Pharmacogenetics of Statin Therapies Daniel I. Chasman, Ph.D. Division of Preventive Medicine Brigham and Women’s Hospital Johanna and Ralph DeStefano Personalized Health Care Conference OSU Medical Center Columbus, OH Oct 6, 2011
Disclosure Funding for this research provided by AstraZeneca Celera
Background and research questions Background There is large inter-individual response to statin therapy as measured by LDL-C reduction, a strong predictor of risk reduction. Some of this variation may be correlated with genetic variation. Research questions What genes, in the entire genome, carry common genetic variation associated with LDL-C lowering on statin therapy? What are the magnitudes of these effects? Are there interactions involving these gene variants? To what extent do the genetic effects explain variation in inter-individual statin response?
Some previous genetic analyses of LDL-C lowering with statin treatment Candidate gene analysis1 HMGCR – target of statin therapy APOE – major apolipoprotein of VLDL, IDL, chylomicrons LDLR – LDL receptor ABCG5/8 – sterol transporters CYP7A1 – cytochrome P450 family metabolizing enzyme ABCG2 – transporter in liver, kidney Genome-wide association studies (GWAS)2 CLMN association in GWAS of PRINCE, CAP, and TNT (pravastatin, simvastatin, atorvastatin) GRIK4 association in GWAS of TNT (atorvastatin) SLCO1B1 association myopathy GWAS of SEARCH (simvastatin) References 1JAMA. 2004 291:2821, ATVB 2010 30:1485, Circ. 2008 117:1537; Athero. 2004 175:287; Am J Cardi. 2004 93:104. Athero. 2001 158:183. Circulation Cardiovascular genetics 2010 doi: 10.1161. 2PLoS One. 2010 5:e9763 , N Engl J Med. 2008 ;359:789, Circ Cardio Genet. 2009 2:173.
Known pharmacologic pathways for statin therapy temporal sequence of statin pharmacology CYP’s ABCB1 ABCG2 SLCO1B1 HMGCR LDLR APOE ABCG5/8
Genome-wide association study (GWAS) Focus on single nucleotide polymorphisms (SNPs), the most prevalent form of genetic variation in people SNPs typically have two alleles, the major allele (≥50% in the population) and the minor allele (<50%) In a single experiment, examine all common SNPs at once. For 1% allele frequency, approx. 1 million SNPs. Require very stringent significance, e.g. p < 5 x 10-8 Test for association of the minor allele with LDL-C response among individuals taking statin
Population with genome-wide data from JUPITER JUPITER trial enrolled 17,802 participants with LDL-C < 130mg/dL and C-reactive protein (CRP) ≥ 2mg/L for primary prevention with random allocation to rosuvastatin (20 mg/day). Treatment highly effective in this population1 Genotyping on the Illumina Omni 1M Quad platform by Illumina 8,782 of the 12,649 JUPITER participants with consent and genotype had verified European ancestry Compliance limits sample to 6,934 SNPs excluded when failing Hardy-Weinberg equilibrium test at P < 10-6, with the exception of rs7412 at APOE (E2 v. E3) 820,411 SNPs pass QC with minor allele frequency > 1% 1N Engl J Med. 2008 359:2195.
Clinical characteristics of study sample(all European ancestry)
Defining LDL-C response to statin therapy Absolute LDL-C response: LDL-C at 12 months – LDL-C baseline Fractional (%) LDL-C response: LDL-C at 12 months – LDL-C baseline = absolute ΔLDL-C LDL-C baseline LDL-C baseline Statistical power: JUPITER sample with genome-wide genetic information is the largest to date with a single statin administered at a single dose
Genome-wide association of baseline LDL-C ~820K SNPs
Genome-wide association ofLDL-C lowering with rosuvastatin < Absolute LDL-C reduction Fractional LDL-C reduction >
Genome-wide association of LDL-C lowering with placebo < Absolute LDL-C reduction Fractional LDL-C reduction >
Magnitude of effects: best SNP at each locus LDL-C lowering # high LD Baseline LDL-C
Distribution of effect by genotype
Total genetic effect: proportion of variance explained at genome-wide loci “●” indicates locus with genome-wide association (p<5x10-8) For comparison, age, BMI, sex, smoking status, region explain: 3.5% of absolute LDL-C response 3.7% of fractional LDL-C response
Genes from genome-wide analysis PCSK9 (chr. 1) Serine protease with functions in LDLR protein degradation ABCG2 (chr. 4) Widely-expressed (hepatic, renal, elsewhere) transporter studied for multi-drug resistance phenotype in chemotherapy (as BCRP). Variation also associated with plasma urate levels. Effects observed in candidate analysis of LDL-C lowering with rosuvastatin*. LPA (chr. 6) Apolipoprotein(a) component of Lp(a). Plasma Lp(a) levels almost entirely determined by genetic variation at LPA. LDL-C includes contribution from cholesterol in Lp(a) particles. APOE (chr. 19) Major apolipoprotein component of VLDL, IDL, chylomicrons. *Circ Cardiovasc Genet. 2010 Jun 1;3(3):276-85.
