Sarcoma Service, Department of Surgery, Scientific Institute for Research, Hospitalization, and Health Care (IRCCS) Foundation, National Cancer Institute, Milan, Italy
Corresponding author: Alessandro Gronchi, MD, Sarcoma Service, Department of Surgery, Fondazione IRCCS Istituto Nazionale dei Tumori, via G. Venezian 1, 20133 Milan, Italy; Fax: (011) 39-02-2390-2404; firstname.lastname@example.org
We thank Pauline Weijers-Koster for expert technical assistance.
A role for the serine to phenylalanine substitution at codon 45 (the S45F mutation) in the catenin (cadherin-associated protein) β-1 (CTNNB1) gene as a molecular predictor of local recurrence in patients with primary, sporadic desmoid tumor (DT) has been reported. To confirm the previous data, the authors evaluated the correlation between CTNNB1 mutation type and local recurrence in this multi-institutional, retrospective study.
Patients with primary, sporadic DT who underwent macroscopic complete surgical resection were included. Recurrence-free survival (RFS) analyses were conducted using the Kaplan-Meier method and log-rank tests to compare strata.
In total, 179 patients were identified, including 65% females and 35% males (median age, 39 years; median tumor size, 7 cm). Most DTs were located in the abdominal/chest wall (42%) followed by extra-abdominal sites (40%) and intra-abdominal sites (18%). All patients underwent either R0 resection (62%) or R1 resection (38%), and most underwent surgery alone (80%). The tyrosine to alanine substitution at codon 41 (T41A) was the most frequent mutation (45%), but the S45F mutation was more prevalent in extra-abdominal DTs compared with other sites (P < .001). At a median follow-up of 50 months, 86% of patients remained alive without disease. The estimated 3-year and 5-year RFS rates were 0.49 and 0.45, respectively, for patients who had tumors with the S45F mutation; 0.91 and 0.91, respectively, for patients who had wild-type tumors; and 0.70 and 0.66, respectively, for all others (P < .001). A similar trend was observed for patients who underwent surgery alone (P < .001). On multivariable analysis, mutation remained the only factor that was prognostic for local recurrence.
Desmoid tumors (DTs) are rare mesenchymal lesions with an annual incidence of 5 cases per million. Lacking metastatic capacity, DTs are very infiltrative and possess a high risk of local recurrence despite adequate surgical resection with negative margins.[2, 3] Repeated excisions, if necessary, often result in significant treatment-related morbidity. Most sporadic DTs (85%) harbor specific mutations in the gene encoding for β-catenin located on the short arm of chromosome 3 (catenin [cadherin-associated protein] β-1 [CTNNB1]); the remainder do not harbor any mutation in CTNNB1 (wild-type [WT] tumors). It is noteworthy that 3 different point mutations in 2 different codons (41 and 45) have been identified more frequently in mutated samples: ACC to GCC in codon 41 (replacement of threonine by alanine [T41A]); TCT to TTT in codon 45 (replacement of serine by phenylalanine [S45F]), and TCT to CCT in codon 45 (replacement of serine with proline [S45P]); these 2 codons are targets for phosphorylation by glycogen synthase kinase 3-β (GSK-3β) and casein kinase-1 (CK1), respectively.
Among solid tumors, it is unusual to demonstrate only 3 specific mutations in CTNNB1; most other neoplasms that harbor exon 3 CTNNB1 mutations have wider mutational variability at multiple critical codons,[7, 8] suggesting that these specific CTNNB1 mutations may be critical in the development of desmoid-type fibromatosis; and the type of CTNNB1 gene mutation may affect the levels β-catenin signaling, thereby having an impact on desmoid formation and/or recurrence. The importance of these mutations in DT pathogenesis remains conjectural, even if a crucial role is hypothesized because of their high prevalence. Different outcomes based on CTNNB1 mutational status have been demonstrated retrospectively in a surgically treated, single-institution patient experience; patients who had DTs with the S45F mutation had worse recurrence-free survival (RFS) compared with those who had either tumors with the T41A mutation or WT tumors. A similar study failed to demonstrate the same significance but indicated that patients who had mutated DTs had poorer RFS compared with those who had WT tumors; however, a trend toward worse outcomes for S45F-mutated DTs was demonstrated. In a recent study, investigators reported that the CTNNB1 mutation spectra concerned mainly 3 point mutations in 2 codons (41 and 45) even if, in 3% of patients, other point mutations in exon 3 of the CTNNB1 gene were identified.
