RESULTS

Generation of an inducible, liver-specific Mtf1 knockout mouse line
Using a homologous recombination strategy, mice were obtained with a modified Mtf1 allele Mtf1loxP where exons 3 and 4, encoding four of the six zinc fingers of the DNA-binding domain, are flanked by loxP sites (Figure 1a). Mice homozygous for the Mtf1loxP allele were further crossed with animals of the Cre recombinase transgenic line Mx-cre. Cre recombinase is expressed in this line under the control of the mouse Mx1 gene promoter, which is inducible by administration of interferon alpha or beta, or synthetic double-stranded RNA pI–pC (32). Cre-mediated deletion was reported to be complete in the liver, while varying in other tissues, ranging from 94% in spleen to 8% in brain (32). After Cre-mediated deletion of exons 3 and 4, which results in a frameshift and premature translation stop, no functional MTF-1 protein can be produced.
For induction of Cre recombinase, Mtf1 conditional knockout mice harboring the Mx-cre transgene (Mtf1Mx-cre) received four intraperitoneal pI–pC injections at 3 day intervals (pI–pC induction); control littermates without transgene (Mtf1loxP) received similar injections. Using RT–PCRs (Figure 1b), a shortened product was obtained with RNA from Mtf1Mx-cre livers, indicating a successful excision of exons 3 and 4 of Mtf1 in these animals. On close examination, a very faint band similar in size to full-length signal was also observed in those mice, probably due to a low amount of residual full-length Mtf1 mRNA. The level of functional MTF-1 protein was examined by EMSA (Figure 1c): MTF-1 protein–DNA complex was detectable with liver protein extract from an Mtf1loxP control mouse, but no functional MRE-binding protein was observed with an Mtf1Mx-cre sample. Thus, deletion of exons 3 and 4 of Mtf1 in the liver of Mtf1Mx-cre mice was virtually complete. All examined liver-specific knockout mice were viable under laboratory conditions and appeared normal.

MTF-1 target gene search
For the identification of MTF-1 target genes, we compared the liver transcript profiles of mice with and without functional Mtf1 gene that had been mock-treated or exposed to cadmium (n = 3 per genotype and respective treatment).
In a first screen, the transcripts were analyzed with a differential display-based method, called amplification of double-stranded cDNA end restriction fragments (ADDER) (33). Thereby an overwhelming number of signals was obtained for the two stress-inducible metallothioneins (Mt1 and Mt2), due to the abundance of their transcripts both in mock-treated and especially in cadmium-treated livers that harbored a functional MTF-1 gene (data not shown). This result confirmed the importance of MTF-1 for both basal and metal-induced expression of metallothionein genes.
In a second approach, the gene expression profile in livers of the above mentioned mice was compared by Affymetrix GeneChip® Mouse Genome 430 2.0 Arrays (Table 1). When analyzing the probe array data of livers from mock-treated Mtf1Mx-cre and Mtf1loxP mice, an at least 2-fold, reliable downregulation of expression was detected in Mtf1Mx-cre livers for 13 Affymetrix GeneChip® probe sets corresponding to 11 characterized genes (Table 1, a). Seven of these genes contain one or more MRE core consensus sequence TGCRCNC within a segment of 1000 bp upstream of the transcription start. For 26 probe sets corresponding to 24 different characterized genes, a 2-fold or higher, reliable upregulation was detected in Mtf1Mx-cre livers (Table 1, b); 17 of these 24 genes contain MRE core consensus sequences in the upstream region. The data set for livers of cadmium-treated Mtf1Mx-cre and Mtf1loxP mice revealed an at least 2-fold, reliable downregulation in Mtf1Mx-cre livers for 21 probe sets corresponding to 16 different characterized genes (Table 1, c); 10 of these contain MRE core consensus sequences in their upstream region. For 9 probe sets corresponding to 9 different characterized genes, an at least 2-fold, reliable upregulation was detected (Table 1, d); five of them contain MRE motifs. In addition to characterized genes, ESTs and RIKEN cDNA sequences were also found in the comparison of Mtf1Mx-cre and Mtf1loxP livers to be differentially expressed (Supplementary Table 1). Downregulation of Mt1 and Mt2 was detected in Mtf1Mx-cre livers for both conditions (though the level of significance for the downregulation of Mt1 in mock-treated animals was above 0.05; data not shown).
For all MTF-1 target genes characterized so far, such as Mt1, Mt2 and Znt1, MTF-1 exerts its transcriptional activation activity via standard MRE sequences located proximal to the transcription start (4,5,8,18,19). Even a specific search for MTF-1 binding sites by selection from a pool of double-stranded oligonucleotides with random sequences yielded no new binding motif for MTF-1 in addition to the known MREs (34). Thus, an MRE sequence is to date the only indication for a direct MTF-1 target gene, and four MRE-containing target gene candidates were further analyzed.

