leucine-free complete medium containing 10 ␮Ci of l-[4,5-3 H]-leucine and 10 ␮M unlabeled leucine. Cells were then again washed three times with phosphate-buffered saline and thoroughly dissolved in 1 ml 0.2 M NaOH. One hundred microlitres aliquots were precipitated by the addition of 100 ␮l 20% trichloroacetic acid, and assayed for total protein

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

content by a bicinchoninic acid based protein assay (Pierce, USA) and for the incorporation of [3 H]-leucine into protein using a GF/C filter assay. The rate of protein synthesis is expressed as the amount of [3 H]-leucine incorporated into 1 ␮g protein during the 1-h incubation period (cpm ␮g−1 h−1 ). Protein concentration was also used as an estimate for cell density. Biotinylated cRNA samples were prepared according to the Affymetrix GeneChip protocol. In short, cells at 90% confluency were washed with phosphate buffered saline and directly lysed in RNA-Bee solution (Tel-Test, USA). RNA purification was performed with the RNeasy mini kit (Qiagen, Germany). With the intensity ratio of the 28S/18S rRNA bands being over 1.8, the integrity of the RNA was confirmed. Twenty micrograms of total RNA was used to perform double-stranded cDNA synthesis with an oligo T7 –dT24 primer and SuperScript kit (Invitrogen Life Technologies, USA). The cDNA was purified with Phase Lock Gel-reaction tubes (Eppendorf, Germany) and the complete pellet was used for a simultaneous 6-h in vitro transcription and labeling reaction using the MegaScript T7 kit (Ambion, USA) and biotin11-CTP/biotin-16-UTP (Perkin-Elmer Life Sciences, USA). Finally, 20 ␮g fragmented cRNA was used for hybridisation on Affymetrix HG-U133 chips, according to the manufacturer’s instruction. Affymetrix’ Microarray Suite 5.0 Software (MAS) was used to determine the percentage of transcripts present and the 3 /5 intensity ratios for ␤-actin and GAPDH. 2.2. Probe level analysis (PLA) and gene set enrichment analysis (GSEA) Most data analyses were performed in R (www.rproject.org) using the Bioconductor functions (www.bioconductor.org). Intensity data were corrected for background arising from optical noise as well as from non-specific hybridisation, according to the procedure developed by Wu and Irizarry [11] and annotations were added using the annotation package hgu133a, Version 1.3.1 [12]. The arrays were normalized using the VSNtransformation [13]. To determine differential expression of the genes on the array, the pre-processed data were handled using the algorithms described in Liu et al. [14] which makes use of the feature that each transcript is represented by 11 25-mer perfect-match

45

probes. The scripts for data processing are available at www.medgencentre.nl/pla. In short, the 11 probeintensities of each transcript from a reference hybridisation (VW6 or VW7) were compared to the eleven probe-intensities of the corresponding transcript from a test-hybridisation (VM48, VM50, ␳0 1 or ␳0 2) using paired Wilcoxon statistics [15] to calculate a onetailed p-value. The signals from the mismatch probes were not subtracted. A perturbation factor was set at 1.1 to exclude transcripts with very small changes in their intensity levels [14]. The most conservative onetailed p-value (the one closest to 0.5) was assigned as the change-value. If the p-values after perturbation are on opposite sides of 0.5, the transcript will be classified as ‘not changed’ and it will be assigned a change-value of 0.5. For all transcripts in pair-wise comparisons the test will assign a change-value between zero and one. A change-value of 0 indicates that all probe-intensities of a test-transcript are higher than the corresponding probe-intensities of the reference-transcript. Conversely, a change-value of 1 corresponds to an overall decrease. All change-values between these two extremes indicate a partial or complete overlap of the probe-intensity patterns. When there is a totally random pattern (at noise-level intensity) the change-value will be 0.5 and the corresponding transcript is considered as not changed in mRNA concentration. Subtle changes are considered to be of biological relevance if the change is consistent within a gene set. Therefore, as a second approach, gene set enrichment analysis was performed on all 22,283 transcripts to pinpoint the most coherently changed gene sets per comparison [16]. Gene sets were derived from the Kyoto encyclopaedia of genes and genomes (KEGG) and from the gene ontology consortium (GO). Expression changes were calculated with signal-to-noise ratios.

