Heme oxygenase 1 is expressed in murine erythroid cells where it controls the level of regulatory heme

Heme is a complex of iron with protoporphyrin IX, which is essential for the function of all aerobic cells. Heme serves as the prosthetic group of numerous hemoproteins (eg, hemoglobin, myoglobin, cytochromes, guanylate cyclase, nitric oxide synthase) and plays an important role in controlling protein synthesis and cell differentiation.¹  However, when left unguarded, non–protein-bound heme promotes free radical formation and lipid peroxidation, resulting in cell damage and tissue injury.^2,3  Hence, cellular heme levels are tightly controlled; this is achieved by a fine balance between heme biosynthesis and catabolism.

The highest amounts of organismal heme (75% to 80%) are present in circulating red blood cells (RBCs) whose precursors synthesize heme with rates that are at least 1 order of magnitude higher than those in the liver, the second most active heme producer in mammals. Differences in iron metabolism and in the genes encoding 5-aminolevulinic acid synthase (the first enzyme in heme biosynthesis) are responsible for the variation in the regulation and rates of heme synthesis in erythroid and nonerythroid cells.¹ 

Heme is degraded by heme oxygenases (HOs) yielding iron, bilirubin, and carbon monoxide.^4,5  Heme from RBCs is fated to be catabolized in splenic and hepatic macrophages after erythrophagocytosis of senescent RBCs. There are 2 isoforms of HOs: HO-1 and HO-2. The constitutive form, HO-2, is membrane-bound protein found at highest levels in the brain and testis.⁶  The inducible form, HO-1, is a 32.8-kDa membrane-bound enzyme found at the highest concentrations in the liver and spleen. HO-1 is increased in animal tissues after treatment with its natural substrate heme and various metals, xenobiotics, endocrine factors, and synthetic metalloporphyrins⁶  as well as several agents that cause oxidative damage.^7-9  HO-1 promotes protection through the removal of the dangerous prooxidant “free” heme, generated in stress conditions, and releases biliverdin and bilirubin, which are metabolites with antioxidant properties.^3,10 

HOs, in particular HO-1, have been extensively studied in hepatocytes and other nonerythroid cells.^4,5  Ironically, even though RBCs possess the greatest amounts of heme, virtually nothing is known about the expression of HO-1 in developing RBCs. Moreover, it is unknown whether HO-1 plays any role in hemoglobin synthesis under physiological or pathological conditions.

The aim of our study was to determine whether developing RBCs express HO-1 and, if so, characterize its role in physiology and possibly pathophysiology of erythropoiesis. We have demonstrated that HO-1 is present in mouse erythroid cells (Ter119⁺ and fetal liver [FL] cells), and show that its expression is enhanced during erythroid differentiation. Interestingly, we have also observed that HO-1 overexpression in murine erythroleukemia (MEL) cells impairs and that its absence, in FL obtained from HO-1^−/− mice, enhances hemoglobinization, respectively. Our results indicate that an appropriate level of HO-1 is essential for efficient erythropoiesis.

FVB.129S2(B6)-Hmox1^tm1Poss/J mice were obtained from The Jackson Laboratory (Bar Harbor, ME) and bred in our animal facility. All animal studies were approved by the McGill Animal Care Committee. Pregnant animals were euthanized on embryonic day 12.5 post coitum and the embryos were genotyped to identify those that were HO-1^−/−. Because there is no truly erythroid cell line available, we used primary erythroid cells from control and HO-1^−/− mice whenever possible. Livers from appropriate animals were processed as described in the following section. For reticulocytes studies, wild-type (HO-1^+/+) and HO-1 knockout (HO-1^−/−) mice were injected for 3 days with phenylhydrazine (25 mg/kg). After 2 days of rest, the mice were euthanized and the blood was collected by cardiac puncture. Thiazole Orange staining was performed to determine the percentage of reticulocytes in both HO-1^+/+ and HO-1^−/− samples.¹¹  The percentage of reticulocytes determined by thiazole orange staining was used to normalize the results.

