Erythropoiesis Suppression Is Associated with Anthrax Lethal Toxin-Mediated Pathogenic Progression

Anthrax is a disease caused by the bacterium Bacillus anthracis, which results in high mortality in animals and humans. Although some of the mechanisms are already known such as asphyxia, extensive knowledge of molecular pathogenesis of this disease is deficient and remains to be further investigated. Lethal toxin (LT) is a major virulence factor of B. anthracis and a specific inhibitor/protease of mitogen-activated protein kinase kinases (MAPKKs). Anthrax LT causes lethality and induces certain anthrax-like symptoms, such as anemia and hypoxia, in experimental mice. Mitogen-activated protein kinases (MAPKs) are the downstream pathways of MAPKKs, and are important for erythropoiesis. This prompted us to hypothesize that anemia and hypoxia may in part be exacerbated by erythropoietic dysfunction. As revealed by colony-forming cell assays in this study, LT challenges significantly reduced mouse erythroid progenitor cells. In addition, in a proteolytic activity-dependent manner, LT suppressed cell survival and differentiation of cord blood CD34+-derived erythroblasts in vitro. Suppression of cell numbers and the percentage of erythroblasts in the bone marrow were detected in LT-challenged C57BL/6J mice. In contrast, erythropoiesis was provoked through treatments of erythropoietin, significantly ameliorating the anemia and reducing the mortality of LT-treated mice. These data suggested that suppressed erythropoiesis is part of the pathophysiology of LT-mediated intoxication. Because specific treatments to overcome LT-mediated pathogenesis are still lacking, these efforts may help the development of effective treatments against anthrax.

After binding to one of 2 known cellular receptors, tumor endothelium marker-8 (TEM8) or capillary morphogenesis protein-2 (CMG2) [12], PA is cleaved using a cellular furin-like protease, oligomerized, and used to deliver LF into cells [6,13]. LT could cause high mortality in experimental mice and rats [14,15]; however, the pathogenic mechanism of LT that leads to animal death is not well understood. Previous studies have indicated that LT treatments induce TNF-a-independent hypoxia and shock [15]. Because systemic hypoxic symptoms cause animal deaths [15], LT-induced anemia may be involved. Anthrax-mediated hemolytic anemia has been reported in several clinical studies [16][17][18]. An LT-induced hemolysis of human blood was also demonstrated in vitro [19]; however, its impact has not been clarified in vivo. Because LF can disrupt all 3 MAPK pathways [8], and that the ERK [20,21], JNK [22], and p38 [23] pathways are all critical to erythropoiesis, we hypothesized that LT may block erythropoiesis and further exacerbate the anemic symptoms. This study clarified the potential involvements of hemolytic anemia and suppressed erythropoiesis in LT-treated C57BL/6J mice. In addition, because a significant induction of serum erythropoietin (EPO) is associated with extensive systemic hypoxia after LT Figure 1. Effects of LT on hemolysis and in vitro erythroid colony-forming cell assay. An image of LT (200 ng/ml)-induced hemolysis in vitro was compared with a saline-treated control group (A). After peripheral blood cells were incubated with or without LT at 37uC for 2 hours, plasma hemoglobin levels (mg/dL) were measured using Drabkin's reagent (B). The experimental outline of the in vivo hemolysis assay is shown, in which untreated and saline-treated groups served as negative control groups (C). The RBC counts (D) and the levels of cell-free plasma hemoglobin (E) were administration in mice [15], we hypothesized that the EPO release may be a self-rescue response. We also evaluated whether enhancing erythropoiesis using EPO treatments may ameliorate LT-induced anemia, and thus, reduce the mortality of LT-treated mice.

