Activated TNF-α/NF-κB signaling via down-regulation of Fas-associated factor 1 in asbestos-induced mesotheliomas from Arf knockout mice

Abstract

The human CDKN2A locus encodes 2 distinct proteins, p16(INK4A) and p14(ARF) [mouse p19(Arf)], designated INK4A (inhibitor of cyclin dependent kinase 4) and ARF (alternative reading frame) here, that are translated from alternatively spliced mRNAs. Human ARF is implicated as a tumor suppressor gene, mainly in association with the simultaneous deletion of INK4A. However, questions remain as to whether loss of ARF alone is sufficient to drive tumorigenesis. Here, we report that mice deficient for Arf are susceptible to accelerated asbestos-induced malignant mesothelioma (MM). MMs arising in Arf (+/−) mice consistently exhibit biallelic inactivation of Arf, but, unexpectedly, do not acquire additional recurrent genetic alterations that we previously identified in asbestos-induced MMs arising in Nf2 (+/−) mice. Array CGH analysis was used to detect a recurrent deletion at chromosome 4C6 in MMs from Arf (+/−) mice. A candidate gene in this region, Faf1 (FAS-associated factor 1), was further explored, because it encodes a protein implicated in tumor cell survival and in the pathogenesis of some human tumor types. We confirmed hemizygous loss of Faf1 and down-regulation of Faf1 protein in a series of MMs from Arf (+/−) mice, and we then showed that Faf1 regulates TNF-α-mediated NF-κB signaling, a pathway previously implicated in asbestos-induced oncogenesis of human mesothelial cells. Collectively, these data indicate that Arf inactivation has a significant role in driving MM pathogenesis, and implicate Faf1 as a key component in the TNF-α/NF-κB signaling node that has now been independently implicated in asbestos-induced oncogenesis.

The CDKN2A (INK4a/ARF) locus encodes 2 distinct proteins translated from alternatively spliced mRNAs; p16(INK4A), designated as INK4A (inhibitor of cyclin dependent kinase 4) here, is encoded by exons 1α, 2, and 3. The alternate product p14(ARF), dubbed ARF (alternative reading frame protein) here, is specified by exons 1β, 2, and 3 (1, 2). Amino acid sequences of INK4A and ARF are unrelated, because exons 1α and 1β show no homology and exon 2/3 sequences are translated in different reading frames.

The CDKN2A locus is among the most commonly mutated genomic sites in human cancer (2). Point mutations or deletions specifically affecting exon 1α of INK4A are not uncommon, although intragenic mutations affecting exon 1β of ARF are seldom, if ever, observed (2, 3). Overall, the high frequency of concurrent INK4A and ARF loss has made it difficult to assess the contribution of ARF to human tumorigenesis.

We previously demonstrated that the CDKN2A locus is homozygously deleted in most malignant mesothelioma (MM) cell lines and in many MM tumor specimens (4). FISH analysis revealed a high incidence (≈50 to 75%) of homozygous deletions in frozen MM specimens or MM cells cultured for ≤5 days (5, 6). Reexpression of INK4A in MM cells resulted in cell cycle arrest, cell death, as well as tumor suppression and regression (7). In contrast to the established role of INK4A, the involvement of ARF in MM is less understood. However, adenovirus-mediated transfer of ARF in human MM cell lines has been shown to induce G₁-phase cell cycle arrest and apoptotic cell death (8), suggesting that ARF, like INK4A, is an important target of 9p21 deletions in MM.

Arf knockout mice represent an invaluable resource to test the relevance of ARF to MM pathogenesis. The Arf mouse model used here was generated by replacing the exon 1β of Arf with a Pgk-Neomycin cassette, leaving Ink4a sequences intact (9). Arf null (−/−) mice are highly prone to tumors, developing undifferentiated sarcomas, carcinomas, and tumors of the nervous system within 1 year of age; heterozygous Arf (+/−) mice develop tumors at a lower rate (20%) and longer latency.

