Genetic variation at insulin-like growth factor 1 (IGF1) has been linked to prostate cancer risk. However, the specific predisposing variants have not been identified. In this study, we fine-mapped the IGF1 locus for prostate cancer risk in African Americans. We conducted targeted Roche GS-Junior 454 resequencing of a 156-kb region of IGF1 in 80 African American aggressive prostate cancer cases. Three hundred and thirty-four IGF1 SNPs were examined for their association with prostate cancer risk in 1,000 African American prostate cancer cases and 991 controls. The top associated SNP in African Americans, rs148371593, was examined in an additional 3,465 prostate cancer cases and 3,425 controls of non-African American ancestry—European Americans, Japanese Americans, Latinos, and Native Hawaiians. The overall association of 334 IGF1 SNPs and prostate cancer risk was assessed using logistic kernel-machine methods. The association between each SNP and prostate cancer risk was evaluated through unconditional logistic regression. A false discovery rate threshold of q < 0.1 was used to determine statistical significance of associations. We identified 8 novel IGF1 SNPs. The cumulative effect of the 334 IGF1 SNPs was not associated with prostate cancer risk (P = 0.13) in African Americans. Twenty SNPs were nominally associated with prostate cancer at P < 0.05. The top associated SNP among African Americans, rs148371593 [minor allele frequency (MAF) = 0.03; P = 0.0014; q > 0.1], did not reach our criterion of statistical significance. This polymorphism was rare in non-African Americans (MAF < 0.003) and was not associated with prostate cancer risk (P = 0.98). Our findings do not support the role of IGF1 variants and prostate cancer risk among African Americans. Cancer Epidemiol Biomarkers Prev; 23(9); 1928–32. ©2014 AACR.

Prostate cancer is the most common cancer in U.S. men. African Americans have the highest incidence rate of prostate cancer and at least twice the mortality rate of disease in comparison with other racial/ethnic groups (1). Insulin-like growth factor 1 (IGF1) is a potentially interesting candidate gene for prostate cancer as it stimulates cellular proliferation and inhibits apoptosis. Men with higher levels of circulating IGFI have been reported to have an increased risk of prostate cancer compared with men with lower levels of IGFI (2–5). Past research and previous studies have linked genetic variation at IGF1 to prostate cancer risk (2, 3, 6); however, the specific predisposing IGF1 variants have not been identified. Detailed fine-mapping of the IGF1 locus may refine the genetic signal and aid in prioritizing risk variants for further follow-up and functional studies. Moreover, studying African Americans is an efficient means of localizing predisposing alleles given their high rates of prostate cancer and lower levels of linkage disequilibrium. These features provide for greater resolution in identifying IGF1 risk alleles and evaluating their effects among a population with the greatest burden of disease. Here, we conducted a fine-mapping study of the IGF1 locus and prostate cancer risk among African Americans.

Study subjects

The Multiethnic Cohort (MEC) Study is a large population-based cohort study of more than 215,000 men and women from Hawaii and Los Angeles. The cohort is composed predominantly of individuals from five racial/ethnic groups: African Americans, Native Hawaiians, Japanese, Latinos, and Whites. Further methodological details of this study are provided elsewhere (7). Briefly, incident prostate cancer cases were identified by cohort linkage to population-based Surveillance, Epidemiology and End Results cancer registries covering Hawaii and California. Information on stage of disease and Gleason grade at the time of diagnosis was also collected from the cancer registries. Aggressive prostate cancer was defined as either regional, metastatic disease, or localized disease with Gleason grade >8. Controls had no diagnosis of prostate cancer and were randomly selected from the control pool of participants that provided blood specimens for genetic analysis. Controls were frequency matched to cases by age (±5 years) and ethnicity. For this study, our African American and non-African American case–control studies of prostate cancer nested in the MEC included 1,098 cases and 1,081 controls and 3,480 cases and 3,447 controls, respectively. This study was approved by the institutional review boards at the University of Hawaii, the University of Southern California, and the California Prevention Institute of California.

