Hum Genet (1992) 88:439-446
9 Springer-Verlag1992
Polymorphisms in the apolipoprotein (apo) AI-CIII-AIV gene cluster: detection of genetic variation determining plasma apo AI, apo CIII and apo AIV concentrations H. PauI-Hayase 1, M. Rosseneu 2, D. Robinson 3, J. P. Van Bervliet 3, J. P. Deslypere 4, and S. E. Humphries 1 1Chafing Cross Sunley Research Centre, 1 Lurgan Avenue, Hammersmith, London W6 8LW, UK 2Departments of Clinical Chemistry and Paediatrics, A-Z St Jan, Brugge, Belgium 3BUPA Medical Research, Battle Bridge House, 300 Grays Inn Road, London WCIX 8DU, UK 4Department of Endocrinology, University Hospital, Ghent, Belgium Received August 29, 1990 / Revised July 30, 1991
Summary. We have examined the associations between levels of plasma apolipoprotein (apo) AI, apo CIII and apo AIV and genetic variation in the apo AI-CIII-AIV gene cluster in 162 boys and young men from Belgium aged from 7 to 23 years. Genotypes were determined for six restriction enzymes XmnI, PstI, SstI, PvuIIA-CIII, PvuIIB-AIV and XbaI, and for the G to A substitution at - 7 5 bp in the 5' region of the apo AI gene. The polymorphism most strongly associated with apo AI levels was the G to A substitution (P = 0.025, R z x 100 = 3.6%) confirming previous observations. The polymorphism most strongly associated with apo CIII levels was that of PvuIIA-CIII (P = 0.023, R 2 • 100 = 2.9%) in the apo CIII gene. This novel association must be interpreted with caution until it has been confirmed in an independent sample. The polymorphism associated with the largest effect on apo AIV levels was that detected with XbaI in the apo AIV gene, but this association was not statistically significant. Previously reported associations between the SstI polymorphism and triglyceride levels, and between the PstI polymorphism and apo AI levels, were weakly detected in the present sample. Our results show that variation associated with some of the polymorphisms in the apo AI-CIII-AIV cluster makes a small, but statistically significant, contribution to the determination of apo AI and apo CIII levels in this sample of young men and boys. These observations may, in part, explain reported associations between polymorphisms in this gene cluster, differences in plasma lipid and lipoprotein levels, a n d prevalence of coronary artery disease.
Introduction A large number of epidemiological studies have demonstrated that low serum levels of high-density lipoprotein Offprint requests to: S. E. Humphries
cholesterol (HDL-C) and apolipoprotein AI (apo AI) are risk factors for the development of coronary artery disease (CAD) (Miller et al. 1977; Norum et al. 1982; Yaari et al. 1981). Various environmental factors can significantly affect plasma HDL-C levels, including smoking, exogenous gonadal steroids, alcohol intake, stress, infection, the amount of exercise, and some drug therapy (Haffner et al. 1985; Berg et al. 1986; Hartung et al. 1986). On the other hand, the heritability of HDLC and apo AI has been estimated in various studies to be between 0.43 and 0.66 (Hamsten et al. 1986; Kuusi et al. 1987). Recently, evidence for a major gene determining individual differences in quantitative levels of apo AI was demonstrated in a population-based sample by applying biometrical techniques (Moll et al. 1989). Altered HDL-C levels have also been associated with rearrangement of the AI gene (Karathanasis et al. 1987), with specific amino acid substitutions in the apo AI protein (e.g., Eckardstein et al. 1989) and with a defect in the gene for the cholesterol ester transport protein (CETP) (Brown et al. 1989). However, these mutations are rare and the study of common genetic factors that determine HDL-C and apo AI levels is important. The gene coding for apo AI is on the long arm of chromosome 11, next to the genes for Apo CIII and Apo AIV (Karathanasis et al. 1983a). Apo AI is the major protein of mature H D L and is synthesized in the liver and intestine. Apo AI can activate lecithin cholesterol acyltransferase (LCAT), which converts unesterified cholesterol to its esterified form in plasma. Apo AI may also be the ligand for an H D L receptor involved in the removal of cholesterol from peripheral tissue (Oram et al. 1987; Schmitz et al. 1985). This would allow subsequent transport of excess cholesterol from tissues to the liver, a mechanism referred to as "reverse cholesterol transport". Apo CIII is synthesized mainly in the liver and to a lesser degree in the intestine and is a constituent of very-low-density lipoprotein (VLDL) particles and H D L particles (Herbert et al. 1982). In vitro, apo CIII
440 has b e e n s h o w n to i n h i b i t the activities of b o t h lipop r o t e i n lipase ( B r o w n et al. 1972; Q u a r f o r d t et al. 1982) a n d h e p a t i c lipase ( K i n n u n e n et al. 1983). A p o A I V is s y n t h e s i z e d p r i m a r i l y in the i n t e s t i n e ( G r e e n et al. 1980; K a r a t h a n a s i s et al. 1986a, b) a n d r e p r e s e n t s a m a j o r prot e i n c o n s t i t u e n t of n e w l y secreted i n t e s t i n a l triglyceriderich l i p o p r o t e i n s ( c h y l o m i c r o n s ) ( G r e e n et al. 1979). The possible i n v o l v e m e n t of apo A I V in the activation of L C A T ( S t e i n m e t z a n d U t e r m a n n 1985) a n d the influence of the L C A T r e a c t i o n o n the d i s t r i b u t i o n of apo A I V a m o n g different l i p o p r o t e i n s ( D e l m a t r e et al. 1983) indicate that apo A I V m a y also have a n i m p o r t a n t role in the m e t a b o l i s m of H D L . M a n y R F L P s in the apo A I - C I I I - A I V gene cluster have b e e n i d e n t i f i e d a n d used as genetic m a r k e r s in b o t h f r e q u e n c y a n d association studies ( H u m p h r i e s 1988; Lusis 1988). W e h a v e , t h e r e f o r e , used these m a r k e r s to e x a m i n e the association b e t w e e n genetic v a r i a t i o n in this gene cluster a n d the levels of p l a s m a apo A I , apo C I I I a n d apo A I V , in a s a m p l e of boys a n d y o u n g m e n who are free of a d u l t diseases, are n o t taking m e d i c a t i o n , a n d are s o c i o - e c o n o m i c a l l y a n d genetically h o m o g e n e o u s .
Materials and methods Subjects A sample of 162 boys and young men aged from 7 to 23 (mean age: 10.7) were recruited from a local football club in Brugge, Belgium, for the study. Blood samples were taken after 10-14 h of fasting.
Lipid, lipoprotein, apoprotein and testosterone measurement The lipid assays were based on enzymatic-colorimetric methods, available commercially (Boehringer Mannheim, FRG). For the cholesterol assay the method used was that of Allain et al. (1974) and for triglycerides that of Bucolo and David (1973). LDL cholesterol was calculated using the Friedewald equation (Friedewald et al. 1972). The lipoproteins were separated from plasma and fractionated to LDL and HDL either by density-gradient ultracentrifugation (Rosseneu et al. 1983) or by gel filtration on a Sepharose 6B or a Superose column (Vercaemst et al. 1983). Quantitation of apoproteins, apo AI, apo AIV and apo CIII, was by sandwich ELISA (Bury and Rosseneu 1988; Rosseneu et al. 1988). The method used for the testosterone assay was as reported by Vermeulen (1973).
