SEARCH

SEARCH BY CITATION

Keywords:

  • molecular heterogeneity;
  • primary colorectal cancer;
  • genetic and epigenetic alterations;
  • lymph node metastases;
  • liver metastases

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Note Added in Proof
  8. REFERENCES

BACKGROUND:

Colorectal cancer (CRC) prognosis and survival are strictly related to the development of distant metastases. New targeted therapies have increased patient survival, but the objective response rate is still very limited, partially because of a traditional focus on designing treatment according to the molecular profile of the primary tumor regardless the diversity between the primary tumor and metastases. The objective of this study was to evaluate the presence of molecular heterogeneity during metastatic progression and its potential impact on clinical treatment.

METHODS:

The authors analyzed v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) codon 12 mutations, the v-Raf murine sarcoma viral oncogene homolog B1 (BRAF) thymine to adenine substitution at codon 1788, and tumor protein 53 (p53) mutations and investigated promoter methylation of Ras association (RalGDS/AF-6) domain family member 1 protein (RASSF1a), E-cadherin, and cyclin-dependent kinase inhibitor 2A (p16INK4a) in 101 primary CRCs (67 stage III and 34 stage IV) and related lymph node and liver metastases.

RESULTS:

Lymph node metastases were characterized by fewer alterations compared with primary tumors and liver metastases, especially KRAS (P = .03) and p16INK4a (P = .05). Genetic changes, when detectable in metastases, mostly were retained from the primary tumor, whereas epigenetic changes more frequently were acquired de novo. Overall, 31 distinct CRC molecular profiles were detected, none of which characterized a particular tumor stage. When the metastatic lesions also were included in the profiles, there were 53 distinct molecular profiles in 67 patients with stage III disease and 34 distinct molecular profiles in 34 patients with stage IV disease.

CONCLUSIONS:

Lymph node and liver metastases appear to originate in clonally different processes, with more molecular alterations occurring in distant metastases than in lymph node metastases and with elevated heterogeneity of the primary tumor. Thus, potential prognostic targets should be carefully evaluated for their heterogeneity in both primary tumors and distant metastases to avoid erroneous misclassification. Cancer 2013. © 2012 American Cancer Society.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Note Added in Proof
  8. REFERENCES

Colorectal cancer (CRC) is the second leading cause of cancer-related mortality in Western populations and 1 of the best characterized models of multistep carcinogenesis. Alterations in oncogenes, such as the v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS), and in tumor suppressor genes, such as adenomatous polyposis coli (APC), deleted in CRC (DCC), and tumor protein 53 (p53),1 are consistently involved in CRC pathogenesis. Overall survival is highly dependent on disease stage at diagnosis, and the estimated 5-year survival rates range from 85% to 90% for patients with stage I disease to <5% for patients with stage IV disease. Although it is evident that the main problem in the treatment of CRC is not the eradication of the primary tumor but, rather, the formation of incurable metastases,2 to date, there are no clinical, pathologic, or molecular markers to identify patients who are at risk of developing distant metastases.

Currently, 2 monoclonal antibodies, cetuximab and panitumumab, are used as second-line or third-line chemotherapy for patients with metastatic CRC.3, 4 Both molecules bind the epidermal growth factor receptor (EGFR), leading to inhibition of its downstream signaling pathways—rat sarcoma/extracellular signal-regulated kinase/mitogen-activated protein kinase (RAS/ERK/MAPK) and phosphatidylinositol 3-kinase/phosphatase and tensin homolog/v-akt murine thymoma viral oncogene homolog (PI3K/PTEN/AKT)—and providing clinical benefits. Because it has been demonstrated that EGFR downstream-effector alterations are predictive of nonresponse, KRAS analysis has been routinely introduced into molecular pathology laboratories, and mutations in this gene have become excluding criteria for the administration of these therapies to patients with metastatic CRC. Although this decision is expected to ameliorate the therapeutic index in this selected population, the objective response rate remains very limited (12% and 17% for cetuximab and panitumumab, respectively5).

