Use of mouse models to study the mechanisms and consequences of RBC clearance

Abstract

Mice provide tractable animal models for studying the pathophysiology of various human disorders. This review discusses the use of mouse models for understanding red-blood-cell (RBC) clearance. These models provide important insights into the pathophysiology of various clinically relevant entities, such as autoimmune haemolytic anaemia, haemolytic transfusion reactions, other complications of RBC transfusions and immunomodulation by Rh immune globulin therapy. Mouse models of both antibody- and non-antibody-mediated RBC clearance are reviewed. Approaches for exploring unanswered questions in transfusion medicine using these models are also discussed.

Introduction: Mice as a tractable experimental model

Although the best way to understand human disease is to study humans, this is often practically difficult and ethically unacceptable [1]. Mice became an important tool for studying human pathophysiology because of their relative ease of genetic manipulation, small size, relatively low cost, short gestation and fecundity. In addition, resources, such as the Mouse Phenome Database (http://www.jax.org/phenome), provide phenotypic and genotypic information on approximately 40 common inbred strains, thereby providing researchers with convenient tools for selecting the appropriate model. The increasing availability of better methods for studying mouse physiology in vivo also facilitates their usefulness and dominance as animal models.

Genetics

The complete genome sequences of several common inbred mouse strains are available [2]. Despite the common assertion that ‘mice are not human,’ there appear to be only ~300 unique genes separating mice and humans [3], with ~99% of human genes having a mouse homologue and a highly conserved gene order [1]. Thus, mice and humans are actually quite similar. However, there are differences that must be considered when using mice for preclinical studies of human disease. For example, although mouse models are important for immunologic experimentation in vivo, there are key differences in both innate and adaptive immunity (for review, see [4]). In particular, human blood is neutrophil rich (50-70%), whereas mouse blood is lymphocyte rich (75-90%) [5]. As another example, immunoglobulin IgG subclasses and Fc-receptors are somewhat different (for review, see [6] and [7]), complicating crossspecies correlations of these molecules [4]. In addition, mice and humans express toll-like receptor orthologues differently [8]. Mouse complement function is also believed to be ‘weak’ in comparison with humans and other species [9,10]. Finally, transfused allogeneic white-blood-cells are cleared at different rates in mice and humans [11]. To overcome some of these limitations, ‘humanized’ mice, defined as either immunodeficient mice repopulated with human immune cells or mice genetically engineered to express human genes, are useful for studying human cells and biologic processes in vivo [12]. Although there are caveats to using ‘humanized’ mice, they provide another promising preclinical model for testing therapeutic approaches.

The genetic variability in any inbred mouse strain is almost negligible, allowing study of disease-causing mutations on an isogenic background [1]. Thus, identifying genetic loci regulating a disease or phenomenon is facilitated by genetic analysis of experimental mouse models of human disease. By crossing one mouse strain expressing the trait of interest with one resistant to that trait, and then genetically mapping the offspring using linkage analysis, one can identify quantitative trait loci associated with the trait of interest. For example, this technique identified the complement factor 5 and Fc-gamma receptor (FcγR) IIb genes as susceptibility alleles in a collagen-induced arthritis model [13]. In the future, the ‘Collaborative Cross’ [14], currently being developed by the Complex Trait Consortium, will provide the scientific community with a large ‘outbred’ set of recombinant inbred strains designed for complex trait analysis. In addition to enabling quantitative trait loci studies, this resource will allow for gene-environment interaction analysis and for validating predictive genetics. In the current context, these resources may help shed light on the role of genetic variation in important issues in transfusion medicine; these include the genetic contribution to differences in alloimmunization rates, in the clinical severity of transfusion reactions and in transfused RBC survival in vivo after prolonged storage in vitro.

Measuring RBC clearance

RBC transfusions are a simple type of adoptive transfer and the simplest type of tissue transplantation, thereby allowing testing of multiple variables. One of the most classical, common and useful experimental approaches in transfusion medicine is the measurement of RBC clearance, alternatively known as a ‘RBC survival study’. RBC clearance studies can be performed both in vitro and in vivo. Several types of genetically engineered mice, which do or do not express a defined RBC antigen, can be used for modelling allogeneic RBC transfusions and performing RBC survival studies (see Table 1). Mouse strains expressing differing, genetically encoded, blood group antigens could also be used for this purpose. In addition to exploiting genetic differences, the RBCs themselves can be modified or labelled ex vivo and then used to measure RBC clearance.

Table 1.

