TGFBI Deficiency Predisposes Mice to Spontaneous Tumor Development

Tumor growth and metastasis is a multistep process involving cell adhesion, proteolytic enzyme degradation of the extracellular matrix (ECM), and motility factors that influence cell migration (1, 2). Integrins are cell surface adhesive receptors composed of α- and β-chain heterocomplexes. Both subunits transverse the membrane and mediate the physical and functional interactions between cell and its surrounding ECM, thus serving as bidirectional transducers of extracellular and intracellular signals, which ultimately lead to regulation of adhesion, proliferation, differentiation, antiapoptosis, and tumor progression (3, 4).

TGFBI was first identified in a human lung adenocarcinoma cell line (A549) treated with transforming growth factor-β (5). This gene encodes a highly conserved 683–amino acid protein that contains a secretary signal sequence and four internal homologous domains, the last of which contains an RGD (Arg-Gly-Asp) motif that can serve as a ligand recognition site for integrins (5). TGFBI product is a component of ECM in lung and mediates cell adhesion and migration through interacting with integrin via integrin receptors: α3β1, αvβ3, and αvβ5 (6–10). It is ubiquitously expressed in human normal tissues; however, down-regulation or lost expression of this gene has been found in a list of human tumor cell lines, including lung, breast, colon, prostate, and leukemia, as well as in human primary lung and breast tumor specimens (11–14). CpG island hypermethylation in the promoter region, one of the mechanisms by which tumor suppressor genes are inactivated in human cancers, correlates with the silencing of TGFBI promoter and its subsequent down-expression (15). In vitro studies have implicated its role in maintaining microtubule stability and inhibiting tumorigenicity and tumor angiogenesis (12, 13, 16–19), suggesting a tumor suppressor function in vivo. To test this hypothesis, we have generated a TGFBI-null mouse model. The results showed, for the first time, that TGFBI loss promotes cell proliferation through aberrant activation of cyclic AMP–responsive element binding protein (CREB)-cyclin D1 pathway and predisposes mice to spontaneous tumor development.

Generation of TGFBI-null mice. The TGFBI locus was PCR cloned using a genomic 129 DNA as template. Linearized targeting vector DNA (70 μg) was electroporated into 129Sv/Ev embryonic stem (ES) cells. Heterozygous targeting ES cells were identified by Southern blot and two different targeted clones were microinjected into C57BL/6J blastocytes. Chimerical male mice were produced and mated with C57BL/6J females. Germ-line transmission of the targeted TGFBI allele was verified by Southern blot analysis of tail DNA from F1 offspring. All mice studied for spontaneous tumor development were the F2 generation of 129Sv/Ev × C57BL/6J crosses. Mice were genotyped by Southern analysis and PCR (primers and conditions available on request). Detailed necropsy and histology are described in Supplementary Data. Protocols were approved by the Institutional Animal Care and Use Committee (IACUC) of the Columbia University. Animal procedures were conducted in compliance with IACUC.

Cell cycle analysis. For S-phase reentry experiment, confluence-arrested cells were plated at 2 × 10⁶ per 100-mm dish in medium containing 0.1% serum for 24 h before adding medium containing 10% serum. Following serum stimulation, cells were pulsed with bromodeoxyuridine (BrdUrd) for 1 h at end of the time point and analyzed by FITC-BrdUrd Flow kit (BD Pharmingen).

Tumor cohorts and carcinogen treatment. For the 7,12-dimethylbenz(a)anthracene (DMBA)–treated cohort, 3- to 5-d-old mice were topically treated with 50 μL of 0.5% DMBA in acetone (Sigma) and monitored for up to 6 mo. Mice were visually examined weekly and were killed if any individual tumor reached a diameter of 5 mm or at the termination of the experiments. Skin tumors will be counted, excised, and examined by staining with H&E.

Chromatin immunoprecipitation analysis. Chromatin immunoprecipitation (ChIP) assays were carried out as previously described (20) and detected with quantitative PCR (Roter-Gene RG-3000A; ref. 21). Primers for mouse cyclin D1 gene were 5′-CCGGCTTTGATCTCTGCTTA-3′ (forward) and 5′-CGCGGAGTCTGTAGCTCTCT-3′ (reverse). Primers used for negative control were 5′-AGGTGGAGAAACACCACCAC-3′ (forward) and 5′-CGGTTTGCCCAAGAAAAATA-3′ (reverse).

