Topoisomerase 1(TOP1) gene copy number in stage III colorectal cancer patients and its relation to prognosis

Abstract

Purpose

A Topoisomerase 1 (Top1) poison is frequently included in the treatment regimens for metastatic colorectal cancer (mCRC). However, no predictive biomarkers for Top1 poisons are available. We here report a study on the TOP1 gene copy number in CRC patients and its association with patient prognosis and tumor cell proliferation.

Experimental design

The study included TOP1 and CEN‐20 fluorescence in situ hybridization (FISH) analyses on formalin fixed paraffin embedded (FFPE) tissue sections from 154 stage III CRC chemonaïve patients. The frequencies of aberration in the TOP1 gene copy number, the CEN‐20 copy number and the TOP1/CEN‐20 ratio were analyzed and associated with overall survival (OS), time to recurrence (TTR) and in a subgroup analysis of rectal cancer patients only with time to local recurrence (LR in RC). Moreover, the TOP1 and CEN‐20 copy numbers were correlated with the tumor Ki67 proliferation index.

Results

35.7% of the tumors had an increased TOP1 copy number above 4n gene copies per cell and 28.6% and 9.7% had a TOP1/CEN‐20 ratio ≥1.5 or ≥2.0, respectively. The TOP1 copy number and the TOP1/CEN‐20 ratios were separately added into multivariate analyses as continuous variables, in which also age, gender, primary tumor location and Ki67 status were added as covariates. In contrast to the TOP1/CEN‐20 ratio, the TOP1 copy number was significantly associated with OS (HR: 0.62; 95% CI: 0.42–0.90; p = 0.01). Neither the TOP1 copy number nor the ratio was significantly associated with TTR and only the TOP1/CEN‐20 ratio was significantly associated with LR in RC (HR: 0.25; 95% CI: 0.08–0.83; p = 0.02). No significant correlation was found between the TOP1 copy number and proliferation, while a weak and inverse correlation between the CEN‐20 copy number and proliferation was observed.

Conclusions

This study showed that increased TOP1 gene copy numbers are frequent findings in cancer cells in stage III CRC tumors but unrelated to the proliferative status of the tumors. The association with prognosis is important to consider when planning and analyzing future studies investigating TOP1 as a potential predictive biomarker for Top1 poisons.

Highlights

• ► 35.7% of CRC stage III tumors have TOP1 copy number.
• ► >4n.
• ► TOP1 copy number was associated with OS (HR: 0.62; p = 0.01).

Abbreviations

CEN-20

centromere 20

CC

colon cancer

CI

confidence interval

CRC

colorectal cancer

DFS

disease free survival

5FU

5-fluorouracil

FFPE

formalin-fixed paraffin-embedded

FISH

fluorescence in situ hybridization

GISTIC

Genomic Identification of Significant Targets in Cancer

HR

hazard ratio

Ki67

Antigen KI-67

LR in RC

local recurrence in rectal cancer

mCRC

metastatic CRC

PFS

progression free survival

OS

overall survival

REMARK

REporting recommendations for tumor MARKer prognostic studies

RC

rectal cancer

SN38

7-Ethyl-10-hydroxy-camptothecin

SD

standard deviation

TTR

time to recurrence

Top1

DNA Topoisomerase 1 protein

TOP1

DNA Topoisomerase 1 gene

Top2α

DNA Topoisomerase II alpha protein

TOP2A

DNA Topoisomerase 2 gene

1. Introduction

In early as well as late stage colorectal cancer (CRC) there is a need to identify and validate predictive markers, which will allow for a personalized treatment approach where the individual patient upfront receives the treatment with the highest likelihood of a beneficial effect.

Most predictive biomarkers also carry a prognostic value, which may disturb the interpretation of the clinical value of such biomarkers (Nielsen and Brunner, 2011). One good example of this is the gene for topoisomerase IIα (Top2α), TOP2A. Breast cancer patients with TOP2A gene aberrations (amplifications or deletions) have a significantly worse prognosis than patients with normal TOP2A copy numbers (Knoop et al., 2005). At the same time, these patients have an increased likelihood of obtaining a beneficial effect from treatment with Top2α poisons. For example, treating TOP2A gene aberrated breast cancer patients with adjuvant chemotherapy containing an anthracycline (Top2α poison) will significantly increase disease free survival (DFS) and overall survival (OS) of these patients and the TOP2A aberrant patients will now have DFS and OS comparable to TOP2A normal patients treated with an anthracycline (Knoop et al., 2005; Di et al., 2011; Nielsen and Brunner, 2011). However, without a control group receiving non‐anthracycline containing chemotherapy, the predictive value of the TOP2A copy number determinations for anthracycline treatment could be overlooked (Tubbs et al., 2009; Harris et al., 2009; Nielsen and Brunner, 2011).

