An essential cell-autonomous role for hepcidin in cardiac iron homeostasis.

Hepcidin is the master regulator of systemic iron homeostasis. Derived primarily from the liver, it inhibits the iron exporter ferroportin in the gut and spleen, the sites of iron absorption and recycling respectively. Recently, we demonstrated that ferroportin is also found in cardiomyocytes, and that its cardiac-specific deletion leads to fatal cardiac iron overload. Hepcidin is also expressed in cardiomyocytes, where its function remains unknown. To define the function of cardiomyocyte hepcidin, we generated mice with cardiomyocyte-specific deletion of hepcidin, or knock-in of hepcidin-resistant ferroportin. We find that while both models maintain normal systemic iron homeostasis, they nonetheless develop fatal contractile and metabolic dysfunction as a consequence of cardiomyocyte iron deficiency. These findings are the first demonstration of a cell-autonomous role for hepcidin in iron homeostasis. They raise the possibility that such function may also be important in other tissues that express both hepcidin and ferroportin, such as the kidney and the brain.

Introduction

As a constituent of hemoproteins, iron-sulphur proteins and other functional groups, iron is essential for cellular functions. Conversely, excess iron participates in cytotoxic Fenton-type chemical reactions. Thus, both iron deficiency and iron overload are detrimental to the cell. Therefore, the healthy functioning of tissues requires tight control of intracellular iron levels. These in turn are dependent both on cellular homeostatic pathways controlling iron uptake, usage, and storage, and on systemic pathways controlling iron levels in the plasma. In mammals, cellular iron homeostasis is controlled by the Iron Regulatory Proteins IRPs. Intracellular iron levels control the degradation of IRP2 and the conformational switch that confers the RNA-binding function of IRP1. IRPs in turn control the levels of iron uptake proteins such as transferrin receptor 1 (TfR1) and divalent metal transporter (DMT1), and the iron storage protein ferritin (Rouault and Klausner, 1997; Rouault, 2006). Systemic iron homeostasis is controlled by the hepcidin/ferroportin axis at the sites of iron entry into the circulation. Ferroportin (FPN), which is encoded by the Solute Carrier Family 40 Member 1 (Slc40a1), is the only known mammalian iron export protein and mediates iron release into the circulation from duodenal enterocytes, splenic reticuloendothelial macrophages and hepatocytes, the sites of iron absorption, recycling and storage respectively (Ganz, 2005; Donovan et al., 2005). FPN-mediated iron release is antagonized by the hormone hepcidin, also known as hepcidin antimicrobial peptide (HAMP). Produced primarily in the liver, hepcidin binds to and induces internalization of FPN, thereby limiting iron release into the circulation and its availability to peripheral tissues (Nemeth et al., 2004; Qiao et al., 2012). The importance of the HAMP/FPN axis is illustrated by diseases of systemic iron overload such as hemochromatosis and β-thalassaemia, where hepcidin production is impaired (Camaschella, 2005; Musallam et al., 2012), and in anaemia of chronic disease where hepcidin production is inappropriately elevated (De Falco et al., 2013; Nemeth and Ganz, 2014).

Other than the liver, hepcidin is also found in tissues with no recognized role in systemic iron homeostasis, including the heart (Merle et al., 2007), the brain (McCarthy and Kosman, 2014), the kidney (Kulaksiz, 2005) and the placenta (Evans et al., 2011). The function of this extra-hepatic hepcidin remains unknown, but one hypothesis is that it is involved in local iron control. Relevant to this hypothesis are our recent findings that FPN is also present in the heart, that it is essential for cardiomyocyte iron homeostasis and that its cardiomyocyte-specific deletion leads to fatal cardiac iron overload in mice (Lakhal-Littleton et al., 2015). Therefore, we hypothesised that cardiac HAMP regulates cardiac FPN, and that such regulation is important for local iron control and for cardiac function.

To test this hypothesis, we generated two novel mouse models; the first with a cardiomyocyte-specific deletion of the Hamp gene, and the second, with cardiomyocyte-specific knock-in of Slc40a1 C326Y, that encodes a FPN with intact iron export function but impaired HAMP binding (Schimanski et al., 2005; Drakesmith et al., 2005). We studied cardiac function and iron homeostasis longitudinally in both models and report that both develop fatal cardiac dysfunction and metabolic derangement as a consequence of cardiomyocyte iron deficiency. This occurs against a background of otherwise normal systemic iron homeostasis. Both cardiac dysfunction and metabolic derangement are prevented by intravenous iron supplementation.