A p o A I V phenotype Apo AIV phenotyping was performed using a rapid micromethod based on isoelectric focusing (pH: 5-7) of delipidated plasma samples, followed by immunoblotting on nitrocellulose filter using a polyclonal anti-apo AIV antiserum (Knijff et al. 1988).
with 32p-labelled DNA probes using standard methods (Feinberg and Vogelstein 1983). The filters were washed successively at 65~ in 3 x SSC and 1 x SSC, 0.1% SDS for 30 min and exposed to X-ray film for 2-7 days at -70~
Polymorphisms XmnL XmnI digestion and hybridization with an apo AI 2.2 kb probe gives bands of 8.3 kb (X1 allele), 6.6 kb (X2 allele, presence of site) or 8.0 kb (X3 allele). The XmnI polymorphism is located in the 5' flanking region about 3.7 kb from the CAP site of the apo AI gene. The X3 allele is created by a 30 bp deletion approximately 4.0 kb 5' to apo AI (Coleman et al. 1986). Strictly speaking this is a separate polymorphism of the cluster.
G to A substitution. This substitution is located at position - 7 5 bp between the CACAT sequence and the TAAATA box of the apo AI gene. The substitution destroys an MspI cutting site and was identified by the polymerase chain reaction (PCR) and Msp! digestion. A 1 lag sample of DNA was amplified by PCR using i unit of TaqI polymerase (Cetus) in 100 gg of reaction mixture consisting of Cetus polymerase buffer, plus 1 gM of the oligonucleotides 5' AGGGACAGAGCTGATCCTI'GAACTCTTAAG3' and 5' TTAGGGGACACCTAGCCCTCAGGAAGAGCA3' (Saiki et al. 1988). The cycling reactions were carried out in a programmable heat block (Cambio) set to heat the samples at 95~ for 5 min (to denature DNA), and anneal at 55~ for 2 rain and at 70~ for 2 min. The cycle was repeated for 40 times at 95~ for i min, at 55~ for i min and at 72~ for 2 min. Then 20 gg of this amplified DNA fragment was incubated with five units of MspI at 37~ for 2 h after addition of low-salt buffer. Fragments were separated on 4% NuSieve agarose (FMC Bioproducts, Rockland, USA) and the fragments were visualized on a UV transilluminator. The presence of the MspI site results in four fragments of 67 bp, 110 bp, 48 bp and 207 bp, while in the absence of the site three fragments are produced of 177 bp (A allele) 48 bp and 207 bp. Pstl. Digestion with PstI and hybridizaton with an apo AI 2.2 kb probe shows a two-allele polymorphism with bands at 2.2 kb (P1 allele, presence of site) or 3.2 kb (P2 allele). The polymorphic site is located between the apo AI and the CIII gene (Kessling et al. 1985). SstI. The SstI polymorphic site, which is located in the 3' noncoding region of the apo CIII gene (Rees et al. 1985), was identified by PCR and subsequent SstI digestion. A 1-gg sample of DNA was amplified by PCR using 1 unit of TaqI polymerase (Cetus) in 100 ~tl of reaction mixture consisting of Cetus polymerase buffer, plus 1 gM of oligonucleotides 5' CATGGTTGCCTACAGGAGTTC 3' and 5' TGACCTTCCGCACAAAGCTGT 3'. The cycling reactions were set to heat the samples at 95~ for 5 min, to anneal at 50~ for 1 min and to extend at 72~ for 2 min. This cycling reaction was repeated 40 times. Then 20 gl of the amplified DNA fragment was incubated with five units of SstI at 37~ for 2 h. Fragments were separated on a 4% Nusieve Agarose gel and were visualized on a UV transilluminator. The presence of the SstI site yields two fragments of 225 bp and 371 bp, while in the absence of the site one fragment of 596 bp is observed.
D N A analysis
PvuH. PvuIIA-CIII: PvuII digestion and hybridization with an apo CIII 1.0 kb PvuII fragment probe detects bands of 1.0 kb (VA1 allele) or 0.87 kb (VA2 allele, presence of site). The PvuII
Genomic DNA was prepared from leucocytes obtained from a 10 ml whole blood sample using a Triton X-100 lysis method (Kunkel et al. 1977). From 2 to 5 lag of DNA was digested with the restriction enzymes XmnI, PstI, SstI, PvuII, MspI, XbaI (Anglian Biotech Ltd.), electrophoresed in 0.8-1.0% agarose gels, transferred onto Hybond-N nylon membrane (Amersham) and hybridized
polymorphism probably results from a single point mutation in the first intron of the apo CIII gene (Coleman et al. 1986). PvuIIBAIV: PvuII digestion and hybridization with a 1.05 kb PstI fragment from the apo CIII-apo AIV intergenic region (Kessling et al. 1988b) gives bands of 4.3 kb (VB1 allele) or 3.5 kb (VB2 allele, presence of site). The PvuII polymorphism is located in this apo CIII-apo AIV intergenic region (Oettgen et al. 1986).
441
Xbal. XbaI digestion and hybridization with an apo AIV 0,7 kb EcoR[ fragment of a 3' cDNA probe detects bands of 23.1 kb (X1 allele) or 9.0 kb (X2 allele). The XbaI polymorphism is located in the second intron of the apo AIV gene (Karathanasis et al. 1986b). Statistical analysis Statistical analyses were carried out using the statistical package for social sciences (SPSS/PC+) on a Tardor PCA 80 microcomputer (Norusis 1986). Testosterone itself showed a strongly positive skew to the distribution, and this was therefore log transformed before analyses. Other traits were normally distributed. The sample specimens were collected on two different dates, and significant differences in trait levels were observed between these two samples. These were allowed for by performing two-way analyses of covariance, using two factors (RFLP and date of assay) and four covariates: age, age 2, body mass index (BMI) and log
Table 1, Mean (+ SD) lipid, lipoprotein and apoprotein levels in 162 young Belgian males. R 2 x 100 = percent of sample variance explained by concomitants
Triglyceride Cholesterol LDL Apo B HDL Apo At Apo CIII ApoAIV ApoAII
Mean + SD mg/dl
R 2 x 100%
52.2 172.1 110.2 59.1 51.7 121.4 11.6 14.8 26.4
5.6 1,7 3.6 12,3 21.7 3.6 13.7 10.6 4.3
+ + + + + + + + +
APO A-I 5'
Xm n I
A P O CIII "
3'
3' "
Pstl
G-A
14,7 34,0 31.6 12,5 10.6 12.5 2.8 4.8 3.2
Sstl
5'
PvullA
5'
PvullB
APO AIV = 3'
Xbal
1 A
B
I__l
I
2.2kb
C
D
I L.J
~
1.0kb 1.05 kb
I
0.7kb
Fig. 1. Restriction endonuclease map of the apo AI-CIII-AIV gene cluster. Position of the variable cutting sites for XmnI, G-A substitution, PstI, SstI, PvuIIA, PvuIIB and XbaI are shown. A Apo AI 2.2 kb genomic probe; B apo CIII 1.0 kb genomic probe; C apo CIII-AIV 1.05 kb genomic probe; D apo AIV 3' cDNA probe
(testosterone). Two-way interactions between each RFLP and the date of assay were non-significant. The proportion of the phenotypic variance (R 2 x 100 value) associated with each RFLP was estimated by a regression that included covariates, in addition to the factors. Because of the small sample size, it was not appropriate to adjust separately for the effect of covariates in different genotype classes. Statistical significance was accepted at P < 0.05. The calculation of the standardized pairwise linkage disequilibrium value (delta value) was determined as described by Chakravarti et al. (1984). The 95% confidence limits of delta were calculated using Z transformation (Sokal and Rohlf 1983).
