nces 263 (2007) 113–117www.elsevier.com/locate/jns

Journal of the Neurological Scie

PDE4D and ALOX5AP genetic variants and risk for IschemicCerebrovascular Disease in Sweden

Konstantinos Kostulas a,⁎, Solveig Gretarsdottir b, Vasilios Kostulas a, Andrei Manolescu b,Anna Helgadottir b, Gudmar Thorleifsson b, Larus J. Gudmundsson b, Unnur Thorsteinsdottir b,

Jeffrey R. Gulcher b, Kari Stefansson b, Jan Hillert a

a Department of Neurology, Neuro-angiological Research Center, Karolinska Institutet, Karolinska University Hospital Huddinge, S-141 86 Huddinge, Swedenb Department of deCODE Genetics, Sturlugata 8, Reykjavik, Iceland

Received 6 November 2006; received in revised form 6 June 2007; accepted 20 June 2007Available online 25 July 2007

Abstract

Background: Genetic variants in Phosphodiesterase 4D (PDE4D) and 5-lipoxygenase activating protein (ALOX5AP) have been shown toconfer risk of Ischemic Cerebrovascular Disease (ICVD) in Iceland. We investigated whether these variants associate with ICVD in Sweden.Methods: Previously published PDE4D and ALOX5AP gene variants were genotyped for cases (685) and controls (751). In PDE4D thisconsisted of SNP41, SNP45 and microsatellite AC008818-1 and in ALOX5AP four SNPs that define the HapA haplotype.Results: The PDE4D SNPs, showed a non-significant risk in the ICVD group which increased for the Large Artery Atherosclerosis subtype(SNP45: RR=1.43, P=0.063, SNP41: RR=1.57, P=0.018). The SNP haplotype GA (SNP45, SNP41) showed an increased risk for LAA(RR=1.58, P=0.016) and the combined LAA and Cardioembolism (CE) (RR=1.34, P=0.031) subgroups. As the SNPs are in strong LD,this haplotype corresponds to the complement of the protective haplotype in the Icelandic study. No allele of the microsatellite marker,showed association to stroke or any subtype and nor did the Icelandic PDE4D at-risk haplotype (GA0). We did not confirm the associationbetween ALOX5AP HapA haplotype and ICVD, but a non-significant risk was observed in the LAA subtype.Conclusion: Our PDE4D findings although non-significant considering the number of markers and phenotypes tested, are consistent with theassociation observed in the original study, with a trend observed in the whole ICVD group, which was strengthened in the stroke subtypeLAA and the combined group of LAA and CE stroke. This supports the notion that PDE4D contributes to the risk of developing stroke.© 2007 Elsevier B.V. All rights reserved.

Keywords: PDE4D; ALOX5AP; Ischemic stroke; Genetics

1. Introduction

A number of different pathogenetic events may lead toischemic cerebrovascular disease (ICVD), including athero-sclerosis, small vessel disease, cardiac arrhythmias, and hyper-coagulation. In turn, these conditions are also dependent on avariety of different underlying conditions some of which areinherited and others are due to exogenous factors. Thus, ICVDis a typical multifactorial disease and several genetic factorsmay contribute to the risk of the disease.

⁎ Corresponding author. Tel.: +46 8 585 87027; fax: +46 8 774 4822.E-mail address: konstantinos.kostulas@karolinska.se (K. Kostulas).

0022-510X/$ - see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.jns.2007.06.042

A large group of candidate genes have been studied inICVD. The results have varied, possibly due to factors suchas differences in inclusion criteria, definition of stroke andsubtypes, age, ethnic origin and number of patients.

