Genetic variants associated with white blood cell count among individuals with sickle cell disease

Abstract

Background:

Sickle cell disease (SCD) is a Mendelian disorder characterized by a point mutation in the β-globin gene that leads to sickling of erythrocytes. Several studies have shown that absolute neutrophil count is strongly associated with clinical severity of SCD, suggesting an apparent role of white blood cells (WBC) in SCD pathology. However, the mechanism by which genetic variants lead to white blood cells count differences in SCD patients remains unclear.

Methods:

Genome wide association (GWA) analyses were carried out among a cohort of 2,409 Brazil SCD participants. Association of white blood cells (WBC) count and genetic markers were investigated in homozygous sickle cell anemia (HbSS) participants and compound heterozygous sickle cell hemoglobin C (HbSC) participants.

Results:

GWA analysis showed that variants in genes TERT, ACKR1 and FAM3C are associated with WBC count variation. The well-studied association between WBC count and Duffy null phenotype (variant in ACKR1) in healthy populations was replicated, reinforcing the influence of the SNP rs2814778 (T>C) in WBC count.

Conclusion:

Genetics plays an important role in regulating WBC count in patients with SCD. Our results point to possible mechanisms involved in WBC count variation and as increased WBC count is associated with more severe SCD, these results could suggest potential therapeutic targets for individuals with SCD.

Graphical Abstract

Several studies have shown that absolute neutrophil count is strongly associated with clinical severity of SCD, suggesting an apparent role of white blood cells (WBC) in SCD pathology. This genome wide association (GWA) analyses, carried out among a cohort of 2,409 Brazil SCD participants, showed that variants in genes TERT, ACKR1 and FAM3C are associated with WBC count variation. The well-studied association between WBC count and Duffy null phenotype (variant in ACKR1) in healthy populations was replicated, reinforcing the influence of the SNP rs2814778 (T>C) in WBC count. Genetics plays an important role in regulating WBC count in patients with SCD and our results point to the possible mechanisms involved in WBC count variation and potential therapeutic targets for patients with SCD.

Introduction

Sickle cell disease (SCD) is caused by a variety of mutations in the β-globin gene, resulting in the production of sickle-shaped red blood cells (RBCs). Disease severity varies widely and is in part associated with individual genetics. For example, patients with sickle cell anemia (SCA) who carry two copies of hemoglobin S (HbS; SCA patients commonly referred to as HbSS patients) usually present more severe clinical phenotypes compared to patients who carry one copy of HbS and one copy of hemoglobin C (HbC; patients commonly referred to as HbSC patients).

Despite SCD being a genetic disorder that primarily leads to hemoglobin S (HbS) polymerization, which alters RBC morphology, white blood cells (WBC) also play a role in disease progression (reviewed in references (1-3)). Early observational studies have shown that high neutrophil count is associated with severity of clinical manifestations (4), elevated white blood cell count is positively correlated with early death (5), clinically overt strokes (6), silent cerebral infarction (7), and acute chest syndromes (8) in patients with SCD. The observation that sickle RBCs bind to neutrophils both in vitro (9) and in vivo in Berkeley mice (10,11) suggests that neutrophils may be directly involved in vascular occlusion. Recent studies have linked inflammatory cytokines and inflammatory responses to clinical complications including pulmonary hypertension (12) and leg ulcers, (13) suggesting another plausible mechanism for how WBC may be involved in SCD clinical manifestations.

Hydroxyurea has been widely used to treat severe clinical complications such as vaso-occlusive pain episodes and acute chest syndrome in patients with SCD. (14,15) One mechanism of action of hydroxyurea is through inducing fetal hemoglobin (HbF) production. (16,17) Interestingly, some patients who have no increase in HbF after hydroxyurea therapy experience clinical benefits (14) and most patients who respond well to treatment have a decrease in neutrophil count. (15) This suggests a plausible causal relationship between WBC count and SCD clinical complications. Studying genetic factors that affect WBC may have implications for optimizing hydroxyurea as achieving maximum tolerated dose may be challenging in patients with lower baseline WBC.