3. Results 3.1. Cell characteristics and quality controls The cybrid cell lines selected for gene expression profiling were comparable to each other with respect to cell morphology, doubling time (range 13–18 h) and cell viability (∼4% trypan blue stained cells). The average oxygen consumption of the wild-type

46

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

cybrids was 1.3 ± 0.3 fmol/(min cell) (n = 10), while this value for the 100% 3243 mutant cybrids was 0.1 ± 0.1 fmol/(min cell) (n = 5). ␳0 cells did not respire at all. Mitochondrial copy numbers of the wild type and the 3243 mutant cybrids were comparable (real time PCR data not shown). For mRNA expression profiling the cells were grown under standardized conditions to ∼90% confluence. Quality control was performed at each step of RNA isolation and cRNA synthesis and no variations were detected in yields and length distribution of the products. Before starting PLA and GSEA analysis we demanded that the fraction of detectable or ‘present’ genes and the 3 /5 hybridisation ratio fell within Affymetrix’ boundaries (<10% variation and <3, respectively). For the six hybridisations, the mean percentage ‘present’ transcripts were 47.5 ± 0.9% and the 3 /5 ratios were 1.2 ± 0.09 for ␤-actin and 0.9 ± 0.04 for GAPDH. Quantitative RT PCR data for 10 different mRNAs correlated very well qualitatively and to large extent quantitatively with chip hybridisation and PLA data (not shown), which is in accord with recent literature comparing different gene expression platforms [17,18]. 3.2. Probe level analysis An intensity frequency histogram was generated using all the transcripts from all six hybridisations after correction for optical noise and non-specific hybridisation. Based on this histogram, the threshold for minimal intensity was set to 25 . Thirty percent of the 22,283 transcript represented on the HG-U133A microarray was expressed above threshold in at least one cell type. The average expression of these 6691 transcripts was 516. The frequency histogram of the PLA change-values in the comparison VM48 versus VW6 is shown in Fig. 1 as an example. It illustrates that the majority of transcript are not changed in expression and that most of

Fig. 1. Frequency histogram of the PLA change-values. This histogram depicts the amount of transcripts with a change in expression level between 3243 mutant clone VM48 and wild-type clone VW6. Values between 0 and 0.5 indicate an increase, values between 0.5 and 1 a decrease. A pertubation factor of 1.1 has been used.

the remaining transcripts have a change-value of zero or one. For a given transcript to be included in the list of differentially expressed transcripts, we demanded that all four change-values associated with it are either all >0.5 (decreased) or all <0.5 (increased). With this criterion, 553 of the 6691 transcripts (8%) were considered differentially expressed when comparing the 3243 mutant to wild-type cells. The number of differentially expressed transcripts in the comparison between ␳0 cells and wild-type cells was more than three-fold higher (28%). Notably, in the comparison of ␳0 and 3243 mutant 1581 transcripts (24%) were found to be changed (Table 1). The 553 transcripts found in the 3243 mutants versus wild-type comparison do not constitute a full subset of the 1869 transcripts of the ␳0 versus wild-type comparison. The single-nucleotide A3243G substitution and the absence of mtDNA apparently reschedule the nuclear gene expression in different ways.

Table 1 Number of transcripts changed in two-way comparisons Two-way comparison

# Changed transcripts

Relative amount (%)

# Down-regulated

# Up-regulated

3243 Wt ␳0 Wt ␳0 3243

553 1869 1581

8.3 27.9 23.6

295 719 1031

258 1150 550

The transcripts changed in the 3243 vs. wild-type comparison are not a complete subset of any other comparison.

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

A three-way cell type comparison technique was used to identify the transcripts that behave similarly in the two types of dysfunctioning cells (A3243G and ␳0 ) as well as the transcripts that behave distinctively. Note that in this comparison with duplicate hybridisation experiments, eight change-values are to be considered: two A3243G cell clones versus two wild-type cell clones and two times ␳0 versus the two wild-type cell clones. As a result the stringency for inclusion of transcripts in the lists of differentially expressed transcripts is further increased, because for generation of the ‘common change’ list it was demanded that all eight (2 × 4) change-values indicate a change in the same direction. Similar to generate the ‘specific’ lists, it was demanded that none of the four corresponding change-

47

Fig. 2. Three-way comparison. This Venn diagram shows the number of transcripts that are changed in common and specifically when compared to wild-type cybrids cells. The three groups shown in this figure correspond to the six supplemented lists, where the up- and down-regulated transcripts are presented separately.