SNP discovery and selection

We used RainDance Technologies Custom Primer Library Design, and utilized Roche GS-Junior 454 next-generation sequencing technology to target and resequence 156 kb of IGF1 (including 50 kb downstream and 25 kb upstream, chromosome 12: 102,741,896–102,898,083, human genome assembly 18) in pooled samples of 80 African American prostate cancer cases with aggressive disease (8 pools of 10 samples each). Variant analysis was performed with Roche Amplicon Variant Analysis (AVA) software. For the eight pools, a total of 395 SNPs were identified as high-quality variants (maximum variant allele frequency (VAF) > 5% with maximum minor allele frequency (MAF) > 10% in regions of possible off-target reads). To increase our coverage of SNPs in this region, we selected 316 additional SNPs that had an MAF of >1% in Yorubans (YRI) from the 1000 Genomes Project, resulting in a total of 711 SNPs. These 711 SNPs were scored for Illumina GoldenGate assay design; 259 SNPs did not meet assay criteria of a sufficient design score and/or located within >60 base pairs of a neighboring SNP, resulting in 452 SNPs. To reach 384 SNPs for the Illumina GoldenGate genotyping assay, an additional 68 SNPs were excluded based on lower design scores and proximity to neighboring SNP.

Case–control genotyping

All assays were undertaken by laboratory personnel blinded to prostate case–control status. For African Americans, the Illumina GoldenGate assay was used to genotype 384 IGF1 variants in 1,098 African American prostate cancer cases and 1,081 controls. We excluded subjects with a call rate of <95% (n = 43) and those missing genetic ancestry estimates (n = 145), resulting in a final analysis set of 1,000 African American cases and 991 African American controls. Of the 384 variants selected for Illumina genotyping, 38 SNPs failed genotyping (call rate < 95%) or were not in Hardy–Weinberg equilibrium (P < 0.0001). Of the 38 SNPs that failed Illumina genotyping, 32 SNPs were prioritized for genotyping by the Taqman assay on the OpenArray platform. Between the Illumina GoldenGate and Taqman OpenArray genotyping, a total of 384 variants were genotyped in African American cases and controls, of which 6 SNP assays failed, 1 SNP was not in Hardy–Weinberg equilibrium (P < 0.0001), 6 SNPs were excluded due to a low call rate (<95%), and 41 were monomorphic. For the 13 SNPs that did not pass the quality control (QC), data were abstracted for 4 SNPs from a previous genome-wide association study (GWAS) of prostate cancer among these subjects (8). This resulted in a final analysis set of 334 SNPs with an average genotyping success rate of 99.83% and average genotype concordance rate for QC duplicates (∼2% of all samples) of 98.5%.

For replication in non-African American samples, the Taqman assay was used to genotype the top associated SNP (rs148371593) in 3,480 prostate cases and 3,447 controls of European, Hawaiian, Japanese, and Latino ancestry. We excluded subjects (n = 37) that had no genotyping call, resulting in a total of 3,465 prostate cases and 3,425 controls. For non-African Americans, the average genotyping success rate was 99.5%, and the genotype concordance rate for QC duplicates (∼2% of all samples) was 99.3%.

Ancestry estimation

For African Americans, to correct for genetic ancestry, we included in the regression analysis the first two eigenvectors from principal components analysis of previously collected GWAS data (genomic inflation factor λ = 1.08; ref. 8). For non-African Americans, 93 ancestry informative markers that captured the major continental genetic diversity (9) were previously genotyped (10), principal components of ancestry were estimated by EIGENSTRAT (11), and the first four eigenvectors were included in the regression analysis.

Association testing

For African Americans, to test for the overall effect of 334 IGF1 SNPs together, we performed a logistic kernel-machine test, which is better powered than the omnibus test in the presence of correlation between SNPs (6). To examine single SNP association with prostate cancer, we conducted unconditional logistic regression, adjusting for age, family history, and genetic ancestry. The P values were estimated by a 1-degree-of-freedom Wald test for trend. To correct for multiple testing, we applied the false discovery rate (FDR) method and used a threshold of q < 0.1 to determine statistical significance (5). All analyses were run using the R software (7) with the packages lme4 for the regression models, GenABEL to compute the genomic inflation factor λ, and SKAT for the kernel-machine test (12). Manhattan and correlation plots were also done using the R software.

The average ages for African American prostate cancer cases and controls were 69.3 and 69.8 years, respectively (Supplementary Table S1). Among all cases, 89.6% presented with localized disease and the remaining had regional/distant disease. Study characteristics for the non-African American subjects are presented in Supplementary Table S1.