Genetic variations in two genes have recently beenshown to associate to common forms of stroke in Iceland,the Phosphodiesterase 4D gene (PDE4D) and 5-Lipoxygen-ase activating protein (ALOX5AP) [1,2]. Both these geneswere identified through genome-wide linkage analysis andsubsequent case–control association studies in an Icelandicpopulation. Two haplotypes in PDE4D, spanning the firstexon of isoform PD4D7, were found to confer risk and pro-tection, respectively, primarily in a subgroup of cardiogenic

Table 1Evaluation and clinical data (%) of Swedish ICVD patients

All ICVD(n=685)

Stroke(n=524)

TIA(n=161)

Background parameters:Age 69.0

(±11.3)69.7(±11.4)

66.9(±10.9)

Current smoking 25 24 27Ischemic heart disease 27 28 25Myocardial infarction 13 13 10Atrial fibrillation 20 23 10Hypertension a 52 52 55Hyperlipidemia b 21 18 30Diabetes mellitus c 19 20 16Intermittent claudication 6 6 6Cholesterol N7.0 mmol/L d 8 7 13Triglycerides N2.3 mmol/Ld 8 8 10

Evaluation data:CT performed 97 98 94MRI performed 16 18 9Duplex ultrasound 79 75 91ECG 97 97 96Echocardiography 38 36 45a If patients (a) had a history of hypertension or (b) no history of

hypertension but were being treated for hypertension.b If patients were on lipid-lowering medication.c If patients (a) had a history of diabetes or (c) no history but were being

treated for diabetes.d Lipid levels were measured fasting.

Table 2TOAST subtyping, % (n), of Swedish ICVD cohort

All ICVD(n=685)

Stroke(n=524)

TIA(n=161)

Large artery atherosclerosis 21 (144) 16 (85) 37 (59)Cardio-embolic 23 (161) 26 (137) 15 (24)Small arterial occlusion 10 (69) 13 (69) 0 (0)Other aetiology 5 (27) 3 (17) 6 (10)Multiple causes 4 (31) 5 (29) 1 (2)Unknown, complete workup 10 (72) 8 (43) 18 (29)Unknown, incomplete workup 26 (181) 27 (144) 23 (37)

TOAST = Trial of Org 10172 in Acute Stroke Treatment.ICVD = Ischemic cerebrovascular disease.

114 K. Kostulas et al. / Journal of the Neurological Sciences 263 (2007) 113–117

(i.e. cardioembolic) and carotid stroke (i.e. large artery athero-sclerosis) patients. Likewise a haplotype spanning ALOX5AP,HapA, was shown to confer risk of both myocardial infarction(MI) and stroke in Iceland [2].

The aim of this study was to determine whether the variantsin PDE4D and ALOX5AP that associates with stroke inIceland also confer risk of ischemic cerebrovascular disease inSwedish ICVD patients. For PDE4D we tested the at-risk andprotective variants (spanning PDE4D7) and for ALOX5AP theHapA haplotype.

2. Materials and methods

2.1. Ischemic Cerebrovascular Disease patients

A total of 685 patients (mean age±SD: 68.3±11.3 years,56% men) with ischemic stroke or TIA (collectively labelledICVD) attending the stroke unit or the stroke outpatient clinicat Karolinska University Hospital, Huddinge unit in Stock-holm, Sweden were recruited from 1996 to 2002 as part of anongoing genetic epidemiology study, the South StockholmIschemic Stroke Study (SSISS). All patients had clinicallyrelevant investigations performed, including brain imagingwith computed tomography (CT) or/and magnetic resonanceimaging (MRI) as well as ancillary diagnostic investigationsincluding duplex ultrasonography of the carotid and vertebralarteries, echocardiography, MR-angiography, CT-angiogra-phy and standardized blood tests. Clinical data and commoninvestigations are listed in Table 1.

Patients were classified into ICVD subtypes according tothe Trial of Org 10172 in Acute Stroke Treatment (TOAST)classification by a physician reviewing original imaging anddata [3]. The TOAST subtypes include a) Large ArteryAtherosclerosis (LAA), b) Cardioembolism (CE), c) SmallArterial Occlusion (SAO) i.e. lacunar infarction, d) Otheraetiologies (OA) and e) Unknown cause (including multiplepotential causes). Frequencies of TOAST subtypes in ourcohort are shown in Table 2.