Although patient-specific factors may contribute to therapy resistance, the rate of low treatment success remains incompletely explained.6 The lack of translational success may be caused in part by a traditional focus on designing treatment according to the molecular profile of the primary tumor,7 considering metastases as the end stage of tumorigenesis. The potential molecular and functional differences between primary tumors and distant metastases have serious implications for the treatment of neoplastic disease. In fact, the metastatic site can vary in sensitivity to cytostatic agents8 and may be a major impediment to the success of chemotherapy for systemic disease. Moreover, the cell population within individual lesions may progress independently, developing variant sensitivities to common therapeutic modalities and entailing serious implications for the development of new therapeutic regimens.8

With the objective of gaining a better understanding of the presence of molecular alterations during metastatic progression and their potential impact on clinical treatment, we studied the mutational status of KRAS, v-Raf murine sarcoma viral oncogene homolog B1 (BRAF), and p53 and the epigenetic status Ras association (RalGDS/AF-6) domain family member 1 protein (RASSF1a), E-cadherin, and cyclin-dependent kinase inhibitor 2A (p16INK4a) in a series of metastatic CRCs and their related lymph node and liver metastases. We previously published a comprehensive study on the molecular changes in these genes between a set of primary metastatic and nonmetastatic tumors9; for the current study, we decided to expand that analysis to lymph node and liver metastases. This study allowed us to evaluate the frequency of alteration for each gene in a specific neoplastic site and its trend during metastatic progression. Moreover, the molecular comparison of 3 neoplastic tissues from same patients allowed us to establish the presence and the amount of molecular heterogeneity during metastatic spread, to speculate how this heterogeneity between primary tumor and local and distant metastases potentially may influence clinical decision making, and to suggest which tissue sample is the most suitable for molecular analysis.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Note Added in Proof
  8. REFERENCES

Patients

This study included 101 microsatellite-stable metastatic CRCs (T3/T4) that were surgically removed between 1997 and 2002 and their corresponding lymph node metastases (n = 101) and liver metastases (n = 34). In particular, we analyzed 67 stage III CRCs and matched lymph node metastases and 34 synchronous stage IV CRCs with both lymph node and liver metastases. All patients had ≥21 lymph nodes available, as recommended by international guidelines. For each patient, the most representative paraffin blocks from the primary tumor, lymph node metastases, and liver metastases were selected for the study (3 blocks for each patient). If the neoplastic component in liver metastases was <50% of the whole tissue, then we performed a manual microdissection to minimize tumor contamination from normal cells. Liver metastasis samples were obtained from hepatic resection in 25 patients and from intraoperative biopsy in 9 patients. Seven patients were not considered fit for adjuvant treatment. The other 27 patients received a 5-fluoruracil-based therapy, and only 1 patient received a therapy based on the use topoisomerase-I plus, an inhibitor of thymidylate synthase. None of the patients who had synchronous liver metastases received neoadjuvant therapy.

DNA Extraction

Tissue sections were deparaffinized as previously reported9 using 100% xylene (Sigma Chemical Company, St. Louis, Mo) followed by 100% ethanol. The pellet was then resuspended in a buffer containing proteinase K (Finnzyme, Espoo, Finland), and DNA was extracted with phenol-chloroform (Sigma Chemical Company) followed by ethanol precipitation and resuspention in 100 μL of water. DNA was quantified spectrophotometrically, and 200 ng were used as a template for each polymerase chain reaction (PCR).

