Genetic loss of Tmprss6 alters terminal erythroid differentiation in a mouse model of β-thalassemia intermedia

In β-thalassemia intermedia (TI), circulating levels of the hepcidin hormone are inappropriately low in relation to hepatic iron stores, promoting excessive absorption of dietary iron. Additionally, precipitation of unpaired α-globin chains causes oxidative damage that increases erythroid precursor (EP) apoptosis and shortens erythrocyte lifespan.1 TMPRSS6 is a hepatic protease that inhibits bone morphogenetic protein (BMP) signaling for hepcidin production.2 In the Hbb^th3/+ TI mouse model, abolishing or reducing expression of TMPRSS6 has been shown to attenuate iron loading and improve anemia.53 Here, we define the consequences of genetic loss of Tmprss6 on erythroid maturation in Hbb^th3/+ mice on a C57BL/6 genetic background. We show that Tmprss6 loss in Hbb^th3/+ mice decreases mean corpuscular hemoglobin (MCH), decreases membrane-bound globins in erythrocytes, attenuates erythroid expansion in bone marrow (BM) and spleen, and produces a terminal erythroid maturation profile indistinguishable from Tmprss6 mice. Additionally, Tmprss6 loss in Hbb^th3/+ mice decreases apoptosis and reactive oxygen species (ROS) in late EP and erythrocytes, raises peripheral red blood cell (RBC) count, decreases reticulocytes, and decreases renal and hepatic Epo mRNA, indicating greater efficacy of erythropoiesis. These findings demonstrate that genetic loss of Tmprss6 determines terminal erythroid differentiation in Hbb^th3/+ mice.

By breeding Tmprss6 and Hbb^th3/+ mice of C57BL/6 genetic background, we found that homozygous Tmprss6 loss in Hbb^th3/+ mice produced organ iron deficiency, limited splenomegaly, raised hepatic expression of hepcidin (Hamp) and other BMP target genes (Id1, Atoh8), and lowered serum iron and transferrin saturation (Online Supplementary Table S1 and Online Supplementary Figures S1-S4). Organ non-heme iron, serum iron, and transferrin saturation levels were similar in Hbb^th3/+Tmprss6 and Hbb^+/+Tmprss6^−/− mice, and hepatic Hamp, Id1, and Atoh8 in Hbb^th3/+Tmprss6 mice matched or exceeded levels in Hbb^+/+Tmprss6^−/− mice. Heterozygous Tmprss6 loss only modestly altered iron-related parameters of Hbb^th3/+mice.

Hemoglobin and hematocrit levels were not improved by homozygous Tmprss6 loss in Hbb^th3/+ mice (Figure 1A and B and Online Supplementary Tables S1-S3). However, homozygous Tmprss6 loss in Hbb^th3/+ mice altered erythrocyte morphology (Online Supplementary Figure S5A), raised RBC, and lowered mean corpuscular volume (MCV), MCH, red cell distribution width, and reticulocytes (which likely contributes to the lower MCV) (Figure 1C-F and Online Supplementary Tables S1-S3). While Hbb^th3/+Tmprss6 spleens showed disrupted follicular architecture with red pulp EP expansion, follicular architecture was retained in Hbb^th3/+Tmprss6 spleens (Online Supplementary Figure S5B). Red cell indices, erythrocyte morphology, and splenic histology of Hbb^th3/+Tmprss6^−/− and Hbb^+/+Tmprss6^−/− mice were similar. Erythrocyte α-globin precipitates were moderately and severely reduced by heterozygous and homozygous Tmprss6 loss in Hbb^th3/+ mice, respectively (Figure 1G), despite persistent globin chain mRNA imbalance in BM and spleen (Online Supplementary Figure S6A and B).

Figure 1.Genetic loss of Tmprss6 in Hbb^th3/+ mice produces red blood cell indices consistent with iron-restricted erythropoiesis and reduces membrane-bound globins in erythrocytes. Shown for 8-week-old female mice of selected Hbb-Tmprss6 genotypes are mean values obtained from analyses of hemoglobin level (A), hematocrit (B), RBC (C), MCV (D), MCH (E), and reticulocyte count as a percentage of total red blood cells (F). Error bars represent SD. *P<0.05; **P<0.005. (G) Shown is analysis of membrane-bound globins prepared from peripheral red blood cells of mice of selected Hbb-Tmprss6 genotypes (males, 8-9 weeks old). Globins were fractionated by triton-acetic acid-urea gel electrophoresis and visualized by Coomassie stain. Sample loading was adjusted by RBC (from complete blood count), so that 1.5 × 10⁸ erythrocytes are represented per lane. Std, globin standard. For panels A-E, 7 Hbb^+/+Tmprss6^+/+, 9 Hbb^+/+Tmprss6^−/−, 7 Hbb^th3/+Tmprss6^+/+, and 6 Hbb^th3/+Tmprss6^−/− mice were analyzed. Reticulocytes were counted in 3-5 female mice per genotype. Mice were fed a 270 p.p.m. iron diet in panels A-F and a 200 p.p.m. iron diet in panel G.