Topoisomerase 1 (Top1) has recently been suggested as a predictive biomarker for the effect of irinotecan, a Top1 poison frequently used in combination therapy of metastatic CRC (mCRC) (Braun et al., 2008). In one study (Braun et al., 2008), the authors reported a significant association between high Top1 immunoreactivity as determined in formalin‐fixed paraffin‐embedded (FFPE) sections from the primary tumor and a benefit (i.e. longer progression free survival (PFS)) for patients receiving irinotecan or oxaliplatin based chemotherapy but not for patients receiving 5‐fluorouracil (5FU) alone. In a retrospective case control study (Kostopoulos et al., 2009) investigating Top1 immunoreactivity in FFPE sections from Dukes' stage B and C CRC tumors receiving adjuvant treatment with 5FU or 5FU plus irinotecan, the investigators reported that patients with Top1 high protein expressing tumors had improved OS. However, because of short follow‐up time on patients receiving adjuvant 5FU plus irinotecan, the study had limitations in differentiating between the potential prognostic and/or a predictive value of Top1 (Kostopoulos et al., 2009). The authors raised the question whether high Top1 protein expression was a favorable prognostic marker by itself, or whether the prognosis could be further improved by adding irinotecan to patients with Top1 high expression tumors in the adjuvant treatment of CRC.

Reproducible quantification of protein levels in FFPE tissue may be difficult to obtain, due to variation in the handling of samples and justify the search for alternative methods to establish relationships between Top1 and the clinical response to irinotecan. Counting gene copy numbers with fluorescence in situ hybridization (FISH) technology is appealing, since DNA is more stable and is not as sensitive to the preanalytical sampling, handling, processing and storage of the samples, as is protein. Furthermore, the gene signals are easily quantified on a cell‐to‐cell basis.

We recently reported on TOP1 FISH analyses on 50 FFPE primary CRC tissues and 10 human CRC cell lines (Romer et al., 2012). A reference probe representing centromere 20 (CEN‐20) was also included. More than 80% of the clinical samples (CRC stage III) demonstrated an increased TOP1 gene copy number as compared with the mean TOP1 gene copy number + 3 standard deviations (SD) in unaffected colorectal mucosa. Approximately 2/3 of the samples had an increased TOP1/CEN‐20 ratio as compared with the non‐affected mucosa. Furthermore, our study on cell lines suggested an association between TOP1 gene copy number or the TOP1/CEN‐20 ratio with sensitivity to the Top1 poison irinotecan (SN38), but not to oxaliplatin.

However, in order to bring TOP1 copy number determination forward as a predictive biomarker for Irinotecan treatment a number of additional analyses have to be performed. We here report on a study that aims to answer three questions: 1) Can the high frequency of TOP1 gene aberrations found in our previous study (Romer et al., 2012) be confirmed in a larger CRC patient population, since a certain frequency is required to make it clinically relevant as a predictive biomarker for irinotecan treatment? 2) Does the TOP1 copy number associate with patient prognosis? i.e. the course of the disease in untreated patients, since as mentioned above, a prognostic component of a biomarker requires a control group in order to study its predictive value, and 3) Is there an association between TOP1 copy number and cancer cell proliferation? This is important because TOP1 gene copy numbers in tumors may be overestimated if more cells are in the DNA synthesis phase of the mitosis due to a higher proliferation rate in the cells.

The present study includes TOP1 and CEN‐20 FISH analyses on FFPE tissue sections from 154 stage III CRC patients who did not receive adjuvant chemotherapy. The TOP1 copy number, the CEN‐20 copy number and the TOP1/CEN‐20 ratios were analyzed and associated with overall survival (OS), time to recurrence (TTR) and in a subgroup analysis of rectal cancer patients with time to local recurrence (LR in RC). The TOP1 and CEN‐20 copy number per cell were further correlated with the proliferation index (Ki67 immunoreactivity) of the individual tumors.

The REMARK guidelines (McShane et al., 2006) were followed whenever applicable.

2. Materials and methods

2.1. Patients and tumor tissue sections

From a total of 774 CRC patients (stage I–IV) enrolled in the RANX05 randomized clinical study (Nielsen et al., 1998) from April 1991 until August 1993 in Denmark, we selected all stage III patients (n = 223), where it was possible to obtain FFPE tumor tissue (n = 186). Out of the corresponding 186 FFPE blocks, nine did not contain any tumor tissue and 23 were not suitable for FISH analyses due to presence of large amounts of necrotic tissue, insufficient fixation etc. This left 154 FFPE samples for this study (Figure 1) consisting of 93 male and 61 female patients. The median age at surgery was 69 years (range 33–86 years). Eighty‐two patients were registered with colon cancer (CC) and 72 with rectal cancer (RC). Fifty of these samples were also included in a previous paper describing the basics of TOP1 copy number determinations (Romer et al., 2012). The distribution of the 154 patients by age, gender and tumor localization (RC or CC) did not vary significantly from that observed in the 223 stage III cancer patients in the original study.

Figure 1.

CONSORT flow diagram describing the selection of the samples analyzed in the study.

The study was conducted in accordance with the Helsinki II Declaration and was approved by the National Board of Health, Denmark (2760‐419‐1989), the Data Protection Agency (1991‐1110‐751) and The Ethics Committee for Copenhagen and Frederiksberg Hospitals. (KF 01‐2045/91). All patients had histologically verified stage III adenocarcinomas of the colon or rectum. The study was performed at a time when adjuvant radio‐ and/or chemotherapy was not part of the standard therapy in Denmark to patients with CRC (all stages), including metastatic disease. The patients were all randomized to receive Ranitidine or placebo twice daily for up to five years. In the original study no effect of Ranitidine on overall survival was demonstrated (Nielsen et al., 1998).