Our findings demonstrate that, at least in the cardiomyocyte, endogenously-derived HAMP plays an essential role in cellular, rather than systemic iron homeostasis. It does this through the autocrine regulation of cardiomyocyte FPN. Disruption of this cardiac HAMP/FPN leads to fatal cardiac dysfunction.

Currently, there is considerable clinical interest in strategies that target the HAMP/FPN axis for the treatment of systemic iron overload and iron deficiency. Our findings suggest that these strategies may additionally alter cardiac iron homeostasis and function. Other than the heart, both HAMP and FPN are also found in the brain, kidney and placenta (McCarthy and Kosman, 2014; Kulaksiz, 2005; Evans et al., 2011; Rouault, 2013; Moos and Rosengren Nielsen, 2006; Bastin et al., 2006; Wolff et al., 2011). A pertinent question is the extent to which our findings in the heart extend to those tissues.

Results

Hamp expression and regulation in the heart

Hamp mRNA levels were approximately 30 fold lower in the adult mouse heart than in the liver (Figure 1A). Next, we examined the regulation of cardiac Hamp mRNA and HAMP protein in response to dietary iron manipulation, having first established that this dietary manipulation altered cardiac and liver iron levels (Figure 1—figure supplement 1). In both tissues, Hamp mRNA levels were decreased by the iron-deficient diet (Fe 2–5 ppm) and increased by the iron-loaded diet (Fe 5000 ppm) (Figure 1A). At the protein level, while changes in hepatic HAMP protein mirrored changes in its mRNA levels, cardiac HAMP protein was increased by the iron-deficient diet and unaffected by the iron-loaded diet (Figure 1B).

Figure 1.

Hepcidin expression and regulation in the heart.

(A) Relative Hamp mRNA expression in heart and liver of adult C57BL/6 mice, under control conditions and after provision of low or high iron diets. *p=0.047, 0.001 respectively relative to control hearts, †p = 0.006, 0.019 respectively relative to control livers. (B) Corresponding immunohistochemical staining for HAMP in heart and liver. (C) Relative Hamp mRNA expression in primary adult mouse cardiomyocytes cultured under control conditions or in presence of FAC or DFO. *p=0.023, 0.001 and 0.014 respectively relative to control. †p = 0.024, 0.037, 0.016 and 0.037 respectively relative to control at the same timepoint. (D) Corresponding HAMP protein levels in supernatants of primary cardiomyocytes. DFO treatment was carried alone (DFO) or presence of Furin inhibitor (DFO+CMK). *p=0.002, 0.020, 0.028, 0.014, 0.015 respectively relative to control at the same timepoint. (E) Relative Hamp expression in heart and liver of 3 month old Hamp and Hamp mice. *p=0.018 relative to cardiac Hamp in Hamp controls. (F) Corresponding immunohistochemical staining for HAMP in heart and liver. All values are plotted as mean ± SEM. Scale bar = 20 µm. n = 3 per group unless otherwise stated.

DOI: http://dx.doi.org/10.7554/eLife.19804.003

Total elemental iron levels in heart and liver of C57BL/6 mice, under control conditions and after provision of low (Fe 5 ppm) or high iron (Fe 5000 ppm) diets from weaning for six weeks. *p=0.037 and 0.033 respectively relative to control heart, †p = 0.010 and 0.005 relative to control liver. n = 3. Data are represented as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.004

(A) Relative Furin mRNA in primary adult mouse cardiomyocytes under control conditions and following treatment with DFO or FAC. *p=0.004, 0.001 and 0.001 respectively relative to control at the respective timepoint. (B) Relative Furin mRNA in hearts of mice provided control diet or iron-deficient diet (2–5 ppm) or iron-loaded diet (5000 ppm) from weaning for six weeks. *p=0.015 relative to control diet. n = 3. Data are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.005

Relative Hamp mRNA expression in primary adult mouse cardiomyocytes under control conditions and following treatment with DFO or FAC, in presence or absence of Furin inhibitor CMK. n = 3. Data are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.006

