The bone morphogenetic protein pathway is active in human colon adenomas and inactivated in colorectal cancer

Abstract

BACKGROUND.

Transforming growth factor β (TGFβ) is important in colorectal cancer (CRC) progression. Bone morphogenetic proteins (BMPs), a subgroup within the TGFβ superfamily, recently also have been implicated in CRC, but their precise role in CRC has yet to be investigated.

METHODS.

The authors used a tissue microarray and immunohistochemistry of BMP receptors and signal transduction elements in adenomas and CRC specimens to elucidate the role of BMP signaling in CRC carcinogenesis.

RESULTS.

The adenoma specimens expressed all 3 BMP receptors (BMPRs) (BMPR type 1a [BMPR1a], BMPR1b, and BMPR2) and expressed SMAD family member 4 (SMAD4); and 20 of 22 adenomas (90.9%) exhibited active BMP signaling, as determined by nuclear phosphorylated SMAD1,5,8 (pSMAD1,5,8) expression. In contrast, pSMAD1,5,8 nuclear staining was present in 5 CRC specimens (22.7%) but was lost in 17 CRC specimens (77.3%; cancer vs adenoma; P < .0001). The earliest loss of pSMAD1,5,8 nuclear staining was detected in regions of high-grade dysplasia/carcinoma in situ within adenomas. CRCs showed frequent loss of BMPR2 (P < .0001) and SMAD4 (P < .01) compared with adenomas. Negative expression of BMPR2 was observed more frequently in earlier stage cancers (Dukes stage B) than in advanced cancers (Dukes stage C; P < .05).

CONCLUSIONS.

Taken together, the current results indicated that loss of BMP signaling correlates tightly with progression of adenomas to cancer and occurs relatively early during cancer progression. Cancer 2008. © 2007 American Cancer Society.

Colorectal cancer is the second leading cause of cancer death in the U.S. and Europe.1 The transformation of colon epithelial cells to cancer follows a predictable progression of histologic and cytologic changes, which are the results of genetic alterations that occur during the adenoma-carcinoma sequence of colorectal cancer (CRC) formation.2 Accumulated genetic changes and the selection of advantageous mutations underlie the development of neoplasia and drive the initiation and progression of CRC. On the basis of the association of these mutations with different steps in the multistep progression model of CRC formation, these mutations may provide unique insights into the process of tumorigenesis and may be used as molecular markers for the detection of colon adenomas or colon carcinomas.

It is believed that the transforming growth factor β (TGFβ) signaling pathway plays a central role in CRC, particularly in tumor progression, invasion, and metastasis.3 The TGFβ superfamily consists of the TGFβ, activin, and bone morphogenetic protein (BMP) subfamilies. The TGFβ superfamily plays an important role during embryogenesis and regulates many processes, including cellular proliferation, differentiation, and apoptosis.4

BMPs and their receptors have been investigated previously in other solid tumors. The expression of BMP receptors (BMPRs) correlates with tumor grade in human prostate cancer,5 and loss of BMPR2 may be a prognostic marker in patients with prostate cancer.6 The important evidence pointing to a likely role of BMPs in colonic neoplasia is the finding that BMPR type 1a (BMPR1a) and SMAD family member 4 (SMAD4) frequently are mutated in juvenile polyposis, an inherited syndrome that carries a high risk of developing CRC.7–9 This also is supported by a transgenic mouse model of juvenile polyposis, the villin-Noggin mouse, in which the BMP signaling is abrogated completely, and the mouse develops neoplasia.10 We recently demonstrated that BMP acts as a tumor suppressor that promotes apoptosis in mature colonic epithelial cells; therefore, perturbations in BMP signaling may lead to increased tumorigenesis.11

BMPs initiate signaling by binding cooperatively to transmembrane serine (Ser)-threonine kinase receptor types 1 and 2, triggering the phosphorylation and activation of the type 1 receptor by the type 2 receptor kinase. The activated type 1 receptor phosphorylates SMAD1, SMAD5, and SMAD8, and this permits their association with SMAD4. This heterodimeric complex then translocates to the nucleus and activates gene transcription.

The objectives of the current study were to determine whether and where BMP signaling is disrupted during the adenoma-carcinoma sequence of CRC formation. For this purpose, we performed immunohistochemistry using a tissue microarray (TMA) on material from 22 patients with simultaneous colorectal adenomas and carcinomas and determined the expression of BMP receptors and signal transduction elements specific for the BMP pathway. We report that the expression of BMPRs is intact and that the pathway is active at the adenoma stage of colon cancer progression. Conversely, BMPR2 and SMAD4 frequently are lost in cancer specimens, and BMP signaling is lost in the majority of CRCs.