Of the 334 IGF1 SNPs identified by sequencing African Americans, 8 SNPs were novel and not found in the 1000 Genomes (13) database (one was found in cases only with an MAF of 0.001, and the remaining 7 SNPs had an MAF in controls ranging from 0.0005 to 0.021; Supplementary Tables S2 and S3), and 74 SNPs were rare with an MAF of <0.01 (Supplementary Tables S1 and S2). In a kernel-machine test, there was no overall cumulative effect of the 334 IGF1 SNPs on prostate cancer risk among African Americans (P = 0.13). In single SNP analysis, 20 SNPs were nominally associated with prostate cancer risk at P < 0.05 (Fig. 1 and Table 1). However, none were statistically significant (q < 0.1) after performing FDR correction for multiple testing (5). The top association was with SNP rs148371593, located in the 3′ region of IGF1, for which the minor A allele (MAF = 0.027) was found to have an inverse association with prostate cancer risk in African Americans (OR, 0.83; 95% CI, 0.75–0.93; P = 0.0014). SNP rs148371593 was weakly or not correlated (r2 < 0.45) with other SNPs in our set (Supplementary Fig. S1). In a non-African American sample set of 3,465 prostate cases and 3,425 controls, only 7 heterozygotes for rs148371593 were observed—2 Hawaiian (1 case and 1 control) and 5 of Latino ancestry (2 cases and 3 controls). The minor A allele was absent in European and Japanese ancestry populations, and it was extremely rare in Latino and Native Hawaiian populations (MAF of 0.003 and 0.001, respectively). Combining Latinos and Native Hawaiians, no significant association was observed (OR, 0.98; 95% CI, 0.53–1.82; P = 0.95; data not shown).

Figure 1.

Association between IGF1 and prostate cancer risk among African Americans. P values below 0.05 are plotted in red. The two pink dots mark rs7978742 and rs7965399, respectively, the two variants originally reported by the MEC to be associated with prostate cancer in ref. (2). The green dot marks variant rs7136446, which was found to be associated with prostate cancer risk and higher circulating levels of IGFI (6, 14), and the two aquamarine dots mark variants, rs1520220 and 10735380, which were found to be associated with higher IGFI circulating levels and higher prostate cancer risk (16, 17).

Figure 1.

Association between IGF1 and prostate cancer risk among African Americans. P values below 0.05 are plotted in red. The two pink dots mark rs7978742 and rs7965399, respectively, the two variants originally reported by the MEC to be associated with prostate cancer in ref. (2). The green dot marks variant rs7136446, which was found to be associated with prostate cancer risk and higher circulating levels of IGFI (6, 14), and the two aquamarine dots mark variants, rs1520220 and 10735380, which were found to be associated with higher IGFI circulating levels and higher prostate cancer risk (16, 17).

Close modal
Table 1.

Associations for 20 IGF1 SNPs and prostate cancer risk among African Americans at P < 0.05