Here, we show that Arf (+/−) mice exposed to asbestos are predisposed to MM, with a significantly shorter tumor latency compared with wild-type littermates. The murine MMs exhibit biallelic inactivation of the remaining wild-type Arf allele, although they do not exhibit the typical profile of tumor suppressor gene inactivation observed in human MMs. Array (a)CGH analysis led us to identify hemizygous loss of the Faf1 (FAS-associated factor 1) locus, resulting in aberrant TNF-α-induced NF-κB signaling, a pathway previously implicated in asbestos-induced oncogenesis and MM cell survival (10, 11).

Results

Arf (+/−) Mice Are Susceptible to Asbestos-Induced MM.

After repeated injections of crocidolite asbestos, we found markedly accelerated MM development in asbestos-treated Arf (+/−) mice, compared with wild-type littermates (Fig. 1A), whereas none of the control TiO₂-treated mice developed MM. The median latency for detection of MM in Arf (+/−) mice was 42 weeks after initial asbestos exposure, compared with 56 weeks in wild-type mice. A log-rank test demonstrated that Arf (+/−) mice had decreased survival times, compared with wild-type mice (P = 4.63E-13).

Fig. 1.

Arf (+/−) mice exhibit decreased MM latency compared with wild-type littermates. (A) Comparison of survival in asbestos-treated Arf (+/−) and wild-type mice, depicted by Kaplan–Meier survival curves. Arf (+/−) mice showed significantly shorter survivals than wild-type mice based on the log-rank test (P = 4.63e-13). One Arf (+/−) mouse and 5 wild-type mice had no obvious tumors and were excluded as censored observations. (B) Summary of MMs arising in asbestos-treated Arf (+/−) and wild-type mice. (C) Representative histopathology of MMs from Arf (+/−) mice treated with asbestos. (Top) Noninvasive epithelial MM growing on the abdominal surface of the diaphragm. Malignant cells exfoliate from the surface of the tumor. (Middle) Bloody tumor ascites. Malignant epithelial MM cells grow in suspension within the peritoneal cavity. (Bottom) MMs grow as solid tumor spheroids within the peritoneal cavity and may attach to the parietal and visceral serosal linings. Sections were stained with hematoxylin and eosin.

Epithelial, mixed, and sarcomatoid histologies were observed in asbestos-treated Arf (+/−) and wild-type mice, although epithelial morphology predominated (Fig. 1B). The growth and histological features of the MMs in this series of Arf (+/−) mice are similar to asbestos-induced tumors in previous chronic carcinogenicity assays in wild-type and Tp53-deficient (+/−) mice (12, 13). Epithelial MMs frequently presented with tumor ascites, spheroids, and diffuse peritoneal seeding on the serosal lining (Fig. 1C). In general, the MMs in Arf (+/−) mice did not show extensive invasion of the lymphatics and skeletal muscle, unlike tumors from Tp53 (+/−) or Nf2 (+/−) mice. Overall, 8 of 25 MMs from the asbestos-treated Arf (+/−) mice exhibited invasion into the lymphatics or muscle tissue, and the incidence of invasion was equivalent to that of MMs from wild-type mice (9 of 22 MMs).

MMs from Arf (+/−) Mice Exhibit Biallelic Inactivation.

MM cells from ascites or peritoneal lavage were cultured. We found biallelic inactivation of Arf in MMs from Arf (+/−) mice, demonstrated by the loss of the wild-type allele in all 11 primary MM cultures analyzed, along with the fact that the mutant knockout allele was retained (not deleted) in 10 of the same 11 tumors (Fig. 2A). In comparison, loss of both wild-type Arf alleles was observed in 3 of 7 MMs from asbestos-treated wild-type mice, and down-regulation of Arf expression was observed in 2 additional MMs in this series, bringing the total to 5 of 7 (71%) MMs from wild-type littermates with loss of functional Arf (Fig. 2A). Immunoblot analysis of a subset of tumor cell lysates from Arf (+/−) and wild-type mice confirmed the RT-PCR results (Fig. 2B).

