UHRF2 as a genetic correlate of hospitalization in sickle cell disease

Sickle cell disease (SCD), or sickle cell anemia, is the most prevalent genetic disease and represents a significant health burden worldwide. In 2021, an estimated 7.74 million people worldwide were living with SCD.1 In Brazil, the birth prevalence is 1 in 1000, resulting in about 3500 new cases annually. Around 45 000 people are living with SCD in Brazil, representing a significant public health challenge. Hydroxyurea (hydroxycarbamide) and chronic transfusion therapy (CTT) are the main disease-modifying treatments.2-5 Haematopoietic stem cell transplant6, 7 and gene therapy8, 9 may be curative but are not yet widely available. SCD is characterized by haemolysis, chronic inflammation and vaso-occlusive events, which are associated with multiorgan damage. Although most of the children will survive into adulthood, relative to children without SCD, childhood mortality is high.1 Up to 5% of children with SCD may die due to complications such as acute chest syndrome (ACS), sepsis, splenic sequestration, stroke and aplastic crisis.10, 11 Life expectancy in adulthood is also reduced. Adults with SCD often experience chronic pain and organ damage, with end-stage organ failure being a potential cause of death. In Brazil, individuals with SCD died at a median age of 32.0 years, which is 37 years younger than the median age of death in the general population (69.0 years).12 In our large Brazilian cohort of individuals with SCD—part of the Recipient Epidemiology and Donor Evaluation Study-III (REDS-III) international programme—hospitalization was independently associated with an increased risk of death.13 The phenotypic expression of SCD varies widely. Modifiers of SCD severity have been identified, mainly the β-globin genotype (such as HbSS, HbSC or others), the co-inheritance of α-thalassaemia trait and the amount of HbF production.14 However, the clinical manifestations of SCD vary even within the same β-globin genotype group receiving identical disease-modifying treatments, highlighting the potential role of other genetic factors underlying disease complications. Although pain is commonly used as a proxy for SCD severity, it represents only part of the disease burden. Given the range of other clinical complications, hospitalization is common and associated with significant disease mortality. The rate of hospital readmissions, which is a quality-of-care indicator, is also significantly higher among patients with SCD in comparison with other patient groups.15 Disease-modifying treatments, which reduce the morbimortality and the occurrence of the most important clinical complications of SCD, are known to prevent hospitalizations and to reduce the length of hospital stay in SCD.16, 17 Given the above, the frequency of hospitalizations in the SCD patient population can be considered an interesting marker of disease severity. Our objective was to identify the genetic markers associated with SCD disease severity using hospitalizations as a marker of severity. To this end, a Genome-wide association study (GWAS) was conducted in a large, well- Brazilian cohort. Participants were drawn from the REDS-III Brazil SCD cohort, part of the National Heart, Lung, and Blood Institute (NHLBI)'s REDS-III. A total of 2793 individuals with SCD were enrolled. Clinical, demographic and laboratory data were collected following as previously described.18 Homozygous SS and HbSβ0-thalassaemia individuals were selected for the analysis and all-cause frequent hospitalization as a marker of SCD disease severity based on the diverse phenotypes leading to hospitalization. Data on hospitalizations were collected during study enrolment and two consecutive follow-up visits (2013–2015). Hospitalization was defined as ≥24 h of hospital stay. For each hospitalization, a maximum of three diagnoses were registered. We defined a composite dichotomous predictor of participants with ≥3 inpatient hospitalizations in the study period compared to participants with <3. Whole-genome sequencing, performed by the NHLBI Trans-Omics for Precision Medicine (TOPMed) programme, has been described elsewhere.19 Association analysis, as well as Q–Q and Manhattan plots, was performed using ENCORE (https://encore.sph.umich.edu). The SAIGE (Scalable and Accurate Implementation of GEneralized mixed model)20 logistic mixed model was used. Adjustments were made for hydroxyurea (hydroxycarbamide) treatment and the first 10 principal components. A genome-wide p-value threshold of 5 × 10−8 was applied. Only single nucleotide variants (SNVs) with a Minor allele frequency greater than 0.1% were included. LocusZoom21 was used to generate LocusZoom plots. To investigate whether a secondary signal existed beyond the lead SNV, a stepwise conditional analysis using GCTA-COJO22 was conducted. To identify putative causal variants within the UHRF2 gene region, variant-level statistical fine-mapping using Sum of single effects (SuSIE) regression (SuSiE-R)23 with summary statistics was performed. SNVs that reached genome-wide significance were tested using the chi-square test for associations with mortality during the study period. Of the entire cohort, a subset of 1931 participants with HbSS or HbSβ0 genotypes had available data on whole-genome sequencing, hospitalization history, age, sex and hydroxyurea (hydroxycarbamide) treatment and were included in this analysis. Of these, 387 (20%) had ≥3 hospitalizations in the ~3-year follow-up period who were compared to 1544 other cohort members. The majority of the participants (44.3%) did not have any hospitalizations in the 3-year period. One hundred and seventy-one participants had 5 or more hospitalizations. There were no significant differences in the sex and age distributions of participants with ≥3 hospitalizations compared to participants with <3 hospitalizations (54% vs. 49.7% female and 18.8 years vs. 20.6 years respectively). There were also no significant differences in haemoglobin levels, lactate dehydrogenase, platelet counts or the prevalence of stroke, leg ulcers and priapism between the two groups. However, participants with ≥3 hospitalizations were more likely to be treated with hydroxyurea (hydroxycarbamide) and received more transfusions in addition to having a higher prevalence of complications such as ACS, hospitalization by vaso-occlusive pain, avascular necrosis and kidney disease, suggesting