By an established method that characterizes stages of erythroid maturation,6 the percentage of total cells that correspond to EP (i.e. cells in regions I-IV) was higher in Hbb^th3/+Tmprss6^+/+ mice than wild-type in both BM (55% vs. 31%; P<0.0005) (Figure 2A) and spleen (47% vs. 3%; P<0.00005) (Online Supplementary Figure S7A). Tmprss6 loss in Hbb^th3/+ mice lowered the percentage of total cells that correspond to EP in BM (43%; P<0.005) and spleen (6%; P<5×10). The fact that Hbb^th3/+Tmprss6^+/+ and Hbb^th3/+Tmprss6^−/− females maintained similar hemoglobin levels, even though Hbb^th3/+Tmprss6^−/− females showed attenuated splenomegaly and fewer EP in BM and spleen, indicates that Tmprss6 loss improves effectiveness of erythropoiesis in Hbb^th3/+ mice.

Figure 2.Genetic loss of Tmprss6 alters the erythropoietic composition of BM of Hbb^th3/+ mice. Erythroid (TER-119⁺) cells from BM of mice of selected Hbb-Tmprss6 genotypes were grouped by stage of erythroid maturation based on forward scatter and CD44 expression, as illustrated in Online Supplementary Figure 9. This approach enables gating of populations that correspond morphologically to proerythroblasts (region I), basophilic erythroblasts (region II), polychromatic erythroblasts (region III), orthochromatic erythroblasts and immature reticulocytes (region IV), and mature erythrocytes (region V).(6) (A) Graphed is the mean percentage of total BM cells composed of TER-119⁺ cells distributed in regions I, II, III, IV, and V, respectively. (B) Graphed on a log scale is the mean fluorescence intensity of CD71 expression in BM erythroid cells distributed in regions I-III. (C) Graphed is the mean percentage of apoptotic (annexin V+) erythroid cells distributed in regions IV and V. Graphed is the mean fluorescence intensity of BM erythroid cells distributed in region IV (D) and region V (E) generated following incubation with the ROS indicator 5-(and-6)-chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate, acetyl ester (CM-H2DCFDA). Error bars represent SD. For all panels, tissues from female mice (8-11 weeks old) were analyzed. *P<0.05; **P<0.005. For panel A, 5 Hbb^+/+Tmprss6^+/+, 6 Hbb^+/+Tmprss6^−/−, 4 Hbb^th3/+Tmprss6^+/+, and 6 Hbb^th3/+Tmprss6^−/− mice were analyzed. For panel B, 4 Hbb^+/+Tmprss6^+/+, 5 Hbb^+/+Tmprss6^−/−, 3 Hbb^th3/+Tmprss6^+/+, and 5 Hbb^th3/+Tmprss6^−/− mice were analyzed. For panel C, 5 Hbb^+/+Tmprss6^+/+, 6 Hbb^+/+Tmprss6^−/−, 4 Hbb^th3/+Tmprss6^+/+, and 5 Hbb^th3/+Tmprss6^−/− mice were analyzed. For panels D and E, 4 Hbb^+/+Tmprss6^+/+, 5 Hbb^+/+Tmprss6^−/−, 3 Hbb^th3/+Tmprss6^+/+, and 4 Hbb^th3/+Tmprss6^−/− mice were analyzed. Mice were fed a 200 p.p.m. iron diet.

In Hbb^th3/+Tmprss6^−/− and Hbb^+/+Tmprss6^−/− mice, the percentage of total cells that correspond to EP was remarkably similar in BM (P=0.59) and in spleen (P=0.95). In each early EP region (I-III) in BM, transferrin receptor (CD71) expression was significantly higher in Hbb^th3/+Tmprss6^−/− mice than Hbb^th3/+Tmprss6^+/+ mice, paralleling the level of hypoferremia, while CD71 expression in Hbb^th3/+Tmprss6^−/−and Hbb^+/+Tmprss6^−/− mice was similarly elevated (Figure 2B). In splenic early EP, CD71 expression by genotype was more variable (Online Supplementary Figure S7B).