2.2. Fluorescence in situ hybridization

The TOP1/CEN‐20 probe mixture was developed by Dako Denmark A/S (Glostrup, Denmark) and used as previously described (Romer et al., 2012). A fluorescence microscope (Zeiss AX10) with a Texas Red/FITC double filter was used for quantifying red (TOP1) and green (CEN‐20) signals. In each of the 154 CRC tumors, the TOP1 and CEN‐20 copies were counted in sixty non‐overlapping cancer cell nuclei. In a subgroup consisting of 105/154 randomly selected patient samples, the TOP1 and CEN‐20 copies were counted separately in sixty nuclei per section in the unaffected epithelium adjacent to the malignant tumors in order to establish the diploid numbers of normal colorectal mucosa. A TOP1/CEN‐20 ratio was calculated both for the normal colorectal mucosa and for the malignant tumors.

2.3. Immunohistochemistry for Ki67

Nuclear accumulation of Ki67 protein was assessed by immunohistochemistry, applying the murine monoclonal antibody clone MIB‐1 (Key et al., 1993) (Dako Denmark A/S) and the ADVANCE™ HRP (Dako Denmark A/S) detection system according to instructions from the manufacturer. In brief, epitope retrieval was performed in a citrate buffer (pH 6.0) heated in a microwave oven for 10 min. Endogenous peroxidase activity was blocked by 1% hydrogen peroxide for 10 min and the slides were incubated with the primary antibody (1:50 dilution) for 30 min at room temperature. After detection, visualization was performed applying DAB+ (Dako Denmark A/S). Negative and positive controls were obtained simultaneously using an irrelevant murine IgG1 antibody, clone DAK‐GO1 (Dako Denmark A/S) and tumor sections with known Ki67 immunoreactivity, respectively. Furthermore, internal controls were present in the included samples with reaction of proliferating crypt cells in the non‐malignant mucosa adjacent to the tumor.

For the evaluation of Ki67 immunoreactivity, individual tumor cells were regarded as positive if staining was localized to the nucleus at any level of intensity. A microscope (Olympus BX51) with a fitted digital camera (Olympus UC30) and the software analySIS getIT 5.1 (Olympus) were applied to generate digital images for evaluation. From each slide twelve non‐overlapping digital images of the tumor area at ×400 magnification (UplanApo 40× objective lense from Olympus), were randomly selected. Tumor cell counting was performed using a manual cell counter and an “unbiased counting frame” (Gundersen et al., 1988) displayed on a monitor. For each sample, counting frames from all digital images were assessed successively, including the one where the total number of tumor cells rounded 300. Since all tumor cells in this last frame were included, this approach resulted in a total number of assessed cells slightly exceeding 300 since all tumor cells in the last frame were included. A Ki67 index was calculated for each sample as the percentage of positively stained tumor cells out of all counted tumor cells.

2.4. Statistical analyses

From the mean value of the TOP1 copy number per cell in the normal colorectal mucosa theoretical ranges for de‐ and increased TOP1 copy numbers were calculated. Using a multiplex of the haploid TOP1 copy number to define five groups (gene loss/<diploid (<2n), normal/diploid (2n), one extra gene/triploid (3n), two extra genes/tetraploid (4n) and more than two extra genes/“highploid” (>4n)) ranges around these points were calculated as ±0.5 times the same haploid gene copy number. The same principles were used to calculate the status of CEN‐20 using the mean CEN‐20 copy number generated from the normal mucosa. This methodology has previously been used to calculate CEN‐17 status in breast cancer tumors (Nielsen et al., 2012).

The endpoints considered for the prognostic part of the study were OS defined as time to death by any cause, TTR defined as time to any event related to primary colorectal cancer, e.g. time to loco‐regional recurrence, distant metastases or death by CRC and for LR in RC defined as time to local recurrence in rectal cancer patients only (American Society for Clinical Oncology (www.asco.org) definitions). Spearman's rank correlations were used as measures of association between continuous covariates and tests of location were done using the Wilcoxon or Kruskal–Wallis rank sum tests. Survival probabilities were estimated by the Kaplan–Meier method and comparisons were tested by the log rank test. The Cox proportional hazards model was used for univariate as well as multivariate analysis of the survival endpoints. Results are presented by the hazard ratio (HR) with 95% confidence intervals (CI) and the p‐value. The baseline covariates included in the multivariate analyses are gender, age (for a 10 year difference) and primary tumor localization (colon or rectum) and Ki67. The assumption of proportional hazards and linearity where applicable were assessed by Schoenfeld and Martingale residuals. No significant departure from these assumptions was detected. Separate multivariate Cox proportional hazard models were constructed for the TOP1 copy number per cell and the TOP1/CEN‐20 ratio as these two covariates were highly correlated. The covariates TOP1 copy number per cell, the CEN‐20 copy number per cell, the TOP1/CEN‐20 ratio and Ki67 were log transformed (base 2) resulting in HRs for a twofold difference in the copy number or the ratio. All calculated p‐values were two‐sided and considered significant at the 0.05 level.