Fig. 2.

MMs arising in Arf (+/−) exhibit biallelic inactivation of Arf, but do not acquire other tumor suppressor losses typical of human MMs. (A) Composite of genomic PCRs and RT-PCRs for p19Arf exon 1β, p16Ink4a exon 1α, p15Ink4b, Nf2, and p53 from primary MM cells (passage ≤5); β-actin and 18S protein were controls for template DNA and RNA, respectively. Open circles indicate absence, and solid circles indicate retention of at least 1 allele. Note that MMs from all Arf (+/−) mice were confirmed to retain the mutant knockout allele. (B) Immunoblot of MM cells showing retention of residual p16Ink4a in MMs from all Arf (+/−) mice except tumor 124 (shown in A to have loss of p16Ink4a, p19Arf, and p15Ink4b). Only the MM from wild-type mouse 132 retained Arf protein; this tumor instead had loss of p53. Expression of Nf2 (merlin) was observed in all MMs; β-actin was a loading control.

Our previous deletion mapping of human MM cell lines uncovered frequent homozygous deletions of chromosome 9p21 (4), which contains INK4A, ARF, and p15(INK4B), referred to here as INK4B. We found that MM cells from asbestos-treated Arf (+/−) mice seldom had homozygous deletions of Ink4a or Ink4b. In Arf (+/−) mice, absence of p53 was not observed in any of the 11 MMs from Arf (+/−) mice (Fig. 2 A and B). One of the MMs from a wild-type mouse showed loss of Tp53, but retention of Arf (wild-type mouse 132). Overall, the reciprocal association between Arf and Tp53 loss was similar to that observed in previous studies of Nf2-deficient (+/−) mice and human MMs.

A Recurrent Deletion in MM Cells from Arf-Deficient Mice Implicates Faf1.

We performed aCGH on primary MM cells from 4 Arf (+/−) mice to identify recurrent genomic imbalances connected with MM in these mice. Several inconsistent genomic changes, primarily loss or gain of whole chromosomes, were observed. The only recurrent change was a focal loss of an ≈2-Mb segment in chromosome 4, band C6, which was identified in all 4 MM cultures. This region is distal to a homozygous loss of the Cdk2na locus observed in MM cells from wild-type mouse 104 (Fig. 3A), identified in Fig. 2 as having loss of Ink4a/Arf DNA, RNA, and protein. Closer examination of the chromosome 4C6 deletion revealed that it was a copy number variant (CNV), i.e., a germ-line polymorphism (Fig. 3B). However, a hemizygous region of loss was detected in the distal shoulder of the CNV peak. A candidate target gene located in the region of hemizygous loss (Fig. 3C), Faf1, was further pursued, because it encodes a factor that regulates cell survival (14, 15).

Fig. 3.

Array CGH of murine MMs reveals recurrent genomic loss in chromosome 4, band C6, near a CNV distal to the Cdk2na (Ink4/Arf) locus. (A) Chromosome 4 view of representative MMs from Arf (+/−) mice showing CNV (arrow) distal to homozygous deletion of the Ink4a/Arf locus observed in MM cells from wild-type littermate 104. Dots indicate oligonucleotides on CGH Analytics scatter plot with negative values. (B) Log2 ratios confirming CNV (arrow) in chromosome 4. Log2 ratios of 2 to −2 are considered normal variances (Upper). Approximate base pair locations in chromosome 4 are indicated. Plot of CNVs in the region indicates that there actually may be 2 separate peaks in this region (Lower). Normal chromosome copy number is 2, and values of 0 indicate deletion of the region. (C) Representative view of proximal side of the peak in band 4C6 indicating a hemizygous loss of Faf1 (Agilent CGH Analytics 3.4 gene view). Profile corresponds to genomic DNA from MM cells from Arf (+/−) mouse 128 hybridized to Agilent 244K chip. Line corresponding to the moving average was calculated with preset linear algorithm, 1-Mb window. Other genes in vicinity are Dmrta2 and Elavl4/HuD.

