The Duffy-null genotype and risk of infection

Abstract Many medical treatments, from oncology to psychiatry, can lower white blood cell counts and thus access to these treatments can be restricted to individuals with normal levels of white blood cells, principally in order to minimize risk of serious infection. This adversely affects individuals of African or Middle Eastern ancestries who have on average a reduced number of circulating white blood cells, because of the Duffy-null (CC) genotype at rs2814778 in the ACKR1 gene. Here, we investigate whether the Duffy-null genotype is associated with the risk of infection using the UK Biobank sample and the iPSYCH Danish case-cohort study, two population-based samples from different countries and age ranges. We found that a high proportion of those with the Duffy-null genotype (21%) had a neutrophil count below the threshold often used as a cut-off for access to relevant treatments, compared with 1% of those with the TC/TT genotype. In addition we found that despite its strong association with lower average neutrophil counts, the Duffy-null genotype was not associated with an increased risk of infection, viral or bacterial. These results have widespread implications for the clinical treatment of individuals of African ancestry and indicate that neutrophil thresholds to access treatments could be lowered in individuals with the Duffy-null genotype without an increased risk of infection.


Introduction
It has long been recognized that individuals with an African or certain Middle Eastern ancestries often have reduced numbers of white blood cells, specifically neutrophils, compared with those with European ancestries (1). When neutrophil counts are <1.5 × 10 9 /L and the individual has no serious or recurrent infections, this condition is termed benign ethnic neutropenia. The Duffy-null (CC) genotype at rs2814778, in the Atypical Chemokine Receptor 1 (ACKR1) gene, previously known as FY and DARC, has been robustly associated with reduced neutrophil counts in individuals of African ancestry (2,3) and is considered to be the cause of benign ethnic neutropenia (4). The Duffynull genotype confers an evolutionary advantage by protecting against the malaria parasite Plasmodium vivax infection (5), and thus it is highly prevalent in geographical areas previously endemic for malaria, such as sub-Saharan Africa. The Duffy-null genotype has a prevalence of ∼80% in Black African/Caribbean populations in the UK (6), ∼65% prevalence in African Americans (3), and is very rare in individuals of European ancestry.
Recent studies have indicated that the Duffy-null genotype causes an altered neutrophil morphology that leads to neutrophils egressing from circulating blood into tissues and thus causing neutropenia (7,8). This mechanism is thought to be clinically benign because the production and functioning of neutrophils is not reduced and so their ability to fight infection remains unchanged (4). It has therefore been assumed that the Duffy-null genotype does not lead to increased rates of infection despite its association with reduced neutrophil counts. However, there have only been a few studies assessing infection outcomes in small clinical cohorts (<50 participants) of clinically diagnosed individuals with benign ethnic neutropenia (9)(10)(11), and none to date for the Duffy-null genotype (12).
Disparities in access to health care and treatment outcomes, including mortality, between individuals of African and European ancestries living in Europe and the United States are well documented (13) and unrecognized benign neutropenia could add substantially to this disparity. For example, individuals of African ancestry have been shown to have poorer clinical outcomes and access to medications for which neutropenia is a barrier (14,15). Thus, establishing that the neutropenia because of the Duffy-null genotype is benign has important clinical implications across the world for many treatments such as chemotherapy, immunosuppressant therapy, organ transplantation and antipsychotics such as clozapine (16). To address this knowledge gap, the aim of this study is to establish whether individuals with the Duffy-null genotype have increased rates of infection in two population-based samples from different countries and age ranges; the UK Biobank (17) and the iPSYCH Danish case-cohort study (18).

