SEARCH

SEARCH BY CITATION

Keywords:

  • liver metastases;
  • colorectal carcinoma;
  • copy number change;
  • single nucleotide polymorphism array

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SUPPORT
  8. CONFLICT OF INTEREST DISCLOSURES
  9. REFERENCES

BACKGROUND

Metastatic dissemination is the most frequent cause of death in patients with sporadic colorectal cancer (sCRC). It is believed that the metastatic process is related at least in part to a specific background of genetic alterations accumulated in cells from primary tumors, and the ability to detect such alterations is critical for the identification of patients with sCRC who are at risk of developing metastases.

METHODS

The authors used high-resolution, 500-K single nucleotide polymorphism arrays to identify copy number alteration profiles present at diagnosis in primary tumors from patients with metastatic (n = 23) versus nonmetastatic (n = 26) sCRC.

RESULTS

The results revealed a characteristic pattern of copy number alterations in metastatic sCRC tumors that involved losses of 23 regions at chromosomes 1p, 17p, and 18q, together with gains of 35 regions at chromosomes 7 and 13q.

CONCLUSIONS

In line with expectations, the copy number profile investigated involved multiple genes that were associated previously with sCRC (ie, SMAD2) and/or the metastatic process (ie, podocalyxin-like [PODXL]), and it also was associated with a poorer outcome. Cancer 2014;120:1948–1959. © 2014 American Cancer Society.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SUPPORT
  8. CONFLICT OF INTEREST DISCLOSURES
  9. REFERENCES

Sporadic colorectal cancer (sCRC) is the second leading cause of death from cancer in the Western world.[1] Up to 50% of all patients with sCRC eventually will develop metastases, and their 5-year overall survival rate is approximately 50% to 60%.[2] Once metastases (ie, mostly liver metastasis) have occurred, a complete cure is unlikely,[2] and up to two-thirds of patients with sCRC who die have evidence of liver metastasis.[3] Accumulating evidence indicates that sCRC metastasis may emerge in the context of a specific genetic tumor background associated or not with other genetic alterations, further affecting cellular control of growth and proliferation.[4] The discovery of those specific genetic alterations that would contribute toward identifying patients who are at risk of harboring or developing metastases could contribute significantly to the development of new strategies for the diagnosis and management of the diseases.

In recent years, multiple recurrent chromosomal abnormalities identified in primary tumors have been associated with metastatic CRC.[4, 5] Among others, these include numerical gains of the long arm of chromosome 8 (8q), 13q, and 20q and losses of the short arm of chromosome 1 (1p), 8p, 17p, 18q, and 22q. In a recent study, we used interphase fluorescence in situ hybridization (iFISH) to further establish that 17p deletion (del[17p]) involving the 17p11.2 breakpoint region and del(22q) were highly prevalent cytogenetic alterations among patients with primary sCRC who had synchronous liver metastasis.[6] However, the identification of specific genes targeted by such chromosomal alterations has proven difficult, partially because of technical issues. In recent years, the development of genome-wide approaches like high-density single nucleotide polymorphism (SNP) arrays has made it possible to identify small regions of chromosomal gains and losses with much greater resolution, down to 2.5 kb.[5]

In the current study, we used 500-K SNP arrays (mean distance between interrogated SNPs = 5.8 kb) to map genetic alterations that were present already at diagnosis in primary tumors from 49patients with sCRC (23 metastatic tumors vs 26 nonmetastatic tumors). Our main objective was to define differential copy number alteration (CNA) profiles between metastatic and nonmetastatic tumors for a better understanding of the genetics of the metastatic process in sCRC. To evaluate the reproducibility of the SNP-array results, we performed parallel iFISH analyses of the same tumor samples using 8 probes directed against the chromosomes that are most frequently altered in sCRC.

MATERIALS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SUPPORT
  8. CONFLICT OF INTEREST DISCLOSURES
  9. REFERENCES

Patients and Samples

Surgical specimens from 49 patients who were diagnosed with sCRC between June 2000 and September 2007 were collected prospectively and included in this study after informed consent was obtained from each patient. All patients underwent surgical resection of tumor tissues at the Department of Surgery of the University Hospital of Salamanca (Salamanca, Spain) and were diagnosed and classified according to World Health Organization criteria[7] before they received any treatment. The study was approved by the Ethics Committee of the University Hospital of Salamanca. The median follow-up at the time the study was closed was 97 months (range, 24-124 months). Of 49 patients, 23 (47%) had liver metastases (group 1) identified either at the time of colorectal surgery (n = 14) or during the first year after diagnosis (n = 9). The other 26 patients (53%) had tumors without metastatic dissemination during long a follow-up (group 2). The clinical and laboratory data from these patients are summarized in Table 1.

Table 1. Clinical and Biologic Characteristics of Patients With Metastatic (N = 23) Versus Nonmetastatic (n = 26) Sporadic Colorectal Carcinomaa
 No. of Patients (%)  
CharacteristicMetastatic sCRC, n = 23Nonmetastatic sCRC, n = 26PNo. of Patients (%): All sCRC, n = 49
  1. Abbreviations: 5-FU, 5-fluorouracil; CEA, carcinoembryonic antigen; NA, not applicable; NS, statistically nonsignificant (P > .05); OS, overall survival; sCRC, sporadic colorectal cancer.