Allelic loss of Faf1 in MMs from Arf (+/−) mice was validated by semiquantitative genomic PCR (Fig. 4A). Interestingly, immunoblot analysis showed that Faf1 was down-regulated in most (6 of 7) MMs from Arf (+/−) mice, as well as a subset (4 of 7) of MMs from wild-type mice (Fig. 4B). Frequent down-regulation of FAF1 protein expression was also observed in 7 of 7 human MM cell lines when compared with normal mesothelial cells, and loss of FAF1 staining occurred in 12 of 14 human MM tumors [supporting information (SI) Fig. S1].

Fig. 4.

Confirmation of Faf1 down-regulation in asbestos-induced mouse MM cells. (A) Semiquantitative genomic PCR of Faf1/Gapdh levels in MM cells from wild-type and Arf (+/−) mice used for aCGH analysis. (B) Immunoblot of Faf1 levels in MMs from wild-type and Arf (+/−) mice. Asterisk(s) indicate cell lines used for semiquantitative genomic PCR analysis in A.

Faf1 Correlates with TNF-α-Induced Nuclear p65 and NF-κB Activity in MM Cells.

Previous reports have implicated Faf1 as a negative regulator of NF-κB through inhibition p65/RelA nuclear localization (14, 15). To evaluate whether Faf1 down-regulation correlates with increased nuclear p65 localization, MM cells were stimulated with TNF-α, a known inducer of NF-κB in mesothelial cells. Nuclear p65 levels were evaluated by cellular fractionation, followed by immunoblot analysis (Fig. 5B). At all times tested, Faf1-negative MM cells from 2 Arf (+/−) mice exhibited increased nuclear p65, compared with Faf1-positive MM cells from a wild-type mouse (Fig. 5 B and C). Importantly, there was no difference in total p65 protein between the 3 MM cultures, suggesting that Faf1 status correlates specifically with TNF-α-induced nuclear p65 levels (Fig. 5A). The cyclic increase and decrease of p65 in the nucleus over time in all 3 MM cultures (Fig. 5 A and B) is consistent with other reports in which a bimodal, temporal-regulated signal has been observed that is indicative of the strength and duration of NF-κB signaling in response to stimuli (16, 17). Increased nuclear p65 levels were also detected in Faf1-negative versus Faf1-positive MM cells by using immunofluorescence analysis to monitor p65 localization in cells stimulated with TNF-α (Fig. S2). Together, these data correlate decreased expression of Faf1 with increased TNF-α-induced nuclear p65 in MM cells.

Fig. 5.

Faf1-deficient MMs exhibit more TNF-α-induced p65 nuclear accumulation and NF-κB activity. (A) Immunoblot of Faf1, p65, and actin in MMs from wild-type mouse 104 (Faf1-positive) and Arf (+/−) mice 110 and 129 (Faf1-negative). (B) Immunoblot of p65 and lamin B in nuclear extracts from TNF-α-treated (0, 0.5, 1, and 4 h) MM cells from tumors 104, 110, and 129. (C) Semiquantitation of nuclear p65 by immunoblotting (3 independent experiments) for TNF-α-treated cells. (D) Luciferase assay of NF-κB activity in MM cells treated with TNF-α for 24 h.

We next examined whether loss of Faf1, as a consequence of increased nuclear p65, correlates with increased NF-κB activity. Using a NF-κB-luciferase reporter construct, we evaluated relative NF-κB activity in Faf1-positive versus Faf1-negative MM cells stimulated with TNF-α. As shown in Fig. 5D, we observed increased NF-κB-luciferase activity in both TNF-α-stimulated and TNF-α-unstimulated Faf1-deficient cells, compared with that observed in Faf1-expressing cells.