Although surgery is considered the mainstay of treatment, an emerging, more conservative strategy (including wait and see approaches and miscellaneous medical therapeutic options) has recently been proposed, suggesting the heterogeneous behavior of DTs and the difficulties in standardizing treatments. Consequently, we undertook the current retrospective, multi-institutional study to test whether β-catenin may be a potential prognostic marker of aggressiveness in DTs, thereby informing upfront therapeutic options.
MATERIALS AND METHODS
Inclusion Criteria and Clinical Database
Patients with a diagnosis of primary, sporadic DT who underwent surgery at 1 of 4 sarcoma referral centers (IRCCS Foundation National Cancer Institute, Milan, Italy; The University of Texas MD Anderson Cancer Center [UTMDACC], Houston, Tex; Gustave Roussy Institute, Paris, France; and Leiden Medical Center, Leiden, the Netherlands) between April 1998 and March 2011 were included in this study. We note that none of the UTMDACC patients who were included in the current study were part of the previously published patient cohort. Informed consent was obtained from each patient for this study.
Patients with familial adenomatous polyposis-related DT, recurrent disease at presentation, or incomplete surgical resection were excluded. Data retrieved included sex; anatomic site; tumor size; date of surgery; margin status, including microscopic positive (R1) and negative (R0) margins; results of CTNNB1 mutational analysis; type of other treatments eventually received before and/or after surgery; date of recurrence; treatment of recurrence; date of last follow-up; and status at last follow-up. Anatomic sites were classified as extra-abdominal (including extremity and head/neck), abdominal/thoracic wall, and intra-abdominal. Tumor size was defined as the greatest DT dimension in the surgical specimen reported by the original pathologists. Surgical excisions were considered macroscopically complete in the absence of gross residual disease. All macroscopically complete resections were classified according to the closest surgical margin, which was microscopically categorized as positive (R1, tumor within 1 mm from the inked surface) or negative (R0, absence of tumor within 1 mm from the inked surface).
Nonsurgical treatments were administered in the primary or recurrent phase of the disease on an individualized basis or as part of clinical trials. They included radiotherapy, chemotherapy (frequently methotrexate and vinorelbine), medical therapy/hormone agents (tamoxifen, toremifene), and nonsteroidal anti-inflammatory drugs (cyclooxygenase-2 inhibitors).
CTNNB1 Mutational Analysis
Mutational analyses were performed on formalin-fixed, paraffin-embedded tissues at each participating institution using the same standardized technique. Microscopic dissection of 7 μm methylene blue-stained tissue sections (with >80% tumor cells) allowed the precise separation of neoplastic and normal tissues. Genomic DNA was extracted using the Qiamp DNA Kit (Qiagen, Chatsworth, Calif) according to the manufacturer's instructions. CTNNB1 mutational status was analyzed using polymerase chain reaction (PCR) on 100 ng of genomic DNA. Exon 3 of the CTNNB1 gene was amplified using the specific primers ex3F (forward: 5′-ATGGAACCAGACAGAAAAGC-3′) and ex3R (reverse: 5′-GCTACTTGTTCTTGAGTGAAG-3′) under the following PCR conditions: 30 seconds at 95°C, 30 seconds at 56°C, and 30 seconds at 72°C for 35 cycles. The PCR products were purified by using ExoSAP (Affymetrix, Santa Clara, Calif) and were sequenced on a 3500 Dx Genetic Analyzer (Applied Biosystems, Foster City, Calif). Each sequence was performed at least twice starting from an independent amplification reaction and was evaluated using ChromasPro software (Technelysium Pty Ltd., South Brisbane, Queensland, Australia). Three different mutations were identified at codon 3 of the gene encoding for β-catenin, CTNNB1 (T41A, S45F, and S45P). Patients without a specific mutation had their tumors classified as WT. Because the number of patients with the S45P mutation was small (n = 12), for the purposes of statistical analysis, the T41A and S45P mutations were grouped together and termed “other.”