In experiments in which primary cells would not suffice, we settled for using MEL cells. MEL cells were cultured as described in the supplemental Methods on the Blood Web site. All of the experiments were conducted with both uninduced MEL cells (0 hours) as well as cells treated with 1.8% dimethylsulfoxide (DMSO) to induce differentiation and hemoglobinization.¹²  Primary erythroid cells were cultured as described.¹³  Briefly, cells were grown from FLs that were obtained from embryonic day 12.5 embryos of wild-type and HO-1 knockout (FVB.129[B6] background) mice. Primary FL cells were kept either in an undifferentiated state (0 hours) or induced to terminal differentiation (24, 48, or 72 hours). The medium composition to maintain the cells in such states is described in further detail in the supplemental Methods. Where indicated, cells were treated with 100 µM of the HO-1 inhibitor, tin-protoporphyrin IX (SnPP) (Frontier Scientific, Logan, UT), and 0.4 mM of heme synthesis inhibitor, succinyl acetone (SA) (Sigma, St Louis, MO).

We generated MEL cells that overexpress HO-1 as described in supplemental Methods. MEL cell clones showing highest levels of HO-1 expression (MELHO-1) were used in the experiments. MEL cell clones carrying only the plasmid, without HO-1 complementary DNA insert, were used as control.

Nondifferentiated and differentiated cells were washed twice with phosphate-buffered saline/2% fetal calf serum and stained with fluorescence-labeled antibodies against transferrin receptor (TfR) (CD71- fluorescein isothiocyanate; Pharmingen, San Diego, CA). Surface marker expression was analyzed by flow cytometry (FACSCalibur; BD Biosciences, Mississauga, Canada) as shown in supplemental Figure 1. Geometric means were used to quantify the changes in TfR expression in the different conditions tested.

Femora and humeri from wild-type mice were collected and their bone marrows extracted. Bone marrow cells were processed following the protocol described in supplemental Methods. Ter119⁻ and Ter119⁺ cells were sorted by flow cytometry (FACSVantage SE; BD Biosciences).

Cellular heme content was assayed as described previously.¹⁴  The experiment results were normalized by protein concentration. Hemoglobin assay was performed as previously.^15,16  The quantification results were normalized by cell number and cell volume. Both protocols are described in detail in supplemental Methods.

Statistical analysis was done using the software SPSS v15.0 (IBM Software, Markham, Canada). Data were evaluated applying analysis of variance and Student t tests. Error bars of graphs represent the standard deviation of 3 independent experiments (n = 3).

We used the erythroid marker Ter119¹⁷  to isolate erythroid cells from mouse bone marrow using a fluorescence-activated cell sorter. As expected, Ter119⁺ cells showed a high expression of globin messenger RNA (mRNA) and protein, indicating their erythroid origin (Figure 1A,C). We also found that HO-1 mRNA (Figure 1B) and protein (Figure 1C) expressions were increased 20-fold and 2.5-fold, respectively, in Ter119⁺ cells (representing all cells of erythroid lineage) as compared with Ter119⁻ cells (representing the bone marrow–derived, nonerythroid cells).

Using primary erythroid cells isolated from mouse FL cells at 12.5 days, we evaluated HO-1 mRNA levels during erythropoietin (EPO)-induced erythroid differentiation. As shown in Figure 1D, HO-1 mRNA levels significantly increase during 48 and 72 hours of differentiation. Hence, we examined whether heme, whose synthesis is well known to increase throughout erythroid differentiation,¹  could be involved in the enhancement of expression of HO-1 in the differentiating FL cells. We demonstrated (Figure 1D) that SA, a specific inhibitor¹⁸  of 5-aminolevulinic acid dehydratase (the second enzyme in heme biosynthesis), added 24 hours before cell harvest, blocked the increase in HO-1 mRNA expression at 48 and 72 hours. Moreover, the expression of HO-1 protein, which became detectable by western blot analysis at 48 hours of differentiation, was suppressed by SA (Figure 1E-F). Thus, as in virtually all cell types,¹⁹  heme induces HO-1 expression in erythroid cells.

We were surprised that the upregulation of HO-1 expression observed in this experiment (Figure 1D-F) did not interfere with hemoglobinization. To further examine this issue, we generated a MEL cell line stably overexpressing HO-1 (see “Material and methods”). As shown in supplemental Figure 2, MELHO-1 cells express 14-fold higher levels of HO-1 mRNA (supplemental Figure 2A) and 10-fold more HO-1 protein (supplemental Figure 2B) as compared with controls. After 72 hours’ incubation with DMSO, MELHO-1 cell pellets exhibited a lower degree of redness (Figure 2A). Moreover, the HO-1 overexpression resulted in a significant decrease in heme levels after 72 hours of differentiation (Figure 2B). In agreement with these observations, we found that MELHO-1 cells also contain lower levels of globin after 48 hours of differentiation (Figure 2C-D). Interestingly, MELHO-1 ferritin levels increased after 48 hours of differentiation (Figure 2C-D). This increase was partially blocked (Figure 2C-D) by SnPP, an HO-1–specific inhibitor.²⁰ 