Lethal Toxins Induced Hemolysis in Mice
Following previous studies using human blood [19], we reproduced in vitro LT-mediated hemolysis experiments using mouse blood. We found that LT treatments (200 ng/ml) had a lesser effect in mice ( Figure 1A and 1B) compared to humans ( [19] and authors' unpublished results). In vivo analyses indicated that LT treatments gradually reduced RBC counts of mice prior to death ( Figure 1C, experiment outline, and Figure 1D). When hemolysis occurs within the blood circulation, hemoglobin is normally cleared by the hemoglobin-scavenging mechanisms [24,25]. When a massive hemolysis is severe enough to saturate the hemoglobin-scavenging mechanism capacity, the plasma level of the cell-free hemoglobin increases [26]. To investigate the level of hemolysis in LT-mediated anemia, plasma hemoglobin levels of LT-treated (1.5 mg/kg; a lethal dose) mice were measured. Compared to the untreated and saline-treated groups, the level of plasma hemoglobin did not increase significantly after LT treatments ( Figure 1C, experiment outline, and Figure 1E). Based on these results, LT-induced hemolysis might not be the only reason low RBC counts were induced. Because LT suppresses all 3 MAPK pathways [8], and that ERK [20,21], JNK [22], and p38 [23] are critical to erythropoiesis, we hypothesized that the production and the maturation of the RBC (i.e., erythropoiesis) might be affected by LT. To ascertain whether LT could influence erythropoiesis of primary erythroid precursor cells, isolated mouse BMs were treated with an erythrocytic differentiation-inducing cocktail that contained native RBC synthesis-promoting cytokine EPO, and then treated with or without LT. Erythroid progenitor cells [i.e., burst-forming unit-erythroid (BFU-Es) colonies] appeared after 7 to 14 days in untreated groups, whereas the colonies were significantly suppressed in LT-treated groups, especially at the high dose ( Figure 1F, for 200 ng/ml groups). These results suggest that LT is able to block erythropoiesis of primary erythroid precursor cells.

Apoptosis is Involved in LT-catalytic Activity-dependent Erythropoiesis Suppression
To investigate the suppression mechanism on erythropoiesis, in vitro erythroid differentiation of cord blood-derived CD34 + hematopoietic stem cells (HSC) in a 16 day course was analyzed.
To determine the effects of LT on various differentiation stages, cells were divided into 9 groups [ Figure 2B (1-9), 2C (1-9), and 2D (1-9)], and then subjected to vehicle [diluents: cell culture medium; Figure 2B (1-5), 2C (1)(2)(3)(4)(5), and 2D (1-5)] and LT [ Figure 2B (6-9), 2C (6-9), and 2D (6-9)] groups. Cells were treated with vehicle or LT on Days 0, 4, 8, and 12 ( Figure 2B-2D, 2-5 vehicle groups, 6-9 LT groups). Toxin treatments were conducted for 4 days for each of LT-treated groups ( Figure 2B-2D: Group 6, Days 0-4; Group 7, Days 4-8; Group 8, Days 8-12; Group 9, Days 12-16). After treatments, surface markers of erythrocytic progenitor cells, such as CD235a (glycophorin A, GPA) and CD71 (the transferrin receptor), were examined using flow cytometry. CD71 is expressed by a wide variety of hematopoietic cells, including BFU-Es, CFU-Es, and proerythroblasts, but it is not present on mature erythrocytes [27], whereas GPA is a late erythroid marker expressed on erythroblast cells and mature erythrocytes, but not on the earliest precursor cells [28]. The data revealed that cell size (FSC), cell granularity (SSC), and the percentage of GPA + and GPA + /CD71 + cells were gradually increased in the R1 region during differentiation ( Figure 2A, experiment outline, Figure 2B, 2C, and 2E). In contrast, the populations of erythrocytic progenitor cells ( Figure 2C-2F, GPA + and GPA + /CD71 + ) were gradually shifted from the R1 ( Figure 2B, larger cells) to the R2 region ( Figure 2B, smaller cells) after LT treatments at varied differentiation times (Figure 2B-2F; E, F, * p,0.05, ** p,0.01). To further explore the enzymatic activity requirement of LT on erythropoiesis suppression, recombinant PA (rPA), and recombinant LF (rLF), and E687A mutant LF (rLF E687A ) [which form recombinant wild-type LT (rLT) and catalytic mutant LT (rLT E687A ), respectively] were produced and purified [29]. In agreement with the treatments using LT, rLT induced a similar cellular response that shifted the populations of the erythrocytic progenitor cells from the R1 to the R2 