Results T h e m e a n v a l u e s for fasting p l a s m a lipids, l i p o p r o t e i n s and apolipoproteins concentrations of the sample of y o u n g m e n (n = 162) a r e s u m m a r i z e d in T a b l e 1. F o r i n d i v i d u a l s w i t h t e s t o s t e r o n e levels b e l o w 100 ng/dl ( r o u g h l y up to age 14 y e a r s ) , t h e r e was a s t r o n g p o s i t i v e correlation between the concentration of testosterone and high-density lipoprotein cholesterol (HDL-C) and a p o A I levels (n = 137, r = 0.41, P < 0.001 a n d r = 0.25 P < 0.01, r e s p e c t i v e l y ) , b u t in i n d i v i d u a l s with t e s t o s t e r o n e levels a b o v e 100 ng/dl a n e g a t i v e c o r r e l a t i o n w a s o b s e r v e d (n = 25, H D L - C r = - 0 . 2 7 N S a n d A p o A I r = - 0.24 NS, r e s p e c t i v e l y ) . T h u s l i p i d s , l i p o p r o t e i n s a n d a p o l i p o p r o t e i n s w e r e a d j u s t e d f o r a g e , age L, B M I , a n d l o g ( t e s t o s t e r o n e ) . T h e p r o p o r t i o n o f t h e s a m p l e variance of e a c h m e a s u r e m e n t which is e x p l a i n e d by t h e s e c o v a r i a t e s is s h o w n in T a b l e 1. A m a p o f t h e a p o A I - C I I I - A I V g e n e c l u s t e r is s h o w n in Fig. 1, i n d i c a t i n g t h e p o s i t i o n o f t h e v a r i a b l e sites u s e d in this study. F o r all r e s t r i c t i o n f r a g m e n t l e n g t h p o l y m o r p h i s m s ( R F L P s ) , g e n o t y p e d i s t r i b u t i o n s d i d n o t differ f r o m t h a t e x p e c t e d for H a r d y - W e i n b e r g p r o p o r t i o n s . T h e f r e q u e n c i e s o f t h e r a r e alleles for all R F L P s a r e s h o w n in T a b l e 2, a n d d o n o t differ significantly f r o m reported values of samples from the UK population. The d e l t a v a l u e s for t h e p a i r w i s e l i n k a g e d i s e q u i l i b r i u m a r e s u m m a r i z e d in T a b l e 3. This s h o w s t h a t t h e r e is s t r o n g l i n k a g e d i s e q u i l i b r i u m ( d e l t a = 0.72) b e t w e e n t h e A allele ( G to A s u b s t i t u t i o n ) a n d t h e X2 allele ( X m n I R F L P ) , b u t no d e t e c t a b l e d i s e q u i l i b r i u m b e t w e e n t h e G to A substitution and any of the other RFLPs. As has been reported previously, detectable linkage disequil i b r i u m was o b s e r v e d b e t w e e n t h e P v u I I A R F L P a n d PvuIIB R F L P ( K e s s l i n g et al. 1988b) a n d b e t w e e n t h e SstI R F L P s a n d t h e XbaI-apo A I V R F L P ( A n t o n a r a k i s et al. 1988). P a i r w i s e analysis r e v e a l e d n o d e t e c t a b l e
Table 2, Relative frequency of the rare allele and the percentage sample variance associated with each RFLP. * P < 0.05
Rare allele frequency
XmnI
G to A
PstI
SstI
PvuIIA-CIII
PvuIIB-AIV
XbaI-AIV
AIV phenotype
0.13
0.18
0.08
0.11
0.20
0.06
0.16
AIV-2 AIV-3
2.3 1.0 0.5
0.4 0.0 2.0
0.5 2.9* 0.6
2.9* 1.7 0.4
2.2 1.7 0.4
1.1 0.9 3.0
R 2 • 100 values Apo AI Apo CIII Apo AIV
2.2 0.5 0.8
3.6* 0.4 0.4
0.11 0.04
442 linkage disequilibrium between the apo A I V phenotypes and any of the polymorphisms. On inspection, both the apo AIV-2 and A I V - 3 phenotypes were found to be associated with the c o m m o n haplotype G-P1-S1-VA1VB1-X1 (G to A substitution, PstI, SstI, P v u I I A , PvuIIB, XbaI); however, the haplotypes G-P1-S2-VA1VB1-X2-AIV-2, A-P1-S2-VA1-VB1-X2-AIV-2, G-P1S2-VA1-VB1-X2-AIV-3, and G-P1-S1-VA2-VB1-X1AIV-3 were also observed ambiguously in two individuals. The contribution of RFLPs to the sample variance in apo A I , apo CIII and apo A I V levels are summarized in Table 2. The significant associations observed in the whole sample were also observed in groups of individuals whose testosterone levels were below and above 100 ng/dl, although in some cases differences were not significant because of the small sample size. The RFLPs associated with the largest effect on apo A I levels were the G to A substitution followed by the P v u I I B - A I V RFLP; on apo CIII levels the PvuIIA-CIII RFLP was associated with the most pronounced effect, while the SstI R F L P and the XbaI-
Table 3. Pairwise delta values for linkage disequilibrium at the apo
AI-CIII-AIVgene cluster (n = 162). The estimates of the pairwise delta values between alleles at the apo AIV locus, as defined by protein phenotypes and RFLPs, were from 0.03 to -0.18 RFLP GA-XrnnI GA-XbaI-AIV GA-PstI GA-SstI GA-PvulI-AIV GA-PvuII-CIII PvulI-CIII-PvuII-AIV SstI-XbaI-AIV
Delta 0.72* 0.15 -0.12 -0.16 -0.11 -0.22 0.46a 0.55a
95% Confidence limit 0.71- 1.00 0.04- 0.34 -0.31- 0.07 -0.35- 0.03 -0.17--0.05 -0.28--0.16 0.30- 0.70 0.42- 0.82
a Detectable disequilibrium
A I V R F L P showed the largest effect in the case of apo A I V levels. The mean values of apo AI, apo CIII and apo A I V , adjusted for covariates and date of assay in groups of individuals with different genotypes, for these RFLPs were compared by analysis of covariance in Table 4. For the PvuIIA-CIII, the G to A substitution and the SstI polymorphism, the number of homozygous individuals for the rare allele was low (seven, two and two respectively) and data from these individuals were combined with those for individuals heterozygous for the RFLP. The G to A substitution showed the strongest association with apo A I level (F = 5.13, P = 0.025). The concentration of apo A I followed a gene dosage effect, with individuals with genotype A A having among the highest apo A I level (130.0 + 9.7 mg/dl, data not shown). The mean level of apo A I in individuals with at least one A allele was 4.5% higher than in individuals with the G G genotype. The G to A substitution was associated with 3.6% of the variance in apo A I concentration in this sample. A significant association (F = 4.53, P = 0.035) was also observed between apo A I level and the PvuIIBA I V RFLP. The mean value of apo A I in individuals with the VB1B1 genotype was 5.8% higher than in those having the VB1B2 genotype (no individuals of VB2VB2 genotype were observed), and was associated with 2.9% of the variance of the apo A I levels. By multiple regression the combined effect of the G to A substitution and the P v u I I B - A I V R F L P accounted for 6.8% of the phenotypic variance in apo A I level; thus, these two polymorphisms showed an approximately additive effect. There was a weak association between apo A I level and the PstI genotype. The mean apo A I level in individuals with the PIP2 genotype was 4.5% lower than in individuals with the P1P1 genotype, but the difference did not reach statistical significance (P = 0.056). The polymorphism most strongly associated with the apo CIII level was the PvuIIA-CIII R F L P in the apo CIII gene. The mean value of apo CIII in those with at least one VA2 allele was 9.0% lower than the mean
Table 4. Mean values of apo AI, apo CIII or apo AIV in individuals with different G to A substitution, PstI, PvuIIA-CIII, PvuIIB-AIV,
XbaI-AIV genotypes and AIV-phenotypes RFLP
Genotype
No.