Table 2 Transcripts changed more than two-fold Affy-code

R/W

M/W

R/M

Description

Opposite effect 39729 at 200734 s at 218275 at 213421 x at 201227 s at

2.16 2.89 2.04 2.00 1.93 (1.99)

0.44 0.93 0.90 0.99 0.92 (0.98)

4.86 3.12 2.27 2.01 2.10 (2.23)

Peroxiredoxin 2 ADP-ribosylation factor 3 Solute carrier family 25 (mit, carrier; dicarboxylate transporter), member 10 Protease, serine, 3 (mesotrypsin)a NADH dehydrogenase (ubiquinone) 1 beta subcomplex, 8, 19 kDa

1.53 0.73 0.48 0.49 0.33 0.44

5.27 2.07 1.49 1.46 1.39 1.13

0.29 0.35 0.32 0.34 0.24 0.38

Argininosuccinate synthetase Chaperonin containing TCP1, subunit 3 (gamma)a Annexin A1 Transforming growth factor, beta receptor II (70/80 kDa)b Hairy/enhancer-of-split related with YRPW motif 1 Protein tyrosine phosphatase, non-receptor type 12

4.50 (1.50) 3.01 2.91 2.77 2.53 3.10 2.38 2.27 2.31

4.08 (1.47) 3.68 2.48 2.39 2.27 2.20 2.03 1.87 1.71

1.09 (1.02) 0.81 1.17 1.16 1.12 1.41 1.17 1.21 1.36

Collagen, type VI, alpha 2b Collagen, type VI, alpha 3b Biglycanb Glycogen synthase 1 (muscle) Putative G-protein coupled receptor Collagen, type IV, alpha 2b Platelet-activating factor acetylhydrolase, isoform Ib, alpha subunit 45 kDa Tyrosyl-tRNA synthetase Calmodulin 3 (phosphorylase kinase, delta)

0.50 0.44 0.50 0.40 (0.58) 0.50 0.46

0.45 0.53 0.69 0.60 (0.62) 0.77 0.79

1.10 0.84 0.72 0.66 (0.94) 0.66 0.58

Wilms tumor 1 associated protein ATP binding protein associated with cell differentiation Ubiquitin-specific protease 10a Ubiquitin-conjugating enzyme E2D 3 (UBC4/5 homolog, yeast)a Thymopoietin YY1 transcription factor

207076 s at 213506 at 201012 at 208944 at 44783 s at 202006 at Common effect 209156 s at 201438 at 213905 x at 201673 s at 220642 x at 211964 at 200815 s at 212048 s at 200623 s at 203137 203008 209137 200667 203432 201901

at x at s at at at s at

W = 100% wild-type cybrid cells, M = 100% A3243G mutant and R = depletion of mtDNA. Data in brackets are calculated from real time PCR data. a Transcripts associated with ubiquitinylation. b Transcripts associated with extracellular matrix formation.

48

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

values of the other mutant versus wild-type comparison indicate a change in the same direction. All transcripts that did not exactly meet one of these two criteria were rejected. The results are presented as a Venn diagram in Fig. 2, which shows that 229 transcripts have a similar change of expression in both mitochondrial mutant cell lines. Interestingly, 155 transcripts are changed specifically in ␳0 cells whereas 73 additional transcripts are specifically altered in 3243 mutant cells. The relative changes in gene expression due to mtDNA mutations appear to be rather mild with maximal differences of ∼5-fold and no on/off situations. The transcripts with a ≥2-fold change in any of the three cell-type comparisons are listed in Table 2. The quantitative RT PCR data of three differentially expressed genes are included. Tables S1–S6, Supplementary material, contain the PLA data of all 457 differentially expressed transcripts. 3.3. Gene set enrichment analysis In contrast to PLA, which analyses differences in expression of individual transcripts, gene set enrichment analysis pinpoints sets of genes that are coherently