Genetically engineered mice with useful RBC antigens

Mouse line	Tissue distribution	Characteristics	Reference
Duffy (Fyb)-transgenic	All cells	Used to model alloimmunization. Mice initially form IgM, and then IgG,  anti-Fyb antibodies. Mice produce anti-Fyb, -Fy3 and -Fy6 antibodies.	[55]
Human glycophorin A  (hGPA)-transgenic	RBCs only	Many monoclonal anti-hGPA antibodies available. Incompatible transfusion  produces haemolytic transfusion reactions in mice accompanied by  cytokine storm.	[19],[52],[53],[54]
Hen egg lysozyme  (HEL)-transgenic	All cells	Mice passively immunized with anti-HEL followed by transfusion with  HEL-RBCs induced HEL-specific antigen loss. HEL-specific T-cell  receptor-transgenic mice enhance alloimmunization to transfused  HEL-RBCs. Alloimmunization to HEL-RBCs requires a spleen and is  augmented by certain types of inflammation.	[17],[22],[57],[127]
HEL-Ovalbumin-Duffy  (HOD)-transgenic	RBCs only	Immunologically useful model due to presence of a well-characterized  combination of antigens with specific immunological tools, such as  HEL- and ovalbumin-specificT-cell receptor- and HEL-specific B cell  receptor-transgenic mice. Refrigerator-stored leukoreduced HOD RBCs are  more immunogenic than fresh leukoreduced RBCs.	[128],[59]
H-transferase-  transgenic	All cells	Results in decreased expression of the α-Gal xenoantigen and increased  expression of the normally cryptic sialylated Tn and Forssman antigens.  Can be used to develop model of Bombay O RBCs.	[60]
Duffy antigen/receptor  for chemokines  (DARC) knockout	All cells	DARC is required for Plasmodium vivax infection of RBCs. Serves as  chemokine sink. LPS challenge produces increased inflammation. Can be  used to develop a haemolytic transfusion reaction model with the KO mice  as recipients of incompatible transfusions from wild-type donors. The  reactions in these mice may be more severe when compared to those in  wild-type mice.	[62]
alpha 1,  3galactosyltransferase  (α-Gal) knockout	All cells	Lack of α-Gal expression, which is universally expressed in mice. Can be used  to develop a carbohydrate antigen-based haemolytic transfusion reaction  model, with the K0 as recipients of incompatible transfusions from  wild-type donors.	[61]
Green fluorescent  protein  (GFP)-transgenic	All cells	Used to develop a mouse blood bank model of RBC storage. Useful for  performing RBC survival studies without the need to label or wash RBCs  pretransfusion.	[129],[20]

Studies can also be performed in vitro using phagocytic cells, which can be either immortalized mouse cell lines (e.g. RAW 264.7 or J774 cells) or primary cultures of cells obtained directly from mice (e.g. peritoneal exudate macrophages, bone marrow-derived macrophages, liver Kupffer cells). Macrophages are incubated in vitro with treated or untreated RBCs and phagocytosis is quantified. Phagocytosis can be measured in several different ways; for example, the number of phagocytes with ingested RBCs can be counted in several microscopic fields, along with the total number of phagocytes in those fields. The phagocytic index can then be defined as either the total number of RBCs ingested by 100 phagocytes or the total number of phagocytes with at least one ingested RBC per 100 phagocytes. Phagocytosis can also be quantified by first using hypotonic lysis to remove non-ingested RBCs followed by spectrophotometric measurement of the haemoglobin in a given number of phagocytes [15]. Finally, RBCs can be labelled, and the label used to measure clearance both in vitro and in vivo.

Radioactive chromium (i.e. 51-Cr) is the most common label used for RBC survival studies in vivo [16]. This not only allows facile measurement of RBC clearance, but also allows determination of the organ distribution of the cleared RBCs. Flow cytometric approaches offer additional benefits, including not requiring radioactivity, allowing evaluation of whether an antigen is still present on the RBC surface (e.g. the antigen-loss phenomenon [17]), and allowing evaluation of RBC coating by antibodies and/or complement (i.e. the direct antiglobulin test).

Flow cytometric approaches can be either immunological or non-immunological. Immunological flow cytometry survival studies involve either haptenating RBCs with moieties such as dinitrophenyl, trinitrophenyl and 4-hydroxy5-iodo-3-nitrophenylacetyl, followed by detecting and counting haptenated RBCs using labelled hapten-specific antibodies. However, some haptens induce antibody responses, thereby limiting their use in long-term survival studies [18]. Similar flow cytometric approaches can detect circulating transgenic RBCs (Table 1) using antigen-specific antibodies. By this approach, anti-hen egg lysozyme (HEL) antibodies in wild-type mice induced non-haemolytic antigen loss of HEL expressed on the surface of transfused HEL-transgenic RBCs [17]. By flow cytometry, the incompatible HEL-transgenic RBCs circulated with a normal lifespan, exhibiting selective loss of HEL antigen, but not other antigens. This illustrates an important advantage of flow cytometry: using multiple antibodies with different fluorescent labels provides significantly more information than other methods, such as agglutination or radioactive labelling.