SNPPositionaMAbMAFcOR (95% CI)P
rs148371593 102773400 A/G 0.027 0.83 (0.75–0.93) 0.001 
chr12_102748461 102748461 A/G 0.001 1.43 (1.10–1.90) 0.008 
rs78360701 102762488 A/G 0.001 1.42 (1.10–1.90) 0.012 
rs5742687 102812676 G/A 0.001 1.42 (1.10–1.90) 0.012 
rs140149239 102888147 A/C 0.004 1.27 (1.10–1.50) 0.013 
rs140231439 102753563 G/A 0.009 1.18 (1.00–1.40) 0.016 
rs138668850 102871977 A/C 0.025 0.87 (0.78–0.98) 0.018 
rs7965399 102891686 G/A 0.267 1.04 (1.00–1.10) 0.030 
rs76845058 102838572 G/C 0.009 0.81 (0.67–0.98) 0.033 
rs146567847 102864270 A/G 0.009 1.17 (1.00–1.30) 0.034 
rs5742681 102813915 G/C 0.004 0.69 (0.49–0.97) 0.035 
rs73187920 102752878 A/C 0.046 0.92 (0.84–0.99) 0.037 
rs60741636 102792368 C/G 0.008 0.81 (0.65–0.99) 0.039 
rs77458341 102894716 G/A 0.272 1.04 (1.00–1.10) 0.043 
rs17878342 102835435 G/A 0.024 0.89 (0.79–1.00) 0.043 
rs35638230 102742243 A/C 0.012 1.14 (1.00–1.30) 0.046 
rs11111256 102761471 A/G 0.046 0.92 (0.85–1.00) 0.046 
rs61125097 102895637 G/A 0.271 1.04 (1.00–1.10) 0.047 
rs56060310 102835566 G/A 0.009 0.82 (0.67–1.00) 0.047 
rs11111285 102895256 G/A 0.271 1.04 (1.00–1.10) 0.049 
SNPPositionaMAbMAFcOR (95% CI)P
rs148371593 102773400 A/G 0.027 0.83 (0.75–0.93) 0.001 
chr12_102748461 102748461 A/G 0.001 1.43 (1.10–1.90) 0.008 
rs78360701 102762488 A/G 0.001 1.42 (1.10–1.90) 0.012 
rs5742687 102812676 G/A 0.001 1.42 (1.10–1.90) 0.012 
rs140149239 102888147 A/C 0.004 1.27 (1.10–1.50) 0.013 
rs140231439 102753563 G/A 0.009 1.18 (1.00–1.40) 0.016 
rs138668850 102871977 A/C 0.025 0.87 (0.78–0.98) 0.018 
rs7965399 102891686 G/A 0.267 1.04 (1.00–1.10) 0.030 
rs76845058 102838572 G/C 0.009 0.81 (0.67–0.98) 0.033 
rs146567847 102864270 A/G 0.009 1.17 (1.00–1.30) 0.034 
rs5742681 102813915 G/C 0.004 0.69 (0.49–0.97) 0.035 
rs73187920 102752878 A/C 0.046 0.92 (0.84–0.99) 0.037 
rs60741636 102792368 C/G 0.008 0.81 (0.65–0.99) 0.039 
rs77458341 102894716 G/A 0.272 1.04 (1.00–1.10) 0.043 
rs17878342 102835435 G/A 0.024 0.89 (0.79–1.00) 0.043 
rs35638230 102742243 A/C 0.012 1.14 (1.00–1.30) 0.046 
rs11111256 102761471 A/G 0.046 0.92 (0.85–1.00) 0.046 
rs61125097 102895637 G/A 0.271 1.04 (1.00–1.10) 0.047 
rs56060310 102835566 G/A 0.009 0.82 (0.67–1.00) 0.047 
rs11111285 102895256 G/A 0.271 1.04 (1.00–1.10) 0.049 

aChromosomal location based on NCBI build 37/hg19.

bMajor/minor allele.

cMinor allele frequency among controls.

In this fine-mapping study of prostate cancer risk among African Americans, we examined 334 IGF1 variants—8 of which were novel—and observed no statistically significant association between IGF1 variants and prostate cancer risk after correcting for multiple testing. The top association, a less common SNP, rs148371593, did not replicate in non-African American samples, for which it was found to be rare among Latinos and Native Hawaiians and nonexistent among Japanese and European Americans.

Numerous studies have indicated a possible involvement of IGF1 in prostate development and associated higher circulating levels of IGFI with an increased risk of developing prostate cancer (2–5). In our previous haplotype-based study (2, 3, 6), we identified two upstream IGF1 variants associated with prostate cancer risk in the MEC Study (2). These two variants, rs7978742 and rs7965399, were originally reported to be associated with overall prostate cancer risk (P = 0.003 for both; ref. 2). In this study, with a larger sample size of African Americans, we observed nominal positive associations for these SNPs in African Americans (rs7978742, P = 0.03; rs7965399, P = 0.05), which was in line with our previously reported pattern of associations.

Subsequent studies confirmed an association at the IGF1 locus and prostate cancer risk in populations of European ancestry, in particular with variants rs7136446 and rs2033178 (13), and rs4764695 (14, 15). We observed a marginal nominal association with rs7136446 (P = 0.06) and could not confirm an association with the other two reported variants (P > 0.05) in African Americans.