Results
In the UK Biobank sample, 7450 (1.53%) individuals had the CC (Duffy-null) genotype at rs2814778, 4525 (0.93%) had the TC genotype and 475 348 (97.54%) had the TT genotype. As expected, the great majority of individuals with the CC genotype reported their ethnicity as Black African/Caribbean (86.0%, Supplementary Material, Table S1). A total of 7644 individuals with a self-reported Black African/Caribbean ethnicity were selected for our primary analysis (57.05% female, mean age of 51.91 years at recruitment), of which 6363 (83.24%) had the CC genotype. Analysis of genetic principal components supports the finding that the CC genotype is not completely congruent with self-reported Black African/Caribbean ethnicity or our definition of African ancestry (Supplementary Material, Fig. S1). Supplementary Material, Table S2 details the country of birth for  individuals with the CC genotype in UK Biobank. A total of 283/77880 (0.36%) individuals in the iPSYCH sample had the CC (Duffy-null) genotype. Of these, 228 (80.57%) had 2 parents of African origin, 27 (9.54%) had 2 parents of Middle Eastern origin and 28 (9.89%) had parents of mixed or another origin. We selected 281 individuals for inclusion in the study that had 2 parents of African ancestry specifically from countries at risk for malaria (42.45% female), of whom 217 (77.2%) had the CC genotype and 64 (22.8%) the TC/TT genotype for rs2814778 (Supplementary Material, Table S3). The majority of study individuals was born between 1995 and 2005 and were followed for an average of 7.4 years (Supplementary Material, Fig. S2 and Table S4).  Table 1A provides comparative proportions of individuals with the CC and TC/TT genotypes that fall below certain ANC thresholds. Individuals with the CC genotype were significantly more likely to have an ANC between 0 and 2.0 × 10 9 /L, the current UK threshold for normal adult range of ANC, (odds ratio (OR) = 25.46; 95% CI = 14.39, 45.06; P = 1.06 × 10 −28 ) and between 0 and 1.5 × 10 9 /L (OR = 45.88; 95% CI = 12.09, 174.07; P = 1.86 × 10 −8 ) ( Table 1B). The results remained consistent in analyses including all ethnicities (Supplementary Material, Results).

Duffy-null genotype and risk of infection
Rates of infection for individuals with the CC (Duffy-null) and TC/TT genotype for rs2814778 in the UK Biobank and iPSYCH samples are detailed in Table 2. In UK Biobank, the CC genotype did not increase the risk of infection in individuals with a Black African/Caribbean ethnicity (rate ratio (RR) = 0.96; 95% CI = 0.82, 1.13; P = 0.61) or the number of infections per subject (RR = 0.95; 95% CI = 0.83, 1.08; P = 0.42). In the iPSYCH cohort, individuals with the CC genotype did not have an increased risk of infection (RR = 0.97; 95% CI = 0.65, 1.45; P = 0.88).
In the UK Biobank, we similarly found no association between genotype and specific types (bacterial, viral) or anatomical sites (respiratory, skin, gastrointestinal) of infection (Fig. 2, Table 2). Furthermore, individuals with the CC genotype in comparison to the TC/TT genotype in UK Biobank were not significantly more likely to have died from an infection-related illness (0.20% (n = 13) vs. 0.16% (n = 2); RR = 1.30; 95% CI = 0.33, 5.18; P = 0.71). All UK Biobank results remained consistent in analyses including all ethnicities (Supplementary Material, Results and Table S5).
All UK Biobank analyses controlled for sex, age at recruitment, Townsend deprivation index at recruitment and the first 20 genetic principal components to control for population   Association between the Duffy-null genotype and absolute neutrophil count (ANC) in UK Biobank participants. A provides comparative proportions of individuals with the CC and TC/TT genotypes that fall below certain ANC thresholds. Columns represent ANC thresholds, values for individuals with the CC (Duffy-null) and TT/TC genotypes, respectively, for rs2814778 with (i) a self-reported Black African/Caribbean ethnicity, and (ii) self-reported 'other' (non-Black African/Caribbean) ethnicity. B details the association between the CC (Duffy-null) genotype and an ANC < 2.0 × 10 9 /L and 1.5 × 10 9 /L. CI, confidence interval.

