Copy number variation of E3 ubiquitin ligase genes in peripheral blood leukocyte and colorectal cancer.

Given that E3 ubiquitin ligases (E3) regulate specific protein degradation in many cancer-related biological processes. E3 copy number variation (CNV) may affect the development and prognosis of colorectal cancer (CRC). Therefore, we detected CNVs of five E3 genes in 518 CRC patients and 518 age, gender and residence matched controls in China, and estimated the association between E3 gene CNVs and CRC risk and prognosis. We also estimated their interactions with environmental factors and CRC risk. We find a significant association between the CNVs of MDM2 and CRC risk (amp v.s. wt: odds ratio = 14.37, 95% confidence interval: 1.27, 163.74, P = 0.032), while SKP2 CNVs may significantly decrease CRC risk (del v.s. wt: odds ratio = 0.32, 95% confidence interval: 0.10, 1.00, P = 0.050). However, we find no significant association between the CNVs of other genes and CRC risk. The only significant gene-environment interaction effects are between SKP2 CNVs and consumption of fish and/or fruit (P = 0.014 and P = 0.035) and between FBXW7 CNVs and pork intake (P = 0.040). Finally, we find marginally significant association between β-TRCP CNVs and CRC prognosis (amp v.s. wt, hazard ratio = 0.42, 95% confidence interval: 0.19, 0.97, P = 0.050).

Colorectal cancer (CRC) is the second most common cancer in women and the third most common in men worldwide1. In 2012, the World Health Organization estimated that about 1,360,000 new CRC cases occurred worldwide. In addition, 694,000 deaths from CRC were estimated worldwide, accounting for 8.5% of all cancer deaths, and making CRC the fourth most common cause of death from cancer2. Although the relative 5-year survival rate of European CRC patients increased between 1930 and 20103, that 5-year survival rate was only 30–65% worldwide4.

Genetic susceptibility has a well-established role in the etiology of CRC56. Accumulating evidence supports the hypothesis that copy number variation (CNV) is a molecular biomarker for CRC risk and prognosis7. DNA CNVs, as structural variants, can be small: microscopic or submicroscopic; or they can be large: deletions, duplications or insertions, often larger than 1 kb89.

CNVs in the E3 ubiquitin ligases (E3) of the ubiquitin-proteasome system (UPS) have been associated with CRC risk and prognosis101112. E3 plays a critical role in the specific protein degradation of UPS, which has an essential regulatory role in cell cycle progression, cell proliferation, differentiation, apoptosis, angiogenesis and cell signaling pathways1314. The two main subfamilies of E3s are RING and HECT domain containing E3s15. As members of RING E3s, FBXW7, MDM2, SKP2 and β-TRCP have been associated with abnormal expression in some malignancies including blood, breast, colon and prostate16171819. As a HECT E3 ligase, NEDD4-1 was also proposed to play a vital role in a number of human cancers, including CRC2021.

FBXW7 serves as a tumor suppressor gene (p53-dependent)22, and loss of FBXW7 has been associated with CRC risk and poor prognosis23. MDM2, functioning as an oncogene, is amplified in approximately one-third of all human carcinomas including CRC24. Increased expression of SKP2 has been significantly associated with poor tumor differentiation and poor prognosis in CRC18. Overexpression of β-TRCP has also been observed in many tumors, such as CRC25, pancreatic cancer26, and breast cancer27. NEDD4-1, as a HECT E3 ligase, is highly expressed in both colorectal and gastric tumor tissues20.

Studies of the CNVs of FBXW7, MDM2, SKP2, β-TRCP and NEDD4-1 genes are mainly limited to intestinal cancer cell lines and clinical pathological tissues1011121923252829303132. In addition, most studies focus on gene expression; the impact of germline CNVs of these five genes on CRC risk and prognosis are not fully understood. Therefore, we conducted a case-control study to explore associations between the CNVs of FBXW7, MDM2, SKP2, β-TRCP and NEDD4-1 genes and CRC risk. We also followed up with cases to study the association between the CNVs of these five genes and CRC prognosis in China.

Results

Characteristics of study subjects

The basic characteristics of the 518 CRC patients and the 518 gender, age, and residence matched controls are summarized in Table 1. However, 32 pairs of our samples were unable to be genotyped in one of the five genes, so gender was not equally distributed in cases and controls (P = 0.002). Education (P < 0.001), occupation (P < 0.001) and family history of other cancers (P < 0.001) were also differently distributed in cases and controls. Of the 518 CRC cases, 262 (57.8%) were colon cancer, 191 (42.2%) were rectal cancer. Gender, occupation, education and family history of cancer were adjusted in the following analysis.

Table 1

Basic characteristics of cases and controls.

