Severe anaemia, iron deficiency, and susceptibility to invasive bacterial infections

Severe anaemia and invasive bacterial infections remain important causes of hospitalization and death among young African children. The emergence and spread of antimicrobial resistance demand better understanding of bacteraemia risk factors to inform prevention strategies. Epidemiological studies have reported an association between severe anaemia and bacteraemia. In this review, we explore evidence that severe anaemia is associated with increased risk of invasive bacterial infections in young children. We describe mechanisms of iron dysregulation in severe anaemia that might contribute to increased risk and pathogenesis of invasive bacteria, recent advances in knowledge of how iron deficiency and severe anaemia impair immune responses to bacterial infections and vaccines, and the gaps in our understanding of mechanisms underlying severe anaemia, iron deficiency, and the risk of invasive bacterial infections.


Introduction
Invasive bacterial infections remain important causes of paediatric in-hospital admission and mortality, and public health interventions remain imperfect. Invasive bacterial infections causing bacteraemia account for 6-15% of febrile hospital admissions with case fatality rates of 5-28% 1-5 and a higher probability of post-discharge mortality 6 . In sub-Saharan Africa, the commonest invasive bacterial pathogens are Streptococcus pneumoniae, Staphylococcus aureus, nontyphoidal Salmonella (NTS), Haemophilus influenzae, and Escherichia coli [1][2][3]7 . In this region, susceptibility to invasive bacterial infections is sustained by a high prevalence of acquired comorbidities such as malnutrition, infections (malaria, and human immunodeficiency virus (HIV) infection), and sickle cell disease 2,[8][9][10] . A common feature among these comorbidities is the fact that they also cause severe anaemia ( Figure 1) 11 , and accumulating evidence suggests a strong association between severe anaemia and increased risk of bacteraemia among African children (Table 1).
Severe anaemia is a moderate-to-severe public health problem in many countries 11 , and is associated with detrimental acute and long-term effects [12][13][14] . In sub-Saharan Africa, severe anaemia is prevalent among pre-school children [15][16][17] , and is present in 6-28% of febrile hospital admissions with case fatality rates of 4-10% [18][19][20][21][22] . Nearly one in five (18.8%) children with severe anaemia in developing countries die within 6 months of hospital discharge 13,22 . However, the biology of severe anaemia is complex. Several distinct mechanisms (including haemolysis, red blood cell production failure, and blood loss) lead to severe anaemia 11 . Additionally, underlying aetiologies of severe anaemia, including infections (malaria, helminth infections, tuberculosis, and HIV infection), nutritional factors (acute/chronic malnutrition, and iron/folate/vitamin B 12 deficiency) and haemoglobinopathies (such as sickle cell disease) often coexist in a single patient or population ( Figure 1). Although the World Health Organization has provided standard cut-offs for severe anaemia 23 , definitions often vary across studies and populations (Table 1). There are no internationally recommended cut-offs for diagnosing severe anaemia in children <6 months of age 23 , and higher cut-offs (haemoglobin <9.0 g/dL) have been suggested for neonates 17 .

Severe anaemia disrupts iron homeostasis
Iron is an important nutrient required for the survival of nearly all aerobic organisms. The unique ability of iron to serve as an electron donor and acceptor renders it inimitable for various metabolic and physiological pathways, including cellular respiration, oxygen transport and DNA synthesis. In mammalian cells, iron is a co-factor for the assembly of functional iron-sulphur (Fe-S) cluster proteins, ribonucleotide reductases, and haem-binding proteins 24 . The majority of the iron is intracellular, complexed within the porphyrin ring of haem or sequestered in ferritin, a heteropolymer capable of storing about 4,500 iron atoms. Despite its numerous benefits, excess "free" iron is toxic through its formation of reactive oxygen species (ROS) or hydroxyl radicals. As such, iron metabolism is highly regulated through control of its absorption, mobilization, storage, and recycling.
Iron is essential for haem and haemoglobin synthesis. In human adults, approximately 2.5×10 6 red blood cells (RBCs) are produced per second 25 . Accordingly, about 20-25 mg of iron is utilized for erythropoiesis daily. Most of the iron required for erythropoiesis is obtained from recycling of senescent and damaged RBCs by splenic red-pulp macrophages.
In some conditions such as haemolysis, liver Kupfer cells and bone marrow erythroid island macrophages participate in iron  binds to other ligands, such as albumin and citrate, to form the non-transferrin-bound iron (NTBI) pool (also known as labile plasma iron, or "free" iron) 38 .