Non-immunologic flow cytometric approaches also exist for tracking circulating RBCs. Lipophilic fluorescent dyes, such as 3,3′-dihexadecyloxacarbocyanine perchlorate (DiO) and chloromethylbenzamido 1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate (DiI), among others, can label RBCs [19]. In addition, RBCs can be labelled with biotin ex vivo and detected by flow cytometry using fluorescently labelled streptavidin. Virtually, all mouse RBCs can also be simultaneously biotinylated in vivo [16], thereby differentiating those RBCs from subsequently produced new RBCs; this is useful for enriching young or old RBCs and also for determining RBC lifespan. Finally, a recently described approach, which avoids labelling and washing steps pretransfusion, involves transfusing transgenic, green fluorescent protein (GFP)-expressing RBCs [20]; although the recombinant GFP did not alter RBC circulatory lifespan, other rheologic properties were not examined.

Organ sites of RBC clearance and cells involved

In mice, FcγR-mediated phagocytosis of IgG-coated RBCs predominantly occurs in the spleen; complement C3b-coated RBCs are rapidly sequestered in the liver. Some C3b-coated RBCs are phagocytosed, a process mediated by specific complement receptors [e.g. complement receptor 3 (CR3)], whereas other RBCs are modified by a C3b inactivator and released back into the circulation [21]. In addition, experimental splenectomy shows that although the spleen is required for alloimmunization to HEL-transgenic RBCs [22], it is not required for antigen loss of HEL-transgenic RBCs [17].

Mouse models can also be used to study the cells involved in phagocytosis. To investigate the role of macrophages, they can be depleted with liposomal clodronate [23] or gadolinium chloride [24]. A transgenic mouse generated using a CD11b- or CD11c-diphtheria toxin receptor construct also allows for conditional macrophage and/or dendritic cell depletion after diphtheria toxin injection [25,26]. Transgenic mice that express an inducible suicide gene also allow for conditional depletion of cells in the monocyte-macrophage system [27]. In addition, macrophage function can be inhibited with inert colloids, thorotrast, or various lipids [28]. In particular, ethanol decreases complement-dependent phagocytosis [21]. In contrast, macrophage function is enhanced by lipopolysac-charide (LPS); LPS induces ‘inside-out’ signalling of the CR3 integrin, thereby increasing complement-mediated phagocytosis of sheep RBCs by Kupffer cells [15]. In addition, glucan, a neutral polysaccharide, stimulates phagocytosis, increasing sheep RBC clearance in mice by fourfold [29]. Finally, increased phagocytosis accompanies viral and bacterial infections, in both laboratory animals and humans [30].

Antibody-mediated RBC clearance

Mouse models have been used for decades to study antibody-mediated mechanisms of RBC clearance. We will first review the general considerations for immune RBC clearance and then discuss mouse models of alloantibody-mediated clearance, followed by those describing auto-antibody-mediated clearance.

General considerations

After IgG antibodies bind to RBCs, the exposed Fc regions initiate phagocytosis by binding to FcγRs on the macrophage surface [31]. This induces a complex signalling cascade, initiated by FcγR clustering and tyrosine phosphorylation of the immunoreceptor tyrosine-activation motif in the cytoplasmic domain of activating FcγRs. This cascade results in cytoskeletal rearrangement and ingestion of antibody-coated RBCs into phagosomes, where acidification and proteolysis release haemoglobin and haeme; the latter is metabolized by haeme-oxygenase, producing biliverdin, carbon monoxide and iron (for a review of FcγR signalling, see Ref. [31]). Mouse models were critical in determining the pathways involved in this process. For example, the roles of the tyrosine kinase, Syk and other Src family tyrosine kinases in FcγR-mediated phagocytosis were demonstrated in the respective knockout (KO) mice (reviewed in [32]). Pharmacologic inhibitors of this pathway were also evaluated in mice. For example, a Syk inhibitor ameliorated both autoimmune haemolytic anaemia (AIHA) and autoimmune thrombocytopenic purpura (ITP) in mouse models; this was also efficacious in humans with chronic refractory ITP [33], highlighting the relevance of mouse models for preclinical studies of novel therapies.

