Carcinogenesis vol.32 no.3 pp.318–326, 2011
doi:10.1093/carcin/bgq245
Advance Access publication November 18, 2010
Associations between genetic variation in RUNX1, RUNX2, RUNX3, MAPK1 and eIF4E
and risk of colon and rectal cancer: additional support for a TGF-b-signaling pathway
Martha L.Slattery, Abbie Lundgreen, Jennifer S.Herrick,
Bette J.Caan1, John D.Potter2 and Roger K.Wolff
Department of Internal Medicine, University of Utah Health Sciences Center,
295 Chipeta Way, Salt Lake City, UT 84108, USA, 1Division of Research,
Kaiser Permanente Medical Care Program, Oakland, CA 94612, USA and
2
Cancer Prevention Program, Fred Hutchinson Cancer Research Center,
Seattle, WA 98109-1024 , USA
To whom correspondence should be addressed. Tel: þ1 801 585 6955;
Fax: þ1 801 581 3623;
Email: [email protected]
The Runt-related transcription factors (RUNX), mitogen-activated
protein kinase (MAPK) 1 and eukaryotic translation initiation
factor 4E (eIF4E) are potentially involved in tumorigenesis. We
evaluated genetic variation in RUNX1 (40 tagSNPs), RUNX2
(19 tagSNPs), RUNX3 (9 tagSNPs), MAPK1 (6 tagSNPs), eIF4E
(3 tagSNPs), eIF4EBP2 (2 tagSNP) and eIF4EBP3 (2 tagSNPs) to
determine associations with colorectal cancer (CRC). We used
data from population-based studies (colon cancer n 5 1555 cases,
1956 controls; rectal cancer n 5 754 cases, 959 controls with
complete genotype data). Four statistically significant tagSNPs
were identified with colon cancer and three tagSNPs were identified with rectal cancer. Whereas the independent risk estimates
for each of the tagSNPs ranged from 1.21 to 1.52, the combined
risk was greater than additive for any of the three combined highrisk genotypes {combined risk range 1.98 [95% confidence interval (CI) 1.45, 2.70] for eIF4E, RUNX1 and RUNX3 to 3.32 [95% CI
1.34, 8.23] for eIF43, RUNX2 and RUNX3}. For rectal cancer, the
strongest association was detected for the combined genotype of
RUNX1 and RUNX3 (odds ratio 1.87 95% CI 1.22, 2.87). Associations with specific molecular tumor phenotypes showed consistent and strong associations for CIMP1/MSI1 tumors where the
risk estimates were consistently >10-fold and lower confidence
bounds were over 3.00 for high-risk genotypes defined by RUNX1,
RUNX2 and RUNX3. For CIMP1/KRAS2-mutated colon tumors,
the combined risk for high-risk genotypes of RUNX2, eIF4E and
RUNX1 was 7.47 (95% CI 1.58, 35.3). Although the associations
need confirmation, the findings and their internal consistency underline the importance of genetic variation in these genes for the
etiology of CRC.
Introduction
The Runt-related transcription factors (RUNX) are thought to play an
important role in carcinogenesis in addition to their role in normal
development (1). The three RUNX genes, RUNX1, RUNX2 and
RUNX3, although widely expressed, have tissue-specific properties
(2). Of the three, RUNX3 has been shown to be specifically associated
with gastrointestinal tract development (3). Studies in RUNX3 knockout mice have shown defects in apoptotic response to transforming
growth factor (TGF)-b; RUNX2 transgenic mice have been shown to
be hypersensitive to TGF-b (4). All three RUNX genes have potential
for involvement in colorectal cancer (CRC) etiology given their role
in signaling cascades mediated by TGF-b and bone morphogenetic
Abbreviations: BMP, bone morphogenetic protein; CI, confidence interval;
CIMP, CpG island methylator phenotype; CRC, colorectal cancer; eIF4E,
eukaryotic translation initiation factor 4E; MAPK, mitogen-activated protein
kinase; MSI, microsatellite instability; NF-jB, nuclear factor-kappa B; OR,
odds ratio; RUNX, runt-related transcription; SNP, single nucleotide polymorphisms; TGF, transforming growth factor.
protein (BMP) (4–7); all three RUNX genes have been shown to bind
Smads that are also involved in the TGF-b signaling pathway (8–10).
Mitogen-activiated protein kinase (MAPK) 1, also known as extracellular signal-regulated kinase 2, is involved in eukaryotic signal
transduction. MAPK1 has been shown to activate RUNX2 (11). Like
RUNX, MAPK1 is involved in the TGF-b-signaling pathway including Smad signaling (12,13) Eukaryotic translation initiation factor 4E
(eIF4E) is a translational regulator that acts downstream of Akt and
mTOR, promoting Akt’s action in tumorigenesis (14). eIF4E has been
shown to play a key role in cell growth and has been reported to be
overexpressed in colon tumors (15). Expression of eIF4E in human
colon cancer cells has been shown to promote the TGF-b stimulation
of adhesion molecules (16).
Together, Runx, MAPK1 and eIF4E appear to be integral parts in
the regulation of the TGF-b, Smad and BMP signaling pathways, each
of which plays an important role in the etiology of colon and rectal
cancer. However, little is known about how genetic variation in these
genes relate to colon and rectal cancer. Furthermore, although other
genes in the TGF-b-signaling pathway have been linked to CpG island
methylator phenotype (CIMPþ) and microsatellite instability (MSIþ)
tumors, it is unknown whether genetic variation in these genes contributes to specific molecularly defined phenotypes colorectal cancer
(17). In this study, we evaluated the association between genetic variability in these genes and colon and rectal cancer and with specific
tumor molecular phenotypes. We further report how these genes interact with other genes in the TGF-b-signaling pathway.
Methods
Two study populations are included in these analyses. The first study, a
population-based case–control study of colon cancer, included cases (n 5
1555 with complete genotype data) and controls (n 5 1956 with complete
genotype data) identified between 1 October 1991 and 30 September 1994
(18) living in the Twin Cities Metropolitan Area or a seven-county area of
Utah or enrolled in the Kaiser Permanente Medical Care Program of Northern
California (KPMCP). The second study, with identical data collection methods, included cases with cancer of the rectosigmoid junction or rectum (n 5
754 cases and n 5 959 controls with complete genotype data) who were
identified between May 1997 and May 2001 in Utah and at the KPMCP
(19). Eligible cases were between 30 and 79 years of age at the time of diagnosis, living in the study geographic area, English speaking, mentally competent to complete the interview and with no previous history of CRC and no
previous diagnosis of familial adenomatous polyposis, ulcerative colitis or
Crohn’s disease. Cases who did not meet these criteria were ineligible as were
individuals who were not black, white or Hispanic for the colon cancer study.
