Interferon-gamma polymorphisms and risk of iron deficiency and anaemia in Gambian children

Background: Anaemia is a major public health concern especially in African children living in malaria-endemic regions. Interferon-gamma (IFN-γ) is elevated during malaria infection and is thought to influence erythropoiesis and iron status. Genetic variants in the IFN-γ gene (IFNG) are associated with increased IFN-γ production. We investigated putative functional single nucleotide polymorphisms (SNPs) and haplotypes of IFNG in relation to nutritional iron status and anaemia in Gambian children over a malaria season. Methods: We used previously available data from Gambian family trios to determine informative SNPs and then used the Agena Bioscience MassArray platform to type five SNPs from the IFNG gene in a cohort of 780 Gambian children aged 2-6 years. We also measured haemoglobin and biomarkers of iron status and inflammation at the start and end of a malaria season. Results: We identified five IFNG haplotype-tagging SNPs ( IFNG-1616 [rs2069705], IFNG+874 [rs2430561], IFNG+2200 [rs1861493], IFNG+3234 [rs2069718] and IFNG+5612 [rs2069728]). The IFNG+2200C [rs1861493] allele was associated with reduced haemoglobin concentrations (adjusted β -0.44 [95% CI -0.75, -0.12]; Bonferroni adjusted P = 0.03) and a trend towards iron deficiency compared to wild-type at the end of the malaria season in multivariable models adjusted for potential confounders. A haplotype uniquely identified by IFNG+2200C was similarly associated with reduced haemoglobin levels and trends towards iron deficiency, anaemia and iron deficiency anaemia at the end of the malaria season in models adjusted for age, sex, village, inflammation and malaria parasitaemia. Conclusion: We found limited statistical evidence linking IFNG polymorphisms with a risk of developing iron deficiency and anaemia in Gambian children. More definitive studies are needed to investigate the effects of genetically influenced IFN-γ levels on the risk of iron deficiency and anaemia in children living in malaria-endemic areas.


Abstract
: Anaemia is a major public health concern especially in Background African children living in malaria-endemic regions. Interferon-gamma (IFN-γ) is elevated during malaria infection and is thought to influence erythropoiesis and iron status. Genetic variants in the IFN-γ gene ) (IFNG are associated with increased IFN-γ production. We investigated putative functional single nucleotide polymorphisms (SNPs) and haplotypes of IFNG in relation to nutritional iron status and anaemia in Gambian children over a malaria season.
We used previously available data from Gambian family trios to Methods: determine informative SNPs and then used the Agena Bioscience MassArray platform to type five SNPs from the gene in a cohort of IFNG 780 Gambian children aged 2-6 years. We also measured haemoglobin and biomarkers of iron status and inflammation at the start and end of a malaria season.
We  ; Bonferroni adjusted P = 0.03) and a trend towards iron deficiency compared to wild-type at the end of the malaria season in multivariable models adjusted for potential confounders. A

Amendments from Version 1
To address the comments and suggestions made by the reviewers,

