Activin Type II Receptor Restoration in ACVR2-Deficient Colon Cancer Cells Induces Transforming Growth Factor-β Response Pathway Genes

The activin type II receptor (ACVR2) gene encodes the type II subunit of the activin receptor complex. The type II subunit is essential to activin-mediated signaling; the extracellular binding domain binds to activin, and the intracellular kinase domain activates the type I subunit that activates SMADs (reviewed in ref. 1). Our previous studies detected very frequent frameshift mutations in the A8 tract of exon 10 of ACVR2 in gastrointestinal cancers with frequent microsatellite instability (MSI-H; colon cancers, 58%; gastric cancers, 44%; ref. 2). Another study identified biallelic mutation of ACVR2 in 86% of MSI-H colon and pancreatic cancer xenografts and cell lines (3). Loss of ACVR2 protein was also reported in the majority of MSI-H tumors harboring frameshift mutation at the polyadenine tract of exon 10 of ACVR2 (4).

Activin signaling is involved in the regulation of apoptosis, differentiation, proliferation, and cell migration in many tissues, including epithelium, lymphocytes, prostate cancer, breast, vascular endothelium, and liver (reviewed in ref. 1). The activin ligand binds to a heterodimeric transmembrane activin-receptor complex with serine/threonine kinase activity that consists of type I and type II subunits (reviewed in ref. 1). This receptor complex belongs to the transforming growth factor (TGF)-β receptor family and, as does the TGF-β receptor complex, takes the SMAD family of proteins as its downstream signal transducers (5). On binding to activin, the activin-receptor complex phosphorylates SMAD2 and SMAD3 in the cytoplasm, resulting in their activation. Phosphorylated SMADs form a complex with SMAD4 and activate transcription of downstream genes (6).

Some members of the activin signaling pathway have been implicated as tumor suppressor genes. SMAD4 has been reported as a tumor suppressor in human pancreatic and colon cancers (7). SMAD2 is mutated in colon and lung cancers (5). Similarly, mutational inactivation of the activin type I receptor gene (ACVR1) have been observed in pancreatic cancers (8). SMAD3 null mice develop metastatic colon cancers (9). A dominant negative mutant ACVR2 also abolishes activin-mediated erythroid differentiation (10). Finally, a recent study showed that activin signaling exerts growth-suppressive effects in colon cancer cells (11).

Microsatellite instability occurring within coding regions underlies tumorigenesis in cancers with frequent microsatellite instability (MSI-H cancers; ref. 12). Frequent frameshift mutations in MSI-H cancers have been reported in TGFBR2 as well as ACVR2 (2, 13). To characterize the influence of ACVR2 gene frameshift mutation on MSI-H colon cancer cells, we analyzed changes in global molecular phenotype after re-expression of constitutively active wild-type (wt)-ACVR2 using the following: (a) microarray-based gene expression profiling; and (b) analysis of phospho-SMAD2 (p-SMAD2) expression.

The entire coding region of wt-ACVR2 was PCR-amplified and cloned into the plasmid vector pEF6/V5-His-TOPO (Invitrogen, Carlsbad, CA). The plasmid construct was purified with the HiSpeed Plasmid Midi Kit (Qiagen, Valencia, CA) and analyzed by sequencing and restriction enzyme digestion to confirm the sequence and direction of the insert.

The MSI-H colon cancer cell line HCT-15 was used for transfection experiments. HCT-15 carries a biallelic ACVR2 frameshift mutation at an A8 tract in exon 10 and is inactivated in TGFBR2 by mutations in both alleles (deletion at an A10 tract in exon 3 and a missense mutation in exon 5: L452P; ref. 14). HCT-15 were cultured in growth medium at 37°C with 5% CO₂. Cells were reseeded in six-well plates at a density of 2.5 × 10⁵ cells per well one day before transfection. Lipofectamine 2000 (Invitrogen) was used to transfect 4 μg of the expression plasmid encoding wt-ACVR2 as well as control plasmid containing the LacZ gene (pEF6/V5-His-TOPO/lacZ). Ten stable wt-ACVR2 and eight stable vector-control transfectants were selected by expanding single cells in growth media containing 14 μg/mL Blasticidin S HCl (Invitrogen).