Characteristic	No. of Case (%)	No. of Controls (%)	Pvaluea
Age	60.45 ± 11.22	59.78 ± 10.64	0.955
Gender			0.002
Male	299 (57.7)	249 (48.1)
Female	219 (42.3)	269 (51.9)
BMI (kg/m2)	23.80 ± 3.80	24.10 ± 4.36	0.441
Educationa			<0.001
Primary school and below	288 (58.6)	258 (51.3)
Junior middle school	103 (21.0)	117 (23.3)
Senior middle school and above	100 (20.4)	128 (25.4)
Occupationa			<0.001
White collar	90 (17.9)	64 (12.8)
Blue collar	260 (51.7)	319 (63.9)
Both	153 (30.4)	116 (23.3)
Family history of cancera			<0.001
Yes	38 (10.2)	218 (43.1)
No	334 (89.8)	288 (56.9)
Location of primary tumora
Colon	262 (57.8)
Rectum	191 (42.2)
Stage of Dukesa
I	52 (10.6)
II	246 (50.3)
III	159 (32.5)
IV	32 (6.6)
I+II	298 (60.9)
III+IV	191 (39.1)

aMissing data on subjects, education 27 cases, 15 controls; occupation, 15 cases, 19 controls; family history of cancer, 146 cases, 12 controls; tumor location, 65 cases.; stage of Dukes, 29 cases. bP < 0.05 was considered statistically significant.

Copy number variation and CRC risk

The FBXW7, MDM2, SKP2, β-TRCP, and NEDD4-1 CNVs were in Hardy-Weinberg equilibrium in all controls. Table 2 shows the CNV frequencies of the five genes and the relationship between the CNVs of the five genes and CRC risk.

Table 2

Associations between CNVs and the risk of CRC.

Gene	No. of Cases (%)	No. of Controls. (%)	Odds Ratioa	95% Confidence Interval	PValueb
MDM2
Wt	487 (96.0)	499 (98.4)	1.00
Del	11 (2.2)	6 (1.2)	3.76	0.69, 20.61	0.127
Amp	9 (1.8)	2 (0.4)	14.37	1.27, 163.74	0.032
Amp v.s. del + wt			14.40	1.26, 164.81	0.032
Del + amp v.s. wt			6.35	1.67, 24.19	0.007
SKP2
Wt	452 (95.4)	433 (91.4)	1.00
Del	13 (2.7)	25 (5.2)	0.32	0.10, 1.00	0.050
Amp	9 (1.9)	16 (3.4)	0.32	0.10, 1.01	0.052
Amp v.s. del + wt			0.33	0.11, 1.02	0.055
Del + amp v.s. wt			0.32	0.14, 0.72	0.006
FBXW7
Wt	410 (84.9)	400 (82.8)	1.00
Del	37 (7.7)	36 (7.5)	1.29	0.61, 2.70	0.506
Amp	36 (4.4)	7 (9.7)	0.64	0.34, 1.23	0.181
Amp v.s. del + wt			0.63	0.33, 1.20	0.162
Del + amp v.s. wt			0.82	0.52, 1.42	0.557
β-TRCP
Wt	465 (93.2)	469 (93.8)	1.00
Del	5 (1.0)	8 (1.6)	0.71	0.11, 4.41	0.710
Amp	29 (5.8)	23 (4.6)	1.59	0.72, 3.52	0.252
Amp v.s. del + wt			1.59	0.72, 3.51	0.254
Del + amp v.s. wt			1.40	0.70, 2.88	0.363
NEDD4-1
Wt	431 (89.0)	448 (92.6)	1.00
Del	2 (0.4)	6 (1.2)	0.59	0.08, 4.55	0.614
Amp	51 (10.6)	30 (6.2)	1.44	0.73, 2.87	0.297
Amp v.s. del + wt			1.44	0.72, 2.87	0.297
Del + amp v.s. wt			1.31	0.39, 2.51	0.409

aAdjusted for gender, occupation, education, and family history of cancer. bP < 0.05 in the conditional logistic regression analysis was considered statistically significant.

We observed significant associations between MDM2 amplification and increased CRC risk (amp v.s. wt: ORadjusted = 14.37, 95% CI: 1.27, 163.74, P = 0.032; amp v.s. del + wt: ORadjusted = 14.40, 95% CI: 1.26, 164.81, P = 0.032). We observed marginally significant association between SKP2 deletions and CRC risk (del v.s. wt: ORadjusted = 0.32, 95% CI: 0.10, 1.00, P = 0.050). While there was no significant association between SKP2 amplification and CRC risk in the amp v.s. del + wt model (amp v.s. del + wt model: OR = 0.33, 95% CI: 0.11, 1.02, P = 0.055). However, we observed no significant associations between FBXW7, β-TRCP or NEDD4-1 CNVs and CRC risk.

Abnormal copy number additive model and CRC risk

In the abnormal copy number additive model, MDM2 CNVs are significantly associated with increased CRC risk (del + amp v.s. wt: ORadjusted = 6.35, 95% CI: 1.67, 24.19, P = 0.007). In the additive models, SKP2 CNVs also significantly decrease CRC risk (del + amp v.s. wt: ORadjusted = 0.32, 95% CI: 0.14, 0.72, P = 0.006).

