Transforming Growth Factor-β Receptor Type I Gene Is Frequently Mutated in Ovarian Carcinomas¹

OC³ is a very aggressive type of cancer among women. Each year, 23,400 women in the United States will be diagnosed with OC, and more than half of them will die from the disease (1). OCs, particularly recurrent OCs, are frequently resistant to growth inhibition by TGF-β (2, 3, 4), suggesting that alterations in the TGF-β response are involved with OC progression (3).

The TGF-β signal is transduced by two transmembrane serine-threonine kinase receptors. TGF-β binds first to TβR-II homodimers, which then form heterotetramers with two TβR-I molecules. As a consequence, the TβR-II kinase phosphorylates and activates TβR-I (5). Activated TβR-I then phosphorylates Smad2 and Smad3, which in turn heterotrimerize with Smad4. These Smad complexes then translocate to the nucleus, bind to DNA in a sequence-specific manner, and regulate gene transcription, leading to cell cycle arrest (5, 6, 7, 8).

Alterations in these key effectors of the TGF-β signaling pathway have been identified in numerous human tumors and account for the loss of TGF-β responsiveness (9, 10, 11, 12, 13). For instance, TβR-II mutations are frequent in hereditary nonpolyposis colorectal cancer, and homozygous deletion of the Smad4 gene is frequent in pancreatic cancer (11, 13). In ovarian cancers, TGF-β resistance is frequent, perhaps exceeding 75% of the cases (2, 4). Code-altering mutations in TβR-II have been reported in a minority of human OCs (14) suggesting that, in ovarian carcinogenesis, alterations in other components of TGF-β signaling may play a role in the loss of TGF-β responsiveness. We have shown recently that TβR-I is frequently inactivated by mutation in metastatic breast cancers and head and neck cancers (15, 16). We therefore sought to explore whether TβR-I mutations may be involved in ovarian carcinogenesis. Our data indicate that 33% of OCs show code-altering, somatic mutations. Moreover, our data also indicate that 23% of OCs harbor a 9-bp deletion of exon 1 that has been defined previously as a germ-line deletion (17, 18). Together, these data suggest that alterations in TβR-I are frequent in OCs and may be intimately involved in the frequent loss of TGF-β responsiveness that typifies OC.

Paraffin-embedded OC tissues are archived at Wood Hudson Cancer Research Laboratory. The tissues were obtained at surgery at St. Elizabeth Medical Center (Covington/Edgewood, KY). Institutional Review Board approval for this study was received from St. Elizabeth Medical Center. A board-certified pathologist (J. P.) reviewed a H&E-stained section from each block of ovarian tissue and graded the tumors according to WHO classification and Nomenclature of Ovarian Neoplasms. We analyzed 30 cases with a diagnosis of OC including tumors of low malignant potential to grade IV. The average age of patients at diagnosis was 62.2 years.

The basic technical approach has been described previously (19). Briefly, a single 8–10-μm-thick paraffin section was stained and manually microdissected for each tumor sample. Using companion H&E-stained slides as a reference, tumor cells were microdissected with a fine-point surgical blade (no. 11) under an inverted microscope. Adjacent normal tissue of the same individual was dissected the same way to determine the somatic or germ-line nature of genetic changes.

Microdissected samples were deparaffinized (three washes with xylene for 30 min each) and rehydrated in decreasing concentration of alcohol (19). DNA was extracted with Instagene Chelex matrix solution according to the manufacturer’s instructions (Bio-Rad, Hercules, CA), containing 60 μg of proteinase K in a shaking incubator at 37°C overnight. Samples were boiled for 10 min, vortexed, and centrifuged at 7000 × g for 5 min. Approximately 2–8 μl of supernatant were used for PCR amplification.

A sensitive “cold” PCR-SSCP was used to screen for mutations (20). Primers used for PCR amplification of the 9 exons of the TβR-I gene were targeted to intronic sequences to amplify the complete coding sequence of each exon. Primer sequences were described previously (15). The PCR reaction was performed in a volume of 20 μl containing 500 nm unlabeled primers and 2 units of AmpliTaq Gold DNA polymerase (PE Applied Biosystems, Branchburg, NJ). After an initial 10-min denaturation at 95°C, PCR was run for 50 cycles of 95°C for 1 min, 55°C for 40 s, and 72°C for 1 min, followed by a 5-min final extension at 72°C. For PCR amplification of the GC-rich exon 1 of TβR-I, we added 2 m betaine in the PCR mixture. For SSCP analysis, a 5-μl aliquot of amplified PCR product was mixed with 15 μl of loading buffer (12.5 μl of 10× TBE buffer, 2 μl of 15% Ficoll, 0.1% bromphenol blue and xylene cyanol, and 0.5 μl of methyl mercury hydroxide), denatured by heating at 75°C for 3 min, and quenched on ice. The single-stranded DNA fragments were then resolved using precast 20% TBE acrylamide gels on a Novex X-cell II Thermoflow apparatus (Novex, San Diego, CA), with the gel temperature maintained precisely at 10°C throughout the run. Bands were visualized by staining the gel in a 1:10,000 dilution of SYBR Green II (Molecular Probes, Inc., Eugene, OR) for 20–30 min, and photographic documentation was carried out by using a camera coupled with a SYBR Green band pass filter.