Direct cell counting was done as follows: Three wt-ACVR2-transfected clones and three control vector-transfected clones were plated at a density of 5 × 10³ cells per well in 96-well plates in growth media in triplicates at day 0 and counted every day over a 5-day period.

Bromodeoxyuridine (BrdUrd) incorporation assay was done as follows: The experiment was done with Cell Proliferation ELISA BrdUrd (colorimetric) Kit (Roche Applied Science, Indianapolis, IN) according to the manufacturer’s protocol. Detailed protocol is available in the Supplementary Methods.

The rabbit polyclonal antibody against ACVR2 has been produced by Washington Biotechnology (Baltimore, MD). Briefly, two rabbits were twice immunized with a synthetic peptide (nwekdrtnqtgvepcy; 36–51) corresponding to the extracellular domain of ACVR2. Six weeks later, the antibodies were affinity chromatography purified from the antisera of the rabbit. Rabbit polyclonal antibodies against SMAD2/3 and p-SMAD2 were purchased from Upstate Biotech (Lake Placid, NY) and Cell Signaling Technology (Beverly, MA; used for activin stimulation experiments). Rabbit polyclonal antiactin antibodies were purchased from Santa Cruz Biotechnology Biotech (Santa Cruz, CA).

Cell lysates were pelleted and then resuspended in 200 μL cell lysis buffer [NaCl 149 mmol/L, NP40 0.01%, Tris 50 mmol/L (pH 7.8), and protease inhibitor cocktail 0.5% (Sigma, St. Louis, MO)]. The protein concentration was determined with the BCA Protein Assay Kit (Pierce, Rockford, IL) with human serum albumin as a standard. The samples were electrophoresed in 10% NuPAGE gel (Invitrogen) and transferred onto polyvinylidene difluoride membrane (Invitrogen). The membranes were immunoblotted with anti-ACVR2 polyclonal antibody (1:5,000 dilution), anti-Smad2/3 antibody (2 μg/mL), and antiphosphorylated Smad2 (1:1,500 dilution). Target protein bands were visualized with ECL Western Blotting detection kit (Amersham Pharmacia Biotech, Piscataway, NJ). The antiactin antibodies were used as a loading control.

For activin stimulation experiments, wt-ACVR2- and control vector transfectants were plated in six-well plates at a density of 10⁶ cells per well and cultured for 24 hours. The cells were then starved overnight in serum-free medium before the stimulation with 10 or 100 ng of recombinant activin A (Calbiochem, San Diego, CA). The protein extracts from both untreated and treated cells were obtained after 1 or 2 hours of stimulation with PhophoSafe Extraction Buffer (Novagen, San Diego, CA). HeLa cells treated with 100 ng of TGF-β1 (Roche Applied Science) were used as a positive control. Thirty 30 micrograms of total protein extracts were electrophoresed and transferred as described above. The membranes were immunoblotted with anti-phospho–SMAD2 (1 μg/mL) or anti-SMAD2/3 (1 μg/mL) antibody.

Total RNA was extracted with TRIzol reagents (Invitrogen) and was treated with RNase-free DNase on the RNeasy columns (Qiagen).

ACVR2 expression was measured with TaqMan method-based real-time quantitative RT-PCR as described in the Supplementary Methods. β-actin was used as a normalization control. The cDNA from untransfected HCT-15 cells was used as a quantification standard. The formula for normalization was as follows: ratio of sample to reference cDNA = [ACVR2(s)/ACVR2(r)]/[(β-actin(s)/(β-actin(r)], where ACVR2(s) and ACVR2(r) were expression levels of ACVR2 in the samples and reference cDNA, respectively, and β-actin(s) and β-actin(r) were β-actin RNA expression levels in the samples and reference. For validation of microarray results, real-time quantitative one-step RT-PCR analysis was done with Quantitect SYBR Green RT-PCR kit (Qiagen) on iCycler (Bio-Rad, Hercules, CA). Normalization to β-actin expression level was done as well.

The sequences of all of the primers and probes are shown in the Supplementary Table 2. Detailed real-time quantitative PCR methods are available in the Supplementary Methods.

