Combined liver–cytokine humanization comes to the rescue of circulating human red blood cells

Abstract

In vivo models that recapitulate human erythropoiesis with persistence of circulating red blood cells (RBCs) have remained elusive. We report an immunodeficient murine model in which combined human liver and cytokine humanization confer enhanced human erythropoiesis and RBC survival in the circulation. We deleted the fumarylacetoacetate hydrolase (Fah) gene in MISTRG mice expressing several human cytokines in place of their murine counterparts. Liver humanization by intrasplenic injection of human hepatocytes (huHep) eliminated murine complement C3 and reduced murine Kupffer cell density. Engraftment of human sickle cell disease (SCD)–derived hematopoietic stem cells in huHepMISTRGFah^−/− mice resulted in vaso-occlusion that replicated acute SCD pathology. Combined liver–cytokine–humanized mice will facilitate the study of diseases afflicting RBCs, including bone marrow failure, hemoglobinopathies, and malaria, and also preclinical testing of therapies.

Human red blood cells (huRBCs), one of the most common cell types in the body, have been under intense genetic selection throughout human evolution, the deleterious consequences of which place a heavy burden on many human populations and health care systems (1). Species-specific differences in the erythropoietic program (2) demand in vivo models of human erythropoiesis (3), yet no humanized immunodeficient mouse models exist with persistent mature huRBCs in peripheral blood (PB), the effective readout of functional human erythropoiesis and RBC integrity (4-6). Comprehensive analysis of hematopoietic and RBC disorders, such as sickle cell disease (SCD), require mature RBCs in the circulation both for disease manifestations and to test therapies.

Liver humanization rescues human RBCs in circulation

MISTRG mice carry knock-ins for the human cytokines granulocyte-monocyte and macrophage colony-stimulating factor (M-CSF), interleukin-3 (I), thrombopoietin (T), and for signal regulatory protein alpha (S), the receptor for the “don’t eat me” signal regulatory protein CD47, in the Rag2^−/−Il2rg^−/− (RG) background (7, 8). Engraftment of MISTRG mice with fetal liver (FL) CD34⁺ cells resulted in robust leukocyte (huCD45⁺) engraftment in PB (~30%) and bone marrow (BM) (~50%) 8 to 12 weeks after intrahepatic injection into newborn livers. Erythroid (huCD235a⁺) engraftment in BM was modest (~5%), and huRBCs were absent in PB (fig. S1, A to C), suggesting destruction of huRBCs in the circulation, as previously described (4-6). To determine the kinetics and sites of huRBC destruction, we injected fluorescently labeled human and murine (mu) RBCs into MISTRG mice (fig. S1D) and confirmed rapid preferential clearance of huRBCs from circulation (Fig. 1A and fig. S1E). Infused labeled huRBCs but not muRBCs accumulated in liver and spleen, reaching a peak at 60 min after infusion (Fig. 1B and fig. S1F), suggesting sequestration and destruction by murine phagocytes. Previous studies have shown that huRBCs in the murine host are coated with murine complement 3 (muC3), which likely targets huRBCs to murine macrophages and other phagocytes (9). The huRBCs but not muRBCs infused into MISTRG mice were rapidly coated with muC3 (Fig. 1C). huRBCs isolated from the liver costained with the murine macrophage marker F4/80⁺ (Fig. 1D), suggesting sequestration of huRBCs by murine liver phagocytes. We performed intravital imaging of the liver immediately after injection of fluorescently labeled human or murine RBCs together with dextran to define blood vessels (fig. S1G). We observed that huRBCs (movie S1), but not muRBCs (movie S2), became rapidly and persistently trapped within the liver vasculature. Thus, the mouse liver represents one of the major sites of huRBC sequestration and destruction.

Fig. 1. Liver humanization prolongs the survival of infused huRBCs in circulation through elimination of muC3 and reduction of murine macrophages.