To determine whether Faf1 directly regulates NF-κB signaling in MM cells, we reexpressed Faf1 in Faf1-deficient cells, and evaluated nuclear p65 levels and NF-κB activity in response to TNF-α. MM 110 and 129 cells from Arf (+/−) mice were nucleofected with Faf1 or empty vector (Fig. 6A). Faf1-positive MM 104 cells from wild-type mice were included as a control. Reexpression of Faf1 resulted in reduced nuclear p65 levels in Faf1-deficient cells, with levels similar to those observed in Faf1-positive cells (Fig. S3). Collectively, these data directly link Faf1 with control of nuclear p65 levels during TNF-α stimulated NF-κB signaling.

Fig. 6.

NF-κB activity is regulated by reexpression or knock down of Faf1 in MM cells. (A) reexpression in Faf1-deficient MM cells reduces NF-κB activity. (Left) Immunoblot of Faf1, p65, and β-actin levels in MM cells, from wild-type mouse 104 and Arf (+/−) mice 110 and 129, nucleofected with pcDNA3.1-Faf1 or pcDNA3.1 alone. (Right) Luciferase assay comparing relative NF-κB activity in MM cells treated with TNF-α for 24 h. (B) Knock down of Faf1 increases NF-κB activity in Faf1-positive MM cells treated with TNF-α for 24 h. (Left) Immunoblot of Faf1 and β-actin levels in MM cells from wild-type mouse nucleofected with siControl or siFaf1 (40 pmol each). Relative knock down is expressed as the KD (knock down) ratio and was calculated by densitometry of Faf1 levels/β-actin levels. (Right) Luciferase assay of relative NF-κB activity in TNF-α-treated Faf1-positive MM 104 cells nucleofected with siControl or siFaf1 no. 1, 2, or 3.

We next evaluated whether reexpression of Faf1 reduces activation of NF-κB activity in Faf1-deficient MM cells. Consistent with nuclear p65 levels, reexpression of Faf1 in MM 110 and 129 cells resulted in reduced NF-κB luciferase activity, compared with that seen in control cells (Fig. 6A). NF-κB activity was reduced in MM cells reexpressing Faf1, suggesting that Faf1 levels directly regulate NF-κB signaling in MM cells.

To address whether Faf1 loss is sufficient to increase NF-κB signaling, siRNA was used to knock down Faf1 in Faf1-positive MM 104 cells. Using 3 different RNAi sequences, we were able to knock down Faf1 levels by up to 63% (Fig. 6B). TNF-α-induced NF-κB luciferase activity was evaluated to determine whether knock down of Faf1 is sufficient to up-regulate NF-κB signaling. As shown in Fig. 6B, higher NF-κB activity was observed with all 3 siRNA used when compared with cells nucleofected with control siRNA, independent of TNF-α treatment. Together, these data implicate Faf1 as an important regulator of the NF-κB pathway, whose loss may contribute to aberrant NF-κB signaling and tumor progression in our mouse model of asbestos-induced MM.

Discussion

A paradigm for MM tumor progression is undefined, although it is clear that recurrent loss/inactivation of several tumor suppressor genes is likely to be involved in a multistep genetic progression cascade (18). In particular, the CDKN2A/ARF and CDKN2B loci on human chromosome 9 and the NF2 locus on chromosome 22 are known to be frequently inactivated in human MMs (4).

There is considerable interest in employing genetically defined mouse models to better understand the significance of specific tumor suppressor genes in the pathogenesis of MM. Asbestos-treated Nf2 (+/−) mice have been shown to have increased susceptibility and accelerated onset of asbestos-induced MM, compared with control mice, and the tumors consistently show loss of the remaining wild-type Nf2 allele (19, 20). We reported that Nf2 (+/−) mice exposed to asbestos fibers recapitulate molecular features of human MM, including activation of Akt signaling and biallelic inactivation of the tumor suppressor genes Ink4a, Arf, Ink4b, and Nf2 (20). Similar inactivation of tumor suppressor genes in MMs from Nf2-deficient mice was reported by others (21).