Ranges of neutrophil counts and risk of infection
Lastly, we investigated the risk of infection for different ranges of ANC in individuals of all ethnicities in UK Biobank. Figure 3 and Table 3  We were not able to detect any statistical difference between the rates of infection in individuals with the CC genotype and an ANC between 1.0 and 1.5 × 10 9 /L (22.0% (71/323)) compared with individuals with a normal ANC who had the CC genotype (RR = 1.17; 95% CI = 0.92, 1.48; P = 0.21) or the TC/TT genotype (RR = 1.03; 95% CI = 0.79, 1.33; P = 0.85). However the low number of subjects with an ANC <1.5 × 10 9 /L precludes us from making firm conclusions about risk of infection at these very low ANC counts (Table 3).
Supplementary Material, Table S6 details the associations between viral infections and different ranges of ANC. We found evidence indicating that individuals with low ANC have an increased risk of viral infection but that this did not differ by Duffy-null genotype. In addition, we were not able to detect any statistical difference in the risk of infection between individuals with the CC and TC/TT genotypes that had similar ANC, for example between 1.0 and 1.5 × 10 9 /L (Supplementary Material, Table S7).

Discussion
In two samples covering different countries and age ranges, we found that the CC (Duffy-null) genotype at rs2814778, which causes a reduction in the number of circulating neutrophils, does not increase the risk of serious infection. Further studies are required to establish the risk of infection at very low ANCs, but these findings could have widespread implications for the clinical management of individuals of African ancestry.

Main findings
Using the large UK Biobank sample, we confirmed that the Duffy-null genotype is strongly associated with reduced ANC in individuals with a self-reported Black African/Caribbean ethnicity. This has already been documented in several large metaanalyses (2,3), and we extend these findings to show that these results are consistent in individuals of other ethnicities.
Despite this large effect on ANC, we found no evidence from two independent samples that individuals with the Duffy-null genotype had an increased risk of infection. This was true for both viral and bacterial infections. Furthermore, we found no evidence of an increased risk of death from infection in those with the Duffy-null genotype. These findings provide an empirical base for the widespread assumption that the on average reduced ANC in those with the Duffy-null genotype is a benign phenomenon. Recent experimental work indicates that the lower ANC seen in individuals with the Duffy-null genotype is caused not by a reduced production of neutrophils but rather by the neutrophils egressing from circulating blood into tissues (4,7,8).
It has also been assumed that particularly low ANC (<2.0 × 10 9 /L) in those with the Duffy-null genotype is benign in nature, but there has been little evidence to confirm this assumption (9,16,19). We found that the Duffy-null genotype was not associated with an increased risk of infection for individuals with an ANC between 1.5 and 2.0 × 10 9 /L, whereas the TC/TT genotype did show an increase in infection risk. However, the low number of subjects with an ANC < 1.5 × 10 9 /L precludes us from making firm conclusions about risk of infection at these lower ANC levels, although the findings presented do not provide evidence of significantly increased rates of infections. We found that individuals with low ANC appeared to have an increased risk of viral infection, but this did not differ by Duffynull genotype, and given that viral infections can lower ANC, this finding could be because of reverse causality if the infection was current. This analysis is further limited by the use of a single ANC measurement collected at the time of study recruitment. To conclusively address this important clinical question and to provide safe ANC thresholds that could be used in clinical settings, a prospective design is required to determine infection risk at the time of the low neutrophil count.