a more severe phenotype of these individuals (demographic and clinical history are summarized in Table 1). A total of 2955 hospitalizations were recorded, with up to three diagnoses reported for each hospitalization (Figure 1). Most hospitalizations were related to SCD pain, and the second most frequent diagnosis was ACS (Figure 2). GWAS was performed comparing the groups with ≥3 hospitalizations to those with fewer than 3, adjusting for hydroxyurea (hydroxycarbamide) treatment. Figure 3 provides a summary of the GWAS results. The Q–Q plot (Figure 3) showed no evidence of bias in p-values and the λGC (=1.097) indicated no genomic inflation or additional confounding factors. A total of 20 SNVs (Table 2 and Figure S1) reached genome-wide significance and are located in chromosome 9 near or in gene ubiquitin-like with PHD and ring finger domains 2 (UHRF2) (Figure S1). The SNVs also showed significant associations with mortality during the study period (Table S1), underscoring the potential relevance of this locus to disease severity (Figures 2 and 3). From the conditional analysis, no additional independent significant SNVs were identified after conditioning on the lead SNV. SuSiE-R identified two 95% credible sets, with a posterior inclusion probability of 1, each containing a single SNV: rs201474308 and rs1330381, respectively (Figure S2), suggesting the presence of at least two independent signals. UHRF proteins take part in important biological processes, including maintenance of DNA methylation, cell cycle regulation and DNA damage repair.24, 25 There is a notable body of evidence linking the UHRF family of proteins with cellular regulation, including development, gametogenesis and carcinogenesis.26 UHRF2 has been extensively associated with gene expression regulation,27 including haematopoiesis.28 In a mouse model, a UHRF2 knockout genotype showed an impaired ability of haematopoietic stem/progenitor cells (HSPCs) to reconstitute haematopoiesis (Table 2). The role of UHRF proteins in maintaining DNA methylation is important to highlight. Dysregulation of the methylation process caused by either genetic mutations or environmental stimuli is potentially associated with multiple diseases and accelerates ageing.29 A discrepancy between epigenetic age and chronological age, that is called ‘epigenetic age acceleration’, underlies higher mortality risk and age-related conditions in humans.30 Additionally, DNA methylation is associated with HbF levels and the response to hydroxyurea (hydroxycarbamide), involving the methylation of γ-globin genes by DNA methyltransferases (DNMT3A, DNMT3B, DNMT1 and MBD family).31 Both UHRF1 and UHRF2 interact with DNMT1, DNMT3a, DNMT3b and G9a.32 On average, individuals with SCD have at least a 20-year reduction in survival compared to the overall population, with a life expectancy of 52.6 years.33 UHRF2 has been identified as a potential biomarker for elevated tricuspid regurgitation velocity in SCD.34 The observed association in Desai et al. suggests a possible role for UHRF2 in the pathogenesis of pulmonary hypertension, one of the more severe organ complications presented by patients with SCD. Alternatively, UHRF2 may take part in pulmonary vaso-occlusions, which underlie the occurrence of pulmonary hypertension. Cardiopulmonary complications are a major cause of death in adults with SCD, and elevated estimated pulmonary artery systolic pressure has been associated with increased mortality.35 Our identified association between the UHRF protein family and SCD severity suggests epigenetics as a factor underlying SCD burden, potential organ damage and longevity. Identifying the genetic background of SCD severity is important to the identification and prognosis of individuals who may be candidates for transformative therapies such as stem cell transplantation or gene therapy. If confirmed in other large cohorts, our finding of variations in UHRF2 as a potential marker of disease severity, associated epigenetic modulation and pulmonary hypertension may help identify these higher-risk individuals. We explored genetic factors associated with frequent hospitalizations in individuals with SCD, using this outcome as a proxy for disease severity. The identified association between UHRF2 and SCD severity helps our understanding of the genetics underlying an unfavourable disease course in SCD. Further studies are needed to investigate mechanisms and pathways through which the UHRF2 influences SCD severity. Mina Cintho Ozahata cleaned, analysed and interpreted the data, and drafted the manuscript. Isabel cleaned and analysed the data, and approved the final manuscript. Monike Oliveira, Cláudia Maximo, Alessandra Ferraz, André Rolim Belisário and Daiane Souza collected the data, contributed to data interpretation, and reviewed and edited the manuscript. Shannon Kelly, Ester C. Sabino and Brian Custer supervised the protocol design, and reviewed, edited and approved the manuscript. Carla L. Dinardo conceptualized the study, and supervised, reviewed, edited and approved the manuscript. The Article Processing Charge for the publication of this research was funded by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior - Brasil (CAPES) (ROR identifier: 00x0ma614). Molecular data for the Trans-Omics in Precision Medicine (TOPMed) programme was supported by the National Heart, Lung, and Blood Institute (NHLBI). WGS for ‘NHLBI TOPMed: REDSIII’ (phs001468) was performed at the University of Washington, Seattle, WA (HHSN268201500015C). REDS-III was supported by the National Institutes of Health, the National Heart, Lung, and Blood Institute under Grants HHSN2682011-00001I, -00002I, -00003I, -00004I, -00005I, -00006I, -00007I, -00008I and -00009I. The authors have no conflict of interest with regard to the publication of this manuscript. This research was approved by all relevant ethical committees/IRBs in both the US and Brazil, including the Brazilian National Committee of Ethics in Research, local ethical committees at each participating site as well as the IRB of record for Vitalant Research Institute. All participants (or guardians of participants <18 years) signed informed consent to participate in the REDS-III Brazil SCD cohort and allowed their DNA and data to be used in GWAS. 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. Figure S1. Figure S2. Table S1. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.