Microarray analysis of three wt-ACVR2-transfected and three control vector-transfected clones was done as described previously (15). Briefly, amplified RNA was obtained from 20 to 50 μg of total RNA from each clone with a T7-based protocol and was labeled with Cy5 with random primers and reverse transcriptase. The reference sample was a pool of amplified RNAs from eight human malignant cell lines labeled with Cy3. Each Cy5-labeled specimen probe and the Cy3-labeled reference probe were cohybridized to a microarray slide containing 8,064 sequence-verified human cDNA clones. Probe preparation and hybridization was done individually for each sample clones. After hybridization, each slide was scanned with a GenePix 4000A dual-color slide scanning system (Axon Instruments, Union City, CA).

We performed both within-slide and between-slide normalization before analysis with LOWESS curve-fitting methods (15). Significance analysis of microarrays was applied to the Lowess-normalized log-scaled data to select genes that were significantly differentially expressed between wt-ACVR2-transfected and control vector-transfected clones (15). Significantly differential expression was determined based on significance analysis of microarrays score, a score assigned to each gene on the basis of its fold change of average expression levels between the wt-ACVR2 transfectants and the control vector transfectants relative to the cumulative SDs of expression levels for wt-ACVR2 transfectants and control vector transfectants. In this study, only genes with a significance analysis of microarrays score higher than 2 (up-regulation in wt-ACVR2 transfectants) or smaller than −2 (down-regulation in wt-ACVR2 transfectants) were classified as significant.

Genes previously related to TGF-β or activin signaling were identified by online database searches. The following web sites were used for this search: PubMed (http://www.ncbi.nim.nih.gov), Stanford SOURCE (http://genome-www5.stanford.edu/cgi-bin/source/sourceResult), Human Genome Browser Gateway (http://genome.ucsc.edu), and GeneCards (http://bioinfo.weizmann.ac.il/cards-bin/cardsearch.pl).

We established a model system for studying wt-ACVR2 function by reconstituting its activity in HCT-15 colon cancer cells, which are ACVR2-deficient secondary to a native biallelic nonsense mutation. We then analyzed the global molecular phenotype of these cells before and after ACVR2 reconstitution. In addition, we evaluated the impact of ACVR2 mutation and restoration on its known downstream effector, SMAD2.

After stable transfection and selection of 18 single-cell clonal transfectants (10 ACVR2- and eight control vector-transfected clones), ACVR2 mRNA levels were measured by real-time quantitative RT-PCR analysis. Measurements in all of the 10 wt-ACVR2 stable transfectants confirmed that ACVR2 mRNA levels were higher than in all of the eight control vector-transfected clones, as shown in Fig. 1,A. Next, ACVR2 protein expression level was evaluated by Western blotting. ACVR2 protein was detectable in all of the 10 wt-ACVR2 transfectants, and higher levels of ACVR2 protein were observed in all of the 10 wt-ACVR2 transfectants than in vector control transfectants (Fig. 1 B). On the basis of the data for ACVR2 mRNA and protein levels, the three wt-ACVR2 transfectants exhibiting the highest mRNA expression levels were selected for additional analyses.

Restoration of wt-ACVR2 function in HCT-15 colon cancer cells resulted in slower cell growth measured by direct cell counting. The difference in growth between wt-ACVR2 transfectants and control vector transfectants was statistically significant for days 3, 4, and 5 (P < 0.05, Student’s t test; Fig. 2). The calculated doubling time was longer in the wt-ACVR2-transfected cells (20 hours) than in the controls (17.4 hours). Additionally, BrdUrd incorporation rate during DNA replication was significantly decreased in wt-ACVR transfectants compared with controls (P < 0.05, Student’s t test; data not shown).

The effect of ACVR2 mutation and restoration on the known target downstream effector, SMAD2, was analyzed by Western blotting. Cell lysates from wt-ACVR2-transfected and untransfected (i.e., native mutant) HCT-15 cells were analyzed for expression of both total SMAD2 and its phosphorylated form (p-SMAD2). The p-SMAD2 expression was higher in wt-ACVR2-transfected than in untransfected HCT-15 cells, whereas total SMAD2 protein levels were identical in both (Fig. 3,A). We also evaluated the induction of SMAD2 phosphorylation by activin stimulation. A dose- and time-dependent increase in p-SMAD2 expression was observed in response to activin in wt-ACVR2 transfectants, whereas no effect of activin stimulation on p-SMAD2 level was observed in control vector transfectants (Fig. 3 B).