KRAS and BRAF Mutations: Restriction Fragment-Length Polymorphism-Polymerase Chain Reaction Analysis

Mutations at codon 12 of the KRAS gene and the BRAF thymine to adenine substitution at codon 1788 (BRAFc.1799T[RIGHTWARDS ARROW]A) mutation were detected by restriction fragment-length polymorphism-PCR, as previously reported.9 This method identifies KRAS codon 12 mutations but not the specifically altered nucleotide. Positive controls (the SW480 cell line for KRAS and the HT29 cell line for BRAFc.1799T[RIGHTWARDS ARROW]A) and negative controls (human placental DNA) for mutations and controls for carryover DNA contamination were included in every experiment. To address the sensitivity of our molecular assay, we performed a titration experiment using a serial dilution of a mutated cell line (SW480), and the results indicated that this method could detect <5 mutated alleles in 100. In case of discordant results between a primary tumor and metastases, we verified the results by sequencing the wild-type site.

p53 Mutation

Single-strand conformation polymorphism

We used single-strand conformation polymorphism (SSCP) to identify samples with altered p53 migratory patterns according to the presence of mutations or polymorphisms. Initial PCR fragments were generated from genomic DNA using primer sequences and PCR conditions as previously described.9

DNA that contained wild-type p53 (human placental DNA) was always included as a control. For SSCP analysis, the PCR product was denatured at 96°C for 10 minutes, then loaded onto 37.5% Tris/borate/ethylene diamine tetraacetic acid (EDTA) polyacrylamide gels and electrophoresed for 1 hour at 100 volts and for 4 to 5 hours at 300 volts. Gels were silver stained with DNA Silver Staining Kit (Amersham, Buckinghamshire, United Kingdom) according to the manufacturer's recommendations.

Immunohistochemistry

To analyze p53 expression, formalin-fixed, paraffin-embedded tissue sections (2 μm) were used, deparaffinized, and exposed to an antigen retrieval system (1 mM EDTA, pH 8; at 98°C for 30 minutes) before they were incubated with the specific p53 antibody (1:1000 dilution; Ab-2; Calbiochem, Cambridge, Mass). Endogenous peroxidase was blocked with 3% hydrogen peroxide for 20 minutes at room temperature. Primary mouse monoclonal antibody was applied for 1 hour at room temperature. Reactive sites were identified with secondary antibody (horseradish peroxidase rabbit/mouse; Envision; DAKO, Carpinteria, Calif) for 30 minutes. Immunoperoxidase staining using diaminobenzidine as the chromogen was carried out (diaminobenzidine plus chromogen ×50; DAKOCytomation, Carpinteria, Calif). The slides were counterstained with Harris hematoxylin (DiaPath, Microstain Division, Martinengo, Italy). An abnormal SSCP pattern coincidental with p53 nuclear immunoreactivity was considered as indicative of tissues with p53 mutation.9

RASSF1a, E-Cadherin, and p16INK4a methylation

Genomic DNA (1 μg of each sample) was treated with 10 mM hydroquinone and 3 M sodium bisulphite (Sigma Chemical Company) at 50°C for 16 hours, as previously reported.10 Modified DNA was purified using the Wizard DNA purification resin according to the manufacturer's instructions (Promega, Milan, Italy). PCR was performed separately with methylation-specific and unmethylated primers for each gene.9 Unmethylated (human placental) DNA and methylated DNA (LoVo, COLO320, and HepG2 cell lines for RASSF1a, E-cadherin, and p16INK4a, respectively) were used as controls. All samples that exhibited PCR products for a methylated DNA sequence on 2% agarose gels also exhibited an unmethylated DNA sequence, probably related to tumor heterogeneity or to the occurrence of intermingled non-neoplastic cells. Because it was not possible to eliminate the contamination by normal cells (the presence of inflammatory, structural, and connective cells in the paraffin tissues), a sample was classified as methylated whenever a band corresponding to the molecular weight of the methylated PCR product had a thickness and staining intensity equal to or greater than the molecular weight of the unmethylated PCR product. To address the sensitivity of our molecular assay, we performed a titration experiment using serial dilutions of a methylated cells line (LoVo), and the results indicated that this method can detect less than <1 methylated allele in 100.