G to A
GG GA and AA P1P1 PIP2 BIB1 BIB2 AIA1 A1A2 and A2A2 SIS1 $1S2 and $2S2 X1X1 X1X2 X2X2 AIV-11 AIV-12 AIV-13
101 43 134 23 137 17 101 53 104 29 lll 20 13 86 27 10
PstI PvuIIB-AIV PvuIIA-CIII SstI XbaI-AIV
AIV-phenotypes
Trait Apo AI Apo AI Apo AI Apo CIII Apo AIV Apo AIV
Apo AIV
Level mg/dl 119.8 125.0 122.0 116.5 121.9 115.2 12.2 11.1 15.9 12.4 15.2 13.5 16.0 14.7 15.2 18.7
F Value
P Value
5.13
0.025
3.70
0.056
4.53
0.035
5.29
0.023
2.70
0.101
2.80
0.065
2.52
0.199
443 value of apo CIII in individuals with the VA1A1 genotype. These differences were statistically significant (F = 5.29, P -- 0.023; Table 4). Since apo CIII was significantly correlated with the plasma triglyceride (r -- 0.35, P < 0.001), the total cholesterol (r = 0.46, P < 0.001) and the HDL-C (r = 0.56, P < 0.001) levels in this sample, weak associations of the VA2 allele with total cholesterol, triglyceride and HDL-C concentrations were observed (data not shown). There were no statistically significant associations between the SstI RFLP and the levels of either apo CIII or triglyceride. The two individuals homozygous for the $2 allele had a mean apo CIII level 21.0% higher (16.2 and 12.5 mg/dl), and a triglyceride level 14.1% higher (64.0 and 54.5 mg/dl), and an apo AIV level 27.3% lower (11.2 and 10.2 mg/dl), than the sample mean. There was no significant association between the level of apo AIV and any of the polymorphisms tested. Table 4 shows that individuals heterozygous for the polymorphism have the lowest mean value of apo AIV concentration (11.2% lower than those with an XIX1 genotype and 15.6% lower than those with an X2X2 genotype). These differences were not significant and may be due to chance alone. Three apo AIV isoforms were observed in this sample, AIV-1, AIV-2 and AIV-3, at frequencies similar to that reported in other studies (Menzel et al. 1988; Knijff et al. 1988). The percentage of the variance in apo AIV level associated with the AIV phenotype was the highest (3.0%) compared to the other RFLPs, and the mean level of apo AIV in individuals with an apo AIV-1.3 phenotype was 21.0% higher than the combined mean level of the AIV-I.1 and the apo AIV-1.2 phenotypes. This difference was not statistically significant (F = 2.8, P = 0.099). Discussion
Before the onset of sexual maturation, boys and girls have similar levels of HDL-C. However, after puberty lower HDL-C, and higher total cholesterol and triglyceride concentrations have been observed in boys in several large epidemiological studies (Morrison et al. 1979; Orchard et al. 1981; Berenson et al. 1981). Therefore, in any adolescent sample it is important to restrict analysis to the same gender, and to consider the effects of sex hormones as determinants of these changes. We observed inverse correlations between both HDL-C and apo AI concentrations and testosterone levels above 100 ng/dl. The mechanism for the decrease of HDL-C and apo AI concentration in males during puberty is not well understood. Despite the small physical distance (maximum 12 kb) between the G to A substitution and the other RFLPs, with the exception of the X m n I RFLP, significant linkage disequilibrium was not observed (Table 2). However, Thompson et al. (1988) have reported that very large samples are required in order to detect significant negative linkage disequilibrium among RFLPs. Therefore, it cannot be ruled out that any of the RFLPs may be acting as a genetic marker for functionally important
sequence differences at a location elsewhere in the gene cluster. In this small sample, we observed no significant linkage disequilibrium between the a16oAIV protein isoforms and any of the RFLPs. This suggests that the mechanisms of the reported effects associated with RFLPs of the apo AI-CIII-AIV gene cluster on levels of plasma apoproteins, lipoproteins and lipids are unlikely to result from linkage disequilibrium combined with the reported effects on lipid metabolism of different apo AIV phenotypes (Menzel et al. 1988; Knijff et al. 1988). In the present study we have focused on polymorphisms in the apo AI-CIII-AIV gene cluster, previously reported to be associated with differences in plasma lipids, lipoproteins and apoproteins, and examined their relationship only with differences in apolipoprotein levels. A summary of such reported associations is shown in Table 5. In our study, the G to A substitution in the 5' flanking region of the apo AI gene was associated with the strongest influence on plasma apo AI concentration, confirming the observations on men from Bristol (Jeenah et al. 1990) and on females from Italy (Pagani et al. 1990) (Table 5). The region between nucleotides -256 bp and -192 bp upstream of the transcription start site of the apo AI gene is known to be essential for expression in both intestinal and liver cells (Sastry et al. 1988). The A substitution at - 7 5 bp is between the CACAT sequence and the T A A A T A box of the transcriptional start site of apo AI gene and creates a 6-bp perfect repeat (CAGGGC) which has homology to known nuclear-protein binding sites (Kadonaga et al. 1986; Jones et al. 1985). Although studies performed with the "G" allele have not identified proteins binding at the - 7 5 bp region (Sastry et al. 1988), the possibility remains of a binding site for a positive transcription factor being created by the A substitution. Our hypothesis is that this binding may have a direct effect on apo AI gene transcription in the liver or intestine. An alternative hypothesis is the existence of linkage disequilibrium between the G to A substitution and functionally important DNA changes elsewhere in the apo AI-CII-AIV gene cluster which alter the expression of the apo AI gene, and this possibility cannot be ruled out. Two previous reports (Table 5) that the rare allele (VB2) of the P v u l I B - A I V RFLP was associated with lower concentrations of apo AI were confirmed strongly in our sample. There is an additive effect on the apo AI levels associated with the G to A substitution and the P v u I I B - A I V RFLP in the absence of detectable linkage disequilibrium between them. This suggests that the P v u I I B - A I V RFLP is associated with an independent effect on variation of apo AI levels. Over the last few years there have been conflicting reports of the association between the PstI RFLP and levels of apo AI and HDL-C (Table 5). We observed that the P2 allele was associated with a lower plasma concentration of apo AI, although this was not statistically significant. The discrepancies between studies may be explained by differences in sample selection (e.g., CAD patients or healthy individuals), but it cannot be ruled out that the different associations are due to chance alone.