Luciferase activity assay and constructs. Mouse embryonic fibroblasts (MEF) were transiently cotransfected with luciferase reporter and dominant-negative plasmid. Forty-eight hours later, relative luciferase activity was determined by Dual Luciferase Activity kit (Promega). Relative luciferase activity was normalized to concentration of cell lysates. The relative fold induction or suppression represents the relative intensity of the experimental sample divided by the relative intensity of the medium control.

Reverse transcription-PCR method and primers. The expression of cyclin D1 gene was analyzed by quantitative real-time reverse transcription-PCR (RT-PCR; Applied Biosystems 7300) using the procedures described previously (14). The primer sets are as follows: 5′-TCGTGGCCTCTAAGATGAAG-3′ and 5′-TTTTGGAGAGGAAGTGTTCG-3′ for mouse cyclin D1 and 5′-AAGGTCATCCCAGAGCTGAA-3′ and 5′-CTGCTTCACCACCTTCTTGA-3′ for mouse glyceraldehyde-3-phosphate dehydrogenase.

To verify the deletion of exons 4 to 6 in TGFBI^−/− MEFs, a pair of oligonucleotide primers specific for the upstream and downstream regions of exons 4 to 6 was designed. The primer sequences are as follows: upstream, 5′-CTCATGCGACTGCTGACCCTCGCTCTG (55-81); downstream, 5′-CAGCACACATGGCTGACTTCAGGATGTG (973-1000). cDNAs, synthesized from the total RNAs of wild-type (WT) and TGFBI^−/− MEFs, were used as the templates for PCR amplification. Using this approach, a 945-bp PCR product was detected in WT, whereas a 475-bp PCR product was identified in TGFBI^−/− MEFs due to the deletion of exons 4 to 6.

Antibodies and recombinant mouse TGFBI protein. Antibodies used in Western blots include mouse monoclonal cyclin D1 (Santa Cruz), rabbit phospho-CREB (p-CREB; Ser¹³³) monoclonal antibody (mAb; Cell Signaling), mouse β-actin mAb (Sigma), anti-CD3 (T-cell marker), and anti-CD45R/B220 (B-cell marker; BD Pharmingen). Rat anti-mouse TGFBI mAb and mouse recombinant TGFBI protein were purchased from R&D Systems.

Tumor development in mice lacking TGFBI gene. To explore the physiologic function of TGFBI and its role in tumorigenesis, TGFBI-deficient mice were generated by homologous recombination. The correct targeting resulted in the replacement of exons 4 to 6 of TGFBI gene in mice with a neomycin resistance gene (Fig. 1A) and was identified by Southern analysis (Fig. 1B). TGFBI^−/− MEFs still expressed TGFBI mRNA but the level was ∼6-fold lower than in WT MEFs (data not shown). Moreover, deletion of exons 4 to 6 in TGFBI^−/− MEFs was shown by RT-PCR (Fig. 1C), and absence of TGFBI protein was revealed by Western blot (Fig. 1D).

TGFBI^−/− mice arose from crosses of TGFBI^+/− mice at expected Mendelian frequency and showed a slower postnatal development with a 13.5 ± 3% lower body weight than that of sex-matched TGFBI^+/+ littermates from 2- to 6-month age (n = 10; Supplementary Fig. S1). Histologic surveys of liver, lung, kidney, stomach, intestine, and testis (n = 10 per genotype; age, 26 weeks) did not reveal morphologic abnormalities. However, 2 of 10 TGFBI^−/− mice showed splenomegaly that was identified as B-cell hyperplasia (Supplementary Fig. S2).

