Evidence for HLA-related susceptibility for stroke in children with sickle cell disease

Stroke is one of the most devastating complications in children with sickle cell anemia (SCA). Approximately 8% to 10% of children with SCA will suffer a symptomatic stroke, while 17% more will have evidence of asymptomatic cerebral infarction with magnetic resonance imaging (MRI).1-3 Even these latter, clinically occult lesions correlate with significant neuropsychological deficits.4,5 The majority of strokes in children with SCA (symptomatic or asymptomatic) are the result of infarction, with intracranial hemorrhage becoming relatively more common later in life.6 The most common location of infarction is in the distribution of the anterior portion of the circle of Willis.7,8 Arteriography or magnetic resonance angiography often demonstrates progressive narrowing of these vessels with collateral vessel development in a pattern similar to moyamoya disease.9,10 Histopathologically, sickle cell vascular lesions show a pattern of smooth muscle proliferation with overlying endothelial damage.7,11 The abnormal adherence of sickle red blood cells to the endothelium, as well as the altered rheology of sickle red blood cells, undoubtedly contributes to the endothelial injury in these vessels.12-15 Why every child with SCA does not develop these histologic changes is unknown and suggests the contribution of other environmental and genetic factors.

Anecdotally, it has been noted that stroke often occurs in siblings with SCA,18-20 and an increased risk of stroke in siblings with SCA was recently suggested by a sib-pair study.17Studies involving candidate genes known to influence coagulation have failed to show an association with cerebral infarction in SCA.18-20 Because particular HLA types have been associated with other vascular diseases characterized by endothelial changes,21-25 we examined whether specific alleles in the HLA region were associated with stroke in 53 children with SCA.

There were 53 patients with SCA from Children's Hospital Oakland in the study. Twenty-two patients demonstrated MRI-documented evidence of cerebral infarction and had a mean age of 12.9 ± 5.6 years (median, 13 years) at the time of the study. Only patients with evidence of cerebral infarction were included in the “stroke” arm of the study. Of these patients, 17 had a history of clinically overt stroke, while the remaining 5 had asymptomatic cerebral infarction. The mean age at the time of stroke in the clinically symptomatic patients was 7.2 ± 3.5 years (median, 6 years). The control group consisted of 31 SCA patients who had a negative MRI scan and no history of clinical stroke, with a mean age of 14.3 ± 7.5 years (median, 13 years). The control group patients, by design, were older in order to reduce the possibility of including in the control group a patient who would subsequently develop a stroke. Patient identities were blinded and study numbers assigned to all collected samples. To maintain anonymity, only MRI status, history of overt stroke, and patient age were linked with individual HLA results. HLA results in the 2 sickle cell disease patient groups were then compared. The entire sample of SCA patients was compared with previously published HLA results from an African American population to confirm that the SCA patient population was representative of the general African American population.26 The study was reviewed and approved by the Institutional Review Board of Children's Hospital Oakland.

All SCA patients underwent MRI scanning at Magnetic Imaging Affiliates (Alta Imaging Medical Group, Oakland, CA) on a GE 1.5 Tesla Magnet clinical imaging system (Siemens, Erlangen, Germany) with the use of Cigna MRI with 8.2 software. Standard techniques used included spin-echo and fast spin-echo imaging, a T1-weighted pulse sequence, a T2-weighted axial pulse sequence, and an intermediate coronal pulse sequence. Contrast agents were not used in any of the MRI examinations.

MRI scanning was performed in these patients for 1 of 3 reasons: suspicion of cerebral infarction, frequent headaches, or as part of enrollment in other clinical studies. MRI scans were read by experienced neuroradiologists who were unaware of the study. MRI scanning was classified as positive if any area of focal infarction was identified. Patients with atrophy, hemorrhage without infarction, and other noninfarctive lesions were excluded from the study.