region, compared to the untreated, vehicle, and other recombinant protein control groups ( Figure 2B-2F vs. Figure 3B-3F). The catalytic mutant LT (rLT E687A ) did not have such effects ( Figure 3B-3F). Because sub-G1 dead cells are smaller than normal cells, propidium iodine (PI) staining was further used to verify whether cell death was involved. In a representative PIstaining analysis, LT and recombinant LT (rLT) elicited 52% and 61% of sub-G1 death cells, respectively [ Figure 4A, Day 16-LT (R1+R2) groups and Day 16-rLT (R1+R2) groups vs. untreated control groups Day 0 (R1+R2) with only 9% sub-G1 and Day 16 (R1+R2) only 11% sub-G1 cells]. Among these 52% [Day 16-LT (R1+R2) groups] and 61% [Day 16-rLT (R1+R2)] sub-G1 cells, 44% and 57% of which belong to R2 population [ Figure 4A, Day 16-LT (R2), and Day 16-rLT (R2) groups], while only 8% and 4% belong to R1 population (data not shown) (aforementioned Figure 4A results are from one representative experiment, averaged results are showed in Figure 4D). This indicates that the major populations of cell in R2 region are actually dead after LT treatments (92% and 95% of R2 region; data not shown). We further used annexin V and active casepase-3 antibody to characterize whether these are apoptotic cells. Our data indicates that after LT and rLT treatments, 70% to 76% of total cells (weighting 94% and 95% of R2 population) became annexin V positive, and 47% to 53% of total cells became active-caspase-3 positive cells ( Figure 4D). This indicates that those LT-elicited hypoploid cells in R2 region are apoptotic cells. Our unpublished data also indicate that there are approximately 80% of R2-region cells expressing erythroid markers GPA on their surfaces. At the same time, about 61% GPA + cells are annexin V positive. To verify whether LT preferentially kills those differentiated cells, we performed LT-treatments of expanded CD34 + cells (4-day treatment, a same time course of aforementioned differentiating cells) for comparisons. Intriguingly, we did not observe significant elicitation of apoptotic cells in R2 region (our unpublished data). These results suggest that LT can't kill undifferentiated CD34 + determined 22, 44, and 66 hours after the mice were treated with LT. The in vitro erythroid colony-forming cell assay was performed by incubating murine BM cells with or without LT (200 ng/ml or 20 ng/ml). Colonies of progenitor cells were counted on Days 7 and 14 post-erythroid differentiation initiation (F). Untreated BM cells were used as a control. ** p,0.01, comparisons between groups are indicated. Data are reported as mean 6 SD. doi:10.1371/journal.pone.0071718.g001 precursor cells. Our data collectively suggest that LT preferentially kills GPA + -differentiating erythroid precursor cells but not GPA 2 , CD34 + progenitors. As a result, our data collectively suggest that anthrax LT can suppress erythropoiesis in part by killing the erythroid progenitors through induction of caspase-3 dependent apoptosis. In addition, this suppression is dependent on the catalytic activity of LT.

Erythropoietin Treatments Reduced LT-mediated Mmortality, Aanemia, and Erythropoiesis Suppression in Mice
Upregulated plasma EPO levels were previously shown to coincide with hypoxic tissue damages in experimental mice [15]. EPO is a 30,400 dalton glycoprotein that positively regulates the BM erythroid cell proliferation, differentiation, and survival [30]. To investigate whether the induction of EPO was only a consequence of an acute phase response or a physiological counteraction to overcome LT-mediated erythrocyte suppression, recombinant EPO was administrated 1-3 days prior to LT treatments ( Figure 5A). If LT-mediated hemolysis anemia and erythropoiesis suppression are associated with LT-mediated pathogenic progression, then increasing erythropoiesis using EPO treatments may have ameliorative effects. We found that a single injection of a lethal dose of LT (1.5 mg/kg, LF : PA = 1: 5) in mice resulted in 100% mortality (6/6 deaths) within 130 hours ( Figure 5B). In contrast, the same dosage of LT induced only 50% mortality (3/6) in EPO-treated mice within 130 hours ( Figure 5B; statistically significant, p = 0.029). Our previous study has found that LT treatments also induce thrombocytopenia and megakaryopoiesis suppression. These defects play certain roles in LTinduced lethal pathogenesis [31]. In addition, treatments of TPO