Basal expression of Sepw1 depends on MTF-1
Sepw1 was found in microarray analysis to be significantly downregulated in livers from cadmium- and mock-treated Mtf1Mx-cre mice (Table 1, a and c). SEPW1 is a selenocysteine-containing protein that binds glutathione (35) and is thought to act as an antioxidant in vivo (36).
Sepw1 expression in livers of pI–pC-induced, mock- or cadmium-treated Mtf1Mx-cre and Mtf1loxP mice was further analyzed by semiquantitative RT–PCRs and S1 analysis (Figure 2a and b). In accordance with microarray data a slight, if any, upregulation of Sepw1 transcription was observed in livers from Mtf1loxP mice upon cadmium treatment. The basal level was reduced in livers from mock- and cadmium-treated Mtf1Mx-cre mice, indicating that MTF-1 is important for the basal expression of Sepw1.
Three MRE core consensus sequences were found in the region upstream of the mouse Sepw1 transcription start (Figure 2c). Two of them in opposite orientation overlap almost completely proximal to the transcription start (MRE1, −40 bp), the third one is located further upstream (MRE2, −527 bp). Specific binding of MTF-1 to Sepw1 MRE1 but not MRE2 oligonucleotide was observed in EMSA with liver protein extract from an Mtf1loxP control mouse (Figure 2d). As a control, no binding to MRE1 was detected with extract from a pI–pC-induced Mtf1Mx-cre mouse, confirming that the bandshift was indeed dependent on the presence of MTF-1.

Cadmium response of Ndrg1 depends on MTF-1
Ndrg1 was significantly downregulated in microarrays of liver transcripts from cadmium-treated Mtf1Mx-cre mice compared to similarly treated Mtf1loxP control mice (Table 1, c). Ndrg1 probably has some role in stress response since various stimuli, including hypoxia and nickel compounds, activate expression of rodent Ndrg1 and/or its human ortholog (37–40).
The Ndrg1 microarray results were confirmed with semiquantitative RT–PCRs (Figure 3a): for Mtf1loxP control livers, a clear increase of Ndrg1 expression was observed after cadmium exposure; in livers from Mtf1Mx-cre mice, this cadmium response was not detectable, while basal expression was similar to controls. This indicates that cadmium-induced expression of Ndrg1 depends on MTF-1.
Five MRE core consensus sequences are located upstream of the mouse Ndrg1 transcription start (Figure 3b). Four of them are clustered (MRE1 to MRE4, −138 to −332 bp), the fifth one is located farther upstream (MRE5, −883 bp). EMSA was performed to test whether MTF-1 is interacting with some or all of the four proximal MRE sequences (Figure 3c). Separate oligonucleotides were tested for MRE1 and MRE2, whereas one oligonucleotide spanning both sequences was used for MRE3 and MRE4 (MRE3,4). No complex was seen with MRE1, but specific MTF-1 complexes were observed for both the MRE2 and MRE3,4 oligonucleotides with liver protein extract from an Mtf1loxP mouse. As expected, no bandshift was observed with protein extract from a mouse lacking MTF-1 (Mtf1Mx-cre).

Cadmium response of Csrp1 depends on MTF-1
Csrp1 was found in microarray analyses to be significantly downregulated in cadmium-treated Mtf1Mx-cre mice compared to Mtf1loxP mice (Table 1, c). CSRP1 is a member of the evolutionary conserved CRP family of proteins that have been implicated in myogenesis and cytoskeletal remodeling (41,42).
Semiquantitative RT–PCRs confirmed the microarray results, namely, that Csrp1 expression is elevated in Mtf1loxP livers upon cadmium exposure (Figure 4a). In contrast, no cadmium response was detectable in livers from Mtf1Mx-cre mice, suggesting that MTF-1 is required for cadmium induction of Csrp1.
Three MRE core consensus sequences were found upstream of the Csrp1 transcription start (MRE2 to MRE4, −56 to −366 bp), one was found immediately downstream (MRE1, +7 bp; Figure 4b). Specific binding of MTF-1 was observed with EMSA for MRE2 oligonucleotide and protein extract from an Mtf1loxP liver, but not an Mtf1Mx-cre liver extract lacking MTF-1, confirming the participation of MTF-1 in the complex (Figure 4c).