  2. a

    Survival analysis was performed using the Kaplan-Meier method and the significance of the differences between survival curves were assessed using a 1-sided log-rank test.

Age: Median [range], y68 [48–80]67 [38–83]NS68 [38–83]
Sex    
Women8 (35)7 (27)NS15 (31)
Men15 (65)19 (73) 34 (69)
Tumor localization    
Rectum7 (31)3 (11)NS10 (20)
Left colon12 (52)14 (54) 26 (53)
Right colon4 (17)9 (35) 13 (27)
Histopathologic grade    
Well differentiated13 (56)19 (73)NS32 (65)
Moderately differentiated8 (35)6 (23) 14 (29)
Poorly differentiated2 (9)1 (4) 3 (6)
Histopathologic tumor classification    
pT10 (0)1 (4).051 (2)
pT20 (0)6 (23) 6 (12)
pT316 (70)11 (42) 27 (55)
pT47 (30)8 (31) 15 (31)
Histopathologic lymph node status    
pN07 (30)26 (100)< .00133 (67)
pN110 (45)0 (0) 10 (20)
pN26 (26)0 (0) 6 (12)
Tumor size: Median [range], cm5 [2.5–9.0]5 [2.5-14.0]NS5 [2.5-14.0]
Serum CEA: Median [range], ng/mL58.3 [0.8-1484.0]3.1 (0.6-18.0]< .0017 [0.6-1484.0]
Adjuvant treatment  NA 
No treatment2 (12)8 (35) 10 (26)
5-FU4 (25)12 (52) 16 (41)
Tegafur2 (12)3 (13) 5 (13)
Capacitabine2 (12)0 (0) 2 (5)
Capecitabine + oxaliplatin1 (7)0 (0) 1 (2)
Tegafur + irinotecan2 (12)0 (0) 2 (5)
5-FU, irinotecan, + bevacizumab3 (20)0 (0) 3 (8)
No. of deaths (%)18 (78)2 (8)< .00120 (41)
OS: Median [range], mo26 [1-96]Not reached< .001Not reached

After the histopathologic diagnosis was established, samples from representative areas of the primary tumors that exhibited macroscopic infiltration were either fixed in formalin and embedded in paraffin or frozen in liquid nitrogen and then stored either at room temperature or at −80°C, respectively. From these tissue samples, sections were cut from 3 different areas representative of the tumor tissue used for the SNP-array studies, and they were placed over poly L-lysine–coated slides. For SNP-array studies, tumor DNA was extracted from freshly frozen tumor samples that contained ≥65% epithelial tumor cells. In turn, normal DNA was extracted from matched peripheral blood (PB) leukocytes from the same patient. For both types of samples, DNA was extracted using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions.

SNP-Array Studies

Each DNA sample derived from primary tumors and normal PB leukocytes was hybridized to 2 different 250-K Affymetrix SNP mapping arrays (NspI and StyI SNP arrays; Affymetrix, Santa Clara, Calif). For this purpose, 250 ng of DNA per array were used according to the instructions of the manufacturer. Fluorescence signals were detected using the Affymetrix GeneChip Scanner 3000. Average genotyping call rates of 94.4%, 95.6%, and 97.6% were obtained from primary tumors corresponding to the metastatic tumors (group 1), the nonmetastatic tumors (group 2), and normal PB DNA samples, respectively.

To identify copy number changes throughout the whole tumor genome, the aroma.affymetrix algorithm was used, following the CRMA version 2 method in the R software package, as described elsewhere[8] (R Foundation for Statistical Computing, Vienna, Austria; available at: http://www.aroma-project.org, accessed March 18, 2014). The array normalization steps we used have been described previously by Munoz-Bellvis et al.[5] Next, data from the 250-K StyI and 250-K NspI arrays were integrated into a single database, and raw copy number values were calculated as transformed log2 values of the following ratios calculated for each patient: group 1 primary tumor/normal PB, group 2 primary tumor/normal PB, and group 1 primary tumor/group 2 primary tumor.

To identify DNA regions with similar copy number values, the circular binary segmentation method with default parameters was used exactly as implemented in the open-source DNAcopy Bioconductor software package (Bioconductor Project Core Team, Fred Hutchison Cancer Research Center, Seattle, Wash)[9]; the criteria used to define single-point copy number changes were based on P values ≤ .01 for ≥7 markers per DNA segment, and median segment values were assigned to each probe using smoothed values. For the identification of CNAs (gains or losses), a threshold was established based on the changes observed in the fluorescence intensity of sequential DNA segments for group 1 primary tumor versus PB, for group 2 primary tumor versus PB, and for group 1 versus group 2 primary tumor samples (log2 ratio cutoff values of >0.09 and <−0.09 for copy number gains and losses, respectively). For each tumor sample analyzed, the specific frequencies of both copy number gains and losses per SNP were established and plotted along individual chromosomes. Common altered regions were identified based on the empirical frequency distribution of gains and losses among group 1 and group 2 primary tumor samples, respectively, by grouping contiguous SNPs that had adjusted P values < .01 (false discovery rate [FDR] correction was based on the Benjamini and Hochberg procedure).[10] Moderated t test analyses were used to identify significant differences in copy numbers between metastatic and nonmetastatic samples (FDR-corrected P values < .05). Significant regions were defined as contiguous sets of at least 7 SNPs that differed significantly. A P value was assigned to the region by using the worst FDR-corrected P value among the SNPs included in that region. Genes coded in these regions were identified using the Ensembl release 53 tool (available at: http://www.ensembl.org, accessed March 18, 2014).

iFISH Studies

To evaluate the reproducibility of the SNP-array results and to assess the impact of background noise using this technique, iFISH analyses of the same tumor samples were performed in parallel, using 8 probes directed against an identical number of regions from 8 different human chromosomes (Table 2).