Several GWA studies in European, (18-23) Asian, (24-26),African or African American (21-23,27,28) and Hispanic populations, (29) have demonstrated the contribution of genetics to WBC count. Most noticeably, a GWA study that included more than 170,000 participants (from the UK Biobank study, (30) the UK BiLEVE study, (31) and the INTERVAL study (32)) identified 180 loci associated with total WBC count and 764 loci associated with at least one WBC index.(18) Arguably, one of the most successful stories on functional studies following GWA-identified variants is the one addressing the mechanisms of benign ethnic neutropenia (BEN), that identified the association between reduced WBC count and the Duffy antigen promoter variant rs2814778 (ACKR1: c.−67T>C). This variant is most prevalent among African populations (1000 genome African C allele frequency [AF] 0.964) and is rare among other populations (1000 genome European AF 0.006, American AF 0.08, and Asian AF 0). A functional study focusing on this variant explained the mechanism by which this variant leads to neutropenia. (33) Despite these successes, these studies enrolled mostly healthy participants who are not exposed to chronic baseline inflammation; therefore, our knowledge remains limited on how genetics are associated with WBC regulation under inflammatory exposure. We address specifically the inflammatory exposure of patients with SCD in this study. Several GWA studies have aimed to investigate alloimmunization susceptibility, (34,35) vaso-occlusive pain, (36) and fetal hemoglobin levels (37,38) among patients with SCD, but less is unknown about the genetic variation that contributes to WBC count. To identify genetic factors associated with WBC count in patients with an inflammatory pathophysiology such as SCD, we performed a GWA analysis in Brazilian individuals with HbSS and HbSC.

Methods

Study Population

The Brazil SCD cohort was established as part of the NHLBI-funded Recipient Epidemiology and Donor Evaluation Study-III (REDS-III). (40) In total, 2,793 participants enrolled into the study with SCD type diagnosed in the treating hemocenters and independently confirmed by pyrosequencing. HbSS was the most common genotype (1973 participants) followed by HbSC (644 participants), Sβ0 (83 participants), and Sβ+ (80 participants). (40)

Protection of human subjects was conducted according to federal regulations and protocols applicable to all human subject research and was supervised by NHLBI’s REDS-III data coordinating center (DCC, RTI International, Rockville, MD). All study procedures were approved by Brazilian Center’s IRB, Vitalant Research Institute, and RTI International.

Data collection and instruments

Clinical and laboratory data were collected according to standardized procedures previously described. (40) Clinical outcomes were defined according to standardized definitions of SCD phenotypic manifestations. (41) Steady-state WBC counts were measured for routine clinical care using automated analyzers. Steady state was defined as no hospitalizations, blood transfusions (for those not on chronic blood transfusion) or clinical complications that may affect the laboratory parameters in the one month prior to measurement of white blood cell count.

Data collected in the Recipient Epidemiology and Donor Evaluation Study (REDS-III) is available through the American National Institute of Health database of Genotypes and Phenotypes (dbGaP) upon request.

Statistical Analyses

Qualitative variables were shown as frequencies, and quantitative variables by mean (if it was normally distributed) or median (otherwise). The quantitative variable’s normality was assessed by Shapiro-Wilk test. The WBC count was transformed using log plus one function to reach the normality. WBC values outside mean ± 4 standard deviations were considered outliers and removed. The association between selected characteristics and the transformed WBC count were evaluated using a linear regression model. The R (42) version 4.1.2 was used to perform the analysis, considering significant p<0.05.

Genome-wide association analysis

Blood collection (40), sequence data processing, quality control and genotype calling were described elsewhere, (43) as whole-genome sequencing was accomplished by the Trans-Omics in Precision Medicine (TOPMed) program supported by the National Heart, Lung and Blood Institute (NHLBI). Of the total of 2617 study participants with HbSS and HbSC, 2409 had whole genome sequence and WBC data available after quality control procedures. Association analysis was conducted using ENCORE (University of Michigan), a web-based tool which allows the execution of GWA analysis of uploaded phenotypes against TOPMed genotypes. A linear mixed model was used, with adjustment for kinship and all previously identified covariates that were found to be associated with WBC counts in addition to the 10 first principal components to correct for population stratification.