The process of iron absorption and recycling is highly adapted to maintain RBC production and optimal oxygen delivery to tissues. Severe anaemia, by acute or chronic blood loss or haemolysis, impairs this process. Rapid recovery from severe anaemia is thought to confer an evolutionary advantage 39 , probably due to the critical role of haemoglobin in oxygen delivery to tissues. As such, an expanded erythropoietic drive up to eight-times the baseline is observed during chronic haemolysis or repeated haemorrhage 40 . These high erythropoietic rates demand for increased iron supply. To achieve this, iron regulation by hepcidin is transcriptionally suppressed by erythroferrone (ERFE), a hormone produced by erythroblasts upon stimulation by erythropoietin (EPO) 41,42 . Other mediators such as platelet-derived growth factor (PDGF)-BB and soluble transferrin receptors could also suppress hepcidin production during severe anaemia 43,44 . Due to the evolutionary benefit of RBC restoration, hepcidin downregulation is observed to occur despite the presence of other underlying inflammatory or infectious mediators 45-47 .
In sub-Saharan Africa, severe anaemia is primarily caused by infectious agents (malaria, HIV infection, hookworms, and bacteraemia), sickle cell disease, and nutritional deficiencies (iron/folate/vitamin B12) 11 . The nature and intensity of the disruption of iron homeostasis, therefore, varies depending on the underlying severe anaemia aetiology. For example, absolute iron deficiency, due to poor dietary iron uptake or inhibited iron absorption, is characterised by low levels of iron both in the plasma and hepatic stores. To restore RBCs, children with absolute iron deficiency require exogenous iron supplementation, but this may also predispose them to infections and risk of mortality 48 . On the other hand, infectious agents are often associated with increased haemolysis (malaria, bacteraemia, and HIV infection) or haemorrhage (hookworm infections). During mild-to-moderate haemolysis, macrophages efficiently phagocytose damaged RBCs, break down haemoglobin and haem, and recover iron to sustain de novo erythropoiesis. Scavenger haemoproteins (such as haptoglobin and hemopexin) and receptors shuttle extracellular haemoglobin and haem to macrophages for clearance and detoxification 49 . Pathological haemolysis, as expected with various severe anaemia aetiologies, results in sustained release of haemoglobin and labile haem which potentially surpass shuttle protein and macrophage clearance capacity. This results in increased plasma haemoglobin, labile haem, and iron, which are not only toxic, but can also accelerate the growth of invasive pathogens. Unliganded haem and haemoglobin can translocate across endothelial barriers into subendothelial and perivascular spaces and the lymph fluid 50 . Furthermore, due to the need for iron in expanded erythropoiesis, iron regulation is impaired in severe anaemia. African children with severe anaemia and concomitant infections have markedly low hepcidin levels despite having elevated markers of inflammation 45-47 . This might also contribute to increased plasma iron levels, and potentially increase the pathogenesis of invasive bacterial infections.

Iron is required for anti-bacterial immune responses
Iron is an important factor in the immune response to invasive infections. Iron deficiency protects against malaria 51,52 , while iron supplementation has been associated with increased risk of malaria, diarrhoea, and other infections 48,53 . However, these results have not been universally observed, and studies are often confounded or have marked biases 54 . Little is known about the association of iron supplementation with bacteraemia incidence or mortality in endemic regions. Considerable literature has been published on how hepcidin sequesters iron away from invasive bacterial infections, especially in cell and mouse models 55,56 . This iron restriction, a main component of "nutritional immunity" 57 , is effective against extracellular bacteria 58-60 , but could be detrimental for intracellular infections 61 . Prolonged hepcidin upregulation also leads to mild-moderate anaemia of chronic disease 62,63 . In African children, anaemia may impair the development of immune responses from a young age 64,65 . Nonetheless, our knowledge of the effects of iron and anaemia (of varying severities) on cellular and non-cellular immune processes against bacterial infections remains incomplete.