can reduce LT-mediated mortality from 95.8% to 46.7% in mice [31]. This evidence prompted us to test whether combined treatments using EPO and TPO could achieve further amelioration on the lethal pathogenesis. Our data revealed that pretreatments using EPO and TPO could exert a complete (100%) rescue of the mice from LT-induced lethality ( Figure 5B, TPO+EPO+LT groups). All surviving mice remained alive for up to 2 months (authors' unpublished data). To verify whether reduced mortality caused by EPO pretreatments is associated with ameliorated suppression on erythropoiesis, a BFU-E colony assay was performed using BM cells from LT-and/or EPO-treated mice ( Figure 5C, experimental outline). Analysis data revealed that the colony formation of BM-derived BFU-E was suppressed in LTtreated groups, while such suppression is significantly ameliorated by EPO pretreatments ( Figure 5D, LT vs. EPO+LT groups). RBC counts, hemoglobin, and hematocrits can help diagnose anemia [32]. These parameters were measured to determine whether EPO rescue is associated with ameliorated anemia. Experimental mice were treated with or without LT and/or EPO, and the respective hematopoietic parameters such as RBC counts, hemoglobin, hematocrits, white blood cell (WBC) counts, and platelet counts were measured at various times for comparison ( Figure 6A, experimental outline). The data revealed that all of these parameters except the WBC counts were reduced after LT treatments ( Figure 6B-6F, untreated vs. LT groups), whereas EPO significantly ameliorated all LT-mediated suppressions except the low platelet counts ( Figure 6B-6D and 6F, LT vs. EPO+LT groups). To verify whether the ameliorative effect of EPO on anemia can be attributed to an increase in erythropoiesis, BM cells of mice were isolated from femurs and tibiae at 69 hours post-LT treatments ( Figure 7A, experimental outline). The surface expression of CD71 and TER-119 was measured to determine the maturation status of erythrocytes by using flow cytometry [33,34]. CD71 is expressed in a wide variety of hematopoietic cells, including BFU-Es, CFU-Es, and proerythroblasts [27], whereas TER-119 is a late erythroid marker expressed on erythroblast cells [35]. These 2 markers allow researchers to distinguish among erythroid cell populations at 4 differentiation states: CD71 high -TER-119 med ( Figure 7B, Region 1, R1), CD71 high TER-119 high ( Figure 7B, Region 2, R2), CD71 med TER-119 high ( Figure 7B, Region 3, R3), and CD71 low TER-119 high ( Figure 7B, Region 4, R4) [36]. Analysis data revealed that the percentage of total erythroblasts was reduced after LT treatment ( Figure 7B, R1-to-R4 regions, and Figure 7C, untreated vs. LT), especially in the R2 and R3 populations ( Figure 7D, R2 and R3 groups, untreated vs. LT) in BM. EPO pretreatment significantly increased the cell numbers in the R1 and R2 regions ( Figure 7D, LT vs. EPO+LT). These results further suggest that LT suppresses erythropoiesis in mouse BM, and that EPO treatments can ameliorate this suppression ( Figure 7B-7D).

Discussion
Hemoglobin in RBCs is the principle molecule that serves as an oxygen carrier [37]. Suppressions of RBC counts, hemoglobin, and hematocrits by LT treatments (Figure 6B-6D) may theoretically contribute to the systemic hypoxia observed in LT-treated mice [15] and anthrax patients [16,18]. LT-induced hemolysis is one of the potential causes of low RBC counts. However, the basal level of cell-free hemoglobin detected in mice plasma cannot fully explain the drastic drops in RBC counts and hematocrits ( Figure 1D, 6B, and 6D). Because approximately 2.5 days (or 60 hours) are required to accomplish erythropoiesis in murine BM [38], LT-mediated suppression on erythropoiesis can theoretically be achieved before lethality occurs ( Figure 5B, 78.5-126.5 hours).
The elicitation of apoptosis of primary CD34 + -derived erythroblasts during in vitro differentiation (Figure 2, 3, 4) suggests that the suppressed erythropoiesis in BM may be caused by cell death of CD71 high TER-119 high (R2) and CD71 med TER-119 high (R3) populations ( Figure 7B and 7D). The suppressive effect of LT on erythropoiesis also occurs in the spleen ( Figure S1), a minor erythropoiesis site that is primarily induced under stress [39,40]. These occurrences are in agreement with our colony-forming unit cell assays (in vitro, Figure 2F; in vivo, Figure 5D) and a previous report [41].