The prognostic analyses focused mainly on the effects of β-catenin mutation. The study endpoint was RFS; and the event time was computed from the date of surgery to the date of relapse or death, whichever occurred first, or was censored at the date of last follow-up assessment in event-free patients. RFS curves were estimated using the Kaplan-Meier method and were compared statistically using the log-rank test. Multivariable Cox model analyses also were conducted incorporating β-catenin mutation along with the covariates sex, tumor site, tumor size, and margin status. All covariates except tumor size were modeled as categorical variables using dummy variables, whereas tumor size was modeled as continuous variable using a 3-knot restricted cubic spline to obtain flexible fits and to allow for its prognostic effect as a variable that was not equivalent in every part of the range. Penalized maximum likelihood estimation methods were applied to control for possible over-fit bias; the latter was likely to occur in the multivariable analysis because of the small sample and the low number of events compared with the number of model estimates, with substantial censoring of survival times. Two-sided P values below the conventional .05 threshold were considered statistically significant. All statistical analyses were performed using the software packages SAS (SAS Institute Inc., Cary, NC) and R (R Foundation for Statistical Computing, Vienna, Austria; available at: http://www.r-project.org; accessed February 2013) for penalized maximum likelihood estimations.
Patient demographics and disease characteristics for the entire population and for specific mutational subgroups are displayed in Table 1. In total, 179 patients were included, including 65% females and 35% males. The median age was 39 years (interquartile range, 31-52 years). The majority of DTs were located in the abdominal/chest wall (42%) followed by extra-abdominal sites (40%) and intra-abdominal sites (18%). The median DT size was 7 cm (interquartile range, 4-10 cm).
Table 1. Patient and Tumor Characteristics
No. of Patients (%)
Total, n = 179
S45F, n = 39
WT, n = 48
Others, n = 92
Abbreviations: CT, chemotherapy; NSAIDs, nonsteroidal anti-inflammatory drugs; RT, radiotherapy; S45F, replacement of serine by phenylalanine in the catenin (cadherin-associated protein) beta 1 (CTNNB1) gene; W&S, wait and see; WT, wild type.
Extra-abdominal sites included extremity and head/neck.
Treatment could be administered as combination therapy (ie, surgery and RT) or in a sequential fashion if the first treatment was not effective (ie, hormone therapy followed by NSAIDs).
All patients underwent complete surgical excision (R0/R1), and 62% of patients underwent R0 resection. The vast majority underwent surgery alone, and 20% received other treatments preoperatively or postoperatively. The most common nonsurgical treatment was radiotherapy (9%). The T41A mutation was the most frequent (45%), and the S45F mutation was significantly more prevalent in extra-abdominal sites compared with other locations (P < .001). Moreover, the percentage of S45F-mutated DTs increased with increasing tumor size (P = .007). A correlation between mutation and margin status was not observed (P = .871).
Patients with recurrence (n = 48) commonly underwent surgery (46%), and combination or sequential treatments were received by 37% of patients. At a median follow-up of 50 months (interquartile range, 28-84 months), 86% of patients were alive without disease; only 2 patients died during follow-up, both because of nondesmoid causes; and, among the patients who developed recurrence, more than half (52%) were alive without any disease.