Innate immune responses
The skin and mucosal membranes provide a physical barrier against invasive bacterial infections. Breaches in these barriers (through burns, lacerations, abrasions, and wounds) provide access of bacteria, both commensal and pathogenic, to the bloodstream. Data suggest that severe anaemia promotes breaches in gut mucosal integrity. Severely anaemic neonatal mice had a persistent increase in intestinal permeability, which has been postulated to be due to abnormalities in epithelial adherens junctions 66 or decreased expression of the tight junction protein zonula occludens-1 67 . A similar increase in gut permeability has been reported in Kenyan children with severe malaria anaemia 68 , although the precise mechanisms remain speculative. This gut barrier dysfunction, combined with possible microbiota dysbiosis during anaemia 53 , feasibly underpin invasion of gut pathogens in children with severe anaemia ( Figure 2). The skin and mucosal membranes also play a role in systemic homeostasis through iron losses in the form of sweating, hair growth, and epithelial sloughing. A recent Mendelian randomization study of 48,972 European individuals found an association between higher iron status and risk of skin bacterial infections 69 , possibly through increased iron supply to skin pathogens. Some antimicrobial peptides, such as lactoferrin and lipocalin-2 (also known as siderocalin or neutrophil gelatinase-associated lipocalin (NGAL)) sequester iron from bacterial pathogens on epithelial surfaces 70,71 . It is plausible that iron status also influences secretion of other epithelial antimicrobial molecules, such as cathelicidins, cryptdins, and β-defensins, but there is limited research in this subject.
Phagocytic leukocytes (macrophages and neutrophils) detect, ingest, and kill invasive bacterial pathogens. Pattern recognition receptors (PRRs) on phagocytic leukocytes detect bacterial pathogen-associated molecular patterns (PAMPs) such as lipopolysaccharide (LPS), unmethylated CpG DNA, and flagellin. The PRR-PAMP interaction induces a cascade of signalling events that mediate effector functions such as enhanced phagocytosis, macro-autophagy, proinflammatory cytokine and chemokine production, release of extracellular traps, antigen presentation, and generation of reactive oxygen species (ROS), reactive nitrogen intermediates and antimicrobial peptides 72 . Iron status influences these PRR-PAMP interactions. For instance, ligation of the Toll-like receptor (TLR)-4, the LPS receptor, induces hepcidin production by macrophages and neutrophils 73 . On the other hand, low intracellular iron levels in macrophages are associated with impaired activation of TLR-4 and reduced inflammatory cytokine expression 74 .
Neutrophils, by virtue of their numbers and antimicrobial armamentarium, have the strongest antibacterial capacity. Consequently, the effects of iron status on neutrophil numbers and function will significantly influence bacterial control. Recent studies in mouse models found that neutrophil production and function are more iron demanding than production of other leukocytes 75,76 . Iron chelation and hepcidin-mediated iron deficiency substantially reduced neutrophil numbers, enhanced neutrophil extracellular traps (NETs), inhibited pharmacologically stimulated granulopoiesis, and suppressed antibacterial responses in mice 76 . On the other hand, iron overload and increased labile haem availability, which are characteristic of haemolytic severe anaemia, also impair neutrophil antibacterial functions 77,78 . Mouse models of severe malarial anaemia have also reported haemolysis-associated reduction in neutrophil influx and lower production of proinflammatory cytokines 78-80 . This suggests a U-shaped relationship between iron status and neutrophil function, where both iron deficiency and iron excess are detrimental. In humans, iron deficiency is associated with neutropenia 81 , neutrophil hyper-segmentation 82,83 , reduced neutrophil oxidative burst 84 , and impaired NETs release 85,86 . Iron overload, characteristic of β-thalassemia major, is associated with impaired neutrophil phagocytosis and bactericidal activity 87 , although this effect was not observed in iron overloaded hereditary hemochromatosis patients with low hepcidin levels 88 . Previous studies found that monocytes from patients with HFE-associated haemochromatosis had low iron levels 89,90 , but little remains known on neutrophil iron status in these patients. Nonetheless, HFE-deficient mice demonstrated impaired neutrophil recruitment, suggesting a role for HFE in neutrophil responses 91 .
Macrophages play a key role in nutritional immunity by controlling systemic and intracellular iron availability to invasive bacterial infections. Iron status itself influences macrophage function. Macrophages display a continuous spectrum of polarization, which influences their function. M1 macrophages ("classically-activated") display the proinflammatory phenotype: they produce inflammatory cytokines (including interleukin (IL)-1β, IL-6, IL-12, IL-18, and tumour necrosis factor-alpha (TNF-α)), sequester iron into ferritin, and have potent antimicrobial capabilities. Increased intracellular iron promotes polarization to M1 phenotypes, and the molecular signature of proinflammatory M1 macrophages show increased activation of iron sequestration genes 92 . Consistent with these observations, haemolysis and haem accumulation in hepatic macrophages trigger a proinflammatory phenotype in a mouse model of sickle cell disease 93 . Although this phenotype is important for the control of extracellular bacteria, iron accumulation in macrophages is associated with the proliferation and persistence of intracellular pathogens such as Salmonella enterica spp., Listeria spp. and Mycobacterium tuberculosis 94,95 , while mechanisms of iron efflux from macrophages control these infections 96 . M2 macrophages ("alternatively activated") display the "healing" phenotype: they produce anti-inflammatory cytokines (IL-10 and transforming growth factor-beta (TGF-β)), actively release intracellular iron, and are involved in inflammation resolution and tissue repair. Iron release from M2 macrophages promotes cell proliferation, immune regulation, and matrix remodelling, consistent with their healing function. Acute iron deprivation of human macrophages is associated with reprogramming macrophages to the M2 phenotype in in vitro studies 97 .