Table 2. Primary Colorectal Cancer (n = 49): Correlation Between the Numerical Changes Detected by Each Individual Fluorescence in Situ Hybridization Probe Used and the Copy Number Changes Identified for the Corresponding Single Nucleotide Polymorphisms (SNPs) Through SNP Array Studies
Chromosomal Region Identified by iFISH (iFISH Probe)DNA ProbeaR2P
  1. Abbreviations: iFISH, interphase fluorescence in situ hybridization; R2, coefficient of correlation; SG, SpectrumGreen probe; SO, SpectrumOrange probe;

  2. a

    The commercial source for all DNA probes was Vysis Inc. (Downers Grove, Ill).

7q31LSI D7S522 (SO)/CEP 7 (SG)0.60< .0001
8p12LSI LPL (SO)0.59< .0001
13q14LSI 13 (RB1) 13q14 (SO)0.58< .0001
14q32LSI IGH (SG)/BCL2 (SO)0.69< .0001
17p13.1LSI TP53 (17p13.1) (SO)0.65< .0001
18q21LSI IGH (SG)/BCL2 (SO)0.52.0001
20q13LSI ZNF217 (20q13.2) (SO)0.64< .0001
22q11LSI 22 (BCR) (SG)0.62< .0001

Additional Statistical Methods

For all continuous variables, mean values with standard deviations and ranges were calculated using the SPSS software package (SPSS version 15.0; SPSS Inc., Chicago, Ill; for dichotomic variables, frequencies were reported. To evaluate the statistical significance of differences observed between groups, the Student t test and the Mann-Whitney U test were used for continuous variables, depending on whether they did or did not display a normal distribution; for qualitative variables, the chi-square test was applied (Cross-Tab; SPSS Inc.). Survival analyses were performed using the Kaplan-Meier method, and the significance of differences in survival between different groups was assessed using a 1-sided log-rank test. Hazard ratios were calculated for each particular chromosomal region using a Cox proportional hazards model as implemented in the “survival” R package (R Foundation for Statistical Computing) using the Efrom approximation to control for ties in survival times. The assumption for proportional odds was tested using the partial likelihood-ratio test. Right-censored observations were considered those observations in which the event of death had not occurred at the date of last contact, ie, the date of the censored observation. These right-censored observations were taken into account in the estimation of the partial likelihood as follows:

  • display math

where δi = 1 if t(i) is a time-of-death event, and δi = 0 if t(i) is the time of censoring. Statistical significance was considered present if the P value (or the Pearson-corrected P value) was < .05. Only those iFISH probes with >12 SNPs localized in the iFISH mapped region (Table 2) were used for correlation studies, and the copy number status was identified by the SNP array (gain vs loss vs no change) for those SNPs localized at each iFISH region. P values < .01 were associated with statistical significance.

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SUPPORT
  8. CONFLICT OF INTEREST DISCLOSURES
  9. REFERENCES

Clinical and Biologic Characteristics of Liver Metastatic Versus Nonmetastatic sCRC

Overall, patients who had sCRC with liver metastases had a greater frequency of lymph node metastases (P < .001) and abnormally increased serum carcinoembryonic antigen levels (P < .001) than patients who had sCRC without metastases (Table 1). From the prognostic point of view, those who had sCRC with liver metastases also had more deaths associated with significantly shortened overall survival (median, 26 months vs not reached, respectively; P < .001). By contrast, no significant differences were observed between patients with versus without liver-metastatic sCRC regarding patient age, sex, tumor localization, histologic grade, or size.

Differences in Copy Number Alterations of Primary Tumors Between Metastatic and Nonmetastatic sCRC

Overall, many different CNAs were common in both metastatic and nonmetastatic primary sCRCs. These included deletions involving chromosomes 3p, 4, 5q, 8p, 10q, 15p, and 22q; whole chromosome 14 losses; and gains of chromosomes 1q, 2p, 3q, 5p, 6p, 8q, 11p, and 20q (Fig. 1). However, differences in the relative percentage of tumors that had some of these specific copy number changes also were observed between the 2 tumor groups (Tables 3 and 4). These mostly included differences in the frequency of deleted regions in chromosomes 1p, 17p, and 18q together with gains of chromosomes 7 and 13q, all of which had a greater incidence (P < .05) among primary tumors from patients who presented with metastatic versus nonmetastatic disease. A comparison between the specific copy number changes that were present differentially in both groups of tumors revealed a total of 58 altered chromosomal regions, all of which were identified more frequently (P < .05) in metastatic tumors versus nonmetastatic tumors. From these, 23 copy number changes corresponded to interstitial deletions at 1p36 (n = 1), 1p33 (n = 1), 17p12 (n = 1), 18q11 (n = 1), 18q11 (n = 3), 18q21(n = 7), 18q12 (n = 3), 18q21 (n = 9), 18q22 (n = 4), and 18q23 (n = 1); and 35 corresponded to regions of gain at 7p22 (n = 6), 7p15 (n = 1), 7p14 (n = 1), 7p13 (n = 1), 7p12 (n = 3), 7p11 (n = 1), 7q11 (n = 2), 7q22 (n = 1), 7q32 (n = 3), 7q33 (n = 1), 7q34 (n = 2), 13q22 (n = 2), 13q31 (n = 4), 13q32 (n = 2), 13q33 (n = 3), and 13q34 (n = 2).