444 Table 5. Comparison of published association studies between apolipoprotein levels and apo AI-CIII-AIV RFLPs in healthy individuals. NS, Not significant; ~, associated with high levels of the trait; ~, associated with low levels of the trait
Polymorphisms
Traits
Rare allele association
Sample (age in years) b
Reference
MspI
Apo AI Apo AI Apo AI Apo AI
T P < 0.006 ]' NS ]" P < 0.025 T P < 0.025
136 F (45-59) 108 M (27-51) 96 M (45-59) 162 M (7-23)
Pagani et al. (1990) Pagani et al. (1990) Jeenah et al. (1991) This study
PstI
Apo AI Apo AI Apo AI Apo AI Apo AI
'~ F < 0.050 None ~ NS ,LNS ,~ P < 0.056
110 M (45-59) 118 M (> 55) 103 M (> 60) 145 MF (53-40) 162 M (7-23)
Kessling et al. (1988a) Paulweber et al. (1988) Wile et al. (1989) Ordovas et al. (1991) This study
SstI
Apo AI Apo AI Apo AI Apo AI Apo CIII Apo CIII Apo AIV Apo AIV
~ NS None "~NS ,~ NS T P < 0.040 T NS ~ NS ~ NS
145 MF 154 MF 58 F 162 M 154 MF 162 M 79 MF 162 M
(53-40) (> 60) (> 20) (7-23) (>60) (7-23) (> 60) (7-23)
Ordovas et al. (1988) Shoulders et al. (1988) Anderson et al. (1989) This study Shoulders et al. (1991) This study Shoulders et al. (1991) This study
PvuIIA-CIIP
Apo AI Apo AI Apo CIII Apo AI Apo AI Apo AI Apo AI Apo CIII
,LP < ~ NS ,LP < ,~ NS ~ NS ,LP < ~P < ,LP
9 Springer-Verlag1992
Polymorphisms in the apolipoprotein (apo) AI-CIII-AIV gene cluster: detection of genetic variation determining plasma apo AI, apo CIII and apo AIV concentrations H. PauI-Hayase 1, M. Rosseneu 2, D. Robinson 3, J. P. Van Bervliet 3, J. P. Deslypere 4, and S. E. Humphries 1 1Chafing Cross Sunley Research Centre, 1 Lurgan Avenue, Hammersmith, London W6 8LW, UK 2Departments of Clinical Chemistry and Paediatrics, A-Z St Jan, Brugge, Belgium 3BUPA Medical Research, Battle Bridge House, 300 Grays Inn Road, London WCIX 8DU, UK 4Department of Endocrinology, University Hospital, Ghent, Belgium Received August 29, 1990 / Revised July 30, 1991
Summary. We have examined the associations between levels of plasma apolipoprotein (apo) AI, apo CIII and apo AIV and genetic variation in the apo AI-CIII-AIV gene cluster in 162 boys and young men from Belgium aged from 7 to 23 years. Genotypes were determined for six restriction enzymes XmnI, PstI, SstI, PvuIIA-CIII, PvuIIB-AIV and XbaI, and for the G to A substitution at - 7 5 bp in the 5' region of the apo AI gene. The polymorphism most strongly associated with apo AI levels was the G to A substitution (P = 0.025, R z x 100 = 3.6%) confirming previous observations. The polymorphism most strongly associated with apo CIII levels was that of PvuIIA-CIII (P = 0.023, R 2 • 100 = 2.9%) in the apo CIII gene. This novel association must be interpreted with caution until it has been confirmed in an independent sample. The polymorphism associated with the largest effect on apo AIV levels was that detected with XbaI in the apo AIV gene, but this association was not statistically significant. Previously reported associations between the SstI polymorphism and triglyceride levels, and between the PstI polymorphism and apo AI levels, were weakly detected in the present sample. Our results show that variation associated with some of the polymorphisms in the apo AI-CIII-AIV cluster makes a small, but statistically significant, contribution to the determination of apo AI and apo CIII levels in this sample of young men and boys. These observations may, in part, explain reported associations between polymorphisms in this gene cluster, differences in plasma lipid and lipoprotein levels, a n d prevalence of coronary artery disease.
Introduction A large number of epidemiological studies have demonstrated that low serum levels of high-density lipoprotein Offprint requests to: S. E. Humphries
cholesterol (HDL-C) and apolipoprotein AI (apo AI) are risk factors for the development of coronary artery disease (CAD) (Miller et al. 1977; Norum et al. 1982; Yaari et al. 1981). Various environmental factors can significantly affect plasma HDL-C levels, including smoking, exogenous gonadal steroids, alcohol intake, stress, infection, the amount of exercise, and some drug therapy (Haffner et al. 1985; Berg et al. 1986; Hartung et al. 1986). On the other hand, the heritability of HDLC and apo AI has been estimated in various studies to be between 0.43 and 0.66 (Hamsten et al. 1986; Kuusi et al. 1987). Recently, evidence for a major gene determining individual differences in quantitative levels of apo AI was demonstrated in a population-based sample by applying biometrical techniques (Moll et al. 1989). Altered HDL-C levels have also been associated with rearrangement of the AI gene (Karathanasis et al. 1987), with specific amino acid substitutions in the apo AI protein (e.g., Eckardstein et al. 1989) and with a defect in the gene for the cholesterol ester transport protein (CETP) (Brown et al. 1989). However, these mutations are rare and the study of common genetic factors that determine HDL-C and apo AI levels is important. The gene coding for apo AI is on the long arm of chromosome 11, next to the genes for Apo CIII and Apo AIV (Karathanasis et al. 1983a). Apo AI is the major protein of mature H D L and is synthesized in the liver and intestine. Apo AI can activate lecithin cholesterol acyltransferase (LCAT), which converts unesterified cholesterol to its esterified form in plasma. Apo AI may also be the ligand for an H D L receptor involved in the removal of cholesterol from peripheral tissue (Oram et al. 1987; Schmitz et al. 1985). This would allow subsequent transport of excess cholesterol from tissues to the liver, a mechanism referred to as "reverse cholesterol transport". Apo CIII is synthesized mainly in the liver and to a lesser degree in the intestine and is a constituent of very-low-density lipoprotein (VLDL) particles and H D L particles (Herbert et al. 1982). In vitro, apo CIII
440 has b e e n s h o w n to i n h i b i t the activities of b o t h lipop r o t e i n lipase ( B r o w n et al. 1972; Q u a r f o r d t et al. 1982) a n d h e p a t i c lipase ( K i n n u n e n et al. 1983). A p o A I V is s y n t h e s i z e d p r i m a r i l y in the i n t e s t i n e ( G r e e n et al. 1980; K a r a t h a n a s i s et al. 1986a, b) a n d r e p r e s e n t s a m a j o r prot e i n c o n s t i t u e n t of n e w l y secreted i n t e s t i n a l triglyceriderich l i p o p r o t e i n s ( c h y l o m i c r o n s ) ( G r e e n et al. 1979). The possible i n v o l v e m e n t of apo A I V in the activation of L C A T ( S t e i n m e t z a n d U t e r m a n n 1985) a n d the influence of the L C A T r e a c t i o n o n the d i s t r i b u t i o n of apo A I V a m o n g different l i p o p r o t e i n s ( D e l m a t r e et al. 1983) indicate that apo A I V m a y also have a n i m p o r t a n t role in the m e t a b o l i s m of H D L . M a n y R F L P s in the apo A I - C I I I - A I V gene cluster have b e e n i d e n t i f i e d a n d used as genetic m a r k e r s in b o t h f r e q u e n c y a n d association studies ( H u m p h r i e s 1988; Lusis 1988). W e h a v e , t h e r e f o r e , used these m a r k e r s to e x a m i n e the association b e t w e e n genetic v a r i a t i o n in this gene cluster a n d the levels of p l a s m a apo A I , apo C I I I a n d apo A I V , in a s a m p l e of boys a n d y o u n g m e n who are free of a d u l t diseases, are n o t taking m e d i c a t i o n , a n d are s o c i o - e c o n o m i c a l l y a n d genetically h o m o g e n e o u s .
Materials and methods Subjects A sample of 162 boys and young men aged from 7 to 23 (mean age: 10.7) were recruited from a local football club in Brugge, Belgium, for the study. Blood samples were taken after 10-14 h of fasting.
Lipid, lipoprotein, apoprotein and testosterone measurement The lipid assays were based on enzymatic-colorimetric methods, available commercially (Boehringer Mannheim, FRG). For the cholesterol assay the method used was that of Allain et al. (1974) and for triglycerides that of Bucolo and David (1973). LDL cholesterol was calculated using the Friedewald equation (Friedewald et al. 1972). The lipoproteins were separated from plasma and fractionated to LDL and HDL either by density-gradient ultracentrifugation (Rosseneu et al. 1983) or by gel filtration on a Sepharose 6B or a Superose column (Vercaemst et al. 1983). Quantitation of apoproteins, apo AI, apo AIV and apo CIII, was by sandwich ELISA (Bury and Rosseneu 1988; Rosseneu et al. 1988). The method used for the testosterone assay was as reported by Vermeulen (1973).
