Introduction

Liver resection remains the only curative treatment option for patients with colorectal liver metastases (CLMs). Clinicopathologic factors, such as number and diameter of CLM, concomitant extrahepatic metastases, carcinoembryonic antigen level, and surgical margins, are known to be associated with survival following CLM resection.1,2 Evidence suggests that somatic gene alterations in RAS, TP53, and SMAD4 are associated with survival following CLM resection and provide additional data that augments decision on treatment sequencing and patient selection.3,4,5,6,7,8,9,10,11,12 However, the impact of rare alterations on oncologic outcomes has been recently described, such as those in BRAF, and thus their detection is critical for surgical decision-making and informed discussion on prognosis for patients with CLM.13,14

F-box/WD repeat–containing protein 7 (FBXW7) is a tumor suppressor gene implicated in the degradation of mediators of cell cycle progression. A previous study with extensive genetic analysis showed that FBXW7 was altered in various human tumor types, with an overall alteration frequency of approximately 6%.15 Of these, FBXW7 alteration was found in 35% of cholangiocarcinoma, 31% of T cell acute lymphocytic leukemia, 10% of colorectal cancer, and 9% of endometrial cancer.15 For patients with metastatic colorectal cancer, overall survival (OS) was significantly worse in patients with FBXW7 alteration than in patients with FBXW7 wild-type.16 However, for patients who undergo resection of CLM, the prognostic role of FBXW7 has not been reported. We hypothesized that FBXW7 alterations would negatively impact survival for patients with resected CLM. Within this context, the primary aim was to evaluate the survival impact of FBXW7 alteration for patients undergoing CLM resection.

Materials and Methods

Study Population

We identified patients who underwent initial CLM resection in the Department of Surgical Oncology at The University of Texas MD Anderson Cancer Center from 2001 to 2016, from a prospectively maintained database. Patients who had genetic sequencing data more than 46 genes were included. Demographic and clinicopathologic characteristics, and survival outcomes, were collected. This study was approved by the institutional review board.

Surgical Management of CLM

As previously described,9 our group performs preoperative chemotherapy followed by liver resection and postoperative chemotherapy in most patients with CLM. Preoperative chemotherapy generally consists of oxaliplatin- or irinotecan-containing regimens plus bevacizumab and is administered for 4 cycles. Postoperatively, 8 cycles of the same regimens without bevacizumab are administered.17 CLMs are deemed resectable if negative surgical margins can be achieved while preserving an adequate standardized future liver remnant volume18 If the future liver remnant is insufficient, preoperative portal vein embolization and two-stage hepatectomy are used. 19 Patients are followed after CLM resection with axial imaging every 3–4 months for the first 2 years and every 4–6 months for the subsequent 3 years.20

Somatic Gene Alteration Profiling

As previously described,21 tumor DNA was isolated from 5-mm-thick unstained sections on the basis of tumor tissue blocks or slides from primary colorectal cancer or CLM specimens. Macrodissection was performed in cases of low tumor cellularity. Next-generation sequencing was performed with an AmpliSeq gene panel related to cancer (Supplementary Table 1) using the Ion Torrent Personal Genome Machine (Life Technologies, CA) in a Clinical Laboratory Improvement Amendment–certified molecular diagnostic laboratory.22

Definitions

We defined synchronous metastases as metastases diagnosed within 12 months of primary tumor diagnosis and a positive surgical margin as the presence of tumor cells within 1 mm of the transection line. Primary tumors were staged according to the AJCC Cancer Staging Manual, eighth edition.23

Statistical Analysis

KRAS and NRAS alterations were grouped in a single category, RAS alteration, and analyzed as previously described24,25 and are supported by the fact that survival after CLM resection was worse in patients who had metastatic colorectal cancer and NRAS alteration.26,27,28