changed in their expression. When using the Kyoto encyclopaedia of genes and genomes and the gene ontology consortium data bases as predefined input gene sets, GSEA calculates for each cell type comparison an enrichment score for all gene sets [16], ranging from 0 to well over 300. Both KEGG and the GO consortium provide a controlled vocabulary to describe gene products, which are grouped together according to process, pathway or localization. In the case of KEGG, the transcripts on the HG-U133A chip are divided into 107 pathway-specific gene sets. The GO consortium on the other hand, makes use of three different classifications according to biological process (GO-BP), cellular component (GO-CC) and molecular function (GO-MF). GO-CC divides the transcripts into 356, GOBP into 1317 and GO-MF into 1461 gene sets. For all of these predefined gene sets, enrichment scores were calculated for all three comparisons (Table S7, Supplementary material). Gene sets with an ES above a threshold of 300 may be considered as biologically significant since they represent 2.5% of all 3241 gene sets. Representative gene sets of these 82 gene sets are depicted in Table 3. Three types of effect may be dis-

Table 3 Enrichment scores for biological clustering groups GSEA

3243 wt (ES)a

␳0 wt

␳0 3243

Common effect MF: ubiquitin-protein ligase activityb MF: RNA binding activity KEGG: ribosomeb CC: collagen

316↓ 533↓ 360↓ ↑231

321↓ 583↓ 306↓ ↑353

217↓ 196↓ ↑260 ↑121

A3243G specific MF: actin binding activity MF: Ca2+ binding activity CC: basement membrane KEGG: alanine and aspartate metabolism

↑428 ↑333 ↑327 ↑326

↑173 ↑138 ↑135 ↑199

255↓ 157↓ 192↓ 138↓

␳0 Specific BP: amino acid transport CC: nuclear pore BP: mitosis

↑195 227↓ 202↓

↑326 535↓ 367↓

↑302 500↓ 321↓

Opposite effect CC: mitochondrionb KEGG: ubiquinon biosynthesis BP: TGF-␤ receptor signaling pathway

271↓ 215↓ ↑142

↑341 ↑237 191↓

↑584 ↑310 322↓

a The threshold for the enrichment score (ES) is set at 300. The arrows indicate either a global decrease or increase of the transcripts in the cluster. b The three criteria ‘ubiquitin-protein ligase activity’, ‘ribosome’ and ‘mitochondrion’ have similar results for comparable criteria in all of the other groups.

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

tinguished: mutant-specific effects, effects common to both ␳0 and A3243G and opposite effects. 3.4. Correlation of GSEA and PLA Both the PLA and GSEA techniques indicate moderate but significant effects of mtDNA mutations on the nuclear expression profile of osteosarcoma cells. Because it is reasonable to assume that relatively mild effects of mtDNA mutations on nuclear gene expression are of biological interest, we have chosen to focus on coherently changed gene sets using the GSEA method and to use PLA as confirmation. The KEGG gene set ‘ribosome’ was found downregulated in both the 3243 mutant (ES = 360) and in ␳0 cells (ES = 306). The three corresponding GO groups gave similar results. The average effect on ribosomal protein gene expression was small (<10%) but consistent (p < 0.01), and always found more strongly downregulated in 3243 mutant cells than in ␳0 cells. The mRNA expression of both the mitochondrial and the cytoplasmic ribosomal proteins was affected, but the expression of cytoplasmic protein genes for the small ribosomal subunit 40S (GO:0005843) seems to be affected most (Fig. 3A). To analyse whether this small reduction in expression of cytosolic ribosomal protein genes has any biological consequence we measured

49

the rate of cytosolic protein synthesis as reflected by [3 H]-leucine incorporation in these cell lines. Both the A3243G and ␳0 cells do indeed show an unambiguous reduction in protein synthesis rate compared to the wild-type cell lines as shown in Fig. 3B. The proteins involved in ubiquitin-mediated proteolysis were also found coherently down-regulated in both mutant cells. Within the KEGG data base one can find the ‘ubiquitin-mediated proteolysis’ gene set. However, in the GO data base, these genes are found rather scattered throughout the gene sets: within GO-MF four different gene sets with ubiquitin-related proteins can be recognized: ‘ubiquitin-protein ligase activity’, ‘ubiquitin C-terminal hydrolase activity’, ‘ubiquitin-specific protease activity’ and ‘ubiquitinconjugating enzyme activity’. Within GO-BP there are two gene sets present: ‘ubiquitin cycle’ and ‘ubiquitindependent protein catabolism’. All these gene sets have an ES > 300 for down-regulation in ␳0 and an ES ∼ 300 for down-regulation in 3243 cells. In line with this, two transcripts involved in ubiquitin-mediated proteolysis were found >2-fold down-regulated according to PLA (Table 2). The extracellular matrix proteins were found to be up-regulated in both mutant cell lines. In the GO-CC data base, the group ‘collagen’ clearly emerged for the ␳0 cells and to a lesser extent also for the 3243 mutant

Fig. 3. (A) Comparison of the average mRNA levels of all 47 proteins of the cytoplasmic small ribosomal subunit. (*) The change is highly significant for both ␳0 and 3243 mutants according to a standard paired t-test (p = 3.10−3 and 3.10−8 , respectively). (B) Cytoplasmic protein synthesis is lower in both mutant cell lines. Because protein synthesis slows down when cells reach confluence [2], its rate is shown as a function of cell density, which is expressed as the amount of protein per cm2 substratum. For gene expression profiling, cells were harvested at a density of ∼40 ␮g protein/cm2 .

50

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

cell. This result was confirmed by PLA, where three different collagens as well as the pericellular matrix protein biglycan can be found strongly up-regulated, this time with the 3243 mutant showing a much larger effect compared to the ␳0 cells (Table 2). The KEGG defined gene set ‘oxidative phosphorylation’ was found specifically up-regulated in ␳0 cells (ES = 384). The enrichment score for the oxidative phosphorylation genes is even larger (ES = 424) in the ␳0 versus 3243 comparison. Comparable enrichment scores were found for GO-MF ‘NADH dehydrogenase activity’, GO-CC ‘mitochondrion’ and GO-BP ‘oxidative phosphorylation, NADH to ubiquinone’.

4. Discussion With the aim to elucidate pathways involved in mitochondrial-nuclear genome cross-talk, we have undertaken a genome-wide analysis of the alterations in nuclear gene expression in response to two different types of mtDNA mutations that both provoke mitochondrial dysfunction such as loss of mtDNA encoded proteins and respiration [8,19]. The transmitochondrial fusion technique developed by King and Attardi [6] permitted the generation of cybrid cell lines with the mtDNA genome being the only variable, thus excluding variations due to mtDNA haplotype and nuclear background [20]. Depletion of mtDNA or introduction of MIDD-derived A3243G mtDNA in 143B osteosarcoma cells led to failure of the cells to respire. As a control we used cells containing wild-type mtDNA derived from the same MIDD patient. Relative mRNA abundances were measured using Affymetrix GeneChip technology on two independently grown clones of each cell type. With a novel probe level analysis technique we distinguished changes in the expression of individual genes that are common to the two types of mutation as well as mutation-specific ones (Table 2). Overall, the magnitude of changes in expression was moderate; only a handful genes reached ∼4-fold changes. The genes and associated pathways identified by probe level analysis overlapped considerably with those found independently by gene set enrichment analysis, which is well suited for identifying consistent changes in gene sets involved in given pathways [16]. Our main conclusion, therefore, is that the mutational mtDNA status is sensed