Statistical Analysis

Data are described as numbers and percentages or means and standard deviations, as appropriate. Differences in the frequencies of variables were tested using the chi-square test (Pearson or Fisher, as appropriate) or the McNemar test with the continuity correction, and differences in mean levels of variables were tested using the t test or the Wilcoxon test. A P value of .05 was considered statistically significant. Data analysis was done using the STATA (version 9.0; Stata Corporation, College Station, Tex).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Note Added in Proof
  8. REFERENCES

In studies like our current investigation, the experimental procedure addressing whether or not the site of metastasis differs from the primary tumor should be as robust as possible to be convincing. For this reason, after our first analysis, we repeated the molecular analysis in all cases with discordant KRAS codon 12 and BRAFc.1799T[RIGHTWARDS ARROW]A results and, for the remaining genes, in randomly selected cases with discordant results. Furthermore, considering the current importance of KRAS/BRAF mutations for therapy in patients with CRC, an additional tumor sample was analyzed from all cases with discordant results. In addition, we observed that all 37 CRCs that carried p53 mutations also had nuclear accumulation of the protein on immunohistochemical analysis (Fig. 1).

thumbnail image

Figure 1. These are images from a representative immunohistochemical analysis of p53 protein expression according to gene mutation status. (A) Single-strand conformation polymorphism analysis of exon 6 of p53 in colorectal cancer tissues reveals an altered pattern of migration in Patient 76, M = molecular weight marker. (B) This photomicrograph reveals the nuclear accumulation of p53 protein in cancer tissues from Patient 76.

Download figure to PowerPoint

After this preliminary control for the robustness of our data, we analyzed the prevalence of alteration for each single gene in the whole series without any stage distinction. Table 1 indicates that KRAS codon 12 mutations were similar in primary tumors (32%) and liver metastases (38%) but were decreased in lymph node metastases (19%; P = .03). Among the epigenetic changes, p16INK4a methylation was greater in primary tumors (41%) than in lymph node metastases (27%; P = .05) or liver metastases (26%). In primary CRCs, p16INK4a methylation was associated with a lack of protein expression (Fig. 2). For the other genes under study, there were not statistically significant differences between primary tumors and metastatic sites.

Table 1. Overall Frequencies and Comparison of Genetic and Epigenetic Changes in Stage III and IV Colorectal Cancers, Lymph Node Metastases, and Liver Metastases
 No. of Patients (%)a
   Primary TumorsLN Metastases 
Gene ChangesPrimary Tumors n = 101LN Metastases, n = 101Stage IIII, n = 67Stage IV, n = 34Stage III, n = 67Stage IV, n = 34LV Metastases, n = 34
  • Abbreviations: B-Raf, v-Raf murine sarcoma viral oncogene homolog B1; E-Cad, cadherin 1, epithelial; K-Ras, V-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog; LN, lymph node; LV, liver; P16IN4a, cyclin-dependent kinase inhibitor 2A; p53, tumor protein 53; RASSF1a, Ras association (RalGDS/AF-6) domain family member 1 protein.

  • a

    The frequency of gene changes was compared between tumors using the Fisher exact test to calculate all P values.

  • b

    P = .03.

  • c

    P = .04.

  • d

    P = .06.

  • e

    P = .05.

  • f

    P = .03.

  • g

    P = .06.

  • h

    P = .09.

K-Ras32 (32)b19 (19)b19 (28)c13 (38)9 (13)cd10 (29)d13 (38)
B-Raf5 (5)4 (4)2 (3)3 (9)1 (1.5)3 (9)5 (15)
p5337 (37)34 (34)21 (31)16 (47)20 (30)14 (41)14 (41)
p16IN4a41 (41)e27 (27)e28 (42)f13 (38)g16 (24)f6 (18)g9 (26)
E-Cad38 (38)27 (27)26 (39)12 (35)18 (27)9 (26)11 (32)
RASSF1a26 (26)24 (24)21 (31)5 (15)h17 (25)7 (21)12 (35)h
thumbnail image