Various mechanisms also regulate phagocytosis to control whether a RBC is internalized or not. For example, a threshold level of FcγR clustering is required to initiate phagocytosis [31], implicating antibody coating density and antigen density in RBC internalization. In addition, an inhibitory receptor, FcγRIIb, with a cytoplasmic domain immunoreceptor tyrosine-inhibitory motif, recruits phosphatases that inhibit the signalling required for phagocytosis. Therefore, the ratio between inhibitory and activating FcγRs affects phagocytosis. The IgG subclass opsonizing RBCs also affects phagocytosis by differentially binding to activating and inhibitory FcγRs. Mice have four FcγR types (FcγRI, IIb, III and IV) [7]; as examples, the high-affinity FcγRI preferentially binds IgG2a, medium-affinity FcγRIV binds IgG2a and IgG2b and low-affinity FcγRIIb and FcγRIII bind IgG1, IgG2a and IgG2b (note: mice lack FcγRIIa) [34]. In addition, IgG is glycosylated at asparagine 297 in the Fc region. Deletion of this N-glycan abolishes FcγR binding, and variations in terminal monosaccharide expression (i.e. sialic acid or fucose) affect IgG-FcγR interactions [34]. For example, terminal sialylation promotes an anti-inflammatory effect dependent on FcγRIIb function, potentially explaining some beneficial effects of intravenous immunoglobulin [35]. Finally, exogenous factors influence phagocytosis. For example, pro-inflammatory cytokines (e.g. tumour necrosis factor-α), complement fragment C5a and LPS enhance FcγR expression, thus enhancing phagocytosis [34]. Mouse models are ideal for studying these issues because of the availability of KO mice (e.g. of Fcγ-chain, FcγRIIb, or C5) and of monoclonal antibodies of chosen specificity, isotype, subclass and glycosylation. For example, Fcc-chain KO mice, lacking FcγRI and FcγRIII, are resistant to haemolytic anaemia induced by anti-mouse RBC antibodies [36].

Complement

Before proceeding to alloantibody- and autoantibody-induced RBC clearance models, the role of the complement system will be discussed. Complement-mediated RBC destruction causes intravascular and/or extravascular haemolysis. IgM antibodies, in particular, activate the classical complement pathway to completion, producing intravascular haemolysis [37,38]. Antibodies that only fix complement to the C3 stage produce extravascular haemolysis by RBC phagocytosis in the liver and spleen [37,38]. Complement opsonization is primarily because of C3b (generated by the classic, alternative, or lectin pathways) and C3bi (the factor I cleavage product of C3b) [39]. At least one receptor recognizes C3b (i.e. CR1), and at least three receptors recognize C3bi (CR2, CR3 and CR4) [39]. In addition, the recently described CRIg receptor, only expressed on macrophages, recognizes both C3b- and C3bi-opsonized particles [40].

The CR3 integrin (also known as CD11b/CD18, Mo-1, α_Mβ₂, or Mac-1) mediates phagocytosis of complementopsonized RBCs [41]. CR3 requires prior activation (i.e. ‘inside-out’ signalling) to initiate phagocytosis [42]. Although RBC-binding IgG antibodies may or may not fix complement, complement-fixing IgG antibodies can enhance RBC clearance because of synergism between FcγR- and complement-mediated signalling [43]. To evaluate the role of complement in RBC clearance in mice, various KO strains are available (e.g. C3 KO, C4 KO, C5 KO, or CR3 KO). In addition, cobra venom factor massively activates complement and, therefore, can deplete mice of complement. Novel therapeutics can also be evaluated using these models; for example, CR1-based inhibitors prevent complement-mediated immune haemolysis [44].

Alloantibody-mediated RBC clearance

In contrast to human RBCs, major histocompatibility complex (MHC) antigens are abundantly expressed on mouse RBCs. Despite MHC differences between FVB, C57BL and BALB/c mice, there are no ‘naturally occurring’ anti-MHC antibodies that cause clearance of ‘incompatible’ RBCs [45,46]. Although there are approximately 10 non-MHC mouse blood group systems [45], and although transfused, incompatible mouse RBCs are cleared rapidly [46,47], these polymorphisms have neither been identified at the nucleotide or amino acid level nor exploited extensively. In addition, to date, there is no mouse model of haemolytic disease of the newborn (HDN). Nonetheless, female mice from one inbred strain mated with males from a different strain develop MHC-specific anti-RBC alloantibodies [48]. Although there are no differences in litter size [49], this approach could be explored further. Indeed, recent technological advances in Doppler echocardiography allow measurement of mouse foetal blood velocity [50]. In addition, mouse HDN models could be constructed using the strains described below.