Both the TOP1 and CEN‐20 copy number per cell were correlated with the Ki67 indices and Spearman's rank correlation coefficients with p‐values were calculated.

All statistical analyses were performed using the SAS (v9.2, SAS Institute, Cary, NC, USA).

3. Results

3.1. FISH analyses of TOP1 and CEN‐20 copy numbers

FISH analyses for TOP1 and CEN‐20 were successfully carried out in all 104 new samples. Results were pooled with data from the 50 samples from our previous study (Romer et al., 2012) resulting in a total of 154 cases. No major heterogeneity in the distribution of the TOP1 and CEN‐20 signals were observed within the individual tumors. The observer was the same for both studies and the same scoring methodologies were used. Results for each tumor are given in Supplementary Table 1. The mean TOP1 copy number ± SD in the tumors was 3.17 ± 1.21 (1.42–7.42) and the mean CEN‐20 copy number was 2.28 ± 0.60 (1.05–3.67). The TOP1/CEN‐20 ratio was 1.39 ± 0.42 (mean) with the range 0.99–2.93. In a total of 105 samples (including the previously published data from 50 of the samples) TOP1 and CEN‐20 signals were counted in the epithelial cells of the normal mucosa. The mean TOP1 copy number per cell was 1.57 (1.28–1.86) and the mean CEN‐20 copy number per cell was 1.52 (1.21–1.83) resulting in a mean TOP1/CEN‐20 ratio of 1.03 (0.93–1.13). From this mean value of TOP1 the theoretical range for normal diploid TOP1 status in the current setting was calculated as 1.18 to 1.96 copies per cell (equaling 1.57 ± 0.5 times the haploid number 0.785). The theoretical ranges for decreased and increased TOP1 copy number together with the tumor distribution according to this definition are presented in Table 1A.

Table 1.

A and B: Calculated theoretical ranges for TOP1 (A) and CEN‐20 (B) status. For TOP1 the ranges are defined as gene status and for CEN‐20 as ploidy/somic status. From only one probe of interest interphase FISH is not able to distinguish between the different underlying mechanisms for extra signals. Ranges are calculated from the FISH results in the normal colorectal mucosa as a multiplex of the mean haploid value of TOP1 or CEN‐20 copy number per cell ±0.5 times the same mean haploid copy number per cell. The frequency is shown in sample numbers (n) and in percentage (%).

A
TOP1 gene status per cell	TOP1 copies/whole nuclei	TOP1 gene signals/truncated nuclei	TOP1 gene signals/truncated nuclei	Frequency
		Average	Range	n (%)
Loss of gene copy	1	0.79	<1.18	0 (0)
Normal/diploid	2	1.57	[1.18–1.96)	29 (18.8)
1 extra gene copy	3	2.36	[1.96–2.75)	35 (22.7)
2 extra gene copies	4	3.14	[2.75–3.54)	35 (22.7)
>2 extra gene copies	> 4		≥3.54	55 (35.7)
B
CEN‐20 status per cell	CEN‐20 copies/whole nuclei	CEN‐20 signals/truncated nuclei	CEN‐20 signals/truncated nuclei	Frequency
		Average	Range	n (%)
Haploid/Monosomic	1	0.76	<1.14	1 (0.6)
Diploid/Disomic	2	1.52	[1.14–1.90)	50 (32.5)
Triploid/Trisomic	3	2.28	[1.90–2.66)	60 (39.0)
Tetraploid/Tetrasomic	4	3.04	[2.66–3.42)	36 (23.4)
”Highploid”/”Highsomic”	> 4		≥3.42	7 (4.5)

Using this definition the number of tumors in the normal diploid range was only 29 (18.8%). The remaining 125 tumors (81.2%) all had an increased TOP1 gene copy number above the normal diploid range. Fifty‐five (35.7%) of the tumors were in the “highploid” range (>4n) indicative of more than two extra TOP1 copies per cell. No tumors had a TOP1 copy number below the diploid range. As for TOP1 the ranges for CEN‐20 status were calculated using the mean value of the CEN‐20 copy number (1.52) obtained from reading the normal mucosa. In Table 1B the distribution of the tumors according to the CEN‐20 status is presented. Fifty (32.5%) of the tumors had a CEN‐20 copy number within the normal diploid range, while 103 (66.9%) had CEN‐20 numbers in the ranges indicative of increased ploidy levels of which 7 (4.5%) were in the range of “highploidy” (>4n). Only one tumor (0.6%) had a CEN‐20 number indicative of allelic loss. Using the TOP1/CEN‐20 ratio to describe the tumors 15 (9.7%) had a ratio ≥2.0 while 44 (28.6%) had a ratio ≥1.5. No tumors had a ratio ≤0.8, which together with the finding that no tumors had a TOP1 copy number per cell below the diploid range indicate that deletions of the TOP1 locus is an infrequent event. Figure 2 shows the correlation between the TOP1 and CEN‐20 copy number per cell and the concordance between using the TOP1 copy number alone or the TOP1/CEN‐20 ratio to describe increased TOP1 copy number. Of the 55 tumors with TOP1 copy numbers suggestive of more than two extra gene copies, the ratio was ≥1.5 in 35 cases (63.6%), whereas the ratio was ≥2.0 in only 15 tumors (27.3%). Out of the 44 tumors with a ratio ≥1.5 or of the 15 tumors with a ratio ≥2.0, 35 (77.3%) and 15 (100%), respectively, had TOP1 copy numbers consistent with more than two extra gene copies. Comparing the dataset obtained when using cutoff values for the ratio of ≥1.5 and for the TOP1 copy number per cell of ≥3.54 (>4n), the kappa value for concordance between these was 0.57.