A rapid reporting system was used to identify cases within months of diagnosis.
Controls were matched to cases by sex and by 5 years age groups. At
KPMCP, controls were randomly selected from membership lists; in Utah,
controls 65 years were randomly selected from the Health Care Financing
Administration lists and controls ,65 years were randomly selected from
driver’s license lists. In Minnesota, controls were selected from driver’s license
and state-identification lists. Study details have been previously reported
(20,21).
Interview data collection
Data were collected by trained and certified interviewers using laptop computers. All interviews were audiotaped as described previously and reviewed
for quality control purposes (22). The referent period for the study was 2 years
prior to diagnosis for cases and selection for controls. Detailed information was
collected on diet, physical activity, medical history, reproductive history, family history of cancer, regular use of aspirin and nonsteroidal anti-inflammatory
drugs and body size.
Tumor marker data
We have previously evaluated tumors for CIMP, MSI, TP53 mutations and
KRAS2 mutations (23–26) and were therefore able to evaluate variation in
the specified genes in relation to molecularly defined subsets of CRC. Details
Ó The Author 2010. Published by Oxford University Press. All rights reserved. For Permissions, please email: [email protected]
318
Additional support for a TGF-b signaling pathway
for methods used to evaluate epigenetic and genetic changes have been described (23–26). Because of the rarity of MSIþ rectal tumors (27), we did not
evaluate MSI in rectal tumors.
TagSNP selection and genotyping
TagSNPs were selected for genes RUNX1, RUNX2, RUNX3, MAPK1, eIF4E,
eIF4EBP2 and eIF4EBP3 using the following parameters: linkage disequilibrium blocks using a Caucasian linkage disequilibrium map with r2 5 0.8;
minor allele frequency . 0.1; range 5 1500 bp from the initiation codon
to þ1500 bp from the termination codon and one single nucleotide polymorphisms (SNP)/linkage disequilibrium bin. All markers were genotyped using
a multiplexed bead array assay based on GoldenGate chemistry (Illumina, San
Diego, CA). A genotyping call rate of 99.85% was attained. Blinded internal
replicates represented 4.4% of the samples. The duplicate concordance rate
was 100% genotyping of other genes along the candidate pathway, which were
assessed for their interactive effects with RUNX, MAPK1 and eIF4E were
genotyped on the same platform. Table I describes tagSNPs associated with
colon or rectal cancer, whereas supplementary Table 1 (available at Carcinogenesis Online) has a listing of all tagSNPs included on the platform.
Statistical methods
All statistical analyses were performed using SASÒ version 9.2 (SAS Institute,
Cary, NC) unless otherwise stated. We report odds ratios (ORs) and 95% confidence intervals (CIs) derived from multiple logistic regression models for colon
and rectal cancer separately based on minimal adjustments for age, sex, race and
study center. Stepwise regression models were used to identify the tagSNPs and
Table I. Description of Runx, eIF4E and MAPK1 genes in the study population
Major/minor
Symbol
Alias
Location
SNP
Allele
MAF-NHW
MAF-Hispanic
MAF-AA
FDR (HWE)
EIF4E
CBP
EIF4E1
EIF4EL1
MGC111573
4EBP2
4E-BP3
ERK
ERK2
ERT1
MAPK2
P42MAPK
PRKM1
PRKM2
p38
p40
p41
p41 mapk
AML1
AML1-EVI-1
AMLCR1
CBFA2
EVI-1
PEBP2aB
4q21–q25
rs11727086
rs12498533
A/G
A/C
0.26
0.42
0.19
0.46
0.06
0.20a
0.93
0.93
10q21–q22
5q31.3
22q11.21
rs7078987
rs250425
rs11913721
rs8136867
rs2298432
rs9610375
A/G
C/T
A/C
A/G
C/A
G/T
0.46
0.24
0.41
0.47
0.38
0.46
0.43
0.25
0.43
0.37
0.26
0.49
0.13
0.12
0.29
0.36
0.07
0.4
0.93
0.93
0.9
1
1
0.97
21q22.3
rs1474479
rs2248720
rs8134380
rs2071029
rs2242878
rs2834645
rs2300395
rs1981392
rs11702779
rs7279123
rs11701453
rs7280028
rs2268281
rs2834650
rs1883067
rs7750470
rs10948238
rs1321075
rs2819863
rs12333172
rs12208240
rs2819854
rs1316330
G/A
A/C
A/T
G/A
C/T
T/C
C/T
T/C
G/A
C/T
G/C
T/C
A/G
C/T
A/G
T/C
C/T
C/A
G/C
C/T
G/A
C/T
G/T
0.35
0.49
0.44
0.14
0.19
0.23
0.3
0.4
0.35
0.25
0.21
0.19
0.16
0.1
0.08
0.19
0.4
0.14
0.1
0.2
0.08
0.48
0.25
0.27
0.46a
0.36
0.22
0.18
0.15
0.25
0.49a
0.46
0.21
0.16
0.16
0.29
0.07
0.07
0.22
0.36
0.24
0.07
0.18
0.12
0.49a
0.17
0.32
0.24
0.17
0.35
0.05
0.05
0.19a
0.42
0.38
0.47
0.21
0.4
0.27
0.02
0.02
0.36
0.34a
0.2
0.03
0.05
0.04
0.41a
0.05
0.96
0.96
0.14
1
0.98
1
1
1
1
1
1
0.62
1
1
1
0.91
0.95
0.95
1
1
1
0.98
1
rs2135756
rs2236850
rs906296
rs6672420
rs6688058
A/G
T/C
C/G
A/T
G/A
0.5
0.44
0.23
0.48
0.13
0.45
0.44
0.2
0.37a
0.19
0.41
0.14a
0.28
0.48
0.16
0.96
0.62
1
0.83
0.89
EIF4EBP2
EIF4EBP3
MAPK1
RUNX1
RUNX2
RUNX3
RP1-166H4.1
AML3
CBFA1
CCD
CCD1
MGC120022
MGC120023
OSF2
PEA2aA
PEBP2A1
PEB2A2
PEBP2aA
PEBP2aA1
RP3-398I9.1
AML2
CBFA3
FLI34510
MGC16070
PEBP2aC
6p21
1p36
a
major/minor allele differs from NHW population. FDR (HWE), false discovery rate adjusted P value for Hardy–Weinberg Equilibrium test. Minor allele
frequency (MAF) based on control population; HWE based on NHW control population (sample sizes range from 2519 to 2652).