Clinical implications
The findings from this study could have widespread implications for the clinical treatment of individuals with an African ancestry living in Europe and the United States, whose neutrophil count should be interpreted in the light of genotype at rs2814778. The current threshold for a 'normal' neutrophil count in the UK and United States is based on available normative data rather than an association with pathology or adverse effects, and is based on patients with a European ancestry. In this study, we found that 21% of individuals with the Duffy-null genotype had a neutrophil count below the currently accepted range (<2.0 × 10 9 /L) at any one time in contrast to 1.1% of those with the TC/TT genotype. Thus, these thresholds will not be appropriate for individuals with the Duffy-null genotype. Neutropenia is a contra-indication to many treatments, for example certain types of chemotherapy, immunosuppressant therapy, organ transplantation and treatment with the antipsychotic drug clozapine (16).  Association of infection in individuals with the CC genotype and a low ANC in contrast to individuals with the CC and TC/TT genotype with a normal ANC (2.0-7.5 × 10 9 /L). Columns represent firstly for low ANC groups: genotype at rs2814778, ANC range, number of subjects in said group and number of infections. The comparative group consisted of individuals with an ANC in the range of 2.0-7.5 × 10 9 /L and columns represent: genotype at rs2814778, number of subjects and number of infections. RR, rate ratio, in reference to low ANC group; SE, standard error; P = association P-value.
Our findings suggest that it is likely that many individuals with an African ancestry seeking treatment in countries where they are a minority could be denied access to treatment because of a benign neutropenia caused by the Duffy-null genotype. This has important consequences; previous research shows that individuals of African ancestry living in Europe and the United States have poorer clinical outcomes and access to medications for which neutropenia is a barrier such as chemotherapy (14,15,20). Our findings suggest that ANC thresholds could be lowered to 1.5 × 10 9 /L in individuals with the Duffy-null genotype without an increased risk of infection. For example, clozapine treatment in the UK requires a baseline ANC > 2.0 × 10 9 /L and a recent study of clozapine users found that across longitudinal neutrophil count measures, 55.4% of patients with the Duffy-null genotype had an ANC <2.0 × 10 9 /L at some point during their treatment (6). Clozapine requires regular monitoring of ANC to aid detection of a rare but potentially serious side effect of agranulocytosis. In the UK and United States, alternative monitoring thresholds for individuals with benign ethnic neutropenia have been incorporated into treatment pathways, allowing more individuals of African ancestry to access the treatment (19,21). If a prospective study could prove that individuals with the Duffy-null genotype and a low ANC did not have increased risk of infection, this strategy could be replaced with Duffy-null genotyping and extended to other treatments to aid interpretation of neutropenia in individuals of African ancestry. These adjustments could improve the racial disparities seen in healthcare outcomes.
Our findings indicate that the Duffy-null genotype, a marker of African ancestry, or self-reporting a Black African/Caribbean ethnicity, is not associated with increased rates of infection whereas in the UK Biobank sample, we observed a strong association of both age and social deprivation with increased rates of infections. We would stress that our study does not include analysis incorporating coronavirus disease-2019 infections and that these findings cannot be assumed to apply to infections other than those included in the analysis.

Strengths and limitations
A key strength of this study is the inclusion of two independent studies with distinct characteristics that in combination give a high level of confidence in our findings. The UK Biobank sample allows for large numbers of individuals to be studied and has a wide range of data available. However, there is evidence of ascertainment bias in UK Biobank and thus the sample cannot be considered representative of the general population (22). Furthermore, given the age of UK Biobank participants was between 40 and 65 years old at the time of recruitment, sample representativeness could be limited by survivor bias; it is possible that Duffy-null carriers that were at high risk of serious infection could have suffered from disproportionate mortality. However, if this were the case we would observe lower than expected rates of individuals with the Duffy-null genotype, which we did not find. The iPSYCH cohort does not suffer from the limitations of UK Biobank, since it is a population-based birth cohort and is thus representative of the population in Denmark. However, the proportion of the iPSYCH sample with the Duffy-null genotype was low and thus more fine-grained analyses, such as specific types and sites of infection could not be conducted.
Our findings are limited to populations of African ancestry living in Europe and thus may not be generalizable to the Duffynull carriers living in other countries. It is thus important to conduct studies of this nature in other populations. Another limitation is that our findings relate to serious infections only and it is possible that those with the Duffy-null genotype could have more frequent minor infections that do not require contact with hospitals or secondary care services.

Conclusions
In two samples from different countries and age ranges, we found that the Duffy-null genotype, which causes an on average reduction in ANC, does not increase the risk of serious infection. This could have widespread implications for the clinical management of individuals of an African ancestry, whose neutrophil count should be interpreted in the light of genotype at rs2814778.

Samples
Study individuals were from two samples covering different age ranges; the UK Biobank (17) and the iPSYCH Danish case-cohort study (18). The UK Biobank is a large prospective populationbased cohort study of approximately 500 000 individuals aged between 40 and 69 who were recruited from across the UK between 2006 and 2010 (17). The North West Multi-Centre Ethics Committee granted ethical approval to UK Biobank and this study was conducted under project number 13310. Primary analyses included individuals who self-reported a Black African or Black Caribbean ethnicity (UK Biobank field ID: 21000). All ancestries were included in secondary analyses.
The iPSYCH study is a population-based sample of 78 000 individuals born between 1981 and 2005 in Denmark and has been previously described (18). Ethical approval for this study was provided by the Danish Scientific Ethics Committee, the Danish Health Data Authority, the Danish data protection agency and the Danish Neonatal Screening Biobank Steering Committee. We selected individuals for this study who had two parents of African origin, specifically from countries at current or historical risk of malaria (listed in Supplementary Material, Table S8), as defined by the Danish Civil Registration System (23).