Table 3. Chromosomal Regions That Exhibited a Significantly Different Frequency of Copy Number Losses in Metastatic (n = 23) Versus Nonmetastatic (n = 26) Sporadic Colorectal Cancer Tumors Once Genotyped on the Affymetrix 500K Single Nucleotide Polymorphism-Array Platform and Their Impact on Overall Patient Survivala
   Percentage of LossesMedian OS, mo  
Lost Chromosomal Region (bp)No. of SNPsMean Fold ChangeNonmetastatic Tumors, %Metastatic Tumors, %PCases With Chromosomal LossCases Without Chromosomal LossPHRP for HR
  1. Abbreviations: bp, base pairs; HR, hazard ratio; NS, statistically nonsignificant (P > .05); OS, overall survival; SNPs, single nucleotide polymorphisms.

  2. a

    Survival analysis of was performed using the Kaplan-Meier method, and the significance of differences between survival curves was assessed using a 1-sided log-rank test. The HR was computed as the exponential of the Cox regression model coefficient for each particular lost chromosomal region. The right censored observations were considered those observations in which the event of death had not occurred by the last contact, taken as the date of the censored observation.

1p36 (20,557,292-20,618,072)10−0.742765.053995NS2.3NS
1p33 (49,620,448-50,587,318)67−0.812761.054093.052.4.05
17p11 (13,916,156-14,521,723)130−1.203187.053797.023.2.02
18q11 (21,640,528-21,805,604)36−1.324283.043997NS2.5NS
18q11 (22,569,906-26,178,219)566−1.503883.053997NS2.5NS
18q11 (26,188,429-27,728,214)311−1.523387.053797.013.6.02
18q12 (27,969,319-29,628,991)321−1.683187.023796.0054.1.01
18q12 (29,673,911-40,483,322)1751−1.563887.023797.013.6.02
18q12 (40,571,702-41,726,633)349−1.503887.033797.0074.4.02
18q21 (42,119,752-44,304,862)360−1.434291.033797.0074.3.02
18q21 (44,765,784-44,818,557)8−1.473887.043796.023.4.03
18q21 (44,879,066-45,497,910)80−1.554291.033799.0016.8.01
18q21 (45,614,026-46,598,496)308−1.494287.043797.0074.3.02
18q21 (46,912,256-53,316,730)1395−1.584278.043997.042.7.05
18q21 (53,658,175-55,508,632)324−1.324278.053796NS2.1NS
18q21 (55,803,729-57,740,851)364−1.544278.053797.042.7.05
18q21 (57,744,781-58,390,267)110−1.514293.053797.042.7.05
18q21 (58,432,935-64,897,804)830−1.623583.023797.042.7.05
18q22 (64,908,767-70,255,162)1210−1.593583.033797.023.0.04
18q22 (70,331,521-70,892,013)101−1.624291.023799.014.0.03
18q22 (71,062,175-71,979,380)313−1.524296.0439101.0026.3.01
18q22 (71,991,789-72,194,339)32−1.5342100.0239101.0026.3.01
18q23 (73,115,995-75,227,087)289−1.594296.033799.014.0.03
Table 4. Chromosomal Regions That Exhibited a Significantly Different Frequency of Copy Number Gains in Metastatic (n = 23) Versus Nonmetastatic (n = 26) Sporadic Colorectal Cancer Tumors Once Genotyped on the Affymetrix 500K Single Nucleotide Polymorphism-Array Platform and Their Impact on Overall Patient Survivala
   Percentage of Gains Median OS, mo   
Gained Chromosomal Region (bp)No. of SNPsMean Fold ChangeNonmetastatic Tumors, %Metastatic Tumors, %PNo. With Chromosomal GainNo. Without Chromosomal GainPHRP for HR
  1. Abbreviations: bp, base pairs; HR, hazard ratio; NS, statistically nonsignificant (P > .05); OS, overall survival; SNPs, single nucleotide polymorphisms.

  2. a

    Survival analysis was performed using the Kaplan-Meier method, and the significance of differences between survival curves was assessed using a 1-sided log-rank test. The HR was computed as the exponential of the Cox regression model coefficient for each particular lost chromosomal region. The right censored observations were considered as those observations in which the event of death had not occurred by the last contact, taken as the date of the censored observation.