Equal numbers of fluorescently labeled huRBCs (labeled with carboxyfluorescein diacetate succinimidyl ester) and muRBCs (violet) were premixed and injected retro-orbitally into mice. (A and B) Ratios of infused huRBCs to muRBCs in PB (n = 15) (A) and in liver, spleen, lung, and BM (B) at designated time points (n = 8). (C) muC3 staining of infused huRBCs (red) and muRBCs (blue) 5 min after infusion into MISTRG mice (n = 11). APC, cells were stained with allophycocyanin. (D) Representative flow cytometry plots and histograms of murine F4/80 staining of infused huRBCs (red) and muRBCs (blue) collected from MISTRG livers at 60 min after infusion (n = 8). (E) Comparison of huRBC survival in PB of MISTRGFah (n = 6), muHepMISTRGFah (n = 8), and huHepMISTRGFah (n = 8) mice. (F) Representative flow plots and histograms of muC3 staining of infused huRBCs and muRBCs in PB at 5 min and 60 min after infusion. (G and H) Representative histologic images showing F4/80⁺ murine macrophages (G) and quantification of F4/80⁺ area (H) in the livers of muHepMISTRGFah (n = 6) and huHepMISTRGFah (n = 6) mice at 12 weeks after hepatocyte engraftment. Scale bars, 100 μm; original magnification, 10×. Histological images were quantified with ImageJ, and individual mice are represented by symbols. Data (means ± (SEM) are representative of three independent experiments with three different donors. P values were determined by Mann-Whitney U test: **P < 0.01; ***P < 0.001; ****P < 0.0001.

The liver produces numerous proteins essential for the innate immune defense, including complement, and contains a large number of tissue macrophages (10, 11). We deleted the fumarylacetoacetate hydrolase (Fah) gene by using CRISPR-Cas9 in MISTRG to generate MISTRGFah^−/− (MISTRGFah) mice and confirmed deletion of Fah by protein immunoblotting (fig. S2A). We successfully humanized the liver in MISTRGFah mice with adult huHep as previously described (12, 13) by intrasplenic injection of adult human hepatocytes and gradual withdrawal of 2-(2-nitro-4-trifluoromethylbenzoyl)-1,3-cyclohexanedione (NTBC) in the drinking water, resulting in the death of Fah^−/− murine hepatocytes and regeneration with human hepatocytes. Murine hepatocyte (muHep)–engrafted MISTRGFah mice (muHepMISTRGFah) served as controls. Human plasma albumin (huALB) concentrations reached a plateau around 8 weeks after huHep engraftment (fig. S2B). Human albumin concentrations were validated as surrogate measurements for human hepatocyte regeneration, with levels above 4 mg/ml indicating that 80 to 95% of hepatocytes were human as determined by FAH histology (fig. S2D). To determine the effects of liver humanization on huRBC survival, we repeated infusion of fluorescently labeled RBCs and compared huHepMISTRGFah and muHepMISTRGFah with MISTRGFah mice (fig. S1D). Survival of huRBCs was significantly prolonged in the PB of huHepMISTRGFah mice compared with those in MISTRGFah and muHepMISTRGFah mice, with persistence of huRBCs for >72 hours (Fig. 1E). Only a small percentage of huRBCs in PB of huHepMISTRGFah mice were coated with muC3, compared with the high muC3 positivity of huRBCs in muHepMISTRGFah mice (Fig. 1F). Infused huRBCs isolated from the livers of huHepMISTRGFah mice no longer costained with murine F4/80⁺, suggesting reduced targeting of huRBCs to murine phagocytes in the liver (fig. S2, E and F). In addition, the overall numbers of F4/80⁺ macrophages in huHepMISTRGFah livers were significantly decreased in comparison with those in muHepMISTRGFah livers (Fig. 1, G and H). Thus, liver humanization prolongs the survival of infused huRBCs and decreases muC3 concentrations, murine phagocyte numbers in the liver, the coating of huRBCs with muC3, and their subsequent destruction.

huRBCs circulate in engrafted huHepMISTRGFah mice

To determine whether liver humanization would spare huRBCs produced within the murine host, we measured huRBCs in circulation in human FL hematopoietic stem and progenitor cell (HSPC)–engrafted, liver-reconstituted huHepMISTRGFah and muHepMISTRGFah mice (fig. S3A). The huHepMISTRGFah mice showed higher human leukocyte (huCD45⁺) contributions to PB cells at 6, 8, and 10 weeks (fig. S3B). Additionally, huHepMISTRGFah mice displayed persistent circulating huRBCs (huCD235a⁺) that represented up to 12% of the total RBCs in PB as early as 4 weeks and until at least 12 weeks after engraftment The huRBCs remained absent in muHepMISTRGFah mice (<0.1% of total RBCs) (Fig. 2, A and B). Qualitative analysis of huRBCs, which was somewhat limited in muHepMISTRGFah mice because of the very low numbers of circulating huRBCs, showed improved maturity with a shift toward a huCD71⁻huCD235⁺ (Fig. 2, C and D) and huCD49d⁻Band3⁺ (Fig. 2, E and F) phenotype and near-complete enucleation of circulating huRBCs (Fig. 2, G and H) in huHepMISTRGFah mice. Neither huHepMISTRGFah mice nor muHepMISTRGFah mice harbored circulating huRBCs positive for surface muC3, consistent with rapid clearance of C3-coated RBCs by the innate immune system (fig. S3C). Lack of muC3 coating and survival of huRBCs in huHepMISTRGFah mice suggested that liver humanization abrogated muC3 expression. Indeed, muC3 protein concentrations in plasma were significantly reduced in huHepMISTRGFah mice (fig. S3D). Absent muC3 expression (fig. S3E) with acquired huC3 expression (fig. S3F) confirmed the proposed mechanism by liver humanization in huHepMISTRGFah mice. Liver macrophages are composed of liver-resident Kupffer cells and BM-derived recruited macrophages (14). We assessed whether liver humanization would alter liver macrophage composition by preferentially recruiting human HSPC-derived macrophages. Indeed, liver regeneration by human hepatocytes in huHepMISTRGFah mice resulted in replacement of murine F4/80⁺ macrophages by human CD68⁺ macrophages, whereas muHepMISTRGFah mice still contained a significant number of murine macrophages in addition to human macrophages (fig. S3, G to I). Thus, huRBCs are spared from destruction in huHepMISTRGFah mice, at least in part by elimination of complement-mediated opsonization and their recognition by the murine innate immune system.