Figure 2.

Distribution of the 154 stage III CRC tumors according the number of TOP1 copies per cell and CEN‐20 copies per cell. The red solid line mark a cutoff value for the TOP1 copy number of >4n (≥3.54 TOP1 copies per cell). The gray solid lines mark levels for TOP1 copy number of <2n (<1.18), 2n (1.18–1.96), 3n (1.96–2.75) and 4n (2.75–3.54). The stippled line, dotted line and the solid line separate the tumors using the TOP1/CEN‐20 ratio with ≥1.5 and ≥2.0, respectively.

3.2. Associations between variables and the basic patient and tumor characteristics

The TOP1 copy number per cell, the CEN‐20 copy number per cell and the TOP1/CEN‐20 ratio were correlated to patient age at operation, gender and tumor localization (Table 2). All three FISH variables were significantly associated with localization showing with higher values in rectal cancer (p = 0.004, p = 0.013 and p = 0.004, respectively). No significant associations were found between the FISH variables and gender, but an inverse and week association was found between the CEN‐20 copy number and patient age at operation (r = −0.16, p = 0.04).

Table 2.

Median values and ranges of the FISH variables, TOP1 copy number cell, CEN‐20 copy number per cell and the TOP1/CEN‐20 ratio, according to gender and primary tumor localization. p‐Values (Wilcoxon rank sum test) are for the association between variables and the basic patient and tumor characteristics. The correlations between patient age at operation and the variables are demonstrated with Spearman's rank correlation coefficients with calculated p‐values.

		TOP1 copy number: Median (min–max)		CEN‐20 copy number: Median (min–max)		TOP1/CEN‐20 ratio: Median (min–max)
Gender	M	3.13 (1.42–7.33)	p = 0.08	2.35 (1.05–3.67)	p = 0.14	1.28 (0.99–2.93)	p = 0.06
	F	2.97 (1.42–7.42)		2.12 (1.38–3.67)		1.20 (1.01–2.82)
Localization	CC	2.87 (1.50–5.45)	p = 0.004	2.06 (1.37–3.67)	p = 0.013	1.16 (0.99–2.93)	p = 0.004
	RC	3.37 (1.42–7.42)		2.49 (1.05–3.67)		1.33 (0.99–2.82)
Age		r = −0.15, p = 0.06		r = −0.16, p = 0.04		r = −0.07, p = 0.40

The TOP1 gene copy number correlated significantly with the CEN‐20 number (r = 0.72; p < 0.0001) and also with the ratio (r = 0.78; p < 0.0001).

Ki67 immunoreactivity was determined in order to assess whether proliferation was correlated with the TOP1 and/or CEN‐20 copy numbers. One sample was excluded since there was no tumor tissue left in the paraffin block for this staining. All sections from the remaining 153 samples showed immunoreactivity for Ki67 in the tumor cell nuclei with proliferation indices ranging from 3% to 92%. Figures 3A and B show the correlation between the CEN‐20 copy number per cell (Figure 3A) or the TOP1 copy number per cell (Figure 3B) and the Ki67 proliferation index. The correlation was weak and inverse, but significant between CEN‐20 copy number and the Ki67 index, with a Spearman's rank correlation coefficients of r = −0.224 (p = 0.005). No correlation was observed between the TOP1 copy number and the Ki67 index (r = −0.131; p = 0.107).

Figure 3.

A–B. Relationship between percentage of Ki67 positive cells and CEN‐20 copy number (Figure 3A, Spearman's rank correlation coefficient: r = −0.224; p = 0.005) and between percentage of Ki67 positive cells and TOP1 copy number/cell (Figure 3B, Spearman's rank correlation coefficient: r = −0.131; p = 0.107) in 153 CRC samples.

3.3. Univariate analysis

Among the 154 CRC patients, 48 (18 CC and 30 RC) experienced local recurrence, 47 (25 CC and 22 RC) were diagnosed with distant metastases and 115 (54 CC and 61 RC) died during the observation period. Among the 115 deaths, 68 were registered as death by CRC. The median observation time was 8.0 years (range: 6.9–9.0).