7p22 (149,081-292,350)361.432752.053784NS2.0NS
7p22 (761,753-1,712,875)761.692752.053384NS2.4NS
7p22 (1,808,417-1,887,352)91.482757.053784NS2.1NS
7p22 (5,048,563-5,622,361)431.413161.053684NS1.9NS
7p22 (6,389,871-6,578,919)291.503161.043784NS1.9NS
7p22 (6,689,136-7,007,136)81.582757.053786NS2.2NS
7p15 (24,587,207-26,302,447)471.152361.053388.052.5.05
7p14 (29,780,510-31,105,857)2211.092361.043388.052.5.05
7p13 (43,303,432-46,005,205)3661.281957.053388.052.5.05
7p12 (47,208,833-48,137,406)1941.162357.043786NS2.0NS
7p12 (50,165,544-50,224,267)161.092761.043689NS2.2NS
7p12 (50,325,753- 51,266,512)2381.162357.042886.032.7.029
7p11 (54,502,479-57,882,795)4161.201952.052688.013.1.01
7q11 (62,935,623-64,090,571)1261.011548.052886.042.6.04
7q11 (76,348,155-77,513,727)770.821948.052484.023.0.02
7q22 (98,997,269-99,301,754)271.031948.042488.013.3.009
7q32 (127,629,005-127,818,534)440.912352.053388.042.6.04
7q32 (128,535,614-29,234,535)770.932352.053388.042.6.04
7q32 (130,847,348-131,391,475)1470.841952.052889.023.0.02
7q33 (134,493,159-134,546,072)260.981952.042889.023.0.02
7q34 (137,691,869-139,874,587)2890.952348.043288.042.6.04
7q34 (142,623,691-142,837,893)251.231952.032889.023.0.02
13q22 (73,603,130-73,627,939)101.893574.043796.013.2.02
13q22 (74,962,410-76,366,765)421.473174.043796.032.8.03
13q31 (79,803,335-79,845,948)111.652774.023695.013.2.02
13q31 (90,792,026-90,867,353)72.032774.053695.0083.6.01
13q31 (92,522,499-92,673,304)311.622370.033693.0073.5.01
13q31 (92,852,858-92,917,452)231.592365.053790.022.9.02
13q32 (93,877,956-96,656,619)4861.462365.053789NS2.4NS
13q32 (97,234,644-100,306,243)5771.522765.043988NS2.1NS
13q33 (102,999,018-105,259,754)5691.462761.053793.052.4.05
13q33 (105,618,714-108,251,446)6971.542765.053990NS2.2NS
13q33 (108,337,015-109,286,583)2041.532765.053792.032.7.04
13q34 (109,466,205-112,946,386)6091.581961.043689.042.6.04
13q34 (112,963,304-114,092,980)1131.851552.053788NS2.1NS
image

Figure 1. Copy number changes were detected in patients with metastatic (n = 23) versus nonmetastatic (n = 26) sporadic colorectal cancer genotyped on the Affymetrix 500-K single nucleotide polymorphism-array platform. This summary plot illustrates the frequency of copy number gains (plotted in red above zero values in the x-axis) and losses (plotted in green below zero values in the x-axis) identified in patients with metastatic primary sporadic colorectal tumors (dark colors) and in patients with nonmetastatic tumors (light color) for the whole human genome. Gains of chromosomes 7 and 13 and losses of chromosomes 1p, 17p, and 18 (indicated by arrows) were observed more frequently (P < .05) in metastatic tumors versus nonmetastatic tumors.

Download figure to PowerPoint

Impact of CNAs on Overall Survival in Patients With sCRC

From the 23 interstitial deletions described above that were observed more frequently in metastatic versus nonmetastatic primary sCRC tumors, 19 deletions (83%) were associated with a poorer outcome (P < .05) (Table 3); furthermore, 3 of these 19 interstitial deletions, all of which were localized to chromosome 18 (base pairs [bp], from start to end)—18q21 (bp 44,879,066-45,497,910), 18q22 (bp 71,062,175-71,979,380 and bp 71,991,789-72,194,339), were strongly associated with poorer overall survival (P = .01; hazard ratio, >6) (Table 3). Regarding chromosomal gains, 23 of 35 gains (66%) also had an impact on the overall survival of patients with sCRC (Table 4); of these, gains of chromosome 7p11 (bp 54,502,479-57,882,571) and chromosome 13q31 (bp 90,792,026-90,867,535 and bp 92,522,499-92,673,304) had a stronger association with overall survival (P = .01; hazard ratio, >3) (Table 4).

Correlation Between the Chromosomal Alterations Detected by SNP Arrays and iFISH

To evaluate the consistency of chromosomal alterations identified by the SNP arrays, iFISH analyses were performed in parallel for a total of 8 chromosomal regions. Overall, our results revealed a high degree of correlation (r2: mean, 0.61; range, 0.52-0.69) between both methods (Table 2).

Genes With an Adverse Impact on Patient Outcome

Genes coded within the chromosomal regions that exhibited a greater frequency of alteration in primary metastatic tumors versus primary nonmetastatic tumors, in association with an adverse impact on overall survival, included a total of 52 and 26 genes mapped in the deleted and gained chromosomal regions, respectively (Table 5). It is noteworthy that, almost 50% of the 52 deleted genes were previously associated with colorectal cancer (n = 24), and another 50% (n = 24) have been related to the metastatic process. Concerning genes involved in the gained chromosomal regions, it was reported previously that 15 of 26 genes were involved in colorectal cancer (eg glioblastoma amplified sequence [GBAS]) and/or the metastatic process (eg, podocalyxin-like [PODXL]). Figure 2 indicates that such deleted (eg genes coded at chromosome 18q21) or gained (eg genes coded at chromosomes 7p11 and 7q22) cancer-associated genes are directly related to well established genetic markers of sCRC, such as epidermal growth factor receptor (EGFR), the SMAD family of genes, and the deleted in colorectal carcinoma (DCC) gene.[35, 45, 46] In addition, in the literature, it was reported previously that several of these genes exhibited deletion and loss of expression or gain and overexpression in metastatic sCRC (Fig. 2).