Table 1 summarizes genetically engineered mice that have been, or could be, used to study alloantibody-mediated RBC clearance. Because mouse monoclonal antibodies can be generated by immunizing mice with the antigen of interest (e.g. the MIMA-19 mouse monoclonal anti-human Fy^a was generated using human Fy^b-transgenic mice immunized with Fy^a [51]), these mouse models provide well-controlled systems for testing hypotheses regarding alloantibody-mediated RBC clearance.

hGPA-transgenic model

Human glycophorin A (hGPA) is the most abundant human RBC membrane glycoprotein. The development of hGPA-transgenic mice, expressing hGPA on mouse RBCs [52], led to a mouse model for studying haemolytic transfusion reactions. Thus, non-hGPA-transgenic recipients could be actively immunized to hGPA or passively immunized by infusing IgG or IgM anti-hGPA antibodies. Passive immunization with IgG1 or IgG3 anti-hGPA monoclonal antibodies induces rapid clearance of transfused, incompatible, hGPA-transgenic RBCs [53]. Transfusions into Fcγ-chain and/or complement C3 KO mice show that complement and activating FcγRs are important for clearing incompatible RBCs. IgG-mediated clearance of transfused RBCs induces cytokine storm in this model [19]. In addition, passive immunization with an IgM anti-hGPA monoclonal antibody produces complement-dependent intravascular haemolysis of transfused incompatible RBCs, accompanied by gross haemoglobinuria [53]. Finally, when used to study alloimmunization, this approach suggests that CD4⁺CD25⁺ regulatory T cells are important in regulating the immune response [54].

Fy^b-transgenic model

In a similar model using human Fy^b-transgenic mice [55], wild-type mice transfused with Fy^b-transgenic RBCs developed IgM, and then IgG, alloantibodies to human Fy^b, Fy3 and Fy6. In addition, transfused incompatible Fy^b-transgenic RBCs had decreased survival in alloimmunized mice. Thus, transfusion of antigen-positive RBCs provides a mouse model of alloimmunization, useful for studying alloimmunization mechanisms and the consequences of RBC clearance.

HEL-transgenic model

Surprisingly, transfusion of HEL-transgenic RBCs into wild-type recipients passively immunized with anti-HEL antibodies leads to disappearance of the HEL antigen from the RBC surface rather than RBC clearance, a phenomenon termed ‘antigen loss’ [17,56]. Antigen loss requires the presence of activating receptors, but not the spleen. In addition, HEL-transgenic mice were used to study alloimmunization, which is spleen dependent [22]. Finally, this model led to the discovery that inflammation provoked by certain molecules [e.g. poly(I:C)], but not others (e.g. LPS), enhances alloimmunization to transfused HEL-transgenic RBCs [57,58].

Other models

The HEL-Ovalbumin-Duffy (HOD)-transgenic RBC model is useful for studying alloimmunization because the well-characterized HEL and ovalbumin antigens are tethered to the membrane by Duffy, a multipass transmembrane protein. Transfusion of 14-day stored, leukoreduced HOD RBCs enhances alloimmunization to the HEL antigen when compared to transfusion of freshly collected leukoreduced RBCs [59].

H-transferase transgenic mice express the carbohydrate blood group H antigen on RBCs and have decreased expression of the α-gal xenoantigen [60]. Transfusion of H-transgenic RBCs could potentially be used to model Bombay O incompatible transfusion reactions. Alternatively, α-gal KO mice could be used to model antibody-mediated RBC clearance to a carbohydrate antigen [61], in which the KO mice would be transfused with incompatible, α-gal expressing, wild-type RBCs. Mouse Duffy KO mice could similarly be used to model clearance of wild-type RBCs with KO mice as transfusion recipients [62].

Autoantibody-mediated RBC clearance

Studies using monoclonal antibodies from mice with AIHA provided important insights regarding the roles of antibody isotypes, IgG subclasses, complement and FcγRs in antibody-mediated RBC clearance. Several mouse strains develop AIHA, which has been studied most with New Zealand Black (NZB) mice. Nearly all NZB mice develop RBC autoantibodies and AIHA by 6 months of age [63]. Band 3, the RBC anion exchanger, is the dominant autoantigen; however, deletion of Band 3 in NZB mice only shifts the dominant autoantigen, presumably to glycophorin [64]. Many pathogenic anti-RBC monoclonal autoantibodies were described; IgG subclass switch variants of one of these, 34-3C, were used to assess the roles of IgG subclass, FcγRs, and complement C3 in AIHA in vivo. IgG2a and IgG2b autoantibodies, which interact with FcγRIII and FcγRIV and activate complement, are most potent at inducing AIHA [65].