Dendritic cells (DCs) are professional antigen presenting cells that create a bridge between the innate and adaptive immune responses. Once DCs phagocytose a pathogen, they become activated and transform to mature DCs, which migrate via blood to the spleen or via the lymphatic fluid to lymph nodes. Mature DCs upregulate expression of cell surface co-receptors (such as CD40, CD80 and CD86), which enhances their ability to activate CD4 + T cells, CD8 + T cells and B cells. Depending on their non-cognate costimulatory factors, DCs can also induce tolerogenic responses (unresponsiveness). Iron status can influence the numbers and function of DCs, and hence their ability to activate the adaptive immune responses. In in vitro studies, iron deprived monocytes differentiate into DCs with low expression of costimulatory molecules and are unable to activate T cells 98 . In contrast, labile iron impairs the activation of hypoxia-inducible factor (HIF)-1α, a transcription factor critical in DC effector function. Sequestration of iron in ferritin counteracts this effect and promotes HIF-1α activation of DCs in response to LPS 99 . In haemolytic anaemia, DCs also differentiate into macrophages to enhance erythrophagocytosis and iron recycling, reducing their numbers 100 . This haem-induced differentiation of DCs to erythrophagocytes contributes to a secondary immunodeficiency in children with severe haemolytic anaemia. During bleeding or injury to the gut, DCs secrete hepcidin to sequester iron away from the microbiota in mouse models that helps with tissue healing 101 . Nonetheless, there remains a general lack of research defining the influence of iron status, haem stress, and haemolysis on DCs in normal and disease states.
Non-cellular antimicrobial factors (such as complement, collectins, pentraxins, and ficolins) also form an integral part of the innate antibacterial immune responses. They bind to the bacterial PAMPs and trigger elimination of the invading organisms by direct lysis, agglutination, inhibition of growth, capsular swelling, opsonization, and activation of phagocytosis among other mechanisms 102,103 . Individuals with deficiencies in these factors, such as complement deficiencies, are more susceptible to bacterial infections. However, very little is known about the influence of iron status/anaemia on these antimicrobial factors. In healthy Asian adults, serum iron and ferritin levels were found to correlate with complement C3 and C4 levels 104 , which conflicts with findings from an older study on 20 children which found no effect of iron deficiency on complement C3 levels 105 . Since complement is an important bactericidal mediator, more definitive studies are needed to understand how its function is influenced by an individual's anaemia and iron status.

Adaptive immune responses
To successfully protect against invasive bacterial infections, the host must elicit efficient B and T cell responses. CD4 + T cells modulate responses by macrophages, neutrophils, CD8+ T cells and B cells. Cytotoxic CD8 + T cells directly kill bacteria-infected cells using molecules such as granzyme and perforin. B cells are key antibacterial cells by inducing antibody-dependent protection. Iron deficiency adversely affects the development and function of B and T cells. In some animal and cell models, iron deficiency was found to reduce the proportion of mature T cells, inhibit T cell proinflammatory cytokine release, and impair the proliferation and activation of T cells 106-110 . Loss of genes encoding iron regulatory proteins (IRP1 and IRP2) impairs iron uptake, proliferation, and effector functions of T cells 106 . Additionally, activated T cells express more than one million copies of the transferrin receptor (TfR1) within 24 hours of antigen encounter 106 , indicating an important role of transferrin-bound iron in T cell effector functions. Accordingly, in vitro blocking of TfR1 with antibodies inhibits polyclonal proliferation of human B and T lymphocytes 107 . Absence of TfR1 or mutations in the TfR1 gene, TFRC, is associated with complete arrest of T cell differentiation and defects in lymphocyte activation 106,109 . Iron deficiency induced by chelation modulates the activity of DNA and histone demethylases, which impairs B cell proliferation 110 .