HAMP protein was measured by ELISA in supernatants of primary adult cardiomyocytes, derived from Hamp or Hamp mice and cultured under control conditions or in presence of FAC or DFO. n = 3. Data are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.007

(A) Staining with HAMP antibody (Abcam ab30760) in liver and heart of C57BL/6 mice is completely abrogated by co-incubation with hepcidin-25 blocking peptide (Abcam ab31875). Scale bar = 20 µm. (B) Loss of HAMP staining in Hamp;;Myh6.Cre+ hearts (Figure 1F) is consistent with the antibody detecting HAMP1 and not HAMP2, because Hamp2 mRNA expression is not altered in Hamp mice relative to Hamp controls, either with control diet or iron-deficient diet (six weeks from weaning). *p=0.007 and 0.047 relative to Hamp control under respective diet. n = 3 per group. Values are plotted as mean ± SEM. N.S=not significant.

DOI: http://dx.doi.org/10.7554/eLife.19804.008

HAMP standard from mouse hepcidin ELISA (E91979Mu, Uscn) was diluted either in the kit’s own standard diluent, or in unconditioned growth medium alone, or containing 100 µmol/L DFO or 500 µmol/L FAC. ELISA was performed as per manufacturer’s instructions.

DOI: http://dx.doi.org/10.7554/eLife.19804.009

Figure 1—figure supplement 1.

Cardiac and liver iron following dietary iron manipulation.

Total elemental iron levels in heart and liver of C57BL/6 mice, under control conditions and after provision of low (Fe 5 ppm) or high iron (Fe 5000 ppm) diets from weaning for six weeks. *p=0.037 and 0.033 respectively relative to control heart, †p = 0.010 and 0.005 relative to control liver. n = 3. Data are represented as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.004

To explore further the regulation of cardiac hepcidin by iron, we isolated primary adult cardiomyocytes from mice, then carried out a timecourse treatment with the iron chelator desferroxamine DFO, or with ferric citrate FAC. Under control conditions, Hamp mRNA was upregulated following addition of fresh cardiomyocyte culture medium (cardiomyocytes are cultured for 2 hr in serum-free medium prior to this). Relative to control cardiomyocytes at the respective timepoint, Hamp mRNA was increased by FAC from 4 hr of treatment, and decreased by DFO at 4, 8 and 16 hr of treatment (Figure 1C). HAMP protein, measured by ELISA in supernatants was also increased following addition of fresh medium. Relative to control cardiomyocytes at the respective timepoint, HAMP protein in supernatants was increased by DFO as early as 2 hr, but remained unchanged by FAC at all timepoints (Figure 1D). Thus, the direction of response of the Hamp mRNA and HAMP protein to iron levels in vitro mirrored the responses seen in vivo.

Next, we aimed to understand the mechanisms underlying increased HAMP secretion in DFO-treated cardiomyocytes. To this end, we investigated the role of the prohormone convertase Furin, which in hepatocytes, is required for cleavage of the propeptide and the release of the mature HAMP peptide (Valore and Ganz, 2008). Furin expression has been reported in the heart (Beaubien et al., 1995; Ichiki et al., 2013), and we also found that its expression was upregulated in the hearts of mice provided an iron-deficient diet and in cardiomyocytes treated with DFO (Figure 1—figure supplement 2). When we measured HAMP in supernatants of cardiomyocytes treated with the Furin inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone (CMK), we found no increase in HAMP release following DFO treatment (Figure 1D). We confirmed that Hamp mRNA levels in cardiomyocytes were not altered by CMK treatment (Figure 1—figure supplement 3). Together, these results indicate that Furin upregulation mediates increased HAMP secretion from iron-deficient cardiomyocytes.

Figure 1—figure supplement 2.

Furin regulation by iron.

(A) Relative Furin mRNA in primary adult mouse cardiomyocytes under control conditions and following treatment with DFO or FAC. *p=0.004, 0.001 and 0.001 respectively relative to control at the respective timepoint. (B) Relative Furin mRNA in hearts of mice provided control diet or iron-deficient diet (2–5 ppm) or iron-loaded diet (5000 ppm) from weaning for six weeks. *p=0.015 relative to control diet. n = 3. Data are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.005

Figure 1—figure supplement 3.

Relative Hamp mRNA expression in cardiomyocytes following treatment with Furin inhibitor CMK.