MATERIALS AND METHODS

Selection of Patient Material

An overview of the clinicopathologic data is provided in Table 1. Formalin-fixed, paraffin-embedded tissues from 22 patients who had adenomas and CRCs between 2002 and 2004 were used for the compilation of the TMA. The inclusion criteria were the presence of an adenoma and carcinoma in a colon resection specimen from the same patient and paraffin-embedded tissue specimens of approximately the same age (specimens that were prepared between 2002 and 2004) to prevent staining artifacts. Blocks were selected from the archives of the Pathology Department of the Academic Medical Center in Amsterdam. The study was approved by the investigators' Institutional Review Board. Our study included 12 men (54.6%) and 10 women (45.4%), and the patients ranged in age from 56 years to 92 years (mean age ± standard deviation, 74.5 ± 10.42 years; median, 76.5 years). The primary carcinomas (localization was known for all tumors) originated in the colon (n = 19 tumors) and the rectum (n = 3 tumors). The adenomas were classified histologically into 4 types; tubulovillous, villous, tubular, and serrated. An additional series of adenomas (n = 13 tumors) with regions of high-grade dysplasia or that contained carcinoma in situ was processed as standard whole tissue sections for immunostaining.

Table 1. Characteristics of Patients and Tumors

Patient			Adenoma			Cancer
Patient no.	Sex	Age, y	Histologic type	Size, cm	Grade of dysplasia	Location	Grade	Duke class
LGD indicates low-grade dysplasia; G, well-differentiated cancer, P, poorly differentiated cancer; M, moderately differentiated cancer.
1	Woman	56	Tubulovillous	4.5	LGD	Colon	G	C
2	Man	71	Serrated	<1	LGD	Colon	P	A
3	Man	77	Villous	3.5	LGD	Rectum	G	A
4	Woman	56	Villous	2.8	LGD	Colon	M	B
5	Woman	81	Serrated	<1	LGD	Colon	P	C
6	Woman	70	Tubulovillous	1.5	LGD	Rectum	M	C
7	Man	81	Tubulovillous	<1	LGD	Colon	M	B
8	Man	66	Villous	3	LGD	Colon	M	C
9	Man	72	Villous	<1	LGD	Colon	M	C
10	Man	83	Villous	<1	LGD	Colon	M	B
11	Woman	60	Tubular	<1	LGD	Colon	M	B
12	Man	79	Tubulovillous	<1	LGD	Rectum	M	B
13	Woman	88	Tubular	<1	LGD	Colon	P	B
14	Man	87	Tubulovillous	3.5	LGD	Colon	M	B
15	Man	70	Tubulovillous	<1	LGD	Colon	M	C
16	Man	87	Tubular	1.2	LGD	Colon	M	B
17	Woman	83	Tubulovillous	<1	LGD	Colon	P	B
18	Man	79	Tubular	<1	LGD	Colon	M	C
19	Man	62	Tubulovillous	<1	LGD	Colon	P	B
20	Woman	79	Tubulovillous	<1	LGD	Colon	P	C
21	Woman	72	Villous	<1	LGD	Colon	M	A
22	Woman	92	Tubulovillous	3	LGD	Colon	G	A

TMA Construction

For the TMA construction, we used 1 hematoxylin and eosin-stained slide from each block to define representative tumor regions. Tissue cylinders with a diameter of 0.6 mm were punched from the tumor areas of each block and brought into a recipient paraffin block by using a Manual Tissue Arrayer (MTA-1; Beecher Instruments, Sun Prairie, Wis). To overcome the problem of tissue microheterogeneity and to increase the number of evaluable patients, the TMA included 3 cores of tissue from each cancer specimen and 2 cores from each adenoma from the same patient. Thus, the analysis of protein expression represents the mean of the staining of 3 different cores from the same cancer and staining of 2 different cores from the same adenoma. For each patient, 1 tissue core from normal colon was included in TMA as a control.

Immunohistochemistry

TMA blocks were sectioned (4 μm), deparaffinized, blocked for endogenous peroxidase activity by immersion in 0.3% hydrogen peroxide in methanol for 20 minutes, and heat treated at 100°C in Tris, pH 9.0, for 10 minutes (antigen retrieval). Nonspecific binding sites were blocked with 5% normal goat serum for 10 minutes followed by incubation for 1 hour with the primary antibody at room temperature. Primary rabbit polyclonal antibodies to BMPR2 kindly were provided by Professor Ten Dijke (Dutch Cancer Institute, Amsterdam, the Netherlands) and were used at a concentration of 1:400. The specificity of the antibodies has been reported previously.12 Mouse monoclonal antibodies to SMAD4 were from Santa Cruz Biotechnology (Santa Cruz, Calif; dilution, 1:1600). The Powervision+ Poly-HRP detection system (ImmunoVision Technologies, Daly City, Calif) was used to observe the antibody binding sites with3,3-diamino-benzidine (DAB)+ as a chromogen. Sections were counterstained with hematoxylin. Negative control sections for all antibodies were processed in an identical manner after omitting the primary antibody and showed no staining.