Introduction
Malaria and iron deficiency are major public health problems for children living in sub-Saharan Africa. The majority (94%) of the 405,000 global deaths due to malaria in 2018 occurred in sub-Saharan Africa, where up to 24% of the population have malaria parasitaemia at any given time 1,2 . In this region, iron deficiency (ID) and anaemia are highly prevalent 3,4 , and may lead to impaired brain development 5,6 , while iron deficiency anaemia (IDA) is a leading cause of years lived with disability in African children 7 . Increasing evidence suggests that malaria may be contributing to ID and IDA 8,9 . Previous studies reported that the prevalence of ID and IDA increased over the malaria season in Gambian and Kenyan children 10,11 , and decreased with the interruption of malaria transmission in the Kenyan highlands 12 .
Multiple factors may contribute towards the development of ID and IDA following a malaria infection. One such factor is interferon-gamma (IFN-γ), which is induced during acute and persistent malaria infection 13 . Among other type 1 responses, IFN-γ is involved in regulating erythropoiesis [14][15][16] and iron-regulatory proteins [17][18][19][20] . IFN-γ has also been reported to increase the expression of hepcidin 17 , and divalent metal transporter 1 (DMT1) 19 , while suppressing ferroportin 18,19 , ferritin 20 , and transferrin receptors 19,20 . This regulation of iron proteins may be aimed at starving invading pathogens of iron, a critical nutrient for pathogen growth, but could also play an important role in the pathogenesis of ID and IDA. Indeed, higher IFN-γ levels have been reported in Kenyan children with severe malarial anaemia 21 .
Single nucleotide polymorphisms (SNPs) in the IFN-γ gene (IFNG) on chromosome 12q14 have been associated with increased production of 23 , and with susceptibility to severe malaria 24,25 . Despite evidence that malaria induces the production of IFN-γ 13 and that this cytokine influences iron regulation 17-20 , it is not known whether variation in the IFNG gene influences the risk of ID and IDA among children in malaria-endemic areas. We investigated SNPs and haplotypes in the IFNG gene locus in relation to nutritional iron status and anaemia in a cohort of 780 Gambian children prior to and at the end of a malaria season, using an approach based on informative SNPs and Agena Bioscience MassArray platform typing.

Study area
The study was conducted in ten rural villages in the West Kiang region of The Gambia at the start (July 2001) and end (December 2001/January 2002) of a malaria season, as previously described 26 . Malaria incidence is highly seasonal in The Gambia, with the majority of cases occurring between September and December. The study participants were from the Mandinka and Fulani ethnic groups. All of the Fulani children were located in a single village and ethnic group was accounted for in all analyses by adjusting for village.

Study design
We used previously collected data from a cohort of 780 children aged two to six years, recruited at the start of a malaria season as previously described 26 . All children had a clinical examination and a blood sample collected in the morning between 6 and 11 am for full blood count, malaria film, and biomarkers of iron status and inflammation at the start and end of the malaria season. Children with pyrexia (temperature ≥37.5°C) had appropriate clinical investigations, clinical treatment and a blood sample taken 2 weeks later after recovery from illness. All children received a 3-day course of mebendazole for possible hookworm infection at recruitment.

SNPs and haplotype construction
Genotypes were determined on whole-genome amplified DNA (primer extension pre-amplification) by the Agena Bioscience MassArray platform (formerly SEQUENOM) using matrixassisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry as previously described 25 . Details of the primer sequences and assays are given in Extended datafiles 1 and 2 29 . The most informative haplotype-tagging SNPs (htSNPs) to type in Gambian subjects were identified by analysing the pattern of linkage disequilibrium (LD) in the IFNG gene loci using previously available data from 32 Gambian family trios 25,30 . The PHASE program (http://stephenslab.uchicago. edu/software.html) version 2.1 was used to infer haplotypes from the genotypes of the study population and estimate the frequency of each inferred haplotype 31 . The entropy maximization method was used to identify htSNPs that described >90% of the observed haplotypic diversity in this gene region 30 . The HaploXT program (http://www.sph.umich. edu/csg/abecasis/GOLD/docs/haploxt.html) was used to estimate pairwise LD statistics. Sickle cell (HbS, rs334) and glucose-6-phosphate deficiency (G6PD) deficiency (rs1050828 and rs1050829) were also genotyped using the Agena Bioscience MassArray platform.

Definition of terms
Inflammation was defined as ACT >0.6 g/L. ID was defined as ferritin <12 µg/L or <30 µg/L in the presence of inflammation or <15 µg/L in children ≥5 years 32 , anaemia as Hb<11.0 g/dL (or Hb <11.5 g/dL in children ≥5 years) and IDA as ID plus anaemia 33 .