Relative Hamp mRNA expression in primary adult mouse cardiomyocytes under control conditions and following treatment with DFO or FAC, in presence or absence of Furin inhibitor CMK. n = 3. Data are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.006

Having established that hepcidin is found in cardiomyocytes, we then aimed to define its function. To this end, we generated cardiomyocyte-specific Hamp knockout mice Hamp by crossing in-house conditional Hamp floxed (fl) mice with mice transgenic for Cre recombinase under the control of cardiomyocyte-specific promoter Myosin Heavy Chain 6 (Myh6.Cre+). Hamp mRNA (Figure 1E) and HAMP protein (Figure 1F) in the hearts of Hamp mice were significantly reduced compared to the hearts of Hamp controls. Furthermore, compared to Hamp cardiomyocytes, levels of HAMP protein in the supernatants of cardiomyocytes from Hamp mice were either greatly reduced or undetectable, both under baseline conditions and following treatment with DFO or FAC (Figure 1—figure supplement 4). Near complete ablation of the cardiac Hamp mRNA and HAMP protein in Hamp mice confirmed that cardiomyocytes were the primary site of hepcidin expression in the heart. Liver Hamp mRNA (Figure 1E) and HAMP protein (Figure 1F) were not different between Hamp and Hamp controls, consistent with the cardiac-specific nature of Hamp gene deletion.

Figure 1—figure supplement 4.

HAMP in supernatants of primary cardiomyocytes.

HAMP protein was measured by ELISA in supernatants of primary adult cardiomyocytes, derived from Hamp or Hamp mice and cultured under control conditions or in presence of FAC or DFO. n = 3. Data are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.007

Also consistent with this cardiac-specific deletion, Hamp mice had normal levels of liver iron stores and circulating markers of iron homeostasis when compared to Hamp controls, demonstrating that loss of cardiac hepcidin did not affect systemic iron homeostasis. In addition, circulating HAMP levels were not reduced in the serum of Hamp mice, suggesting that cardiac hepcidin does not contribute significantly to circulating HAMP levels (Table 1).

Table 1.

Indices of systemic iron in six month old Hamp and Hamp;Myh6.Cre+ mice.

n = 6 per group. All values are shown as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.010

	Hampfl/fl	Hampfl/fl;Myh6.Cre+
liver total elemental iron (ng/mg tissue)	96.3 ± 12.2	88.7 ± 19.2
liver ferritin (µg/mg total protein)	0.65 ± 0.04	0.64 ± 0.05
serum iron (µmol/L)	28.60 ± 7.20	31.50 ± 8.40
serum ferritin (mg/L)	1.81 ± 0.04	1.88 ± 0.29
hemoglobin (g/L)	122.7 ± 11.5	116.0 ± 11.9
serum hepcidin (µg/L)	23.5 ± 7.6	23.9 ± 10.40

Fatal cardiac abnormalities in Hamp;Myh6.Cre+ mice

To determine the effects of loss of cardiac hepcidin, we first assessed the cumulative survival of Hamp mice and Hamp littermate controls over a period of 52 weeks. Significantly greater mortality was observed amongst Hamp mice, with only 29% of animals surviving to 52 weeks, compared with 90% of Hamp controls. The median survival of Hamp mice was 28 weeks, whereas the majority of Hamp controls were still alive at 52 weeks (Figure 2A).

Figure 2.

Fatal cardiac abnormalities in Hamp;Myh6.Cre+ mice.