To assess the tumor suppressor activity of TGFBI in vivo, a large cohort of TGFBI^−/− (n = 54), TGFBI^+/− (n = 75), and TGFBI^+/+ (n = 48) animals generated from crosses of TGFBI^+/− mice was observed for the development of spontaneous malignancies for up to 20 months. Mice were sacrificed for complete necropsies either at earlier time due to clinical features of systemic illness (weight loss, inactivity, ruffling of fur, and hunched posture) or when reaching end of the observation period. From ages of 9 to 16 months, >20% of TGFBI^−/− mice died of systemic illness, whereas all TGFBI^+/+ mice were still alive. To determine the cause of death, the moribund TGFBI^−/− mice between ages of 9 and 16 months were sacrificed and subjected to detailed histopathologic analysis. Four of 12 mice developed malignancies, including one invasive lung adenocarcinoma and three lymphomas, one of which was a highly disseminated lymphoma infiltrating liver and lung tissues (Fig. 2A). Others died of unidentified causes with no detectable tumors. Survival of heterozygotes was similar as TGFBI^+/+ mice, and only one died at the end of 16 months without detectable tumor burden. By the end of 20 months, 8.3% (4 of 48, lung adenocarcinoma and lymphoma) of TGFBI^+/+ mice, 13.3% (10 of 75, uterus histiocytic sarcoma, liver tumor, lymphoma, and lung adenocarcinoma) of heterozygotes, and 37.0% (20 of 54) of TGFBI^−/− mice had developed tumors (P < 0.01 for TGFBI^−/− versus heterozygotes and TGFBI^+/+ mice, χ² test; Fig. 2B; Supplementary Table S1). The tumor incidence in heterozygotes is higher than in WT mice but did not reach statistical significance (P > 0.05, χ² test). Southern blot–based genotyping analysis showed that the second WT allele of TGFBI gene was retained in all the 10 tumors derived from heterozygous mice. However, 3 of 10 tumors displayed a dense methylation pattern in the TGFBI promoter identified by bisulfite sequencing (Supplementary Fig. S3). Tumor-free survival in TGFBI^−/− mice was significantly lower than in heterozygote and TGFBI^+/+ mice (P < 0.01, log-rank test; Fig. 2C).

An increased skin tumor induction in mice with TGFBI deficiency. In skin carcinogenesis assays, we treated TGFBI^+/+, TGFBI^+/−, and TGFBI^−/− mice with a single dose of DMBA, a chemical carcinogen, on the dorsal skin 3 to 5 days after birth. The treated mice were checked weekly and monitored for up to 6 months. Ten of 23 TGFBI^−/− mice developed skin tumors by 2.5- to 6-month age, 2 of which formed skin tumors at multiple sites. In contrast, only 2 of 21 TGFBI^+/− mice and 1 of 25 TGFBI^+/+ mice developed tumors during 6-month observation period. Skin tumor incidence in TGFBI^−/− mice was significantly higher (P < 0.01, χ² test) than in TGFBI^+/+ and heterozygous mice (Fig. 2D). Therefore, mice with TGFBI deficiency are prone to the development of both spontaneous malignancies and DMBA-induced skin tumors.

Early passage (P2) of TGFBI^−/− MEFs exhibits an increased frequency of chromosomal aberrations. To clarify whether disruption of TGFBI resulted in an increased frequency of chromosomal aberrations, TGFBI^−/− and WT MEFs at passage 2 were treated with 0.05 μg/mL colcemid for 3 to 6 hours. Chromosomal metaphases were prepared from the treated cells, hybridized with Cy3-conjugated (C3TA2)₃ peptide nucleic acid probe (Applied Biosystems), and counterstained with 4′,6-diamidino-2-phenylindole (DAPI) solution followed by the previously reported procedures (22). Digital images were recorded using Zeiss Axioplan 2 microscope with a multicolor image analysis system (Fig. 3A). Various types of chromosomal aberrations in TGFBI^−/− MEFs were shown in Fig. 3B (arrows). Overall, 43.75% (7 of 17) of metaphases prepared from TGFBI^−/− MEFs contained chromatid breaks, centric fragments, or chromosomal breaks, whereas only 13.3% (2 of 15) metaphases from WT MEFs contained only centric fragments (Fig. 3D).

