Thrombospondin-1 Gene Polymorphism is Associated with Estimated Pulmonary Artery Pressure in Patients with Sickle Cell Anemia

To the Editor

Sickle cell disease (SCD) is known to promote end-organ damage over time. Patients with especially high rates of intravascular hemolysis in SCD have decreased nitric oxide (NO) bioavailability, which is associated with vasculopathy and pulmonary hypertension (PAH) in SCD, and increased morbidity and mortality [1, 2]. Thrombospondin-1 (TSP1), a multifunctional glycoprotein present in platelet alpha granules and an important contributor to blood cell adhesion and vaso-occlusion, also suppresses NO production through its cognate receptor CD47 [3, 4]. Additionally, TSP1 is highly expressed in the plasma and pulmonary vessels of PH patients, as well as in SCD patients during steady state and then rises further during an acute vasoocclusive episode [5-7].

Since TSP1 suppresses NO bioavailability, we hypothesized that specific genetic variants of this protein may be associated with variation in risk of PAH in sickle cell anemia (SCA) as measured by tricuspid regurgitant jet velocity (TRV), an echocardiographic marker of pulmonary artery pressure.

We explored this hypothesis utlizing a test cohort of SCA patients with available clinical phenotyping data from the screening phase of the multicenter walk-PHaSST trial (Treatment of Pulmonary Hypertension and Sickle Cell Disease with Sildenafil Therapy, ClinicalTrials.gov identifier NCT00492531)[8]. During the screening process of this trial, transthoracic Doppler echocardiography and peripheral blood DNA specimens were collected from patients with SCD at steady-state.

A total of 10 THBS1 single nucleotide polymorphisms (SNPs) (rs1478604, rs1478605, rs2228262, rs753599, rs17633107, rs2228261, rs1051442, rs11070220, rs2228263, rs3743125) including haplotype tagged SNPs, coding variants, and SNPs potentially regulating gene expression and alternative splicing were selected for analysis. These SNPs were genotyped for the participating subjects using Taqman SNP genotyping and a 7900 DNA analyzer (ABI, Foster City, CA).

A total of 406 SCA patients ranging in age from 12 to 70 years and 51% female were included in the test cohort. TRV was analyzed as a categorical variable using logistic regression analysis, and subjects were grouped based on the following clinically relevant cut-off points: 2.5 m/sec, 2.7 m/sec, and 3 m/sec. All SNPs were tested in dominant, recessive, and additive genetic models.

Among the 10 tested THBS1 SNPs, univariate regression analyses showed TRV ≥ 2.5 was independently negatively associated with both rs1478604 (minor allele frequencies (MAF) 0.291)) and rs1478605 (MAF 0.286) with OR 0.45 (95% CI 0.19, 1.08; p=0.069) and OR 0.33 (95% CI 0.12, 0.88; p=0.017), respectively, under a recessive model, although the association with rs1478604 did not reach statistical significance. We found an independent association of TRV with age, gender, plasma amino-terminal pro-brain natriuretic peptide (NT-proBNP) and serum creatinine, consistent with previously published results. We performed a sensitivity analysis for NT-proBNP, and found an increase in statistical significance in the model for rs1478605 when NT-proBNP was included (p=0.008 versus 0.023). Both rs1478604 and rs1478605 were significantly negatively associated with TRV ≥2.5 m/sec even after controlling for all of these variables (p=0.034 and p=0.008, respectively).

To seek validation, we performed similar analyses in an independent smaller cohort of 160 SCA patients enrolled in a prospective study at Duke University Medical Center, the University of North Carolina Health Care System, and Grady Health System at Emory University evaluating risk factors for decreased survival [9]. The cohort is similar in basic demographic and disease-specific characteristics to the walk-PHaSST cohort (Supplemental Table). Although the findings from this smaller validation cohort did not demonstrate statistical significance, encouragingly, the associations of these two SNPs with TRV trend in the same direction as our findings under a recessive model with OR 0.439 (CI 0.136, 1.42; p=0.169) and OR 0.645 (CI 0.196, 2.12; p=0.470), respectively, in an adjusted model controlling for age, gender and creatinine (Table I). Although adjustment for NT-proBNP had increased significance in the test cohort analysis, NT-proBNP was not available in this validation cohort.

Table I. Association of TRV with THBS1 SNPs rs1478604 and rs1478605^#.

SNP	Test Cohort (N=406)		Validation Cohort (N=160)
rs1478605	OR (95% CI)	P-value	OR (95% CI)	P-value
Unadjusted Model	0.33 (0.12, 0.88)	0.017	0.74 (0.30, 1.83)	0.5083
Adjusted Model*	0.26 (0.086, 0.76)	0.008	0.65 (0.20, 2.12)	0.470
rs1478604
Unadjusted Model	0.45 (0.19, 1.08)	0.069	0.43 (0.16, 1.18)	0.102
Adjusted Model*	0.37 (0.15, 0.95)	0.034	0.44 (0.14, 1.42)	0.169

^#Based on a recessive model.

^*Test cohort adjusted for age, gender, NT-proBNP, and creatinine. Validation cohort adjusted for age, gender, and creatinine.

Our results suggest that genetic variants in the THBS1 gene are associated with variation in pulmonary artery systolic pressure as estimated by echocardiography, a known predictor of PAH in patients with SCA. A recent case study found that TSP1 and CD47 expression were increased in the pulmonary vasculature and vascular cells cultured from the same in a SCD patient with PAH [10]. Additionally, mice lacking TSP1 and CD47 were protected from hypoxia-induced pulmonary vascular and right ventricular remodeling and showed fewer alterations in right ventricular systolic pressure compared to controls. These published results provide biological plausibility to our findings, as protective alterations in the THBS1 gene hypothetically contribute to lower expression of TSP1 in SCA patients.

Since both rs1478605 and rs1478604 are proximal to the transcription start site, they are potentially functional in transcriptional regulation of the THBS1 gene. Bioinformatics analysis for rs1478605 identified a putative allelic binding site for transcription factor CP2, which also regulates the alpha-globin gene. Although no differential allele-specific transcription factor binding site was identified for rs1478604, the close location to the transcription start site suggests a potential effect on THBS1 transcription. Alternatively, because the two SNPs are in high linkage disequilibrium (D′= 0.94), it is possible that rs1478605 may be the functional variant of these two SNPs that regulates THBS1 gene expression.

While the demographics of both cohorts are similar, it is possible that the results in the test cohort may be skewed by the enrollment of a higher percentage of patients with elevated TRV in the screening phase of the walk-PHaSST study than seen in the validation cohort. Additionally, NT-proBNP is clinically associated with PAH, and our sensitivity analysis shows that it does contribute to significance of our statistical model and primary outcome. The unavailibility of NT-proBNP data may be limiting the power to detect a statistical association of TRV and the THBS1 SNPs in the validation cohort.

Our genomic findings support a hypothetical model of TSP1 involvement in the pathophysiology of SCA, consistent with other published lines of evidence. Genetic variations in the promoter region may affect the transcription regulation of THBS1 gene and lead to altered TSP1 expression in SCA patients. Functional characterization of the identified promoter variants in THBS1 transcription regulation is warranted. Identification of protective genetic variants in THBS1 regulation may provide clinical translation for PAH risk stratification in SCA patients.