Despite reports of sex steroid receptor and COX2 expression in desmoid-type fibromatosis, responses to single agent therapy with anti-estrogens and non-steroidal anti-inflammatory drugs are unpredictable. Perhaps combination pharmacotherapy might be more effective in desmoid tumors that co-express these targets. Clearly, further understanding of the signaling pathways deregulated in desmoid tumors is essential for the development of targeted molecular therapy. Transforming growth factor–β (TGFβ) and bone morphogenetic proteins (BMP) are important regulators of fibroblast proliferation and matrix deposition, but little is known about the TGFβ superfamily in fibromatosis. A tissue microarray representing 27 desmoid tumors was constructed; 14 samples of healing scar and six samples of normal fibrous tissue were included for comparison. Expression of selected receptors and activated downstream transcription factors of TGFβ family signaling pathways, β–catenin, sex steroid hormone receptors and COX2 were assessed using immunohistochemistry; patterns of co–expression were explored via correlational statistical analyses. In addition to β–catenin, immunoreactivity for phosphorylated SMAD2/3 (indicative of active TGFβ signaling) and COX2 was significantly increased in desmoid tumors compared with healing scar and quiescent fibrous tissue. Low levels of phosphorylated SMAD1/5/8 were detected in only a minority of cases. Transforming growth factor–β receptor type 1 and androgen receptor were expressed in both desmoid tumors and scar, but not in fibrous tissue. Estrogen receptor–β was present in all cases studied. Transforming growth factor–β signaling appears to be activated in desmoid-type fibromatosis and phosphorylated SMAD2/3 and COX2 immunoreactivity might be of diagnostic utility in these tumors. Given the frequency of androgen receptor, estrogen receptor–β and COX2 co-expression in desmoid tumors, further assessment of the efficacy of combination pharmacotherapy using hormonal agonists/antagonists together with COX2 inhibitors should be considered.
Desmoid-type fibromatosis (desmoid tumor) is a fibroblastic-myofibroblastic tumor with a rather unpredictable clinical course.[1-3] Whereas some tumors are slow growing and may regress spontaneously, others grow rapidly and recur repeatedly after complete surgical resection. There are few clinicopathological predictors of disease recurrence. Even the prognostic significance of positive surgical resection margins is controversial.[1, 2, 4, 5] Despite lacking metastatic potential, most desmoids are locally aggressive and cause significant morbidity. Other than wide surgical re-excision and radiation therapy, there is no standard treatment for recurrent disease, a testimony to the limited efficacy of most adjuvant treatment regimens.[2-6]
Expression of estrogen receptor–β and COX2 has been documented in many cases of deep fibromatosis,[7-11] but response rates to single agent anti-estrogen therapies and non-steroidal anti-inflammatory drugs are limited and unpredictable. Co–expression of target proteins might account for responses to combined steroidal and non-steroidal anti-inflammatory drugs in some patients.[12, 13] Since interactions between cell signaling pathways are complex and often redundant, aberrancy in one signaling pathway is often corrected through compensatory (de)regulation of another pathway. This might not only explain why efficacy of single pharmaceutical agents is limited, but also suggests that other pathways might represent additional targets for inhibitor therapy in this disease.[1, 14]
Myofibroblast proliferation, differentiation, migration, angiogenesis and extracellular matrix synthesis during wound healing or fibrosis are processes driven by a variety of paracrine signals, including transforming growth factor–β (TGFβ).[15-18] It is well known that the majority of desmoid tumors harbor mutations of either the β-catenin (CTNNB1) or adenomatous polyposis coli genes of the Wnt signaling pathway, which are thought to deregulate proliferation and invasiveness of fibroblasts.[6, 14, 19] That TGFβ stimulates CTNNB1-mediated transcription suggests that TGFβ might be an important factor in desmoid tumorigenesis.[20-23] Active TGFβ signaling has been reported in desmoid myofibroblasts,[21, 24, 25] and expression of TGFβ-related cytokines such as bone morphogenetic proteins (BMP) has also been described in desmoid tumors.