Immunohistochemistry for Phosphorylated SMAD1,5,8, BMPR1a, and BMPR1b

Sections were deparaffinized first then immersed in 0.3% H₂O₂ in methanol for 20 minutes. Antigen retrieval was performed by boiling slides for 10 minutes in 0.01 mol/L sodium citrate (pH 6.0). Nonspecific binding sites were blocked with TENG-T (10 mmol/L Tris, 5 mmol/L ethylenediamine tetraacetic acid, 0.15 mol/L NaCl, 0.25% gelatin, and 0.05% [volume/volume] Tween 20, pH 8.0) for 30 minutes. Slides were incubated with primary rabbit polyclonal antibodies to phosphorylated SMAD1,5,8, which recognizes the double-phosphorylated forms of SMAD1 (Ser463/465), SMAD5 (Ser463/465), and SMAD8 (Ser426/428) (Cell Signaling, Beverly, Mass), either at a concentration of 1:50, or with goat polyclonal antibodies to BMPR1a (R&D) at a concentration of 1:100, or with mouse monoclonal antibodies to BMPR1b (R&D) at a concentration of 1:1000, overnight at 4°C in phosphate-buffered saline (PBS) with 0.1% Triton and 1% bovine serum albumin and incubated with biotinylated secondary goat antirabbit, goat antimouse, or rabbit antigoat antibodies (Dako, Glostrup, Denmark) at a concentration of 1:200 at room temperature for 1 hour in PBS with 10% human serum. Slides were then incubated with streptavidin-biotin-horseradish peroxidase (Dako) for 1 hour, and peroxidase activity was detected with “Fast DAB” (Sigma Chemical Company, St. Louis, Mo).

TMA Analysis

The cellular localization and pattern of immunoreactivity were examined by 2 investigators independently in a blinded fashion. BMPR1a, BMPR1b, and BMPR2 expression levels were graded according to both the intensity of staining and the number of cells stained, as shown in Table 2. Only staining of the cell membrane or intracytoplasmic staining was assessed. Scores of 0 and 1 were considered negative for BMPR expression, and scores of 2 and 3 were considered positive.

Table 2. Scoring System for Bone Morphogenetic Protein Receptor Types 1a, 1b, and 2

Intensity of staining	Percentage of cells stained
	<10%	10–30%	>30%
No staining	0	0	0
Weak staining	0	0	1
Moderate staining	0	1	2
Strong staining	0	2	3

For the evaluation of SMAD4 and pSMAD1,5,8 expression, only nuclear staining was assessed and was scored as shown in Table 3. Scores of 0 and 1 were considered negative for SMAD expression, and scores of 2 and 3 were considered positive.

Table 3. Scoring System for SMAD Family Member 4 (SMAD4) and Phosphorylated SMAD2,5,8

Intensity of staining	Percentage of cells stained
	<10%	10–30%	30–50%	>50%
No staining	0	0	0	0
Weak staining	0	0	1	1
Moderate staining	0	1	2	3
Strong staining	1	2	3	3

Statistical Analysis

Statistical analyses were performed with the Statistical Package of Social Science (SPSS) version 11.5 for Windows (SPSS Inc, Cary, NC). The chi-square test and Fisher exact tests were used as appropriate. The McNemar test was used to compare differences between adenomas and cancers. P values <.05 were considered statistically significant.

RESULTS

Expression of BMPRs and Signal Transduction Elements in the Normal Human Colon and Adenomas

The expression of BMPRs, SMAD4, and pSMAD1,5,8 was investigated in specimens of normal colon and adenomas by using immunohistochemistry (Fig. 1). In normal and adenomatous tissue, the expression of BMP pathway components was localized predominantly in the epithelial cells. SMAD4 was expressed in both epithelial and stromal cells. The cellular staining pattern for BMPRs was both membranous and cytoplasmic. BMPRs were expressed in all normal and adenoma samples (Figs. 1 and 2). With SMAD4 and pSMAD1,5,8, both cytoplasmic and clear nuclear staining were observed. Positive nuclear staining for SMAD4 was observed in all normal and adenoma specimens. pSMAD1,5,8 was expressed in the nucleus of all normal specimens and in 20 of 22 adenomas (adenomas vs normal: P = .5; not significant), suggesting that the BMP pathway is active in normal colon tissue and in the majority of adenomas (Fig. 2) (Table 4).