Figure 2. These are images from a representative immunohistochemical analysis of cyclin-dependent kinase inhibitor 2A (p16INK4a) protein expression according to gene methylation status. (A) Results are shown in methylated (m) and unmethylated (u) samples of a positive control (Pos. Ctrl.) (human placental DNA treated in vitro with an excess of SssI methyltransferase, which generates a completely methylated DNA), colorectal cancer (CRC) tissues from Patients 53 and 47, and a negative control (Neg. Ctrl.) (untreated human placental DNA). (B) These photomicrographs from an immunohistochemical analysis of CRC tissues for p16INK4a expression reveal (Left) unmethylated CRC tissue from Patient 53 with positive p16INK4a expression and (Right) methylated CRC tissue from Patient 47 with negative p16INK4a expression.

Download figure to PowerPoint

To easily compare the molecular trends observed in our series, we subdivided our samples according to tumor stage (III vs IV) and sites of metastases (lymph node and liver). Thus, we compared (Table 1) the percentage of alterations for each individual gene in: 1) primary tumors (stage III or stage IV) versus the related lymph node metastasis, 2) primary tumors versus liver metastasis, 3) stage III versus stage IV lymph node metastasis, and 4) stage IV lymph node metastasis versus liver metastasis.

KRAS mutations and p16INK4a methylation decreased between stage III CRCs (28% and 42%, respectively) and the corresponding lymph node metastases (13% [P = .04] and 24% [P = .03], respectively). In a comparison between stage IV CRC and lymph node metastases, only KRAS, E-cadherin, and p16INK4a abnormalities decreased, but the decreases were without statistical significance. The comparison between stage IV primary tumors and corresponding liver metastases revealed an increase in RASSF1a methylation (15% vs 35%; P = .09) and a decrease in p16INK4a methylation (38% vs 26%). In lymph node metastases from stage III and IV CRCs, KRAS mutations increased from 13% to 29%, respectively (P = .06). Generally, in lymph node metastases from stage III and IV CRCs, mutation changes (KRAS, BRAFc.1799T[RIGHTWARDS ARROW]A, and p53) tended to increase, whereas RASSF1a, E-cadherin, and p16INK4a methylation remained essentially unaltered. KRAS, p16IN4a, and RASSF1a alterations increased from lymph node metastases to liver metastases originating from stage IV CRC.

Next, we analyzed individual genes for molecular concordance or discordance between primary tumors (overall and at different stages) and metastases (Fig. 3). Overall, 13 lymph node metastases from primary CRCs with KRAS mutations had the wild-type KRAS codon 12 (P = .001). All 19 KRAS mutations that were detected in lymph nodes notes also were detectable in the primary tumor.

thumbnail image

Figure 3. The rate of concordant (+/+ and − −) and discordant (+/− and −/+) genetic and epigenetic changes are illustrated in primary colorectal cancer (CRC) (stages III and IV) and in corresponding lymph node (LN) and liver (LV) metastases. KRAS indicates v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog; BRAF, v-Raf murine sarcoma viral oncogene homolog B1; RASSF1a, Ras association (RalGDS/AF-6) domain family member 1 protein.

Download figure to PowerPoint

Sixteen patients had concordant p16INK4a methylation in the primary tumor and lymph node metastases, whereas 31 patients had discordant methylation (25 patients had unmethylated lymph nodes but an methylated primary CRC, and 6 patients had methylated lymph nodes but an unmethylated primary CRC; P = .001). To further address the issue of possible contamination by nontumor cells leading to false-negative results, we revised the extent of tumor involvement in lymph nodes from patients who had discordant results between lymph node metastases and the primary tumor. We observed that the percentage of micrometastases (small foci of tumor cells) in which the rate of false-negative results may be higher did not differ statistically between concordant versus discordant cases for all genes under study (results not shown).