The estimated 3-year and 5-year RFS rates were 0.496 (95% confidence interval [CI], 0.354-0.695) and 0.458 (95% CI, 0.316-0.664), respectively, for patients with the 45F mutation; 0.911 (95% CI, 0.831-0.998) and 0.91 (95% CI, 0.831-0.998), respectively, for those with WT tumors; and 0.701 (95% CI, 0.606-0.811) and 0.661 (95% CI, 0.559-0.781), respectively, for those who had the mutations designated “others” (P < .001) (Fig. 1A). When the “others” group was analyzed separately (45P and 41), the estimated 3-year and 5-year RFS rates could be superimposed. A similar trend was observed when only the patients who underwent surgery alone were considered (P < .001) (Fig. 1B). In particular, for this group, the estimated 3-year and 5-year RFS rates were 0.441 (95% CI, 0.280-0.692) and 0.385 (95% CI, 0.229-0.650), respectively, for patients with the 45F mutation; 0.97 (95% CI, 0.924-1) and 0.97 (95% CI, 0.924-1), respectively, for those with WT tumors; and 0.702 (95% CI, 0.602-0.819) and 0.64 (95% CI, 0.532-0.771), respectively, for those who had the mutations designated “others.”
In contrast, there was no difference in RFS for patients who underwent surgery and received other treatments (P = .564) (Fig. 1C). The estimated RFS rate at both 3 years and 5 years was better for the S45F group (0.606; 95% CI, 0.368-0.998) and the “others” group (0.802; 95% CI, 0.587-1.00) but was worse for the WT group (0.714; 95% CI, 0.447-1.00).
The median time to relapse for those who developed recurrent disease was similar among patients with different mutations. In particular, the median time to relapse was 13 months for the entire population and 13 months and 14 months for those with 41 and 45F mutations, respectively. Sex, margin status, and tumor size were not identified as significant prognostic markers of recurrence (data not shown). A significant trend for site of origin as a prognostic marker was identified: patients who had extra-abdominal desmoids had poorer outcomes (P = .009) (Fig. 1D). On multivariable analysis, CTNNB1 mutation remained the only significant prognostic factor for relapse (Table 2).
Table 2. Multivariable Analysis
Abbreviations: CI, confidence interval; HR, hazard ratio; S45F, replacement of serine by phenylalanine in the catenin (cadherin-associated protein) beta 1 (CTNNB1) gene; WT, wild type.
S45F vs other
Other vs WT
Sex: Male vs female
Margin status: R1 vs R0
Tumor size: 10 cm vs 4 cm
Extra-abdominal vs abdominal/chest wall
Intra-abdominal vs abdominal/chest wall
In this series of 179 patients with DTs who underwent surgery at 4 referral centers over a 10-year time span, CTNNB1 gene mutation was the only significant prognostic factor for relapse. Patients who had DTs harboring the S45F CTNNB1 mutation had poorer RFS compared with those who had DTs harboring T41A/S45P mutations or the WT CTNNB1 gene (Fig. 1). These data reproduce the previously reported UTMDACC results.
It is noteworthy that a difference in RFS between each mutation group was evident when we compared our multicenter series with the previous UTMDACC single-center experience, with outcomes that generally were worse in the latter study (RFS rate: 23% vs 46%, respectively, for those with the S45F mutation; 57% vs 66%, respectively, for those with the T41A mutation; and 65% vs 91%, respectively, for those with WT tumors). The better trend in the current multicenter series was probably related to the exclusion of patients with recurrent disease, who are known to have a higher risk of further recurrences compared with patients who have primary disease. Moreover, we observed a specific correlation between the type of mutation and the site of DT origin (a predominance of the S45F mutation was observed in extremity DTs), perhaps explaining why patients who have DTs of the extremity usually have a worse outcome compared with patients who have DTs of other anatomic sites.