Of the two variants (rs6220 and rs7136446) reported significantly associated with higher levels of circulating IGFI by Johansson and colleagues (6), rs6220 was not examined in this study, though one correlated SNP, rs5009837 (r2 = 0.91 among African Americans), was tested in our study and was not associated with prostate cancer risk (P = 0.47). The second variant, rs7136446, showed a nominal borderline association with prostate cancer risk in our study (P = 0.06). Two additional SNPs rs1520220 and rs10735380, previously associated with high circulating levels of IGFI (16, 17), were not associated with prostate cancer risk in our study (P > 0.05).

Several GWASs of prostate cancer have failed to observe an association with IGF1 and prostate cancer (18). The absence of an IGF1 association in GWAS of prostate cancer coupled with our findings here would suggest that common IGF1 variants do not play a role in prostate cancer susceptibility.

Sufficient statistical power is a major challenge of fine-mapping studies that aim to identify less common and rare risk alleles. While our study was able to observe nominal associations with variants that have an MAF ranging from <0.01 to 0.27, reduced study power is a limitation of our study. For less common variants with an MAF of 0.03, we had 80% to detect odds ratios of 1.60 among our African American sample, while for more common variants with an MAF of 0.20, we had 80% power to detect smaller odds ratios of 1.24.

Despite these limitations, the identification of novel and less common variants may be useful in further investigating the difference in prevalence and risk of prostate cancer between African Americans and other racial/ethnic groups. Large study populations of diverse ancestral populations are needed to confirm associations for less common variants.

No potential conflicts of interest were disclosed.

Conception and design: D.O. Stram, D. Taverna, I. Cheng

Development of methodology: E.E. Giorgi, S.D. Turner, M. Tirikainen, D. Duggan, I. Cheng

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. Lum-Jones, M. Tirikainen, D. Duggan, L. Le Marchand, I. Cheng

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.E. Giorgi, D.O. Stram, D. Taverna, S.D. Turner, F. Schumacher, A. Lum-Jones, D. Duggan, I. Cheng

Writing, review, and/or revision of the manuscript: E.E. Giorgi, D.O. Stram, D. Taverna, S.D. Turner, F. Schumacher, C.A. Haiman, A. Lum-Jones, M. Tirikainen, D. Duggan, B.E. Henderson, L. Le Marchand, I. Cheng

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Taverna, A. Lum-Jones, C. Caberto, I. Cheng

Study supervision: I. Cheng

This work was supported by the V Scholar Award (I. Cheng) from the Jim Valvano Foundation for Cancer Research. E.E. Giorgi is supported by the Center for Nonlinear Studies, LANL, through Laboratory Directed Research and Development (LDRD) funds, number 201110434DR.