Results

Overall Distribution of WBC Counts in the Study Population

There were 2793 participants included in the overall REDS-III Brazil SCD cohort (1973 HbSS, 644 HbSC, 83 HbSβ0-thalassemia, 80 HbSβ+-thalassemia and 13 with other rare compound hemoglobin mutations). Due to low numbers of HbSβ-thalassemia and other genotypes in the cohort, this analysis included only the HbSS and HbSC participants with available whole genome sequencing data. Both genotype groups were similar in age with median and mean ages of 15 and 20.7 years in the HbSC participants and 17 and 20.0 years in the HbSS participants, respectively. The distribution of sexes was also comparable in the two groups, with males accounting for 44.4% of HbSC participants and 45% of the HbSS participants. Regarding treatment, 7.3% of HbSC participants were treated with hydroxyurea, whereas 38.9% of HbSS participants were treated. Additionally, 4.97% of HbSC participants had received RBC transfusion in the last 12 months, compared to 37.6% of HbSS participants. The median WBC count of analyzed participants was 10,600 cells/μL (mean 10,978 cells/μL). Participants with HbSC had lower WBC (mean 8,835 cells/μL; median 8,510 cells/μL) than participants with HbSS (mean 11,685 cells/μL; median 11,343 cells/μL; p < 0.0001) (Table 1).

Table 1.

Demographic and clinical variables associated with WBC count. WBC count is shown as cells/μL. P-value was calculated using age as a numerical, not categorical variable.

Characteristic	WBC count (median)	P-value
SCD genotype		<0.0001
SC	8,510
SS	11,343.33
Sex		0.017
female	10,800
male	10,400
Age		<0.0001
0 - 4	13,692
5 - 9	11,533
10 - 17	10,731
18 -29	10,685
30 - 39	9,800
40 - 49	8,220
50 - 59	8,308
60 +	7,318
Enrollment center
Hemominas HBH	9,430	<0.0001
Hemominas JFO	10,200	0.29
Hemominas MOC	11,266.67	0.006
Hemope	11,140	<0.0001
Hemorio	11,293.33	<0.0001
ITACI SP	9,983.33	0.24
Recent hydroxyurea use (yes)	9,877.92	<0.0001
Recent transfusion	11,903.33	<0.0001
Ever transfused	10,905	<0.0001
Number of transfusions		<0.0001
Number of hospitalized pains		<0.0001

WBC Counts Associated with Age, SCD genotype, Hydroxyurea Use, Hemocenter, and Transfusion

Overall distribution of WBC count results among participants with HbSS and HbSC was right skewed as shown in Figure 1a. Log plus one transformation brought the data to approximately normal distribution (Figure 1b). WBC count was associated with age (p < 0.0001), enrollment center (p < 0.0001), SCD genotype (p < 0.0001), recent hydroxyurea (p < 0.0001) use and recent transfusion history (p < 0.0001) (Table 1).

Figure 1.

Distribution of leukocyte results. a. Distribution of raw WBC counts in HbSS (blue) and HbSC (red) participants. The QQ plot insert shows the distribution of WBC count data (x-axis) as compared to quantile of normal distribution (y-axis), blue line is a 45-degree line that goes through origin. b. Distribution of log plus one transformed WBC counts in HbSS (blue) and HbSC (red) participants. QQ plot shows that log plus one transformation effectively brings leukocyte results in approximately normal distribution.

Multivariate regression was used to model association between WBC counts and demographic and clinical variables. WBC count was negatively associated with age, with only slight decrease as age increases (Table 2). Current use of hydroxyurea was associated with significant decrease of WBC counts (p < 0.0001). The effect size of hydroxyurea use was −0.104, suggesting that the mean log transformed WBC count result was 0.104 lower in participants who recently used hydroxyurea compared to those who did not. The raw WBC count was approximately 14% lower in the recent hydroxyurea use group. The effect of age and hydroxyurea use were similar among participants with HbSS and HbSC as shown in the stratified models (Table 2).

Table 2.

Multivariate modeling the effect of age, sex, hemo centers, hydroxyurea use, transfusion, and hospitalized pain episodes on leukocyte results. Three statistical models were conducted, one jointly analyzing HbSS and HbSC participants, one restricted to HbSS only participants, and one restricted to HbSC only participants. Shown in the table are effect sizes of each variable on the log transformed leukocyte results after adjusting for other covariates in the model (Beta column), the equivalent percent difference after reverse transformation (Pct. Diff. column), and the corresponding p-values (P column).