by the nuclear genome and reacted upon in a common and a mutation-specific way. The common responses likely originate from a failure of oxidative phosphorylation. They include down-regulation of ribosomal protein genes and upregulation of extracellular matrix genes, the latter also observed in respiration-deficient cybrids containing dimer mtDNA [4]. The presence of the A3243G mutation and the absence of mtDNA also seem to induce mutation-specific signals towards the nuclear genome. A similar view has emerged from studies in yeast [21] where different respiratory chain inhibitors and the absence of mtDNA also induced different transcriptional responses. Taken together, it is conceivable that nuclear transcription is reprogrammed by the mitochondrial DNA status through different sensing mechanisms, similar to retrograde signalling in yeast [3]. On its turn, mtDNA replication and transcription is stimulated by the nuclear co-activator PGC-1, the nuclear respiratory factors, the general transcription factor Sp1 and mitochondrial transcription factor TFAM [22,23], emphasizing the mutual dependence and interplay between both genomes. We were intrigued by the small but consistent change in expression of the ribosomal proteins. Therefore we conducted an experiment to confirm these data on protein synthesis activity. Indeed the protein synthesis rate, in both the A3243G mutant and the ␳0 cells, is significantly lower compared to wild-type cells. The view emerges that cells with defective energy supply down-regulates energy demanding processes like protein synthesis [24]. In accordance, genes involved in protein breakdown are also down-regulated, possibly in an attempt to compensate for the danger of an unbalance in protein metabolism. A study in yeast recently revealed that the abundance of ribosome biogenesis factors, which are more often co-ordinately downregulated, is controlled at the level of mRNA stability [25]. Since insulin production is highly depending on ribosome activity, it is tempting to speculate that a decreased rate of protein synthesis in the pancreatic ␤ cell might contribute to the pathogenesis of the diabetic phenotype. Notwithstanding the limitations of a model system like cultured osteosarcoma cells for the diabetic phenotype, it is remarkable that mitochondrial dysfunction often associates with diabetes and at least in the case of the A3243G mutation this is due to a reduction in insulin secretion capacity of the

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

pancreatic beta cell [26]. Mitochondrial dysfunction is strongly linked with development of diabetes type 2 [27,28]. Expression studies with muscle tissue from diabetic humans [16,29] and mice [30,31] show mild down-regulation of nuclear OXPHOS genes, likely a consequence of hyperglycemia. In contrast, expression studies with muscle tissue and fibroblast cells from mitochondrial disease patients show an increased expression of nuclear OXPHOS genes [32,33]. As such, up-regulation may be considered a compensatory effect for the mitochondrial dysfunction. Accordingly, we found an up-regulation in OXPHOS genes in the mtDNA depleted ␳0 cell line, but no significant effect of OXPHOS gene expression was seen in the cells with the diabetogenic A3243G mutation. Taken together these observations, indicate the existence of at least two different mechanisms for OXPHOS gene regulation. In conclusion, our data clearly demonstrate the presence of both general and specific communication routes between the mtDNA and the nuclear genomes and may contribute to the identification of pathways determining the specificity of mitochondrial diseases.

[3] [4]

[5]

[6]

[7]

[8] [9]

[10]

Acknowledgements We thank Marchien van de Sande for her excellent overall technical assistance. The microarray hybridisations were performed by Eveline Mank at the Leiden Genome Technology Center.

[11] [12]

[13]

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.mrfmmm.2005.02.002.

[14]

[15]

References [1] J.V. Leonard, A.H. Schapira, Mitochondrial respiratory chain disorders I: mitochondrial DNA defects, Lancet 355 (2000) 299–304. [2] G.M. Janssen, J.A. Maassen, J.M. van Den Ouweland, The diabetes-associated 3243 mutation in the mitochondrial tRNA(Leu(UUR)) gene causes severe mitochondrial dysfunc-

[16]

