Reduced expression of transforming growth factor beta type I receptor contributes to the malignancy of human colon carcinoma cells.

Transforming growth factor β (TGFβ) type I (RI) and type II (RII) receptors are essential for TGFβ signal transduction. A human colon carcinoma cell line, designated GEO, is marginally responsive to TGFβ and expresses a low level of RI mRNA relative to colon carcinoma cells, which are highly responsive to TGFβ. Hence, the role of RI as a limiting factor for TGFβ sensitivity and the contribution of low RI levels to the malignant phenotype of GEO cells were examined. Stable transfection of a tetracycline-regulatable rat RI cDNA increased TGFβ1 binding to RI and resulted in increased growth inhibition by exogenous TGFβ1. In contrast, although stable transfection of an RII expression vector into the same GEO cells increased TGFβ1 binding to RII, growth inhibition by exogenous TGFβ1 was not altered. This indicated that the low level of RI is a limiting factor for the growth-inhibitory effects of TGFβ in GEO cells. RI-transfected cells were growth-arrested at a lower saturation density than GEO control cells. They also showed reduced growth and clonogenicity in plating efficiency and soft agarose assays, whereas RII-transfected cells did not show any differences from the NEO control cells in these assays. Tetracycline repressed RI expression in transfected cells and reversed the reduction in plating efficiency of RI-transfected clones, confirming that growth effects were due to increased RI expression in transfected cells. TGFβ1 neutralizing antibody stimulated the proliferation of RI-transfected cells but had little effect on GEO control cells, indicating that increased autocrine-negative TGFβ activity also resulted from increased RI expression. Tumorigenicity in athymic nude mice was significantly delayed in RI-transfected cells. These results indicate that low RI expression can be a limiting factor for response to exogenous TGFβ, as well as TGFβ autocrine-negative activity, and that reduction of RI expression can contribute to malignant progression.

Transforming growth factor ␤ (TGF␤) type I (RI) and type II (RII) receptors are essential for TGF␤ signal transduction. A human colon carcinoma cell line, designated GEO, is marginally responsive to TGF␤ and expresses a low level of RI mRNA relative to colon carcinoma cells, which are highly responsive to TGF␤. Hence, the role of RI as a limiting factor for TGF␤ sensitivity and the contribution of low RI levels to the malignant phenotype of GEO cells were examined. Stable transfection of a tetracycline-regulatable rat RI cDNA increased TGF␤ 1 binding to RI and resulted in increased growth inhibition by exogenous TGF␤ 1 . In contrast, although stable transfection of an RII expression vector into the same GEO cells increased TGF␤ 1 binding to RII, growth inhibition by exogenous TGF␤ 1 was not altered. This indicated that the low level of RI is a limiting factor for the growth-inhibitory effects of TGF␤ in GEO cells. RItransfected cells were growth-arrested at a lower saturation density than GEO control cells. They also showed reduced growth and clonogenicity in plating efficiency and soft agarose assays, whereas RII-transfected cells did not show any differences from the NEO control cells in these assays. Tetracycline repressed RI expression in transfected cells and reversed the reduction in plating efficiency of RI-transfected clones, confirming that growth effects were due to increased RI expression in transfected cells. TGF␤ 1 neutralizing antibody stimulated the proliferation of RI-transfected cells but had little effect on GEO control cells, indicating that increased autocrine-negative TGF␤ activity also resulted from increased RI expression. Tumorigenicity in athymic nude mice was significantly delayed in RI-transfected cells. These results indicate that low RI expression can be a limiting factor for response to exogenous TGF␤, as well as TGF␤ autocrine-negative activity, and that reduction of RI expression can contribute to malignant progression.
Transforming growth factor ␤s (TGF␤s 1 ) are a family of multifunctional cytokines that regulate many aspects of cellular function including proliferation, differentiation, adhesion, and migration (1)(2)(3). The role of TGF␤ in growth regulation is of particular interest in cancer. TGF␤ has been shown to be an autocrine-negative growth factor, as evidenced by stimulation of growth of several cell lines treated with TGF␤-neutralizing antibody (4 -8). Loss of autocrine TGF␤ activity and/or responsiveness to exogenous TGF␤ appears to provide cells with a growth advantage leading to malignant progression. For example, loss or reduction of TGF␤ response has been shown to be associated with tumor development and progression in a number of cancer cell lines as well as the conversion of colonic adenomas to carcinomas (9 -12). Previous work in our laboratory showed that suppression of autocrine TGF␤ activity by constitutively repressing endogenous TGF␤ expression without eliminating the ability to respond to exogenous TGF␤ led to a more progressed phenotype in colon cancer cells (6,7). This suggested that loss of autocrine TGF␤ function was a key feature in the development of malignant properties of these colon carcinoma cells. Restoration of TGF␤ responsiveness and/or autocrine TGF␤ activity by reconstitution of the autocrine TGF␤ loop led to reduced malignancy in several carcinoma cell lines (8,10,13).