In the univariate survival analyses (Table 3A) the HRs for the association between the TOP1 copy number per cell and all end points were below 1.0, indicating that continuously increasing values of the TOP1 copy number were associated with longer survival times. However, the only significant HR for the TOP1 copy number per cell was demonstrated in relation to OS (HR: 0.71; 95% CI: 0.50–0.99; p = 0.04). For all endpoints, the HRs for the CEN‐20 copy number per cell were insignificant and closer to 1.0 than for the TOP1 copy number, indicating a weak or no prognostic association. Just as for the TOP1 copy number alone, HRs for the TOP1/CEN‐20 ratio were below 1.0 for all endpoints, suggesting that higher ratios were associated with longer survival times. Only the association between the ratio and LR in RC was significant (HR: 0.27; 95% CI: 0.09–0.84; p = 0.02).

Table 3.

A–C: HR, 95% CI and p‐values (Wald test) for the Cox proportional hazard models with the TOP1 copy number per cell, the CEN‐20 copy number per cell, the TOP1/CEN‐20 ratio and Ki67 treated as continuous variables. A shows the results from the univariate models. B–C show the results from the two separate multivariate models. *Female. **Rectal tumor localization. ***Age at operation (for 10 year differences).

A
Variable	OS			TTR			LR in RC
	HR	95% CI	p‐Value	HR	95% CI	p‐Value	HR	95% CI	p‐Value
TOP1 per cell	0.71	0.50–0.99	0.04	0.88	0.61–1.27	0.50	0.56	0.29–1.09	0.09
CEN‐20 per cell	0.73	0.45–1.20	0.21	1.09	0.63–1.88	0.77	0.99	0.37–2.65	0.98
TOP1/CEN‐20	0.68	0.41–1.11	0.12	0.72	0.42–1.24	0.23	0.27	0.09–0.84	0.02
B
Covariates	OS			TTR			LR in RC
	HR	95% CI	p‐Value	HR	95% CI	p‐Value	HR	95% CI	p‐Value
TOP1 per cell	0.62	0.42–0.90	0.01	0.72	0.48–1.07	0.10	0.51	0.24–1.08	0.08
Gender*	0.70	0.47–1.03	0.07	0.80	0.52–1.23	0.30	0.81	0.37–1.75	0.59
Localization**	1.99	1.36–2.91	0.0004	2.24	1.45–3.45	0.0003	–	–	–
Age***	1.50	1.24–1.81	<0.0001	1.23	1.00–1.50	0.05	1.06	0.74–1.53	0.75
Ki67	0.87	0.73–1.04	0.13	0.86	0.71–1.04	0.13	0.97	0.67–1.40	0.86
C
Covariates	OS			TTR			LR in RC
	HR	95% CI	p‐Value	HR	95% CI	p‐Value	HR	95% CI	p‐Value
TOP1/CEN‐20	0.61	0.36–1.03	0.07	0.61	0.34–1.09	0.09	0.25	0.08–0.83	0.02
Gender*	0.74	0.51–1.09	0.13	0.82	0.54–1.25	0.36	0.83	0.39–1.76	0.62
Localization**	1.88	1.29–2.74	0.001	2.18	1.43–3.34	0.0003	–	–	–
Age***	1.53	1.27–1.84	<0.0001	1.24	1.02–1.52	0.03	1.09	0.76–1.56	0.65
Ki67	0.92	0.77–1.10	0.35	0.89	0.74–1.08	0.25	1.07	0.75–1.54	0.70

3.4. Multivariate analyses

Since no prognostic characteristic for the CEN‐20 copy number per cell was found in the univariate survival analyses only the TOP1 copy number per cell and the TOP1/CEN‐20 ratio were included in multivariate analyses. However, due to the high degree of correlation between these two variables, separate multivariate Cox proportional hazard models were made for each (Tables 3B and C). The association between the TOP1 copy number per cell and the OS was even more pronounced than for the univariate analysis with a decreased HR (HR: 0.62; 95% CI: 0.42–0.90; p = 0.01). The effect on TTR and LR in RC, as seen on decreased HRs in relation to both endpoints, was also more pronounced but results were still insignificant. The HR for the association between the TOP1/CEN‐20 ratio and LR in RC also retained its significance (HR: 0.25; 95% CI: 0.08–0.83; p = 0.02). The HRs for the association between the ratio and both OS and TTR decreased, again indicating a higher degree of effect, although still statistically insignificant (p = 0.07 and p = 0.09, respectively). In both multivariate models, tumor localization was highly significantly associated with both OS and TTR with rectal tumor localization, indicating poor prognosis. As expected, age at surgery was significantly associated with OS but not with LR in RC. No significant associations were found between gender and the endpoints. Ki67 did not have any prognostic characteristic in both multivariate analyses, since it was not significantly associated with OS, TTR and LR in RC.