MTF-1 inhibits expression of Slc39a10
Slc39a10 was detected in microarray analysis to be significantly upregulated in livers from both mock- and cadmium-treated Mtf1Mx-cre mice compared to control animals (Table 1, b and d). SLC39 proteins are members of the Zrt- and Irt-like protein (ZIP) family of metal ion transporters that transport, with no known exception, metal ion substrates across cellular membranes into the cytoplasm (43,44).
In accordance with microarray data, semiquantitative RT–PCRs showed a downregulation of Slc39a10 expression in livers of Mtf1loxP mice upon cadmium exposure. In samples from Mtf1Mx-cre mice, the basal expression was significantly increased; cadmium treatment still resulted in a decrease of Slc39a10 expression (Figure 5a). It cannot be judged by this experiment whether the degree of cadmium-induced reduction of Slc39a10 transcription was identical for Mtf1Mx-cre and Mtf1loxP mice or lower in the absence of MTF-1. In microarray analysis, the degree of the downregulation was either comparable to the one in control livers or lower, depending on the considered Affymetrix GeneChip® probe set (data not shown). The results indicate that MTF-1 is involved in repression of the basal expression of Slc39a10. It might also participate in the cadmium response of this gene, but it is apparently not exclusively responsible.
One MRE core consensus sequence was found just upstream of the mouse Slc39a10 transcription start (MRE1, −21 bp), another one directly downstream (MRE2, +17 bp; Figure 5b). Specific binding of MTF-1 was observed in EMSA analysis for MRE2 with liver protein extract from an Mtf1loxP but not from an Mtf1Mx-cre mouse, while no binding was detected with MRE1 (Figure 5c).

Cadmium-responsive, MTF-1-independent genes
Finally, we also identified a number of cadmium-responsive genes that were independent of MTF-1 presence, by comparing the probe array data of all cadmium-treated mice with the data of all mock-treated mice, irrespective of the genotype (Table 2). An at least 2-fold, reliable upregulation was observed after cadmium exposure for 31 probe sets corresponding to 21 different characterized genes (Table 2, a). For 2 probe sets corresponding to 2 characterized genes, an at least 2-fold downregulation was detected (Table 2, b).
Several genes involved in the metabolism of the antioxidant glutathione were found to be upregulated by cadmium exposure, namely the genes encoding the catalytic subunit of glutamate-cysteine ligase (Gclc) that is the rate limiting enzyme in de novo synthesis of glutathione (45); glutathione reductase 1 (Gsr), the reducing enzyme for oxidized glutathione (45); and glutathione-S-transferase, mu 4 (Gstm4), which is a member of the glutathione-S-transferase supergene family of detoxification enzymes (45). In all of these cases, induction was confirmed by semiquantitative RT–PCRs (data not shown). Gclc, also referred to as heavy chain subunit of gamma-glutamylcysteine synthetase (Ggcs-hc), had been discussed previously as a target gene of MTF-1 (6). Our expression data indicate that Gclc is induced by cadmium but, at least in the adult mouse liver, not dependent on MTF-1.
To analyze the role of the glutathione system in the cellular cadmium response, mouse embryonic fibroblasts with and without functional Mtf1 were treated with cadmium in combination with BSO, a specific inhibitor of glutamate-cysteine ligase (31), and cell viability was assessed by a colorimetric assay based on the tetrazolium salt MTT (Figure 6). Increasing concentrations of BSO or cadmium alone were to some extent cytotoxic for the examined cell lines. Treatment with both BSO and cadmium resulted in an enhanced lethality particularly for the cells without functional Mtf1, indicating that a depletion of glutathione together with a lack of Mtf1 impair an efficient anti-cadmium defense. Thus, adequate glutathione supply as well as MTF-1 and its target genes are essential for the survival of the cell under cadmium stress.
Besides genes related to the glutathione pathway, several other stress-related genes were upregulated upon cadmium exposure, including genes for thioredoxin reductase 1 (Txnrd1), one of the reducing enzymes of the antioxidant thioredoxin (46); KDEL endoplasmic reticulum protein retention receptor 2 (Kdelr2) participating in ER stress response (47); and the anti-apoptotic Bcl2-associated athanogene 3 (Bag3) involved in stress-induced apoptosis (48).