In addition, frequency of micronuclei was also examined in the early passage of MEFs (P2). Twenty-four hours after plating, cells were fixed with acetone/methanol (1:1) for 10 minutes and stained with 0.03 mg/mL acridine orange in the dark for 10 minutes. A total of 3,000 cells were counted for each experiment and three independent assays were performed. Numbers of micronuclei were recorded in each cell type using Nikon fluorescence microscope. Figure 3C showed multiple micronuclei found in TGFBI^−/− MEFs (arrows). Micronuclei frequency in TGFBI^−/− MEFs was 4.7-fold higher than that in WT MEFs, with 0.128 and 0.027 micronuclei per cell, respectively (Fig. 3D).

An accelerated G₁-S progression and cyclin D1 up-regulation in TGFBI^−/− MEFs (passage 18). To investigate molecular mechanism(s) of tumorigenesis, we characterized MEFs derived from TGFBI^−/− and TGFBI^+/+ littermates. Long-term in vitro growth of MEFs was assayed by a 3T3 protocol. TGFBI^−/− MEFs showed a higher growth at early passage (P2) but grew significantly faster than TGFBI^+/+ MEFs after overcoming the senescence (Fig. 4A). This prompted a comparison of the kinetics of S-phase entry in serum-stimulated quiescent cells. Using BrdUrd incorporation assay, quiescent TGFBI^−/− MEFs were consistently found to enter into S phase in advance of TGFBI^+/+ MEFs on serum stimulation (Fig. 4B).

We then examined the expression patterns of proteins related to G₁-S progression. Cyclin D1 was identified to be significantly up-regulated in TGFBI^−/− MEFs (Supplementary Fig. S4). Moreover, cyclin D1 induction was substantially higher in quiescent TGFBI^−/− MEFs at 4 hours after serum stimulation relative to TGFBI^+/+ cells (Fig. 4C). However, TGFBI reconstitution in TGFBI^−/− MEFs by infection with retroviral vector containing WT TGFBI gene resulted in a marked suppression of cyclin D1 expression (Fig. 4D) and subsequent inhibition of cell growth measured by BrdUrd incorporation (Supplementary Fig. S4).

Aberrant activation of CREB and an increased binding activity of p-CREB to cyclin D1 promoter in the absence of TGFBI in TGFBI^−/− MEFs (passage 18). A transcriptional mechanism seems to be involved in cyclin D1 up-regulation because TGFBI^−/− MEFs showed a 5.6-fold higher level of cyclin D1 mRNA than WT cells (Fig. 5A). Cyclin D1 promoter region contains several established or potential binding sites for the transcriptional factors (23). Thus, induction of transcription factors was examined by Western blots in quiescent WT and TGFBI^−/− MEFs in response to serum stimulation. Only CREB was identified to be aberrantly activated in TGFBI^−/− cells (Fig. 5B), which is further substantiated by a luciferase reporter assay showing that relative pCRE promoter activity in TGFBI^−/− cells was over 15-fold higher than WT cells; however, it can be suppressed significantly by a dominant-negative CREB (DN-CREB) vector (Fig. 5C and D).

To determine whether aberrant activation of CREB in TGFBI^−/− cells results in an enhanced binding activity of p-CREB to cyclin D1 promoter, a quantitative PCR-based ChIP assay was used to quantify the bindings of CREB and p-CREB to cyclin D1 promoter in both WT and TGFBI^−/− cells. As shown in Fig. 6A, ratio of p-CREB/CREB binding to cyclin D1 promoter is over 5-fold higher in TGFBI^−/− cells than in WT cells, suggesting that binding activity of p-CREB to cyclin D1 is significantly increased in TGFBI^−/− cells.

TGFBI deficiency correlates with CREB activation and leads to cyclin D1 up-regulation. To determine whether CREB activation is responsible for cyclin D1 up-regulation in TGFBI^−/− cells (passage 18), a luciferase assay was used to examine the specific suppression of cyclin D1 promoter activity by DN-CREB. The cyclin D1 promoter activity in TGFBI^−/− cells was over 10-fold higher than in WT cells (data not shown); however, it could be inhibited to the similar level of WT cells by DN-CREB (Fig. 6B). These data clearly suggested that CREB activation is involved in cyclin D1 up-regulation in TGFBI-null cells.

