Analysis of baseline and cisplatin-inducible gene expression in Fanconi anemia cells using oligonucleotide-based microarrays

The autosomal recessive disorder Fanconi anemia (FA) is clinically characterized by bone marrow failure, predisposition to cancer and various developmental abnormalities [1]. FA is genetically heterogeneous, and thus far eight complementation groups have been described. Six of the identified FA genes (FANCA, -C, -D2, -E, -F, and -G; [2–8]) encode proteins that are unique and lack apparent homology to other proteins or to each other. In addition, there are no conserved motifs present in these FA proteins, hampering understanding of their molecular function. Recently, biallelic mutations were found in the BRCA2 gene in patients belonging to complementation groups FA-D1 and one FA-B patient [9]. At the cellular level, FA is characterized by increased spontaneous genomic instability and hypersensitivity to DNA crosslinking agents, e.g. cisplatin and mitomycin C (MMC) [reviewed in [1]]. Multiple studies have also shown defects in FA cells related to the interferon-signaling pathway. Compared to control cells, FA cells express constitutively high levels of ISGF3 gamma, IRF-1, p21 ^waf1/Cip1 and MxA [10, 11].

Studies on the FA proteins have shown that five FA proteins, FANCA, FANCC, FANCE, FANCF, and FANCG, bind to each other to form a nuclear core complex [Reviewed in [12] and [13]]. FANCD2 is a nuclear protein that requires activation by mono-ubiquitination. In mutant FA cells that lack one of the FA core complex proteins, FANCD2 is not activated by ubiquitination, suggesting that these FA proteins exert a common molecular function in the nuclear compartment of the cell [14]. Whether BRCA2 is involved in the same pathway or exerts a separate function is currently unknown [15].

Crosslinking agents are widely used for the treatment of various types of cancer and are thought to exert their cytotoxic effect predominantly through irreversible binding with DNA. How the FA proteins are related to this cytotoxic effect, e.g. by functioning in processes such as DNA repair, cell cycle control or protection, is unknown. Upon treatment with a discriminating dose of crosslinking agent that will only transiently arrest the growth of wild type cells, FA cells arrest in the late S or early G2-phase of the cell cycle, and ultimately undergo cell death [16–21]. In contrast to normal cells, FA cells fail to inhibit replicative DNA synthesis after treatment with crosslinking agents. While normal cells will arrest in S-phase, FA cells continue replication and subsequently arrest at a later cell cycle check point [22, 23]. This hypersensitivity to crosslinking agents is the hallmark of the FA phenotype.

Many studies have indicated that in vitro treatment of cells with cisplatin affects the expression of specific genes involved in various molecular processes such as transcription, DNA repair, apoptosis, and cell cycle regulation. Examples are induction of c-jun, c-fos, ercc1, gadd45, gadd153, and p21 ^waf1/Cip1 [24–28]. Some of these genes respond within hours after treatment of cells, e.g. c-jun, whereas others show maximum changes in expression after 24 to 72 hours, e.g. p21 ^waf1/Cip1, gadd45 and gadd153.

Recently developed techniques, cDNA microarrays and oligonucleotide expression probe arrays, enable the systematic analysis of expression of thousands of genes in a single experiment [reviewed in [29]]. These techniques are useful to evaluate possible differences in gene expression profiles between FA and control cells, either at baseline or after crosslinker treatment. To study this, there are a number of experimental approaches to consider. For example, one might compare FA cells with cells from healthy controls. This might necessitate studying a large number of both types of cells in order to reduce false positives due to genetic differences in the cell lines that are unrelated to FA. Another approach would be to study cell lines derived from FA patients that are compared with the same cell lines corrected by transfection with the complementing cDNA. The use of such isogenic cell lines would presumably reduce the number of false positives and therefore reduce the number of cell lines that need to be studied. However, this might also result in a non-physiologic expression of the FA protein in transfected cells, which might influence the expression profiles. Therefore, we used the latter approach in combination with a unique cell line couple, one FA-like cell line and one wild-type, which are both derived from the same (mosaic) patient [30]. This cell line couple has the advantage of being both isogenic and expressing physiologic levels of the FA gene.

The aim of this study was to compare the expression profiles of FA cells defective in one of the FA core complex proteins with isogenic control cells at baseline, as well as in response to treatment with discriminating concentrations of crosslinking agents. Differences in expression profiles of known genes might shed light on the cellular function of the FA proteins.

The FA proteins have a caretaker function and a defective FA pathway results in chromosomal instability, a phenotype that is aggravated by exposure of cells to cross-linking agents [1]. Here we have tried to identify differences in gene expression between FA and control cells that might shed light on the molecular role of the FA proteins. Although recent findings of BRCA2 mutations in a subset of FA patients and studies on the FANCD2 protein suggests a link between the FA pathway, BRCA1, and DNA repair [9, 14], the exact function of the FA core complex remains elusive.

We examined samples using oligonucleotide arrays harboring probes directed to approximately 12,000 different transcripts. Various types of crosslinking agents are used in FA research, such as mitomycin C, diepoxybutane, psoralen with UV, and cisplatin [1]. In the present study, we used cisplatin because the effect of this agent on gene expression has been well described and there is no need for metabolic activation of the drug (unlike mitomycin C). Relatively late effects, such as cell cycle arrest and apoptosis, have been documented for FA [16–23]. Our main interest, however, was to determine the effect on early response genes that preceded cell cycle arrest. Therefore, cells were pulse treated for one hour with cisplatin and samples taken shortly thereafter. The concentration of cisplatin used was chosen to achieve the highest discrimination between FA and control cells in growth inhibition assays. Although this cytotoxic dosage is relatively low compared to those used in prior studies of cisplatin-induced gene expression changes, we reasoned that a higher dosage would lead to indiscriminate growth inhibition in both FA and control cells [17, 18, 20]. In addition, these concentrations were found to be sufficient for inactivating tyrosine phosphorylation of CDC2, 24 hours post treatment, as assessed by Western blotting (data not shown).