In humans, genetic studies have found defective ex-vivo proliferation of B and T cells in patients with a hypomorphic mutation in the TFRC gene 111 . Ex vivo addition of high concentrations of iron citrate rescued the proliferative defects. Iron deficiency is also associated with thymic atrophy, but findings are inconsistent 112 . Two case-control studies, both with 40 children as cases, reported opposite findings on the effects of iron deficiency anaemia on B and T cell immunity 113,114 .
One study found that cases (children with iron deficiency anaemia) had lower CD4 + counts and CD4/CD8 ratios, but similar immunoglobulin levels compared with the controls 113 , while another study reported that cases had lower IgG levels but similar CD4/CD8 ratios compared with the controls 114 . These contrasting findings can possibly be attributed to different definitions of iron deficiency, anaemia, and controls, and confounding by other infections and modifiable biological factors. Moreover, these findings underpin the complexity of understanding the effects of anaemia on immunity in heterogenous human populations (unlike mouse model studies) and indicate the need for more systematic approaches.
The cytokine milieu during a bacterial infection guides the functional polarization of CD4 + T cells, and subsequent antibody responses. A Th1 and Th17 skewed response (characterised by interferon-gamma (IFN-γ), IL-17, and IL-22 secretion) is required to control invasive bacterial pathogens. IFN-γ activates macrophages and induces polarization towards the proinflammatory M1 phenotype, while IL-17 and IL-22 recruit neutrophils to the site of bacterial infection. The M1 macrophages not only sequester iron away from invading pathogens, but also possess potent antimicrobial activity. Polymorphisms associated with increased IFN-γ production promote iron deficiency phenotypes 115 . Interestingly, iron deficiency has been associated with reduced serum IFN-γ, IL-6, and TNF-α in animal and human studies 108,114,116 , but these findings are not universally observed 117 . Contrary to cell model studies 108 , a cross-sectional study reported increased proportions of lymphocytes producing IFN-γ, TNF-alpha and IL-6 in participants with iron deficiency 117 . These differences may be due to variations in study set-ups, severities of iron deficiency/ anaemia among the subjects, age, concomitant infections and/or underlying nutritional status. Many questions remain unanswered on the interaction of iron and adaptive immunity. For example, does iron supplementation in areas with high infectious disease burden improve immune responses? How do different severities of iron deficiency and anaemia affect B and T cell subset heterogeneity and long-term immune memory? Can these effects be reversed by iron supplementation?

Iron status determines host-bacteria interactions
The competition for iron between the mammalian host and invading bacterial pathogens is important in determining the course and outcome of an infection 55,56,118 . Not only is iron essential for the immune system, but iron is also required for bacterial growth and proliferation. Invading bacterial pathogens can access iron from various sources, including haem, haemoproteins, ferritin, transferrin-bound iron, non-transferrin-bound iron (NTBI), and lactoferrin 119 . During a bacterial infection, the host traffics iron away from the site of infection, into intracellular iron stores (ferritin) to protect against extracellular pathogens or out of cells to protect against intracellular pathogens 55 . The effectiveness of this iron sequestration is dependent on the bacterial niche and the bacterial iron acquisition strategies (Figure 3). At the centre of this nutritional immunity is hepcidin, which blocks iron recycling by splenic macrophages and iron absorption by duodenal enterocytes and suppresses plasma iron concentrations 120 . Hepcidin-mediated "hypoferraemia of infection" is effective against extracellular bacteria 58-60,121 , but has been shown to promote the growth and proliferation of intracellular pathogens 94,122 . Other mechanisms that work independently or in collaboration with the hepcidin-FPN axis to redistribute iron away from invading pathogens include induction of lactoferrin, lipocalin-2, calprotectin, and the natural resistance-associated macrophage protein-1 (NRAMP1, also known as Slc11a1) 70,71,123-125 . Impaired iron regulation is associated with increased susceptibility to bacterial pathogens such as Vibrio vulnificus and Escherichia coli in iron overloaded patients with haemochromatosis and thalassemia 126,127 .
