Abstract
Phenotypic heterogeneity is a well known characteristic of sickle cell anemia. Patients have different rates of hemolysis-related complications, like pulmonary hypertension, priapism and leg ulceration, and viscosity/vasoocclusion-related complications, like painful episodes, acute chest syndrome and osteonecrosis; they also have variation in levels of HbF and hematocrit. To integrate individual disease variables into a global measure of severity, we developed a Bayesian network model that described the complex associations of 25 clinical and laboratory variables, deriving a score that we used to define disease severity (0, least severe to 1, most severe) as the risk of death within 5 years (Sebastiani et al, Blood 2007). This initial network, validated in 2 unrelated patient populations, did not incorporate the genetic heterogeneity that is likely to modulate its components. Accordingly, we studied the association of single nucleotide polymorphisms (964 SNPs) in candidate genes (315 genes) using a Bayesian beta regression model of the severity score in 741 HBB glu6val homozygotes, aged more than 18 years. Forty-three SNPs in about 25 genes were associated with disease severity. Some associated SNPs tag genes that affect nitric oxide and oxidative biology and the endothelium, such as NOS1, ASS, KL, HMOX1, ECE1, KDR, FLT1. Homozygosity for an intronic SNP in ECE1 is associated with a increase of severity (OR=3.5). As expected, some associations were consistent with our previous findings. For example, the same SNP in ECE1 and TGFBR3, that was highly predictive of severity, was also strongly associated with sickle cell stroke (Sebastiani et al, Nature Genet 2005). Also, the association with severity of genes in the TGF-beta signaling pathway, including BMP6 and TGFBR3, were also associated with individual disease complications. Other associated genes play a less obvious role in the pathobiology of disease, e.g., HAO2, but are very strongly associated with the phenotype of severity (probability of a chance association, for HAO2, 10−6). Several of the genes associated with severity, including KL, PRKCA, FLT1 and MET have been related to aging, as suggested by gene expression profiling and studies in model organisms for aging. In genome-wide studies of the genetic basis of exceptional longevity, we found associations with some of the same genes that were associated with severity in sickle cell anemia. Perhaps increased oxidative stress, and the relentless progression of vasculopathy in sickle cell anemia, cause accelerated tissue damage that is modulated by a set of genes similar to those involved in the normal aging process. We suggest that the disease severity score can be used as a phenotype integrating many features of the disease, for genetic association studies. As we add the results of unbiased genome-wide association studies to capture polymorphisms not included in candidate gene studies, we can develop a predictive network with even greater reliability than one using only clinical and laboratory variables. Such networks might also identify pathways that could be targeted to alter the course of disease.
A network model to predict the risk of death in sickle cell disease