51

tion without a strong decrease in protein synthesis rate, J. Biol. Chem. 274 (1999) 29744–29748. R.A. Butow, N.G. Avadhani, Mitochondrial signaling: the retrograde response, Mol. Cell 14 (2004) 1–15. K.M. Clark, T.A. Brown, M.M. Davidson, L.C. Papadopoulou, D.A. Clayton, Differences in nuclear gene expression between cells containing monomer and dimer mitochondrial genomes, Gene 286 (2002) 91–104. R. Delsite, S. Kachhap, R. Anbazhagan, E. Gabrielson, K.K. Singh, Nuclear genes involved in mitochondria-to-nucleus communication in breast cancer cells, Mol. Cancer 1 (2002) 6. M.P. King, G. Attardi, Human cells lacking mtDNA: repopulation with exogenous mitochondria by complementation, Science 246 (1989) 500–503. J.M. van Den Ouweland, H.H. Lemkes, W. Ruitenbeek, L.A. Sandkuijl, M.F. de Vijlder, P.A. Struyvenberg, J.J. van de Kamp, J.A. Maassen, Mutation in mitochondrial tRNA(Leu)(UUR) gene in a large pedigree with maternally transmitted type II diabetes mellitus and deafness, Nat. Genet. 1 (1992) 368–371. H.T. Jacobs, Disorders of mitochondrial protein synthesis, Hum. Mol. Genet. 12 (Suppl. 2) (2003) R293–R301. J.M. van Den Ouweland, P. Maechler, C.B. Wollheim, G. Attardi, J.A. Maassen, Functional and morphological abnormalities of mitochondria harbouring the tRNA(Leu)(UUR) mutation in mitochondrial DNA derived from patients with maternally inherited diabetes and deafness (MIDD) and progressive kidney disease, Diabetologia 42 (1999) 485–492. K. Szuhai, J. Ouweland, R. Dirks, M. Lemaitre, J. Truffert, G. Janssen, H. Tanke, E. Holme, J. Maassen, A. Raap, Simultaneous A8344G heteroplasmy and mitochondrial DNA copy number quantification in myoclonus epilepsy and ragged-red fibers (MERRF) syndrome by a multiplex molecular beacon based real-time fluorescence PCR, Nucleic Acids Res. 29 (2001) E13. Z. Wu, R.A. Irizarry, Preprocessing of oligonucleotide array data, Nat. Biotechnol. 22 (2004) 656–658. G. Liu, A.E. Loraine, R. Shigeta, M. Cline, J. Cheng, V. Valmeekam, S. Sun, D. Kulp, M.A. Siani-Rose, NetAffx: Affymetrix probesets and annotations, Nucleic Acids Res. 31 (2003) 82–86. W. Huber, A. von Heydebreck, H. Sultmann, A. Poustka, M. Vingron, Variance stabilization applied to microarray data calibration and to the quantification of differential expression, Bioinformatics 18 (Suppl. 1) (2002) S96–S104. W.M. Liu, R. Mei, X. Di, T.B. Ryder, E. Hubbell, S. Dee, T.A. Webster, C.A. Harrington, M.H. Ho, J. Baid, S.P. Smeekens, Analysis of high density expression microarrays with signedrank call algorithms, Bioinformatics 18 (2002) 1593–1599. F. Wilcoxon, Individual comparisons by ranking methods, Biometrics Bull. 1 (1945) 80–83. V.K. Mootha, C.M. Lindgren, K.F. Eriksson, A. Subramanian, S. Sihag, J. Lehar, P. Puigserver, E. Carlsson, M. Ridderstrale, E. Laurila, N. Houstis, M.J. Daly, N. Patterson, J.P. Mesirov, T.R. Golub, P. Tamayo, B. Spiegelman, E.S. Lander, J.N. Hirschhorn, D. Altshuler, L.C. Groop, PGC-1alpha-responsive genes involved in oxidative phosphorylation are coordinately downregulated in human diabetes, Nat. Genet. 34 (2003) 267– 273.