The study was approved by the Bioethics Committee ofKarolinska Institutet (Dnr 286/96 and 08/02).

2.2. Controls

The controls used in this study are population-based con-trols recruited from the same region in central Sweden as thepatients, representing the general population in this area.In total 751 Swedish individuals (mean age±SD: 46.8±14.9 years, 59% men) were recruited, either blood donors(recruited in 2001) or healthy volunteers (collected in 1990–1994) recruited by the clinical chemistry department at theKarolinska University Hospital to represent a normal refer-ence population. Information for gender and age at samplingwas available for the controls.

2.3. PDE4D and ALOX5AP genotyping

Markers defining the previously published PDE4Dand ALOX5AP variants were genotyped for both cases andcontrols. In PDE4D this consisted of two SNPs, SNP41(rs12153798) and SNP45 (rs12188950) and one microsatel-lite marker, AC008818-1. In ALOX5AP the four SNPs defin-ing the HapA are: SNPs SG13S25 (rs17222814), SG13S114(rs17222772), SG13S89 (rs17216522), and SG13S32(rs17222828). The genotype distribution of all SNPs wasfound to be in Hardy–Weinberg equilibrium both in cases andcontrols.

2.4. Statistics

For single-marker association and haplotype analysis, weused the program NEMO [1], which handles missing geno-types through a likelihood procedure, using the expectation-

Table3

SNPandhaplotypeassociationof

theIcelandicat-riskandprotectiv

ePDE4D

haplotypein

SwedishICVD

andTOASTsubtypes

Alleles

Controlsn=75

1ICVD

n=68

5LAA

n=14

4CEn=16

1LAA/CEEn=30

5SAO

n=69

Frq

Frq

RR

CI

PFrq

RR

CI

PFrq

RR

CI

PFrq

RR

CI

PFrq

RR

CI

P

SNP45

G0.844

0.85

21.07

0.87–1.31

0.540

0.88

51.43

0.98

–2.08

0.063

0.855

1.09

0.78

–1.53

0.620

0.869

1.23

0.94–1.61

0.132

0.819

0.84

0.53–1.33

0.452

SNP41

A0.830

0.85

11.16

0.94–1.41

0.180

0.88

41.57

1.08

–2.27

0.018

0.851

1.17

0.84

–1.64

0.359

0.867

1.33

1.02–1.74

0.038

0.823

0.96

0.60–1.52

0.848

At-risk

GA

0.829

0.849

1.15

0.94–1.40

0.180

0.88

41.58

1.09

–2.30

0.016

0.851

1.18

0.85

–1.66

0.327

0.867

1.34

1.03–1.76

0.031

0.816

0.92

0.58–1.46

0.723

At-risk

GA

00.300

0.290

0.95

0.81–1.13

0.550

0.29

50.98

0.74

–1.29

0.881

0.314

1.07

0.82

–1.39

0.610

0.305

1.02

0.83–1.26

0.817

0.266

0.85

0.58–1.26

0.421

Protective

AG

0.154

0.14

60.94

0.76–1.15

0.530

0.115

0.71

0.49

–1.03

0.072

0.145

0.93

0.66

–1.30

0.663

0.131

0.82

0.63–1.08

0.154

0.174

1.15

0.72–1.84

0.555

Protective

AG

−8

0.150

0.14

50.96

0.78–1.18

0.680

0.111

0.71

0.48

–1.03

0.074

0.145

0.96

0.68

–1.35

0.799

0.129

0.84

0.63–1.10

0.197

0.174

1.19

0.74–1.90

0.474

Sho

wnaretheallelic

frequenciesof

themajor

allelesof

SNP4

5andSNP4

1andhaplotypefrequenciesof

theat-riskandprotectiv

ehaplotypes

(Frq),theRelativeRisk(RR),ConfidenceIntervals(CI),and

thetwo-

sidedP-value

(P)for

theassociationwith

stroke

andits

subphenotypes.ICVD=Ischem

icstroke

orTIA

,LAA=largeartery

atherosclerosis,CE=cardioem

bolism,S

AO=sm

allarterialocclusion

.Genotyp

eyieldfor

themarkers

wereas

follo

ws(m

arkername:

genotyped#controls/#

patientswith

ICVD);SNP4

1:42

8/565,

SNP45

:70

1/67

3,AC0088

181-1:

751/68

5.