Figure 1. These photomicrographs illustrate expression of SMAD family member 4 (SMAD4), bone morphogenetic protein receptor type 2 (BMPR2), and phosphorylated SMAD1,5,8 (pSMAD1,5,8) in normal human colon, in human colonic adenomas, and in human colon cancer specimens.

Figure 2. Schematic representation of the expression of bone morphogenetic protein (BMP) pathway components in 22 patients with adenomas and colorectal cancer. BMPR1a indicates BMP receptor type 1a; SMAD4, SMAD family member 4; pSMAD1, phosphorylated SMAD1; A, adenoma; C, cancer; white box, retained expression; black box, loss of expression.

Table 4. Expression of Bone Morphogenetic Protein Pathway Components in Adenomas and Colorectal Cancer Specimens

Component	Adenoma		Cancer		P
	Positive	Negative	Positive	Negative
BMPR1a indicates bone morphogenetic protein receptor type 1a; SMAD4, SMAD family member 4; pSMAD1,5,8, phosphorylated SMAD1,5,8.
BMPR1a	22	0	18	4	.125
BMPR1b	22	0	19	3	.25
BMPR2	22	0	10	12	<.0001
SMAD4	22	0	12	10	.002
pSMAD1,5,8	20	2	5	17	<.0001

Expression of BMPRs and Signal Transduction Elements in Human CRC Specimens

To determine whether or not a significant portion of patients with CRC have reduced levels of BMPRs compared with adenomas, cancer tissues from the same patients were evaluated for BMPR expression by using immunohistochemistry (Figs. 1 and 2). The frequency of loss of expression of BMPR1a and BMPR1b did not differ significantly between cancers and adenomas (Table 4). We observed that cancers had frequent loss of BMPR2 and SMAD4 expression compared with adenomas (P < .0001 and P = 0.002, respectively) (Table 4). Seventeen of 22 CRC samples (≈80%) had negative nuclear pSMAD1,5,8 staining, suggesting reduced BMP-pathway activity in these cancers (cancer vs adenomas: P < .0001). We demonstrated a strong, statistically significant association between defective expression of BMPRs or SMAD4 and negative nuclear staining for pSMAD1,5,8 (P = .024) (Table 5), suggesting that the abnormal expression of the BMPRs and SMAD4 influenced the activity of the BMP pathway in CRC.

Table 5. Association Between Nuclear Localization of Phosphorylated SMAD1,5,8 and Expression of Components of the Bone Morphogenetic Protein Pathway

BMP pathway (n=22)	pSMAD1,5,8 Expression: no. of specimens (%)*		P
	Negative	Positive
pSMAD1,5,8 indicates phosphorylated SMAD family members 1, 5, and 8; BMP, bone morphogenetic protein. * Percentages in parentheses refer to the percentage of the total number of specimens. † Defective means negative staining in 1 or more of BMP receptor type 1a (BMPR1a), BMPR1b, BMPR2, or SMAD4.
Defective†	16 (72.7)	2 (9.1)	.024
Normal	1 (4.5)	3 (13.6)

We also demonstrated a significant association between loss of BMPR2 and Duke stage. Negative expression of BMPR2 was more frequent in Dukes stage B tumors (80%) compared with Duke stage C tumors (25%) (see Table 6). No significant differences were observed for the remaining clinicopathologic or biologic variables (sex, age, and location and grade of tumors).

Table 6. Expression of Bone Morphogenetic Protein Receptor 2 in Colorectal Carcinomas in Relation to Duke Stage

BMPR2 expression	No. of patients (%)			P
	Duke stage A	Duke stage B	Duke stage C
BMPR2 indicates bone morphogenetic protein receptor type 2.
Score 0	1 (25)	2 (20)	2 (25)	.028
Score 1	1 (25)	6 (60)	0 (0)
Score 2	0 (0)	1 (10)	0 (0)
Score 3	2 (50)	1 (10)	6 (75)
Total	4 (100)	10 (100)	8 (100)

Expression of BMPRs and Signal Transduction Elements in Adenomas With High-grade Dysplasia/Carcinoma in Situ

Next, we examined pSMAD1,5,8 staining in a series of adenomas that contained areas of high-grade dysplasia/early carcinomas to determine the morphologic stage at which the loss of BMP pathway activity could be detected. In 10 of 13 patients, loss of nuclear pSMAD1,5,8 staining was observed. There was a clear difference in pSMAD1,5,8 nuclear expression between highly dysplastic/cancer loci and the surrounding adenomatous tissue (Fig. 3). This fits well with our results obtained from the TMA and suggests that BMP signaling is lost at the advanced adenoma/early cancer stage.