The distribution of EP among regions I-IV in both BM and spleen was affected by genotype (P<0.0001 and P<0.0001, respectively) (Figure 2A and Online Supplementary Figure S7A). In Hbb^+/+Tmprss6^+/+ BM, the relative size of each EP population (as a percentage of total BM cells) increased progressively from regions I to IV. By contrast, in Hbb^th3/+Tmprss6^+/+ mice, the percentage of BM EP was lower in region IV (20%) than region III (23%), compatible with the increased α-globin-mediated apoptosis described in Hbb^th3/+ region IV EP.7 Indeed, Tmprss6 loss reduced the percentage of apoptotic erythroid cells in regions IV and V in BM (Figure 2C) and spleen (Online Supplementary Figure S7C) of Hbb^th3/+ mice; Tmprss6 loss also reduced ROS levels in these populations (Figure 2D and E and Online Supplementary Figure S7D and E). Nevertheless, in Hbb^th3/+Tmprss6^−/− BM, the percentage of EP in region IV (15%) remained lower than in region III (19%), suggesting that factors other than apoptosis regulate the size of the region IV EP population in iron deficiency. The distribution of EP among regions I-IV in Hbb^th3/+Tmprss6^−/− mice was remarkably similar to that of Hbb^+/+Tmprss6^−/− mice in BM and spleen (P=0.58 and P=0.95, respectively), demonstrating that erythroid maturation was dictated by Tmprss6 genotype, rather than effects of impaired β-globin production.

Reactive oxygen species promote cell membrane damage and impair erythrocyte deformability, decreasing oxygen-delivering capacity.8 Consistent with improved tissue oxygenation due to a reduction in erythrocyte ROS, Tmprss6 loss in Hbb^th3/+ mice caused a significant reduction in erythropoietin (Epo) mRNA in kidney (Figure 3A) and liver (Online Supplementary Figure S8). However, the high serum erythropoietin levels of Hbb^th3/+ mice were not lowered by Tmprss6 loss (Figure 3B), perhaps because Tmprss6 loss reduces the number of EP available to consume erythropoietin.

Figure 3.Genetic loss of Tmprss6 lowers renal Epo expression and alters erythroid gene expression in Hbb^th3/+ mice. Graphed for selected Hbb-Tmprss6 genotypes are mean values obtained from analyses of Epo mRNA relative to β-actin (Actb) in total kidney RNA (A), serum erythropoietin (B), Erfe mRNA relative to Hba in total BM RNA (C), Gdf15 mRNA relative to Hba in total BM RNA (D), Ahsp mRNA relative to Hba in total BM RNA (E), and Erfe mRNA relative to Hba in total spleen RNA (F). Error bars represent SD. Eight-week-old female mice were analyzed in panels A-B, and 8-to-9-week-old male mice were analyzed in panels C-F. For panels A, C, D, E, and F, ratios of mRNA expression are normalized to an Hbb^+/+Tmprss6^+/+ mean value of 1. *P<0.05; **P<0.005. For panel A, kidneys from 6 Hbb^+/+Tmprss6^+/+, 9 Hbb^+/+Tmprss6^−/−, 5 Hbb^th3/+Tmprss6^+/+, and 6 Hbb^th3/+Tmprss6^−/−mice were analyzed. For panel B, serum from 6 Hbb^+/+Tmprss6^+/+ mice and 7 mice of all other genotypes was analyzed. For panels C-F, the number of mice analyzed was as follows: 7 Hbb^+/+Tmprss6^+/+, 7 Hbb^+/+Tmprss6^−/−, 5 Hbb^th3/+Tmprss6^+/+, and 7 Hbb^th3/+Tmprss6^−/−(8 in panel F). Mice were fed a 200 p.p.m. iron diet.

In Hbb^th3/+Tmprss6^−/− and Hbb^+/+Tmprss6^−/− mice, BM levels of specific mRNA expressed during terminal erythroid maturation (Erfe, Gdf15, Ahsp) were higher than corresponding levels in Hbb^th3/+Tmprss6^+/+ mice when normalized to α-globin to adjust for genotype-specific differences in BM erythroid content (Figure 3C-E). The respective BM expression of Erfe, Gdf15, and Ahsp was similar in Hbb^th3/+Tmprss6^−/− and Hbb^+/+Tmprss6^−/− mice. In spleen, Erfe levels in Hbb^th3/+Tmprss6^−/− and Hbb^+/+Tmprss6^−/− mice were higher than Hbb^th3/+Tmprss6^+/+ mice when adjusted for erythroid content (Figure 3F).