The impact of specific mutations on the pathogenesis of DTs and their predisposing role in local relapse remains uncertain. Only 3 different point mutations in 2 different codons (41 and 45) have been identified to date in DTs. In particular, specific amino N terminal residues (serine 45 [Ser45] by CK1α, and threonine 41 [Thr41], Ser37, and Ser33 by GSK-3β, in that sequential order) represent the phosphorylation targets that mediate the degradation of β-catenin, a crucial step in maintaining low β-catenin cytoplasmic levels. To serve as an explanation for the predisposition toward local recurrence of specific mutations, the role of CK1α in the maintenance of β-catenin equilibrium ultimately may prove to be relevant. The identification of 2 different types of point mutations at position 45 and the correlation between 45F and aggressive behavior suggests, but does not prove, a potential critical role of phosphorylation mediated by CK1α. In this context, phosphorylation in the position 45 amino acid residue by CK1α comprises the first step in β-catenin ubiquitination, a process that apparently cannot begin if a mutation in this specific residue is present.
Another study that examined the potential role of β-catenin as a prognostic marker evaluated 101 surgically excised primary and recurrent extra-abdominal DT samples obtained through Conticanet (the Connective Tissue Cancer Network, which includes France, Belgium, and Switzerland) and reported that, with at least 5 years of follow-up available for all patients, the CTNNB1 mutation rate was 83%. In that study, 5-year RFS was significantly shorter in patients who had mutated DTs compared with those who had WT CTNNB1; a nonsignificant trend toward enhanced recurrence in patients with DTs was reported for those with the S45F mutation versus those with the T41A mutation. Several reasons may account for this lack of statistical significance, including the limited size of the patient sample, which necessitated collection from several institutions; the inclusion of both primary and recurrent tumors; and the exclusion of intra-abdominal DTs. Moreover, that study did not specify whether or not patients received nonsurgical treatments, a potentially confounding factor, as illustrated by our Figure 1C in the current study. A recent retrospective report described a correlation between specific mutations in exon 3 of CTNNB1, particularly S45F, and the risk factors for recurrence in pediatric patients that potentially could be used as prognostic factors.
The multicenter study reported here included a larger number of only primary DTs and confirmed the predictive utility of CTNNB1 mutational status, as established previously by Lazar et al. Surgical margins were not prognostic, as demonstrated previously, although this remains an area of debate. Various investigative series have demonstrated a role for surgical margins as prognostic for DT relapse,[15, 16] whereas others have failed to do so.[3, 4, 17] This inconsistency may be related to the heterogeneity of patients included in the studies (intra-abdominal and extra-abdominal tumors; sporadic and familial adenomatous polyposis-associated DTs as well as primary and recurrent DTs; and additional, inconsistent, nonsurgical treatments). Moreover, because the vast majority of those published series included a relatively large time span as well as both indolent and more aggressive DTs, the role of surgical margins in predicting the propensity for recurrence is difficult to definitively address. It was recently demonstrated that, at 5 years, 50% of patients with primary disease who received “treatment” that consisted of observation alone had yet to progress. Indeed, tumor biology may trump surgical excision margins in predicting the risk of local recurrence. To resolve whether or not surgical margins have prognostic relevance, ideally, only patients who progress under an observation-alone approach should be evaluated.
Although, to our knowledge, this is the largest clinical study to date addressing the role of CTNNB1 gene mutational status as prognostic for DT recurrence, the inherent limitation of this retrospective analysis needs to be overcome by prospective validation; 2 twin, prospective, observational studies have now been initiated in Italy and France to address the potential role of CTNNB1 gene analysis as predictor of tumor aggressiveness. Furthermore, biologic studies focusing on the mechanism by which a specific mutation can drive aggressive DT behavior are crucially needed. A difference in the gene and microRNA expression patterns between specific mutation groups and/or alternative pathways of β-catenin regulation level can be hypothesized, and such studies also are underway.
This current series confirmed that patients with S45F-mutated primary and completely resected DTs exhibit a greater tendency for local recurrence. With the increasing implementation of a “watchful-waiting” policy as part of DT management, it will be important to determine whether mutation type predicts the course of such untreated patients.
This work was supported in part by an Italian Association for Cancer Research (AIRC) fellowship grant (to C.C.), by a grant from the Desmoid Tumor Research Foundation (supporting D.L.), and by the Ricerca Finalizzata 2009 Ministero della Salute (supporting A.G.).