In erythroid precursors, heme acts as a translational regulator by modulating the activity of the heme-regulated eIF2α kinase, HRI.²¹  During heme deficiency, HRI is active, which in turn leads to the inhibition of globin translation. We demonstrated that after 48 hours of stimulation of erythroid differentiation (incubation with DMSO), MELHO-1 cells contained increased levels of eIF2α-P compared with controls (Figure 2E-F). This indicates that HO-1 overexpression decreases the regulatory heme pool, which subsequently leads to HRI activation. Moreover, the observed decrease of globin protein levels in differentiated MELHO-1 cells (Figure 2C-D) further supports the conclusion that activated HRI inhibited globin translation.

The rate of iron acquisition from transferrin limits and therefore controls the rate of heme synthesis in erythroid cells.¹  The expression of TfRs, the gatekeepers for physiological iron acquisition, is tightly regulated¹  during erythroid differentiation. After 72 hours of DMSO-mediated differentiation, TfR mRNA expression was decreased by 5.4-fold in MELHO-1 cells compared with control cells (Figure 3A). Additionally, TfR protein levels were also reduced after 48 hours of differentiation compared with controls (Figure 2C-D). We also found a significant reduction of TfR on MELHO-1 cell membranes compared with control cells at 0 hours as well as 48 and 72 hours of differentiation (Figure 3B). The reduction in TfR on the cell membrane significantly decreased the ability of MELHO-1 cells to take up ⁵⁹Fe from ⁵⁹Fe-transferrin (Figure 3C) and incorporate ⁵⁹Fe into heme (Figure 3C).

HO-1 could interfere with the formation of heme and; therefore, we tested the integrity of heme biosynthesis by supplying cells with ⁵⁹Fe-salicylaldehyde isonicotinoyl hydrazone (⁵⁹Fe-SIH), a membrane-permeable iron chelate that has the ability to supply iron for ferrochelatase^22-24  via bypassing the physiological transferrin-TfR pathway. There was no difference in iron incorporation into heme between MELHO-1 and control cells when the physiological iron delivery pathway was bypassed using ⁵⁹Fe-SIH (Figure 3D). Hence, these data indicate that HO-1 overexpression leads to a diminished iron supply for heme biosynthesis as a result of decreased TfR expression. Because in erythroid cells heme is needed to maintain a high rate of TfR synthesis,²⁵  the hemoglobinization defect in MELHO-1 is likely caused by the reduction in TfR levels. Additionally, we observed a slight decrease in the iron-regulatory protein 1 (IRP1) RNA-binding activity in MELHO-1 with no difference in IRP2 levels when compared with control (supplemental Figure 3). These data suggest only a small contribution, if any, of the iron-responsive element/IRP system to the reduction in TfR levels and the increase in ferritin levels in MELHO-1 cells (Figures 3A-B and 2C-D).

The results of the preceding experiments led us to examine several parameters of erythroid differentiation in FL cells isolated from HO-1 knockout embryos (FL/HO-1^−/−). Figure 4A-B show that, in FL/HO-1^−/− cells, HO-1 was not detectable at mRNA or protein levels. Interestingly, FL/HO-1^−/− cell pellets exhibited a higher degree of redness than wild-type cell pellets at 0, 24, and 48 hours of differentiation (Figure 4C). Indeed, erythroid cells lacking HO-1 have a higher hemoglobin content, as determined by its direct measurement (Figure 4D).

As expected, we found that FL/HO-1^−/− cells express higher amounts of TfR mRNA, in particular at 24 and 48 hours (Figure 5A), and exhibit a slight increase in TfR protein levels (Figure 5D-E). Flow cytometry analysis revealed an increase in TfR presence on the membrane of FL/HO-1^−/− cells, in particular at 0 and 24 hours after EPO treatment (Figure 5B). Figure 5C shows a significant increase in globin mRNA in FL/HO-1^−/− cells at 0 and 48 hours. The equal levels of globin mRNA at 24 hours can likely be explained by the presence of overriding, heme-independent transcription factors that drive globin gene expression during the early stages of differentiation in both cell types equally. FL/HO-1^−/− cells contain more globin protein (Figure 5D-E), although the differences in its levels are smaller than in the case of globin mRNAs. This probably indicates that translation is limiting in the overall process of globin gene expression. On the other hand, ferritin protein levels were reduced in FL/HO-1^−/− cells (Figure 5D-E), suggesting that the increased flux of iron into heme leads to a decrease in ferritin translation. An additional factor that could cause a decrease in ferritin synthesis would be a decrease in “free” iron levels caused by HO-1 deficiency.

