S100B has pleiotropic effects on vaso-occlusive manifestations in sickle cell disease

To the Editor,

Acute pain crisis is the dominant manifestation of vaso-occlusion in sickle cell disease (SCD)¹ and the major cause of morbidity. Frequent vaso-occlusive crises associate with progressive organ damage and bone infarction known as avascular necrosis (AVN). The molecular etiologies underlying the development of vaso-occlusive complications are not well understood. In this study, we assessed the genetic factors contributing to acute pain crisis and AVN, applying a prioritized analysis focusing on expression quantitative trait loci (eQTL) in SCD. The study involved 1162 patients from two cross-sectional cohorts and an additional 303 patients from two expression profiling cohorts. We found allelic heterogeneity in eQTL of S100B, encoding a damage-associated molecular pattern molecule.² The A alleles of the tag single nucleotide polymorphisms (SNPs) of the eQTL, rs2154586 and rs2070435, both associated with higher S100B expression. The A allele of the former SNP associated with AVN whereas the A allele of the latter SNP with reduced rather than increased acute pain episodes. Serum S100B concentrations also correlated independently with AVN and with reduced acute pain episodes.

Phenotype definition:

We examined pain episodes for which patients went to the ER or were admitted to hospital over 12 months prior to a baseline visit, in 650 Walk-PHaSST adolescent and adult patients and 484 PUSH pediatric patients (Supplemental Table 1). To reflect the frequency of crisis naturally based on calendar year, quarter, and month while accommodating the distribution of count data, numbers of severe pain episodes in the past year were converted into four ordered groups defined as 1) no episodes, 2) 1-4 episodes, 3) 5-12 episodes, and 4) >12 episodes, herein referred to as pain frequency. Pain frequency had a strong association with history of acute chest syndrome, as expected,³ and showed a consistent trend of association with symptomatic AVN at hip or shoulder confirmed by radiography (Supplemental Figure 1). Older age associated with increased pain frequency (OR=1.5, P=6.5×10⁻⁵) and with AVN (OR=4.3, P=1.8×10⁻⁶). Indeed, 28 out of the 31 patients with AVN in the pediatric PUSH cohort were 12 years or older.

eQTL mapping:

Genetic regulation of gene expression plays an important role in complex traits, as indicated by the enrichment of trait-associated SNPs in eQTL.^4,5 Such enrichment exhibits tissue-, disease-, and population-specificity.⁶ Gene expression in peripheral blood mononuclear cells (PBMCs) was markedly different between sickle cell anemia patients and healthy African Americans,⁷ suggesting its relevance in disease etiology. We mapped eQTL in a cohort of 61 adult patients from Howard University and the University of Chicago who were microarray-profiled for PBMC gene expression and genome-wide SNPs (Supplemental Methods).⁷ At a false discovery rate of 5%, 1004 independent SNPs, obtained by linkage disequilibrium (LD) clumping per gene expression trait (r² >0.3), herein referred to as expression quantitative trait nucleotides (eQTNs), were identified for 880 genes.

Allelic heterogeneity in eQTL of S100B underlies its pleiotropic effects on vaso-occlusive complications:

Among the eQTNs, 865 for 763 genes with minor allele frequency >0.1 and imputation quality r² >0.9 in Walk-PHaSST and PUSH were further analyzed. As the eQTL mapping was carried out genome-wide, the prioritized analysis represents a de novo approach for hypothesis generating. Pain frequency and AVN were regressed on allele dosage by the ordinal logistic model and the binary logistic model, respectively, adjusting for covariates of age, gender, hemoglobin β genotype severity, and population stratification. The A allele of an eQTN of S100B (rs2154586), located in its last intron, associated with higher S100B expression (β=0.40, P=1.6×10⁻⁶) (Figure 1A) and with AVN in meta-analysis (OR=2.0, nominal P=1.4×10⁻⁵, Bonferroni-corrected P=0.012) (Supplemental Table 2, Figure 1B). The result is consistent with an inflammation promoting effect of S100B at localized high dose.⁸

Figure 1.

Pleiotropic effects of S100B expression on AVN and pain frequency. (A) The relative expression levels of S100B are plotted according to the A allele dosage of the two eQTNs for the Howard expression cohort. (B) The number of non AVN and AVN patients in Walk-PHaSST and PUSH are plotted according to the A allele dosage of rs2154586. (C) The number of patients in each pain group in Walk-PHaSST and PUSH are plotted according to the A allele dosage of rs2070435. In B and C the relative proportion of patients is shown above each bar. (D, E) The −log₁₀ P-values of regional association with S100B expression in the Howard expression cohort (upper panels) and with AVN (lower panel D) and pain frequency (lower panel E) in Walk-PHaSST and PUSH are shown, along with recombination rate. In D and E, the color of points depicts the strength of LD between regional SNPs and the two eQTNs; the point bordered by yellow diamond denotes rs2154586 and by yellow square denotes rs2070435. Relative exon positions and transcriptional directions of S100B and DIP2A are depicted as vertical bars and arrows, respectively, at the bottom of the Figure.