115K. Kostulas et al. / Journal of the Neurological Sciences 263 (2007) 113–117

maximization algorithm as a computational tool to estimatehaplotype frequencies. Under the null hypothesis, the af-fected individuals and controls were assumed to have iden-tical haplotype frequencies. Under the alternative hypotheses,the candidate at-risk haplotype was allowed to have a higherfrequency in the affected individuals than in controls, andthe ratios of frequencies of all other haplotypes were assumedto be the same in both groups. Likelihoods were maximizedseparately under both hypotheses, and a corresponding1-degree-of-freedom likelihood ratio statistic was used toevaluate statistical significance. The same method is alsoused when we test the markers individually for association inorder to ensure that the same data/individuals were used togenerate results for all the tests, allowing meaningful com-parisons among the different markers/haplotypes. As some ofthemarkers did not have high genotype yield, genotypes fromother markers in the same gene were used to provide somepartial information on the missing genotypes [4]. Instead oftesting an allele A of amarker individually for association, wetest all haplotypes that include allele A and other nearbymarkers as a group for association to the disease. This methodis especially useful for the markers in PDE4D, which are invery strong linkage disequilibrium.

The relative risk of a haplotype is calculated using themultiplicative model, in which the risk corresponding to eachof the two chromosomes multiplies. All P-values presentedare two-sided. Allelic rather than carrier frequencies for mi-crosatellites, SNPs and haplotypes are presented. To adjustfor the number of subgroups/haplotypes tested we use simu-lations. In each simulation we randomly assign individuals tothe study cohorts, preserving the number of individuals ineach sub-group, and repeat the all the association tests. Thisprocedure was repeated 1000 times and an adjusted P-valuecalculated as the number of simulations that gave a P-value,for any of the haplotype/marker tested, equal to or lower thanthe P-value for the original cohort.

3. Results

3.1. The Swedish ICVD population

The ICVD patients conformed to the Swedish NationalStroke Registry in frequencies of the registered risk factors(atrial fibrillation, diabetes mellitus, hypertension and smok-ing) for Karolinska University Hospital in Huddinge [5].Frequencies of TOAST-subtypes were similar to recently pub-lished cohorts [6].

3.2. Association to the PDE4D genetic variants

We first tested the association of SNP45, SNP41 andAC008818-1 to Swedish ICVD and to the different TOASTsubtypes as was done in the Icelandic study (Table 3 andSupplementary Table 1). The population allele frequenciesof the SNPs were similar in Sweden and Iceland. Howevera difference was observed in the frequency of two alleles

Table 4Linkage disequilibrium calculations of microsatellite AC008818-1 andSNP41 and SNP45 in PDE4D

R2 D′ Allele (marker)

Iceland 0.873 0.952 −8 (AC008818-1) G (SNP41)Sweden 0.804 0.912Iceland 0.914 0.982 −8 (AC008818-1) A (SNP45)Sweden 0.863 0.955Iceland 0.921 0.967 A (SNP41) G (SNP45Sweden 0.870 0.990

R2 correlation coefficient, D′ linkage disequilibrium coefficient.