Statistical analyses
Statistical analyses were conducted using STATA 15.1 (Stata-Corp. College Station, Texas, USA). Categorical data were expressed as proportions with corresponding percentages. Pearson chi-squared test was used to compare the prevalence of malaria and iron status (ID, IDA and anaemia) at the start and end of the malaria season. Changes in haemoglobin levels and markers of iron status over the malaria season were assessed using the paired t-test. Biological data that were not normally distributed were log-transformed, and geometric means were calculated from original untransformed values.
Log-transformed markers of iron status and risk of ID, IDA and anaemia were analysed using univariable and multivariable linear and logistic regression models, as appropriate. Multivariable regression models were adjusted for age (grouped by year), sex, village (which also acted as a proxy for ethnic group), malaria parasitaemia and ACT at the start and end of the malaria season. The Bonferroni correction for multiple testing 34 was applied when the five SNPs and six haplotypes were considered individually as independent factors. For multivariable analyses, P values are noted as adj. P for non-Bonferroni corrected analyses and as Bonferroni adj. P for multivariable analyses that are Bonferroni corrected, and for univariable models P values are similarly presented as P or Bonferroni P if Bonferroni corrected. All analyses were considered statistically significant at P <0.05.

Ethics
Individual written informed consent was obtained from children's parents or guardians and the study was approved by The Gambian Government and the Medical Research Council Ethics Review Committee (874/830). adj. P = 0.05) and children with G6PD deficiency had lower haemoglobin levels (9.7 g/dl [95% CI 9.4, 10.0]) than those with G6PD wild-type genotype (10.0 g/dl [95% CI 9.9, 10.2]; adj. P = 0.01). The prevalence of ID and IDA increased over the malaria season (from 20.6% to 31.6% and from 11.9% to 21.7%, respectively), as previously reported 11,26 . Individual markers of iron status also reflected an increase in ID over the malaria season. We found that hepcidin and hepcidin/ferritin ratio decreased while the TSAT/hepcidin ratio increased across the malaria season in keeping with the need for increased erythropoiesis and increased rates of iron absorption at the end of the malaria season. Table 1 summarises the characteristics of the study population and their iron status at the beginning and end of the malaria season.

Associations with iron and anaemia
We found that the IFNG SNPs were not associated with ID, IDA or anaemia at the start of the malaria season in multivariable logistic regression analyses adjusted for age, sex, village, ACT and malaria parasitaemia following Bonferroni adjustment (Extended datafile 5 29 ).
The IFNG SNPs were similarly not significantly associated with ID, IDA or anaemia at the end of the malaria season after Bonferroni correction. The IFNG+2200C allele (rs1861493) was associated with trends towards increased risk of anaemia ( TSAT (%) 12.9 (12.4, 13.4) 11.6 (11.1, 12.1) Hb (g/dL) 10.6 (10.5, 10.7) 10.0 (9.8, 10.1) * Frequency and percentages are shown. ** Geometric means and 95% CIs are shown. a Iron deficiency was defined as ferritin <12µg/l (or ferritin <30µg/l in the presence of inflammation or <15 µg/L in children ≥ 5years) 32 ; b iron deficiency anaemia as iron deficiency a and anaemia c ; c anaemia as haemoglobin<11.0 g/dL or haemoglobin <11.5 g/dL in children ≥ 5years. ZnPP, zinc protopophyrin; TSAT, transferrin saturation; sTfR, soluble transferrin receptor; UIBC, unsaturated iron binding capacity; and Hb, haemoglobin. as well as a trend towards ID in other markers of iron status at the end of the malaria season in adjusted linear regression models (Table 3).

IFNG haplotypes
Haplotype analysis identified ten haplotypes (four with less than 1% population frequency) resolved by SNPs at nucleotide positions -1616, +874, +2200, +3234 and +5612 in the IFNG gene locus (Figure 1). Six haplotypes accounted for most of the variation. The wild-type haplotype (haplotype 1, IFNG-CATTC) was present at a frequency of 35% in the Gambian children, while haplotype 6 (IFNG-CACTC), uniquely identified by the IFNG+2200 SNP, was present at a frequency of 7%. Using haplotype 1 as the reference, haplotype 6 was associated with a trend towards increased risk of ID (  (Table 5).