In Table 2, the CRC molecular profiles are provided to highlight the peculiar molecular combinations that characterized CRCs with and without distant metastases. We had 31 different CRC molecular profiles (101 patients) ranging from 0 alterations up to 5 alterations. The average number of molecular alterations in individual patients was similar between those with stage III CRC (1.75 alterations) and those with stage IV CRC (1.82 alterations). We did not detect any combinations that characterized a particular disease stage. Surprisingly, the CRC molecular profile without molecular alterations was well represented (1 of the most frequent) in all analyzed CRCs.

Table 2. The 31 Individual Molecular Profiles of Colorectal Cancers Under Study
  1. Abbreviations: BRAF, v-Raf murine sarcoma viral oncogene homolog B1; CRC, colorectal cancer; KRAS, V-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog; LN, lymph node metastases; P16, tumor protein 16 (cyclin-dependent kinase inhibitor 2A); P53, tumor protein 53; RASSF1a, Ras association (RalGDS/AF-6) domain family member 1 protein.

inline image

In Table 3 the molecular profiles (primary CRC plus lymph node metastases) of 67 patients with stage III disease are reported. We identified 53 different individual molecular profiles. The same individual molecular profile was shared by at least 2 patients in each group with 1 alteration (7 distinct molecular profiles in 11 patients), 2 alterations (7 distinct molecular profiles in 12 patients), and 3 alterations (14 distinct molecular profiles in 15 patients). Within the groups that had ≥4 alterations, all individual molecular profiles differed. The molecular profile (primary CRC plus lymph node and liver metastases) was different in all 34 patients with stage IV disease (Table 4).

Table 3. Individual Molecular Profiles of 67 Stage III Colorectal Cancers and Corresponding Lymph Node Metastases
  1. Abbreviations: BRAF, v-Raf murine sarcoma viral oncogene homolog B1; CRC, colorectal cancer; KRAS, V-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog; LN, lymph node metastases; P16, tumor protein 16 (cyclin-dependent kinase inhibitor 2A); P53, tumor protein 53; RASSF1a, Ras association (RalGDS/AF-6) domain family member 1 protein.

inline image
Table 4. Individual Molecular Profiles of 34 Stage IV Colorectal Cancers, Lymph Node Metastases, and Liver Metastases
  1. Abbreviations: BRAF, v-Raf murine sarcoma viral oncogene homolog B1; CRC, colorectal cancer; KRAS, V-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog; LN, lymph node metastases; LV, liver metastases; P16, tumor protein 16 (cyclin-dependent kinase inhibitor 2A); P53, tumor protein 53; RASSF1a, Ras association (RalGDS/AF-6) domain family member 1 protein.

inline image

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Note Added in Proof
  8. REFERENCES

Notwithstanding the considerable improvements in surgical management and targeted therapies, metastatic spread, mostly to the liver, still represents the major cause of mortality among patients with CRC. The poor survival of patients with metastatic disease and the low response rate, even for new targeted therapies like cetuximab and panituxumab,11 point out the need for improved screening techniques and optimized treatment. In fact, although patient-specific factors may contribute to drug resistance, the rate of low treatment success remains incompletely explained. One hypothesis is that CRC metastases are genetically heterogeneous from the primary neoplasm. Thus, a standard approach to the treatment of metastatic disease using the same therapy regimen for the primary tumor and for all patients is unlikely to be uniformly effective.

In this study, we explored KRAS, BRAF, and p53 mutation status and RASSF1a, E-cadherin, and p16INK4a epigenetic (methylation) status in a series of metastatic CRCs and their related lymph node and liver metastases. This analysis allowed us to establish the heterogeneity of these genes during metastatic spread and to speculate on its potential implications in clinical decision making and in selecting the best suitable tissue sample for molecular analysis. Moreover, the use of tumor samples from the same patients significantly lowered the background variation that originates from individual differences, allowing the comparison of individual molecular profiles between patients with different tumor stages. To obtain high levels of sensitivity and to minimize the occurrence of false-positive results in our assays, we included appropriate negative and positive controls in all experiments both for genetic and epigenetic alterations. Cases with discordant results in genetic/epigenetic alteration patterns between primary CRCs and paired metastases were analyzed in duplicate. To our knowledge, our results well represent the distribution of alterations for these genes in the different sites analyzed; however, we cannot completely exclude the possibility of sampling bias because of the possible genetic/epigenetic heterogeneity of cancer clones within different tumor areas and anatomic sites (ie, primary CRC, lymph node metastases, and liver metastases).