116 K. Kostulas et al. / Journal of the Neurological Sciences 263 (2007) 113–117

of microsatellite AC008818-1, (alleles −8 and 0), both ofwhich showed significant association in the Icelandic cohort(Supplementary Table 1). While none of the five alleles ofAC008818-1 showed significant association to stroke in theSwedish cohort, both common alleles of the SNPs showed aslight increased risk for the whole ICVD patient group andconsistent with the Icelandic results the risk increased for theLAA subtype (RR=1.43, P=0.063, 0.55, and RR=1.57,P=0.018, Padj=0.24, for SNP45 and SNP41 respectively,Icelandic data is presented in Supplementary Table 2). HerePadj is a P-value adjusted for the number of haplotypesand sub-groups tested for both genes using simulation. Thetwo SNPs, and the −8 allele of AC008818-1, are in stronglinkage disequilibrium (LD) (pairwise R2N0.8) although theLD is stronger in the Icelandic cohort (Table 4); hence theyare expected to show similar results.

We then tested for association the predefined protectiveand the at-risk haplotypes (Table 3). The Icelandic at-riskhaplotype is primarily defined by allele 0 of AC008818-1; thisallele does not show any association in the Swedish cohort,neither to ICVD nor any of the subgroups. Consistent with thiswe do not observe association in the Swedish cohort to the at-risk haplotype GA0. In Iceland the protective haplotype couldbe defined by the minor alleles of either of the SNPs (A and Gfor SNP45 and SNP41 respectively) or by the −8 allele ofAC008818-1. When we tested the haplotype, AG-8, the RR inSwedish ICVD is 0.96 and the strongest protective effect isobserved in the LAA group, RR=0.71 (P=0.074, Padj=0.62),Table 4. These results, although non-significant, are consistentwith the association observed in the Icelandic cohort to theprotective haplotype, both with respect to the direction ofthe effect and the stroke subgroup that shows the strongestassociation, although the magnitude of the effect is less thanis observed in the Icelandic cohort (RR=0.71 compared to

Table 5ALOX5AP HapA association in Swedish ICVD and TOAST subtypes

Controls n=709 ICVD n=677 LAA n=145

Frq Freq RR CI P Freq RR CI

Hap A 0.142 0.142 1.00 0.80–1.26 0.98 0.181 1.34 0.94–1

Shown is the haplotype frequency of HapA (Frq) , the relative risk (RR), Confidencits subphenotypes. ICVD = Ischemic stroke or TIA, LAA = large artery atherosclerfor the markers were as follows (marker name: genotyped # controls/# patients wSG13S32: 604/653.

RR=0.57 in Iceland, see Supplementary Table 2 for the Ice-landic results). As mentioned above, we do not observe sig-nificant association to the Icelandic at-risk haplotype (GA0),however, by excluding the microsatellite the two SNPhaplotype, GA, showed increased risk in the LAA and LAA/CE subgroups (Table 3), P=0.016 (Padj=0.21) and P=0.031(Padj=0.35), respectively. This haplotype is close to being thecomplement of the protective haplotype in Sweden as thesetwo SNPs are in very strong LD (R2=0.92).

3.3. Association to the ALOX5AP genetic variants

The ALOX5AP HapA haplotype confers risk of bothMI and stroke in Iceland [2] and has been associated withstroke in Scotland [7]. We tested HapA in our Swedish strokegroup (Table 5). No association was observed to ischemicstroke in general but an increased risk was observed in theLAA group, although non-significant (RR=1.34, P=0.11,Padj=0.75, Table 4).

3.4. Discussion

Our PDE4D findings for the protective variant, althoughnot significant after adjusting for the haplotypes/sub-groupstested, are consistent with the original study, where a trendwas observed in the whole ICVD group, which was strength-ened in the subtype of stroke most closely related to athero-sclerosis i.e. LAA and in the combined group of LAA andCE. No association was observed to SAO also consistent withthe Icelandic results. The at-risk PDE4D variant observed inthe Icelandic study was not associated with stroke in theSwedish cohort; however, the at-risk variant is primarilydefined by the microsatellite AC008818-1 and, unlike inIceland, no allele of the microsatellite was associated withstroke in Sweden.

The ALOX5AP variant, HapA, did not associate to ICVDin Sweden. However, a non-significant risk was seen to theLAA subtype.