Fig. 2. Enucleated, mature huRBCs circulate in huHepMISTRGFah mice.

(A and B) Representative flow cytometry plots (A) and quantitation (B) of human (huCD235⁺) and murine (muTer119⁺) erythroid cells in PB of muHepMISTRGFah (n = 8) versus huHepMISTRGFah (n = 13) mice. PE-Cy7, phycoerythin-Cy7 fluorophore. (C to F) Representative flow cytometry analysis and quantitation of human erythropoietic differentiation on the basis of huCD71 and huCD235a expression (C and D) and huCD49d and Band3 expression (E and F) in PB of muHepMISTRGFah (n = 8) versus huHepMISTRGFah (n = 13) mice. PacBlue, Pacific blue label; PerCP Cy5.5, peridinin chlorophyll protein Cy5.5. (G and H) Representative flow cytometry plots (G) and quantitation (H) of enucleated human RBCs, determined by Hoechst staining of the PB of muHepMISTRGFah (n = 8) versus huHepMISTRGFah (n = 13) mice. FSC-W, forward scatter width. Individual mice are represented by symbols; data are presented as means ± SEM. P values were determined by Mann-Whitney U test: n.s., not significant; *P < 0.05; **P < 0.01; ****P < 0.0001. Data are representative of three independent experiments with three different donors of hepatocytes and FL cells.

Liver humanization enhances human erythropoiesis

Liver humanization prolonged circulating huRBC survival. To determine whether liver humanization also affected generation of huRBCs, we analyzed human erythropoiesis within the BM. Overall, huCD45⁺ engraftment was higher in huHepMISTRGFah than in muHepMISTRGFah mouse BM (fig. S4, A to C) without significant differences in the distribution of myeloid (huCD33⁺), B lymphoid (huCD19⁺), and T lymphoid (huCD3⁺) lineages (fig. S4D). Enhanced engraftment levels translated across all progenitor subpopulations (fig. S4, E to K). Human erythroid engraftment (huCD235a⁺) was significantly higher in huHepMISTRGFah BM (Fig. 3, A to C) and also in the spleen (fig. S5, A to C) with progressive down-regulation of huCD71, acquisition of huCD235a expression (Fig. 3, D and E), down-regulation of huCD49d, and acquisition of Band3 (15) (fig. S5, D and E) in the BM. Thus, liver humanization in huHepMISTRGFah improved human erythroid engraftment and maturation. We hypothesized that reduction in muC3 expression in huHepMISTRGFah mice would also spare human erythroid progenitors and RBCs in the BM. Indeed, muC3 staining was very low in huHepMISTRGFah mice, whereas nearly 100% of human erythroid progenitors within the BM of muHepMISTRGFah were coated with muC3 (fig. S5, F and G).

Fig. 3. Liver humanization enhances human erythropoiesis.