Categorical variables were expressed in numbers and percentages and were compared among groups using Fisher’s exact test or the chi-square test, as appropriate. Continuous variables were expressed as median values with the interquartile range. A Cox proportional hazards model analysis was performed with clinicopathologic factors, somatic genes which were associated with prognosis (BRAF, RAS, TP53, and SMAD4),29 and FBXW7. A Cox proportional hazards model analysis initially included age (continuous variable), sex, primary tumor location, T category, primary lymph node metastasis, prehepatectomy carcinoembryonic antigen level (continuous variable), timing of metastasis (synchronous vs. metachronous), prehepatectomy chemotherapy, extrahepatic disease, number of CLM (continuous variable), largest liver metastasis diameter (continuous variable), surgical margin status (R1 vs. R0), BRAF alteration, RAS alteration, TP53 alteration, SMAD4 alteration, and FBXW7 alteration. A backward elimination with a threshold P value of 0.05 was used to select variables for the final models. Hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated for each factor. We estimated the 5-year OS time and survival curves adjusted for covariates by using direct adjusted survival estimation.30,31 This method uses the Cox regression model to estimate survival probabilities at each time point for each individual and averages them to obtain an OS estimate. The proportional hazards assumption was tested by using Schoenfeld residuals. P ≤ 0.05 was considered to indicate statistical significance. Statistical analysis was conducted with SAS (SAS Institute, Cary, NC).

Results

Study Population

Of 2045 patients who underwent CLM resection during the study period, 476 met inclusion criteria (Supplementary Figure 1). Because genetic sequencing was not frequently performed before 2010, 407 (85.5%) of the 476 patients underwent CLM resection from 2011 to 2016.

Table 1 shows demographic and clinicopathologic characteristics. A total of 431 patients (90.5%) underwent prehepatectomy chemotherapy. Of these, 338 (71.0%) received anti-vascular endothelial growth factor (VEGF) agent-containing regimen, and 37 (7.7%) received anti-epidermal growth factor receptor (EGFR) containing regimen. BRAF, RAS, TP53, SMAD4, and FBXW7 were altered in 11 patients (2.3%), 240 patients (50.4%), 337 patients (70.8%), 51 patients (10.7%), and 27 patients (5.7%), respectively. Of the 27 patients with FBXW7 alteration, 26 (96.3%) had mutation of FBXW7 including 25 single nucleotide variation and 1 duplication, and 1 patient missed the detailed information. No amplification was found in patients with FBXW7 alteration (Fig. 1a). Co-alteration of FBXW7 and other somatic genes are shown in Fig. 1b. The frequency of RAS alteration was significantly higher in patients with FBXW7 alteration than in patients with FBXW7 wild-type (77.8% vs. 44.8%, P = 0.005). The frequencies of BRAF, TP53, and SMAD4 were similar between patients with and without FBXW7 alteration.

Table 1 Demographic and clinicopathologic characteristics in 476 patients who underwent resection of CLM from 2001 to 2016
Full size table
Fig. 1
figure 1

Types of alterations and mutations in FBXW7 (a) and co-alteration of FBXW7 with BRAF, RAS, TP53, and SMAD4 (b). Abbreviation: SNV, single nucleotide variation

Full size image

The median duration of follow-up was 3.1 years (interquartile range, 2.1–4.8 years). During the follow-up period, 170 (35.7%) patients died and 388 (81.5%) patients experienced recurrence, including 24 patients with FBXW7 alteration and 364 patients with FBXW7 wild-type. Recurrence rates in the liver alone, lung alone, and two or more sites were 20.8%, 33.3%, and 29.2% in patients with FBXW7 alteration as compared to 36.5%, 28.9%, and 21.4% in patients with FBXW7 wild-type.

A Cox Proportional Hazards Model Analysis for OS After CLM Resection

We evaluated FBXW7 alteration status in a Cox proportional hazards model analysis, together with reported prognostic somatic gene (BRAF, RAS, TP53, and SMAD4) in this patient group and clinicopathologic factors. A multivariable Cox proportional hazards model analysis revealed that alterations of FBXW7 was an independent predictor of OS together with BRAF, RAS, TP53, and SMAD4 (Table 2). Additionally, prehepatectomy chemotherapy > 6 cycles, number of CLM, and largest liver metastasis diameter were associated with OS (Table 2).