(A) Cumulative survival of Hamp mice (n = 50) and Hamp littermate controls (n = 47) over 52 weeks. (B) Representative H and E longitudinal heart sections from a six month old Hamp mouse and Hamp littermate control. (C) Representative WGA cardiac staining from a six month old Hamp mouse and Hamp littermate control. (D) Quantitation of cardiomyocyte size based on WGA staining in six month old Hamp mice and Hamp littermate controls. *p=0.001 relative to Hamp littermate controls. (E) Relative expression of the hypertrophic gene markers Myh7 and Nppb in hearts of 6 month old Hamp mice and Hamp littermate controls. *p=0.001, 0.047 for the respective gene relative to Hamp littermate controls. (F) Representative cardiac in-situ TUNEL staining from a six month old Hamp mouse and Hamp littermate control. (G) Quantitation of percentage of apoptotic cardiomyocytes based on in-situ TUNEL staining in six month old Hamp mice and Hamp littermate controls *p=0.001 relative to Hamp littermate controls. (H) Representative midventricular Cine MR images of hearts from Hamp mice and Hamp controls at 3, 6 and 9 months of age. (I–K) Cine MRI measurements of LV lumen, at end-systole (LVES), end-diastole (LVED), and of ejection fraction (LVEF) in Hamp mice and Hamp littermate controls at three months (n = 8 per group), six months (n = 11 per group, *p=0.043 for LVES, 0.047 for LVED and 0.020 for LVEF) and nine months (n = 5 per group, *p=0.044 for LVES, 0.042 for LVED and 0.034 for LVEF). p values are relative to Hamp controls of the respective age. All Values are plotted as mean ± SEM. n = 3 per group unless otherwise stated. Scale bar = 50 µm.

DOI: http://dx.doi.org/10.7554/eLife.19804.011

DOI: http://dx.doi.org/10.7554/eLife.19804.012

Source data file for Figure 2I,J and K.

Six-month old mice were sacrificed for assessment of cardiac morphology, which showed gross enlargement of the left ventricle (LV) in Hamp hearts compared to Hamp controls (Figure 2B). Assessment of cardiomyocyte size by wheat germ agglutinin (WGA) staining confirmed that Hamp cardiomyocytes were hypertrophied (Figure 2C–D). This was accompanied by upregulation of expression of hypertrophic gene markers myosin heavy chain (Myh7) and natriuretic peptide precursor (Nppb) (Figure 2E). TUNEL staining for in-situ detection of cell death also showed significantly greater apoptosis in the hearts of Hamp mice than in Hamp controls (Figure 2F–G).

To characterise further the phenotype caused by loss of cardiac hepcidin, we used cine MRI in anaesthetised Hamp mice and Hamp littermate controls at 3, 6 and 9 months of age. Mid-ventricular cine MR images showed no differences between the two genotypes at 3 months of age. At 6 and 9 months of age, cine MR images showed marked enlargement of the LV in Hamp mice compared to Hamp controls (Figure 2H). Formal quantitation of cardiac parameters by cine MRI confirmed enlargement of the LV lumen in Hamp mice, both at end-systole (LVES) (Figure 2I) and at end-diastole (LVED) (Figure 2J), accompanied by a decrease in LV ejection fraction (LVEF) from 62% to 42% (Figure 2K). Other parameters of cardiac performance were not significantly altered between mice from the two genotypes (Table 2). Taken together, histological examination of the hearts and cine MRI studies indicated that Hamp mice developed fatal LV dysfunction with reduced LVEF.

Table 2.

Non-LV parameters of cardiac function are not altered between Hamp and Hamp;;Myh6.Cre+ mice.

Cine MRI measurements of cardiac function in Hamp mice and Hamp controls at three months (n = 8 per group), six months (n = 11 per group) and nine months (n = 5 per group) of age. Values are shown as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.013

	3 months		6 months		9 months
	Hamp fl/fl	Hamp fl/fl; Myh6.Cre+	Hamp fl/fl	Hamp fl/fl; Myh6.Cre+	Hamp fl/fl	Hamp fl/fl; Myh6.Cre+
Average mass (mg)	70.51 ± 7.47	70.38 ± 5.33	72.59 ± 5.38	82.54 ± 11.32	78.30 ± 4.92	83.07 ± 5.44
RVED lumen (µl)	31.37 ± 2.55	26.81 ± 2.30	33.12 ± 3.19	30.63 ± 2.35	38.39 ± 3.88	39.95 ± 3.00
RVES lumen (µl)	7.43 ± 0.79	5.23 ± 0.67	8.70 ± 1.22	8.34 ± 1.06	13.36 ± 2.21	15.51 ± 2.47
RVEF (%)	76.42 ± 1.32	79.88 ± 2.80	73.66 ± 2.50	73.42 ± 2.22	65.20 ± 4.36	61.95 ± 3.53
Stroke volume (µl)	25.47 ± 1.99	22.73 ± 2.11	25.70 ± 2.70	23.76 ± 1.85	29.79 ± 2.30	29.84 ± 1.59
Cardiac output (ml/min)	10.34 ± 1.06	9.62 ± 0.90	10.72 ± 1.07	10.53 ± 1.05	11.46 ± 1.15	11.99 ± 1.38
Heart Rate (bpm)	404.07 ± 18.69	426.90 ± 17.94	419.64 ± 19.05	436.14 ± 14.30	384.84 ± 28.55	400.97 ± 41.20
Heart/body weight ratio x1000	2.80 ± 0.18	3.09 ± 0.26	2.51 ± 0.23	2.98 ± 0.37	2.36 ± 0.28	2.88 ± 0.27