Figure 3. Selective loss of nuclear expression of phosphorylated SMAD family member 1,5,8 (pSMAD1,5,8) is observed in an area of high-grade dysplasia within an adenoma. Inset a: The area of adenoma with positive pSMAD1,5,8 nuclear expression. Inset b: The area of high-grade dysplasia with negative pSMAD1,5,8 nuclear staining.

DISCUSSION

Disturbances in TGFβ superfamily signaling result in various human diseases, including cancer. Several pathway elements act as tumor suppressors in a variety of human tumor types, including colon and pancreatic cancer.13, 14 Mutational inactivation of TGFβ receptor 2 (TGFβR2) and SMAD4 are a frequent event in CRC.15, 16 The potential role of the BMP pathway in colon cancer progression has so far received little attention despite the finding of BMPR1a germline mutations in families with juvenile polyposis, in which affected individuals develop multiple intestinal polyps with a high chance of developing cancer.17

We demonstrated previously that BMP signaling is required for apoptosis in colonic epithelial cells and acts as a tumor suppressor with characteristics to similar to those of TGFβ.11 In the current study, we demonstrated that CRCs, but not adenomas, frequently have reduced expression levels of BMPRs and SMAD4. We observed that the BMP pathway was inactivated in 80% of CRCs, as judged by nuclear pSMAD1,5,8 expression, but was active in adenomas. Selective loss of pSMAD1,5,8 nuclear staining in areas of high-grade dysplasia/carcinoma in situ within adenomas suggests that loss of BMP signaling is an event associated with the progression of adenomas to carcinomas. Taken together, these observations provide an insight into the potential role of BMPs in the adenoma/carcinoma sequence.

In the current study, no BMP pathway component abnormalities were detected in adenomas. This is in agreement with the finding that mutation of TGFβR2 is a late event in adenomas and correlates tightly with progression of these adenomas to cancer.18

We observed that the tumors with attenuated BMPR2 expression were significantly more likely to be Duke stage B cancers than advanced cancers. These results correlate with the previously described association between Duke stage B cancers and TGFβR2 mutations.19

TGFβR2 inactivation is associated with increased survival for patients in the late stages of CRC.20, 21 Despite the well-known tumor suppressor functions of TGFβ, there is considerable evidence that this pathway promotes tumor progression, invasion, and metastasis in the late stages of carcinogenesis.21 Inactivation of TGFβ signaling at a point when cells no longer are sensitive to the growth-inhibitory effect of TGFβ, and when the effects of TGFβ have switched from tumor suppressive to tumor promotional may explain the improved survival of patients with TGFβR2 inactivation. Because BMPs have characteristics similar to those of TGFβ, it is possible that BMPs may also act to promote tumor progression in the late stages of CRC progression. Inactivation of BMP signal transduction through inactivation of BMPR2 initially may release the adenoma cells from BMP-induced growth suppression; but, later, it protects the host from metastasis by switching off the metastasis-enhancing effects of BMPs. Further analysis will be necessary to confirm this hypothesis.

We also assessed whether changes in the protein expression of BMPR2 or SMAD4 influenced the activity of the BMP pathway as judged by pSMAD1,5,8 nuclear staining. Our results indicated that impaired BMPR2 or SMAD4 expression strongly correlates with reduced BMP-pathway activity. The influence of this inactivation of BMP signaling on cancer progression, invasiveness, and (eventually) patient outcome is a subject for further investigation.

We did not observe a correlation between BMP signaling inactivation and cancer location or degree of differentiation, which was been observed for TGFβ (right-sided, moderate or poorly differentiated CRC). This may be because of the small sample size, and more studies will be needed to test these associations.

In summary, the results of the current study demonstrate that BMP signaling is intact in normal colonic epithelium and adenomas of all types but frequently is inactivated in carcinomas, similar to TGFβ signaling (Fig. 4). The change occurs at the advanced adenoma/early carcinoma stage.

Figure 4. Modified representation of the molecular events associated with the polyp-carcinoma sequence according to Fearon and Vogelstein2 with the addition of the bone morphogenetic protein (BMP) pathway acting as a tumor suppressor at the late-advanced adenoma stage. APC indicates antigen-presenting cells; SMAD4, SMAD family member 4; TGFβR2, transforming growth factor β receptor 2.