Correlation between TGFBI deficiency and CREB activation was further established by reconstitution of TGFBI expression in TGFBI^−/− cells. Compared with WT cells, TGFBI^−/− cells showed a significantly higher level of p-CREB and cyclin D1, whereas it could be suppressed to the level of WT cells after supplement with recombinant mouse TGFBI protein in the culture medium at 0.5 μg/mL for 24 hours (Fig. 6C).

Cyclin D1 up-regulation in tumors arising from TGFBI^−/− mice. To define the potential significance of cyclin D1 up-regulation in in vivo tumor development, cyclin D1 protein level was examined by Western blotting in tumor tissues arising from TGFBI^−/− mice. A markedly increased level of cyclin D1 protein was shown in most (10 of 13) of tumor samples examined in relation to their matched WT controls (Fig. 6D).

TGFBI gene has been regionally mapped to chromosome 5q31, a locus often deleted in leukemias, myelodysplastic syndromes, and many human cancers, such as renal cell, esophageal, and lung carcinomas (24–26). In addition, lost expression or down-regulation of this gene has been found in a list of human tumor cell lines as well as in primary human lung and breast cancer specimens (11–14). Specifically, TGFBI down-regulation has been causally linked to an enhanced tumorigenicity and tumor angiogenesis by in vitro studies (12, 13, 18, 19). However, it is not clear whether TGFBI possesses antitumor function in vivo. To this aim, TGFBI-null mice have been generated. The results showed that mice with TGFBI disruption are prone to spontaneous tumors as well as DMBA-induced skin tumors. The most common tumors arising spontaneously in TGFBI^−/− mice were lymphomas (65%, 13 of 20). Other tumor types were bronchial adenocarcinoma, skin invasive adenocarcinoma, liver histiocytic sarcoma, and testis hemangioendothelioma, which are less common in aged 129Sv/Ev, C56BL/6J mice. Compared with the tumors occurred in TGFBI^+/+ and heterozygous mice, spontaneous tumors arisen from TGFBI^−/− mice showed metastatic potential: (a) 7 of 13 lymphomas were classified as disseminated types with infiltration to other nonlymphoid organs, including liver, lung, kidney, and pancreas; (b) 4 of 6 other types of tumors were either invasive or metastatic malignancies; and (c) one TGFBI^−/− mouse developed both invasive lung adenocarcinoma and lymphoproliferative disorder. Hence, TGFBI disruption resulted in a dramatic predisposition to lymphomas and other cancers.

Cell cycle progression through the G₁ phase requires dual signaling from soluble growth factors and adhesion to the ECM (27). Both receptor tyrosine kinase–dependent and integrin-dependent proliferations are regulated through three major G₁-phase targets, including induction of cyclin D1 and down-regulation of the cyclin-dependent kinase inhibitors p21cip1 and p27kip1, which lead to efficient phosphorylation of the retinoblastoma protein and progression to S phase (27–29). Under mitogenic conditions, adhesion promotes G₁-phase progression primarily via up-regulation of cyclin D1 (30). Because TGFBI is an adhesion protein associated with integrin receptor through its RGD motif (8–10), it is expected that an accelerated G₁-S transition in TGFBI^−/− cells is due to the dysregulation of three major G₁-phase targets. This is supported by our findings showing that cyclin D1 is overexpressed in TGFBI^−/− cells. In contrast, TGFBI reconstitution results in a substantially decreased level of cyclin D1 expression and cellular proliferation. Furthermore, cyclin D1 protein level was shown to be elevated in most of tumors isolated from TGFBI^−/− mice, suggesting a critical role of cyclin D1 up-regulation in in vivo tumor progression in the absence of TGFBI protein. It is well documented that cyclin D1 plays a major role in controlling G₁-S progression and is consistently up-regulated in most human cancers (28, 31). Deregulated cell proliferation and increased frequency of spontaneous tumors has been found in transgenic mice with overexpression of cyclin D1, whereas deletion of cyclin D1 protects the mice from tumor induction (32–34). Collectively, these observations together with our findings suggest a critical role of cyclin D1 up-regulation in the enhanced cell proliferation and tumor formation in TGFBI^−/− mice. However, cyclin D2, cyclin D3, cyclin A, and p21CIP1 expression was not strongly affected by loss of TGFBI, although p21CIP1 level seemed lower in TGFBI^−/− MEFs.