Reticulocytes were isolated from both HO-1^+/+ and HO-1^−/− mice injected with phenylhydrazine. When analyzed by flow cytometry, HO-1^−/− reticulocytes showed a significant increase in the levels of TfR present on the cell membrane (Figure 6A). Moreover, HO-1^−/− reticulocytes took up significantly more ⁵⁹Fe-Tf compared with HO-1^+/+ reticulocytes after 15 minutes of incubation (Figure 6B). Additionally, eIF2α-P levels were decreased in HO-1^−/− reticulocytes (Figure 6C), suggesting an increased heme level and consequent reduction in HRI activity. In agreement with that, globin protein levels were higher in HO-1^−/− reticulocytes compared with HO-1^+/+ (Figure 6C).

HO-1 has been extensively studied in hepatocytes and other nonerythroid cells.^4,6,26  In contrast, the literature on HO-1 in erythroid cells is extremely scarce. It is quite surprising that there is so little known about the role of the enzyme that catabolizes heme (HO-1) in erythroid cells that make, by far, the most heme in mammals. We were able to identify only 2 rather noninstructive studies that examined HO-1 mRNA expression in erythroid cells.^27,28  Fujita and Sassa observed a decrease in HO-1 mRNA levels in the first 12 hours of differentiation, at which point hemoglobin synthesis barely commences. In a recently published study, Alves et al were unable to detect HO-1 in K562 cells. However, there K562 cells were ill-defined cells derived from a patient with chronic myelogenous leukemia, exhibiting only a low expression of embryonic and fetal types of hemoglobin.²⁹  Their report also claims an absence of HO-1 expression in bone marrow “erythroid precursors” that were isolated using cell-surface markers, which are not appropriate for detecting hemoglobinized cells. Additionally, bone marrow used in their study was derived from patients with hematological malignancies. In this context, a recent study by Contag and coworkers³⁰  reported disturbed erythropoiesis in mouse spleens after the transplantation of bone marrow cells isolated from HO-1^+/− mice. It is likely that that decreased macrophage HO-1 will diminish heme iron recycling from senescent erythrocytes and cause abnormal erythropoiesis in transplanted mice. Moreover, these authors did not investigate HO-1 expression in erythroblasts of normal or transplanted mice. Because this study did not differentiate between erythroid and macrophage HO-1 function, it is not possible to relate the findings of Contag et al to our current report.

In the present work, we have demonstrated that HO-1 is present in erythroid cells and that its expression is upregulated during erythroid differentiation. When overexpressed in MEL cells, HO-1 impairs the hemoglobin synthesis pathway by inhibiting TfR expression and consequently iron uptake, heme synthesis, and globin expression. The key element in this chain of events is the TfR, whose expression in erythroid cells requires appropriate levels of heme for its transcription and likely also translation²⁵  by modulating HRI activity. On the other hand, HO-1 absence leads to an enhancement of hemoglobin synthesis that is likely caused by increased TfR transcription and translation and, consequently, iron uptake.

Ter119, a molecule associated with glycophorin A, is expressed on erythroid cells at the differentiation stages from early proerythroblast to mature erythrocytes.¹⁷  We demonstrated that the expression of both HO-1 mRNA (Figure 1B) and protein (Figure 1C) was increased 20-fold and 2.5-fold, respectively, in Ter119⁺ cells compared with Ter119⁻ cells. We have also found that HO-1 increases after 48 and 72 hours of differentiation of FL cells (Figure 1D-F) and that this upregulation can be dramatically inhibited by the heme synthesis inhibitor SA (Figure 1D-F). This finding strongly suggests that the enhancement of HO-1 levels in FL cells is mediated by heme, a well-known inducer of this enzyme in numerous cell types.^5,26,31  This is the first report showing that this mechanism operates in erythroid cells as well. Heme is not only the essential component of hemoglobin, it is also a key regulatory molecule involved in the erythroid differentiation process. Considering these facts, it is surprising that erythroid cells not only contain, but even upregulate, HO-1 during their differentiation. Thus, HO-1 appears to be part and parcel of the physiological erythroid differentiation program, but does not seem to, under physiological conditions, catabolize heme associated with hemoglobin.