TGF␤s exert their effects through binding to specific cell surface proteins. Three major types of TGF␤ receptors, type I (RI), type II (RII), and type III (RIII), have been identified in most cells by receptor-affinity labeling assays (1,2). RI and RII are glycoproteins of M r 53,000 and M r 75,000, respectively, whereas RIII is a proteoglycan of M r 280,000 -330,000. TGF␤ signals through a heteromeric complex of RI and RII. Both RI and RII belong to the transmembrane serine/threonine kinase receptor family (14 -17). It has been shown that RII is required for the binding of TGF␤ to RI, whereas RI is required for signal transduction (15,18). The ability of the two receptors to undergo heteromeric interactions and the necessity of phosphorylation of RI by RII kinase to confer signal transduction further reinforces the evidence that both receptors are required for TGF␤ response (18,19). The direct involvement of RI and RII in TGF␤ signal transduction would suggest that loss of functional RI and/or RII expression could contribute to loss of TGF␤ responsiveness, resulting in tumor progression. The lack of response to TGF␤ in some types of cancer and tumor cell lines has been reported to be associated with the loss of RII. Reexpression of RII has restored TGF␤ sensitivity and generated reversal of malignant properties in some of these systems (8, 10, 13, 20 -23). Since RI is as necessary as RII for TGF␤ signal transduction, it would seem likely that disruption of this TGF␤ receptor could also contribute to loss of TGF␤ response and tumor-suppressive activity. Resistance to TGF␤ resulting from RI loss has been shown in mutagenized mink lung cells, and re-expression of this receptor led to regeneration of TGF␤ response (17, 24 -26). However, loss or reduced expression of RI has not been directly demonstrated to have a role in determining malignant properties.
We identified a group of colon carcinoma cell lines that showed reduced RI expression (21). One of these cell lines, designated GEO, has been characterized as being relatively insensitive to TGF␤ inhibition (9). Here we report that reconstitution of RI expression in GEO cells increased TGF␤ growthinhibitory effects and autocrine TGF␤ activity with a concomitant reduction of tumorigenicity in athymic mice. In contrast, increased expression of RII in GEO cells did not affect TGF␤ sensitivity or other GEO biological properties, thus indicating that RI is the limiting factor for TGF␤ signal transduction in these cells.

MATERIALS AND METHODS
Cell Culture-The GEO human colon carcinoma cell line was established in vitro from a primary tumor, as described previously (27). Cells were maintained at 37°C in a humidified atmosphere of 5% CO 2 in McCoy's 5A serum-free medium (Sigma) supplemented with pyruvate, vitamins, amino acids, antibiotics, 10 ng/ml epidermal growth factor, 20 g/ml insulin, and 4 g/ml transferrin (28).
Rat RI Expression Vector Construction and Transfection-The rat TGF␤ RI cDNA (ϳ3 kilobases; Ref. 29) was subcloned into a tetracycline-controllable expression system kindly provided by Dr. H. Bujard at University of Heidelberg, Heidelberg, Germany (30). As described previously (13), this system includes a tTA, which is generated by fusing the tet repressor with the activating domain of virion protein 16 of herpes simplex virus, and an expression vector that consists of a tet operator sequence and a minimal promoter sequence derived from the human cytomegalovirus promoter. When the tetracycline repressor of the tTA binds to tet operators, the virion protein 16 domain of the tTA can activate the minimal promoter to start transcription. Tetracycline can block this activation by preventing the tTA from binding to the tet operator sequence. A neomycin-resistant gene under control of the mouse ␤-globin promoter was subcloned into the tTA-containing plasmid. This plasmid (2 g) and the rat RI expression vector (20 g) were linearized and transfected into GEO cells. Electroporation was carried out at 250 V, 960 microfarads with a Gene Pulser (Bio-Rad). The control cells were similarly transfected with 2 g of NEO-containing plasmid and 20 g of the expression vector without rat RI cDNA. The transfected cells were allowed to grow for 2 days before being subjected to selection with 600 g/ml geneticin (G418 sulfate; Life Technologies, Inc.). Stable cell clones resistant to G418 sulfate were ring cloned after 2-3 weeks and expanded for screening for rat RI expression. The control clones were pooled, expanded, and designated as GEO NEO pool. As a control, TGF␤ RII used in exactly the same regulatable expression system (13) was transfected into the same GEO cells.