In order to graphically illustrate the associations between the TOP1 copy number per cell or the ratio and the clinical endpoints (OS, TTR and LR in RC), Kaplan–Meier survival plots were generated (Figure 4A–F). For the TOP1 copy number per cell, strata was defined by the ranges for gene status described above (<2n, 2n, 3n, 4n, >4n) which resulted in four groups since no tumors had copy numbers <2n. In the subgroup analysis with LR in RC as endpoint, only >4n was chosen to dichotomize the population. For the TOP1/CEN‐20 ratio the population was dichotomized by 1.5, since a cutoff value of 2.0 would result in two unbalanced groups. As seen from Figure 4A the four curves showed a tendency for separation with the TOP1 copy number per cell in relation to OS. However, the p‐value was not significant (log rank test; p = 0.08). Dichotomizing the population by 2.75 (>3n) (i.e. combining the two curves with the highest TOP1 copy numbers and the two with the lowest TOP1 copy numbers, respectively) resulted in a significant separation of the two new curves (log rank test; p = 0.01)(Figure not shown). When the TOP1 copy number per cell was seen in relation to TTR (Figure 4C) no separation of the curves could be observed (log rank test; p = 0.77), which was in line with the result from the univariate Cox proportional hazard model. For the TOP1 copy number per cell and LR in RC (Figure 4E) as well for the TOP1/CEN‐20 ratio in relation to all three endpoints (Figure 4B, D and F), the curves separated with higher copy numbers and higher ratios associated with improved survival. However the separation was insignificant (log rank test; p > 0.05) in these Kaplan–Meier plots.

Figure 4.

A–F. Kaplan–Meier survival plots for the TOP1 copy number per cell and the TOP1/CEN‐20 ratio in relation to all three clinical endpoints. For the TOP1 copy number per cell in relation to OS (A) and TTR (C) strata is defined by the gene copy number ranges, and in relation to LR in RC (E) by 3.54 (>4n). For the TOP1/CEN‐20 ratio a cutoff value of 1.5 dichotomizes the population in relation to all three endpoint, OS (B), TTR (D) and LR in RC (F). The p‐values are for the log rank test and the HRs are for univariate analyses with “high” vs. “low” TOP1 copy numbers or ratios.

4. Discussion

Prognostic and predictive biomarkers have gained increasing interest with the emerging demand for personalized medicine. However, all prognostic and predictive biomarkers must pass a number of well‐described and strict validation steps before they can be introduced into daily clinical oncology practice (Simon et al., 2009).

Studies on 10 human CRC cell lines conducted previously have suggested that the TOP1 copy number per cell was associated with increased in vitro sensitivity to irinotecan (SN38) but not to oxaliplatin (Romer et al., 2012). We also described that increased TOP1 and increased CEN‐20 numbers were frequent findings in CRC tumors (Romer et al., 2012).

We here report on a study aiming to validate that TOP1 gene copy number aberrations or increased TOP1/CEN‐20 ratios are frequent findings in patients with stage III CRC. When we applied an approach to determine the TOP1 gene status based on the gene copy numbers alone, we found that 81.2% of the tumors had increased TOP1 copy number per cell. Setting a more restrictive cutoff value where only tumors with more than two extra copies (>4n) of the gene were selected the frequency remained high at 35.7%. When using the TOP1/CEN‐20 ratio, which is a more conventional approach, the fraction of tumors with TOP1 gene aberrations was still significant with 28.6% ≥ 1.5 and 9.7% ≥ 2.0. No tumors with a TOP1 copy number per cell or a ratio indicative of gene loss were identified. The overall frequency of TOP1 gene aberrations was determined to be sufficient to continue with further analysis in order to resolve whether counting of TOP1 gene copy number is applicable as a predictive biomarker of irinotecan efficacy.

A prognostic component of a putative predictive biomarker may have confounding influence on the interpretation of results from clinical studies comparing the presence of a biomarker and the benefit from therapy (Nielsen and Brunner, 2011). In the present study we aimed to determine if the TOP1 copy number per cell and/or the ratio between TOP1 and CEN‐20 was associated with patient prognosis in stage III CRC. Our study showed that both the TOP1 copy number per cell and the TOP1/CEN‐20 ratio has an association to prognosis; where a higher TOP1 copy number per cell and/or a higher ratio is associated with longer survival times. Our results stress that future studies on associations between TOP1 and CEN‐20 copy number and treatment benefit from Top1 poisons in CRC patients must be performed with caution and must always include a control cohort. Such a study might be the FOCUS trial (Seymour et al., 2007), the two first‐line treatment arms of the CAIRO study (Koopman et al., 2007) for treatment of mCRC or the PETACC III study (van et al., 2009) for adjuvant treatment of primary CRC.