The increase in ferritin levels in MELHO-1 cells (Figure 2C-D) is likely caused by a more efficient translation of this protein resulting from iron liberated from heme as an outcome of HO-1 overexpression. This conjecture was confirmed by our finding that the HO-1 inhibitor, SnPP, decreased ferritin levels in MELHO-1 cells (Figure 2C-D). Intriguingly, SnPP decreased ferritin levels also in control of MEL cells (Figure 2C-D), suggesting that both the synthesis of heme and its catabolism are involved in maintaining appropriate levels of heme in erythroid cells. It is tempting to speculate that, in control cells, HO-1 regulates the level of “uncommitted” heme pool, whereas unphysiologically high levels of HO-1 can also attack heme associated with hemoglobin or, at least, earmarked for globin binding. This hypothesis is supported by our finding that hemoglobin and heme levels decrease in MELHO-1 cells (Figure 2A-B).

We deemed it important to also investigate iron metabolism indices and some aspects of erythroid differentiation in erythroid cells lacking HO-1. We demonstrated that FL cells, obtained from FL/HO-1^−/− mice, accumulated more hemoglobin (Figure 4C-D) when induced to differentiate with EPO. Moreover, after 24 and 48 hours of differentiation, FL/HO-1^−/− had significantly increased TfR mRNA (Figure 5A) and protein (Figure 5D-E) levels; the presence of TfR at the cell membrane increased at 0 and 24 hours of differentiation (Figure 5B). The likely explanation for the enhanced expression of FL/HO-1^−/− cells’ TfR is their heme-mediated increase at both transcriptional and translational levels (reviewed in Ponka and Lok ²⁵ ). There is a tendency for increased uptake of iron and its incorporation into heme in FL/HO-1^−/− cells (supplemental Figure 4), but this difference is statistically significant only at 0 hours. At 0 and 48 hours of differentiation, globin mRNA expression was also significantly higher in FL/HO-1^−/− cells (Figure 5C); this increase was associated with enlarged globin protein levels (Figure 5D-E).

These results suggest that an expansion of an “uncommitted” or “regulatory” heme pool leads to an increase in TfR levels (Figure 5A-B,D-E) and consequently enhanced globin expression (Figure 5C-E). Heme can be expected not only to promote the efficiency of globin translation^21,32  but also its transcription.^33,34  Similarly, TfR levels and iron uptake were increased in HO-1 knockout reticulocytes (Figure 6A-B). We also observed a trend of increased iron incorporation into heme in HO-1 knockout reticulocytes compared with wild-type reticulocytes (Figure 6B). Additionally, eIF2α-P levels were reduced in HO-1 knockout reticulocytes (Figure 6C). Therefore, the effects of HO-1 deficiency on erythroid iron metabolism and heme synthesis are preserved up to the very late stages of the terminal erythroid differentiation.

Compared with control cells, ferritin protein expression is dramatically reduced in both nondifferentiated and differentiated FL/HO-1^−/− cells (Figure 5D-E). This finding strongly suggests that in the absence of HO-1, there is a reduction in the cellular labile iron pool, which is highly likely as a result of increased heme stability.

HO-1–deficient mice developed anemia associated with abnormally low serum iron levels.³⁵  HO-1 deficiency in humans appears as an extremely rare condition, with only 2 live births reported.^36,37  Both patients presented severe anemia.^36,37  Hence, our results showing that FL/HO-1^−/− cells have their hemoglobin synthesis pathway induced (Figures 4C-D and 5A-E) appear to be contradictory to the previous reports.^35-37  However, both mice and human HO-1–deficient phenotypes derive from the essential role of this enzyme in the recycling of hemoglobin iron from senescent erythrocytes by reticuloendothelial macrophages.^35,38  Macrophages recycle approximately 25 mg of iron daily, most of which is then transported to the bone marrow for hemoglobin synthesis in developing erythroid cells.³⁹  Hence, it is not appropriate to compare the phenotype occurring in humans and mice lacking functional HO-1 with our results conducted on FL/HO-1^−/− cells.