Mice lacking Arf are highly prone to tumor development, with undifferentiated sarcomas predominating over lymphomas, unlike Tp53-null mice (9). We report here that Arf (+/−) mice have a markedly increased susceptibility to asbestos-induced MM, compared with wild-type littermates. Interestingly, biallelic inactivation of Arf was not accompanied by loss of Ink4a, providing further support for a significant independent role of Arf loss/inactivation in MM, and is not simply a bystander effect that accompanies loss of Ink4a. These findings are in keeping with the initial characterization of Arf mice, where spontaneous tumors were found to retain expression of Ink4a and Ink4b (9). These findings also are consistent with those of Sharpless et al. (22), who reported that homozygous deletions of Ink4a and Arf have significant and nonredundant roles in suppressing malignant transformation in vivo.

Inactivation of Tp53 was rarely observed in MMs from Arf (+/−) mice. This finding is consistent with the reciprocal association between inactivation of Arf and Tp53 observed in MMs from humans and Nf2 (+/−) mice (20), i.e., inactivation of Arf and retention of p53, or vice versa, is sufficient to promote MM tumorigenesis. In contrast to MMs arising in humans and in Nf2 (+/−) mice, inactivation of Ink4a, Ink4b, and Nf2 was not frequently observed in MMs from Arf (+/−) mice. Although Arf deficiency predisposes to accelerated onset of asbestos-induced MM, the fact that MMs seen in Arf (+/−) mice were not highly invasive suggests that additional somatic genetic alterations are required for extensive invasion and metastasis.

Variation in DNA copy number is increasingly recognized as a factor contributing to genetic diversity (23, 24). CNVs theoretically could affect gene expression or function that contributes to disease susceptibility. In our study, aCGH revealed a striking recurrent CNV in chromosome 4C6, with a less prominent copy number loss near this CNV. The general location of the CNV in band 4C6, in a region that has synteny to human chromosome 1p32.3-1p33, was intriguing given that several CGH studies have implicated loss in/near this chromosomal region in human tumors, including MM. In particular, D1S427-FAF1 was frequently lost in human uterine cervix carcinomas (25). This finding and the fact that Faf1 encodes a factor involved in the regulation of cell survival led us to pursue the possible involvement of this gene in MMs from Arf (+/−) mice. Recurrent loss and down-regulation of Faf1 was observed in our MMs. Other known candidate genes located in the deleted region, i.e., Dmrta2 (Doublesex- and mab-3-related transcription factor A2) and Elavl4/HuD (embryonic lethal, abnormal vision, Drosophila-like 4), were not pursued at this time.

FAF1 is evolutionarily conserved and involved in various biological processes. Although FAF1 was identified as a member of the Fas death-inducing signaling complex (DISC) (26–28), we did not find that Faf1 loss contributed to attenuation of Fas receptor/DISC-mediated caspase cleavage in our mouse MM cells. Thus, we believe loss of Faf1 leading to deregulation of the NFκB pathway may be a more important activity than its contribution to Fas receptor-mediated apoptosis in this murine model of MM. NF-κB activation induced by TNF-α, interleukin-1β, or lipopolysaccharide was shown to be inhibited by FAF1 overexpression by preventing translocation of the RelA (p65) subunit of NF-κB into the nucleus and decreasing its DNA-binding activity on TNF-α treatment (14). Like human FAF1, we showed that mouse Faf1 regulates NF-κB activation induced by TNF-α, preventing translocation of p65 into the nucleus. Both restoration of Faf1 in MM cells and knock down of Faf1 with siRNA in Faf1-positive MM cells had the predicted effect on NF-κB signaling. Also, although we originally identified down-regulation of Faf1 in MMs from Arf (+/−) mice, some MMs from wild-type mice also showed down-regulation of Faf1. We also observed Faf1 down-regulation in human MM cell lines and tumors; thus, reduced Faf1 expression may have implications in human as well as mouse MM.