RNA Analysis-RNase protection assays were performed as described previously (7,8,13) to screen for RI-and RII-expressing clones. The rat RI riboprobe construct was made by inserting a 1-kilobase fragment of rat RI cDNA into pBSK(Ϫ) plasmid (Stratagene cloning system). After linearizing with PvuII, an antisense riboprobe that protects a 143-base pair fragment of rat RI mRNA in an RNase protection assay was synthesized in vitro using T7 RNA polymerase. The RII riboprobe was prepared as described previously (8,13). The 32 P-labeled probes were hybridized with 40 g of total RNA extracted from GEOtransfected cells. The hybridization mixture was then digested with RNase A and RNase T 1 , followed by proteinase K treatment. The protected fragments were analyzed by 6% urea-polyacrylamide gel electrophoresis and visualized by autoradiography. Actin mRNA protected by an actin riboprobe was used to normalize sample loading, as described previously (7).
To study the effect of exogenous TGF␤ 1 on fibronectin and integrin ␣5 subunit expression, exponentially growing NEO pool and RI clones were treated with various concentrations of TGF␤ 1 for 24 h. Total RNA was then extracted for detection of fibronectin and integrin ␣5 mRNA using an RNase protection assay. The fibronectin antisense riboprobe was synthesized as described previously (8). The integrin ␣5 antisense riboprobe plasmid was constructed by subcloning a 405-base pair BamHI-AccI fragment of human ␣5 subunit cDNA into pBSK(Ϫ). The probe was then synthesized in vitro using T3 RNA polymerase. Actin mRNA was used to normalize sample loading.
Receptor Cross-linking-Simian recombinant TGF␤ 1 was purified from conditioned media of transfected Chinese hamster ovary cells, as described by Gentry et al. (31). Purified TGF␤ 1 was iodinated by the chloramine-T method (32) and utilized to visualize TGF␤ receptors after cross-linking with disuccinimidyl suberate, as described previously (8,13).
DNA Synthesis Assay-[ 3 H]Thymidine incorporation was used to determine TGF␤ sensitivity of the control and transfected cells to exogenous TGF␤ treatment (13). The cells were plated in 24-well tissue culture plates at a density of 3.0 ϫ 10 4 cells/well with various concentrations of TGF␤ 1 . Exponential cells were labeled on day 4 with [ 3 H]thymidine (7 Ci; 46 Ci/mmol; Amersham Corp.) for 1 h. DNA was then precipitated with 10% trichloroacetic acid and solubilized in 0.2 M NaOH. The amount of [ 3 H]thymidine incorporated was analyzed by liquid scintillation counting in a Beckman LS7500 scintillation counter.
Plating Efficiency Assay-Plating efficiency assays were performed as described previously (8) to study the effects of increased RI or RII expression on clonogenic potential at a low seeding density. GEO RI and RII transfectants and control cells were plated in 24-well plates at a density of 400 cells/well with McCoy's 5A serum-free medium. TGF␤ 1 neutralizing antibody (R&D Systems) was produced in chickens immunized with purified recombinant human TGF␤ 1 . This antibody neutralizes the biological activity of recombinant human TGF␤ 1 , porcine TGF␤ 1 , and porcine TGF␤ 1 . 2 . TGF␤ 1 is the only isoform expressed by GEO cells. TGF␤ 1 neutralizing antibody was added to the medium at a final concentration of 10 g/ml to determine autocrine TGF␤ activity. Cell colonies were visualized by staining with MTT (Sigma) and then solubilized with dimethyl sulfoxide for optical density measurements.
Soft Agarose Assay-Soft agarose assays were performed as described previously (6,8,13) to compare the clonogenic potential of control cells and RI-and RII-transfected cells in semisolid medium. Briefly, cells were suspended at 3.0 ϫ 10 3 cells/ml in 1 ml of 0.4% SeaPlaque agarose in McCoy's 5A serum-free medium and plated on top of 1 ml of 0.8% agarose in the same medium in 6-well tissue culture plates. Plates were incubated for 2-3 weeks at 37°C with 5% CO 2 in a humidified incubator. Cell colonies were visualized by staining with 0.5 ml of p-iodonitrotetrazolium violet (Sigma).