Several PDE4D and ALOX5AP studies have been pub-lished recently [7–20]. We have compiled the results for thestudies that have tested the two PDE4D SNPs, SNP45 andSNP41, and the HapA ALOX5AP haplotype; and presentthem together with our results in Supplementary Table 3.

A possible limitation of this study is the matching ofthe control and the patient groups. The controls are on aver-age 20 years younger than the patients. This age difference

CE n=158 SAO n=68

P Freq RR CI P Freq RR CI P

.93 0.11 0.128 0.89 0.64–1.83 0.54 0.148 1.08 0.64–1.83 0.78

e Intervals (CI) and the two-sided P-value for the association with stroke andosis, CE = cardioembolism, SAO = small arterial occlusion. Genotype yieldith ICVD); SG13S25: 669/665, SG13S114: 661/661, SG13S89: 435/657,

117K. Kostulas et al. / Journal of the Neurological Sciences 263 (2007) 113–117

between the controls and the patients would, however, onlyaffect the association analysis if the variants tested hadsignificant age dependence in the general population. Thishas been studied in the Icelandic population where largenumber of individuals has been typed for both the PDE4Dand ALOX5AP variants, but no age or gender difference hasbeen observed (unpublished work). To minimise bias due topotential population structure both the controls and thepatients have been recruited from the same geographicalregion within Sweden. Furthermore, the same populationcontrols as used in this study have also been used as part of theNordic Multiple Sclerosis cohort, collected in Stockholm. Inthat study, approximately 60 markers, unlinked to the diseasephenotype and distributed across the genome, were typed in300 patients and controls. With those markers no populationstratification was detected in the cohort (unpublished results).

In conclusion, although we observe a weaker effect in aSwedish cohort than originally reported in Iceland, our worksupports the notion that genetic variants ofPDE4D contributeto the risk of developing ischemic stroke through its effect onstroke on the background of large artery atherosclerosis.However, it is worth pointing out that although the effect isweak the at-risk SNP variant is very common hence this cor-responds to a substantial population attributed risk (e.g. 39%for LAA/CE).

Acknowledgements

We thank the participants who made this study possible;Thorbjörg Jonsdottir for excellent technical assistance andpersonnel at deCODE’s core facilities. This study was sup-ported by grants from the Swedish Medical Research Council(project numbers 11023 and 11220), The 1987 Foundation forStroke Research and Karolinska Institutet.

Appendix A. Supplementary data

Supplementary data associated with this article can befound, in the online version, at doi:10.1016/j.jns.2007.06.042.

References

[1] Gretarsdottir S, Thorleifsson G, Reynisdottir ST, Manolescu A,Jonsdottir S, Jonsdottir T, Gudmundsdottir T, et al. The gene encodingphosphodiesterase 4D confers risk of ischemic stroke. Nat Genet Oct2003;35(2):131–8.

[2] Helgadottir A, Manolescu A, Thorleifsson G, Gretarsdottir S, JonsdottirH, Thorsteinsdottir U, et al. The gene encoding 5-lipoxygenaseactivating protein confers risk of myocardial infarction and stroke. NatGenet Mar 2004;36(3):233–9.

[3] Adams Jr HP, Bendixen BH, Kappelle LJ, Biller J, Love BB, GordonDL, et al. Classification of subtype of acute ischemic stroke: definitionsfor use in a multicenter clinical trial. Stroke 1993;24:35–41.

[4] Amundadottir LT, Sulem P, Gudmundsson J, Helgason A, Baker A,Agnarsson BA, et al. A common variant associated with prostatecancer in European and African populations. Nat Genet May 2006;7.

[5] Asplund K, et al, for the Riks-Stroke Collaboration. Riks-Stroke—aSwedish national quality register for stroke care. Cerebrovasc Dis2003;15:5–7 (Suppl).