A p o A I V phenotype Apo AIV phenotyping was performed using a rapid micromethod based on isoelectric focusing (pH: 5-7) of delipidated plasma samples, followed by immunoblotting on nitrocellulose filter using a polyclonal anti-apo AIV antiserum (Knijff et al. 1988).
with 32p-labelled DNA probes using standard methods (Feinberg and Vogelstein 1983). The filters were washed successively at 65~ in 3 x SSC and 1 x SSC, 0.1% SDS for 30 min and exposed to X-ray film for 2-7 days at -70~
Polymorphisms XmnL XmnI digestion and hybridization with an apo AI 2.2 kb probe gives bands of 8.3 kb (X1 allele), 6.6 kb (X2 allele, presence of site) or 8.0 kb (X3 allele). The XmnI polymorphism is located in the 5' flanking region about 3.7 kb from the CAP site of the apo AI gene. The X3 allele is created by a 30 bp deletion approximately 4.0 kb 5' to apo AI (Coleman et al. 1986). Strictly speaking this is a separate polymorphism of the cluster.
G to A substitution. This substitution is located at position - 7 5 bp between the CACAT sequence and the TAAATA box of the apo AI gene. The substitution destroys an MspI cutting site and was identified by the polymerase chain reaction (PCR) and Msp! digestion. A 1 lag sample of DNA was amplified by PCR using i unit of TaqI polymerase (Cetus) in 100 gg of reaction mixture consisting of Cetus polymerase buffer, plus 1 gM of the oligonucleotides 5' AGGGACAGAGCTGATCCTI'GAACTCTTAAG3' and 5' TTAGGGGACACCTAGCCCTCAGGAAGAGCA3' (Saiki et al. 1988). The cycling reactions were carried out in a programmable heat block (Cambio) set to heat the samples at 95~ for 5 min (to denature DNA), and anneal at 55~ for 2 rain and at 70~ for 2 min. The cycle was repeated for 40 times at 95~ for i min, at 55~ for i min and at 72~ for 2 min. Then 20 gg of this amplified DNA fragment was incubated with five units of MspI at 37~ for 2 h after addition of low-salt buffer. Fragments were separated on 4% NuSieve agarose (FMC Bioproducts, Rockland, USA) and the fragments were visualized on a UV transilluminator. The presence of the MspI site results in four fragments of 67 bp, 110 bp, 48 bp and 207 bp, while in the absence of the site three fragments are produced of 177 bp (A allele) 48 bp and 207 bp. Pstl. Digestion with PstI and hybridizaton with an apo AI 2.2 kb probe shows a two-allele polymorphism with bands at 2.2 kb (P1 allele, presence of site) or 3.2 kb (P2 allele). The polymorphic site is located between the apo AI and the CIII gene (Kessling et al. 1985). SstI. The SstI polymorphic site, which is located in the 3' noncoding region of the apo CIII gene (Rees et al. 1985), was identified by PCR and subsequent SstI digestion. A 1-gg sample of DNA was amplified by PCR using 1 unit of TaqI polymerase (Cetus) in 100 ~tl of reaction mixture consisting of Cetus polymerase buffer, plus 1 gM of oligonucleotides 5' CATGGTTGCCTACAGGAGTTC 3' and 5' TGACCTTCCGCACAAAGCTGT 3'. The cycling reactions were set to heat the samples at 95~ for 5 min, to anneal at 50~ for 1 min and to extend at 72~ for 2 min. This cycling reaction was repeated 40 times. Then 20 gl of the amplified DNA fragment was incubated with five units of SstI at 37~ for 2 h. Fragments were separated on a 4% Nusieve Agarose gel and were visualized on a UV transilluminator. The presence of the SstI site yields two fragments of 225 bp and 371 bp, while in the absence of the site one fragment of 596 bp is observed.
D N A analysis
PvuH. PvuIIA-CIII: PvuII digestion and hybridization with an apo CIII 1.0 kb PvuII fragment probe detects bands of 1.0 kb (VA1 allele) or 0.87 kb (VA2 allele, presence of site). The PvuII
Genomic DNA was prepared from leucocytes obtained from a 10 ml whole blood sample using a Triton X-100 lysis method (Kunkel et al. 1977). From 2 to 5 lag of DNA was digested with the restriction enzymes XmnI, PstI, SstI, PvuII, MspI, XbaI (Anglian Biotech Ltd.), electrophoresed in 0.8-1.0% agarose gels, transferred onto Hybond-N nylon membrane (Amersham) and hybridized
polymorphism probably results from a single point mutation in the first intron of the apo CIII gene (Coleman et al. 1986). PvuIIBAIV: PvuII digestion and hybridization with a 1.05 kb PstI fragment from the apo CIII-apo AIV intergenic region (Kessling et al. 1988b) gives bands of 4.3 kb (VB1 allele) or 3.5 kb (VB2 allele, presence of site). The PvuII polymorphism is located in this apo CIII-apo AIV intergenic region (Oettgen et al. 1986).
441
Xbal. XbaI digestion and hybridization with an apo AIV 0,7 kb EcoR[ fragment of a 3' cDNA probe detects bands of 23.1 kb (X1 allele) or 9.0 kb (X2 allele). The XbaI polymorphism is located in the second intron of the apo AIV gene (Karathanasis et al. 1986b). Statistical analysis Statistical analyses were carried out using the statistical package for social sciences (SPSS/PC+) on a Tardor PCA 80 microcomputer (Norusis 1986). Testosterone itself showed a strongly positive skew to the distribution, and this was therefore log transformed before analyses. Other traits were normally distributed. The sample specimens were collected on two different dates, and significant differences in trait levels were observed between these two samples. These were allowed for by performing two-way analyses of covariance, using two factors (RFLP and date of assay) and four covariates: age, age 2, body mass index (BMI) and log
Table 1, Mean (+ SD) lipid, lipoprotein and apoprotein levels in 162 young Belgian males. R 2 x 100 = percent of sample variance explained by concomitants
Triglyceride Cholesterol LDL Apo B HDL Apo At Apo CIII ApoAIV ApoAII
Mean + SD mg/dl
R 2 x 100%
52.2 172.1 110.2 59.1 51.7 121.4 11.6 14.8 26.4
5.6 1,7 3.6 12,3 21.7 3.6 13.7 10.6 4.3
+ + + + + + + + +
APO A-I 5'
Xm n I
A P O CIII "
3'
3' "
Pstl
G-A
14,7 34,0 31.6 12,5 10.6 12.5 2.8 4.8 3.2
Sstl
5'
PvullA
5'
PvullB
APO AIV = 3'
Xbal
1 A
B
I__l
I
2.2kb
C
D
I L.J
~
1.0kb 1.05 kb
I
0.7kb
Fig. 1. Restriction endonuclease map of the apo AI-CIII-AIV gene cluster. Position of the variable cutting sites for XmnI, G-A substitution, PstI, SstI, PvuIIA, PvuIIB and XbaI are shown. A Apo AI 2.2 kb genomic probe; B apo CIII 1.0 kb genomic probe; C apo CIII-AIV 1.05 kb genomic probe; D apo AIV 3' cDNA probe
(testosterone). Two-way interactions between each RFLP and the date of assay were non-significant. The proportion of the phenotypic variance (R 2 x 100 value) associated with each RFLP was estimated by a regression that included covariates, in addition to the factors. Because of the small sample size, it was not appropriate to adjust separately for the effect of covariates in different genotype classes. Statistical significance was accepted at P < 0.05. The calculation of the standardized pairwise linkage disequilibrium value (delta value) was determined as described by Chakravarti et al. (1984). The 95% confidence limits of delta were calculated using Z transformation (Sokal and Rohlf 1983).