The next question is whether developing RBCs could benefit from having a higher than physiological rate of hemoglobin synthesis; we surmise that they would not. First, hyperchromia of erythrocytes would predictably lead to abnormal membrane deformability.⁴⁰  Second, it was shown that there is a critical mean corpuscular hemoglobin concentration, approximately 22 g/dL, above which DNA replication is not possible.⁴¹  This means that the number of cell divisions of erythroblasts with an artificially enhanced rate of hemoglobinization would be lower; this could disturb erythropoiesis. We have recently embarked on the generation of conditional erythroid-specific Hmox1^−/− mice with the aim of further assessing the role of HO-1 in erythropoiesis.

The results reported here have important ramifications for our understanding of erythroid differentiation, both in normal and some pathological conditions. First, it has been proposed that most mammalian cells contain a “free” or “uncommitted” heme pool, serving both precursor and regulatory functions.¹  This study strengthens the idea that such a pool is also present in erythroid cells^42,43  by demonstrating that the inhibition of HO-1 activity decreases ferritin levels in MEL cells (Figure 2C-D and previous discussion). It has been estimated that the concentrations of “free” heme in young rat reticulocytes are 10 μM or lower.⁴³  As mentioned, the “regulatory” pool of heme is essential for synthesis of globin chains at both the transcriptional and translational levels. Additionally, in erythroid cells, heme serves as a positive feedback regulator that maintains high TfR levels.²⁵ 

Second, the aberrant expression of HO-1 may play a role in the pathophysiology of thalassemia. In the normal assembly of adult hemoglobin (α₂β₂), the synthesis of globin chains is very tightly coordinated. However, in β-thalassemia, production of β-globin decreases and excess α-globin accumulates; in α-thalassemia, this process occurs in reverse.^44,45  Unpaired globin chains that accumulate in both types of thalassemic erythroblasts have heme molecules attached to them. α-subunits, which accumulate in β-thalassemia, are highly unstable and susceptible to autoxidation, likely liberating the heme that would be expected to induce HO-1. Oxidative stress is a key pathogenic factor involved in damaging β-thalassemic erythroblasts,⁴⁶  which also exhibit a significantly higher level of apoptosis.⁴⁷  Indeed, we found that MELHO-1 cells generate significantly higher levels of reactive oxygen species than control cells (supplemental Figure 5A) and display elevated apoptosis (supplemental Figure 5B), suggesting that elevated levels of HO-1 could play a role in the pathophysiology of β-thalassemic erythroblasts. Additionally, our finding of elevated apoptosis in MELHO-1 cells is congruent with an earlier report that the inhibition of heme synthesis induces apoptosis in human erythroid progenitor cells.⁴⁸  Also, membranes of β-thalassemic erythrocytes,^49,50  and likely also erythroblasts, contain iron—a substrate for a Fenton reaction—that can be derived from heme because of the enhanced activity of HO-1.

In conclusion, our results bring to light the importance of HO-1 expression for erythroid development and expand our knowledge about the fine regulation of hemoglobin synthesis in erythroid cells. Our results also indicate that HO-1 plays an important role as a coregulator of the erythroid differentiation process. Moreover, HO-1 expression must be tightly regulated during RBC development, because imbalance of its expression may lead to serious consequences for the hemoglobin biosynthesis pathway (Figure 7A-B).

The authors thank Dr Marc Mikhael for his valuable suggestions and advice. We are also grateful to 4 anonymous reviewers for their valuable comments.

This work was supported in part by the Canadian Institutes for Health Research (D.G.-S., P.P., A.S., M.M.J.), the Czech Science Foundation (P305/11/1745) (M.H.), and by the Education for Competitiveness Operational Programme of the Czech Ministry of Education, Youth and Sports (CZ.1.07/2.3.00/20.0164).

Contribution: D.G.-S. and M.S. performed research, analyzed data, and wrote the manuscript; M.H. and M.M.J. performed research and analyzed data; J.A.B.C. analyzed data; A.S. analyzed data and wrote the manuscript; and P.P. conceived the study, analyzed data, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Prem Ponka, Lady Davis Institute for Medical Research, Jewish General Hospital, and the Department of Physiology, McGill University, 3755 Chemin Cote-Ste-Catherine, Montréal, QC, H3T 1E2, Canada; e-mail: prem.ponka@mcgill.ca.