NHW, non-Hispanic white; AA, African American
319
M.L.Slattery et al.
their inheritance models that contributed uniquely to the overall fit of the model
for colon and rectal cancer; separate stepwise models were used to identify
tagSNPs associated with specific molecular subtypes of tumors. Inclusion in
the regression model was based on a score chi-square significance level of
0.05, whereas exclusion was determined based on a Wald chi-square 0.05 significance level. In addition to the minimal adjustments previously stated, the
subset of SNPs returned from stepwise regression was also used as adjustment
variables. Subsequent interaction analyses were based on tagSNPs identified as
being statistically significant from stepwise regression. Adjusted multiple comparison P values were estimated taking into account all tagSNPs within the gene
using the methods of Conneely and Boehnke (28) implemented in R version
2.11.0 (R Foundation for Statistical Computing, Vienna, Austria).
We evaluated interaction between RUNX1, RUNX2, RUNX3, MAPK1 and
eIF4E and its binding proteins on the one hand and BMP-related genes, TGFb1
and its receptors, Smad3, Smad4, Smad7 and nuclear factor-kappa B (NF-jB) 1
on the other hand given the biologic links to the TGF-b-signaling pathway.
Possible interactions between SNPs and sex, age (30–64 or 65–79), recent
aspirin or nonsteroidal anti-inflammatory drugs use, recent estrogen use and
body mass index (,25, 25–30, .30) also were evaluated because of the
mechanisms hypothesized for these genes. P values for interaction were determined by comparing a full model that included a categorical multiplicative
interaction term to a reduced model without such an interaction term, using
a likelihood ratio test. We evaluated risk estimates based on high-risk genotypes for each SNP as well as for the combined high-risk genotypes for those
tagSNPs that were independently associated with colon or rectal cancer or with
specific tumor subsets. High-risk genotypes were defined as those genotypes
associated with statistically significant increased risk of colon or rectal cancer.
The combined high-risk genotypes were determined based on the risk estimate
for combinations of SNPs using either a dominant or a recessive model and
compared with the referent genotype. Risk estimates were based on sets of
combined high-risk genotypes compared with individuals without any of the
high-risk genotypes designated (shown by asterisks in the table).
Tumors were defined by specific molecular alterations: any TP53 mutation,
any KRAS2 mutation, MSIþ, CIMPþ defined as at least two of five markers
methylated and a combination of CIMPþ/KRAS2-mutated or CIMPþ/MSIþ.
As the proportion of MSIþ tumors in the rectal cases was ,3% (27), we did
not examine these tumor markers. Estimates of risk for molecular tumor phenotypes were made relative to controls.
Results
Of the genes assessed, four significant tagSNPs were identified that
best represented the statistically significant associations with colon
cancer, and similarly three tagSNPs were identified that captured the
association with rectal cancer. One additional RUNX1 and two RUNX2
tagSNPs also were independently, statistically significantly associated
with colon cancer. RUNX1 rs7279123 (OR 1.17 95% CI 1.02, 1.35 for
the CT/TT genotype) and RUNX2 rs12208240 (OR 0.27 95% CI 0.08,
0.97 for the AA genotype) and rs2819854 (OR 1.21 95% CI 1.03, 1.42
for the TT genotype) were associated with colon cancer. Genes shown
in Table II, best represented the independent and combined risk based
on P values, magnitude of the association and frequency of the genotypes. The adjusted P values for multiple comparisons for RUNX1
were all above 0.2; the adjusted P value for RUNX2 rs12333172 was
0.12; the adjusted P value for RUNX3 rs667240 was 0.02; the adjusted
P value for eIF4E was 0.02. Single SNP associations with colon
cancer were generally modest and the combined risks from the
high-risk genotypes were generally greater than that expected for an
Table II. Associations between Runx, MAPK1, eIF4E and colon and rectal cancer
HRG
EIF4E
RUNX2
RUNX1
RUNX3
Colon
rs11727086 (A . G)
rs1233372(C . T)
rs2834645 (C . T)
rs6672420 (A . T)
Controls
AG/GG
1
1
1
1
2
2
2
2
2
2
3
3
3
3
4
1
1
1
2
2
2
3
TT
TT
AA
a
Cases
N
839
65
1160
495
33
517
219
43
17
273
20
7
119
11
3
733
76
1001
457
37
472
212
52
30
281
26
15
130
19
8
1.22
1.52
1.21
1.24
1.62
1.49
1.47
1.76
2.37
1.56
2.13
3.32
1.98
2.69
4.52
(1.07, 1.40)
(1.08, 2.13)
(1.06, 1.39)
(1.07, 1.45)
(1.00, 2.62)
(1.23, 1.81)
(1.18, 1.83)
(1.16, 2.68)
(1.30, 4.34)
(1.26, 1.93)
(1.16, 3.91)
(1.34, 8.23)
(1.45, 2.70)
(1.26, 5.77)
(1.18, 17.39)
OR
(95% CI)
1.43
1.36
1.28
1.24
1.87
1.78
1.64
(1.04, 1.95)
(1.04, 1.78)
(1.02, 1.59)
(0.57, 2.70)
(1.22, 2.87)
(1.23, 2.57)
(0.57, 4.74)
RUNX1
MAPK1
RUNX3
Rectal
rs11702779 (G . A)
rs11913721 (A . C)
rs6672420 (A . T)
Controls
Cases
GG/GA
AA/AC
TT
Nc
Nd
(95% CI)
N
OR
b
777
784
218
630
162
180
133
631
647
210
542
171
177
145
HRG, high-risk genotypes. Shaded ORs indicate risk estimates greater than additive effect. Rows are not mutually exclusive and include all individuals with the
starred high-risk genotype versus those without any of the starred high-risk genotypes.
a
1956 controls total for colon.
b
1555 cases total for colon.
c
959 controls total for rectal.
d
754 cases total for rectal.