Previous studies have indicated that PA binds to the anthrax toxin receptor ATR1 on all lineages of hematopoietic progenitors in the bone marrow, including the erythroid progenitor lineage [41,42], and suppresses hematopoiesis [41]. However, the mechanism involving LT-induced erythropoiesis suppression is unclear. In our study, stroma cells were supplied in the expansion stage but not in the in vitro differentiation stages (Figure 2A and  3A). Thus, an indirect suppressive effect through LT-elicited stroma-derived cytokines, which was suggested to suppress hematopoiesis [41], is not likely to be involved. LT is an MAPK inhibitor and cleaves the N-terminal domain of MKKs/MEKs [8,9], disrupting 3 MAPK pathways. The carboxy-terminal region additional flow cytometry analysis. Representative cell-populations and histograms (B-D) were shown. The total cell number was defined as 100%. *p,0.05, ** p,0.01, comparisons between groups are indicated. Data are reported as mean 6 SD and represent results from 4 independent experiments. doi:10.1371/journal.pone.0071718.g002 at residues 686-690 (HEXXH) of LF contains a putative zincbiding site for its metalloprotease activity [43]. We previously demonstrated that the mutant LF E687A -composed LT has a reduced cytotoxicity and catalytic property against MEK1 and MEK2 ( [29] and authors' unpublished results). Our data suggests that the catalytic activity of LT is required for the suppressive effect on erythropoiesis (Figure 3).
Two major methods for tackling anthrax are currently available: vaccination and antibiotics [44,45]. Animal experiment results show that it takes approximately 6 or more PA immunization cycles to elicit a weak immune response against anthrax. However, the protective immunity does not usually last long (authors' unpublished data). In addition, because the toxic effect of LT is sufficient for lethality, animals may eventually die even after aggressive antibiotic therapies to eliminate the bacteria [46]. These results indicate that a specific treatment against LT is required. In this study, we demonstrated that pretreatments of EPO could reduce mortality in LT-injected mice by ameliorating RBC counts, hemoglobin, hematocrits, and erythropoiesis in BM ( Figure 5, 6,  7). However, EPO contains nonerythropoietic properties, including tissue protection resulting from its ability to bind to injuryinduced lower-affinity EPO receptors (EPORs), distinct from the high-affinity EPORs involved in erythropoiesis [47,48]. In addition to the erythropoietic induction, the tissue-protective property of EPO may be one of the possible mechanisms explaining the rescue of LT-challenged mice [47,49]. Studies have indicated that low levels of EPO (picomolar) are sufficient for erythropoiesis induction, whereas much higher levels of EPO (nanomolar) are required for tissue protection [47,48]. EPO has a half-life of approximately 2 hours in the blood circulation [50]. Accordingly, to exclude the possibility of tissue protection, mice were treated with rhEPO using doses that kinetically declined into a range of concentrations to enhance erythropoiesis rather than tissue protection before LT was administered ( Figure S2) [47]. The rhEPO was almost completely removed from the blood circulation of the mice within 24 hours (Figure S2), at a much faster clearing rate than for humans [47]. These results suggest that tissue protection is not involved in the EPO rescue. Our previous study showed that treatments of TPO could ameliorate LT-mediated mortality in mice [31]. This evidence prompted us to test whether combined treatments using EPO and TPO could also achieve amelioration on the lethal pathogenesis. Our data reveal that pretreatments using EPO and TPO could exert a complete (100%) rescue of the mice from LT-induced lethality ( Figure 5B, TPO+EPO+LT groups). These results indicate the combined treatments of EPO and TPO is much potent than the single treatment of TPO alone [31]. TPO contains erythropoiesisenhancing properties in certain conditions [51], whereas treatments of TPO alone did not significantly resolve the anemic symptoms of LT-challenged mice [ Figure 7D of the reference [31]], indicating that the erythropoiesis-enhancing property of TPO is not sufficient to protect against LT-induced erythropoiesis suppression. Nonetheless, these results suggest that the platelet and RBC are critical cell types for maintaining homeostasis in LTchallenged mice. EPO administration is the well known medical treatment to enhance erythropoiesis in renal anemia [52], anemia of prematurity [53], and anemia with non-myeloid malignancies [54]. TPO treatment has been discontinued due to thrombocytopenia elicited by autoantibodies against endogenous TPO [55]. For clinical view, combined administration of EPO and secondgeneration of thrombopoietic agents, such as romiplostim and eltrombopag, will be used as a potential approach to treat anthrax.
Evidence from other studies has revealed that once anthrax enters a bacteremia stage, it inevitably leads to death, even with aggressive antibiotic therapy that prevents bacterial growth [17]. This is caused by toxin-mediated damages that are accumulated in the body, although the exact mechanism remains unclear [56]. Our data collectively suggested that LT-mediated suppression on megakaryopoiesis and erythropoiesis is part of LT-mediated pathophysiology. Protection from such damage to hematopoietic progenitor cells using specific growth factors can significantly reduce LT-induced animal death. The findings we have documented provide a novel perspective for the potential development of feasible alternative and additional therapeutic strategies against anthrax.

