Baseline patient characteristics
A total of 8,554 participants (57.1% men and 52.9% women) were included in our analysis. Table 1 shows the baseline characteristics of all participants. The mean age of the participants was 52.1 years. Smoking history was obtained using a questionnaire, and smoking status and pack years were used as covariates for association analyses. Smoking status was categorized into never smokers (60%), former smokers (15.4%), and current smokers (24.6%). Among the 8,554 baseline subjects, COPD was observed in 738 subjects (8.6%) (Additional file 5: Table S2). During the 12-year follow-up period, 10.8% (N = 844) of patients developed COPD.
Table 1. Baseline characteristics of male and female subjects included in the study
|
Males
(N = 4,026)
|
Females
(N = 4,528)
|
Total
(N = 8,554)
|
P-value
|
Age (years), mean
|
51.7 ± 8.7
|
52.5 ± 9.0
|
52.1 ± 8.8
|
< 0.001
|
Body mass index (kg/m2), mean
|
24.4±2.8
|
24.9±3.2
|
24.6±3.1
|
< 0.001
|
Height (cm)
|
167.0 ± 5.8
|
153.8 ± 5.5
|
160.0 ± 8.7
|
< 0.001
|
Smoking status, N (%)
|
|
|
|
< 0.001
|
Never smoker
|
805 (20.0)
|
4329 (95.6)
|
5134 (60.0 )
|
|
Former smoker
|
1259 (31.6)
|
55 (1.2)
|
1314 (15.4 )
|
|
Current smoker
|
1962 (48.7)
|
144 (3.2)
|
2106 (24.6 )
|
|
Smoking, pack-year
|
24.2 ± 17.4
|
6.8 ± 9.9
|
22.9 ± 17.6
|
< 0.001
|
FEV1 (% predicted), mean
|
106.4 ± 16.2
|
116.6 ± 17.6
|
111.8 ± 17.7
|
< 0.001
|
FVC (% predicted), mean
|
102.0 ± 13.9
|
107.1 ± 15.0
|
104.7 ± 14.7
|
< 0.001
|
FEV1/FVC (%)
|
78.0 ± 13.9
|
81.7 ± 6.6
|
79.9 ± 7.7
|
< 0.001
|
Data are presented as the mean ± standard deviation or n (%). FEV1, forced expiratory volume in 1 s; FVC, forced vital capacity
Effect of smoking on lung function
Fig. 2a shows the decline in FEV1 with age in males and females. The annual rate of decline was significantly more rapid for male patients (41.73 mL; 95% confidence interval [CI], 41.15–42.31) than for their female counterparts (33.36 mL; 95% CI, 33.12–33.59; P < 0.001). We also compared the annual mean decline in FEV1 among healthy male never smokers, former smokers, and current smokers (Fig. 2b); and the annual rate of decline in FEV1 was fastest for current smokers (46.3 mL; 95% CI, 45.2–46.8; P < 0.001 vs. healthy never smokers and former smokers).
GWIS of FEV1/FVC with KARE data
GWISs were conducted on 3,333,374 SNPs by applying the two DF test to KARE data.Fig. 3 shows the quantile-quantile (QQ) plot and the Manhattan (MH) plot of the two LRT results. The MH plot showed that 89 SNPs associated with FEV1/FVC at the genome-wide significance level were located in FAM13A on chromosome 4. Table 2 shows the ten most significant SNPs from the two LRT results for FEV1/FVC. These selected SNPs were located in the upstream region of FAM13 on chromosome 4 and had similar MAFs. The SNP most associated with FEV1/FVC was the rs2446304 SNP (P = 1.56 ´ 10−10). Fig. 4 shows a regional plot r2 around rs2446304 SNPs created using LocusZoom. FAM13,within the topologically associated domain (TAD), is located between 89.6–90.2 M.