Following the start of a bacterial infection, proinflammatory cytokines such as IL-6, IL-1β and TNF-α induce the production of hepcidin 128 . IFN-γ-mediated macrophage activation enhances intracellular bactericidal capabilities, including generation of ROS and reactive nitrogen intermediates, macro-autophagy, and nutrient deprivation 72 . This includes iron efflux through increased FPN expression 96 , a process that has been observed in mice infected with Salmonella enterica Typhimurium 129 . Interestingly, elevated macrophage intracellular iron inhibits the activity of IFN-γ in a dose-dependent manner 130,131 . Intracellular bacteria are also starved of iron by siphoning out Fe 2+ from the bacteria containing vacuoles. This is thought to involve various iron transporters such as NRAMP1 and FPN 124,132 . However, the precise role of FPN in this process remains unclear, as recent studies have demonstrated that, based on its orientation, FPN transports iron into the Salmonella-containing vacuole 133,134 . This raises questions on the role of iron in the vacuoles -whether increased iron availability benefits intravacuolar bacteria or kills them through ferroptosis or the Fenton reaction. M2 macrophages express higher levels of FPN and have been reported to be the most preferred niche for S. enterica Typhimurium 135 , suggesting an evolutionary advantage of infecting these cells. If FPN does indeed supply iron into the bacteria-containing vacuoles, then it is plausible that the iron benefits the bacteria and might explain the association of NTS with haemolytic severe anaemias in African children 65 . The conflicting findings warrant further studies on intracellular iron homeostasis mechanisms during bacterial infections, especially in the context of iron deficiency and severe anaemia.
Invading bacterial pathogens have evolved numerous mechanisms to acquire iron from the mammalian host. The bacteria sense iron-limiting conditions and upregulate iron-acquisition and virulence genes 56 . This is mediated by transcription factors, such as the iron-dependent ferric uptake regulator (Fur). In iron-limiting conditions, the Fur-mediated repression of iron transporters (such as the ATP-binding cassette (ABC) transporter) and siderophore genes is lifted, allowing enzymes for synthesis of iron acquisition proteins to be expressed 136 . Siderophores, which are low-molecular-weight highly potent iron chelators, can extract Fe 3+ from transferrin and other human chaperone proteins 137 . Bacteria can also acquire iron from haem, haemoproteins and siderophore-like haemophores 56 . Gram positive bacteria, such as Streptococcus pyogenes and S. aureus, produce haemolysins and have membrane receptors that scavenge and bind myoglobin, haemoglobin, and haemoglobin-haptoglobin complexes 138 . Haem from these haemoproteins is then transported through the envelope of bacteria into the cytoplasm, where it is degraded by haem oxygenase or reverse ferrochelatases to produce iron. Haem-responsive activators and receptors for haem, haemoglobin, and haemophores are also found on the outer membrane of some To safeguard against siderophores, the host relies upon lipocalin-2, which strongly binds enterobactin, the prototypical catecholate siderophore of many enteric bacteria, and sequesters the siderophore-iron complex 71 . Mice lacking lipocalin-2 exhibit increased susceptibility to enterobactin-expressing bacteria 143 . The host also sequesters labile plasma iron using lactoferrin, a high affinity Fe 3+ glycoprotein secreted by macrophages and neutrophils 70 . Hemopexin clears labile haem from plasma, and expression of IL-22 after an infection elevates hemopexin levels 144 . Bacterial defence against lipocalin-2 include use of "stealth" siderophores or production of competitive antagonists that bind lipocalin-2. Stealth siderophores, such as salmochelin (S. enterica Typhimurium) and peterobactin (Bacillus anthracis), are structurally modified making them resistant to lipocalin-2 145 . Pathogenic bacteria such as Neisseria meningitidis and Neisseria gonorrhoeae express lactoferrin receptors to bypass its iron restriction 146 . Some bacteria also bind and extract iron from mammalian siderophore-like molecules including 2,5-dihydroxy-benzoic acid (DHBA) and dopamine 147 .
The need for iron by bacterial pathogens presents an opportunity for vaccine development 148 . In the iron "arms race", bacteria have developed numerous conserved proteins that are expressed and exposed on surface membranes during an infection 149 . Merck V710, a novel S. aureus vaccine targeting the highly conserved iron-scavenging protein IsdB, remains the most advanced vaccine targeting bacterial iron proteins 150 . Although the vaccine was safe and demonstrated high immunogenicity in phase I studies 151 , it was terminated during a phase II trial due to safety concerns and low efficacy 152 . Participants receiving the Merck V710 vaccine had a higher mortality rate with post-operative S. aureus compared to those who received placebo 152 . Despite this failure, the V710 vaccine proved that targeting bacterial iron proteins is a feasible vaccine approach against bacteraemia. Other proposed iron protein vaccine candidates include Escherichia coli's

Severe anaemia increases the risk of invasive bacterial infections
An association between severe anaemia and invasive bacterial infections was reported long ago in a series of experiments by Kaye and colleagues in mouse models [155][156][157] . The researchers found that invasive Salmonella spp. was associated with severe anaemia induced by haemolytic processes (such as malaria infection and anti-RBC antibodies), but not non-haemolytic processes (phlebotomy) [155][156][157] . Subsequent mouse model studies have confirmed the association between haemolysis/iron status and Salmonella spp. bacteraemia 77,78,158 . In humans, epidemiological studies have reported strong associations between severe anaemia and invasive bacterial infections (Table 1), particularly with NTS bacteraemia (Table 2). However, interpretation of these data is difficult as the studies were performed in different age-groups, used different definitions for severe anaemia, and were mostly underpowered. Some of the children might also have had HIV infection or undernutrition, which possibly confounded the associations observed in these studies. Additionally, severe anaemia aetiologies are heterogenous and complex, and data is not provided on other key predictors of bacteraemia such as socio-economic status, nutritional deficiencies, caregiver health status, and genetic factors (Figure 1). Whereas iron deficiency is common in sub-Saharan Africa, prevalent in 70% of children in some areas 159 , several studies have found no or negative associations between iron deficiency and severe anaemia 18, 160,161 . Consequently, most of the underlying aetiologies of severe anaemia in this region are haemolytic (except hookworm infections which cause severe anaemia through blood loss) 11,18 . It is, therefore, plausible that the observed association between severe anaemia and invasive bacterial infections is through haemolysis-associated mechanisms.