Gene-environment interactions on CRC risk

We find a significant synergistic interaction effect between SKP2 CNVs and fruit consumption (amp v.s. del + wt: ORi = 13.89, 95% CI: 1.20, 160.57, P = 0.035) (Table 3). In addition, there is a significant interaction effect between the amplification of SKP2 and roughage consumption (≥50 g/week) (amp v.s. del + wt: OReg = 0.18, 95% CI: 0.03, 0.99). We also find significant interaction effects between the amplification of FBXW7 and consumption of roughage (≥50 g/week) or fish (>once/week) (OReg = 0.37, 95% CI: 0.15, 0.91 and OReg = 0.25, 95% CI: 0.07, 0.94, respectively). There were also significant interaction effects between the amplification of NEDD4-1 and consumption of refined grains (>250 g/week) (OReg = 2.83, 95% CI: 1.02, 7.88), Chinese pickled sour cabbage (>twice/month) (OReg = 3.59, 95% CI: 1.23, 10.48), and fatty meats (OReg = 3.60, 95% CI: 1.27, 10.19).

Table 3

Interactions between five gene CNVs and environmental factors on the risk of CRC.

CNV genotypes		Environmental factors		Interaction
SKP2		Roughage (g/week)
		<50	≥50
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Del + wt	1.00	0.58 (0.39, 0.88)
	Amp	0.18 (0.03, 0.98)	0.18 (0.03, 0.99)	1.72 (0.16, 18.74)	0.657
	Wt	1.00	0.62 (0.41, 0.94)
	Del + amp	0.28 (0.08, 1.00)	0.13 (0.04, 0.44)	0.74 (0.13, 4.30)	0.734
		Fruit (times/week)
		<2	≥2
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Del + wt	1.00	0.61 (0.40, 0.94)
	Amp	0.82 (0.03, 0.53)	0.70 (0.15, 3.22)	13.89 (1.20, 160.57)	0.035
	Wt	1.00	0.62 (0.34, 0.98)
	Del + amp	0.09 (0.02, 0.44)	0.33 (0.12, 0.96)	6.10 (0.92, 40.38)	0.061
		Fish (times/week)
		≤1	>1
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	0.30 (0.17, 0.52)
	Del + amp	0.09 (0.03, 0.31)	0.39 (0.08, 2.00)	13.62 (1.70, 109.36)	0.014
MDM2		Refined grains (g/day)
		≤250	>250
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	2.53 (1.59, 4.02)
	Del + amp	13.35 (2.13, 89.49)	5.44 (1.03, 28.86)	0.09 (0.00, 2.14)	0.135
		Fat meat
		No	Yes
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	2.30 (1.48, 3.57)
	Del + amp	11.25 (1.63, 77.25)	8.55 (1.22, 59.75)	0.30 (0.02, 5.09)	0.427
		Egg (/week)
		≤3	>3
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	1.60 (1.08, 2.37)
	Del + amp	9.31 (0.66, 131.06)	7.33 (1.57, 34.30)	0.49 (0.02, 10.06)	0.645
		Chinese pickled sour cabbage (times/month)
		≤2	>2
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	2.05 (1.35, 3.12)
	Del + amp	6.41 (1.16, 35.27)	27.61 (2.12, 259.81)	2.10 (0.10, 44.07)	0.633
		leftovers (times/week)c
		≤3	>3
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	1.873 (1.234, 2.843)
	Del + amp	3.30 (0.37, 29.54)	26.67 (2.62, 271.60)	4.31 (0.18, 100.81)	0.363
FBXW7		Roughage (g/week)
		<50	≥50
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Del + wt	1.00	0.62 (0.41, 0.95)
	Amp	0.57 (0.22, 1.50)	0.37 (0.15, 0.91)	1.04 (0.30, 3.67)	0.946
		Fish (times/week)
		≤1	>1
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Del + wt	1.00	0.34 (0.24, 0.66)
	Amp	0.64 (0.28, 1.44)	0.25 (0.07, 0.94)	0.99 (0.20, 4.88)	0.993
		Refined grains (g/day)
		≤250	>250
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	2.28 (1.37, 3.80)
	Del + amp	0.87 (0.45, 1.67)	2.81 (1.52, 6.86)	1.43 (0.49, 4.19)	0.519
		Fat meat
		No	Yes
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	2.23 (1.35, 3.68)
	Del + amp	0.76 (0.40, 1.47)	2.30 (1.03, 5.11)	1.35 (0.47, 3.82)	0.577
		Pork (g/week)
		≤250	>250
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	1.34 (0.87, 2.06)
	Del + amp	0.46 (0.21, 0.99)	1.92 (0.90, 4.11)	3.13 (1.06, 9.41)	0.040
		Physical exercise
		No	Yes
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	0.06 (0.02, 0.20)
	Del + amp	1.58 (0.66, 3.78)	0.06 (0.01, 0.31)	0.65 (0.10, 4.47)	0.662
NEDD4-1		Refined grains (g/day)
		≤250	>250	interaction
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Del + wt	1.00	2.61 (1.61, 4.24)
	Amp	1.63 (0.56, 4.71)	2.83 (1.02, 7.88)	0.67 (0.15, 2.89)	0.587
	Wt	1.00	2.63 (1.62, 4.27)
	Del + amp	1.48 (0.57,3.86)	2.75 (1.00, 7.55)	0.71 (0.18, 2.81)	0.622
		Chinese pickled sour cabbage (times/month)
		≤2	>2
		OReg (95% CI)1		ORi (95% CI)a	P valueb
	Del + wt	1.00	1.86 (1.21, 2.85)
	Amp	3.13 (0.92, 10.62)	3.59 (1.23, 10.48)	1.79 (0.41, 7.88)	0.444
		Fat meat
		No	Yes
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Del + wt	1.00	2.26 (1.43, 3.58)
	Amp	1.03 (0.35, 3.00)	3.60 (1.27, 10.19)	1.54 (0.35, 6.71)	0.564
	Wt	1.00	2.27 (1.43, 3.60)
	Del + amp	1.00 (0.38, 2.62)	3.30 (1.21, 9.14)	1.46 (0.37, 5.84)	0.590
		Physical exercise
		No	Yes
		OReg (95% CI)a		ORi (95% CI)a	P valueb
	Wt	1.00	0.06 (0.02, 0.20)
	Del + amp	1.58 (0.66, 3.78)	0.06 (0.01, 0.31)	10.27 (0.60, 177.48)	0.109