Discussion
In this study we observed an increase in the prevalence of ID, IDA and anaemia across the malaria season in Gambian children. Dietary iron insufficiency may be an important cause since the malaria season also coincides with the 'hungry season' in The Gambia when there is a scarcity of staple foods. We hypothesized that IFN-γ, a pro-inflammatory cytokine induced during malaria infection 13 , might play a role in influencing the risk of ID and anaemia in children exposed to malaria. In addition to directly reducing erythrocyte half-life 15 , evidence suggests that IFN-γ induces hepcidin and inhibits ferroportin, hence reducing iron absorption and promoting sequestration of iron in macrophages 17,19 . Consequently, high levels of IFN-γ induced during malaria infections may concomitantly lead to ID and anaemia.
We found that the IFNG+2200C (rs1861493) allele, located at intron 3 of the IFNG gene, was associated with reduced haemoglobin levels and a trend towards ID, IDA and anaemia   at the end of a malaria season in multivariable analyses adjusting for potential confounders. We then constructed haplotypes to increase the probability of capturing functional mutations which might reside within a given haplotype. Haplotype 6 (uniquely identified by the IFNG+2200C allele), was associated with reduced haemoglobin levels and TSAT and increased ZnPP levels in keeping with iron deficiency compared to the wild-type haplotype. Haplotype 6 was similarly associated with trends towards increased risk of ID, IDA and anaemia at the end of the malaria season.
So how might the IFNG+2200C genotype and a haplotype uniquely defined by this genotype potentially lead to reduced haemoglobin levels and ID at the end of the malaria season? A possible explanation may be through increasing IFNG gene expression and IFN-γ levels. The IFNG+2200C allele was associated with increased IFN-γ levels in Kawasaki disease patients 23 , although another study in patients with ankylosing spondylitis found no difference in IFN-γ levels by IFNG+2200 genotype 35 . Elevated IFN-γ levels promote dyserythropoiesis, anaemia and iron dysregulation. IFN-γ inhibits proliferation of erythroid progenitor cells by disrupting lineage differentiation, blocking renal production of erythropoietin, inhibiting renal iron reabsorption, and reducing red blood cell half-life [14][15][16]36 . As a type 1 immune response, IFN-γ also induces defensive transcriptional programs within enterocytes resulting in reduced dietary iron absorption 37 . Additionally, IFN-γ promotes iron sequestration in macrophages either directly or through its influence on hepcidin, ferroportin, and DMT1 17-20 . These responses reduce circulating transferrin-bound iron, which is required by Plasmodium parasites for metabolism and proliferation 38 .
In keeping with this, we observed decreased ferritin levels and TSAT and increased ZnPP levels in children carrying the IFNG+2200C haplotype. Lower haemoglobin levels may also translate into reduced amino acid availability for Plasmodium parasites 39 and hence protection against blood-stage parasitaemia. Indeed, we observed that the IFNG+2200C SNP was associated with protection against malaria parasitaemia at the end of the malaria season (adj. OR 0.40; Bonferroni adj. P=0.03).
The influence of the IFNG+2200C allele on haemoglobin and iron status was only observed at the end of the malaria season. We hypothesized that the effects of this SNP may be most marked when expression of IFNG is upregulated, such as during malaria infections 13,24 . This also highlights the influence of gene-environment interactions in promoting disease, in this case ID, IDA and anaemia. It is unlikely that the decreased haemoglobin levels observed in individuals carrying the IFNG+2200C allele was due to increased malaria since these children had reduced prevalence of malaria parasitaemia at the end of the malaria season. It is possible that higher IFN-γ levels, putatively produced by IFNG+2200C carriers, may induce a protective proinflammatory response against malaria 13,40 , but at the expense of iron homeostasis. The IFNG+874TT (rs2430561) genotype, located at the first intron coinciding with the NFkB binding region, has also been associated with higher production of 35,41 . However, studies have found no association between IFNG+874TT and malaria 25 or aplastic anaemia 42 , and in our study we observed a trend towards a decreased risk of ID in these individuals. Further investigations are required on a cellular level to explore putative functional effects of IFNG genotypes on IFN-γ levels and iron status.
To our knowledge, this is the first study examining the role of IFNG gene polymorphisms in relation to iron status. We found that the IFNG+2200C (rs1861493) allele, and a haplotype defined by this allele, were associated with reduced haemoglobin levels and a trend towards ID at the end of the malaria season, a finding that may be due to increased IFN-γ levels 23 . However, our study had a number of important limitations and our findings should be viewed with considerable caution. The study was conducted in a single site and had relatively small numbers (n = 756). Additionally, many of our findings were of marginal significance with wide confidence intervals and lost statistical significance after correction for multiple testing with Bonferroni adjustment. It is also unclear if our findings have clinical relevance at an individual level. The IFNG+2200C SNP may also be in linkage disequilibrium (LD) with another genetic variant within the haplotype that might influence IFN-γ levels and / or measures of iron status in Gambian populations. Finally, we did not measure IFN-γ levels to determine if they differed between genotypes at the end of the malaria season. Thus, our findings need to be examined in larger populationbased studies, in other malaria-exposed populations, and functional assays are needed to identify whether genetic variation in the IFNG gene influences iron status. However, our study supports the hypothesis that preventing and treating malaria infection may improve haemoglobin levels and iron status in African children 8 .