Several groups have shown that TNF-α signaling is important in human MM pathogenesis. TNF-α has been shown to inhibit asbestos-induced cytotoxicity via a NF-κB-dependent pathway (10). A model was proposed whereby exposure to asbestos leads to the accumulation of macrophages and release of TNF-α, whereas asbestos simultaneously induces human mesothelial cells to secrete TNF-α. TNF-α then activates NF-κB to promote mesothelial cell survival and permits cells with asbestos-induced DNA damage to divide rather than undergo apoptosis (10). NF-κB has been shown to be a constitutive survival factor in transformed human mesothelial cells, as well as MM tumor cells (11). Inhibition of NF-κB activity in MM cells through the pleiotropic actions of Bortezomib, an inhibitor of the 20S proteasome that also inhibits IκB degradation, induces cell cycle blockage and apoptosis in vitro, as well as tumor growth inhibition in vivo (11). Also, therapeutic targeting of TNF-α/NF-κB signaling decreases drug resistance and increases cytotoxicity in MM cells (29). Collectively, our findings validate a role for the TNF-α/NF-κB pathway in MM pathogenesis, and the Arf (+/−) model described here should prove useful for preclinical testing of therapies targeting TNF-α/NF-κB signaling in MM tumors.

The mouse model used here targets the p53 pathway through Arf deficiency to accelerate MM development. Similarly, Tp53-deficient mice exhibit accelerated MM after asbestos exposure (12, 13), and hamsters infected with SV40, to inhibit p53 via large T antigen expression, also show markedly increased incidence and accelerated development of MM (30). Therefore, the investigations presented here, together with these former studies, demonstrate that genetic or viral alterations that impair the p53 pathway, directly or indirectly, increase the risk of asbestos-mediated carcinogenesis. These findings may have implications for cancer prevention, because now 3 independent lines of research indicate that an impaired p53 pathway leads to an increased risk of developing MM after asbestos exposure.

Materials and Methods

Animals.

Arf (+/−) breeding pairs were a gift from C. Sherr (St. Jude Children's Research Hospital, Memphis, TN). Mice were housed and treated according to guidelines established by the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Male Arf (+/−) and wild-type littermates were genotyped as described (31).

Treatment of Mice.

Union Internationale Contre le Cancer (UICC)-grade crocidolite asbestos was obtained from SPI Supplies, and TiO₂ particles were obtained from Aldrich Chemicals. We used methodology described previously (20), injecting Arf (+/−) mice i.p. with crocidolite asbestos or control TiO₂. Although not a fiber, TiO₂ has been used in previous i.p. injection and inhalation studies (32), because of low toxicity and poor solubility. For details, see SI Materials and Methods. A total of 26 Arf (+/−) and 27 wild-type mice treated with asbestos, and 13 Arf (+/−) mice and 18 wild-type littermates treated with TiO₂ completed the protocol. Other than MMs in Arf (+/−) mice, 1 rhabdomyosarcoma and 1 lung adenoma were found in the asbestos-treated group, and 1 hemangiosarcoma was found in the TiO₂-treated group. Two Arf (+/−) and 1 wild-type mouse did not complete asbestos treatments, and were not analyzable for cause of death.

Primary Cell Cultures.

Primary murine MM cells were isolated from ascitic fluid and/or lavage of the peritoneal cavity, as previously described (20). Early passage cells from asbestos-treated mice were evaluated for mesothelial markers. MMs from both wild-type mice and Arf (+/−) mice typically expressed a panel of markers as assessed by RT-PCR, including WT1, cytokeratins 18 and 19, N-cadherin, and E-cadherin.

PCR and RT-PCR.

Primers for genomic DNA and RT-PCR for the analysis of Arf, Ink4a, Ink4b, Nf2, and p53, as well as Wt1, Krt18 (cytokeratin 18), Krt19, Cdh1, and Cdh2 were described previously (20). Primers for murine β-actin were purchased from Clontech Laboratories. Primer sets for Faf1 and Gapdh and corresponding PCR conditions are described in SI Materials and Methods.

Western Blottings.

Immunoblot analysis of protein lysates was performed as described elsewhere (33), and in more detail online (SI Materials and Methods); 15–30 μg of protein/sample was subjected to immunoblot analysis. Antibodies included anti-Arf from Abcam, anti-FAF1 (Ab1) from Cell Signaling Technology, anti-FAF1 (Ab2) from BioVision, and anti-Ink4a (M-156), anti-Nf2 (H-260), anti-NF-κB p65 (C-20), anti-lamin B (C20), and anti-β-actin (I-19) from Santa Cruz Biotechnology, and anti-p53 (NCL-p53–505) from Novocastra.

Array-Based CGH.

Array CGH analysis was performed by using 44K and/or 244K Genomic DNA Arrays (Agilent), according to the manufacturer's instructions (SI Materials and Methods). Slides were scanned by using an Agilent scanner. Data were extracted by using Feature Extraction Software from Agilent. Output files were imported into Agilent CGH Analytics for DNA Copy Number Analysis.

Plasmids, siRNA, and Nucleofection.

FAF1 cDNA (clone ID 2958207) was from Open Biosystems. A 1950-bp fragment encoding FAF1 was excised with EcoRI and XhoI, and subcloned into pcDNA3.1. The pcDNA3.1-FAF1 plasmid was nucleofected into mouse MM cells by using an Amaxa Nucleofector II and Cell Line Nucleofector Kit R with program U-030. PathDetect pNFκB-Luc Cis reporter plasmid was from Stratagene. Stealth siRNAs against Mus musculus Faf1 (oligo ID MSS204189, MSS204190, and MSS204188) were from Invitrogen; 400 pmol of each siRNA molecule were nucleofected into mouse MM cells by using Amaxa Cell Line Nucleofector Kit R with program T-20.

Luciferase Reporter Assays.

MM cells were nucleofected with 1 μg of PathDetect pNF-κB-Luc Cis and 0.1 μg of pRenilla-Luc (Promega). At 24-h postnucleofection, cells were treated or not with 50 ng/mL mouse TNF-α (Sigma) for 24 h before being harvested for Dual-Luciferase Reporter Assay (Promega). For Faf1 reexpression, 1 μg of pcDNA3.1-FAF1 or empty plasmid were nucleofected into Faf1-deficient MM cells, along with pNF-κB-Luc Cis and pRenilla-Luc. To determine whether knock down of Faf1 increases NF-κB activity, Stealth siRNAs against Mus musculus Faf1 were nucleofected into MM cells as per manufacturer's instructions. After 48 h, cells were treated or not with 50 ng/mL mouse TNF-α for 24 h before luciferase assay. Cells were harvested in 1X lysis buffer (Promega) and freeze-thawed once before clarifying the lysate by centrifugation at 14,000 × g for 10 min; 20 μL of sample was used for both luciferase assay and Bio-Rad protein assay according to the manufacturer's protocols. NF-κB luciferase values were normalized by both Renilla luciferase activity (transfection efficiency) and protein concentration.

Cell Fractionation.

Cell fractions were isolated with NE-PER Nuclear and Cytoplasmic Extraction Reagents (Pierce Biotechnology), following the manufacturer's recommendations. Nuclear protein extracts were quantitated by using a Bio-Rad protein assay, and equal concentrations for each treatment were subjected to immunoblot analysis. Nuclear p65 levels were semiquantitated by densitometry analysis and normalized to relative lamin B levels as a measure of nuclear extraction efficiency.