Validation No replication, but … Genome-wide standard of significance (p<5x10-8) imposed All loci previously recognized in genetics of statin response literature Winner’s curse probably not a strong influence on effect estimates Associations not merely due to individuals with extreme LDL-C since such individuals were excluded by the trial design No effects at all in placebo
Sub-genome-wide significant loci(5x10-8<P<5x10-6)
IDOL (inducible degrader of LDL receptor)
IDOL (originally named MYLIP)
Sterol responsive ubiquitin-mediated pathway for post-transcriptional regulation (degradation?) of the LDL receptor1 Regulated by LXR Recently associated with baseline LDL-C2 Candidate therapeutic target for “statin-like” regulation of LDL-C levels mediated through the LDL receptor
Unknown function but shares band 4.1 homology with IDOL 1Science 2009 325:100-104 2PLoS Genetics 2009 5:e1000730. Nature 2010 466:707-13.
Candidate associations 1,2locus-wide best SNP for absolute (1) or fractional (2) LDL-C reduction No associations at GRIK4, CLMN, APOB, CYP3A5, CYP2C9
Interaction analysis No interaction among lead SNPs at genome-wide loci No interaction between lead SNPs and other SNPs across genome No interaction with sex No evidence for conditional associations within top loci However, evidence for PCSK9 X LDLR interaction with fractional LDL-C reduction (pint=0.002)
Influence of common genetic variation on rosuvastatin therapy in JUPITER temporal sequence of statin pharmacology CYP’s HMGCR APOB
Predicting LDL-C reductionGenetic score: sum of inherited “risk alleles” absolute LDL-C response fractional (%) LDL-C response
Effects of genetic score Estimates per unit of score, i.e. per inherited allele
Another candidate (KIF6) KIF6 gene non-synonymous variant (rs20455, MAF=34.7%). Minor allele (719Arg) has greater CV risk and greater response to atorvastatin (CARE, WOSCOP). No effect observed in JUPITER. See: Ridker et al. Circ Cardiovasc Genet. 2011 Apr 14. Lack of association may be related to differences between rosuvastatin and other statins
Summary In JUPITER, three (3) loci genome-wide significant association for LDL-C reduction with random rosuvastatin (20mg/dL) allocation: ABCG2, LPA, APOE An additional locus (PCSK9) for LDL-C reduction arises from genome-wide association with baseline LDL-C. Per allele, the lead SNPs are associated with a -5.2 mg/dL (ABCG2) and a +6.2 mg/dL (LPA) change in absolute LDL-C; a -5.1 mg/dL change in fractional LDL-C change (APOE) In total, 2.8% and 6.7% of the variance explained by four loci in absolute and fractional LDL-C reduction respectively A sub-genome-wide association at IDOL is consistent with current understanding of LDL receptor regulation Additional candidate analysis supports a role for variation in SLCO1B1 and LDLR A genetic risk score reveals dependence of median LDL-C response on genetics but only explains a small proportion of the variance No interaction effects with rosuvastatin observed for KIF6 variant
Collaborators and support BWH Paul M Ridker, MD, MPH Audrey Chu, PhD Franco Guilianini, PhD Jean MacFadyen, BS AstraZeneca Fredrik Nyberg, MD, PhD, MPH Bryan Barratt, PhD Support AstraZeneca Celera