Genomic DNA was extracted from 100 μL whole blood by means of the QIAGEN QIAamp 96 Spin Kit. DNA samples were typed at the HLA class II loci (DRB1, DQB1, and DPB1) with the use of the polymerase chain reaction, to amplify a locus-specific second-exon product, and analyzed with the use of sequence-specific oligonucleotide probes in a dot blot format as previously described.32,33Class II haplotypes (DRB1-DQB1-DPB1) were inferred from linkage disequilibrium patterns.26,30,31 Samples were typed for HLA Class I A and B loci at an allelic level with the use of immobilized probe methods as previously described32,33 with modifications (personal communications, Drs R. Apple, T. Bugawan, and H. Erlich, Roche Molecular Systems, Alameda, CA). Class I typing methodology modifications from published methods included amplifying exons 2 and 3 separately in a multiplex reaction, and hybridization to modified and additional probes for exons 2 and 3 on our more optimized immobilized probe strips. For allelic level resolution, some sample types required subsequent dot blot probe hybridization using horseradish peroxidase–labeled exon 2 or 3 probes, and/or group-specific amplifications followed by immobilized probe analysis to distinguish them. Computer software programs were used to identify allelic types from the probe hybridization patterns. HLA alleles were classified according to nomenclature defined by the World Health Organization.34 

Differences in HLA allele distributions between SCA patients with and without MRI-documented evidence of cerebral infarction were examined by means of row-by-column testing for independence using the log likelihood ratio or G test.35 This approach assumes that both alleles are statistically independent but avoids the multiple testing problem inherent in examining antigen frequencies. To deal with the problem of small expected values in contingency table testing of the highly polymorphic HLA loci, a Monte Carlo procedure was applied. The Monte Carlo procedure was used to determine the validity of the χ² distribution in testing a given table without unnecessary binning of rare alleles.36 For each locus tested, 10 000 replicate tables, based on the observed marginal sums of a table, were generated. Row sums were binned into the “combined” class only if the individual row sum was less than 5 for either the case or the control group. When row sums were binned as necessary to achieve at least 5 entries, then the G values for the overall row-by-column test were close to limiting values of the χ² distribution at the 0.10, 0.05, 0.01, and 0.001 quintiles. This resampling procedure is likely to be conservative because any disease effect contributed by a rare allele will be hidden in the combined class. The magnitude of significant individual allele effects for a locus showing overall significance were measured with the odds ratio based on Fisher's exact test, calculated with probabilities.

In 31 of the SCA patients, brain MRI results revealed no evidence of infarction. Of the 22 SCA patients with a positive MRI, 17 (77%) had a history of a clinically overt stroke. Of these 17 patients, 14 (82%) had evidence of large-vessel disease on MRI scans. The remaining 3 had watershed lesions. Five of the 22 patients (23%) with a positive MRI had a clinically silent stroke. Of these 5 patients, 3 (60%) had evidence of large-vessel disease, and the remaining 2 had watershed lesions.

All 53 SCA individuals were typed at high (allelic) resolution for the HLA loci A, B, DRB1, DQB1, and DPB1. Allele frequencies from published control data from African Americans for the HLA class II loci DRB1, DQB1, and DPB1 were compared with the entire sample of SCA patients in order to validate the HLA composition of the study population as a random sample of African Americans. G tests of independence between the SCA and an outside control group were nonsignificant for each of the 3 loci, suggesting that our sample of pediatric patients, preselected for hemoglobin S homozygosity, did not differ from African Americans generally at the HLA loci.

Allele frequency distributions for the MRI-positive and MRI-negative groups were compared for each of the 5 HLA loci. Of the class I loci, HLA-A was not significant (P = .86) (Table1), but the HLA-B locus frequencies did differ between the 2 groups (P = .012) (Table2). Two alleles were responsible for the difference between the groups. B*5301 was in apparent excess (odds ratio [OR] = 3.7; confidence interval [CI], 1.13-12.34;P = .038), and B*4501 was in apparent deficit in the stroke sample (OR = 0.08; CI, 0.01-1.5; P = .04).