Haemolytic anaemias induce sustained release of labile haem and NTBI. Haem impairs the recruitment and function of phagocytic leukocytes, including their ability to kill ingested bacterial through ROS 162,163 . In extreme haemolysis, haem is toxic to tissues, and may cause immune paralysis hence impeding resistance to invasive bacterial infections 163 . HO-1, the haemcatabolizing enzyme, induces tolerogenic effects of the immune system, and impairs resistance to NTS and other pathogens [164][165][166] . Haem breakdown products, biliverdin and carbon monoxide, scavenge antibacterial radical molecules and contribute to anti-inflammatory responses 167 . Increased NTBI levels are associated with increased susceptibility to siderophilic extracellular bacteria in mouse models 59,60 , and we have previously hypothesized that increased NTBI and labile haem increase NTS risk in Kenyan children with severe malaria 46 . We demonstrated that young children with severe malaria anaemia, accounting for more than 50% of admissions with severe anaemia in the study area, had markedly suppressed hepcidin levels compared to other forms of malaria, suggesting that the association between severe anaemia and bacterial infections may be through iron-mediated pathways 46 . Very low hepcidin levels could impair the host's ability to withdraw iron from invading bacterial pathogens as observed in mouse models 58-60 . It is also plausible that the acute need of iron for erythropoiesis promotes polarization of macrophages to the iron releasing M2 phenotype, which has poor antibacterial activity 72 .
Severe anaemia also has direct and profound adverse effects on cellular immune responses to bacteraemia (Figure 4). Early studies found that children with low haemoglobin levels had lower proportions of T cells and humoral bactericidal apabilities 168,169 . Evidence from literature suggests that these disruptions of immune function increase along a spectrum of anaemia severities and is dependent on the underlying severe anaemia aetiology 65 . For example, malarial anaemia reduces neutrophil and macrophage count, impairs oxidative burst capacity, reduces neutrophil bactericidal activity, and promotes iron accumulation in macrophages 77, 158,170,171 . Severe malarial anaemia is also associated with blunted production of IL-12 172 , a cytokine critical for control of NTS bacteraemia 173 . Although not epidemiologically associated with severe anaemia in sub-Saharan Africa 18,160,161 , severe iron deficiency is associated with neutrophil and macrophage dysfunction, cytokine dysregulation, impaired antigen-specific antibody production, reduced mature B-cell populations, thymic atrophy, and poor T cell responses 76,106,110,116 . Consequently, disruption of iron status impairs immune responses to bacterial infections, and as previously discussed, there is possibly a U-shaped relationship between iron status and immune function. In mouse models, iron supplementation of anaemic mice was associated with increased S. enterica Typhimurium colony-forming units in the spleen 174 . Both iron deficiency and iron overload, which are characteristic of different aetiologies of severe anaemia, negatively affect the development, recruitment, and function of various immune cells as discussed in this review. The availability of haem and its products during severe haemolysis exacerbates this immune impairment and may result in immune paralysis in extreme cases. Mechanistic studies of the association between severe anaemia and invasive infections should factor in specific anaemia aetiologies to help in the development of targeted interventions. Complexities arise in situations of overlapping aetiologies, such as haemolysis in children with severe iron deficiency. These circumstances are more common in malaria-endemic and resource limited settings than is appreciated, and little is understood regarding how the body regulates iron or the immune outcomes of bacterial infections in these patients.