Current advances in targeted molecular therapy offer new opportunities to treat fibromatosis patients. A better understanding of the signaling pathways active in fibromatosis is necessary to determine the theoretical potential for inhibitor therapy in this debilitating disease. Selected signaling molecules, particularly those of the TGFβ signaling family, were evaluated in 27 cases of deep fibromatosis and compared with healing scar (a reactive myofibroblastic proliferation) and non-neoplastic fibrous tissue. Co-expression patterns were explored using correlational analyses. Although TGFR1 expression levels are similar in desmoid tumors and scars, phosphorylated SMAD2/3 and COX2 levels are significantly greater in desmoid tumors. In contrast, phosphorylated SMAD1/5/8 was detectable in only a minority of desmoids and scar tissues. While estrogen receptor–β expression is present in all desmoids and scar, levels of androgen receptor were variable in these tissues. In addition to being useful diagnostic markers, some of these proteins might be targeted by specific molecular therapeutic agents, offering new treatment regimens that might be potentially individualized to the expression pattern of an individual tumor.
Materials and Methods
The surgical pathology archives were searched for cases diagnosed as “fibromatosis” or “desmoid tumor” between 1990 and 2005; superficial and intra-abdominal fibromatoses were excluded. A series of healing cutaneous scars obtained from re-excision specimens for non-desmoid-related diseases (n = 14) and non-neoplastic fibrous tissue (n = 6) from reduction mammoplasty or panniculectomy specimens were collected for comparison purposes (Table 1). Clinicopathological information (patient demographic data, tumor location, tumor size, surgical resection margin status, adjuvant therapy administered and recurrence-free and overall survival intervals) was collected via review of electronic medical records. The study protocol conforms to the provisions of the Declaration of Helsinki (1995) and was approved by the Institutional Review Board at Vanderbilt University; the requirement for informed consent was waived by this committee.
Table 1. Sources of healing scars and fibrous tissues
Reason for surgery
Interval since prior surgery/trauma
F, female; M, male; NA, not applicable.
Re-excision of melanoma
Re-excision of melanoma
Re-excision of melanoma
Re-excision of melanoma
Re-excision of trichilemmal carcinoma
Revision of scar from trauma surgery
Revision of scar from traumatic injury
Revision of ventral hernia
Reconstructive surgery after burn injury
Repeat cesarean section
Revision of ventral hernia
Revision of umbilical hernia
Twenty-seven cases of sporadic desmoid-type fibromatosis were available for analysis; none of the study patients had documented evidence of Gardner syndrome or familial adenomatosis polyposis (Table 2). The study group was composed of 14 women (median age, 26 years; range, 4–73 years) and 13 men (median age, 23 years; range, 1–65 years). Sites of involvement included the extremities (n = 7), head/neck region (n = 6), back/thoracic wall (n = 4), shoulder (n = 4) and abdominal walls (n = 3); anatomical location was not specified for three cases. Tumor size ranged from 1.0 cm to 15.0 cm in the greatest dimension (median, 5.8 cm). Gross tumor volume was calculable for 23 cases (median, 56.9 cm3; range, 0.4–1320.0 cm3). The status of the surgical resection margin was known in 22 cases. In 12 cases, the margins were either grossly or microscopically positive (American Joint Committee on Cancer category R1 or R2). Three patients were treated with adjuvant radiation, one was treated with adjuvant chemotherapy, and one received combination chemoradiotherapy.
Table 2. Clinicopathological features of desmoid-type fibromatosis cases
Tumor size (cm)
Surgical resection margins
Local recurrence (months)
Follow up (years)
†Two cycles of vincristine, actinomycin-D and cyclophosphamide; nine cycles of ifosfamide, carboplatin and etoposide; and vinblastine and methotrexate for 6 months. ‡Four cycles of vincristine, actinomycin-D and cyclophosphamide. §Patient developed disease recurrence with associated radiation-induced fibrosarcoma 20 years after initial resection. ¶Only one tumor dimension was recorded. ††Patient died of cholangiocarcinoma. ANED, alive without evidence of disease; AWD, alive with recurrent disease; DOC, died of other unrelated causes; F, female; M, male; NA, data not available; XRT, radiation therapy (45–60 Gy).