Although some anti-RBC autoantibodies can induce AIHA in both wild-type and C3 KO mice [66], complement may still influence haemolysis. For example, in addition to the synergism between CR3 and FcγRs in promoting phagocytosis, C5a signalling via its receptor (i.e. C5aR) upregulates activating FcγR expression, thereby exacerbating IgG-mediated AIHA [67]. Complement regulatory proteins on the RBC surface, such as decay-accelerating factor (CD55), also modulate RBC clearance in mouse AIHA by preventing excessive antibody-mediated complement activation [68].

Non-obese diabetic (NOD) mice are another AIHA model; in addition to developing autoimmune insulin-dependent diabetes, they also develop AIHA with age [69]. Interestingly, NZB and NOD mice have reduced expression of the inhibitory FcγRIIb because of a promoter polymorphism [34]. Interleukin-2 KO mice on a BALB/c background also develop severe AIHA causing death by 5 weeks of age [70]; cytokine dysregulation is implicated as a cause of disease in this model [71].

A classical mouse AIHA model involves repeated immunization with rat RBCs [72]. In addition to eliciting rat-specific xenoantibodies, anti-mouse RBC autoantibodies eventually appear. In this model, CD4+CD25+ regulatory T cells prevent autoantibody formation [73]. In addition, implanting hybridomas derived from these mice induce AIHA in naïve mice [74]. In an alternative model, levodopa, a drug for Parkinson’s disease, induces AIHA in humans and in mice [75]. In summary, AIHA occurs in mice ‘naturally’ (e.g. NZB and NOD strains) or after administration of xenogeneic RBCs or a drug. AIHA can also be studied with greater control by implanting anti-RBC hybridomas or by passive immunization with anti-RBC monoclonal autoantibodies.

Mouse models that simplify genetic and pathogenic factors are important for answering questions regarding the initiation of autoimmune disease [76]. An elegant model involves mice transgenic for immunoglobulin genes comprising the 4C8 anti-RBC autoantibody [77]. In this case, almost all B cells are deleted from the periphery; however, B-1 cells in the peritoneum, which are sequestered from RBCs, remain [77]. Oral LPS administration induces B-1 cell expansion and AIHA, whereas a pathogen-free environment prevents AIHA [76]. Thus, infection may be one factor that incites the induction of AIHA.

Non-antibody-mediated RBC clearance

There are multiple non-antibody-mediated causes of RBC clearance. Although antibodies may participate in some models described below, their role is not as clearly defined as in those described above.

Senescence

A complete understanding of why RBCs have a finite lifespan, after which they are cleared by the monocyte-macrophage system, is still not available. Mouse RBCs have a ~55 day lifespan [20]. Because mature RBCs cannot synthesize new proteins, changes in membrane proteins or phospholipids are probably involved in RBC senescence. Haemoglobin degradation with crosslinking to Band 3, a major RBC membrane-spanning protein, thereby generating a senescent antigen recognized by IgG antibodies, may be a mechanism for clearing senescent RBCs [78-80]. In addition, glycosylation differs between young and old RBCs, with decreased sialylation of older RBCs, suggesting that desialylation is involved in senescent RBC clearance [81]. Indeed, desialylated RBCs are rapidly cleared in vivo [82,83] and bind macrophages in vitro [81]. Desialylated RBCs may be cleared by asialoglycoprotein receptor-mediated phagocytosis (this lectin recognizes D-galactose exposed by desialylation) [81] and/or FcγR-mediated phagocytosis because of increased autoantibody binding to senescent RBCs [84]. Other mechanisms for clearing senescent RBCs probably exist, and phosphatidylserine-mediated mechanisms, described below, may also be involved.