320
Additional support for a TGF-b signaling pathway
Table III. Associations between Runx, MAPK1, eIF4E and colon tumor mutations
HRG
MAPK1
RUNX2
RUNX1
CIMPþ
rs11913721 (A . C)
rs12333172 (C . T)
rs2071029 (G . A)
Controls
CC
1
1
1
2
2
2
3
295
65
559
13
83
17
2
RUNX2
RUNX3
EIF4EBP2
RUNX1
KRAS2
rs10948238 (C . T)
rs6672420 (A . T)
rs7078987 (A . G)
rs8134380 (A . T)
Controls
AA/AT
GG
AA
RUNX2
RUNX1
EIF4EBP3
RUNX3
TP53
rs12333172 (C . T)
rs2248720 (A . C)
rs250425 (C . T)
rs6672420 (A . T)
Controls
AA
RUNX3
MAPK1
MSIþ
rs2242878 (C . T)
rs7078987 (A . G)
rs906296 (C . G)
rs9610375 (G . T)
Controls
(95% CI)
(95% CI)
EIF4EBP2
OR
d
OR
RUNX1
(1.18, 2.05)
(1.01, 1.81)
(1.04, 1.78)
(1.21, 1.94)
(1.37, 3.18)
(1.07, 3.19)
(1.61, 3.91)
(1.21, 2.73)
(1.49, 3.43)
(1.50, 3.35)
(1.08, 4.50)
(2.39, 8.82)
(2.01, 10.25)
(1.90, 6.36)
(2.99, 25.09)
Cases
Cases
1.55
1.35
1.36
1.53
2.09
1.85
2.51
1.81
2.26
2.24
2.21
4.59
4.54
3.48
8.66
(1.12, 2.76)
(1.04, 1.60)
(1.06, 1.60)
(1.17, 1.79)
(1.01, 6.92)
(1.48, 5.00)
(1.73, 7.33)
(1.26, 2.33)
(1.22, 2.48)
(1.40, 2.53)
(2.69, 47.88)
(1.12, 18.55)
(2.19, 15.81)
(1.09, 2.92)
(1.73, 183.42)
(95% CI)
1.76
1.29
1.30
1.44
2.65
2.72
3.56
1.71
1.74
1.88
11.35
4.55
5.89
1.78
17.80
OR
c
29
162
328
167
7
18
14
102
52
108
6
4
9
30
3
GG
(1.04, 1.99)
(1.37, 3.92)
(1.00, 1.73)
(0.07, 4.28)
(1.14, 3.26)
(0.90, 7.00)
65
503
1120
495
11
31
17
273
136
282
3
4
8
79
1
GG
1.44
2.32
1.32
0.55
1.93
2.51
undefined
N
AA/AG
(95% CI)
N
Cases
81
284
88
144
62
18
34
71
113
39
13
28
10
30
8
CC
55
20
94
1
20
5
0
N
AA
N
OR
b
316
1504
413
605
255
69
94
321
469
123
58
77
18
96
16
Cases
N
CT/TT
1
1
1
1
2
2
2
2
2
2
3
3
3
3
4
N
TT
1
1
1
1
2
2
2
2
2
2
3
3
3
3
4
GA/AA
TT
1
1
1
1
2
2
2
2
2
2
3
3
3
3
4
TT
a
e
N
N
638
1514
115
559
485
35
194
90
429
29
25
145
9
23
5
80
156
17
72
64
8
36
14
57
4
7
27
1
3
1
1.60
1.51
1.58
1.60
2.66
2.82
2.68
2.27
2.91
1.77
7.07
5.16
1.39
3.49
6.88
(1.17, 2.17)
(1.00, 2.30)
(0.92, 2.69)
(1.17, 2.18)
(1.44, 4.94)
(1.26, 6.31)
(1.73, 4.14)
(1.13, 4.57)
(1.56, 5.45)
(0.61, 5.15)
(2.38, 21.01)
(2.16, 12.35)
(0.17, 11.29)
(0.88, 13.83)
(0.63, 75.06)
321
M.L.Slattery et al.
Table III. Continued
RUNX2
EIF4E
RUNX1
CIMPþ and KRAS2
rs10948238 (C . T)
rs11727086 (A . G)
rs1474479 (G . A)
Controls
TT
1
1
1
2
2
2
3
GG/GA
N
316
839
1705
125
268
721
103
N
19
40
70
13
19
38
13
RUNX2
RUNX1
MAPK1
RUNX3
CIMPþ and MSIþ
rs1321075 (C . A)
rs2242878 (C . T)
rs8136867 (A . G)
rs906296 (C . G)
Controls
CT/TT
GG
GG
N
1437
638
411
115
463
320
86
146
35
20
105
22
14
7
3
Cases
N
OR
(95% CI)
f
CC
1
1
1
1
2
2
2
2
2
2
3
3
3
3
4
AG/GG
Cases
1.79
1.68
2.53
3.63
3.27
3.38
7.47
(1.04, 3.07)
(1.04, 2.71)
(0.91, 7.00)
(1.81, 7.30)
(1.08, 9.90)
(0.80, 14.31)
(1.58, 35.32)
OR
(95% CI)
1.75
1.94
2.20
2.03
3.30
5.75
3.75
4.21
4.07
3.91
9.75
11.00
10.30
4.36
76.32
(1.04, 2.94)
(1.31, 2.87)
(1.46, 3.33)
(1.10, 3.76)
(1.61, 6.77)
(2.55, 12.94)
(1.72, 8.21)
(2.30, 7.68)
(1.61, 10.33)
(1.09, 14.00)
(3.14, 30.31)
(3.37, 35.96)
(1.78, 59.71)
(0.50, 38.03)
(3.10, 1879.34)
g
90
52
39
13
44
30
12
19
6
3
14
6
2
1
1
HRG, high-risk genotypes. Rows are not mutually exclusive and include all individuals with the starred high-risk genotype versus those without any of the starred
high-risk genotypes.
a
1956 controls total.
b
272 CIMPþ cases total.
c
348 KRAS2 cases total.
d
516 TP53 cases total.
e
185 MSIþ cases total.
f
74 CIMPþ and KRAS2 cases total.
g
108 CIMP and MSIþ cases total.
additive model. Nearly all of the two-way SNP combinations had
an increased colon cancer risk greater than that would be expected
on an additive scale. Combinations of three high-risk genotypes
were generally associated with a 2- to 3-fold increased risk of colon
cancer.