aAdjusted for gender, occupation, education, and family history of cancer. bP < 0.05 in the conditional logistic regression analysis was considered statistically significant. c leftovers: leftovers more than 12 hours.

Gene-environment interactions in abnormal copy number additive model

We observed significant synergistic interactions between SKP2 del + amp genotype and fish intake on CRC risk (del + amp v.s. wt: ORi = 13.62, 95% CI: 1.70, 109.36, P = 0.014) (Table 3). In addition, We also observed significant interaction effects between the del + amp genotype of SKP2 and roughage consumption (≥50 g/week), or fruit (≥twice/week) consumption (del + amp v.s. wt: OReg equal to 0.13 (95% CI: 0.04, 0.44) and 0.33 (95% CI: 0.12, 0.96), respectively). We also find significant interaction effects between the del + amp genotype of MDM2 and consumption of refined grains (>250 g/week) (OReg = 5.44, 95% CI: 1.03, 28.86), fatty meats (OReg = 8.55, 95% CI: 1.22, 59.75), eggs (>3/week) (OReg = 7.33, 95% CI: 1.57, 34.30), Chinese pickled sour cabbage (>twice/month) (OReg = 27.61, 95% CI: 2.12, 259.81) and leftovers (>3 times/week) (OReg = 26.67, 95% CI: 2.62, 271.60). Moreover, we observed a significant interaction between FBXW7 CNVs and pork consumption (>250 g/week) (del + amp v.s. wt: ORi = 3.13, 95% CI: 1.06, 9.41, P = 0.040). We find significant interaction effects between the del + amp genotype of FBXW7 and consumption of refined grains (>250 g/day), fatty meats and physical exercise (OReg = 2.81 (95% CI: 1.52, 6.86), OReg = 2.30 (95% CI: 1.03, 5.11) and OReg = 0.06 (95% CI: 0.01, 0.31), respectively). Finally, we also find significant interaction effects between the del + amp genotype of NEDD4-1 and consumption of refined grains (>250 g/day), fatty meats and physical exercise (OReg = 2.75 (95% CI: 1.00, 7.55) OReg = 3.30 (95% CI: 1.21, 9.14) and OReg = 0.06 (95% CI: 0.01, 0.31), respectively).

Copy number variations and CRC prognosis

323 patients completed the follow-up (Table 4). Of the 323 patients, 186 (57.9%) patients didn’t receive chemotherapy, 45 (14.0%) patients received FOXFOX4-based chemotherapy, 22 (6.9%) received XELOX-based chemotherapy, 44 (13.7%) received LCF-based chemotherapy, 6 (1.9%) received 5-Fu-based chemotherapy and 18 (5.6%) received other chemotherapy treatments after surgery. The mean overall survival (OS) of CRC patients was 75.35 ± 2.26 months. The CEA and CA19-9 level before surgery, Dukes stage, pathological type and metastasis were adjusted for in the analysis of FBXW7, MDM2, SKP2, β-TRCP and NEDD4-1 CNVs and CRC prognosis, due to their significant association with CRC prognosis in the univariate Cox proportional hazards regression.

Table 4

Clinical and pathological features of 323 CRC patients.