Ethics Statement
Cord bloods and the umbilical cords from full-term deliveries were collected by Mennonite Christian Hospital, Hualien, Taiwan. Informed written consents were provided by participants and obtained using protocols approved by the Research Ethics Committee of Mennonite Christian Hospital (Approval ID: 09-12-046-ER and 09-12-047-ER). All human samples were anonymized. The research methods used on the experimental mice were in accordance with the national guidelines of Animal Protection Act (Taiwan) and approved by the Institutional Animal Care and Use Committee, Tzu Chi University (Approval ID: 95017; Project: Potential roles of Bacillus anthracis lethal toxin on erythroid cells).

Toxins
LT was purified and obtained as previously described [57]. Doses of LT refer to 1:5 amounts of LF and PA (i.e., 120 mg LT equals 20 mg LF plus 100 mg PA). Recombinant wild-type LT (rLT) is composed of recombinant PA (rPA) and wild-type LF (rLF), and catalytic mutant LT (rLT E687A ) is composed recombinant PA (rPA) and E687A-mutant LF (rLF E687A ) [29], respectively. These recombinant proteins were produced and purified from Escherichia coli/SG13009 as previously described [29], using Histagged affinity chromatography according to the manufacturer's instructions (Qiagen, Hilden, Germany). . Proteolytic activity of LT component lethal factor is essential for the suppression of erythrocytic differentiation. The timetable of in vitro erythrocytic differentiation using human cord blood-derived CD34 + cells is shown (A). To address the effect with or without proteolytic activity, rLF and rLF E687A were employed, respectively. Flow cytometry analysis of the cell size (FSC) and cell granularity (SSC) are shown at various times (B). The percentage of GPA + (erythrocytic marker) cells in the R1 and R2 regions are shown in (C) and (D), respectively. The percentage of GPA expression cells in the total cells (1610 4 ) in the R1 and R2 regions for different treatments are shown in (E) and (F), respectively. LT and control proteins such as rPA (5 mg/ml), rLF (5 mg/ml), rLF E687A (5 mg/ml), rLT (5 mg/ml rPA +5 mg/ml rLF), and rLT E687A (5 mg/ml rPA +5 mg/ml rLF E687A ) were added into the medium by differentiation on Day 8. The cells were then analyzed using flowcytometry 4 days later. Groups treated with or without a (untreated) vehicle and BSA (10 mg/ml) served as negative controls. Representative cell-populations and histograms (B-D) were shown. Total cell number was defined as 100%. *p,0.05, ** p,0.01, comparisons with untreated groups are indicated. Data are reported as mean 6 SD and represent results from 4 independent experiments. doi:10.1371/journal.pone.0071718.g003 Hemoglobin Detection C57BL/6J mice (males, 6-8 weeks old) were purchased from the National Laboratory Animal Center (Taipei, Taiwan) and maintained in pathogen-free conditions in the experimental animal center at Tzu Chi University. For the in vitro hemoglobin assay, peripheral blood collected from the retro-orbital plexi of C57BL/6J mice were mixed with anticoagulant acid citrate dextrose formula A (ACD-A : 38 mM citric acid, 75 mM trisodium citrate, 139 mM D-glucose, 12.5 mM EDTA) (blood : ACD-A = 8:1) [58] and then incubated with or without LT (200 ng/ml) at 37uC for 2 hours [19]. Plasma hemoglobin was measured using Drabkin's reagent (Sigma-Aldrich, St. Louis, MO, USA) by detecting the absorbance at 540 nm. The standard curve for hemoglobin was obtained by serial dilution of the whole blood from the peripheral blood of untreated mice, in which the hemoglobin concentration was determined using an automated hematology analyzer (KX-21, Sysmex Corporation, Kobe, Japan). Hemoglobin concentrations (mg/dL) of the samples were then obtained through comparisons of the standard curve. To determine LT-induced hemolysis in vivo, C57BL/6J mice were retro-orbitally injected with or without a lethal dose of LT (1.5 mg/kg in 250 ml saline). Untreated and saline-treated mice were used as negative controls. The level of plasma cell-free hemoglobin was measured at 22, 44, and 66 hours post-LT treatments using Drabkin's reagent.