Severe anaemia and iron deficiency impair immune responses to bacterial vaccines
Vaccines are extremely effective public health interventions. Licensed bacterial vaccines in routine immunisation schedules have drastically reduced the incidence and mortality from invasive pathogens such as C. diphtheriae, Bordetella pertussis, Clostridium tetani, H. influenzae type b, S. pneumoniae, and N. meningitidis. There is considerable progress in vaccine development for other bacterial pathogens using emerging technologies such as structural vaccinology, novel adjuvants, reverse vaccinology, and rationally designed bacterial outer membrane vesicles 186,187 . While vaccines have made an unprecedented impact on human health, not everyone benefits from vaccine protection equally. Poor vaccine efficacy is characteristic of low-income populations where severe anaemia and iron deficiency are highly prevalent 188 , and it is plausible that severe anaemia and/or iron deficiency impair vaccine responses. Most licensed bacterial vaccines confer protection by eliciting long-lasting protective antibody responses. Higher vaccine efficacy depends on synergy of B and T cell responses, as well as innate immune responses. However, iron deficiency and severe anaemia have widespread adverse effects on major immune cells required for efficient vaccine response (Figure 4).
Data on the effect of severe anaemia and/or iron deficiency on bacterial vaccine responses remain scarce, and the little available is mostly of poor quality 112,189 . We did not find records of any clinical trial that has investigated the effect of iron supplementation on bacterial vaccine efficacy. Observational studies on typhoid, diphtheria, and tetanus vaccines found no association between iron deficiency anaemia and vaccineinduced antibody titres 105, 168,190,191 , although iron deficient children had a lower T cell response 191 . Nonetheless, the studies included very few subjects and were likely underpowered 189 . In a recent study on a cohort of Kenyan children, iron deficiency and anaemia at the time of routine childhood vaccination were strong predictors of poor responses to diphtheria, pertussis, and pneumococcal vaccines, while moderate and severe anaemia were the strongest risk factors for anti-diphtheria sero-negativity at 18 months after vaccination 192 . This is consistent with a study from Ecuadorian infants that found low anti-diphtheria antibodies following DTP vaccinations among anaemic children 193 . Further studies are needed to confirm and quantify these associations between anaemia/iron deficiency and poor vaccine efficacy in other areas with a high burden of anaemia/iron deficiency.
In the context of severe anaemia and/or iron deficiency, there are some noteworthy concerns about vaccine efficacy that have not been addressed. First, it is not known whether severe anaemia or iron deficiency reduce antibody affinity or mediate wrong specificity of antibodies during germinal centre maturation, and if this is dependent on the severity of anaemia or iron deficiency. Second, while data show that iron is required for T and B cell proliferation activity, it is not known what quantity of serum iron is critical, and whether increased haemolysis or haem stress affects these processes. Germinal centre reactions are fundamental for the generation of high-quality and long-lasting B cell responses, and more studies are needed to understand how anaemia or iron status influences germinal centre responses in humans. More data is also needed to determine the effects of anaemia/iron status on vaccine responses including antibody avidity, T cell polarization, cytokine production, timing of immune responses, and their effects on the microbiome. Retrospective analysis of data from vaccine trials in areas with high anaemia and iron deficiency burden would provide invaluable insights, but this is contingent on such studies not having excluded participants with anaemia or iron deficiency. Standardised measurements and definitions of iron deficiency, accounting for inflammation and malaria 159 , are also required in these regions, as incorrect prevalence estimates may undermine the analysis and interpretation of data.

Conclusion
Severe anaemia and invasive bacterial infections are strongly associated among children living in sub-Saharan Africa, and the association may be due to disruption of iron regulation during severe anaemia. The aetiology and severity of anaemia are important determinants of the course and outcomes of invasive bacterial infections. However, our understanding of the interaction between anaemia, iron, and immune responses remains incomplete. Most of the existing evidence on anaemia, iron and immunity stems from in vitro and mouse model studies, and the findings are mostly conflicting due to differences in experimental set-ups. Extrapolating results from these controlled laboratory set-ups to heterologous human populations in anaemia endemic areas remains challenging.
There is also limited information on the effects of anaemia and iron status on vaccine efficacy, even though areas and populations with high anaemia and iron deficiency are also the ones with poor vaccine efficacy to currently available vaccines. As a result, there is an urgent need for more reliable and systematically collected data from endemic areas to understand how anaemia and iron status influence immune responses to infections and vaccination, and whether this differs by anaemia aetiology and severity.

Data availability
No data are associated with this article.