Results T h e m e a n v a l u e s for fasting p l a s m a lipids, l i p o p r o t e i n s and apolipoproteins concentrations of the sample of y o u n g m e n (n = 162) a r e s u m m a r i z e d in T a b l e 1. F o r i n d i v i d u a l s w i t h t e s t o s t e r o n e levels b e l o w 100 ng/dl ( r o u g h l y up to age 14 y e a r s ) , t h e r e was a s t r o n g p o s i t i v e correlation between the concentration of testosterone and high-density lipoprotein cholesterol (HDL-C) and a p o A I levels (n = 137, r = 0.41, P < 0.001 a n d r = 0.25 P < 0.01, r e s p e c t i v e l y ) , b u t in i n d i v i d u a l s with t e s t o s t e r o n e levels a b o v e 100 ng/dl a n e g a t i v e c o r r e l a t i o n w a s o b s e r v e d (n = 25, H D L - C r = - 0 . 2 7 N S a n d A p o A I r = - 0.24 NS, r e s p e c t i v e l y ) . T h u s l i p i d s , l i p o p r o t e i n s a n d a p o l i p o p r o t e i n s w e r e a d j u s t e d f o r a g e , age L, B M I , a n d l o g ( t e s t o s t e r o n e ) . T h e p r o p o r t i o n o f t h e s a m p l e variance of e a c h m e a s u r e m e n t which is e x p l a i n e d by t h e s e c o v a r i a t e s is s h o w n in T a b l e 1. A m a p o f t h e a p o A I - C I I I - A I V g e n e c l u s t e r is s h o w n in Fig. 1, i n d i c a t i n g t h e p o s i t i o n o f t h e v a r i a b l e sites u s e d in this study. F o r all r e s t r i c t i o n f r a g m e n t l e n g t h p o l y m o r p h i s m s ( R F L P s ) , g e n o t y p e d i s t r i b u t i o n s d i d n o t differ f r o m t h a t e x p e c t e d for H a r d y - W e i n b e r g p r o p o r t i o n s . T h e f r e q u e n c i e s o f t h e r a r e alleles for all R F L P s a r e s h o w n in T a b l e 2, a n d d o n o t differ significantly f r o m reported values of samples from the UK population. The d e l t a v a l u e s for t h e p a i r w i s e l i n k a g e d i s e q u i l i b r i u m a r e s u m m a r i z e d in T a b l e 3. This s h o w s t h a t t h e r e is s t r o n g l i n k a g e d i s e q u i l i b r i u m ( d e l t a = 0.72) b e t w e e n t h e A allele ( G to A s u b s t i t u t i o n ) a n d t h e X2 allele ( X m n I R F L P ) , b u t no d e t e c t a b l e d i s e q u i l i b r i u m b e t w e e n t h e G to A substitution and any of the other RFLPs. As has been reported previously, detectable linkage disequil i b r i u m was o b s e r v e d b e t w e e n t h e P v u I I A R F L P a n d PvuIIB R F L P ( K e s s l i n g et al. 1988b) a n d b e t w e e n t h e SstI R F L P s a n d t h e XbaI-apo A I V R F L P ( A n t o n a r a k i s et al. 1988). P a i r w i s e analysis r e v e a l e d n o d e t e c t a b l e
Table 2, Relative frequency of the rare allele and the percentage sample variance associated with each RFLP. * P < 0.05
Rare allele frequency
XmnI
G to A
PstI
SstI
PvuIIA-CIII
PvuIIB-AIV
XbaI-AIV
AIV phenotype
0.13
0.18
0.08
0.11
0.20
0.06
0.16
AIV-2 AIV-3
2.3 1.0 0.5
0.4 0.0 2.0
0.5 2.9* 0.6
2.9* 1.7 0.4
2.2 1.7 0.4
1.1 0.9 3.0
R 2 • 100 values Apo AI Apo CIII Apo AIV
2.2 0.5 0.8
3.6* 0.4 0.4
0.11 0.04
442 linkage disequilibrium between the apo A I V phenotypes and any of the polymorphisms. On inspection, both the apo AIV-2 and A I V - 3 phenotypes were found to be associated with the c o m m o n haplotype G-P1-S1-VA1VB1-X1 (G to A substitution, PstI, SstI, P v u I I A , PvuIIB, XbaI); however, the haplotypes G-P1-S2-VA1VB1-X2-AIV-2, A-P1-S2-VA1-VB1-X2-AIV-2, G-P1S2-VA1-VB1-X2-AIV-3, and G-P1-S1-VA2-VB1-X1AIV-3 were also observed ambiguously in two individuals. The contribution of RFLPs to the sample variance in apo A I , apo CIII and apo A I V levels are summarized in Table 2. The significant associations observed in the whole sample were also observed in groups of individuals whose testosterone levels were below and above 100 ng/dl, although in some cases differences were not significant because of the small sample size. The RFLPs associated with the largest effect on apo A I levels were the G to A substitution followed by the P v u I I B - A I V RFLP; on apo CIII levels the PvuIIA-CIII RFLP was associated with the most pronounced effect, while the SstI R F L P and the XbaI-
Table 3. Pairwise delta values for linkage disequilibrium at the apo
AI-CIII-AIVgene cluster (n = 162). The estimates of the pairwise delta values between alleles at the apo AIV locus, as defined by protein phenotypes and RFLPs, were from 0.03 to -0.18 RFLP GA-XrnnI GA-XbaI-AIV GA-PstI GA-SstI GA-PvulI-AIV GA-PvuII-CIII PvulI-CIII-PvuII-AIV SstI-XbaI-AIV
Delta 0.72* 0.15 -0.12 -0.16 -0.11 -0.22 0.46a 0.55a
95% Confidence limit 0.71- 1.00 0.04- 0.34 -0.31- 0.07 -0.35- 0.03 -0.17--0.05 -0.28--0.16 0.30- 0.70 0.42- 0.82
a Detectable disequilibrium
A I V R F L P showed the largest effect in the case of apo A I V levels. The mean values of apo AI, apo CIII and apo A I V , adjusted for covariates and date of assay in groups of individuals with different genotypes, for these RFLPs were compared by analysis of covariance in Table 4. For the PvuIIA-CIII, the G to A substitution and the SstI polymorphism, the number of homozygous individuals for the rare allele was low (seven, two and two respectively) and data from these individuals were combined with those for individuals heterozygous for the RFLP. The G to A substitution showed the strongest association with apo A I level (F = 5.13, P = 0.025). The concentration of apo A I followed a gene dosage effect, with individuals with genotype A A having among the highest apo A I level (130.0 + 9.7 mg/dl, data not shown). The mean level of apo A I in individuals with at least one A allele was 4.5% higher than in individuals with the G G genotype. The G to A substitution was associated with 3.6% of the variance in apo A I concentration in this sample. A significant association (F = 4.53, P = 0.035) was also observed between apo A I level and the PvuIIBA I V RFLP. The mean value of apo A I in individuals with the VB1B1 genotype was 5.8% higher than in those having the VB1B2 genotype (no individuals of VB2VB2 genotype were observed), and was associated with 2.9% of the variance of the apo A I levels. By multiple regression the combined effect of the G to A substitution and the P v u I I B - A I V R F L P accounted for 6.8% of the phenotypic variance in apo A I level; thus, these two polymorphisms showed an approximately additive effect. There was a weak association between apo A I level and the PstI genotype. The mean apo A I level in individuals with the PIP2 genotype was 4.5% lower than in individuals with the P1P1 genotype, but the difference did not reach statistical significance (P = 0.056). The polymorphism most strongly associated with the apo CIII level was the PvuIIA-CIII R F L P in the apo CIII gene. The mean value of apo CIII in those with at least one VA2 allele was 9.0% lower than the mean
Table 4. Mean values of apo AI, apo CIII or apo AIV in individuals with different G to A substitution, PstI, PvuIIA-CIII, PvuIIB-AIV,
XbaI-AIV genotypes and AIV-phenotypes RFLP
Genotype
No.