Duffy-null genotype
UK Biobank participants were genotyped on either the UK Biobank Axiom, or the UK BiLEVE Axiom arrays at the Affymetrix Research Services Laboratory. Genotypes for rs2814778 were imputed using the Haplotype Reference Consortium panel (24) after standard quality control procedures. All genetic data were provided by UK Biobank and the imputation and quality control procedures are fully described elsewhere (25).
For the iPSYCH sample, DNA was extracted from neonatal dried blood spot samples obtained from the Danish Neonatal Screening Biobank and genotyped using the Illumina Psy-chChip. Genotypes for rs2814778 were imputed in 10 batches using IMPUTE2 (26) and haplotypes from the 1000 Genomes Project, phase 3 (27).
For both UK Biobank and iPSYCH samples, we selected individuals whose genotype for rs2814778 had been imputed with high confidence (genotype probability thresholds of 0-0.1, 0.9-1.1 and 1.9-2). Given the T allele for rs2814778 is dominant with respect to neutrophil count (2), individuals with the CT and TT genotypes were combined in all analyses.

Outcome measures
Neutrophil count. A total of 478 511 (95.22%) UK Biobank participants had a single absolute neutrophil count (ANC) assay result (UK Biobank field ID: 30140) derived from the blood sample obtained at the initial UK Biobank assessment centre visit at which participants were recruited (between 2006 and 2010).

Infections.
We chose as the primary study outcome a broad definition of infection, defined as any inpatient hospital admission in which at least one International Statistical Classification of Diseases and Related Health Problems (ICD) infection code was recorded (Supplementary Material, Table S9 provides the full list of codes included). Any diagnosis that started with a parent ICD code was included (for example, searching for A40 would also include A40.0, A40.1, A40.2 and so on). This included secondary infections that may not have been the primary reason for hospital contact.
Infections in the UK Biobank sample were extracted using ICD-10 codes from linked National Health Service  Table S9). Death from an infectionrelated illness in UK Biobank was defined as an ICD infection code listed as either the primary or secondary cause of death. Number of infections was defined as the number of hospital contacts associated with any of the ICD-10 codes.
Data for infections in the iPSYCH sample were extracted from the Danish National Patient Registry (28), using ICD-8 and ICD-10 codes listed in Supplementary Material, Table S9.

Analysis
All UK Biobank analyses were conducted controlling for sex (UK Biobank field ID: 22001), age at recruitment (UK Biobank field ID: 21022), Townsend deprivation index at recruitment (UK Biobank field ID: 189) and the first 20 genetic principal components to control for population structure. We compared the distribution of ANC between individuals with the CC (Duffynull) and TC/TT genotypes for rs2814778 in UK Biobank via linear regression (UK Biobank field ID: 22009). Using logistic regression, we also tested the relationship between genotype and an ANC < 2.0 × 10 9 /L and 1.5 × 10 9 /L. Risk of infections were analysed by modelling (i) the occurrence of an infection and (ii) number of infections per subject using Poisson regression and adjusting for the covariates listed previously. The models were unadjusted for observation time since hospital records started in 1997 for all subjects. Our primary analyses included individuals with a self-reported Black African or Black Caribbean ethnicity but all analyses were repeated in individuals of all ethnicities to assess generalizability of the findings.
In the iPSYCH sample, we calculated rate ratios for infections in individuals with the CC and TC/TT genotype in a Cox model using robust standard errors, and adjusting for sex and psychiatric case status on the time to first infection. Subjects were followed until death, loss to follow-up or April 4, 2017 whichever occurred first. Most infections were observed before age 5.

Supplementary Material
Supplementary Material is available at HMG online.