There was no eQTN reaching a Bonferroni-corrected P=0.05 for pain frequency association. The two eQTNs having the most significant associations with pain frequency were an eQTN of DIP2A (rs4347941, C allele OR=0.71, nominal P=1.6×10⁻⁴) and another eQTN of S100B (rs2070435, A allele OR=0.73, nominal P=5.7×10⁻⁴) (Supplemental Table 2, Figure 1C). These two eQTNs were in LD (r²=0.92, D’=0.99). As gene expression levels of S100B and the nearby DIP2A were correlated (ρ=0.63), this genetic locus, located downstream of S100B and within DIP2A, appears to regulate both genes. The A allele of rs2070435 not only associated with higher S100B expression (β=0.34, P=3.1×10⁻⁷) (Figure 1A) but also with higher DIP2A expression (β=0.14, P=2.6×10⁻⁹). The association with reduced nociceptive pain⁹ may be related to the functions of S100B and DIP2A in stimulating growth of peripheral nerve axons¹⁰ and reducing neuronal apoptosis after ischemia.¹¹

In regional associations (Figure 1D, 1E), the SNPs with the greatest significance for AVN (rs376461336, OR=2.1, nominal P=2.2×10⁻⁶) and pain frequency (rs2096507, OR=0.69, nominal P=4.5×10⁻⁵) were in LD with rs2154586 (r²=0.84, D’=1) and rs2070435 (r²=0.83, D’=0.98), respectively. The two eQTNs for S100B, located in two consecutive LD blocks (Supplemental Figure 2), individually explained 35% (rs2154586) and 39% (rs2070435) and jointly explained 57% of S100B expression variation. The largely non-overlapping effects of the eQTNs on S100B expression reflect only weak LD (r²=0.043, D’=0.42) and imply that the two SNPs tag distinct regulatory sites, which may underlie the pleiotropic effects of S100B expression on AVN and pain frequency.

Serum S100B correlated with AVN and reduced acute pain episodes:

We used data of 242 adult patients from the University of Illinois at Chicago (UIC) cohort (Supplemental Table 1) to replicate the findings. The two S100B eQTNs were validated for PBMC expression association in 159 patients with data available (β=0.50, P=4.8×10⁻¹⁸ for rs2154586; β=0.35, P=1.2 ×10⁻¹³ for rs2070435). AVN in UIC was determined by radiography regardless of the presence of symptoms; pain episodes were defined as episodes for which patients went to acute care center or the ER or were admitted to hospital in 1 year. AVN and pain episodes were analyzed by the binary logistic model and negative binomial model, respectively, adjusting for the same covariates as in the discovery phase. Our initial replication effort did not show statistically significant association of rs2154586 with AVN (OR=1.1, P=0.7) or rs2070435 with pain episodes (e^β=0.85, P=0 .2) (Supplemental Table 3), probably due to the small sample size, discrepancy in outcome definition, and variation in local LD structure. Further studies, involving substantially more patients, will be necessary.

We further examined serum S100B protein level by enzyme-linked immunosorbent assay, from 51 UIC patients with variable S100B expression (the top and bottom quartiles of expression level) and with available serum samples. Serum S100B concentration strongly correlated with array gene expression (ρ=0.70, P=3.6×10⁻⁸) (Supplemental Figure 3A). Higher level of serum S100B showed a stronger association with the A allele of rs2154586 (the AVN associated eQTN, β=0.71, P=6.6×10⁻⁵) than with the A allele of rs2070435 (the pain associated eQTN, β=0.39, P=0.013). Similar patterns were observed for their linked SNPs (Supplemental Table 3). Furthermore, higher serum S100B concentration correlated independently with AVN (β=0.68, P=0.018) and with fewer acute pain episodes (β=−0.041, P=0.0082) (Supplemental Figure 3B, 3C) in multivariate analysis.

S100B eQTL associated with venous thromboembolism in general African American population:

To assess the effects of the S100B eQTNs under non-sickle cell conditions, we examined their associations with venous thromboembolism (VTE) in general African Americans. VTE shares certain pathogenetic features with sickle vaso-occlusion.^12,13 Under non-sickle conditions, VTE prevalence is higher in patients with idiopathic osteonecrosis than in the general population.¹⁴ We analyzed VTE, defined as having a blood clot in leg or lung as told by a doctor, among African Americans from Atherosclerosis Risk in Communities, Jackson Heart Study, and Cardiovascular Health Study in dbGaP. The A allele of rs2154586, which associated with AVN in SCD, associated with VTE in the general African Americans (OR=1.4, P=0.0067) (Supplemental Table 4).