CREB has been shown to act as an oncogene and implicated in the development of human endocrine tumors and acute myeloid leukemia (35–37). In this study, CREB was identified to be aberrantly activated after TGFBI disruption. In addition, causal links between TGFBI deficiency and CREB activation, and cyclin D1 up-regulation have been established in our model system, suggesting that signaling pathway from TGFBI to CREB/cyclin D1 is dysregulated after TGFBI disruption and is involved in tumor progression in TGFBI^−/− mice. The exact mechanisms of CREB/cyclin D1 activation downstream of TGFBI remain unknown but may include activation of one or more kinases through an integrin-dependent pathway, including FAK, AKT, protein kinase A, protein kinase C, Ca²⁺/calmodulin-dependent protein kinase, glycogen synthase kinase III, and casein kinase II, which have been shown to regulate CREB phosphorylation (35, 38). We have found that phospho-AKT was substantially elevated in serum-starved early passage (P2) of TGFBI^−/− MEFs at 15 and 30 minutes after serum stimulation when compared with WT cells (Supplementary Fig. S5). Previous studies have shown that CREB can be activated by protein kinase B/AKT (38, 39); therefore, AKT pathway might mediate TGFBI-regulated CREB/cyclin D1 activation in TGFBI^−/− cells.

It is commonly accepted that malignant transformation is a lengthy multistep process and arises through an accumulation of mutations at various genetic loci (40). Genomic instability has been shown to not only initiate tumorigenesis but is at least a factor in tumor progression (40, 41). In the present study, early passage of TGFBI^−/− MEFs showed a significantly increased chromosomal aberration and micronuclei than WT MEFs, suggesting that TGFBI deficiency induces genetic instability. This is supported by other study showing that TGFBI protein is involved in microtubule stability and silencing of its expression contributes to centrosome amplification and enhanced mitotic abnormalities (17). It is well documented that chromosomal instability, which is equated to mitotic defects and consequential chromosome segregation errors, provides a formidable basis for the acquisition of further malignant phenotype during tumor progression (42). It should be noted that 3 of 13 tumor samples isolated from TGFBI^−/− mice did not show cyclin D1 up-regulation. Thus, other signaling pathway(s) instead of CREB/cyclin D1 activation might be aberrantly regulated due to genomic instability.

Although loss of TGFBI expression has been found in different types of primary human cancers (11–14), several studies show that TGFBI is commonly overexpressed in colorectal, renal, and pancreas cancers (43–45). In addition, overexpression of TGFBI promotes metastasis of SW480 colon cancer cells by enhancing extravasation (45). However, we have found a significant down-regulation of TGFBI in HT29 colon cancer cells (12). These data clearly point out that dysregulation of TGFBI expression is tumor cell or type specific. It is highly possible that TGFBI is a double-edged sword whose loss or gain of expression leads to tumorigenesis. Previous studies have shown that TGFBI mutations correlate with human corneal dystrophy (46). However, overexpression of mutant TGFBI induces retinal degeneration, but no corneal phenotype was observed in transgenic mice (47). Similarly, we did not identify any eye phenotypes in TGFBI^−/− mice.

The present studies provide the evidence, for the first time, that loss of TGFBI functions as a tumor suppressor in vivo. Because of frequent loss of TGFBI protein in human cancer cells (11–14), TGFBI and its associated signaling represent the promising targets for anticancer drug discovery.

No potential conflicts of interest were disclosed.

Grant support: NASA NAG2-1637 (Y. Zhao), CA127120 (Y. Zhao), and NIH ES-11804 (T.K. Hei).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Drs. Ze'ev A. Ronai, Howard B. Lieberman, and Dangsheng Li for critical reading; Dr. M.R. Montminy (San Diego, CA) for kindly providing WT CREB and DN-CREB expression plasmids; and Dr. Isabella Screpanti (University La Sapienza, Rome, Italy) for mouse cyclin D1 luciferase reporter plasmid.