G to A
GG GA and AA P1P1 PIP2 BIB1 BIB2 AIA1 A1A2 and A2A2 SIS1 $1S2 and $2S2 X1X1 X1X2 X2X2 AIV-11 AIV-12 AIV-13
101 43 134 23 137 17 101 53 104 29 lll 20 13 86 27 10
PstI PvuIIB-AIV PvuIIA-CIII SstI XbaI-AIV
AIV-phenotypes
Trait Apo AI Apo AI Apo AI Apo CIII Apo AIV Apo AIV
Apo AIV
Level mg/dl 119.8 125.0 122.0 116.5 121.9 115.2 12.2 11.1 15.9 12.4 15.2 13.5 16.0 14.7 15.2 18.7
F Value
P Value
5.13
0.025
3.70
0.056
4.53
0.035
5.29
0.023
2.70
0.101
2.80
0.065
2.52
0.199
443 value of apo CIII in individuals with the VA1A1 genotype. These differences were statistically significant (F = 5.29, P -- 0.023; Table 4). Since apo CIII was significantly correlated with the plasma triglyceride (r -- 0.35, P < 0.001), the total cholesterol (r = 0.46, P < 0.001) and the HDL-C (r = 0.56, P < 0.001) levels in this sample, weak associations of the VA2 allele with total cholesterol, triglyceride and HDL-C concentrations were observed (data not shown). There were no statistically significant associations between the SstI RFLP and the levels of either apo CIII or triglyceride. The two individuals homozygous for the $2 allele had a mean apo CIII level 21.0% higher (16.2 and 12.5 mg/dl), and a triglyceride level 14.1% higher (64.0 and 54.5 mg/dl), and an apo AIV level 27.3% lower (11.2 and 10.2 mg/dl), than the sample mean. There was no significant association between the level of apo AIV and any of the polymorphisms tested. Table 4 shows that individuals heterozygous for the polymorphism have the lowest mean value of apo AIV concentration (11.2% lower than those with an XIX1 genotype and 15.6% lower than those with an X2X2 genotype). These differences were not significant and may be due to chance alone. Three apo AIV isoforms were observed in this sample, AIV-1, AIV-2 and AIV-3, at frequencies similar to that reported in other studies (Menzel et al. 1988; Knijff et al. 1988). The percentage of the variance in apo AIV level associated with the AIV phenotype was the highest (3.0%) compared to the other RFLPs, and the mean level of apo AIV in individuals with an apo AIV-1.3 phenotype was 21.0% higher than the combined mean level of the AIV-I.1 and the apo AIV-1.2 phenotypes. This difference was not statistically significant (F = 2.8, P = 0.099). Discussion
Before the onset of sexual maturation, boys and girls have similar levels of HDL-C. However, after puberty lower HDL-C, and higher total cholesterol and triglyceride concentrations have been observed in boys in several large epidemiological studies (Morrison et al. 1979; Orchard et al. 1981; Berenson et al. 1981). Therefore, in any adolescent sample it is important to restrict analysis to the same gender, and to consider the effects of sex hormones as determinants of these changes. We observed inverse correlations between both HDL-C and apo AI concentrations and testosterone levels above 100 ng/dl. The mechanism for the decrease of HDL-C and apo AI concentration in males during puberty is not well understood. Despite the small physical distance (maximum 12 kb) between the G to A substitution and the other RFLPs, with the exception of the X m n I RFLP, significant linkage disequilibrium was not observed (Table 2). However, Thompson et al. (1988) have reported that very large samples are required in order to detect significant negative linkage disequilibrium among RFLPs. Therefore, it cannot be ruled out that any of the RFLPs may be acting as a genetic marker for functionally important
sequence differences at a location elsewhere in the gene cluster. In this small sample, we observed no significant linkage disequilibrium between the a16oAIV protein isoforms and any of the RFLPs. This suggests that the mechanisms of the reported effects associated with RFLPs of the apo AI-CIII-AIV gene cluster on levels of plasma apoproteins, lipoproteins and lipids are unlikely to result from linkage disequilibrium combined with the reported effects on lipid metabolism of different apo AIV phenotypes (Menzel et al. 1988; Knijff et al. 1988). In the present study we have focused on polymorphisms in the apo AI-CIII-AIV gene cluster, previously reported to be associated with differences in plasma lipids, lipoproteins and apoproteins, and examined their relationship only with differences in apolipoprotein levels. A summary of such reported associations is shown in Table 5. In our study, the G to A substitution in the 5' flanking region of the apo AI gene was associated with the strongest influence on plasma apo AI concentration, confirming the observations on men from Bristol (Jeenah et al. 1990) and on females from Italy (Pagani et al. 1990) (Table 5). The region between nucleotides -256 bp and -192 bp upstream of the transcription start site of the apo AI gene is known to be essential for expression in both intestinal and liver cells (Sastry et al. 1988). The A substitution at - 7 5 bp is between the CACAT sequence and the T A A A T A box of the transcriptional start site of apo AI gene and creates a 6-bp perfect repeat (CAGGGC) which has homology to known nuclear-protein binding sites (Kadonaga et al. 1986; Jones et al. 1985). Although studies performed with the "G" allele have not identified proteins binding at the - 7 5 bp region (Sastry et al. 1988), the possibility remains of a binding site for a positive transcription factor being created by the A substitution. Our hypothesis is that this binding may have a direct effect on apo AI gene transcription in the liver or intestine. An alternative hypothesis is the existence of linkage disequilibrium between the G to A substitution and functionally important DNA changes elsewhere in the apo AI-CII-AIV gene cluster which alter the expression of the apo AI gene, and this possibility cannot be ruled out. Two previous reports (Table 5) that the rare allele (VB2) of the P v u l I B - A I V RFLP was associated with lower concentrations of apo AI were confirmed strongly in our sample. There is an additive effect on the apo AI levels associated with the G to A substitution and the P v u I I B - A I V RFLP in the absence of detectable linkage disequilibrium between them. This suggests that the P v u I I B - A I V RFLP is associated with an independent effect on variation of apo AI levels. Over the last few years there have been conflicting reports of the association between the PstI RFLP and levels of apo AI and HDL-C (Table 5). We observed that the P2 allele was associated with a lower plasma concentration of apo AI, although this was not statistically significant. The discrepancies between studies may be explained by differences in sample selection (e.g., CAD patients or healthy individuals), but it cannot be ruled out that the different associations are due to chance alone.
444 Table 5. Comparison of published association studies between apolipoprotein levels and apo AI-CIII-AIV RFLPs in healthy individuals. NS, Not significant; ~, associated with high levels of the trait; ~, associated with low levels of the trait
Polymorphisms
Traits
Rare allele association
Sample (age in years) b
Reference
MspI
Apo AI Apo AI Apo AI Apo AI
T P < 0.006 ]' NS ]" P < 0.025 T P < 0.025
136 F (45-59) 108 M (27-51) 96 M (45-59) 162 M (7-23)
Pagani et al. (1990) Pagani et al. (1990) Jeenah et al. (1991) This study
PstI
Apo AI Apo AI Apo AI Apo AI Apo AI
'~ F < 0.050 None ~ NS ,LNS ,~ P < 0.056
110 M (45-59) 118 M (> 55) 103 M (> 60) 145 MF (53-40) 162 M (7-23)
Kessling et al. (1988a) Paulweber et al. (1988) Wile et al. (1989) Ordovas et al. (1991) This study
SstI
Apo AI Apo AI Apo AI Apo AI Apo CIII Apo CIII Apo AIV Apo AIV
~ NS None "~NS ,~ NS T P < 0.040 T NS ~ NS ~ NS
145 MF 154 MF 58 F 162 M 154 MF 162 M 79 MF 162 M
(53-40) (> 60) (> 20) (7-23) (>60) (7-23) (> 60) (7-23)
Ordovas et al. (1988) Shoulders et al. (1988) Anderson et al. (1989) This study Shoulders et al. (1991) This study Shoulders et al. (1991) This study
PvuIIA-CIIP
Apo AI Apo AI Apo CIII Apo AI Apo AI Apo AI Apo AI Apo CIII
,LP < ~ NS ,LP < ,~ NS ~ NS ,LP < ~P < ,LP