DNA Content Analysis
Cells were washed twice using phosphate buffered saline (PBS), fixed with ice-cold 70% ethanol in PBS at -20uC for 2 hours, and then washed with PBS and resuspended in 500 ml staining solution containing propidium iodine (PI) (20 mg/ml PI, 0.1% Triton X-100 and 0.2 mg/ml RNase A in PBS) and incubated at room temperature for 30 minutes. The DNA content of the cells was analyzed using a flow cytometer (FACSCalibur, Becton-Dickinson, San Jose, CA, USA).
EPO Rescue Experiments (Mortality, Erythroid Colonyforming Cell Assay, Peripheral Blood Hematopoietic Parameters, and Bone Marrow Analysis) C57BL/6J mice (males, 10-11 weeks old) were treated with recombinant human EPO (rhEPO, Neorecormon H, Roche, Mannheim, Germany) (2 IU/g, in 250 ml saline) and/or murine TPO (CytoLab, Rehovot, Israel) (0.25 mg/mice, in 250 ml saline) twice by using a retro-orbital injection at 72 and 24 hours before administration of a lethal dose of LT (1.5 mg/kg in 250 ml saline, using retro-orbital injection, with approximately 95.8% mortality [31]). Untreated and EPO-treated groups without further LT treatments served as controls. As there is no suitable parameter to predict the death/survival for LT-challenged mice, we use death as an endpoint. The mortality and survival duration of the mice were recorded after LT treatments. The experimental mice were monitored every 4-6 hours. The surveillance showed that LTchallenges reduce behavioral activities of mice before death without body weight loss and any obvious discomfort (such as abnormal behavior and convulsions). Surviving mice were sacrificed by cervical dislocation 69 hours after LT treatments. The BM cells were collected and performed in vivo in the erythroid colony-forming cell assay (Materials and Methods: Erythroid colony-forming cell assay). To measure the hematological parameters [RBC counts, hemoglobin level, white blood cell counts, hematocrits (Hcts), and platelet counts], blood samples were collected 22, 44, and 66 hours from the retro-orbital plexi of the mice after LT treatments. Collected blood was mixed with the anticoagulant ACD-A (blood: ACD-A = 1:9). The hematological parameters were determined using an automated hematology analyzer (KX-21, Sysmex Corporation, Kobe, Japan). To detect erythroblast-specific surface markers, BM cells of mice were collected from femurs and tibiae 69 hours after LT treatments. The femurs and tibiae were flushed using the RPMI-1640 medium supplemented with 20% anticoagulant ACD-A through 30G needles. The cell suspensions were passed through a 55 mm nylon mesh to remove the cell aggregates. Cells were blocked with 5% BSA in an RPMI medium at 37uC for 1 hour and then incubated in a 500 ml RPMI-1640 medium with 1 ml of an FITC-conjugated rat anti-mouse CD71 antibody (BioLegend) and 3 ml of R-Phycoerythrin (R-PE)-conjugated rat anti-mouse TER-119 antibody (BD Immunocytometry System) at 37uC for 1 hour. After PBS washes, the cells were analyzed using a FACSCalibur flow cytometer and CellQuest TM Pro program (Becton-Dickinson, San Jose, CA, USA).

Statistics
Means, standard deviations (SDs), and statistics for quantifiable data were calculated using Microsoft Office Excel 2003 for Windows. Comparisons between groups were made using the twotailed Student t test. A p value of ,0.05 was considered significant. Figure S1 LT-mediated mortality is associated with reduced number of erythroblast cells in spleen. The experimental outline of spleen experiments is illustrated (A). Flow cytometry analysis of spleen cells were performed at 69 hours after LT treatments. To determine the maturation stages of RBC precursors, the erythroblast cells were gated as R1 (CD71 high , TER-119 med ), R2 (CD71 high , TER-119 high ), R3 (CD71 med , TER-119 high ), and R4 (CD71 low , TER-119 high ) in all groups (B). The relative cell population (% of total 1610 5 ) (C) and the cell numbers of all erythroblast cells (R1, R2, R3, and R4) in respective groups are showed (D). *p,0.05, ** p,0.01, comparisons between groups are indicated. Data are reported as mean 6 standard deviation (SD). (TIF) Figure S2 Plasma EPO concentration after EPO injection. The experimental outline of EPO immunoassay is showed (A). EPO concentration of plasma after two sequential EPO injections was examined 30 minutes and 24 hours after the second EPO injection by ELISA (B). Data are reported as mean 6 standard deviation (SD) except the 30-minute group showed as a mean value. EPO concentration was calculated according to a standard curve showed on (C).