To additionally delineate the downstream effects of ACVR2 activation, and to provide insights into candidate downstream pathways discrete from SMAD signaling, we performed gene expression profiling using cDNA microarrays. In these experiments, three clonal wt-ACVR2 transfectants were compared with three clonal control vector transfectants. Significance analysis of microarray was used to select genes that were significantly differentially expressed between these two groups (Table 1; Supplementary Table 1). We eliminated the possibility that modification in gene expression was due to transfection reagents, selection media, or antibiotics by using control vector-transfected clones, rather than parental cells, in this gene expression comparison. ACVR2 expression in wt-ACVR2-transfected cells was the highest of all 8,064 genes on the microarrays. This result was not only expected, but also served as a validation of the cDNA microarray and model system developed in this study.

Genes significantly influenced by wt-ACVR2 transfection are shown in Table 1 (induced genes) and in Supplementary Table 1 (suppressed genes). These included genes induced by growth factors (e.g., RHOB, MAPK6, HGS, and PPAP2B); negative regulators of cell proliferation, such as BTG1, PMP22, or HGS; genes implicated in cellular growth regulation (e.g., CYR61, RHOB, GPC1, JUN, INHA, PPAP2B, and GPC1); and genes involved in intercellular adhesion (e.g., CLSTN1, LAMP2, PVRL3, PVRL2, MLLT4, ARHGDIA). A series of signal transducers, including ARHE, MAPK6, LTB, PPP2R2C, PP1R3A, MAP2K3, RAB6A, as well as several regulators of transcription (e.g., JUN, JUND, FOSB, ATF3, JUNB, EGR1, VGLL1, CEBPA, MSX1, and IRF1) were also induced by wt-ACVR2. Expression of the transcriptional repressor ATF3 was increased by wt-ACVR2. Proapoptotic genes such as NR4A1, DUSP2, and TNFRSF10C were overexpressed after wt-ACVR2 restoration, whereas the antiapoptotic gene BIRC5 was down-regulated.

To validate our microarray results, we performed real-time quantitative RT-PCR analysis. Six genes found to be up-regulated in wt-ACVR2 transfectants by cDNA microarrays (ARHE, ARHGDI, CYR61, FOSB, JUN, and JUND) were analyzed. All six of these genes exhibited increased mRNA levels in wt-ACVR2 transfectants compared with the control vector transfectants, confirming our cDNA microarray results (Supplementary Figure).

The involvement of activin signaling disruption in the origin or progression of human digestive tract cancer has been suspected based on the high ACVR2 and ACVR1 mutation rate in various tumors discovered recently (2, 3, 4, 8). However, mechanisms of activin signaling and the significance of the ACVR2 mutation in human tumorigenesis have not yet been fully elucidated. Activin and TGF-β share the same receptor binding properties, and their receptors exhibit the same substrate specificity, namely phosphorylation and activation of SMAD2 and SMAD3 (5). Because both TGF-β and activin use the same set of SMADs, it is conceivable that they share common regulatory mechanisms. Therefore, the identification of genes transcriptionally regulated by ACVR2 and elucidation of the molecular mechanisms responsible for this transcriptional regulation will bring us closer to a better understanding of both the activin and TGF-β signaling pathways. Gene expression profiling has been used by a variety of investigators to explore genetic events involving TGF-β signaling (16, 17, 18, 19). The current study used this approach to identify new participants in the activin signaling pathway and to explore their commonality with those of the TGF-β signaling pathway.

Initially, we evaluated the effect of restoration of wt-ACVR2 function in a colon cancer cell line, HCT-15, carrying a biallelic ACVR2 frameshift mutation at the mutational hotspot where the majority of human primary MSI-H digestive tract cancers demonstrate ACVR2 frameshift mutation (2, 3, 4). We showed that restoration of activin signaling by wt-ACVR2 transfection in this cell line resulted in increased SMAD2 phosphorylation in response to activin stimulation. This finding establishes that this biallelic mutation impairs signal transduction and implicates loss of ACVR2-mediated SMAD signaling in MSI-H colon cancer. Furthermore, the restoration of wt-ACVR2 function caused decreased cell growth in MSI-H colon cancer cells in vitro.