Tumorigenicity-Tumorigenicity studies were performed as described previously (6,8,13). Briefly, exponentially growing GEO transfectants and control cells were inoculated subcutaneously behind the anterior forelimb of 4-week-old athymic mice (athymic nu/nu-CWRU Cancer Center athymic mouse colony). Mice were maintained in a pathogen-free environment. Growth curves for xenografts were determined by externally measuring tumors in two dimensions. The volume was determined by the following equation  2, 4,  and 6, from the left) and repression of rat RI expression by 0.1 g/ml tetracycline (lanes 3, 5, and 7). B, endogenous and transfected RII mRNA levels in RII clone 37. Actin mRNA levels were used for normalization.

Expression of Transfected RI and RII-To demonstrate that
RI is the limiting receptor in TGF␤ signal transduction in GEO cells, we stably transfected rat RI and human RII cDNA separately into GEO cells that expressed a low level of endogenous RI and normal levels of RII relative to TGF␤-sensitive human colon carcinoma cell lines (data not shown). Several positive clones (designated RI clones 32, 46, and 53 and RII clone 37) with high levels of transfected RI or RII mRNA expression were obtained (Fig. 1). The addition of tetracycline in the culture medium almost completely suppressed the transfected RI mRNA expression (Fig. 1A, from the left, lanes 3, 5, and 7), indicating that the tetracycline-regulatable expression system was functional in GEO cells. Receptor cross-linking assays confirmed that RI clones 32, 46, and 53 expressed higher levels of RI protein, whereas RII clone 37 expressed a higher level of RII protein as compared with the control NEO pool (Fig. 2). The addition of tetracycline in the medium also inhibited RI protein expression (data not shown). The cross-linking assays showed increased TGF␤ binding to RI but not to RII in RI-transfected cells and increased TGF␤ binding to RII but not to RI in RII-transfected cells.
TGF␤ Sensitivity-We next examined whether increased expression of RI or RII alone could affect the sensitivity of the transfectants to TGF␤ growth inhibitory activity. The NEO pool and RI-and RII-transfected cells were treated with increasing concentrations of TGF␤ 1 (0.2-25 ng/ml) for 4 days. Relative to the NEO pool, all RI clones showed increasing inhibition of DNA synthesis in a dose-dependent manner, whereas RII clone 37 did not show any increase in inhibition of DNA synthesis (Fig. 3). These results showed that increased RI expression could render the GEO cells more sensitive to growth inhibition by exogenous TGF␤ 1 , but increased RII expression had no effect. This indicated that RI but not RII is a limiting factor in TGF␤ signal transduction in GEO cells.
Fibronectin and Integrin ␣5 mRNA Expression-We then determined whether increased RI expression could induce expression of ECM molecules. Exponentially growing NEO pool and RI clones were treated with increasing concentrations of TGF␤ 1 for 24 h. RNase protection assays were performed, and fibronectin, integrin ␣5, and actin mRNA levels of the NEO pool and RI clone 32 are shown in Fig. 4. Contrary to our expectations, fibronectin mRNA levels remained unchanged in the NEO pool and RI clone 32 after TGF␤ 1 treatment (Fig. 4A). Similarly, the expression of integrin ␣5 was also not induced by TGF␤ 1 treatment, although the expression level of integrin ␣5 was low in the GEO cells (Fig. 4B). Similar results were also observed in RI clone 46.
Growth Arrest at Low Cell Density-Growth curves for RI clones and the NEO pool were performed to determine whether increased expression of RI would lead to alterations of growth properties of the cells (Fig. 5). Growth rates in the exponential phase were essentially similar for the NEO pool and RI clones, but RI clones 32 and 46 entered log phase (day 6) later than the NEO pool (day 4). All three RI clones had a slightly lower saturation density than the NEO pool.