Characteristics	Patients	%
Age at diagnosis
<50	74	22.9
50–60	112	34.7
60–70	91	28.2
>70	46	14.2
Mean	58.58 ± 10.68
Median survival time (month)	73
Extreme value	0–109
Gender
Male	182	56.3
Female	141	43.7
Location of primary tumera
Colon	108	33.5
Rectum	214	66.5
CEA level (ng/ul)
<5	141	43.7
≥5	182	56.3
CA19-9 level (U/ml)a
<37	240	74.8
≥37	81	25.2
Pathological typea
Protrude type	202	64.7
Infiltrating or ulcerative type	107	34.3
Others	3	1.0
Anastomat on surgerya
Yes	228	71.9
No	75	23.7
Unknown	14	4.4
Stage of Dukesa
I	39	12.1
II	142	44.1
III	119	37.0
IV	22	6.8
I+II	181	56.2
III+IV	141	43.8
Histological type
Adenocarcinoma	249	77.1
Mucinous adenocarcinoma	63	19.5
Other types	11	3.4
Degree of differentiation
Low	49	15.2
Medium	258	79.9
High	3	0.9
Unknown	13	4.0
Chemotherapy treatmenta
No	186	57.9
FOLFOX4	45	14.0
XELOX	22	6.9
LCF	44	13,7
5-Fu	6	1.9
Others	18	5.6
Metastasis
Yes	141	43.7
No	182	46.3
Prognosis
Death	136	42.1
living	146	45.2
Losing follow-up	41	12.7

CA19-9, carbohydrate antigen19-9; CEA, carcinoembryonic antigen; CI, confidence interval. aMissing data on subjects, tumor location, one case; CA19-9 level, two cases; pathological type, 11 cases; anastomat on surgery, six cases; stage of Dukes, 22 cases; chemotherapy treatment, 2 cases.

We find a marginally significant association between β-TRCP CNVs and CRC prognosis (amp v.s. wt, HRadjusted = 0.42, 95% CI: 0.19, 0.97, P = 0.050). In the additive model, β-TRCP CNVs (del + amp) is significantly associated with CRC prognosis (del + amp v.s. wt, HRadjusted = 0.39, 95% CI: 0.17, 0.88, P = 0.023) (Table 5, Fig. 1a,b). In the stratified analyses based on tumor location, the significant association between β-TRCP CNVs and CRC prognosis becomes marginally significant in rectal cancer (amp v.s. wt: HRadjusted = 0.22, 95% CI: 0.06, 0.86, P = 0.029; amp v.s. del + wt,: HRadjusted = 0.22, 95% CI: 0.06, 0.86, P = 0.029; del + amp v.s. wt: HRadjusted = 0.21, 95% CI: 0.06, 0.83, P = 0.026), but not significant in colon cancer (Table 5, Fig. 1c–e). There was no statistically significant association between other gene CNVs and colon or rectal cancer in analyses stratified by tumor location (Table 5).

Table 5

The relationships between gene CNVs and prognosis of CRC.