The profiling analysis revealed differences in expression of a number of genes, many of which are difficult to relate to the FA pathway or are unlikely to be expressed in B-lymphocytes. Only a few genes are related to processes that are possibly linked to the FA defect, such as those involved in cell cycle regulation, chromatin structure, and DNA mismatch repair. However, using Western blot analysis with antibodies directed to the respective gene products, we were unable to confirm differences in protein levels for all these genes. Genes that previously have been shown to respond to cisplatin and that were present on the arrays, i.e. c-jun, c-fos, ercc1, gadd153, gadd45, pcna and p21 ^waf1/Cip1, were not changed under the conditions used, except for an increase in expression of p21 ^waf1/Cip1. Although the change in our experiments was < 2-fold, there was a clear induction of p21 ^waf1/Cip1 expression in FA cells 24 hours post treatment that was not seen in the control cells. Microarray based experiments to discover cisplatin induced differences in gene expression have been performed previously both using cell lines [35] and tissues from rats [36]. These experiments are however difficult to relate to our experiment due to much longer exposure time to similar concentrations of cisplatin used [35] or due to the differences in experimental approach [36]. Although in those experiments many genes were found to be differentially expressed, also p21 ^waf1/Cip1 was found to be induced by cisplatin treatment. The reason for not detecting other cisplatin-induced genes in our experiments might be the relatively low concentrations of cisplatin used. This suggests that p21 ^waf1/Cip1 (compared to other genes) regulation is very sensitive to cisplatin, at least in lymphoblastoid FA cells. p21 ^waf1/Cip1 is known to be a key regulator of G1 cell cycle arrest, but has more recently also been implicated in G2 cell cycle regulation [37]. This observation fits well with the previously observed upregulation of p53 and cyclin B proteins, inactivation of CDC2, and cell cycle arrest that follows exposure to low doses of cross-linking agents [17–19].

Earlier studies reported that FA lymphoblasts express increased levels of transcripts from genes connected to the interferon pathway, IRF-1 and MxA [10, 11]. Although the approaches used were different, these observations are not confirmed by our data. Both genes were detected as being expressed but showed no significant differences between FA and controls. In addition, the constitutive elevated expression of p21 ^waf1/Cip1 observed in FA cells [9] was not found at the level of mRNA in our experiments. However, we did observe a slight increased expression of p21 ^waf1/Cip1 protein in FA cells. This might be explained by posttranscriptional regulation of p21 ^waf1/Cip1 or by a relatively low sensitivity of the expression arrays, incapable of detecting minor increases in mRNA. Interestingly, the observed elevated levels of p21 ^waf1/Cip1 expression in FA cells is reminiscent of a number of other DNA repair defective cell types, i.e. those defective in ATM, ERCC1, BRCA1, and BRCA2 [38–42].

Except for p21 ^waf1/Cip1, these experiments did not reveal differences in gene expression that could explain the basic defect in FA, when looking at baseline or when looking for genes that preceded cisplatin-induced cell cycle changes. There are several possible explanations possible for this. First, no significant changes at the transcriptional level are present in FA cells. It might well be that phenotypic differences of FA cells are predominantly reflected at the posttranslational level. The recent finding of FANCD2 mono-ubiquitination suggests that regulation by protein modification is an important step in the FA pathway [14]. Second, subtle changes in gene expression might remain undetected. The up- or down-regulation of an entire functional pathway may have large effects but be difficult to trace using these types of experiments, particularly when our knowledge of components involved in the various functional pathways is limited. Third, expression of certain genes may be different but not detected either because of lack of sensitivity or absence of the oligonucleotide sequence for those specific genes in the arrays used. Concerning sensitivity, it might be that the response of cells to cross-linker damage is related to signals at a certain stage of the cell cycle (e.g. during replication), resulting in a dilution of the response when using an asynchronous population of cells. The p21 ^waf1/Cip1 results showed that differences in expression of transcripts with high intensity on the arrays were detected even though this was a single data point and the difference was < 2-fold. Therefore, differences < 2-fold in genes with low intensity signals are presumably present but missed because of the increasing levels of noise in that region of intensity, making it difficult to discriminate between true and false positives in a limited number of samples.

Analysis of the expression data resulted in the identification of 10 genes (Table 1) that were found to be consistently differentially expressed between FA and control cells, as well as 59 genes (Table 2) that exhibited different response patterns between FA and control cells after treatment with cisplatin. However, subsequent analysis of selected genes at the protein level by Western blot analysis did not confirm the observed differences in expression. Several limitations of the experimental approach were discussed that might explain these results.

The observed strong cisplatin-induction of p21 ^waf1/Cip1 in FA cells, but not in control cells, suggests that FA cells respond similarly as non-FA cells but in a hypersensitive manner. According to experiments from other investigators [20] cell cycle checkpoints function properly, suggesting that cross-linking agents evoke more damage or create damage that remains unrepaired in cells with a defective FA pathway. Therefore, the observed increase in basal p21 ^waf1/Cip1 levels in FA cells by Western blot analysis most likely reflects a physiologic response to the presence of spontaneous damage in a subpopulation of these cells.