Plating Efficiency Assay-Since the growth curve data suggested that RI affected the growth properties of GEO cells in the lag phase when the cell density was low, we compared the ability of the NEO pool and RI and RII clones to expand and form colonies at a low seeding density. The NEO pool and RI and RII clones were plated in 24-well plates in McCoy's 5A serum-free medium at 400 cells/well. After 2 weeks of incubation, RI clones showed a significant reduction of colony formation as compared with the NEO pool. As shown in Fig. 6A, RI clone 46 had fewer and smaller colonies than the NEO pool, whereas RII clone 37 had a similar number and size of colonies as that of the NEO pool. Absorbance measurements showed that the cloning efficiency of RI clones 32 and 46 were about 25 and 33%, respectively, of that of the NEO pool, whereas RII clone 37 showed no reduction of cloning efficiency relative to the NEO pool (Fig. 6B). The RNase protection assay (Fig. 1) showed that tetracycline could suppress the expression of the transfected RI. To confirm that the reduction of cloning efficiency of RI clones was due to expression of RI, tetracycline was added to the culture medium to rescue the cells from RI effects. The cloning efficiency of RI clones 32 and 46 was stimulated by 67 and 48%, respectively, after tetracycline treatment, whereas the NEO pool showed no significant response (Fig. 7).
Taken together with the growth curves, these data suggest that transfection of RI but not of RII enhanced autocrinenegative TGF␤ activity as well as the inhibitory response to exogenous TGF␤ 1 . To test this hypothesis, TGF␤ 1 neutralizing antibody was used to suppress autocrine-negative activity of TGF␤, as described previously (8). As expected, TGF␤ 1 neutralizing antibody increased colony formation of RI clones, as reflected by an increase in the number and size of colonies. As shown in Fig. 8A, the stimulatory effect of the TGF␤ 1 neutralizing antibody on RI clone 32 but not on the NEO pool cells is significant compared to the control antibody-treated cells. Absorbance measurements showed that the cloning efficiency of RI clones 32 and 46 was increased after neutralizing antibody treatment by 60 and 47%, respectively, whereas that of the NEO pool was not significantly increased (Fig. 8B). Addition of TGF␤ 1 together with the neutralizing antibody reversed this increase in proliferation (Fig. 8), thus demonstrating that the RI expression increased autocrine-negative TGF␤ activity.
Anchorage-independent Growth-The ability to form colonies in soft agarose is reflective of malignant transformation. Therefore, to assess the effect of increased RI or RII expression on the malignant properties of GEO transfectants, we compared the colony formation of the NEO pool and RI and RII clones in soft agarose. RI clone 46 showed a striking reduction in cloning efficiency in semisolid medium compared to the NEO pool, whereas RII clone 37 had a similar cloning efficiency as the NEO pool (Fig. 9). These results indicated that increased expression of RI but not of RII reduced anchorage-independent growth of GEO cells, further demonstrating that RI is the limiting receptor in determining the lack of TGF␤ responsive- ness and/or autocrine TGF␤ in GEO cells.
Tumorigenicity-Exponentially growing NEO pool and RI clones 32 and 53 cells were inoculated into nude mice at 2 ϫ 10 6 cells/site and monitored for the development of xenograft formation. The NEO pool cells formed xenografts in 9 of 10 inoculation sites and grew rapidly. Xenograft formation of RI clones 32 and 53 was delayed compared with the NEO pool (Fig. 10). The time needed to form xenografts of Ͼ100 mm 3 was 6 days for the NEO pool and 10 and 15 days for RI clones 32 and 53, respectively. RI clone tumors grew at a slower rate than the NEO pool tumors. Xenografts of RI clones 32 and 53 were approximately 25 and 40% that of the NEO pool 26 days after inoculation. These results indicated that increased expression of RI in GEO cells reduced in vivo growth and experimentally establishes the potential for RI tumor suppressor activity. DISCUSSION Based on the identification of a subset of colon tumors with reduced RI expression and the demonstration that RII has tumor suppressor activity (21), we hypothesized that loss or reduction of RI expression could lead to an increase in the malignant properties of carcinoma cells. The results from this study support this hypothesis. Increased RI expression in one of the colon carcinoma cell lines with a low level of RI expression increased TGF␤ responsiveness as well as autocrine-negative activity and reduced in vivo malignancy, indicating that low RI expression can be a limiting factor for TGF␤ response and autocrine-negative activity. In addition, these studies experimentally demonstrate the importance of the relative stoichiometry of both receptor types for TGF␤ signal transduction.
We used rat RI cDNA for transfection to differentiate the transfected RI mRNA from the endogenous RI mRNA. Rat RI protein is predicted to be the same size as human RI (29). Rat and human RI show approximately 77% homology in terms of amino acid identities (33). However, 125 I-labeled TGF␤ crosslinking assays indicated a lower molecular weight for rat RI. This may be due to differences in the degree of glycosylation between rat and human RI proteins. Similar species differences in glycosylation have been noted for RII (34). It has been shown that TGF␤ receptors from different species have varying degrees of affinity to one another (35). However, in our system, rat RI had the ability to complex with human RII, bind ligands, and propagate signals downstream, as previously shown by Feng et al. (36).