The concordance of KRAS codon 12 mutations between primary CRCs and lymph node and liver metastases was 87% and 88%, respectively. In line with published concordance rates,12 our data indicate that tumor heterogeneity at different anatomic sites should not be dismissed from the perspective of personalized cancer therapy. Nevertheless, in our series, lymph node metastases were characterized by fewer alterations, especially in p16INK4a and KRAS, compared with primary tumors. This may indicate that neoplastic cells colonizing the lymph nodes already have left the primary tumor before KRAS and/or p16INK4a alterations. Otherwise, assuming the primary tumor is constituted by molecularly heterogeneous populations, we hypothesized that 1 of these cell subsets is responsible for the survival and proliferation of cells in the primary site (mutated KRAS and methylated p16INK4a, as previously reported in our group9), whereas another subset has the ability to colonize lymph nodes (wild-type KRAS and unmethylated p16INK4a, although we cannot exclude the presence of other alterations). When sorted by disease stage (Table 1), KRAS and p16INK4a alterations decreased between primary tumors and lymph node metastases, with a statistically significant decrease in stage III CRCs but not in stage IV CRCs; because, during lymph node dissemination, a stage IV tumor is more prone to retain alterations compared with a stage III tumor.

In liver metastases, we observed an increase in RASSF1a methylation (a high percentage of de novo events) and a decrease in p16INK4a methylation. Such a high frequency of RASSF1a methylation in liver metastases may originate in part from a contamination with adjacent non-neoplastic hepatocytes, which are subjected to an age-related methylation mechanism that is active from age 30 years onward, as previously reported.13 The down-regulation of p16INK4a expression in metastases partially disagrees with our previous data, in which we identified p16INK4a methylation as 1 of the molecular alterations that characterize metastatic tumors.9 Considering our results and some previous indications about the demethylation and elevated expression of p16INK4a protein and mRNA levels at the CRC invasion front,14, 15 it seemed reasonable to hypothesize the presence of different subpopulations at the primary site, where the p16INK4a promoter probably is methylated in the central region but not at the invasive front, from which the metastatic cells depart to colonize secondary organs.

During CRC progression, mutations rarely occur de novo, because most lymph node mutations (47 of 60 mutations; 78%) and liver mutations (31 of 43 mutations; 72%) also were detectable in primary CRCs. By contrast, genuine de novo methylation was less predictable (44 of 76 lymph node metastases [58%], 18 of 51 liver metastases [35%]; P = .0009). Overall, the individual genetic profile in stage III and IV CRCs was shared by at least 2 individuals in 77% and 41% of cases, respectively; whereas the individual epigenetic profile in stage III and IV CRCs was shared by at least 2 individuals in 55% (P = .001) and 15% (P = .003) of cases, respectively. We are tempted to speculate that, although genetic changes may be more tumor-related, epigenetic changes may be more individual-related. This may fit with the different dynamics of genetic and epigenetic gene silencing in cancer. Transcriptional or replicative errors lead to a specific altered protein. In case of a selective advantage, the cells expand clonally to give rise to a tumor in which all cells have the same altered protein.16 In contrast, epigenetically mediated gene silencing occurs gradually and randomly, and it is really a matter of dose. For this reason, we classified a sample as methylated only if the methylated band was as thick as the unmethylated band.