Similar risk estimates were observed for rectal cancer, although
only three tagSNPs in these genes captured the increased risk,
RUNX1, MAPK1 and RUNX3. The following two RUNX1 tagSNPs
were identified as being associated with rectal cancer risk, although
they did not substantially alter the combined risk and were therefore
omitted from the Table II: rs11701453 (OR 1.29 95% CI 1.05, 1.59 for
the GC/CC genotype) and rs7280028 (OR 0.80 95% CI 0.65, 0.99 for
the TC/CC genotype). Half of the combinations of high-risk genotypes presented had an increased rectal cancer risk that was greater
than additive. The adjusted P value for RUNX1 rs11702799 was 0.43,
for RUNX3 rs667240 was 0.17 and for MAPK1 was 0.10.
Various tagSNPs were associated with specific colon tumor molecular phenotypes (Table III). CIMPþ tumors were associated with
variants in MAPK1, RUNX2 and RUNX1 with individual risks ranging
from an OR of 1.32 for RUNX1 rs2071029 to 2.32 for RUNX2
rs12333172. Associations with mutations in KRAS2 and TP53 were
more stable (the result of more cases with these tumor molecular
subtypes). Four tagSNPs best-characterized associations with KRAS2:
RUNX2 rs10948238, RUNX3 rs6672420, eIF4EBP2 rs7078987 and
RUNX1 rs8134380. Associations ranged from a statistically significant OR of 1.35 for RUNX3 to greater than a 4-fold increased risk with
various combinations of high-risk genotypes. Although imprecise,
having all four high-risk genotypes were associated with more than
322
an 8-fold increased risk of colon cancer (OR 8.66 95% CI 2.99,
25.09). Similar levels of risk were seen for TP53 for both independent
and combined risk estimates. RUNX2 rs12333172, RUNX1
rs2248720, eIF4EBP3 rs250425 and RUNX3 rs6672420 best captured
the risk-associated TP53-mutated tumors. Four tagSNPs illustrated
the association with MSIþ colon tumors, RUNX1 rs2242878,
eIF4EBP2 rs70798987, RUNX3 rs906296 and MAPK1 rs9610375.
CIMPþ/KRAS2-mutated tumors were associated with RUNX2
rs10948238, eIF4E rs11727086 and RUNX1 rs1474479. All combinations of two high-risk genotypes were associated with a .3-fold
increased risk of CIMPþ/KRAS2-mutated tumors, although the combination of eIF4E and RUNX1 did not reach statistical significance.
The CIMPþ/MSIþ tumors were associated with RUNX2 rs1321075,
RUNX1 rs2242878, MAPK1 rs8136867 and RUNX3 rs906296. The
risk of each independent tagSNP was associated with 2-fold increased likelihood of this combination of tumor types, having most
combinations of three high-risk genotypes resulted in a statistically
significant 10-fold increased risk.
Associations with rectal tumors were generally less precise than
those observed for colon cancer (Table IV). As with colon cancer,
CIMPþ, KRAS2-mutated, CIMPþ/KRAS2-mutated and TP53mutated tumors had unique associations with the genes under investigation. However, unlike colon cancer, the independent risk associated
with the tagSNPs was generally higher, but the combined risk was
usually additive or less. Risk estimates were highest for CIMPþ,
KRAS2-mutated and CIMPþ/KRAS2-mutated tumors. The strongest
associations were observed for CIMPþ/KRAS2-mutated tumors, with
combinations of RUNX1 rs1474479, eIF4EBP3 rs250425, RUNX2
Additional support for a TGF-b signaling pathway
Table IV. Associations between Runx, MAPK1, eIF4E and rectal tumor mutations
HRG
RUNX1
EIF4EBP3
RUNX2
CIMPþ
rs1981392 (T . C)
rs250425 (C . T)
rs2819863 (G . C)
Controls
TT
1
1
1
2
2
2
3
1
1
1
2
2
2
3
1
1
1
1
2
2
2
2
2
2
3
3
3
3
4
GC/CC
N
N
344
47
156
18
60
7
2
29
6
17
2
10
2
1
EIF4E
RUNX1
RUNX2
KRAS2
rs11727086 (A . G)
rs1474479 (G . A)
rs7750470 (T . C)
Controls
AG/GG
1
1
1
2
2
2
3
TT
AA
CC
Cases
a
Cases
N
N
431
117
32
62
17
4
4
92
34
12
17
7
2
1
EIF4E
RUNX3
RUNX1
TP53
rs12498533 (A . C)
rs6672420 (A . T)
rs8134380 (A . T)
Controls
Cases
AA
TT
TT
N
Nd
280
218
170
57
42
37
11
98
83
67
26
19
20
5
RUNX1
EIF4EBP3
RUNX2
RUNX3
CIMPþ and KRAS2
rs1474479 (G . A)
rs250425 (C . T)
rs2819863 (G . C)
rs906296 (C . G)
Controls
Cases
AA
TT
GC/CC
CC
N
Ne
117
47
156
583
5
21
67
7
28
96
3
3
13
3
1
6
4
9
18
0
2
5
2
3
8
0
0
1
2
0
(95% CI)
1.77
2.20
2.07
2.87
3.70
5.25
9.75
OR
(1.04, 3.01)
(0.90, 5.40)
(1.14, 3.76)
(0.62, 13.32)
(1.64, 8.35)
(1.03, 26.72)
(0.79, 121.07)
(95% CI)
c
OR
b
1.46
1.79
2.18
2.08
2.85
3.31
2.02
OR
1.36
1.40
1.52
1.85
2.10
2.14
2.07
(1.05, 2.03)
(1.17, 2.73)
(1.10, 4.33)
(1.14, 3.81)
(1.13, 7.18)
(0.59, 18.53)
(0.22, 18.80)
(95% CI)
(1.02, 1.81)
(1.03, 1.89)
(1.10, 2.10)
(1.11, 3.08)
(1.17, 3.76)
(1.20, 3.80)
(0.69, 6.21)
OR
(95% CI)
2.95
4.56
4.04
3.90
undefined
8.96
12.45
24.65
18.82
11.95
undefined
undefined
13.02
undefined
undefined
(1.12, 7.77)
(1.47, 14.13)
(1.64, 9.93)
(1.14, 13.34)
(1.71, 46.96)
(2.34, 66.29)
(4.21, 144.43)
(2.94, 120.70)
(2.47, 57.74)
(0.89, 189.91)
HRG, high-risk genotypes. Rows are not mutually exclusive and include all individuals with the starred high-risk genotype versus those without any of the starred
high-risk genotypes.
a
959 controls total.
b
59 CIMPþ cases total.
c
173 KRAS2 cases total.
d
277 TP53 cases total.
e
21 CIMPþ and KRAS2 cases total.
rs2819863 and RUNX3 rs906296 associated with over a 10-fold statistically significant increased risk.