Follow-up information was available for 20 patients. The median follow-up interval was 94 months (range, 10–321 months). During this interval, 10 patients suffered a local recurrence (50%) with a median time to recurrence of 23 months (range, 5–38 months). Four of these patients had multiple local recurrences. Overall, median disease-free survival was 38 months. Six of 11 patients with positive surgical resection margins and four of 10 patients with negative surgical resection margins suffered local recurrences (Fisher's exact test, P =0.67). No patient died from complications of desmoid tumor during the duration of the follow-up interval.
Tissue microarray and immunohistochemistry (IHC)
The original hematoxylin and eosin-stained slides were reviewed and diagnostic tissue was marked for construction of a tissue microarray (TMA) using a manual arrayer (Beecher Instruments, Sun Prairie, WI, USA). Two tissue cores (each 1.5 mm) of representative areas from each of the selected formalin-fixed, paraffin-embedded tissue blocks were used for the array. Non-neoplastic fibrous tissue samples were not included on the TMA; whole tissue sections were stained instead. The antisera and IHC protocols used are presented in Table 3. Antigen retrieval was performed for 25 min in a preheated steamer followed by 15 min cool down to room temperature. Sections were then incubated with primary antibodies for 1 h at room temperature unless otherwise noted. Appropriate positive and negative control sections were included for each antibody and assay run, respectively. Specific membranous staining was suboptimal using antibodies against TGFβ receptor type 2, BMP receptor type 1A and BMP receptor type 2. Therefore, the activated and phosphorylated isoforms of the downstream signal transduction molecules SMAD2/3 (for TGFβ receptor) and SMAD1/5/8 (for BMP receptor) were investigated as indicators of active signaling by these respective pathways.[15, 20, 28]
Table 3. Antibodies and immunohistochemistry protocols used
†Santa Cruz Biotechnology, Santa Cruz, CA, USA; Dako North America Inc., Carpinteria, CA, USA; Invitrogen Corporation, Carlsbad, CA, USA; BioGenex Laboratories Inc., Fremont, CA, USA; Oxford Biomedical Research, Upper Heyford, UK; Abgent Inc., San Diego, CA, USA; Cell Signaling Technology, Danvers, MA, USA. ‡1X Citra buffer, pH 6.0 (Dako North America Inc., Carpinteria, CA, USA). §LSAB2 (Dako North America Inc.); incubation overnight at 4°C. ¶Dako Target Retrieval Solution (Dako North America Inc.). ††EnVision+ (Dako North America Inc.). ‡‡Proteinase K, prediluted (Dako North America Inc.) incubated for 5 min at room temperature. §§Vectastain Elite ABC Kit (Goat IgG), Vector Laboratories, Burlingame, CA, USA. AR, androgen receptor; CTNNB1, β–catenin; ERα, estrogen receptor–α; ERβ, estrogen receptor–β; PR, progesterone receptor; p–SMAD1/5/8, phosphorylated-SMAD1/5/8; p–SMAD2/3, phosphorylated-SMAD2/3; TGFR1, transforming growth factor receptor–β type 1.
The IHC stains were independently reviewed by two of the authors (JHF and JMMC) and the expression of each marker was scored in a semiquantitative manner. As most lesional cells stained uniformly within each TMA core, staining intensity was recorded as either negative (0), weaker than the corresponding positive control (1+), similar to the corresponding positive control (2+) or stronger than the corresponding positive control (3+). For the steroid receptors, the number of positive nuclei were enumerated and scored as follows: 0, ≤5%; 1+, 5–33%; 2+, 34–66%; and 3+, ≥67%.
Median IHC scores were compared using the Kruskal–Wallis test with a conservative Bonferroni correction applied; α = 0.005 was considered statistically significant for this analysis. For markers showing statistical differences among the three diagnostic groups, post-hoc analyses were performed using Dunn's multiple comparison test. The IHC data were compared with clinicopathological findings using standard bivariate methods. Kaplan–Meier recurrence-free survival curves were compared using the log rank test. All statistical analyses were performed with GraphPad Prism v5.02 (GraphPad Software, Inc., La Jolla, CA, USA).