Table 2 Multivariable Cox proportional hazards model analysis for OS in 476 patients*
Full size table

OS Estimates Stratified by Alteration Status of FBXW7

OS curves with and without adjustment for other prognostic factors are shown in Fig. 2. The 5-year OS was significantly lower in patients with FBXW7 alteration than in patients with FBXW7 wild-type: 29.7% vs. 61.2%, P = 0.005. After adjustment for other prognostic factors, the covariate-adjusted 5-year OS remains significantly lower in patients with FBXW7 alteration than in patients with FBXW7 wild-type: 40.4% vs. 59.4%, P = 0.015.

Fig. 2
figure 2

Overall survival (OS) by FBXW7 alteration status. a OS curves. b OS curves after adjustment for somatic gene alteration status (BRAF, RAS, TP53, and SMAD4), prehepatectomy chemotherapy (> 6 cycles vs. ≤ 6 cycles or no prehepatectomy chemotherapy), number of CLM, and largest liver metastasis diameter

Full size image

OS Estimates Stratified by Alteration Status of RAS and FBXW7

Because the frequency of FBXW7 alteration was significantly higher in patients with RAS alteration than in patients with RAS wild-type, we evaluated OS stratified by RAS and FBXW7 alteration (Fig. 3). The 5-year OS was lower in patients with co-alteration of RAS and FBXW7 alteration than in patients with RAS alteration and FBXW7 wild-type (27.1% vs. 53.4%, P = 0.066) and in patients with RAS wild-type (27.1% vs. 67.0%, P < 0.001). After adjustment for other prognostic factors, the covariate-adjusted 5-year OS was significantly lower in patients with co-alteration of RAS and FBXW7 alteration than in patients with RAS alteration and FBXW7 wild-type (26.1% vs. 47.0%, P = 0.036) and in patients with RAS wild-type (26.1% vs. 70.6%, P < 0.001). We repeated the analysis of OS stratified by TP53 and FBXW7 alteration. Similarly, the 5-year OS with and without adjustment for other prognostic factors was lower in patients with co-alteration of TP53 and FBXW7 alteration (Supplementary Figure 2). The 5-year OS without adjustment of patients with alterations in FBXW7, RAS, and TP53 (triple alteration, 17.9%) was worse than alterations in FBXW7 and RAS or FBXW7 and TP53 or RAS and TP53 (double alteration, 38.2%) although the number of patients with alteration in FBXW7, RAS, and TP53 was small (n = 12).

Fig. 3
figure 3

Overall survival (OS) by RAS and FBXW7 alteration status. a OS curves. b OS curves after adjustment for somatic gene alteration status (BRAF, TP53, and SMAD4), prehepatectomy chemotherapy (> 6 cycles vs. ≤ 6 cycles or no prehepatectomy chemotherapy), number of CLM, and largest liver metastasis diameter

Full size image

Discussion

Patients with FBXW7 alteration experienced worse OS after CLM resection compared to FBXW7 wild-type patients. When grouped by RAS and FBXW7, or TP53 and FBXW7 alteration status, the stratification of prognosis was more refined. Of the 12 clinicopathologic factors, only 3 factors (prehepatectomy chemotherapy > 6 cycles, number of CLM, and largest liver metastasis diameter) were associated with OS when assessed with somatic gene alteration status of FBXW7, BRAF, RAS, TP53, and SMAD4. Our findings confirm the prognostic importance of knowing the status of multiple potential somatic gene alterations in CLM patients due to the genomic heterogeneity of colorectal cancer.

FBXW7 is a tumor suppressor gene associated with the Notch signaling pathway (Fig. 4).32 Our study showed that the frequency of FBXW7 alteration was 5.7% in patients with CLM, in line with previous studies that reported frequency rates of 6–10%.15,33 Importantly, this alteration was more frequent among CLM patients than BRAF alteration, which is well recognized to be a poor prognostic marker: the percentage of patients with BRAF alteration was 2.3% in our study and 2–5% in a large series including patients with CLM.29 In line with previous reports,15,34 our study found that the frequency of FBXW7 alteration was significantly higher in patients with RAS alteration than in patients who were RAS wild-type. The Notch pathway is a regulator of cell growth and differentiation.35 Inactivation of FBXW7 causes abnormal accumulation of the intracellular domain of Notch1 and influences cell growth.36 As such, alteration of FBXW7 may result in uncontrolled cell growth and proliferation and thus, a deleterious effect on survival through the Notch pathway. The resultant negative survival impact has been reported for patients with metastatic colorectal cancer.16 Nonetheless, our study is the first to show that OS after CLM resection was significantly worse in patients with FBXW7 alteration than in patients with FBXW7 wild-type.