The use of 3 different samples within the same patient allowed us to analyze all genetic/epigenetic combinations to evaluate the interpatient grade of molecular variability. Considering only primary tumors, we identified 31 different molecular profiles in 101 patients. We were not able to identify any combination that was specifically typical of a particular stage. When we evaluated the individual molecular profiles with the addition of lymph node and liver metastases (Tables 3 and 4), the same molecular profiles (CRC plus metastases) tended to be shared by at least 2 individuals in a small fraction of patients (14 of 67 patients with stage III CRC and 0 of 34 patients with stage IV CRC; 14 of 101 [14%] shared profiles). This high interpatient variability should be taken into account, especially now that research is focused mainly on genetic therapies that target only 1 altered molecule at time instead of broad-spectrum treatments.

New evidence suggests that patients who have alterations in EGFR downstream effectors have a limited clinical response to cetuximab compared with patients who have wild-type tumors. Our analysis demonstrated concordance for KRAS mutations between primary tumors and metastases, as previously reported17, 18: all KRAS mutations that we detected in lymph nodes also were documented in the primary CRC. Nevertheless, a small percentage of liver metastases produced discordant results with respect to the primary tumor: Two patients had KRAS mutations in the primary tumor only, and an additional 2 patients had KRAS mutations in liver metastases only. A similar discordance recently was demonstrated by Vermaat et al.19

In our series, BRAFc.1799T[RIGHTWARDS ARROW]A mutations were rare (only 4% of CRCs) and were never observed in combination with KRAS mutations. Three patients harbored a BRAFc.1799T[RIGHTWARDS ARROW]A mutation exclusively in liver metastases, whereas 1 patient had the mutation only in the primary tumor. According to the guidelines for cetuximab therapy, overall, 8 distant metastases (24%) would go erroneously misclassified. To minimize the possibility of false-negative results because of sample selection, we analyzed additional samples from patients who had discordant KRAS and BRAF patterns between primary tumors and related metastases. We analyzed 8 samples for the BRAFc.1799T[RIGHTWARDS ARROW]A mutation and did not detect any additional mutations. In total, 19 patients had discordant results for KRAS mutations; and, in 4 of those patients, the additional analysis revealed a mutation in a different sample. This corroborates the possibility of molecular heterogeneity not only between primary tumors and related metastases but also within the same lesion, as previously reported in other studies.20, 21 Even if recent data did not confirm the benefit of adding cetuximab to traditional chemotherapeutic regimens,22 and although the response rate is still inexplicably low, even in selected CRC populations, the introduction of these drugs has been 1 of the most promising developments in oncology treatment in the past 5 years. Therefore, in a clinical and diagnostic setting involving these monoclonal antibodies, our results confirm the necessity of evaluating KRAS and BRAF mutation status in distant metastatic sites and not only in the primary tumor.

Screening tools designed to identify individual patients who are at risk from distant metastases and who are resistant to chemotherapy are of extreme necessity. Although further matched tumor studies need to be completed to confirm the current data, the development of biologic therapies should face and take into account the relevant grade of intertumor and intratumor heterogeneity, which determines the simultaneous presence within the same tumor of several molecular clones that, consequently, have different chemoresistant and metastasizing capabilities. Thus, it seems reasonable to conclude from the current study that each molecular target intended to be used for prognosis in patients with CRC should be evaluated carefully for molecular heterogeneity in all available tissue samples.

Note Added in Proof

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Note Added in Proof
  8. REFERENCES

FUNDING SOURCES

This work was supported by the Italian Ministry for University, Scientific, and Technological Research (MURST) (Cofinancial Projects 2003). Dr. Miranda was supported by a fellowship of the Doctorate School of Molecular Medicine, University of Milan, granted by the Institute for In-Patient Treatment and Scientific Studies (IRCCS) Humanitas Clinical Institute (Milan).

CONFLICT OF INTEREST DISCLOSURES

The authors made no disclosures.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. Note Added in Proof
  8. REFERENCES