Table 2. Top ten significant SNPs (SNP, SNP*time 2LRT)
CHR
|
SNP
|
BP
|
Minor
|
MAF
|
HWE
|
MISS
|
Info
|
SNP BETA (SE)
|
SNP P-value
|
Interaction Beta (SE)
|
Interaction P-value
|
LR
|
4
|
rs2446304
|
89825872
|
C/A
|
0.477
|
0.83
|
0.003
|
0.999
|
0.665 (0.104)
|
1.40E-10
|
-0.012 (0.024)
|
0.614241
|
1.56E-10
|
4
|
rs2446303
|
89826058
|
T/C
|
0.477
|
0.83
|
0.003
|
0.999
|
0.665 (0.104)
|
1.40E-10
|
-0.012 (0.024)
|
0.614241
|
1.56E-10
|
4
|
rs2704581
|
89815404
|
A/G
|
0.477
|
0.814
|
0.003
|
0.999
|
0.664 (0.104)
|
1.51E-10
|
-0.012 (0.024)
|
0.611959
|
1.71E-10
|
4
|
rs2464528
|
89817664
|
C/A
|
0.477
|
0.814
|
0.003
|
0.999
|
0.664 (0.104)
|
1.51E-10
|
-0.012 (0.024)
|
0.611959
|
1.71E-10
|
4
|
rs2704582
|
89818521
|
T/C
|
0.477
|
0.814
|
0.003
|
0.999
|
0.664 (0.104)
|
1.51E-10
|
-0.012 (0.024)
|
0.611959
|
1.71E-10
|
4
|
rs2704583
|
89818643
|
A/G
|
0.477
|
0.814
|
0.003
|
0.999
|
0.664 (0.104)
|
1.51E-10
|
-0.012 (0.024)
|
0.611959
|
1.71E-10
|
4
|
rs1246629
|
89819423
|
T/C
|
0.477
|
0.814
|
0.003
|
0.999
|
0.664 (0.104)
|
1.51E-10
|
-0.012 (0.024)
|
0.611959
|
1.71E-10
|
4
|
rs1246631
|
89820287
|
G/A
|
0.477
|
0.814
|
0.003
|
0.999
|
0.664 (0.104)
|
1.51E-10
|
-0.012 (0.024)
|
0.611959
|
1.71E-10
|
4
|
rs1246632
|
89822166
|
C/T
|
0.477
|
0.814
|
0.003
|
0.999
|
0.664 (0.104)
|
1.51E-10
|
-0.012 (0.024)
|
0.611959
|
1.71E-10
|
4
|
rs1246634
|
89823798
|
G/A
|
0.477
|
0.814
|
0.003
|
0.999
|
0.664 (0.104)
|
1.51E-10
|
-0.012 (0.024)
|
0.611959
|
1.71E-10
|
SNP, single nucleotide polymorphism; LRT, likelihood ratio test; CHR, chromosome; BP, physical position; MAF, minor allele frequencies; HWE, Hardy-Weinberg equilibrium; SE, standard error
Effects of time and gene interactions on FEV1/FVC
Fig. 5a shows the QQ plot for the interaction effect tests, in which the variance inflation factor (VIF) was 1.03, while Fig. 5b shows the MH plot for the same factor. To assess the significance of the identified interactions, we applied statistical significance thresholds that are routinely used for GWAS.
We specified a genome-wide significance threshold of 5 ´ 10-8, while applying a more liberal genome-wide suggestive threshold of 1 ´ 10-5 to include interactions with P-values that were slightly above the genome-wide significance threshold but may represent genuine interactions. Additionally, we investigated time interactions, and those that fulfilled a 0.05 P-value significance threshold (nominal significance) were included in the interaction analysis. Data-time interaction effects were analyzed from GWIS with KARE, and the most significant SNPs are summarized in Table 3.
The SNPs highly associated with lung function decline in SNP-by-time interaction were rs75679995 within DNAH11, and rs9991425 located near AADAT and MFAP3L, with P = 2.28 ´ 10−7 and 2.01 ´ 10−6, respectively.
Fig. 6 shows the linkage disequilibrium and the regional plot for the most significant SNP region. The SNPs were located within the DNAH11 region TAD (21.6–22.2 Mb of chromosome 7). DNA sequences within a TAD physically interact with each other more frequently than with sequences outside the TAD, and thus our most significant SNPs may affect DNAH11 expression (Fig. 7). Therefore, our results indicated that DNAH11 may be functionally related to decreased lung function.