1.
Howlader
N
,
Noone
AM
,
Krapcho
M
,
Garshell
J
,
Miller
D
,
Altekruse
SF
, et al
(eds). SEER Cancer Statistics Review, 1975–2011
,
National Cancer Institute
,
Bethesda, MD
, http://seer.cancer.gov/csr/1975_2011/,
based on November 2013 SEER data submission, posted to the SEER web site, April 2014
.
2.
Cheng
I
,
Stram
DO
,
Penney
KL
,
Pike
M
,
Le Marchand
L
,
Kolonel
LN
, et al
Common genetic variation in IGF1 and prostate cancer risk in the multiethnic cohort
.
J Natl Cancer Inst
2006
;
98
:
123
34
.
3.
Schumacher
FR
,
Cheng
I
,
Freedman
ML
,
Mucci
L
,
Allen
NE
,
Pollak
MN
, et al
A comprehensive analysis of common IGF1, IGFBP1 and IGFBP3 genetic variation with prospective IGF-I and IGFBP-3 blood levels and prostate cancer risk among Caucasians
.
Hum Mol Genet
2010
;
19
:
3089
101
.
4.
Price
AJ
,
Allen
NE
,
Appleby
PN
,
Crowe
FL
,
Travis
RC
,
Tipper
SJ
, et al
Insulin-like growth factor-I concentration and risk of prostate cancer: results from the European Prospective Investigation into Cancer and Nutrition
.
Cancer Epidemiol Biomarkers Prev
2012
;
21
:
1531
41
.
5.
Roddam
AW
,
Allen
NE
,
Appleby
P
,
Key
TJ
,
Ferrucci
L
,
Carter
HB
, et al
Insulin-like growth factors, their binding proteins, and prostate cancer risk: analysis of individual patient data from 12 prospective studies
.
Ann Intern Med
2008
;
149
:
461
71
,
W83–8
.
6.
Johansson
M
,
McKay
JD
,
Wiklund
F
,
Rinaldi
S
,
Verheus
M
,
van Gils
CH
, et al
Implications for prostate cancer of insulin-like growth factor-I (IGF-I) genetic variation and circulating IGF-I levels
.
J Clin Endocrinol Metab
2007
;
92
:
4820
6
.
7.
Kolonel
LN
,
Henderson
BE
,
Hankin
JH
,
Nomura
AM
,
Wilkens
LR
,
Pike
MC
, et al
A multiethnic cohort in Hawaii and Los Angeles: baseline characteristics
.
Am J Epidemiol
2000
;
151
:
346
57
.
8.
Haiman
CA
,
Chen
GK
,
Blot
WJ
,
Strom
SS
,
Berndt
SI
,
Kittles
RA
, et al
Genome-wide association study of prostate cancer in men of African ancestry identifies a susceptibility locus at 17q21
.
Nat Genet
2011
;
43
:
570
3
.
9.
Kosoy
R
,
Nassir
R
,
Tian
C
,
White
PA
,
Butler
LM
,
Silva
G
, et al
Ancestry informative marker sets for determining continental origin and admixture proportions in common populations in America
.
Hum Mutat
2009
;
30
:
69
78
.
10.
Cheng
I
,
Kocarnik
JM
,
Dumitrescu
L
,
Lindor
NM
,
Chang-Claude
J
,
Avery
CL
, et al
Pleiotropic effects of genetic risk variants for other cancers on colorectal cancer risk: PAGE, GECCO and CCFR consortia
.
Gut
2014
;
63
:
800
7
.
11.
Price
AL
,
Patterson
NJ
,
Plenge
RM
,
Weinblatt
ME
,
Shadick
NA
,
Reich
D
. 
Principal components analysis corrects for stratification in genome-wide association studies
.
Nat Genet
2006
;
38
:
904
9
.
12.
Wu
MC
,
Lee
S
,
Cai
T
,
Li
Y
,
Boehnke
M
,
Lin
X
. 
Rare-variant association testing for sequencing data with the sequence kernel association test
.
Am J Hum Genet
2011
;
89
:
82
93
.
13.
Abecasis
GR
,
Altshuler
D
,
Auton
A
,
Brooks
LD
,
Durbin
RM
,
Gibbs
RA
, et al
A map of human genome variation from population-scale sequencing
.
Nature
2010
;
467
:
1061
73
.
14.
Johansson
M
,
McKay
JD
,
Stattin
P
,
Canzian
F
,
Boillot
C
,
Wiklund
F
, et al
Comprehensive evaluation of genetic variation in the IGF1 gene and risk of prostate cancer
.
Int J Cancer
2007
;
120
:
539
42
.
15.
Schumacher
F
,
Cheng
I
,
Freedman
ML
,
Mucci
L
,
Allen
N
,
Pollak
MN
, et al
A comprehensive analysis of common IGF1, IGFBP1 and IGFBP3 genetic variation with prospective IGF-I and IGFBP-3 blood levels and prostate cancer risk among Caucasians
.
Hum Mol Genet
2010
;
19
:
3089
101
.
16.
Al-Zahrani
A
,
Sandhu
MS
,
Luben
RN
,
Thompson
D
,
Baynes
C
,
Pooley
KA
, et al
IGF1 and IGFBP3 tagging polymorphisms are associated with circulating levels of IGF1, IGFBP3 and risk of breast cancer
.
Hum Mol Genet
2006
;
15
:
1
10
.
17.
Gu
F
,
Schumacher
FR
,
Canzian
F
,
Allen
NE
,
Albanes
D
,
Berg
CD
, et al
Eighteen insulin-like growth factor pathway genes, circulating levels of IGF-I and its binding protein, and risk of prostate and breast cancer
.
Cancer Epidemiol Biomarkers Prev
2010
;
19
:
2877
87
.
18.
Hindorff
LA
,
Sethupathy
P
,
Junkins
HA
,
Ramos
EM
,
Mehta
JP
,
Collins
FS
, et al
Potential etiologic and functional implications of genome-wide association loci for human diseases and traits
.
Proc Natl Acad Sci U S A
2009
;
106
:
9362
7
.