Table 5. Chromosomal Regions Most Frequently Altered in Metastatic (n = 23) Versus Nonmetastatic (n = 26) Primary Sporadic Colorectal Cancer Tumors Genotyped on the Affymetrix 500K Single Nucleotide Polymorphism-Array Platforma
Altered Chromosomal Region: bpRegion Length, bpChromosomal BandGene Listb
  1. Abbreviations: bp, base pairs; Chr, chromosome.

  2. a

    Only cancer-associated genes coded in the altered chromosomal regions that had an impact on overall survival are listed in the table.

  3. b

    Genes that have been associated with colorectal cancer are indicated in bold, and genes commonly associated with the metastatic process are underlined.

Deletions   
Chr1: 49,620,448-50,587,318966,8701p33ELAVL4
Chr17: 13,916,156-14,521,723605,56717p12COX10
Chr18: 26,188,429-27,728,2141,539,78518q11.2
Chr18: 27,969,319-29,628,9911,659,67218q12.1DSC3, DSC2, DSC1, DSG3, DSG2, TTR, RNF125
Chr18: 29,673,911-40,483,32210,809,41118q12.2MAPRE2, ZNF24, GALNT1, SLC39A6, BRUNOL4, PIK3C3
Chr18: 40,571,702-41,726,6331,154,93118q12.3SYT4
Chr18: 42,119,752-44,304,8622,185,11018q21.1SETBP1, SLC14A2, ATP5A1, KIAA1632
Chr18: 44,765,784-44,818,55752,77318q21.1
Chr18: 44,879,066-45,497,910618,84418q21.1SMAD2
Chr18: 45,614,026-46,598,496984,47018q21.1SMAD7
Chr18: 46,912,256-53,316,7306,404,47418q21.2LIPG, MBD1, ME2, SMAD4, DCC, MBD2, POLI, TCF4
Chr18: 55,803,729-57,740,8511,937,12218q21.32MALT1, GRP, PMAIP1
Chr18: 57,744,781-58,390,267645,48618q21.32
Chr18: 58,432,935-64,897,8046,464,86918q21.33PHLPP1, BCL2, SERPINB5, SERPINB12, SERPINB13, SERPINB4, SERPINB11, SERPINB3, SERPINB7, SERPINB2, SERPINB10, SERPINB8, CDH7, CDH19
Chr18: 64,908,767-70,255,1625,346,39518q22.2CD226, SOCS6
Chr18: 70,331,521-70,892,013560,49218q22.3
Chr18: 71,062,175-71,979,380917,20518q22.3CYB5A
Chr18: 71,991,789-72,194339202,55018q22.3CNDP2
Chr18: 73,115,995-75,227,0872,111,09218q23GALR1
Gains   
Chr7: 24,587,207-26,302,4471,715,2407p15.3DFNA5, NFE2L3, HNRNPA2B1
Chr7: 29,780,510-31,105,8571,325,3477p14.3SCRN1, FKBP14, PLEKHA8, NOD1, GARS, AQP1, GHRHR, ADCYAP1R1
Chr7: 43,303,432-46,005,2052,701,7737p13
Chr7: 50,325,753- 51,266,512940,7597p12.1IKZF1, DDC, GRB10, COBL
Chr7: 54,502,479-57,882,7953,380,3167p11.2SEC61G, EGFR, LANCL2, VOPP1, PSPH, GBAS
Chr7: 62,935,623-64,090,5711,154,9487q11.21
Chr7: 76,348,155-77,513,7271,165,5727q11.23
Chr7: 98,997,269-99,301,754304,4857q22.1CYP3A5, CYP3A7
Chr7: 127,629,005-127,818,534189,5297q32.1
Chr7: 128,535,614-129,234,535698,9217q32.1
Chr7: 130,847,348-131,391,475544,1277q32.3PODXL
Chr7: 134,493,159-134,546,07252,9137q33
Chr7: 137,691,869-139,874,5872,182,7187q34
Chr7: 142,623,691-142,837,893214,2027q34TRPV5, KEL
Chr13: 73,603,130-73,627,93924,80913q22.1
Chr13: 74,962,410-76,366,7651,404,35513q22.2
Chr13: 79,803,335-79,845,94842,61313q31.1
Chr13: 90,792,026-90,867,35375,32713q31.3
Chr13: 92,522,499-92,673,304150,80513q31.3
Chr13: 92,852,858-92,917,45264,59413q31.3
Chr13: 102,999,018-105,259,7542,260,73613q33.1
Chr13: 108,337,015-109,286,583949,56813q33.3
Chr13: 109,466,205-112,946,3863,480,18113q34
image