To establish how these candidate genes and tagSNPs associated
with colon and rectal cancer, we assessed their interaction with other
genes in the TGF-b-signaling pathway, including SNPS for TGFb1,
TGFbR1, BMP2, BMP4, BMPR1A, BMPR1B, Smad3, Samd4,
Smad7 and NFjB1 (Table V). For colon cancer, the following statistically significant interactions were identified: RUNX1 with
323
M.L.Slattery et al.
Table V. Interaction between Runx, MAPK1, eIF4E and other genes in candidate pathway
Gene
SNP
HRG
Pathway gene
SNP
HRG
Combined risk
OR (95% CI)
Interaction P value
Colon cancer
RUNX1
rs2300395
CC
BMPR1B
BMPR1A
TGFBR1
Smad7
TGFBR1
Smad3
Smad7
NFjB1
TGFB1
BMP4
NFkB1
TGFBR1
rs17616243
rs7088641
rs10733710
rs4464148
rs1571590
rs16950687
rs12953717
rs230510
rs4803455
rs17563
rs13117745
rs6478974
CT/TT
TT
AA
TC/CC
GG
GG
TT
TT
AA
TT
CC/CT
TT
1.33 (1.08, 1.64)
1.25 (1.03, 1.52)
1.77 (1.09, 2.89)
1.38 (1.12, 1.71)
3.37 (1.48, 7.71)
1.82 (1.06, 3.13)
2.18 (1.47, 3.24)
1.50 (1.12, 2.01)
1.82 (1.38, 2.42)
1.57 (1.13, 2.19)
3.58 (1.56, 8.19)
1.39 (1.11, 1.74)
0.024
0.042
0.0359
0.019
0.0272
0.013
0.0056
0.0239
0.0476
0.023
0.0061
0.0168
RUNX2
rs7279123
rs2248720
rs10948238
CT/TT
AA
TT
RUNX3
rs6672420
AA
MAPK1
rs8136867
rs2298432
GG
CC
RUNX1
rs1981392
rs1474479
rs11702779
rs8134380
CC
AA
GG/GA
TT
BMPR1A
TGFB1
Smad3
rs2819863
rs7750470
GC/CC
CC
rs6672420
rs2135756
TT
GG
Smad4
Smad3
NFKB1
BMPR1B
NFKB1
rs7088641
rs1800469
rs12708492
rs16950687
rs17293443
rs10502913
rs7163381
rs230510
rs13134042
rs4648110
CC
AA
CT/TT
GG
CC
AA
AA
AA
GA/AA
AA
2.94 (1.11, 7.73)
2.76 (1.11, 6.82)
2.75 (1.49, 5.08)
5.05 (1.67, 15.26)
15.17 (1.95, 117.96)
3.30 (1.02, 10.60)
12.01 (1.51, 95.46)
3.99 (1.67, 9.53)
1.75 (1.26, 2.44)
2.97 (1.04, 8.50)
0.0171
0.0373
0.0108
0.0084
0.0086
0.0496
0.0119
0.0216
0.046
0.0375
RUNX2
RUNX3
Rectal cancer
HRG, high-risk genotype group.
BMPR1B, BMPR1A, TGFbR1 and Smad7; RUNX2 with TGFbR1,
Smad3 and Smad7; RUNX3 with NFjB1 and TGFb1 and MAPK1
with BMP4, NFjb1 and TGFbR1. Risk estimates for SNP combinations varied in magnitude of association. For some combinations
such as RUNX1 and BMPR1B, the combined risk was 1.33 (95% CI
1.08, 1.64; P interaction 0.024), whereas for others such as MAPK1
and NF-jB1, the risk was considerably greater (OR 3.58 95% CI
1.56, 8.19; P interaction 0.005). For rectal cancer, there were also
numerous interactions between candidate genes and other genes in
the TGF-b-signaling pathway: RUNX1 interacted with BMPR1A,
TFGb1 and SMAD3; RUNX2 interacted with Smad4, Smad3 and
NFjB1 and RUNX3 interacted with BMPR1B and NFjB1. Risk
estimates were generally much stronger for rectal cancer than for
colon cancer, with most interactions showing over a 2-fold increase
in risk. Two interactions, RUNX1 and Smad3 and RUNX2 and
Smad3, had over a 10-fold increase in risk (OR 15.17 95%
CI 1.95, 117.96, P interaction 0.003 and OR 12.01 95% CI 1.51,
95.46, P interaction 0.0098, respectively).
Discussion
Our data suggest the importance of RUNX, MAPK1 and eIF4E in the
etiology of both colon and rectal cancer, although associations generally were stronger for colon than for rectal cancer. Multiple genetic
variants appear to have an impact on risk of colon cancer, where associations are greater than that would be expected on an additive scale.
Furthermore, our data support the involvement of these genes in the
TGF-b-signaling pathway given the findings of interaction between
genetic variants in the genes under investigation with other genes in
that pathway. Finally, our data emphasize the importance of this signaling pathway in the development of CRC with a CIMPþ phenotype.
The TGF-b-signaling pathway plays an important role in numerous
conditions including CRC (29). Studies have shown that loss of TGFb growth control is a critical event in tumorigenesis (5). The TGF-b
family is involved in cell proliferation, extracellular matrix synthesis,
angiogenesis, apoptosis and cell differentiation. The TGF-b family of
324
cytokines contains several related growth factors including TGF-b
and its receptors, BMPs and growth differentiation factors. Smads
are important to the pathway because they mediate TGF-b signaling.
Smad 1, 5 and 6 are more responsive to BMP, whereas Smad 2 and 3
are more responsive to TGF-b. Smad 7 plays an inhibitory role in
TGF-b signaling (30). MAPKs, including extracellular signal-regulated
kinases, can induce or modulate the outcome of TGF-b signaling (13).
RUNX genes have been shown to be involved in TGF-b and BMP
signaling. RUNX1 and RUNX3 are involved in carcinogenesis;
RUNX2 is a common target of TGF-b1 and BMP-2 and is induced
indirectly by Smad (8). eIf4E and its binding proteins appear to be
important in converging the TGF-b and AKT signaling pathways.