52

R.S. Jahangir Tafrechi et al. / Mutation Research 578 (2005) 43–52

[17] Y. Wu, P. de Kievit, L. Vahlkamp, D. Pijnenburg, M. Smit, M. Dankers, D. Melchers, M. Stax, P.J. Boender, C. Ingham, N. Bastiaensen, R. de Wijn, D. van Alewijk, H. van Damme, A.K. Raap, A.B. Chan, R. van Beuningen, Quantitative assessment of a novel flow-through porous microarray for the rapid analysis of gene expression profiles, Nucleic Acids Res. 32 (2004) 123. [18] R.F. Chuaqui, R.F. Bonner, C.J. Best, J.W. Gillespie, M.J. Flaig, S.M. Hewitt, J.L. Phillips, D.B. Krizman, M.A. Tangrea, M. Ahram, W.M. Linehan, V. Knezevic, M.R. Emmert-Buck, Postanalysis follow-up and validation of microarray experiments, Nat. Genet. 32 (Suppl.) (2002) 509–514. [19] J.J. Jansen, J.A. Maassen, F.J. van der Woude, H.A. Lemmink, J.M. van Den Ouweland, L.M. t’ Hart, H.J. Smeets, J.A. Bruijn, H.H. Lemkes, Mutation in mitochondrial tRNA(Leu(UUR)) gene associated with progressive kidney disease, J. Am. Soc. Nephrol. 8 (1997) 1118–1124. [20] D.R. Dunbar, P.A. Moonie, H.T. Jacobs, I.J. Holt, Different cellular backgrounds confer a marked advantage to either mutant or wild-type mitochondrial genomes, Proc. Natl. Acad. Sci. U.S.A. 92 (1995) 6562–6566. [21] C.B. Epstein, J.A. Waddle, W. Hale, V. Dave, J. Thornton, T.L. Macatee, H.R. Garner, R.A. Butow, Genome-wide responses to mitochondrial dysfunction, Mol. Biol. Cell 12 (2001) 297–308. [22] Z. Wu, P. Puigserver, U. Andersson, C. Zhang, G. Adelmant, V. Mootha, A. Troy, S. Cinti, B. Lowell, R.C. Scarpulla, B.M. Spiegelman, Mechanisms controlling mitochondrial biogenesis and respiration through the thermogenic coactivator PGC-1, Cell 98 (1999) 115–124. [23] R.C. Scarpulla, Transcriptional activators and coactivators in the nuclear control of mitochondrial function in mammalian cells, Gene 286 (2002) 81–89. [24] D.F. Rolfe, G.C. Brown, Cellular energy utilization and molecular origin of standard metabolic rate in mammals, Physiol. Rev. 77 (1997) 731–758. [25] J. Grigull, S. Mnaimneh, J. Pootoolal, M.D. Robinson, T.R. Hughes, Genome-wide analysis of mRNA stability using tran-

[26]

[27] [28] [29]

[30]

[31]

[32]

[33]

scription inhibitors and microarrays reveals posttranscriptional control of ribosome biogenesis factors, Mol. Cell Biol. 24 (2004) 5534–5547. J.A. Maassen, L.M. ’t Hart, E. Van Essen, R.J. Heine, G. Nijpels, R.S. Jahangir Tafrechi, A.K. Raap, G.M. Janssen, H.H. Lemkes, Mitochondrial diabetes: molecular mechanisms and clinical presentation, Diabetes 53 (Suppl. 1) (2004) S103– S109. P. Maechler, C.B. Wollheim, Mitochondrial function in normal and diabetic beta-cells, Nature 414 (2001) 807–812. B.B. Lowell, G.I. Shulman, Mitochondrial dysfunction and type 2 diabetes, Science 307 (2005) 384–387. M.E. Patti, A.J. Butte, S. Crunkhorn, K. Cusi, R. Berria, S. Kashyap, Y. Miyazaki, I. Kohane, M. Costello, R. Saccone, E.J. Landaker, A.B. Goldfine, E. Mun, R. DeFronzo, J. Finlayson, C.R. Kahn, L.J. Mandarino, Coordinated reduction of genes of oxidative metabolism in humans with insulin resistance and diabetes: potential role of PGC1 and NRF1, Proc. Natl. Acad. Sci. U.S.A. 100 (2003) 8466–8471. V.K. Yechoor, M.E. Patti, R. Saccone, C.R. Kahn, Coordinated patterns of gene expression for substrate and energy metabolism in skeletal muscle of diabetic mice, Proc. Natl. Acad. Sci. U.S.A. 99 (2002) 10587–10592. V.K. Yechoor, M.E. Patti, K. Ueki, P.G. Laustsen, R. Saccone, R. Rauniyar, C.R. Kahn, Distinct pathways of insulin-regulated versus diabetes-regulated gene expression: an in vivo analysis in MIRKO mice, Proc. Natl. Acad. Sci. U.S.A. 101 (2004) 16525–16530. F.H. Van Der Westhuizen, L.P. Van Den Heuvel, R. Smeets, J.A. Veltman, R. Pfundt, A.G. Van Kessel, B.M. Ursing, J.A. Smeitink, Human mitochondrial complex I deficiency: investigating transcriptional responses by microarray, Neuropediatrics 34 (2003) 14–22. A. Heddi, G. Stepien, P.J. Benke, D.C. Wallace, Coordinate induction of energy gene expression in tissues of mitochondrial disease patients, J. Biol. Chem. 274 (1999) 22968–22976.