Suspect bands detected in cold SSCP were first verified by an additional, independent PCR-SSCP analysis. DNA from adjacent normal tissue was included in this process. The confirmed shifts in the SSCP gel were excised for sequencing. After reamplification of the shifted bands, the PCR products were purified using the QIAquick PCR purification kit (Qiagen, Chatsworth, CA). Finally, purified DNA fragments were directly sequenced by the same forward or reverse primers used in the original PCR amplification. The sequencing was done in an automated sequencer (PE Applied Biosystems Model 377) at the DNA Core Facility of the University of Cincinnati (Cincinnati, OH).

To assess the incidence of somatic mutations in ovarian carcinoma, we have microdissected tumor from normal cells and screened all nine exons of the TβR-I gene in a total of 30 OC cases. Ten tumor cases showed confirmed bandshifts that were different from the wild-type bands among exons 2, 3, 4, and 6 of the TβR-I gene (Fig. 1). Most of the shifts were exclusive to tumor tissues and are therefore indicative of somatic mutations. These included two shifts in exon 2, two shifts in exon 3, five shifts in exon 4, and three shifts in exon 6 (Table 1). One case (OC18) showed a strong shift in the tumor and a weak shift in the adjacent normal stroma tissues but not in different non-tumor blocks of the same patient (Fig. 1). Sequencing of these shifted bands revealed that each carries a point mutation, most of which were missense mutations (Fig. 2,A). There was only one silent mutation of exon 4 in case OC6 among all genetic changes found in this study (Table 1). However, this patient harbored another missense mutation in exon 3. Mutations were not found in exons 5, 7, 8, or 9 of the TβR-I gene. Table 1 summarizes the clinical information for all patients studied and the mutations, including their locations at different exons and amino acid substitutions. There was no apparent relationship between OC subtype and TβR-I mutations (Table 1).

Mutations of exons 4 and 6 might affect the receptor kinase domain (Fig. 2,B). Three cases (OC4, OC21, and OC23) had G-to-A transition mutations in exon 4 that resulted in substitutions of serine-to-asparagine at codon 210 and arginine-to-glutamine at codon 225. Case 11 had a missense mutation with substitution of valine-to-methionine at codon 206. The mutations of exon 4 are located in the phosphate-binding loop adjacent to the GS (glycine-serine rich) domain that is critical for activation of TβR-I kinase activity. Case 6 has a silent mutation at codon 203 in exon 4. In Fig. 2 B, we show that mutations of S210N, R225Q, and V206M are located in the ATP binding loop of the receptor kinase domain (21).

In exon 6, three cases showed the same mutation, a C-to-A transversion mutation resulting in proline-to-glutamine substitution at codon 327 (P327Q). Codon 327 is highly conserved in all type I receptor kinases (TβR-I, ActR-IB, ActRI, BMPRI-B, tkv, and TβR-II), which suggests that mutations of this codon (P327Q) may have functional consequences (21). Moreover, this mutation sits right in the E6 loop that is critical for receptor dimerization, as described in studies of the crystal structure of TβR-I (21).

Additional somatic mutations in the transmembrane and extracellular domains were also found in this study. Case 10 had a mutation in exon 2 that results in a threonine-to-isoleucine substitution (T98I) located in the extracellular domain. Case 6 had a mutation inside the transmembrane domain that resulted in a valine-to-isoleucine substitution (V145I). Case 21 had a mutation in exon 3, substituting histidine for tyrosine (H149Y), that might affect the cytoplasmic juxtamembrane region.

A 9-bp, germ-line deletion (heterozygous allelic variant) in exon 1 of TβR-I [del(GGC)₃] was reported in our previous studies of cervical cancer (17, 18) and in a large study of the normal population (22). To evaluate the prevalence of this allelic variant form in OCs, we amplified exon 1 of the TβR-I gene in all 30 cases of OC. Because exon 1 is located in a highly GC-rich region and is relatively difficult to amplify, we optimized the PCR amplification by including 2 m betaine in the PCR mixture. By using sensitive “cold” SSCP methods, we were able to successfully identify the deleted allele from the wild-type allele (Fig. 3). We found that 7 of 30 (23.3%) ovarian cancer cases showed this allelic variant by SSCP. Each of the cases with del(GGC)₃ allele also retained the wild-type allele (Fig. 3). Furthermore, using tissue microdissection, we identified this deletion in both the tumor and the adjacent stroma. In fact, the deleted allele and the wild-type allele were evident at roughly the same intensity in both tumor and stroma, suggesting that this deletion represents the germ-line deletion that had been identified previously (18). DNA sequencing confirmed the presence of both the wild-type and variant del(GGC)₃ alleles. This deletion would eliminate three alanines in the signal peptide of the receptor. Interestingly, the frequency of the del(GGC)₃ allele in this study of OCs was 23.3%, more than twice that reported in the normal population (18, 22).