Of the RUNX genes analyzed, the role of RUNX3 is clearest
biologically. The adjusted P value for rs667240 for colon cancer
was 0.023, further indicating its potential importance. It is a key
element in gastrointestinal tract development and is strongly expressed in that tissue (1). In addition to its involvement in the
TGFb-signaling pathway, RUNX3 has been shown to attenuate Wnt
signaling in intestinal tumorigenesis. It is downregulated in serrated
adenomas and hyperplastic polyps. RUNX3 hypermethylation has
been identified as a key component in CIMPþ CRC (31,32). This is
the first report of an association between genetic variation in this gene
and colon and rectal cancer, particularly CIMPþ tumors to our knowledge. However, we also observed associations between RUNX genes
and TP53, which may be indicative of involvement in other pathways
such as Wnt signaling. This could also explain some of the differences
observed between tagSNPs associated with rectal and colon cancer
overall. At one end of the spectrum are the CIMPþ/MSIþ phenotypes, which are almost unique to colon cancer, whereas rectal tumors
have a higher proportion of TP53 mutations. Differences in associations with different tagSNPs could in part be the result of tumor
molecular phenotype differences for colon and rectal cancer.
MAPKs are major signaling transduction molecules involved in the
regulation of cell proliferation, differentiation and apoptosis (33).
MAPK1 is an extracellular signal-regulated kinase, which, when activated through Raf signaling, modulates gene expression by activating
Additional support for a TGF-b signaling pathway
other transcription factors. In human colon cancer, this pathway includes activated KRAS2 (34). It has been proposed that Ras signaling
can inhibit TGF-b signaling via the mitogen-activated protein pathway
(13). We observed an independent association between MAPK1 and
rectal cancer but not colon cancer. However, for colon cancer, genetic
variation in MAPK1 was associated with a greater likelihood of having
a CIMPþ and/or MSIþ phenotype. We are not aware of other reports
of the association between genetic variation in MAPK1 and risk of
colon or rectal cancer, although there is strong biologic support for an
association.
Expression of the translation initiation factor eIF4E has been shown
to be important in colon tumorigenesis (15). eIF4E overexpression
can cause neoplastic transformation of cells; overexpression of the
inhibitory eIF4E bonding proteins can suppress the oncogenetic properties of cell lines; overexpression of eIF4E has been demonstrated for
many solid tumors including colon cancer (15). We observed a statistically significant association between eIF4E and colon cancer overall
that remained significant at the 0.02 level after adjusting for multiple
comparison. Additionally, we observed that eIF4E bonding proteins
were statistically significantly associated with specific tumor phenotypes, primarily those involving CIMPþ tumors. Again, we are not
aware of reports of genetic variation in eIF4E and its binding proteins
and colon or rectal cancer. Given the biological support for such an
association, we encourage others to evaluate these associations.
The study has many strengths as well as some limitations. Our ability
to examine colon and rectal cancers separately in a well-characterized
dataset that includes tumor characteristics as well as lifestyle factors
and genetic factors is a major strength. Although our sample size is
large, it is limited in power to perform a test/retest analysis. Therefore,
we provide adjusted P values for each gene in order to account for the
number of tests performed. However, there are limitations with presentation of adjusted P values, in that we report risk estimates rather
than P values to indicate associations with our candidate genes. Genes
studied were selected based on their role in a biological pathway.
Although we have specified these as candidate genes, there is little
information on functional SNPs within these genes, hence the use of
tagSNPs. We identified tagSNPs, which we believed were the best
indicators of risk based on stepwise regression models. We evaluated
those tagSNPs together to have a better idea of how genes in the
pathway worked together. Our risk estimates for combined genotypes
in many instances were imprecise; however, several risk estimates had
a lower confidence bound over three. We interpret this as indicating
the importance of these genes in the profile of risk associated with this
pathway and regard the consistency of the patterns of association
similarly. Our results also stress the need for follow-up studies to
validate these findings, to determine which SNPs may be functionally
involved and to test functionality.
It might be expected that once the TGF-b pathway is impaired,
further less functional pathway members would not matter and that
multiple suboptimal proteins would not result in even additivity in
their impact, let alone something greater than additivity. The fact that
we report here greater than additivity is important because it suggests
that there may be limits to robustness. Robustness was initially defined by Waddington in the context of development; he said: ‘ .
developmental reactions, as they occur in organisms submitted to
natural selection are, in general, canalized. That is to say, they are
adjusted so as to bring about one definite end result, regardless of
minor variations in conditions during the course of the reaction’.
Furthermore, he argued that the constancy of the wild-type is evidence
of the buffering of the genotype against minor variations in genetics
and environment (35). More generally, robustness is a property of
systems that is characterized by relative insensitivity to the precise
values of the component parameters (36).
It is not well established how much potentially deleterious variation
can accumulate in a pathway before robustness begins to weaken; data
are especially lacking for cancer. The fact that we find that a larger
number of minor alleles in the same pathway are associated with
a greater than additive risk suggests that, at some point, the integrity
of the pathway becomes increasingly less robust, as a consequence of
which cancer risk begins to rise. It is probably worth noting that the
increasing impairment of a developmentally important pathway may
indicate loss of morphostatic control over tissue architecture rather
than a change involving epithelial mutation (37–39).
In summary, we interpret these findings as an indicator of the importance of these genes in the etiology of colon and rectal cancer. We
infer from this only the biologic significance of the genes and the
pathway, not the specific alleles. The somewhat stronger pattern of
association with CIMPþ tumors suggests that they are particularly
important in that molecular subset.
Supplementary material
Supplementary Table 1 can be found at http://carcin.oxfordjournals.
org/
Funding
National Cancer Institute (CA48998, CA61757). This research also
was supported by the Utah Cancer Registry, which is funded by
Contract #N01-PC-67000 from the National Cancer Institute, with
additional support from the State of Utah Department of Health, the
Northern California Cancer Registry and the Sacramento Tumor
Registry.
Acknowledgements
The contents of this manuscript are solely the responsibility of the authors and
do not necessarily represent the official view of the National Cancer Institute.
We would like to acknowledge the contributions of Sandra Edwards, Roger
Edwards, Leslie Palmer, Donna Schaffer, Dr Kristin Anderson and Judy Morse
for data management and collection.
Conflict of Interest Statement: None declared.
References
1. Lund,A.H. et al. (2002) RUNX: a trilogy of cancer genes. Cancer Cell, 1,
213–215.