(A and B) Representative flow cytometry plots (A) and quantitation (B) of human erythroid progenitors in BM of muHepMISTRGFah (n = 8) and huHepMISTRGFah (n = 13) mice. (C) Representative BM histology images of huCD235 staining from muHepMISTRGFah (n = 8) and huHepMISTRGFah (n = 13) mice. Scale bars, 100 μm; original magnification, 10×. (D and E) Representative flow cytometry plots (D) and quantitation (E) of erythroid lineage differentiation determined on the basis of huCD71 and huCD235 expression in BM of muHepMISTRGFah (n = 8) versus huHepMISTRGFah (n = 13) mice. GPA, glycophorin A. (F) Gating strategy in multispectral imaging flow cytometry of human EBIs in the huHepMISTRGFah mouse. BF, bright field. (G) Representative images of human EBIs. Enucleated reticulocytes in contact with the central macrophage are marked by white arrows. Scale bars, 10 μm; original magnification, 40× Individual mice are represented by symbols, and data are means ± SEM. P values were determined by Mann-Whitney U test: *P < 0.05; **P < 0.01; ****P < 0.0001. Data are representative of three independent experiments with three different donors of hepatocytes and FL cells.

Before the exit of mature RBCs from the bone marrow, their condensed nuclei are removed by macrophages within so-called erythroblastic islands (EBIs), composed of a central macrophage surrounded by erythroid precursors at varying stages of maturation (16, 17). Expression of human M-CSF in MISTRG mice is required for full maturation of monocytes to tissue macrophages (7, 18). We assessed formation of EBIs by multispectral imaging flow cytometry (19) in BM of huHepMISTRGFah mice. EBIs contained numerous fully human EBIs characterized by huCD169⁺huCD14⁺ double-positive central macrophages surrounded by huCD235a⁺ erythroid progenitors of varying sizes, including enucleated huRBCs (Fig. 3F). Quantitation of enucleation by Hoechst staining indicated equal percentages of enucleated human red cells in the BM of both muHepMISTRGFah and huHepMISTRGFah mice (fig. S5, H and I).

Thus, liver humanization increases overall erythroid engraftment and maturation but does not affect terminal enucleation. However, enucleated huRBCs that exit the BM are protected from destruction by the lack of muC3 expression and reduction in murine liver phagocytes in huHepMISTRGFah mice.

Modeling SCD in huHepMISTRGFah mice

Single-nucleotide point mutations in the globin β chain in SCD cause polymerization of hemoglobin under low oxygen tension that results in progressive damage to the RBC membrane. Vaso-occlusion and tissue hypoxia ensue, because of the adhesion of sickle RBCs in postcapillary venules and selective trapping of dense sickled RBCs enhanced by additional factors, such as endothelial activation and leukocyte adherence (20, 21). To determine whether enhanced erythropoiesis and, most notably, circulating huRBCs would be sufficient to replicate SCD in huHepMISTRGFah mice, we engrafted HSPCs from adult SCD patients and age-matched controls into huHepMISTRGFah mice (fig. S6, A and B). Overall, huCD45⁺ engraftment, lineage distribution, and erythroid engraftment in BM (fig. S6, C to E) and PB (fig. S6, F to G) were similar between mice engrafted with normal or SCD HSPCs. Both normal and SCD HSPC-engrafted mice developed mild anemia as previously described (7) (fig. S6, H to E). We observed sickling huRBCs (7 to 11 sickling huRBCs per 100 huRBCs) in the PB of SCD HSPC-engrafted mice (Fig. 4A) but not in normal HSPC-engrafted mice (Fig. 4B). Both SCD (Fig. 4C) and normal (Fig. 4D) HSPC-engrafted mice exhibited EBIs in their respective BM. To determine whether circulating sickle huRBCs would result in vaso-occlusion and associated findings, we compared histological sections of lung, liver, spleen, and kidney from normal and SCD HSPC-engrafted mice. Tissues from SCD HSPC-engrafted mice showed significant abnormalities consistent with SCD vaso-occlusion caused by huRBCs, identified by human-specific anti–hemoglobin A antibody staining (Fig. 4, E to H), which were absent in normal HSPC-engrafted mice. Lungs showed an increase in alveolar macrophages (fig. S7, A and B) associated with alveolar hemorrhage and thrombosis (Fig. 4E). Spleens showed erythroid precursor expansion, sickled erythrocytes in the sinusoids, and vascular occlusion and thrombosis (Fig. 4F and fig. S7, C and D). Liver architecture was disrupted in SCD-engrafted mice, with RBCs in sinusoids and microvascular thromboses (Fig. 4G and fig. S7, E and F). Congestion of capillary loops and peritubular capillaries and glomeruli engorged with sickled RBCs were evident in kidneys (Fig. 4H and fig. S7, G and H) of SCD but not of normal HSPC-engrafted mice.

Fig. 4. Modeling sickle cell disease in huHepMISTRGFah mice.

(A and B) Representative multispectral imaging flow cytometry images of mature enucleated huRBCs (huCD235a⁺, Hoechst⁻) and muRBCs (muTer119⁺, Hoechst⁻) in the PB of SCD (A) and normal (B) BM CD34⁺ cell–engrafted huHepMISTRGFah mice (n = 3). (C and D) Representative images of human EBIs in the BM of SCD (C) and normal (D) BM CD34⁺ cell–engrafted huHepMISTRGFah mice (n = 6 for each BM). (E to H) Representative histologic images of tissues stained with human-specific anti–hemoglobin alpha (HBA) antibody from SCD (n = 6) and normal (n = 6) BM CD34⁺ cell–engrafted huHepMISTRGFah mice. Scale bars, 100 μm; original magnification, 20× Arrowheads in (E) mark lung alveolar hemorrhages and thrombosis, in (F) they mark erythroid precursor expansion, sickled erythrocytes in sinusoids, vascular occlusion, and thrombosis in the spleen, in (G) they mark sickled RBCs in sinusoids and microvascular thrombosis in liver, and in (H) they mark congestion of capillary loops and peritubular capillaries and engorged glomeruli in kidneys of SCD-engrafted huHepMISTRGFah mice. Data are representative of two independent experiments with two different donors of normal and SCD CD34⁺ cells.

Thus, this model allows the successful engraftment of adult HSPCs and establishment of robust erythropoiesis from both normal and SCD patients. Moreover, circulating human RBCs from a sickle cell donor replicate the hallmark of SCD, vaso-occlusion in hematopoietic and nonhematopoietic organs in the murine host.

Discussion

Immunodeficient mouse models have historically been poor hosts for human erythropoiesis. Despite recent improvements of the BM niche by human cytokine knock-in (6, 7) or mutation of the murine host’s stem cell factor receptor (4, 5), huRBCs in PB are universally absent. Elimination of phagocytes by liposomal clodronate results in transient circulation of huRBCs, confirming the involvement of the innate immune system (4, 5, 22). Liposomal clodronate is toxic with significant mortality and provides only temporary benefit due to regeneration of phagocytes within 1 to 2 weeks. A recent study showed that muC3 mediates adherence of huRBCs to mouse phagocytes, yet in vivo depletion of muC3 only improves huRBC survival when phagocytes are concurrently abrogated by clodronate treatment (9). In our studies, we identified the mouse liver as the major site of huRBC sequestration. Liver humanization significantly reduced muC3 synthesis and resulted in reduced density of murine macrophages in the regenerated liver in human HSPC–engrafted huHepMISTRGFah mice. Therefore, liver humanization, likely by elimination of muC3 and reduction of liver Kupffer cells, results in persistent enucleated, mature huRBCs in circulation.

Human HSPC–engrafted huHepMISTRGFah mice show enhanced human erythropoiesis. Unlike all other humanized immunodeficient mouse models, MISTRG mice express human M-CSF, which allows mature functional resident tissue macrophages to populate host tissues (6, 7). We determined that a subset of human macrophages in huHepMISTRGFah mice express the central macrophage marker CD169⁺ (19, 23, 24) and are found within EBIs in close contact with human erythroid progenitors of several differentiation stages. Human erythroid precursors in BM of huHepMISTRGFah mice also lack coating with muC3. Hence, liver humanization combined with cytokine humanization enables study of the role played by the central macrophage in human erythropoiesis in health and disease in an immunologically advantageous context.

SCD is an inherited blood disorder caused by a single point mutation in the β-globin gene, mutating the sixth amino acid glutamine to valine (25). The most common murine models of SCD (26) exclusively express human globins in muRBCs in the background of murine globin knockouts (27-29). Mouse models have notably contributed to the elucidation of many of the mechanisms of SCD pathology. However, they exclusively contain muRBCs and fail to capture the genetic heterogeneity encountered in patients, making a flexible model of human SCD highly desirable. We successfully engrafted SCD HSPCs in huHepMISTRGFah mice and detected circulating, sickling huRBCs in the mouse PB. In addition, we observed pathological changes in the lung, spleen, liver, and kidney that were comparable to changes in patients (30-33) and established SCD mouse models (27, 29).

We herein present our huHepMISTRGFah mouse model with enhanced reconstitution of human erythropoiesis and, more notably, mature circulating huRBCs. Our findings highlight the potential of this model for use in studies of the numerous life-threatening RBC diseases, which involve ≈5% of the population worldwide (34), including alpha and beta thalassemia (35) and SCD (25). Moreover, this model may be important in studies of hematopoietic stem cell diseases, such as myelodysplasia; in erythroleukemia; and in pathologies that intricately link RBCs and the liver, such as malaria. This mouse model may therefore open additional avenues for the study of disease pathophysiological mechanisms and the preclinical screening of therapeutics.