Figure 2. Copy number changes detected in primary tumors from patients with metastatic (n = 23) versus nonmetastatic (n = 26) sporadic colorectal cancer (sCRC) were genotyped on the Affymetrix 500-K single nucleotide polymorphism (SNP)-array platform corresponding to (A) gains of the 7p11 and 7q22 chromosome (Chr.) 7 regions and (B) chromosome 18q21 losses. In both A and B, summary plots illustrating the mean (smoothed) fluorescence intensity log2 values of primary tumor/normal peripheral blood (PB) ratios for the chromosomal regions that had copy number (A) gains and (B) losses are displayed as red lines for nonmetastatic tumors and blue lines for metastatic tumors (ie, plotted above zero values in the x-axis); confidence intervals are indicated as background gray areas. Genes coded at the 7p14.3 (AQP1[11, 12]), 7p11.2 (EGFR[13, 14] and PSPH[15, 16]), 7q22.1 (CYP3A5 and CYP3A7[17, 18]), 7q32.3 (PODXL[19-21]), 18q12.1 (DSC3,[22-24] DSC2,[22, 25] DSC1,[22, 24] and DSG2[26]), 18q21.1 (SMAD2[27, 28] and SMAD7[29]), 18q21.2 (SMAD4,[30-33] DCC,[32-35] and TCF4[36-38]), 18q21.32 (GRP[39, 40]), and 18q21.33 (PHLPP1,[16, 41] BCL2,[42, 43] SERPINB5,[44] and SERPINB2[44]) chromosomal regions that were previously associated with sCRC tumor metastases are listed in italics. Red arrows represent gains and overexpression, and green arrows represent deletion and losses of expression of a given gene, as described in published data sets.

Download figure to PowerPoint

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SUPPORT
  8. CONFLICT OF INTEREST DISCLOSURES
  9. REFERENCES

Most sCRC deaths are associated with metastatic dissemination of the primary tumor. New advances in genetic profiling of sCRCs suggest that primary tumors may already contain tumor cells with metastatic potential.[4] Herein, we describe a comprehensive, detailed map of CNAs present in primary tumors from patients with metastatic versus nonmetastatic sCRC who were followed for a relatively long time, as assessed by high-resolution, 500-K SNP arrays and confirmed by iFISH studies. A high correlation also was observed between the SNP-array results and iFISH analyses performed on the same series of primary tumor samples with regard to the most commonly deleted (eg 17p and 18q) and gained (eg 7 and 13q) chromosomal regions, in line with in our previous findings.[5, 47] To our knowledge, this is the most extensive study in which high-resolution SNP arrays have been used to define and compare the CNA profiles of both groups of sCRC tumors. Overall, we observed that there were many CNAs in common between the 2 groups of tumors, including losses of 3p, 4, 5q, 8p, 10q, 14q, 15q, and 22q and gains of 1q, 2p, 3q, 5p, 6q, 8q, 11p, and 20q. However, in addition to the CNAs that were common to both metastatic and nonmetastatic tumors, multiple CNAs occurred with greater frequency in primary tumors from patients with metastatic disease. These latter CNAs mainly involved regions lost at chromosomes 1p, 17p, and 18q together with regions of gain at chromosomes 7 and 13q.

Previous studies based on lower resolution cytogenetic analyses of primary tumors from patients with metastatic sCRC[48, 49] demonstrated frequent alterations of different regions at chromosomes 1p, 7, 13q, 17p, and 18q.[48, 50-53] Herein, we identified 58 different regions in these same chromosomes that were altered preferentially in primary metastatic tumors versus primary nonmetastatic tumors. Furthermore, approximately 75% of these regions also were associated with poorer overall survival. Altogether, these observations point out the potential relevance of genes coded in these chromosomal regions in the development of sCRC metastasis.

Loss of chromosome 1p is well documented in sCRC.[52] Allelic losses of chromosome 1p have been associated with increased tumor aggressiveness and reduced patient survival.[50, 54] In line with our findings, Ghadimi et al observed a greater frequency of chromosome 1p32-ter loss in metastatic versus nonmetastatic cancer (61% vs 11% of patients) using comparative genomic hybridization.[53] Our results indicate a particularly high frequency of losses involving the 1p33 chromosomal locus, which may harbor relevant tumor suppressor genes like the embryonic lethal, abnormal vision (ELAV)-like neuron specific RNA binding protein 4 (ELAVL4) gene. Of note, Stawski et al recently reported significantly decreased expression of ELAVL4 among male patients with meningiomas who carried deletions at D1S162, supporting the hypothesis that ELAVL4 may act as a tumor suppressor gene in meningiomas.[55]

In line with our previous findings using iFISH,[6] del(17p) also was observed with significantly greater frequency among patients with metastatic tumors versus patients with nonmetastatic tumors and was associated with a poorer prognosis. These results are also in line with the observations of other groups that evaluated the association between del(17p) and outcome in patients with stage II and stage III sCRC[56]; however, apart from tumor protein 53 (TP53), the specific relevant genes involved remain to be identified.