2. Anglin,I. et al. (2004) Runx protein signaling in human cancers. Cancer
Treat. Res., 119, 189–215.
3. Bangsow,C. et al. (2001) The RUNX3 gene—sequence, structure and regulated expression. Gene, 279, 221–232.
4. Blyth,K. et al. (2005) The RUNX genes: gain or loss of function in cancer.
Nat. Rev. Cancer, 5, 376–387.
5. Zhu,B. et al. (2007) Transforming Growth Factor-B and Cancer. John Wiley & Sons, Ltd, San Francisco, CA.
6. Coffman,J.A. (2003) Runx transcription factors and the developmental
balance between cell proliferation and differentiation. Cell Biol. Int., 27,
315–324.
7. Ito,Y. et al. (2003) RUNX transcription factors as key targets of TGF-beta
superfamily signaling. Curr. Opin. Genet. Dev., 13, 43–47.
8. Lee,K.S. et al. (2002) Both the Smad and p38 MAPK pathways play a crucial
role in Runx2 expression following induction by transforming growth factorbeta and bone morphogenetic protein. Oncogene, 21, 7156–7163.
9. Cameron,E.R. et al. (2003) The Runx genes as dominant oncogenes. Blood
Cells Mol. Dis., 30, 194–200.
10. Ito,Y. et al. (2003) RUNX and cancer. Ann. Acad. Med. Singapore, 32, S6–S7.
11. Ge,C. et al. (2009) Identification and functional characterization of ERK/
MAPK phosphorylation sites in the Runx2 transcription factor. J. Biol.
Chem., 284, 32533–32543.
12. Guo,H. et al. (2007) Smad4 and ERK2 stimulated by transforming
growth factor beta1 in rhabdomyosarcoma. Chin. Med. J. (Engl.), 120,
515–521.
13. Javelaud,D. et al. (2006) Smad signal transduction. Interplays between
the smad and map kinase signaling pathways. Protein Cell Regul., 5,
317–334.
14. Wendel,H.G. et al. (2004) Survival signalling by Akt and eIF4E in oncogenesis and cancer therapy. Nature, 428, 332–337.
15. Berkel,H.J. et al. (2001) Expression of the translation initiation factor
eIF4E in the polyp-cancer sequence in the colon. Cancer Epidemiol. Biomarkers Prev., 10, 663–666.
325
M.L.Slattery et al.
16. Rajasekhar,V.K. et al. (2004) Postgenomic global analysis of translational
control induced by oncogenic signaling. Oncogene, 23, 3248–3264.
17. Durst,K.L. et al. (2004) Role of RUNX family members in transcriptional
repression and gene silencing. Oncogene, 23, 4220–4224.
18. Slattery,M.L. et al. (1997) Dietary fats and colon cancer: assessment of risk
associated with specific fatty acids. Int. J. Cancer, 73, 670–677.
19. Slattery,M.L. et al. (2003) Energy balance and rectal cancer: an evaluation
of energy intake, energy expenditure, and body mass index. Nutr. Cancer,
46, 166–171.
20. Slattery,M.L. et al. (1997) Energy balance and colon cancer—beyond physical activity. Cancer Res., 57, 75–80.
21. Slattery,M.L. et al. (2003) Physical activity and colorectal cancer. Am. J.
Epidemiol., 158, 214–224.
22. Edwards,S. et al. (1994) Objective system for interviewer performance
evaluation for use in epidemiologic studies. Am. J. Epidemiol., 140,
1020–1028.
23. Samowitz,W.S. et al. (2002) Prognostic significance of p53 mutations in
colon cancer at the population level. Int. J. Cancer, 99, 597–602.
24. Slattery,M.L. et al. (2000) Associations between cigarette smoking, lifestyle factors, and microsatellite instability in colon tumors. J. Natl Cancer
Inst., 92, 1831–1836.
25. Samowitz,W.S. et al. (2000) Relationship of Ki-ras mutations in colon
cancers to tumor location, stage, and survival: a population-based study.
Cancer Epidemiol. Biomarkers Prev., 9, 1193–1197.
26. Slattery,M.L. et al. (2007) Diet and lifestyle factor associations with CpG
island methylator phenotype and BRAF mutations in colon cancer. Int. J.
Cancer, 120, 656–663.
27. Slattery,M.L. et al. (2009) A comparison of colon and rectal somatic DNA
alterations. Dis. Colon Rectum, 52, 1304–1311.
28. Conneely,K.N. et al. (2007) So many correlated tests, so little time! rapid
adjustment of P values for multiple correlated tests. Am. J. Hum. Genet., 81,
1158–1168.
326
29. Gordon,K.J. et al. (2008) Role of transforming growth factor-beta superfamily signaling pathways in human disease. Biochim. Biophys. Acta, 1782,
197–228.
30. Korchynskyi,O. et al. (1999) Expression of Smad proteins in human
colorectal cancer. Int. J. Cancer, 82, 197–202.
31. Subramaniam,M.M. et al. (2009) RUNX3 inactivation in colorectal polyps
arising through different pathways of colonic carcinogenesis. Am. J.
Gastroenterol., 104, 426–436.
32. Subramaniam,M.M. et al. (2009) Molecular pathology of RUNX3 in human carcinogenesis. Biochim. Biophys. Acta, 1796, 315–331.
33. Wu,G.S. (2007) Role of mitogen-activated protein kinase phosphatases
(MKPs) in cancer. Cancer Metastasis Rev., 26, 579–585.
34. Vial,E. et al. (2003) Elevated ERK-MAP kinase activity protects the FOS
family member FRA-1 against proteasomal degradation in colon carcinoma
cells. J.Cell Sci., 116, 4957–4963.
35. Waddington,C. (1942) Canalization of development and the inheritance of
acquired characteristics. Nature, 150, 563–565.
36. Barkai,N. et al. (1997) Robustness in simple biochemical networks. Nature,
387, 913–917.
37. Potter,J.D. (2007) Morphogens, morphostats, microarchitecture and malignancy. Nat. Rev. Cancer, 7, 464–474.
38. Baker,S.G. et al. (2009) Plausibility of stromal initiation of epithelial cancers without a mutation in the epithelium: a computer simulation of morphostats. BMC Cancer, 9, 89.
39. Baker,S.G. et al. (2010) Research on early-stage carcinogenesis:
are we approaching paradigm instability? J. Clin. Oncol., 28,
3215–3218.
Received September 7, 2010; revised November 8, 2010;
accepted November 10, 2010