CNV Genotypes	Total patients (n = 323)					P Valueb	Colon cancer (n = 108)					P Valuec	Rectal cancer (n = 214)					P Valuec
	Patients (%)	5-Year Survival (%)	3-Year Survival (%)	OS (Mean (SD)) (month)	HR (95% CI)a		Patients (%)	5-year survival (%)	3-year survival (%)	OS (Mean (SD)) (month)	HR (95% CI)a		Patients (%)	5-year survival (%)	3-year survival (%)	OS (Mean (SD)) (month)	HR (95% CI)a
FBXW7
Wt	255 (83.6)	59	68	74.00 (2.53)	1.00		88 (85.4)	62	71	77.32 (4.35)	1.00		166 (82.5)	58	68	72.48 (3.09)	1
Del	27 (8.9)	68	76	82.75 (7.61)	0.62 (0.29, 1.33)	0.220	10 (9.7)	48	70	68.95 (11.96)	1.09 (0.41, 2.88)	0.868	17 (8.5)	81	81	90.74 (9.01)	0.31 (0.07, 1.26)	0.100
Amp	23 (7.5)	61	70	79.35 (8.18)	0.91 (0.46, 1.83)	0.793	5 (4.9)	20	40	35.40 (12.23)	3.39 (1.03, 11.18)	0.045	18 (9.0)	72	78	89.61 (7.71)	0.48 (0.18, 1.29)	0.146
Amp v.s. del + wt					0.95 (0.48, 1.90)	0.888					3.38 (1.05, 11.12)	0.045					0.51 (0.19, 1.38)	0.185
Del + amp v.s. wt					0.75 (0.44, 1.29)	0.299					1.57 (0.74, 3.34)	0.242					0.41 (0.18, 0.92)	0.031
MDM2
Wt	298 (94.6)	60	68	74.28 (2.35)	1.00		99 (96.1)	56	68	73.5 (4.18)	1.00		198 (93.8)	60	69	74.97 (2.84)	1
Del	9 (2.9)	67	67	75.22 (14.09)	0.72 (0.23, 2.29)	0.580	4 (3.9)	75	75	81.25 (20.59)	0.62 (0.08, 4.61)	0.640	5 (2.4)	60	60	55.40 (13.56)	1.13 (0.27, 4.80)	0.867
Amp	8 (2.5)	61	61	60.60 (9.15)	0.80 (0.25, 2.57)	0.707	0						8 (3.8)	61	61	60.60 (9.15)	0.96 (0.30, 3.13)	0.956
Amp v.s. del + wt					0.81 (0.25, 2.61)	0.727											0.96 (0.30, 3.11)	0.949
Del + amp v.s. wt					0.76 (0.33, 1.75)	0.517					0.62 (0.08, 4.61)	0.640					1.03 (0.40, 2.61)	0.957
SKP2
Wt	284 (95.3)	60	70	75.34 (2.39)	1.00		95 (95.0)	58	69	74.9 (44.23)	1.00		188 (95.5)	61	70	75.79 (2.90)	1
Del	6 (2.0)	50	67	68.50 (15.37)	1.31 (0.41, 4.17)	0.658	3 (3.0)	33	67	43.33 (16.85)	3.62 (0.80, 16.31)	0.094	3 (1.5)	67	67	84.00 (15.51)	0.75 (0.10, 5.55)	0.777
Amp	8 (2.7)	42	42	50.03 (11.55)	1.39 (0.51, 3.79)	0.526	2 (2.0)			19.50 (3.89)	4.68 (0.57, 38.39)	0.151	6 (3.0)	46	46	53.83 (12.53)	0.99 (0.30, 3.22)	0.985
Amp v.s. del + wt					1.38 (0.50, 3.78)	0.531					4.47 (0.55, 36.48)	0.162					0.99 (0.31, 3.22)	0.990
Del + amp v.s. wt					1.35 (0.62, 2.91)	0.447					3.92 (1.13, 3.66)	0.032					0.93 (0.33, 2.54)	0.865
β-TRCP
Wt	282 (91.0)	58	68	74.09 (2.42)	1.00		90 (89.1)	58	68	73.97 (4.41)	1.00		191 (91.8)	59	68	74.47 (2.89)	1
Del	4 (1.3)	100	100	90.53 (11.69)		0.962	3 (3.0)	100	100	86.00 (14.69)		0.976	1 (0.5)	100	100	79		0.971
Amp	24 (7.7)	66	74	63.25 (5.44)	0.42 (0.19, 0.97)	0.050	8 (7.9)	50	75	61.00 (8.08)	0.61 (0.18, 2.03)	0.421	16 (7.7)	74	74	64.01 (7.04)	0.22 (0.06, 0.86)	0.029
Amp v.s. del + wt					0.44 (0.19, 1.01)	0.052					0.63 (0.19, 2.08)	0.444					0.22 (0.06, 0.86)	0.029
Del + amp v.s. wt					0.39 (0.17, 0.88)	0.023					0.50 (0.16, 1.57)	0.230					0.21 (0.06, 0.83)	0.026
NEDD4-1
Wt	269 (89.7)	59	68	74.39 (2.46)	1.00		89 (87.9)	56	68	73.35 (4.44)	1.00		181 (90.5)	60	68.4	75.23 (2.94)	1
Del	2 (0.6)	50	50	46.00 (23.34)	1.02 (0.14, 7.66)	0.985	0						2 (1.0)	50	50	46.00 (23.34)	0.85 (0.11, 6.81)	0.878
Amp	29 (9.7)	65	72	74.79 (7.61)	0.82 (0.40, 1.70)	0.597	12 (12.1)	58	67	70.83 (12.27)	1.06 (0.41, 2.76)	0.900	17 (8.5)	77	76	70.12 (8.26)	0.73 (0.23, 2.35)	0.595
Amp v.s. del + wt					0.82 (0.40, 1.70)	0.597					1.06 (0.41, 2.76)	0.900					0.73 (0.23, 2.35)	0.595
Del + amp v.s. wt					0.84 (0.42, 1.67)	0.620					1.06 (0.41, 2.76)	0.900					0.75 (0.27, 2.09)	0.578

CI, confidence interval; HR, hazard ratio; OS, overall survival.

aadjusted for CEA and CA19-9 level before surgery, Dukes stage, pathological type and metastasis. bP < 0.05 in the survival analysis was considered statistically significant. cP < 0.025 in the stratified survival analysis by location was considered statistically significant.

Figure 1

Kaplan–Meier curves of overall survival (OS) according to the five genes CNVs in patients with rectal cancer.

(a) β-TRCP CNVs in CRC; (b) β-TRCP CNVs in combined model in CRC; (c) β-TRCP CNVs in rectal cancer; (d) β-TRCP amplification in rectal cancer; (e) β-TRCP CNVs in combined model in rectal cancer.

Discussion

To our knowledge, this is the first study on the association between germline CNVs of FBXW7, MDM2, SKP2, β-TRCP, NEDD4-1 and CRC risk and prognosis. In this study, MDM2 CNVs significantly increase CRC risk, while SKP2 CNVs significantly decrease CRC risk. We find evidence of three significant gene-environment interactions that increase risk of CRC: SKP2 CNVs interact with consumption of fruit and fish consumption, and FBXW7 CNVs interact with pork consumption. We also observe a significant association between β-TRCP CNVs and CRC prognosis.