Del(18q) has long been observed in sCRC using a broad panel of techniques that vary from conventional cytogenetics[57] and iFISH[4] to comparative genomic hybridization, comparative genomic hybridization arrays, and SNP arrays.[47] Within the long arm of chromosome 18, the 18q21 cytoband was the most frequently altered (range, 50%-70% of CRCs) in those studies.[4, 47, 57] This region contains both the DCC and SMAD genes, which are well established genetic biomarkers of sCRC and typically are associated with advanced disease. Currently, it is well establish that DCC plays a role in inducing apoptosis, whereas SMAD family member 4 (SMAD4 [DPC4]) encodes intracellular transducers of the transforming growth factor-β apoptosis pathway.[35, 58, 59] It is worth noting that deletions at other specific subregions of chromosome 18q also have been associated with a poor prognosis in advanced sCRC.[60] Thus, Liu et al observed that loss of 18q12-qter is an independent prognostic marker that is detected more frequently in stage III/IV (vs stage I) tumors.[61] Similarly, Poulogiannis et al used DNA microarrays to identify DNA copy number losses at 18q12.2 involving a single gene—ie, the BRUNOL4 (Bruno-like 4 splicing factor) gene—as an independent, adverse prognostic factor.[62] In addition to the BRUNOL4 gene, we also observed 5 other recurrently altered genes in this same chromosomal region: 2 of those genes were associated previously with sCRC (eg microtubule-associated protein RP/EB family member 2 [MAPRE2] and zinc finger protein 24 [ZNF24]), another gene was associated recurrently with the metastatic process (eg phosphatidylinositol 3-kinase, catalytic subunit type 3 [PIK3C3]), and the remaining 2 genes were related previously to other malignancies like ovarian cancer (UDP-N-acetyl-α-D-galactosamine, polypeptide N-acetylgalactosaminyltransferase 1 [GALNT1])[63] and esophageal squamous cell carcinoma (solute carrier 14-urea transporter, member 2 [SLC14A2]).[64] It is noteworthy that the greater number of 18q22 genes identified in our series versus that reported by Poulogiannis et al potentially may have a technical explanation because of the use of a higher resolution array (down to 2.5 kb vs 1 Mb) in the 2 series. In addition, the greater number of genes identified also may be related to the loss of other genes that are codeleted in chromosome 18q12. In line with this hypothesis, Knosel et al reported an association between del(18q23) observed with comparative genomic hybridization and both lymphatic and liver metastasis.[65] Storojeva et al used real-time quantitative polymerase chain reaction analysis to establish that del(18q22) involving DNAX accessory molecule 1 (DNAM-1), suppressor of cytokine signaling 6 (SOCS6), and α-7 nictonic receptor (ACHR-7) also was associated with a worse prognosis among patients with sCRC,[66] and we also identified other deleted regions at chromosome 18q (ie, 18q11, 18q12, 18q21, 18q22, and 18q23) that were associated with shorter overall survival.

It is noteworthy that we identified 3 regions of gain in chromosome 7 associated with poor outcome and metastasis: the 7p11, 7q22 and 7q23 chromosomal regions. The former encodes a well establish oncogene, ie, the EGFR gene, which is involved in cell proliferation with antiapoptotic properties.[46, 67] It is noteworthy that different studies have reported coamplification with EGFR of several other adjacent genes, such as the SEC γ subunit (SEC61G) and LanC lantibiotic synthase component C-like 2 (LANCL2) genes in breast cancer[68] and the LANCL2, vesicular-overexpressed in cancer prosurvival protein 1(VOPP1), phosphoserine phosphatase (PSPH), and glioblastoma amplified sequence (GBAS) genes in glioblastoma.[69] In turn, the 7q22 chromosomal region contains the cytochrome p450 family 3, subfamily A, polypeptide 5 (CYP3A5) and CYP3A7 genes, which encode for the P450 enzymes involved in drug metabolism. In line with our findings, overexpression of the CYP3A5 and CYP3A7 genes has been detected recurrently in primary sCRC tumors[17] and metastatic prostate cancer.[70] Finally, special attention should be paid to the PODXL gene coded at chromosome 7q32, because PODXL is involved in cell adhesion, which has been associated with the metastatic process[71] and also with a poor prognosis among patients with sCRC.[21]

With regard to chromosome 13q, in the current study, we identified 9 regions in which 13q gains were associated with metastatic tumors and poorer overall survival. However, we did not identify any coded genes in these regions that were related previously with the oncogenic process; for instance, a clear association between gains of these chromosomal regions and both tumor metastasis and a poor prognosis have been reported previously by our group and others.[5, 47, 62]

In summary, in this study, we used high-resolution SNP-array techniques to describe in detail the genetic alterations most frequently associated with metastatic sCRC and the potential genes involved. Further gene expression profiling and functional studies focused on such genes encoded at chromosomes 1p, 7, 13q, 17p, and 18q and their potential interactions are needed to establish their precise pathogenic role in the metastatic process in patients with sCRC.

FUNDING SUPPORT

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SUPPORT
  8. CONFLICT OF INTEREST DISCLOSURES
  9. REFERENCES

This work was supported in part by grants from the Instituto de Salud Carlos III (ISCIII), Ministerio de Sanidad y Consumo, Madrid, Spain (PI12-02053-FIS); Consejeria de Sanidad, Junta de Castilla y Leon, Valladolid, Spain (BIO-SA02-13); RTICC (RD12-0020-0035-FEDER, RD12-0036-0048-FEDER); Fundación Memoria de Don Samuel Solórzano Barruso, Salamanca, Spain; and Caja de Burgos (Obra Social), Burgos. Dr. Sayagués is supported by a grant (CP05-00321) from the ISCIII, Ministerio de Ciencia e Innovación, Madrid, Spain.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. MATERIALS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. FUNDING SUPPORT
  8. CONFLICT OF INTEREST DISCLOSURES
  9. REFERENCES