Transgenic hepcidin overexpression,7 low-iron diet,7 and transferrin injection9 improve erythropoiesis in TI mouse models; the underlying mechanism(s), while still uncertain, appear to be related to altered iron availability to EP. Because Tmprss6 is expressed in liver but not hematopoietic cells, erythropoietic consequences of Tmprss6 loss likely occur secondary to the primary physiological consequence of Tmprss6 loss: hepcidin elevation leading to hypoferremia. Indeed, early EP of both Hbb^th3/+Tmprss6^−/− and Hbb^+/+Tmprss6^−/− mice showed markedly elevated CD71 expression, suggesting similar sensing of hypoferremia.

Given that treatment of Hbb^th3/+ mice with Tmprss6-siRNA improved erythrocyte survival,4 the ability of Tmprss6 genetic loss to raise RBC in Hbb^th3/+ mice likely reflects, at least in part, an improvement in erythrocyte lifespan. Because RBC and reticulocyte count in both Hbb^th3/+Tmprss6^−/− and Hbb^+/+Tmprss6^−/− mice were higher than wild type, we hypothesize that Tmprss6 loss increases the rate of reticulocyte release into the circulation. However, because decreased erythrocyte lifespan has been reported in patients with iron deficiency anemia,10 it is possible that the erythrocyte lifespan of mice with Tmprss6 loss remains shorter than wild type.

Notably, on a C57BL/6 background, Tmprss6 loss in Hbb^th3/+ mice not only raised RBC but indeed simultaneously lowered MCH; therefore, hemoglobin did not improve. By contrast, on a C57BL/6-Sv129 mixed background, Nai et al. found that Tmprss6 loss increased hemoglobin in Hbb^th3/+ mice.3 Genetic background influences iron parameters, including hepcidin,11 and Tmprss6 loss produced a greater rise in hepcidin in our study. Hbb^th3/+Tmprss6^+/+ hemoglobin levels in our study, which were similar to 2-month old C57BL/6J Hbb^th3/+ mice that were also retro-orbitally-sampled,12 were slightly higher than tail vein levels sampled by Nai et al.3 Hemoglobin levels in Hbb^th3/+ mice decline with age,12 which should be considered when comparing different studies employing this strain.54

Molecular chaperones, ubiquitin-mediated proteolysis, and autophagic pathways have been implicated in the degradation of excess α-globin in erythroid cells. In β-thalassemia, these protective mechanisms may become overwhelmed during erythroid maturation.13 Because the MCH of Hbb^th3/+Tmprss6^−/− mice is lower than Hbb^th3/+Tmprss6^+/+ mice, the absolute amount of excess α-globin per cell is also likely to be lower, enabling more effective cellular handling. Iron deficiency may also induce protective mechanisms;14 indeed, the chaperone α-hemoglobin stabilizing protein (Ahsp) was up-regulated in total BM RNA from mice with Tmprss6 loss.

Hepcidin suppression in TI has been attributed to effects of circulating factors, such as erythroferrone, that are released by erythroblasts.1 The ability of Tmprss6 loss to raise hepcidin in Hbb^th3/+ mice could indicate that: (i) the hepcidin-suppressing factors require TMPRSS6 for activity; or (ii) without functional TMPRSS6, production or efficacy of the hepcidin-suppressing factors is reduced. A recent study found that serum erythroferrone was markedly higher in Hbb^th3/+ mice than Tmprss6^−/−mice, yet higher in Tmprss6^−/− mice than wild-type controls.15 All genotypes in our study, however, showed low or undetectable serum erythroferrone, possibly due to differences in blood collection methods or duration of sample storage (E Nemeth, 2019, personal communication). Notably, erythroferrone does not appear to modulate the hepcidin elevation caused by Tmprss6 loss, as genetic loss of Erfe did not alter hepatic hepcidin mRNA or hematologic parameters in Tmprss6^−/− mice.15

In summary, homozygous Tmprss6 loss in Hbb^th3/+ mice results in terminal erythroid differentiation consistent with iron-restricted erythropoiesis. Our findings have relevance for application of Tmprss6-targeting therapies in TI, suggesting that TMPRSS6 inhibition will require careful titration to avoid exacerbation of anemia by iron restriction.