The overall test of independence for the distributions of alleles at the class II loci was significant for DRB1 (P = .0008) (Table3) and DQB1 (P = .029) (Table4) but not for DPB1 (P = .82) (Table 5). For DRB1, the DR3 alleles, DRB1*0301 and *0302, appeared to be associated with susceptibility to stroke (OR = 5.8; CI, 1.6-20.8; P = .007) (Table 3). The DR15 alleles, DRB1*1501 and *1503, were protective for stroke (OR = 0.21; CI, 0.06-0.72; P = .019). For the DQB1 locus (Table 5), DQB1*0201 was associated with stroke (OR = 3.04; CI, 1.21-7.68; P = .033) while DQB1*0602 appeared protective (OR = 0.25; CI, 0.09-0.70, P = .011).

Subclassification of MRI-positive patients by large- versus small-vessel involvement and overt versus silent stroke did not localize the observed HLA associations to any particular MRI findings.

HLA associations with autoimmune and infectious diseases have been studied for more than 30 years and have provided insight into the pathophysiology of many diseases.37,38 The HLA system serves as a master regulator of the immune system by directing the processing and presentation of foreign antigens to T lymphocytes. While most diseases associated with HLA genes have a recognized immune component, many do not. We present evidence that particular HLA alleles are associated with the presence of cerebral infarction in SCA. SCA is a well-characterized genetic disorder that invariably results from the substitution of valine for glutamic acid at the sixth position of the β globin molecule. The clinical manifestations of this disease are heterogenous, and few predictors for disease severity have been identified. The association of HLA type with stroke provides the first evidence of a multi-genic involvement in a specific manifestation of SCA. The idea that other genes are likely to affect the clinical course of sickle cell patients is not new, but identification of genes and evidence of their involvement have been difficult. High fetal hemoglobin levels are known to ameliorate the severity of sickle cell disease, and determination of the factors that influence fetal hemoglobin expression has been extensively studied. β globin haplotype and the F-cell production locus on the X chromosome have been associated with different patterns of fetal hemoglobin expression in patients with sickle cell disease.37 38 

Considerable attempts have been made to relate stroke risk in SCA to other genetic factors. Studies investigating mutations conferring an increased risk of thrombosis in the general population, including those involving Factor V Leiden, homocysteine, and prothrombin 20210, have failed to show an association with stroke in SCA patients.18-20 Coinheritance of alpha thalassemia was initially reported to decrease risk of stroke in patients with SCA,41,42 but subsequent studies in SCA patients with clinically overt and silent lesions have been unable to show a protective effect of coinherited alpha thalassemia on multivariate analysis.3 43 In these same studies, a low hemoglobin and an increased white blood cell count correlate with an increased risk of infarctive stroke. Strikingly, a higher fetal hemoglobin level, a protective factor in nearly every other manifestation of SCA, does not appear to correlate with stroke risk. The pathophysiology of stroke, therefore, may be uniquely different from that in other manifestations of SCA. A multi-institutional study evaluating the combination of HLA and other clinical features of SCA may lead to the identification of high- and low-risk populations of children with SCA.

Documentation of an association between HLA alleles and risk of stroke in patients with SCA suggests that the HLA system plays a role in the pathophysiology of vascular changes leading to stroke. Despite the strength of the observed association, this study does not preclude the possibility that HLA serves as a marker for another gene in linkage disequilibrium, operating to confer susceptibility to or protection from stroke. For example, the major gene for hemochromatosis, HFE, originally associated with HLA-A3, actually lies 4 megabases telomeric to the HLA-A locus but is nonetheless in strong linkage disequilibrium with A3.44 Larger, confirmatory studies are needed to definitively associate the HLA system, or other genes in linkage disequilibrium with HLA alleles, with stroke risk in SCA. The high prevalence of cerebral infarction in SCA patients suggests that the “risk” gene is unlikely to be a “rare” mutation, but rather a susceptibility gene that predisposes an individual to develop stroke. DR3 and DQ2, the allelic groups most significantly associated with stroke in SCA, are also commonly associated with other diseases.37,38 Interestingly, the DR3 haplotype in this population does not carry the allele B*0801, excluding the involvement of the B8-DR3 haplotype so common in European origin populations.31 Also, an increased risk of stroke has been documented with HLA-B*5301. B*5301 has been reported to confer protection from severe malaria in The Gambia in a similar fashion to the sickle gene mutation.45 