[6] Lovett JK, Coull AJ, Rothwell PM. Early risk of recurrence by subtypeof ischemic stroke in population-based incidence studies. NeurologyFeb 24 2004;62(4):569–73.

[7] Helgadottir A, Gretarsdottir S, St Clair D, Manolescu A, Cheung J,Thorleifsson G, et al. Association between the gene encoding 5-lipoxygenase-activating protein and stroke replicated in a Scottishpopulation. Am J Hum Genet Mar 2005;76(3):505–9.

[8] Lohmussaar E, Gschwendtner A, Mueller JC, Org T, Wichmann E,Hamann G, et al. ALOX5AP gene and the PDE4D gene in a centralEuropean population of stroke patients. Stroke Apr 2005;36(4):731–6.

[9] Bevan S, Porteous L, Sitzer M, Markus HS. Phosphodiesterase 4Dgene, ischemic stroke, and asymptomatic carotid atherosclerosis.Stroke May 2005;36(5):949–53.

[10] Nilsson-Ardnor S, Wiklund PG, Lindgren P, Nilsson AK, Janunger T,Escher SA, et al. Linkage of ischemic stroke to the PDE4D region on5q in a Swedish population. Stroke Aug 2005;36(8):1666–71.

[11] Meschia JF, Brott TG, Brown Jr RD, Crook R, Worrall BB, Kissela B,et al. Phosphodiesterase 4D and 5-lipoxygenase activating protein inischemic stroke. Ann Neurol Sep 2005;58(3):351–61.

[12] Saleheen D, Bukhari S, Haider SR, Nazir A, Khanum S, Shafqat S, et al.Association of phosphodiesterase 4D gene with ischemic stroke in aPakistani population. Stroke Oct 2005;36(10):2275–7.

[13] van Rijn MJ, Slooter AJ, Schut AF, Isaacs A, Aulchenko YS, SnijdersPJ, et al. Familial aggregation, the PDE4D gene, and ischemic stroke ina genetically isolated population. Neurology Sep 14 2005.

[14] Nakayama T, Asai S, Sato N, Soma M. Genotype and haplotypeassociation study of the STRK1 region on 5q12 among Japanese: acase–control study. Stroke Jan 2006;37(1):69–76.

[15] Woo D, Kaushal R, Kissela B, Sekar P, Wolujewicz M, Pal P, et al.Association of Phosphodiesterase 4D with ischemic stroke: apopulation-based case–control study. Stroke Feb 2006;37(2):371–6.

[16] Kuhlenbaumer G, Berger K, Huge A, Lange E, Kessler C, John U, et al.Evaluation of single nucleotide polymorphisms in the phosphodiesterase4D gene (PDE4D) and their association with ischaemic stroke in a largeGerman cohort. J Neurol Neurosurg Psychiatry Apr 2006;77(4):521–4.

[17] Brophy VH, Ro SK, Rhees BK, Lui LY, Lee JM, Umblas N, et al.Association of phosphodiesterase 4D polymorphisms with ischemicstroke in a US population stratified by hypertension status. StrokeMay 4 2006.

[18] Zee RY, Brophy VH, Cheng S, Hegener HH, Erlich HA, Ridker PM.Polymorphisms of the phosphodiesterase 4D, cAMP-specific (PDE4D)gene and risk of ischemic stroke: a prospective, nested case–controlevaluation. Stroke Aug 2006;37(8):2012–7.

[19] Staton JM, Sayer MS, Hankey GJ, Attia J, Thakkinstian A, Yi Q, et al.Association between phosphodiesterase 4D gene and ischaemic stroke.J Neurol Neurosurg Psychiatry Sep 2006;77(9):1067–9.

[20] SongQ, Cole JW,O'Connell JR, StineOC,GallagherM,GilesWH, et al.Phosphodiesterase 4D polymorphisms and the risk of cerebral infarctionin a biracial population: the Stroke Prevention in Young Women Study.Hum Mol Genet Aug 15 2006;15(16):2468–78.