Global molecular phenotyping revealed numerous similarities between signaling via TGFBR2 and via ACVR2. For example, the AP-1 complex members including FOS and JUN have been implicated in signaling initiated by TGF-β (20). In the current study, AP-1 complex members JUN, JUND, JUNB, and FOSB were up-regulated by wt-ACVR2, suggesting that activin and TGF-β signaling share AP-1 involvement as effector mechanisms. Phosphorylated JUN and JUND have higher DNA-binding affinities than their nonphosphorylated counterparts, making them logical targets for the phosphorylation-mediated signaling shared by TGFBR2 and ACVR2 (20). Moreover, JUND is an inhibitor of normal intestinal mucosal growth in vivo and plays a critical role in the negative control of epithelial cell renewal under physiologic and pathological conditions (21). Similarly, JUN is a regulator of transcription, cell growth and maintenance, and interacts with the SMAD3/4 heterodimer (20). Another newly observed similarity between ACVR2 and TGFBR2 signaling concerns the up-regulation of genes induced or activated by growth factors. For example, the Rho protein family member RHOB is rapidly induced by TGF-β, epidermal growth factor, and platelet-derived growth factor (22). In the current study, RHOB was up-regulated in wt-ACVR2 transfected cells. RHOB is an immediate-early gene implicated in growth control as a potent inhibitor of malignant transformation as well as a suppressor of tumor growth and has been suggested to be a novel mechanism of tumor suppression by TGF-β (18, 19, 22). Interestingly, two other genes involved in Rho protein signaling, ARHE and ARHGDIA, were also up-regulated in wt-ACVR2 transfectants. Transcriptional repressor ATF3 is a TGF-β–inducible factor that forms a complex with SMAD3 (23). ATF3 expression was found to be increased by TGF-β but not by bone morphogenetic protein (23). In the current study, ATF3 was overexpressed in wt-ACVR2 transfectants, suggesting that activin is another TGF-β family member that can induce the expression of ATF3. Finally, in the current study, wt-ACVR2 restoration up-regulated a growth factor-inducible immediate-early protein, CYR61, that promotes proliferation, migration, and adhesion, which was known to be similarly up-regulated by TGFBR1 and ACVR1 (19).

Furthermore, the current study revealed a number of genes that have not previously been implicated in TGFBR2 signaling. For example, up-regulation by wt-ACVR2 was observed for the proapoptotic genes NR4A1, DUSP2, and TNFRSF10C in addition to down-regulation of the antiapoptotic gene BIRC5. Increased expression of negative regulators of cell proliferation, including BTG1, PMP22, and TOB2 after wt-ACVR2 restoration was observed as well. Additionally, the following genes were also up-regulated by wt-ACVR2 transfection: MAPK6, a gene activated in response to growth factors through protein phosphorylation; HGS, a negative regulator of cell proliferation that undergoes tyrosine phosphorylation in response to epidermal growth factor and platelet-derived growth factor; and PPAP2B, a growth control gene that is known to be enhanced by epidermal growth factor in HeLa cells (24, 25, 26).

Recently, significant progress has been made toward identifying signal transduction pathways activated by TGF-β receptor family members. However, fewer studies have been published regarding the downstream effectors of activin receptors. In the present study, we focused our attention on ACVR2-regulated genes. We combined examination of the SMAD-mediated signaling pathway with gene expression profiling. Among genes identified as influenced by ACVR2 restoration, some had previously been related to TGF-β receptor signaling, whereas others were not previously suspected in either the ACVR2 or TGFβR2 pathways. The observed strong similarity between the activin and TGF-β signaling systems suggests that activin serves as an alternative activator of downstream TGF-β effectors in addition to SMADs. In addition, we confirmed that restoration of wt-ACVR2 function resulted in decreased cell growth and increased SMAD2 phosphorylation in a MSI-H colon cancer cell line with native biallelic ACVR2 frameshift mutation. Taken together, these data suggest that activin may serve as an alternative activator of SMADs and other downstream TGF-β effectors and that frameshift mutation of ACVR2 may contribute to MSI-H colon tumorigenesis via disruption of alternate TGF-β effector pathways.