HLA alleles may influence the risk of stroke in SCA, but how the immune system is involved in the pathophysiology of sickle cell cerebrovascular lesions is unknown. Immune system involvement in other diseases characterized by endothelial injury and smooth muscle cell proliferation is already well established.46-48 Links between the HLA system and diseases with prominent vascular components, such as atherosclerosis,21,22 pulmonary hypertension,23,24 and moyamoya disease, have been described.25 In Japan, moyamoya disease is often seen in members of the same family, and the development of moyamoya disease is associated with the HLA allele DQB1*0502.25Histopathologically, the vascular changes of moyamoya disease are virtually indistinguishable from those in the cerebral vasculopathy of SCA. The vascular lesions of moyamoya disease show infiltration by T cells and macrophages.48 Particular HLA types have been associated with early onset of atherosclerosis, and T cells and monocytes are prominently involved in the initiation of the atherosclerotic lesion.21 22 

Potential hypotheses regarding the role played by HLA genes in the development of stroke require an explanation of how HLA predisposes to stroke in the context of SCA. Sickle cell disease may generate a stimulus that enables a predisposing HLA allele to propagate the pathophysiological process that results in smooth muscle cell proliferation. In atherosclerosis, the response-to-injury hypothesis holds that a primary injury to the endothelium initiates formation of the atherosclerotic lesion.49 In response to this initial injury, endothelial cells produce chemotactic factors that recruit T cells and monocytes into the subendothelium where they secrete mitogenic factors leading to the proliferation of smooth muscle cells.46 Sickle cell disease could conceivably contribute to cerebral vascular changes by providing a source of continual endothelial injury via the effects of increased shear stress, hypoxia, increased thrombin generation, and the abnormal adherence of sickle red blood cells to the endothelium.13,50 51 

Ongoing endothelial injury in SCA is now well established in vivo with the documentation of activated, circulating endothelial cells in sickle cell patients.14,52,53 In vitro, exposure of vascular endothelium to sickle red cells results in up-regulation in leukocyte adhesion molecule expression and NFκB transcription,54providing a means for the recruitment of T cells or antigen-presenting cells into the arterial wall. In addition, sickle cell–mediated endothelial injury could also augment the response of professional antigen-presenting cells or stimulate the expression of HLA class II molecules on the endothelium. Endothelial cells are not professional antigen-presenting cells, but can be induced to express class II HLA molecules when exposed to a variety of cytokines and injurious agents.55 

Ultimately, the significance of the findings presented here lies in their application to the care of the patient with SCA. The cerebrovascular lesions in SCA occur very early in life, with the highest incidence of symptomatic stroke in 2- to 5-year-olds.43 Once a stroke has been diagnosed, the risk of recurrence is over 70%.6,16,56,57 Chronic transfusion and bone marrow transplantation are effective in preventing further strokes,16,57-59 but the child with a stroke is left with physical and neuropsychological deficits that have a substantial impact on quality of life. Because of the lifelong morbidity from primary stroke, there has been considerable effort to identify patients at risk for stroke and intervene before the stroke occurs. Transcranial Doppler ultrasound is one diagnostic modality that has been shown to be useful in identifying patients at risk for symptomatic stroke,60but many of these patients already have evidence of infarction on MRI scanning.61 Thus, the use of transcranial Doppler to detect patients at risk for stroke may be too late in many cases. Our identification of an association in the HLA region with stroke risk could affect the care of SCA patients substantially as SCA patients at risk for stroke could be identified at birth or even prenatally. Identification of such patients would allow the preventive use of effective therapies, such as chronic transfusion or bone marrow transplantation. Also, elucidation of the genes involved in the cerebral vasculopathy of SCA may suggest other preventive therapies with less morbidity than chronic transfusion and bone marrow transplantation.

We are indebted to Henry Erlich for critical scientific review of this manuscript.