Data availability
Underlying data Havard Dataverse: Replication Data for: Interferon-gamma polymorphisms and risk of iron deficiency and anaemia in Gambian children, https://doi.org/10.7910/DVN/2NKJID 43 . This project contains the following underlying data: • IFNG_final_data_v2 (dataset containing demographic information and results of laboratory assays for participants included in the study).
• IFNG_analysis_KM (contains the codes used for data analysis).

Laura Silvestri
Regulation of Iron Metabolism Unit, Division of Genetics and Cell Biology, San Raffaele Scientific Institute, Milan, Italy In this paper Abuga et al investigate the role of IFNG SNPs in influencing iron deficiency (ID) and iron deficiency anemia (IDA) in Gambian children over a malaria season. They identify a SNP, IFNG+2200C, located in intron 3, associated with a reduced risk of malaria parasitemia, reduced Hb, and a trend towards ID and IDA. The authors conclude that although more studies are needed to assess the role of IFNG SNPs in ID and IDA, this represents the first study that investigates the association of IFNG genetic variants with iron status.
The study is well conducted and of interest. I have only minor comments: Serum hepcidin levels are influenced by circadian rhythm and serum/iron stores. Please indicate at what time of the day blood was drawn. To "normalize" serum hepcidin to body iron concentration, the hepcidin/ferritin or hepcidin/TSAT ratio should be shown in Table 1 and Table 3 (as in Donker et al., Pediatric Blood and Cancer 2019 ).
In the M&M section, the authors claim that some children are carriers of the HbAS, and G6PD deficiency. Is it possible to present the hematological and iron data (in Table 1 and

Are sufficient details of methods and analysis provided to allow replication by others? Yes
If applicable, is the statistical analysis and its interpretation appropriate? I cannot comment. A qualified statistician is required.
Are all the source data underlying the results available to ensure full reproducibility? Yes

Are the conclusions drawn adequately supported by the results? Yes
No competing interests were disclosed.

Competing Interests:
Reviewer Expertise: Laura Silvestri has long-term expertise in the characterization of molecular mechanisms responsible for the pathogenesis of iron/hepcidin-related disorders as Hereditary Hemochromatosis, beta-thalassemia, and iron refractory iron deficiency anemia (IRIDA).