Activated myofibroblasts in desmoid-type fibromatosis and reactive scar tissue differ from quiescent fibroblasts by their relatively more abundant, often somewhat basophilic cytoplasm and open nuclei with dispersed chromatin and inconspicuous nucleoli (Fig. 1). While the growth pattern of desmoid-type fibromatosis is that of broad, sweeping fascicles often encompassing entire medium-power fields, that of scar tissue is generally more disordered and consists of smaller, irregular fascicles. In addition, evidence of prior trauma such as foci of neovascularization and an associated mononuclear cell infiltrate or hemosiderin-laden macrophages are sometimes noted in reparative fibrous proliferations. However, these findings might be subtle, making the distinction between neoplastic and reactive myofibroblastic proliferations difficult. Non-neoplastic or quiescent fibrous tissue was composed of sparse, inconspicuous fibroblastic cells with minimal cytoplasm and small slender inactive-appearing nuclei embedded within dense collagen bundles.
Representative IHC stains and the distribution of IHC intensity scores for each diagnostic category are presented in Figures 2 and 3, respectively. Nuclear CTNNB1 was detected in 70% of desmoid tumors and 14% of hypertrophic scars, but was not seen in quiescent fibrous tissue. Nuclear CTNNB1 immunoreactivity was significantly higher in desmoid tumors compared with both scar tissue and fibrous tissue (Kruskal–Wallis test, P = 0.0003; Dunn's post-tests, P < 0.01). Transforming growth factor–β (TGFβ) receptor type 1 (TGFR1) immunoreactivity was similar in both desmoid tumors and scar, but was not detected in fibrous tissue. However, activated (phosphorylated) SMAD2/3 (p–SMAD2/3) was detected in 96% of desmoid tumors (74% with 2+ or 3+ immunoreactivity) compared with 29% of scar tissue samples (all with 1+ immunoreactivity) and 0% of the fibrous tissue samples. The differences in immunoreactivity were statistically significant between desmoids and both scar and fibrous tissue (Kruskal–Wallis test, P < 0.0001; Dunn's post-tests, P < 0.001). Low levels (1+) of phosphorylated SMAD1/5/8 were detectable in only a minority of desmoids (17%) and scar tissues (14%); these differences were not statistically significant.
COX2 was detectable in 83% of desmoid tumors (52% with 2+ or 3+ immunoreactivity) compared with low levels (1+ only) seen in 21% of scar tissues; COX2 was undetectable in quiescent fibrous tissue (Kruskal–Wallis test, P < 0.0001). Differences in COX2 levels between desmoid and scar (Dunn's post-test, P < 0.001) and desmoid and fibrous tissue (Dunn's post-test, P = 0.001–0.01) both reached statistical significance. Variable levels of androgen receptor immunoreactivity were present in desmoids and scar, but not in fibrous tissue. Whereas all tissue types showed consistent and strong expression of estrogen receptor–β, none of the specimens examined showed specific nuclear staining for estrogen receptor–α or progesterone receptor (data not shown).
The differential diagnosis between desmoid-type fibromatosis and reactive myofibroblastic proliferations such as healing scar can be quite difficult. In this regard, demonstration of nuclear CTNNB1 by immunohistochemistry can be helpful. However, the prevalence of nuclear CTNNB1 in desmoid tumors was only 70% in this series, with a false positive rate of 14%. The sensitivity of p–SMAD2/3 for desmoid tumors is significantly greater than that of CTNNB1 (Table 4; McNemar's test, P =0.016), whereas the specificities of these two markers are not statistically significant (McNemar's test, P =0.63). Although the sensitivity and specificity of COX2 for desmoid tumor are not significantly different than those of CTNNB1 (McNemar's test, P =0.45 and P =1.0, respectively), other parameters of diagnostic accuracy compare favorably with CTNNB1 (Table 4).
Table 4. Performance characteristics of CTNNB1, p–SMAD2/3 and COX2 as diagnostic tests for desmoid-type fibromatosis
Correlation analysis was performed using Spearman rank tests (Table 5). As would be expected if TGFR1 were activated, levels of p–SMAD2/3 and its upstream activator TGFR1 were strongly correlated. TGFR1 and p–SMAD2/3 were also both strongly correlated with COX2 expression. Androgen receptor expression was found to correlate with the presence of p–SMAD2/3 and COX2. Nuclear CTNNB1 showed no correlation with any of the other markers studied. None of the IHC markers correlated with age, greatest tumor dimension or total tumor volume using Spearman rank tests. No IHC marker showed an association with gender or anatomic location using Mann–Whitney or Kruskal–Wallis tests. There were no associations between any IHC marker and recurrence-free survival using log rank tests of Kaplan–Meier survival curves.
Table 5. Spearman correlation analysis between immunohistochemical markers and selected clinicopathological parameters of desmoid tumorsa
Upper right panels, Spearman correlation coefficients; lower left panels, P values. AR, androgen receptor; CTNNB1, β–catenin; GD, greatest tumor dimension; p–SMAD2/3, phosphorylated SMAD2/3; TGFBR1, transforming growth factor–β type 1; TV, total tumor volume.
3.2 x 10−9
The expression of a variety of signaling proteins and activated downstream signal transduction molecules was compared in desmoid tumors, reactive myofibroblasts in healing scars and quiescent fibroblasts in non-proliferative fibrous tissue to identify signal transduction pathways deregulated in desmoid-type fibromatosis. In addition, co–expression patterns were assessed by correlation analysis. Differential patterns of expression of CTNNB1, p–SMAD2/3 and COX2 were observed between desmoid-type fibromatosis, scar and fibrous tissue. Thus, these markers might be useful in distinguishing between desmoid-type fibromatosis and healing scar. The results also suggest that signaling pathways other than the Wnt pathway are abnormally activated in desmoid tumors.
The constituent cell in both desmoid tumor and healing scar tissue is the activated myofibroblast. Therefore, it can be difficult to determine whether a myofibroblastic proliferation is neoplastic or reactive, particularly when evaluating the surgical resection margins of re-excised desmoid tumors. Subtle histological features that are suggestive of a reparative response include relatively small, disordered, irregular fascicles of myofibroblasts, sometimes associated with chronic inflammation, hemosiderin-laden macrophages and areas of neovascularization.
Nuclear CTNNB1 was seen more often in desmoid tumors compared with scar tissue, confirming prior observations.[30, 31] Although nuclear accumulation of CTNNB1 has not been reported in the few examples of non-hypertrophic scars previously studied, it has been demonstrated in myofibroblasts during the proliferative phase of wound healing, in hypertrophic scars and in rare cases of reactive-type fasciitis, suggesting that this signaling pathway is important in the physiological regulation of fibroblast proliferation.[14, 19, 23, 30, 32] Given the occasional presence of nuclear CTNNB1 in non-neoplastic and reactive myofibroblasts, documentation of strong immunoreactivity for p–SMAD2/3 and/or COX2 might be of clinical utility in the differential diagnosis between desmoid-type fibromatosis and these reactive myofibroblastic proliferations in the appropriate clinical setting.
TGFR1 is a serine/threonine kinase receptor that in response to cytokine signaling phosphorylates SMAD2 and SMAD3, which then translocate to the nucleus and regulate the transcription of a number of genes involved in a variety of mesenchymal cell processes, including proliferation, myofibroblastic differentiation, angiogenesis and extracellular matrigenesis.[15, 16, 18, 21, 24, 25, 28] Increased TGFβ activity has been implicated in pathological myofibroblastic proliferations such as keloid formation[33, 34] superficial fibromatoses[35-37] and hypertrophic scar. The TGFβ signaling family is also activated in desmoid tumors.[21, 39, 40] In the present study, TGFR1 staining intensity was similar in both desmoid-type fibromatosis and scar tissue, but it was not detected in quiescent fibrous tissue. In contrast, increased levels of the activated isoform of the TGFR1 downstream signaling molecule p–SMAD2/3 were detected in most desmoid tumors, whereas scar tissue showed low levels in less than one-third of samples. Amini Nik and colleagues previously demonstrated increased levels of p–SMAD2/3 in desmoid tumor cells compared with normal fibrous tissue. Thus, although TGFR1 is present in both desmoid tumor and healing scar tissue, it appears to be activated preferentially (and perhaps aberrantly) in the neoplastic myofibroblasts of desmoid tumors. Despite the reported presence of BMP in desmoid tumors, the BMP receptor signaling pathway does not appear to be activated in desmoid-type fibromatosis or scar tissue, as phosphorylated SMAD1/5/8 was detected in <15% of study cases.
CTNNB1-mediated transcriptional activity can be stimulated by Wnt-independent pathways such as TGFβ signaling.[20, 32] Activation of the TGFβ-receptor pathway has been shown to increase CTNNB1 activity in myofibroblasts during the proliferative phase of wound healing and in desmoid-type fibromatosis in vitro.[21, 22] As both TGFR and atypical Wnt signaling are activated in desmoid tumors, perhaps inhibition of TGFR signaling might be of therapeutic value in desmoid-type fibromatoses despite the absence of correlations between CTNNB1 and TGFR1 or p–SMAD2/3 immunoreactivity seen in the present study.[20-22] In addition to CTNNB1 and p–SMAD2/3, increased levels of COX2 were also more often observed in desmoid tumors than in scar tissues, similar to previous findings.[10, 27] COX2 expression is induced by numerous cytokines and growth factors and is quite common in both inflammatory and neoplastic conditions. Like TGFβ, COX2-mediated prostaglandin signaling might also stimulate CTNNB1-dependent transcription and cell proliferation,[42-44] as inhibition of COX2 reduces proliferation of desmoid tumor cells in vitro.  Correlation analysis performed in the present study demonstrated associations between TGFR1, p–SMAD2/3 and COX2 expression, suggesting that TGFβ activity increases COX2 expression in desmoid tumors, as in other cell types.[41, 46]
Variable levels of androgen receptor expression were present in desmoids and scars, which were significantly higher than in fibrous tissue. The reported prevalence of androgen receptor expression in fibromatosis ranges from <10% to 100%, depending on the methodology used.[11, 47, 48] As shown previously, estrogen receptor–α and progesterone receptor are not detectable in fibromatosis, while expression of estrogen receptor–β is ubiquitous in fibrous and myofibroblastic tissues.[7-9, 11] Similar to previous studies, no associations were observed between sex steroid receptor expression and patient age or sex. There also appears to be no correlation between sex steroid receptor status and response to hormonal therapy.
The androgen receptor modulates Wnt signaling, possibly by binding CTNNB1-related transcriptional complexes in a cell type-specific manner.[49-51] Androgen receptor also appears to inhibit COX2 expression and modulate TGFβ signaling independent of androgen ligands.[52, 53] In the present study, androgen receptor levels were found to correlate with both p–SMAD2/3 and COX2 in desmoid-type fibromatosis. As the androgen receptor can regulate Wnt and TGFβ signaling as well as COX2 expression, perhaps androgen manipulation therapy would be beneficial to fibromatosis patients. Supporting this suggestion is recent evidence from Hong and colleagues, who showed that testosterone increased cell proliferation and CTNNB1 levels in cultured desmoid tumor cells. The frequent co–expression of androgen receptor, COX2 and estrogen receptor–β suggests that specific pharmacotherapeutic agents used in combination might show superior efficacy than single agents alone, particularly when tailored to the expression profile of an individual tumor.
Aberrations of the Wnt signaling pathway are well documented in desmoid-type fibromatosis. In the present study, we showed that TGFβ activity and COX2 expression are also increased in desmoid tumors compared with non-neoplastic myofibroblasts and might be useful ancillary diagnostic markers for desmoid-type fibromatosis. Other experimental evidence of positive interactions between TGFβ and Wnt signaling pathways suggest that the former might be an important factor in the pathogenesis of desmoid tumors, as in other tumor types.[20-22] These interactions might provide additional molecular targets for systemic therapy of intractable desmoid tumors. As signaling molecules such as estrogen receptor–β, androgen receptor and COX2 are frequently co-expressed in desmoid-type fibromatosis, specific pharmacotherapeutic agents used in combination might show superior efficacy than single agents alone. Consideration should be given to combination therapy based on the specific expression profile of an individual tumor, because the complex cross-signaling between these pathways intimate that combined therapy might result in more robust treatment responses.
The present study was supported in part by the National Cancer Institute (5P30 CA068485-15). The authors thank Anthony Frazier and Melissa Downing of the Translational Pathology Shared Resource for technical assistance.