Transfection of RI into GEO cells led to increased autocrine TGF␤ activity with alterations of growth parameters of the cells similar to our previous reports for RII transfection (8,13). Briefly, the transfected cells showed a delay in reaching log phase, a low saturation density, reduced clonogenicity in a plating efficiency assay, and diminished in vivo tumorigenicity relative to the control cells. Despite increased autocrine TGF␤ activity and striking effects on plating efficiency and tumorigenicity, RI-transfected cells showed only a modest enhancement in growth-inhibitory response to exogenous TGF␤.
RI-transfected cells displayed increased sensitivity to TGF␤ growth-inhibitory effects. However, there was no induction of fibronectin and integrin ␣5 expression by TGF␤ 1 in these cells (Fig. 4). Uncoupling of the growth-inhibitory and ECM induction responses to TGF␤ has been observed in several other studies (8,10,35,(37)(38)(39). Although it is still possible that there may be activation of some other ECM-related genes after TGF␤ treatment of these cells, it may be that TGF␤ signal transduc- FIG. 7. Reversal of plating efficiency of GEO RI clones by tetracycline. The GEO NEO pool and RI clones 32 and 46 were plated in 24-well plates at 400 cells/well in the absence or presence of 0.1 g/ml tetracycline. The relative cell number was determined as described in Fig. 6. Reversal by tetracycline is expressed as the percentage increase of absorbance relative to the untreated cells. The values are the means (bars, S.E.) of four replicates.
FIG. 8. Autocrine TGF␤ activity in the GEO NEO pool and RI transfectants. The GEO NEO pool and RI clones 32 and 46 were plated in 24-well plates at 400 cells/well in the presence of 10 g/ml normal chicken IgG or 10 g/ml TGF␤ 1 neutralizing antibody described under "Materials and Methods." Cell colonies were stained and photographed, and the relative cell number was determined as described in Fig. 6. A, the effect of TGF␤ 1 neutralizing antibody in stimulating clonogenicity of RI clone 32 in contrast with the NEO pool in the presence or absence of exogenous TGF␤ 1 . B, the quantitation of the colony formation of the RI clones 32 and 46 and NEO pool cells. Stimulation by TGF␤ 1 neutralizing antibody and inhibition by exogenous TGF␤ 1 are expressed as the percentage change of absorbance relative to normal IgG-treated cells. The values are the means (bars, S.E.) of four replicates. tion pathways for these two types of TGF␤ responses diverge downstream of receptor binding. Thus, the effectors of ECM induction may be less sensitive than growth inhibition effectors after enhancement of RI expression in contrast to other studies involving RII reconstitution (8,10), or there may be defects in the specific transactivating factors associated with transcription of ECM molecules in GEO cells.
There are several possible explanations as to why increased RI expression would lead to elevated TGF␤ sensitivity. One is that an increased amount of RI expression leads to increased RI kinase activity in the presence of excess RII, which can then complex with and phosphorylate RI. Another mechanism could involve a change in the ratio of RII and RI. Transfection of RI could restore the stoichiometric ratio of RII and RI. Vivien et al. (40) showed that TGF␤ mediates responses through the heteromeric RI and RII complex but not through homo-oligomeric RI or RII complexes. This suggests that excessive RII levels may form inactive homomeric complexes, preventing the formation of active heteromeric complexes between RI and RII. Increased RI expression in GEO cells with low endogenous RI expression would change the ratio of RII and RI and thus prevent the sequestration of homomeric RII complexes. Further biochemical analysis of complex formation by RI and RII will be necessary to determine which of these possibilities occurs. However, since overexpression of RII did not lead to additional loss of TGF␤ sensitivity, it would suggest that RI transfection simply leads to increased heteromeric RI/RII complex formation with increased RI activation by RII. FIG. 9. Anchorage-independent colony formation in soft agarose by GEO NEO control and RI and RII transfectants. Exponentially growing cells were suspended at 3.0 ϫ 10 3 cells/ml in 1 ml of 0.4% SeaPlaque agarose in McCoy's 5A serum-free medium and plated on top of 1 ml of 0.8% agarose in the same medium in a 6-well tissue culture plate. Cell colonies were visualized by staining with 0.5 ml of p-iodonitrotetrazolium violet after 2 weeks of incubation.