"
Methods, under Study design: All children had a clinical examination and a blood sample collected in the morning between 6 and 11 am for full blood count, malaria film, and biomarkers of iron status and inflammation at the start and end of the malaria season." b) To "normalize" serum hepcidin to body iron concentration, the hepcidin/ferritin or hepcidin/TSAT ratio should be shown in Table 1 and Table 3

(as in Donker et al., Pediatric Blood and Cancer 2019).
Response: Thank you for the suggestion. We have added hepcidin/ferritin and transferrin saturation (TSAT)/hepcidin ratios to tables 1 and 3. We have also made the following changes: Methods section under laboratory methods: Hepcidin/ferritin and TSAT/hepcidin ratios were also . calculated Results section under Characteristics of participants: We found that hepcidin and hepcidin/ferritin ratio decreased while the TSAT/hepcidin ratio increased across the malaria season in keeping with the need for increased erythropoiesis and increased rates of iron absorption at the end of the 1 the need for increased erythropoiesis and increased rates of iron absorption at the end of the malaria season.
2. In the M&M section, the authors claim that some children are carriers of the HbAS, and G6PD deficiency. Is it possible to present the haematological and iron data (in Table 1 and  HbAA (115.9 [95% CI 110.6,121.3]; adj. P = 0.05) and children with G6PD deficiency had lower haemoglobin levels (9.7 g/dl [95% CI 9.4,10.0]) than those with G6PD wild-type genotype (10.0 g/dl [95% CI 9.9,10.2]; adj. P = 0.01)." informative SNPs and then used the Agena Bioscience MassArray platform to type five SNPs from the IFNG gene in a cohort of 780 Gambian children aged 2-6 years".

2.
The last sentence at the end of the Results section mentions a number of covariates that were adjusted for in the models. One such covariate is 'malaria parasitaemia'. a) Was malaria parasitaemia adjusted for only at the beginning and at the end of the malaria season or did it capture all the malaria cases during the study period?
Response: In the current analysis, we adjusted for malaria parasitaemia detected on blood film at the beginning and at the end of the malaria season, at the same timepoints when iron biomarkers were measured since malaria parasitaemia influences markers of iron status. We did not monitor for malaria cases during the malaria season. Methods, under Statistical analyses: "Multivariable regression models were adjusted for age (grouped by year), sex, village (which also acted as a proxy for ethnic group), malaria parasitaemia and ACT at the start and end of the malaria season." b) Does 'malaria parasitaemia' mean malaria episodes during the study period? Response: Malaria parasitaemia referred to the identification of Plasmodium parasites on the blood film taken from the study participants at either of the two cross-sectional timepoints, i.e. at the beginning and / or end of the malaria season. It does not refer to malaria episodes during the study period.

c) If it only captures the two time points, what would the effect of 'additional episodes'
during the study period have on the results? In malaria endemic areas, repeated infections are more likely to result in anaemia than just the patent infection. The authors need to clarify on this and if there were no malaria cases in-between the two time-points they need to indicate so.
Response: We agree that repeated malaria episodes would increase the risk of anaemia. Based on published data we estimated that the majority of the children would be likely to have had one or more episodes of malaria over the malaria season during the study period. Studies indicate that malaria increases IFN-g levels and that IFN-g regulates erythropoiesis and iron-regulatory proteins including hepcidin. We therefore hypothesized that IFNG SNPs, which might alter IFNG expression, would influence iron status at the end of the malaria season. Thus, we would expect that the effect of 'additional episodes' of malaria would be to increase malaria-induced IFN-g levels and IFN-g -induced iron deficiency and anaemia. To clarify we have updated the Methods section as follows: Methods, under Laboratory methods: "Giemsa-stained thick and thin blood films were examined for Plasmodium falciparum and other Plasmodium species at the start and end of the malaria season." No competing interests were disclosed.

Sarah Atkinson
We thank the reviewer for his helpful comments, which we have addressed point-by-point below: