NOG (NOD.Cg-Prkdc^scid Il2rg^tm1Sug/ShiJic) (3) and NOG-C3^ΔMG2-3 (NOD.Cg-Prkdc^scid Il2rg^tm1SugC3^em1/Jic) mice were maintained at the Central Institute for Experimental Animals (CIEA) under specific pathogen-free conditions. To generate NOG-C3^ΔMG2-3 mice, we used the CRISPR/Cas9 system for genome editing (34, 35). Four different guide RNA (gRNA) sequences were designed to target exons 5, 6, and 7 of the complement 3 gene. All gRNAs were cloned into the PX330 plasmid (36) and cleavage activity was confirmed using the reporter construct pCAGGS-EGxxFP (37). Both plasmids were obtained from Addgene (Watertown, MA). The genomic region containing the target exons was inserted in the middle of the EGFP gene, and the plasmid was transfected into HEK293T cells together with each gRNA-PX330 plasmid. Genome-editing success was evaluated via GFP signal rescue after cleavage of the targeted exon. Two gRNAs (gRNA1 and -4) were selected, and a mixture of the two was used for microinjection into fertilized eggs of NOG mice. The sequences of the gRNAs were as follows: gRNA1 targeting exon 5, 5’-CTTGACAGGAATGCCATCGG-3’gRNA4 targeting exon 7, 5’-CATCGATGACCCAAATGGCC-3’.

All studies involving human participants were reviewed and approved by the research ethics committee of the CIEA. Study participants provided written informed consent. All animal experiments were performed in accordance with institutional guidelines (14038, 17024, and 20043) and were approved by the animal experimentation committee of the CIEA.

Murine PB was collected from the retro-orbital venous plexus using heparinized pipettes periodically under anesthesia. BM cells were obtained by flushing femurs with 1 mL phosphate-buffered saline (PBS). PB and BM samples were diluted 10- and 5-fold with PBS, respectively, and 10 μL of the diluted sample was mixed with an equal volume of antibody mixture (described below). After 20–30 min of incubation in the dark at room temperature, the samples were further diluted 15–20-fold with PBS and analyzed on a FACSCanto™ or LSRFortessa™ flow cytometer (BD Biosciences, San Jose, CA). Flow cytometry data were analyzed using FlowJo (ver. 10.7.1, BD Biosciences).

Anti-human CD235a (glycophorin A)-allophycocyanin/cyanine 7 (APC/Cy7), anti-human CD71-APC and anti-mouse TER-119-phycoerythrin (PE) were purchased from BioLegend (San Diego, CA). Fluorescent isothiocyanate (FITC)-labeled goat anti-mouse complement C3 polyclonal antibody was purchased from MP Biomedicals (Santa Ana, CA). Anti-mouse complement component C1q-PE and anti-mouse complement component C4-biotin antibodies were purchased from Cedarlane (Burlington, ON). Streptavidin-PE (StAv-PE) was purchased from BD Bioscience.

C3 levels in the plasma were measured using an anti-mouse complement C3 ELISA kit (Abcam, Cambridge, UK) according to the manufacturer’s instructions.

We collected PB from healthy donors to obtain hRBCs for transfusion. The plasma and buffy coat were removed after centrifugation (400 × g for 5 min). The pellets were washed with PBS, and the RBCs were adjusted to a concentration of 1 × 10¹⁰ cells/mL. Mice were intravenously (i.v.) injected with hRBCs via the tail vein. For long-term experiments, the hRBCs were intraperitoneally (i.p.) injected once a week for maintenance after the initial transfusion.

To reconstitute the human hematopoietic system, 6–8-week-old NOG-C3^ΔMG2-3, NOG/hGM3 Tg, and NOG-C3^ΔMG2-3/hGM3 Tg mice were irradiated with X-rays at 160 cGy (MBR-1520R-4; Hitachi, Hitachi, Japan). Then, 5.0 x 10⁴ or 2.5 × 10⁴ umbilical cord blood-derived CD34⁺ cells (StemExpress, Folsom, CA), for NOG-C3^ΔMG2-3 mice or for NOG/hGM3 Tg and NOG-C3^ΔMG2-3/hGM3 Tg mice, respectively, were transplanted by i.v. injection the next day; hHSC-NOG-C3^ΔMG2-3, hHSC-NOG/hGM3 Tg, and hHSC-NOG-C3^ΔMG2-3/hGM3 Tg mice were thus obtained.

Clo-lip (400 μL/kg; Hygieia Bioscience, Osaka, Japan) and GdCl₃ (30 mg/kg; G7532, Sigma Aldrich, St. Louis, MO) were i.v.-injected into 6–10-week-old NOG or NOG-C3^ΔMG2-3 mice via the tail vein. The injection volume was adjusted to 200 μL with saline. Control mice were injected with 200 μL of saline. Mice were injected with Clo-lip, GdCl₃, or saline four times at 3–4-day intervals and were transfused with hRBCs on the day after the last injection. For long-term experiments, Clo-lip was i.v.-injected as the first treatment to ensure macrophage depletion. Thereafter, GdCl₃ was administered three times by i.v. injection, at 3–4-day intervals. After the first transfusion of hRBCs, GdCl₃ was administered by i.p. injection every 4–6 days for maintenance. For the induction of hRBC from hHSCs, hHSC-transplanted mice were treated with Clo-lip as described above with 3-4-days intervals at 8 weeks after hHSC-transplantation.

The mouse PB (5 μL) was diluted with 50 μL of PBS containing heparin; 3 μL of the solution was mixed with 30 μL of mouse serum from NOG or NOG-C3^ΔMG2-3 mice and incubated for 30 min at 37°C. Subsequently, samples were stained with FITC-labeled goat anti-C3 polyclonal antibody for 20 min at room temperature. Then, the cells were washed with PBS and analyzed using a FACSCanto flow cytometer.

Mice were administered hRBCs (2–5 × 10⁹) by i.v. injection and the PB was collected 3 h after the injection. The blood (10 μL) was immediately transferred to a tube containing 100 μL of cold PBS with 5 IU/mL heparin and 10 mM EDTA (PBS/heparin/EDTA) to prevent complement activation. The diluted blood (10 μL) was stained with APC/Cy7-anti-human CD235a and FITC-anti-mouse C3 polyclonal antibodies, or APC/Cy7-CD235a and biotin-anti-mouse C1q or -C4 antibodies, followed by StAv-PE for 20 min on ice. After staining, the cells were suspended in 100 μL of cold PBS/EDTA and analyzed using flow cytometry.

Tissues from mice were fixed in 10% neutralized formalin (Mildform, FUJIFILM Wako Pure Chemical, Osaka, Japan). Formalin-fixed tissues were embedded in paraffin and analyzed with either hematoxylin and eosin staining or immunohistochemistry using anti-mouse CD68 antibody (Cell Signaling Technology, Danvers, MA) or anti-F4/80 antibody (BIO-RAD, Hercules, CA). Sections were stained using a fully automated BOND-MAX system (Leica Biosystems, Mount Waverley, Australia). The images were captured using a NanoZoomer S60 scanner (Hamamatsu Photonics, Hamamatsu, Japan).

The statistical significance of the results was determined using two-way repeated-measures ANOVA or Mixed-effects analysis using GraphPad Prism software (ver. 9.0; GraphPad Software Inc., San Diego, CA).

To examine the fate of transfused hRBCs in NOG mice, we i.v.-injected fresh hRBCs into NOG mice and investigated their dynamics in the circulation using flow cytometry. We found that hRBCs were rapidly eliminated from the circulation of NOG mice, even when 5.0 × 10⁹ hRBCs (equivalent to 25%–35% of total RBCs) were administered ( Figures 1A, B ). Mouse innate immune cells, especially macrophages, have been implicated in the rapid clearance of hRBCs (24); hence, we investigated the role of mouse complement molecules in the elimination of hRBCs by mouse innate immune cells. To evaluate the deposition of mouse complement molecules on the hRBC surfaces, PB was collected 3 h after hRBC injection and stained with polyclonal anti-mouse C1q, C3, or C4b antibodies. Significant amounts of mouse C3 fragments were detected on the transfused hRBC surfaces, whereas C1q and C4 were modest ( Figure 1C ).

The deposition of mouse C3 on the surfaces of hRBCs indicates that murine C3 may function as an opsonin to facilitate the recognition of hRBCs by mouse innate immune cells. To evaluate the relevance of C3 to hRBC recognition, we disrupted the mouse C3 gene using CRISPR/Cas9. Among five founder mice, one mouse had a 252-nucleotide (nt) deletion resulting in the total exclusion of exon 6 and exclusion of large parts of exons 5 and 7 (nt 614–865; Supplementary Figure 1 ). The resulting 84-amino-acid (aa) truncation (aa 172–255 in pro C3 molecules) corresponded to a part spanning the MG2-MG3 domains in the C3α chain (38). Despite the in-frame deletion, mouse C3 protein was not detected in the plasma by ELISA ( Supplementary Figure 2A ), indicating the absence of mature functional mouse C3 molecules in the circulation. Indeed, when the serum from NOG or mutant (NOG-C3^ΔMG2-3) mice was incubated with hRBCs in vitro, large amounts of mouse C3 or its derivative fragments were deposited on the hRBC surfaces with the NOG mouse serum. By contrast, no signal was detected when hRBCs were incubated with NOG-C3^ΔMG2-3 mouse serum ( Supplementary Figure 2B ), confirming that NOG-C3^ΔMG2-3 are C3-deficient.

We then assessed the survival of transfused hRBCs in NOG and NOG-C3^ΔMG2-3 mice. To this end, we transfused 5.0 × 10⁹ hRBCs into NOG and NOG-C3^ΔMG2-3 mice and evaluated their dynamics in the circulation using flow cytometry. In NOG mice, the frequency of circulating hRBCs decreased dramatically within a few days ( Figures 1 and 2 ). The survival rates of hRBCs were 64.8% ± 11.2% at 1 day post-transfusion (dpt), 36.5% ± 16.3% at 2 dpt, and 7.7% ± 5.9% at 3 dpt; no hRBCs were detected after 4 dpt ( Figure 2B ). The elimination of transfused hRBCs was slower in NOG-C3^ΔMG2-3 mice than in NOG mice, with hRBC survival rates of 88.8% ± 2.0% at 1 dpt, 73.7% ± 1.2% at 2 dpt, and 59.5% ± 3.4% at 3 dpt ( Figure 2B ). Remarkably, 44.0% ± 5.8% of the transfused hRBCs (equivalent to ~12% of the total RBCs in the mouse blood) were maintained at 4 dpt in NOG-C3^ΔMG2-3 mice ( Figure 2A ); a significant amount of hRBCs could still be detected up to 7 days after a single transfusion ( Figure 2B ). The transfusion of a smaller number of hRBCs (2.5 × 10⁸) was also tested. The frequency of hRBCs 30 min after transfusion was 1.15% ± 0.21% in NOG mice and 3.53% ± 0.42% in NOG-C3^ΔMG2-3 mice, which was statistically significant. At 6 hours post transfusion, there were a few hRBCs in NOG mice, whereas 2.43% ± 0.44% of the total RBCs were hRBCs in NOG-C3^ΔMG2-3 mice ( Figure 2C ). The decrease of the survival rate of hRBCs was more rapid in NOG mice than in NOG-C3^ΔMG2-3 mice ( Figure 2D ), suggesting that C3-dependent elimination mechanisms are independent of the number of hRBCs. The prolonged survival of human RBCs in NOG-C3^ΔMG2-3 mice was observed irrespective of the donor blood type because we obtained similar results with different donor blood types (A-type, O-type and AB-type, data not shown).

Despite the extended hRBC survival in NOG-C3^ΔMG2-3 mice, the survival period of erythrocytes was significantly shorter than their natural lifespan. Hence, we depleted mouse macrophages by treating NOG and NOG-C3^ΔMG2-3 mice with Clo-lip, which has been previously shown to extend hRBC survival. After four Clo-lip injections at 3–4-day intervals, we transfused hRBCs into NOG and NOG-C3^ΔMG2-3 mice. Consistent with the previous reports, Clo-lip extended the survival of hRBCs in NOG mice. We detected 34.7% ± 16.8% and 13.5% ± 12.1% of the initial amount of hRBCs at 4 and 7 dpt, respectively ( Figures 3A, B ). The half-life of hRBCs extended from 1–2 days to 2–3 days in NOG mice; nevertheless, the difference between Clo-lip-treated NOG mice and saline-treated control mice was not statistically significant ( Figure 3B ). By contrast, the extension of hRBC survival was profound in NOG-C3^ΔMG2-3 mice. The survival rates of hRBCs were 71.9% ± 5.6% at 7 dpt, 38.9% ± 12.0% at 13 dpt, and 18.9% ± 10.1% at 21 dpt ( Figure 3C ); hRBC half-life was extended from 3–4 days to 10–13 days ( Figure 3D ). Notably, a significant number of hRBCs could be detected up to 1 month after a single transfusion of hRBCs.

Despite the prolonged hRBC survival in NOG and NOG-C3^ΔMG2-3 mice after Clo-lip treatment, Clo-lip caused significant toxicity. Two of four NOG mice and two of five NOG-C3^ΔMG2-3 mice died during the experiments ( Figure 3 ), and the remaining mice exhibited aberrant phenotypes, including ruffled hair, weight loss, and a hunched posture (data not shown). Hence, alternative chemicals are required to either deplete or suppress mouse macrophages.

GdCl₃ has been shown to deplete or suppress macrophages in rats (31) and mice (33). In this study, we administered 30 mg/kg of GdCl₃ four times at 3–4-day intervals. This dosing was determined based on preliminary results from titration experiments (data not shown). Immunohistochemistry showed that Clo-lip induced severe depletion of F4/80-positive macrophages in the spleen and liver of NOG mice ( Figure 4 ). Recovery of hepatic macrophages was detected at 7 days post Clo-lip treatment, and a few macrophages were detected in the spleen. In contrast to Clo-lip, GdCl₃ did not deplete macrophages in the liver of NOG mice, although there were some morphological changes such as swelling ( Figure 4A ). GdCl₃ also induced a mild reduction of macrophages in the spleen ( Figure 4B ). Notably, GdCl₃ treatment did not cause significant toxicity, in sharp contrast to Clo-lip. GdCl₃-treated mice did not develop overt phenotypic abnormalities, and, although the mice experienced a slight weight loss early after GdCl₃ treatment, mouse weight returned to physiological levels at 1 week after treatment ( Supplementary Figure 3 ).

Importantly, GdCl₃ treatment of NOG and NOG-C3^ΔMG2-3 mice before hRBC transfusion significantly extended the survival of hRBCs. In NOG mice, hRBC half-life was extended from 1–2 days to approximately 4 days after transfusion ( Figures 5A, B ). The extension of hRBC survival was confirmed in NOG-C3^ΔMG2-3 mice, where GdCl₃ increased hRBC half-life from 2–4 days to nearly 8 days. At 12 dpt, 18.3% ± 6.6% of the initial hRBC count was retained, which was equivalent to ~5% of the total RBC count in the mouse blood ( Figures 5C, D ). The effects of GdCl₃ were also confirmed in NOG mice with a smaller number of hRBCs (2.5 × 10⁸) ( Supplementary Figure 4 ).

Notably, GdCl₃ did not cause significant toxicity in NOG or NOG-C3^ΔMG2-3 mice, and no deaths occurred during the experiments (data not shown).

After confirming prolonged survival of hRBCs in NOG-C3^ΔMG2-3 mice, multiple injections of hRBCs and GdCl₃ were administered to maintain the level of hRBCs in the circulation for an extended period. To ensure initial elimination of mouse macrophages, Clo-lip was administered as the initial treatment, followed by three GdCl₃ treatments at 3–4-day intervals. After pretreatment, 5.0 × 10⁹ hRBCs were administered to NOG mice on day 0, and GdCl₃ was administered on days 2, 5, and 8 by i.p. injection. Consecutive administration of GdCl₃ did not prolong the survival of hRBCs in NOG mice. The additional administration of 5.0 × 10⁹ hRBCs by i.p. injection on day 5 transiently increased the amount of hRBCs on the following day; however, they rapidly decreased thereafter ( Supplementary Figure 5 ). NOG-C3^ΔMG2-3 mice showed better results. After pretreatment, 5.0 × 10⁹ hRBCs were administered both by i.v. and i.p. injections on day 0, to increase the initial loading amount. Thereafter, GdCl₃ was i.p.-injected every 4–6 days, and 5.0 × 10⁹ hRBCs were administered by i.p. injection once a week for 3 weeks. The proportion of hRBCs in PB was around 25% after the first injection of hRBCs, and gradually increased with every hRBC injection ( Figure 6 ). The proportion of hRBCs in two mice reached nearly 100% by day 15 and was maintained for 2 weeks. The other two mice showed about 80% chimerism by day 28. During the experiment, one mouse died on day 20; all of the other mice were healthy. These results indicate that NOG-C3^ΔMG2-3 mice are significantly superior with respect to the maintenance of hRBCs compared with NOG mice.

Long-term maintenance of transfused hRBCs in NOG-C3^ΔMG2-3 mice allowed for examination of the differentiation of hRBCs from hHSCs. Our preliminary experiments suggested that Clo-lip, but not GdCl₃, induced a few hRBCs (not greater than 1% in PB) in hHSC-NOG-C3^ΔMG2-3 mice. Thus, we used NOG-C3^ΔMG2-3/hGM-CSF-IL-3 (hGM3) Tg mice to enhance erythropoiesis and human myelopoiesis (15). The proportion of hRBCs in some hHSC-NOG-C3^ΔMG2-3/hGM3 Tg mice reached approximately 8-10% in PB after three-time injections of Clo-lip. However, nearly half of the mice both in the NOG-C3^ΔMG2-3/hGM3 Tg and hHSC-NOG/hGM3 Tg groups died during the experiments, resulting in a large variance and the absence of a statistical significance between these two groups ( Figure 7A ). The BM analysis showed an enhanced development of human erythrocytes in the NOG-C3^ΔMG2-3/hGM3 Tg and hHSC-NOG/hGM3 Tg mice ( Figure 7B ). About 40% of them were CD71⁺ CD235a⁺ immature erythrocytes, indicating the erythropoiesis from hHSC in the BM ( Figure 7C ).