Phosphatidylserine (PS) exposure

Cell membranes contain asymmetrically arranged phospholipids, with PS and phosphatidylethanolamine predominantly found on the inner leaflet and phosphatidylcholine and sphingomyelin on the outer leaflet. This asymmetry is maintained by flipase, an ATP-dependent aminophospholipid translocase. Loss of asymmetry in nucleated cells is a marker of apoptosis, leading to recognition and ingestion by macrophages [85,86]. Macrophages interact with PS by a specific PS receptor or through bridging proteins (e.g. annexin I, milk fat globule EGF-factor 8, growth arrest specific gene 6 protein, β2-glycoprotein 1, protein S, thrombospondin and Del-1) that link apoptotic cells to a cognate receptor [86]. PS exposure on the outer leaflet of RBC membranes increases with RBC ageing [87] and oxidative damage [85] and is a signal for splenic and liver clearance [88]. This suggests a role for PS exposure in RBC senescence and RBC apoptosis, the latter termed ‘eryptosis’ [85]. Mechanisms inducing eryptosis include increased cytosolic calcium, producing cell shrinkage by calcium-dependent activation of potassium (‘Gardos’) channels and by calcium-dependent activation of scramblase, thereby exposing PS [85]. These processes are regulated by prostaglandins, platelet activating factor and ceramide. In addition, eryptosis can be modelled by incubating RBCs with a calcium ionophore and calcium, exposing PS [85]. In contrast, when exposed to a calcium ionophore without calcium, RBCs retain normal membrane asymmetry, thereby serving as an appropriate negative control.

Interestingly, murine (and human) sickle cell disease (SCD) RBCs express PS, potentially because of higher ATP consumption, increased reactive oxygen species production and flipase dysfunction [89]. Increased PS exposure results in rapid RBC clearance [90]; thus, PS exposure during SCD crises may exacerbate haemolysis [91]. In addition, uraemia and inflammation increase RBC PS expression and endothelial cell adhesion, which may explain the increased thrombosis risk in these patients [92]. Finally, irondeficient RBCs have increased cytosolic calcium, increased PS expression and decreased survival [93].

Oxidative damage

Oxidative damage is implicated in damaging RBCs in glucose 6-phosphate dehydrogenase (G6PD) deficiency [94]. Mutant mice with low G6PD levels are available for modelling this common human disorder [95]. Oxidation-induced haemolysis of G6PD-deficient RBCs can also be modelled in mice by phenylhydrazine treatment, which causes oxidative damage and precipitates denatured haemoglobin (i.e. Heinz bodies), leading to RBC ingestion by spleen and liver macrophages [96]. Although the receptor(s) responsible for clearing these oxidatively damaged RBCs is unknown, clearance is inhibited by scavenger receptor ligands, such as oxidized low-density lipoprotein, liposome-encapsulated PS, polyinosinic acid and fucoidan; however, there may be different scavenger receptor systems in the liver and spleen [97]. Using scavenger receptor class A KO mice, this receptor was found to be a minor participant in phagocytosis of oxidatively damaged RBCs [97]. Studies in vitro suggest that lectin-like macrophage receptors recognizing galactosyl residues on these damaged RBCs, as well as FcγRs, are also involved in their ingestion [94,98]. Mouse models are also useful in evaluating methods for preventing this clearance; for example, glucocorticoids effectively reduce clearance of oxidatively damaged RBCs in vivo [99].

Oxidative damage of membrane lipids also occurs in several haematological disorders associated with decreased RBC survival [100]. One product of lipid peroxidation, malonyldialdehyde, itself crosslinks membrane lipids and proteins, leading to decreased RBC deformability and increased RBC clearance [100]. This provides yet another useful mouse model for studying the effects of oxidative damage on RBCs.

Xenogeneic RBCs

Transfusions of xenogeneic RBCs have been used for studying various aspects of RBC clearance. For example, sheep RBCs were extensively used to model the immunomodulation seen by using Rh immune globulin in humans to prevent HDN [101]. The mechanisms for clearing xenogeneic RBCs are complex, involving multiple pathways. For example, because of glycosylation differences, surface carbohydrates may be involved in clearing xenogeneic RBCs [102]. In addition, varying glycosylation can activate complement by the alternative and lectin pathways [10]. The presence of ‘naturally occurring’ xenoantibodies is also mechanistically relevant [44]. Finally, size may matter when xenogeneic RBCs are transfused into mice; for example, human RBCs are twice the diameter of mouse RBCs and may have difficulty navigating narrow murine capillaries.

SCD and thalassaemia

A complete review of mouse models of SCD and thalassaemia is beyond the scope of this article. These diseases exhibit decreased RBC survival because of defects in haemoglobin synthesis. Many mouse SCD models were developed, each with advantages and disadvantages [103,104]). Nonetheless, these SCD models have been instrumental in improving our understanding of vascular [105] and pulmonary complications [106] and for evaluating potential treatments [107]. Similarly, multiple mouse thalassaemia models, with various human mutations, are available [108-116].