We observe a significant association between MDM2 amplification CNVs and CRC risk. However, there are few MDM2 amplification among patients and controls (9 and 2 respectively), which limits statistical power. Because both amplification and deletion of MDM2 can increase CRC risk, the del + amp v.s. wt model can be viewed as a conservative estimate of the effect of MDM2 on CRC risk. The amplification of MDM2 may increase CRC risk by up to 14.40-fold, and the del + amp genotype of MDM2 may also increase CRC risk by 6.35-fold. MDM2 amplification was observed in 26 of 284 (9%) colorectal cancer tissue samples33, 14 of 80 (18%) CRCs tumor tissue samples34 and almost one-third of sarcomas16. MDM2 could promote tumorigenesis by acting as a positive regulator of p53 or independent of p5335. SNP data also indicate that even small differences in MDM2 levels may affect cancer risk36. Moreover, MDM2 also acts as a tumor suppressor through the Akt pathway, inducing the ubiquitination and degradation of NFAT (an invasion-promoting factor), thereby blocking cancer cell motility and invasion37. This could explain the significant association between MDM2 CNVs and increased CRC risk in the del + amp v.s. wt model in our study. There is no significant association between MDM2 CNVs and CRC prognosis. The role of MDM2 in cancer prognosis is controversial, and may be affected by tumor variety and racial differences313839.

We observed that SKP2 CNVs (del + amp) are significantly associated with a 68% decreased risk of CRC. The overexpression of SKP2 was associated with tumor differentiation, malignant transformation, and prognosis of malignant tumors1118. SKP2 gene amplification is commonly observed in metastatic tumors but not in early stage cancers1840. Thus SKP2 gene amplification is likely to be associated with advanced tumor progression. In our study, 60.9% of CRC patients were in stage I or II (Table 1). This may explain the non-significance of the association between SKP2 CNVs and CRC risk. We did observe significant interactions between SKP2 CNVs and fish or fruit consumption. Fish consumption has been reported to have protective effects in CRC41, which may be attributable to the omega-3 polyunsaturated fatty acids (PUFAs) in fish42. Omega-3s function as an anti-inflammatory, and is expected to have a function analogous to aspirin. Aspirin has been shown to reduce the incidence of CRC in both observational studies and randomized trials4344. Dietary fiber in fruit is hypothesized to reduce the risk of CRC. Potential mechanisms for the protective effect dietary fibers include dilution of fecal carcinogens, reduction of transit time of feces through the bowel, and increased production of short chain fatty acids454647.

FBXW7 serves as a substrate adaptor for SCF ubiquitin ligase complex and mediates the recognition and binding of substrate proteins. SCF degrades several proteins with important roles in cell growth, proliferation, differentiation, and survival48. Previous studies have reported a tumor-suppressive function of FBXW7 in colorectal tumor cells or tissues2330, and copy number loss of FBXW7 gene in tumor tissue was reported to be significantly associated with worse CRC prognosis23. The blood level of FBXW7 expression has also been associated with the prognosis of breast cancer patients49. However, we did not observe any significant association between FBXW7 CNVs and CRC risk or prognosis. The study by Chang et al. also found a non-significant association between FBXW7 mRNA expression and CRC risk10, which is consistent with our results. About 6% of tumors harbor FBXW7 loss-of-function variants, with different variants detected in different tumor types. This might reflect tissue-specific roles of FBXW7 substrates48. A significant interaction effect has been observed between FBXW7, pork intake, and increased CRC risk. An updated meta-analysis of all prospective studies showed that the risk of CRC increased by 29% for every 100 g/d of red meat consumed50. The hard muscle fibers and high fat content in red meat may be the source of this association.

We found CRC patients with β-TRCP CNVs, have a better prognosis with a 58–61% OS increase. β-TRCP is the component of the ubiquitin ligase complex targeting β-catenin and NF-ΚB for proteasome degradation, which may contribute to the inhibition of apoptosis and to tumor metastasis25. Moreover, enhanced activity of β-TRCP has been widely observed in colorectal tumor cells and primary tumors1925. The dual function of β-TRCP might explain the significant association between β-TRCP CNVs, and improved CRC prognosis. Different mechanisms of oncogenesis in rectal vs. colon cancer may explain why β-TRCP CNVs are only associated with rectal cancer prognosis in our study51. However we do not observe a significant association between β-TRCP CNVs and CRC risk. Mutations in β-TRCP are rarely detected in CRC, which is consistent with our results5253.

Prior work has indicated that NEDD4-1 may promote tumorigenesis by decreasing PTEN protein level, or through interference with the PI3K/AKT signaling pathway5455. NEDD4-1 is overexpressed in cancer cell lines1250, animal models565758, and in human cancer tissues596061. However, we find no significant association between NEDD4-1 CNVs and CRC risk or prognosis. Meanwhile, there have been no studies focused on the effect of NEDD4-1 CNVs in peripheral blood on CRC risk and/or prognosis. One study indicated that SCF can negatively regulate NEDD4-1 stability, and β-TRCP-mediated destruction of the NEDD4-1 oncoprotein may inhibit cell proliferation and migration62. This suggests that epistatic effects between β-TRCP and NEDD4-1 may modify many signaling pathways. Further research is required to shed light on the relationship between these genes, and any differences that may exist between their functions in the germline vs. their function in tumor.