	Combined model		HbSS participants		HbSC participants
	Beta	P	Beta	P	Beta	P
Intercept	5.136		5.140		4.916
SCD type	−0.136	<0.0001
Age	−0.003	<0.0001	−0.003	<0.0001	−0.001	<0.0001
Sex (female)	−0.004	0.051	−0.006	0.353	0.002	0.177
Hemominas JFO	0.019	0.084	0.022	0.065	0.013	0.512
Hemominas MOC	0.033	0.001	0.045	0.0001	0.010	0.390
Hemope	0.052	<0.0001	0.067	<0.0001	−0.056	0.082
Hemorio	0.043	<0.0001	0.055	<0.0001	0.003	0.095
ITACI SP	−0.0079	0.716	−0.008	0.721	0.005	0.148
Recent hydroxyurea use	−0.104	<0.0001	−0.105	<0.0001	−0.103	<0.0001
Transfused within 12 months	−0.030	<0.0001	−0.032	<0.0001	0.039	0.388
Number of hospitalized pains	−0.00004	0.888	−0.00004	0.873	0.0016	0.255

Transfusion history was also associated with increased WBC counts (Table 1). This may be because patients requiring more frequent transfusions are more severe and the elevated WBC reflects that. Participant transfusion history was coded into three variables: (a) whether a participant had ever received blood transfusion (ever transfused), (b) number of historical transfusions received by a participant (number of transfusions), and (c) whether a participant received a blood transfusion within 12 months prior to enrollment (recent transfusion). We conducted additional statistical modeling to evaluate which of these variables better models the effect to WBC count. Although all three ways of coding transfusion history showed association with WBC counts, recent transfusions were most correlated with elevated WBC counts. Ever transfused and number of historical transfusions were not significantly associated with WBC count when adjusted for recent transfusions (Table 1).

Number of pain episodes hospitalizations were associated with elevated WBC counts when modeled in a univariate model (Table 1) but were not statistically significant when adjusted for transfusion history (Table 2). Therefore, recent transfusion history was used as a covariate in the genetic models.

Genome Wide Association of WBC Counts

Genetic markers associated with WBC count were interrogated using genome wide association analyses, adjusted for the identified demographic and clinical covariates (age, sex, SCD genotype, enrollment center, recent hydroxyurea use, and recent transfusion history). Results are shown in Figure 2. There were 4 genome wide significant hits (P < 5 x 10⁻⁸).

Figure 2.

Manhattan plot and QQ plot for GWA analyses restricted to HbSS participants.

Top variant rs772442796 (p-value 1.14 x 10⁻⁸, beta −2.18) , is a C deletion with minor allele frequency 0.1038% in the study population (0.0219% in Topmed, 0% in African and 0.02% in European ALFA population (39)) located in chromosome 5. This variant is in the intronic region of gene telomerase reverse transcriptase (TERT) (Figure 3). The presence of one copy of the rs772442796 variant was linked to an approximate reduction of 2.18 in the log plus one wbc count.

Figure 3.

LocusZoom single variant association regional association plot for GWA top hit on chromosome 5.

Variant rs2814778 (p-value 1.16 x 10⁻⁸, beta −0.16) in chromosome 1 also reached genomic wide significance. One copy of the rs2814778 variant was associated with an estimated decrease in log plus one wbc count by 0.16. rs2814778 (c.−67T>C) is in the promoter region of atypical chemokine receptor 1 (ACKR1) gene (Figure 4). The C allele of this variant is most prevalent in the African population (1000 genome African allele frequency 0.964, European allele frequency 0.006) and is associated with approximately 20% decrease of WBC count in healthy individuals. The Brazilian population is an admixture population that exhibits both C and T alleles with modest frequency for rs2814778 (616 C/C, 1,290 C/T, and 838 T/T in the study population).

Figure 4.

LocusZoom single variant association regional association plot for GWA top hit on chromosome 1

Two other SNPs reached genome wide significance, SNPs rs151288099 (p-value 2.24 x 10⁻⁸, beta −0.28) and rs60492718 (p-value 3.14 x 10⁻⁸, beta 0.27) in gene FAM3 metabolism regulating signaling molecule C (FAM3C) in chromosome 7 (Figure 5). 8 other SNPs in FAM3C had P-values close to genome wide significance (P < 8 x 10⁻⁸), adding to evidence of association between the gene and WBC. The presence of one copy of the rs151288099 variant was associated with an estimated decrease in log plus one wbc count by 0.28 and the presence of one copy of the rs60492718 variant was associated with an estimated increase in log plus one wbc count by 0.27.

Figure 5.