Fig. 4
figure 4

Overview of NOTCH signaling pathway. Abbreviations: NICD1, intracellular domain of Notch1

Full size image

The Cancer Genome Atlas project has detailed the landscape of somatic gene alteration of colorectal cancer in the context of cancer-related signaling pathways.37 Our group reported that alteration of RAS, TP53, and SMAD4 and co-alteration of RAS and TP53 were associated with worse survival.4,8,9,10,21RAS, TP53, and SMAD4 belong to three cancer-related signaling pathways: the mitogen-activated protein kinase (MAPK) pathway, the p53 pathway, and the transforming growth factor-β (TGF-β) pathway, respectively.37 Because these three pathways are associated with tumor-cell growth, it may be plausible that the malfunction of these pathways influences prognosis in patients with CLM. The information on alterations in these pathways may have high impact on clinical practice because the alterations of RAS and TP53 were found in more than 50% of this patient group.

Alterations of FBXW7 and BRAF are less frequent than alterations of RAS, TP53, and SMAD4. However, we believe that it is important to identify rare deleterious alterations in order to more succinctly predict CLM patients’ prognosis. It is being increasingly recognized that the interplay of multiple altered signaling pathways in CLM may cause deleterious effect and result in observable differences in tumor phenotype, response to therapy, and pattern of recurrence after resection. Therefore, it is imperative that we identify the status of the rare alterations, such as that of BRAF and FBXW7, because they not only allow for prognostication on their own, but when preset in combination with others provide more accurate data for patients with CLM. This is clearly demonstrated in our previous reports8,21 and here with the survival differences in RAS alteration patients with or without a co-alteration in FBXW7. The data presented emphasizes the importance of multiple gene testing as single gene alterations are insufficient for accurate prognostication after CLM resection. Whether these somatic alterations can definitively direct patient selection for surgery and treatment sequencing is an evolving subject. We may use this information to identify patients who have CLM with favorable molecular biology (i.e., wild-type in FBXW7, BRAF, RAS, TP53, and SMAD4) and may be best suited for aggressive surgery and local therapies. For example, patients with poor clinicopathologic factors (e.g., number of CLMs > 10, largest diameter of CLM > 10 cm, multiple primary lymph node metastases, extrahepatic metastases) but with favorable molecular biology may expect oncological benefits using aggressive treatment strategies.

Our study should be understood in the context of limitations. First, the retrospective single-institution design makes it difficult to preclude all biases. Nonetheless, the large size of the study cohort with complete data regarding the status of 46 somatic gene alterations allowed the analysis of patients with FBXW7 alteration. Second, we analyzed patients who had complete data of 5 somatic genes (FBXW7, BRAF, RAS, TP53, and SMAD4). As such, we included only patients who underwent the 46-gene panel test in the study. Third, we did not analyze specific types of genetic alteration because the majority of FBXW7 alterations were single nucleotide variations followed by duplication. Last, we studied patients who underwent CLM resection for a relatively long period from 2001 to 2016. However, this may be a limited impact because genetic sequencing has only been performed with regularity in the past several years, and 85.5% of the patients in the study underwent CLM resection after 2011 and had similar management of CLM. Further study including more patients may elucidate the interaction of multiple alterations in FBXW7 and other somatic genes.

Conclusion

In conclusion, FBXW7 alteration was found in 5.7% of patients undergoing CLM resection and was associated with worse survival. This finding further supports the genetic heterogeneity of colorectal cancer and the importance of determining the status of multiple somatic gene alterations for risk stratification for patients with CLM considering resection.