As in any case-control or prospective survival study, we must consider the limitations of our study. First, recall bias may be inevitable in the collection of data on environment factors. Second, we collected the frequency of soybean, sausage, fried food, and leftovers consumption without collecting information regarding quantity, which limits the statistical power of our analysis of gene-dietary interactions.

We find that MDM2 and SKP2 CNVs are significantly associated with CRC risk. In addition, we observe significant interaction effects between SKP2 CNVs, fish or fruit consumption, and between FBXW7 CNVs and pork intake, and CRC risk. There is a significant association between β-TRCP CNVs and CRC prognosis. Further research with larger sample sizes and more detailed functional evaluation will be required to confirm our results.

Materials and Methods

Subjects

After obtaining informed consent from study subjects, and approval from the Institutional Research Board of Harbin Medical University, we carried out the experiment in accordance with the relevant guidelines, including any relevant details. Informed consent was obtained from all subjects. We identified CRC patients who underwent surgery at the Cancer Hospital of Harbin Medical University, based on pathologic diagnosis without pre-selection. We excluded patients with neuroendocrine carcinoma, malignant melanoma, non-Hodgkin’s lymphoma, gastrointestinal stromal tumors, and Lynch syndrome CRC. From November 1st, 2004 to May 1st, 2010, we recruited 518 primary CRC patients. During the same period, we collected cancer free control subjects from the 2nd Affiliated Hospital of Harbin Medical University. We excluded controls with history of gastrointestinal disease according to self-report. 518 controls matched for age, gender, and residence were recruited.

Data collection

We interviewed each participant face-to-face using a structured questionnaire with questions on demographic characteristics (age, gender, height and weight education and occupation), history of physical exercise, family history of cancer, and dietary status during the 12 months preceding the interview. We collected clinical information from medical records on tumor size, Dukes stage, chemotherapy treatment, histological and pathological types, and level of serum carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (CA19-9). We followed up with 323 patients from November 2004 to March 2014. Overall survival (OS) was defined at the primary end point in our study. Survival time was calculated from the date of cancer diagnosis to death from colorectal cancer or other causes, or the time of follow-up. The date and cause of death of CRC patients were validated through the medical certification of death and the Harbin death registration system.

DNA extraction and CNV detection

We extracted DNA from all 1036 blood samples (518 CRC and 518 controls) using QIAGEN DNeasy Blood & Tissue Kit. We detected FBXW7, MDM2, SKP2, β-TRCP, and NEDD4-1 copy numbers using custom designed TaqMan Copy Number Assays (Supplementary Table S1). The quantitative assays were performed using the 7500 Fast Real-Time polymerase chain reaction machine in 96-well plates with a 10 ul reaction volume containing 20 ng DNA, 5 ul TaqMan Universal PCR Master Mix, 0.5 ul of the CNV assay, and 0.5 ul of the reference RNase P assay (Applied Biosystems, Carlsbad, Calif). The reaction was completed using the following cycling conditions: 95 °C for 15 seconds and 60 °C for 1 minute for 40 cycles. We used one sample with 2 copies of each CNV as a quality control in every 96-well assay plate (Supplementary Fig. 1). CNVs for each sample were detected three times. We analyzed data using 7500 software v2.0.6 (Applied Biosystems) to quantify the amplification cycle, and then imported the data to Copy Caller version 2.0 (Applied Biosystems) to estimate the gene copy numbers in every sample.

Statistical analyses

We calculated the Hardy-Weinberg equilibrium in controls and compared using Fisher’s exact test. We evaluated homogeneity between cases and controls using Student’s t-test for continuous variables and a Chi-squared test for categorical variables. The unbalanced factors between the two groups were controlled for in a multivariable logistic regression for each gene, and in a multivariable logistic regression for gene-environment interactions. We used odds ratios (OR) and corresponding 95% confidence intervals (95% CI) to estimate the associations between FBXW7, MDM2, SKP2, β-TRCP and NEDD4-1 CNVs and CRC risk via conditional logistic regression. We performed crossover analyses to evaluate gene-environment interaction effects on the risk of CRC with four types of OR (ORe, ORg, OReg, ORi). We adjusted the heterogeneous demography characteristics in the conditional logistic regression. We defined 2 copies as the wild type (wt), more than 2 copies as the amplification type (amp) and less than 2 copies as the deletion type(del). Two additive models were applied in the conditional logistic regression analysis: amp v.s. del + wt and del + amp v.s. wt to estimate the association between CNVs CRC risk and prognosis. All statistical tests were two-sided, P value < 0.05 in the overall analysis. Adding a Bonferroni correction, a P value < 0.025 was used in stratified analyses. We used a multiple interpolation method to fill missing values in questionnaire responses (Supplementary Tables S4–S8). All statistical analyses were performed using SAS, version 9.2 (SAS Institute Inc.Cary, NC, USA).

Additional Information

How to cite this article: Bi, H. et al. Copy number variation of E3 ubiquitin ligase genes in peripheral blood leukocyte and colorectal cancer. Sci. Rep. 6, 29869; doi: 10.1038/srep29869 (2016).