LocusZoom single variant association regional association plot for GWA top hit on chromosome 7

In order to test if the SNPs significantly associated with WBC were only associated with WBC as a surrogate marker of disease severity, we tested the association of the SNPs with other markers of disease severity such as 3 or main pain crisis hospitalizations per year, stroke, priapism and acute chest syndrome. However, none of the SNPs associated with WBC were associated with these phenotypes

Discussion

The mutation that causes sickle cell disease only directly affects RBCs, however, it is known that WBCs contribute to the pathophysiology . Sickle cell severity varies significantly from patient to patient and WBC counts are associated with important SCD complications (4-8,12,13), vaso-occlusion (7,9) and mortality (4,5), Therefore, in an effort to help elucidate genetic factors influencing WBC count, and consequently disease severity, a GWAS within the REDS-III Brazil SCD cohort was performed.

The observed median WBC count (10,600 cells/μL with mean 10,978 cells/μL) was congruent with the observation of elevated WBC in SCD patients (5) (normal range 3,500 to 10,500 cells/μL). WBC count was associated with age, recent hydroxyurea use and recent transfusion history. In the genetic analysis, 4 SNPs were found to be associated with WBC count in this cohort.

The GWAS top hit was in TERT, a telomerase protein subunit that adds TTAGGG repeats to chromosome ends with the help of a template coded by gene TERC. (40) Variations in TERT were previously associated with telomere length in TOPMED data. (41) Telomeres are regions of repetitive DNA sequence in the end of chromosomes.

Although TERT has been associated with telomere length, it has also been associated with the development of bone marrow failure. (42,43) The association may be related to the occurrence of short telomeres in patients with aplastic anemia. (44,45) In aplastic anemia, the bone marrow consists mainly of fat and very few hematopoietic cells.(43)

Previous GWAS also found associations between polymorphisms in TERT and platelet (18,46,47), granulocyte (48), neutrophil (46,47) and eosinophil (49) counts. GWAS also found associations between SNPs in TERT and neutrophil percentage of leukocytes, (18) myeloid white cell counts, (18) eosinophil percentage of leukocytes, (49) monocyte percentage of white cells, (49) lymphocyte percentage of leukocytes (49) and myeloproliferative disorder. (50)

A known association of WBC and SNP rs2814778 was replicated in this study. rs2814778 is known to be associated with WBC count and has been replicated across multiple previous GWA studies. (18,20,28,29,51,52) This variant (c.−67T>C) is in the FY*B promoter region of ACKR1 gene responsible for a common phenotype Fy(a-b-) present most in West Africans and rare in Caucasians. (53,54) ACKR1 is also known as the Duffy Antigen Receptor for Chemokines, or DARC. Chemokines attract specific subpopulations of WBCs to inflammation sites. (55) In individuals with HbSS, variant in rs2814778 was associated with end-organ damage (56) and Duffy expression was also associated with WBC count. (56)

Two SNPs in FAM3C reached genome wide significance and 8 other SNPs had P-values close to genome significance (P-values < 8 x 10⁻⁸). Polymorphisms in FAM3C were previously associated with bone mineral density (BMD) and fractures. (50,53,54,56) FAM3C is hypothesized to have roles in epithelial-to-mesenchymal transition (EMT), tumor growth, metastasis, osteoblast differentiation and bone homeostasis. (57-59) In a study with FAM3C knockout mouses, although no differences were found in female mouses, in male knockout mouses the percentage of neutrophils out of WBCs was significantly increased and the percentage of lymphocytes out of WBCs reduced compared to wild-type male mouses. (59) Additionally, a study found an association between high BMD loss increased neutrophil, decreased monocyte and decreased lymphocyte counts. (60)

Conclusion

In this study, we investigated genetic factors influencing WBC count variation in patients with SCD. We replicated a well-known mutation associated with WBC count in healthy people and found two additional genes. These findings show the importance of genetic factors and help elucidate pathways and genes that contribute to WBC count variation and that might influence SCD severity.

Further studies are required to reveal the specific mechanisms involved in WBC count variation, but if confirmed these pathways and genes may be candidate targets for furture SCD treatments..

Figure 6.

Frequency of variants that reached genomic significance.

data availability statement

Data collected in the Recipient Epidemiology and Donor Evaluation Study (REDS-III) is available through the American National Institute of Health database of Genotypes and Phenotypes (dbGaP) upon request.