Table 3. Top ten significant SNPs (interaction effects)
CHR
|
dbSNP.RS.ID
|
BP
|
minor
|
MAF
|
HWE
|
MISS
|
info
|
SNP BETA (SE)
|
SNP P-value
|
Interaction Beta (SE)
|
Interaction P-value
|
LR
|
7
|
rs75679995
|
21880268
|
G/A
|
0.224
|
0.576
|
0.021
|
0.98
|
0.151 (0.125)
|
0.228528
|
-0.149(0.029)
|
2.28E-07
|
1.18E-06
|
7
|
rs74676405
|
21878748
|
A/G
|
0.208
|
0.896
|
0.021
|
0.97
|
0.126 (0.129)
|
0.328204
|
-0.149(0.03)
|
4.37E-07
|
1.89E-06
|
7
|
rs10274154
|
21881548
|
G/A
|
0.215
|
0.949
|
0.005
|
0.992
|
0.083 (0.126)
|
0.511169
|
-0.14(0.029)
|
1.10E-06
|
3.44E-06
|
4
|
rs9991425
|
171175255
|
G/A
|
0.083
|
0.318
|
0.04
|
0.945
|
0.524 (0.191)
|
0.006037
|
-0.206(0.043)
|
2.01E-06
|
6.92E-06
|
4
|
rs10026067
|
171175090
|
A/T
|
0.083
|
0.318
|
0.04
|
0.945
|
0.526 (0.191)
|
0.005779
|
-0.206(0.043)
|
2.10E-06
|
7.06E-06
|
4
|
rs55764142
|
171176154
|
T/G
|
0.083
|
0.392
|
0.04
|
0.945
|
0.522 (0.19)
|
0.006134
|
-0.205(0.043)
|
2.13E-06
|
7.31E-06
|
4
|
rs55999620
|
171176227
|
T/A
|
0.083
|
0.392
|
0.04
|
0.945
|
0.522 (0.19)
|
0.006134
|
-0.205(0.043)
|
2.13E-06
|
7.31E-06
|
4
|
rs4692810
|
171177863
|
T/C
|
0.083
|
0.392
|
0.04
|
0.944
|
0.521 (0.191)
|
0.006284
|
-0.205(0.043)
|
2.21E-06
|
7.63E-06
|
4
|
rs10014884
|
171175473
|
C/A
|
0.083
|
0.392
|
0.039
|
0.945
|
0.525 (0.19)
|
0.005804
|
-0.204(0.043)
|
2.36E-06
|
7.82E-06
|
7
|
rs73279813
|
21889459
|
A/T
|
0.212
|
0.897
|
0.017
|
0.986
|
0.101 (0.128)
|
0.430423
|
-0.138(0.029)
|
2.44E-06
|
9.12E-06
|
SNP, single nucleotide polymorphism; LRT, likelihood ratio test; CHR, chromosome; BP, physical position; MAF, minor allele frequencies; HWE, Hardy-Weinberg equilibrium; SE, standard error
Several identified loci, including MFAP3L and AADAT, were plausible candidates for lung function decline characterized by changes in the FEV1/FVC ratio (Fig. 8). We assessed whether the identified SNPs were associated with gene expression by analyzing whether they were cis-acting expression quantitative trait loci (eQTL) in lung tissue. Two SNPs were cis-eQTLs for MFAP3L and AADAT in muscles and cultured fibroblasts, respectively. The exon-level expressions of MFAP3L and AADAT showed 6.286 and 1 transcript per million kilobases in lung tissues, respectively (Additional file 6: Fig. S1).
Changes in FEV1/FVC with time and rs75679995 (DNAH11) status were plotted using generalized additive models.The estimated FEV1/FVC ratio was generated using gene-by-time analysis and KARE data (Fig. 8). The results suggested that the cohort rs75679995 significantly affected the FEV1/FVC ratios for time-to-gene interactions, resulting in a rapid decline in lung function; the minor allele, G, negatively impacted the increment in time interaction.
Replication of the most significant SNPs in the GENIE cohort of KARE
Results for the most significant SNPs were identified in GWIS. For FEV1/FVC, GWIS was performed on a GENIE cohort, and the results are summarized in Additional file 7: Table S3. SNP and INT are the coefficients for the main SNP and interaction effects between SNP and time, respectively. Overall effects are indicated by the P-values for testing the null hypotheses, H0: βSNP = βSNP−time = 0, obtained using the F-test.