The loss of TGF-β responsiveness is a very common event in OCs, particularly recurrent OCs, occurring in >75% of ovarian cancers. As such, diminished TGF-β responsiveness may be a key event in the genesis and/or progression of ovarian cancer. Defined alterations in TβR-II have been identified in only 25% of OCs (14), suggesting that alterations in other key members of the TGF-β pathway may be involved in OC. We have recently identified frequent alterations in TβR-I in metastatic breast cancer and cervical carcinomas (15, 18). We therefore sought to examine the possibility that TβR-I alterations may be involved in OCs. Our data demonstrate that code-altering mutations and/or a germ-line deletion in TβR-I are prevalent in primary human OCs occurring in 47% of the cases analyzed.

The crystal structure of the cytosolic portion of TβR-I protein has been determined recently. Functionally critical segments include the GS loop, the phosphate-binding loop, and the catalytic segment (21). In this report, we identified five missense mutations in exon 4 and three mutations in exon 6 of the TβR-I gene that alter critical codons within the TβR-I kinase domain and may disrupt receptor kinase activity. The mutations in exon 4 are located in the phosphate-binding loop adjacent to the GS domain and may have detrimental effects on activation of receptor kinase or ATP binding. The proline-to-glutamine substitution (P327Q) in exon 6 occurred in three cases. This residue is highly conserved in all type I receptor kinases and sits within the E6 loop that is critical for receptor dimerization (21). Mutations of this codon may therefore affect receptor dimerization. We have also identified mutations in exons 2 and 3, which are located in the extracellular domain and transmembrane domain, respectively. These mutations may affect ligand binding capability during formation of the ligand/receptor complex. Importantly, the mutations identified in this study were not the same as those identified in our previous studies of breast and head and neck carcinomas, suggesting that tissue-specific factors might influence the accumulation of genetic changes in different cancer types.

Recently, Wang et al. (23) also reported somatic mutations in TβR-I in 31% of primary ovarian carcinomas. However, these were defined as frameshift mutations exclusively in exon 5 of TβR-I. In this report, we did not find any mutations in exon 5. It seems plausible that these differences may reflect differences in the patient populations examined. Nevertheless, both studies indicate that somatic alterations in TβR-I are frequent in human OC, further strengthening the notion that these alterations may be related to the loss of TGF-β responsiveness.

In addition to the somatic mutations, we have also shown a high frequency of allelic variant in exon 1 [germ-line del(GGC)₃]. This deletion has been defined previously as a germ-line mutation in a large study of the normal population and in our study of cervical cancer (18, 22). In this study, 7 of 30 (23.3%) OC cases showed the allelic variant of exon 1. This deletion was evident in the tumor and adjacent stroma, consistent with a germ-line nature. Moreover, the frequency of this deletion is more than twice that found in the normal population (10.6%) in a recent study (22). In the clinical follow-up analysis, two of the patients (OC1 and OC15) with the deletion are among the younger ages in the group analyzed (Table 1). The average ages of the patients with del(GGC)₃ variant was 57.1 but was 63.8 for the rest of patients. Interestingly, six of seven cases with this allelic variant had poorly differentiated or undifferentiated carcinomas (grade 3 or 4; see Table 1). Together these data suggest that this deletion may contribute to, or predispose individuals to, OC. However, a large case-control study will be needed to further explore the association of this allelic variant with clinical outcome.

In each of the cases showing genetic mutations of TβR-I, both the mutant and wild-type alleles are retained (Figs. 1 and 3). It is plausible that haploinsufficiency for a fully functional TβR-1 may contribute to OC. Alternatively, these mutant alleles may act as dominant negatives, disrupting the function of the heterotetrameric TβR-II and TβR-I complexes. It is also possible that expression of the wild-type allele is eliminated by methylation or epigenetic inactivation (24). Additional studies will help dissect the precise role for these TβR-I mutations in the development and progression of ovarian carcinogenesis.

In summary, this report demonstrates that code-altering mutations of TβR-I, as well as a high frequency of a germ-line deletion, are frequent in human OC. The majority of somatic mutations are clustered in critically functional domains of the receptor kinase, suggesting that selection of the genetic changes might be functionally beneficial in OC development. Together, these data indicate that the loss of TGF-β responsiveness that typifies ovarian carcinogenesis and progression may frequently involve alterations in the TβR-I gene.

We thank Dr. Eric Hugo for computational assistance and many helpful suggestions for the project.