CD47

In addition to pro-phagocytic ‘eat me’ signals provided to phagocytes by PS exposure, antibody Fc domain clustering and complement activation, there are also anti-phagocytic ‘don’t eat me’ signals. For example, CD47, a ubiquitously expressed integrin-associated protein, functions as a marker of self on RBC surfaces by inhibiting phagocytosis through binding to signal regulatory protein-α (SIRPα) on macrophages [117]. CD47 attenuates phagocytosis of IgG and/or complement-opsonized RBCs via immunoreceptor tyrosine-inhibitory motif-mediated signalling [118]; however, CD47 is not involved in scavenger receptor-mediated phagocytosis [119]. The density of CD47 expression also regulates phagocytosis, because RBCs heterozygous for CD47 are ingested at levels intermediate between RBCs homozygous or null for CD47 [120]. In addition, CD47-deficient NOD mice have accelerated development of AIHA, suggesting that CD47-SIRPα interactions participate in regulating autoimmunity [121]. CD47 levels also decrease during refrigerated RBC storage in vitro [122], potentially increasing the clearance of transfused stored RBCs. CD47 density on human RBCs is similar to that seen on murine RBCs [123]; however, decreased CD47 expression on human RBCs is not associated with decreased RBC survival [124]. There is also decreased avidity of human CD47 for SIRPα when compared to mouse CD47, along with differences at cell-cell contact areas, suggesting an alternative role(s) for CD47 on human RBCs [123].

RBC storage

In general, the mechanisms by which RBC damage during refrigerated storage in vitro (i.e. the ‘RBC storage lesion’) leads to increased clearance in vivo are not yet clear. In addition, controversy remains regarding whether RBC transfusions after prolonged storage (i.e. near to outdate) have adverse consequences in human patients [125]. Questions were also raised regarding the applicability of rodent RBC storage models to the human setting because of a rat RBC storage system in which RBC survival rapidly declined in vivo as storage time increased in vitro [126]. However, more recently, a mouse RBC storage model was established that may be more relevant [20]. In this case, mouse RBCs stored in CPDA-1 for up to 2 weeks exhibited a 24-h RBC survival of ~64%, which was comparable to human RBCs stored in CPDA-1 for 5 weeks. In addition, in the mouse storage model, there was little increase in PS positive RBCs and only minor changes in RBC CD47 expression, which were similar to stored human RBCs. This model may be a useful platform for studying the mechanisms involved, and the consequences of, transfusing stored RBCs.

Conclusions

Mice are the most widely used species for constructing animal models of human disease. Many mouse models are available for studying various issues in transfusion medicine, in general, and mechanisms of haemolysis, in particular. Some examples of unanswered questions that can be studied with these models are shown in Table 2. To date, mouse experiments provided many important lessons regarding both efferent immunity (e.g. antibody-FcγR interactions, the role of antibody isotype and IgG subclass,the mechanisms of complement activation and its control, and the mechanisms and consequences of eryptosis), afferent immunity (e.g. factors regulating antibody response, such as inflammation), innate immunity (e.g. cytokine storm from transfusion of stored RBCs) and RBC physiology. The future holds the potential for many studies that will then translate into improved practices in human transfusion medicine.

Table 2.

Examples of questions amenable to study with mouse models of RBC clearance

Mechanism

Why are subpopulations of incompatible RBCs resistant to   IgG- and/or IgM-mediated RBC clearance?

What signaling pathways produce the cytokine storm induced by   haemolytic transfusion reactions?

Why are some haemolytic transfusion reactions clinically severe,   whereas others are benign?

What host and donor factors influence RBC alloimmunization to   allogeneic transfusions?

Can a mouse model of HDN be constructed to study immunization,   immunomodulation and pathogenesis?

RBC storage

Can new storage solutions be developed to minimize the RBC storage   lesion?

What biomarkers identify RBCs that are more prone to clearance   in vivo after refrigerated storage in vitro?

Why do RBCs from some donors not perform well in vivo after   refrigerated storage in vitro? Is this genetically determined?

Autoimmunity

Why is Band 3 the dominant autoantigen in NZB mice with AIHA?

What is the mechanism(s) for initiating the production of RBC   autoantibodies?

Antigen loss

Does antigen loss occur more commonly than is currently believed?

What is the mechanism(s) responsible for antigen loss?

What factors determine whether RBC antigens are specifically   removed from the RBC surface or whether the intact RBC is   ingested? What are the roles in antigen loss of the antigen, the   antibody, the macrophage and combinations thereof?

Treatment

What therapies can reduce antibody-mediated clearance and   ameliorate the downstream consequences of this clearance?

Is there evidence that any of the following approaches are beneficial:   corticosteroids, intravenous immunoglobulin, other   immunosuppressants and complement-inhibitors?

AIHA, autoimmune haemolytic anaemia, HDN, haemolytic disease of the newborn, NZB, New Zealand Black.