It is known that CRC tumors often show aneuploidy (Sasaki et al., 1995). Among Dukes' stage A (stage I) tumors, 90% are aneuploid, slightly increasing with tumor stage, thus leaving very few tumors diploid at high stages (Flyger et al., 1999). Aneuploidy has been associated with a poor prognosis (16). Chromosomal rearrangements are often seen in CRC tumor tissue (Ried et al., 1996) and increased copy number of the chromosome 20 q‐arm, either as aneusomy of chromosome 20, whole 20q arms gain or 20q isochromosome formation, has been reported to occur frequently in CRC tumor samples (Carvalho et al., 2009; Sheffer et al., 2009) and has been associated with progression of CRC (Hermsen et al., 2002; Ried et al., 1996). We used the public “Tumorscape database” (www.broadinstitute.org/tumorscape) which contains >3000 samples from 54 cancer subtypes (including cell lines) to estimate the frequency of TOP1 copy number aberrations. This database applies the Genomic Identification of Significant Targets in Cancer (GISTIC) method (Beroukhim et al., 2007), which distinguishes between random background aberrations and significant events. The GISTIC analysis of the 128 CRC samples and 33 cell lines from the database revealed that TOP1 is not significantly focally amplified (Firestein et al., 2008; Beroukhim et al., 2010). To investigate the mechanism of TOP1 copy number increase, we used the integrative genomics viewer (Robinson et al., 2011) (IGV – www.broadinstitute.org/igv) on the same 128 CRC samples. This revealed that a large fraction (>50%) had a broad gain spanning the entire q‐arm of chromosome 20, indicating that the increased copy number of TOP1 could be due to a gain of 20q. This would be in agreement with previous findings (Tsafrir et al., 2006) and with the high correlation between TOP1 and CEN‐20 found in the present study. In some tumors this 20q gain includes the centromere region, which is in line with our CEN‐20 results. This may imply that the calculation of the TOP1/CEN‐20 ratio in these tumors may result in a ratio close to 1, thereby being similar to TOP1/CEN‐20 ratio calculations in patients with diploid copy numbers. Even though the TOP1/CEN‐20 ratio was able to identify a fraction of tumors with increased gene copy numbers the biologically relevant fraction could theoretically be even larger. Whether the TOP1 copy number should be related to the CEN‐20 signal is therefore a central question when validating the potential predictive value of TOP1 FISH measurements.

Cells undergoing DNA synthesis during mitosis have increased gene copy numbers with a maximum of 4 copies just before cell division. Thus, part of the increased TOP1 copy numbers observed in the present study could be explained by high proliferative activity in the tumors. To test this hypothesis, we performed Ki67 analyses on the CRC samples and correlated these results with the FISH results. Although a significant association between CEN‐20 copy number and Ki67 was observed, this association was inverse and weak suggesting that increased CEN‐20 copy number in CRC is not a marker of increased proliferative capacity of the tumor cells. Of equal importance, we did not find any significant associations between the TOP1 copy number and Ki67 immunoreactivity. This suggests that TOP1 gene copy number in CRC is not associated with increased cancer cell proliferation.

The observed association between increased TOP1 copy number and improved patient prognosis is contrary to what we expected and we have at present no simple explanation to this observation.

Ki67 was included in the multivariate analyses, where it was found not to hold any independent prognostic characteristic. This implies that Ki67 did not obscure the prognostic value of the TOP1 copy number.

Our prognostic part of the study has some limitations, the most important being the relative low number of patients (n = 154). However, selecting stage III CRC patients resulted in a high number of events, strengthening the statistical power. Moreover, we selected three clinically relevant endpoints, OS, TTR and LR in RC. Although none of the included patients received adjuvant chemotherapy, we cannot exclude that some of the patients received chemotherapy for mCRC late in the observation period, a fact that might influence OS.

Since this is the first study reporting on the association between TOP1 and CEN‐20 copy numbers and the prognosis of stage III CRC patients, we were not able to search for an optimized cut‐point for TOP1 or the ratio. However, we have shown a continuous relationship between the TOP1 copy number levels as well as the TOP1/CEN‐20 ratio levels and patient outcome, which supports an association to prognosis. We applied a biological approach using calculated ranges for the TOP1 gene copy number status based on results from epithelial cells of the normal colorectal mucosa, thus creating reference values for gene loss vs. normal vs. increased TOP1 copy number.

Finally, it should be stressed as a potential limitation of our study that the included patients had surgical procedures according to old recommendations and no adjuvant therapy, meaning that they are not representative of CRC patients of today. However, the lack of adjuvant therapy makes it possible to determine the “true” association between TOP1 copy number and patient outcome. If a similar study had been conducted on present‐day patients, the anti‐tumor effect of adjuvant 5FU and oxaliplatin could potentially have masked the prognostic effect, or the data could have been misinterpreted as an increased TOP1 gene copy number being predictive of 5FU + oxaliplatin treatment. We would therefore like to emphasize that even though the treatment strategies are different today, we find that this study provides an important contribution when estimating whether TOP1 FISH can be used as a predictive marker in the selection of systemic chemotherapy to patients with metastatic CRC.

Disclosure/Conflict of interest

Kirsten Vang Nielsen and Sven Müller are employees at DAKO A/S.

Nils Brünner is Medical Advisor to DAKO A/S. The remaining authors have no disclosures or conflicts of interest.

Supporting information

The following is the supplementary data related to this article:

Supplementary data

Supplementary data 1.

1.1.

Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.molonc.2012.09.001.

Rømer Maria Unni, Nygård Sune Boris, Christensen Ib Jarle, Nielsen Signe Lykke, Nielsen Kirsten Vang, Müller Sven, Smith David Hersi, Vainer Ben, Nielsen Hans Jørgen, Brünner Nils, (2013), Topoisomerase 1(TOP1) gene copy number in stage III colorectal cancer patients and its relation to prognosis, Molecular Oncology, 7, doi: 10.1016/j.molonc.2012.09.001.