Hamp ;

Hamp ;

Hamp ;

Such changes in cardiac performance could not be attributed to Cre recombinase toxicity in the heart as we have previously shown that Myh6.Cre+ mice have normal cardiac function compared to wild type littermate controls (Lakhal-Littleton et al., 2015).

The role of cardiomyocyte FPN in the phenotype of Hamp;Myh6.Cre+ mice

We examined FPN protein in the hearts of Hamp mice, and found that FPN protein was markedly upregulated compared to Hamp controls (Figure 3A), consistent with the idea that loss of cardiac HAMP was acting through upregulation of cardiomyocyte FPN. In order to test whether cardiac dysfunction arose from upregulation of cardiomyocyte FPN, we engineered mice where the Slc40a1 gene harbours a conditional cardiac-specific C326Y point mutation, which confers HAMP-resistance while conserving the iron export function of FPN (Schimanski et al., 2005; Drakesmith et al., 2005). We confirmed that the Slc40a1 C326Y fl allele produced the C326Y transcript specifically in the heart (Figure 3—figure supplement 1) and that Slc40a1 C326Y mice did not exhibit changes in systemic iron indices (Table 3). As seen in Hamp mice, cardiomyocyte FPN was indeed upregulated in Slc40a1 C326Y mice (Figure 3B).

Figure 3.

The role of cardiomyocyte FPN.

(A–B) Immunohistochemical staining for FPN in the hearts of three month old Hamp, Slc40a1 C326Y mice and respective controls. (C) Cumulative survival of Slc40a1 C326Y mice (n = 36) and Slc40a1 C326Y littermate controls (n = 31) over 52 weeks. (D) Representative H and E-stained longitudinal heart sections from a six month old Slc40a1 C326Y mouse and a Slc40a1 C326Y control. (E) Representative WGA cardiac staining from a six month old Slc40a1 C326Y mouse and Slc40a1 C326Y control. (F) Quantitation of cardiomyocyte size based on WGA staining. n = 3 per group. *p=0.001 relative to Slc40a1 C326Y controls. (G) Relative expression of Myh7 and Nppb in hearts of 6 month old Slc40a1 C326Y mice and Slc40a1 C326Y controls. n = 3 per group. *p=0.032, 0.044 for the respective gene relative to Slc40a1 C326Y controls. (H) Representative cardiac TUNEL staining from a six month old Slc40a1 C326Y mouse and Slc40a1 C326Y control. (I) Quantitation of percentage of apoptotic cardiomyocytes based on TUNEL staining, n = 3 per group. *p=0.0003 relative to Slc40a1 C326Y controls. (J–L) Cine MRI measurements of LVES, LVED and LVEF in Slc40a1 C326Y mice and Slc40a1 C326Y controls at three months (n = 6 per group), six months (n = 6 per group, *p=0.003 for LVES, 0.043 for LVED and 0.001 for LVEF) and nine months (n = 5 per group, *p=0.033 for LVES, 0.047 for LVED and 0.023 for LVEF). P values are relative to Slc40a1 C326Y controls of the same age. (M) Percentage Fe55 efflux in cardiomyocytes from Hamp mice, Slc40a1 C326Y mice and respective controls, in presence or absence of HAMP peptide. *p=0.018, 0.006 and 0.007 respectively (N) Elemental iron levels in cardiomyocyte fractions (CF) from the hearts of Hamp mice, Slc40a1 C326Y mice and their respective controls. n = 4 per group. *p=0.032, 0.044, 0.047 and 0.031 respectively. (O–P) Relative TfR1 (*p=0.038, 0.001) and Fpn (*p=0.039, 0.047) expression in hearts of 3 month old Hamp mice, Slc40a1 C326Y mice and their respective controls. All values are plotted as mean ± SEM. Scale bar = 50 µm.

DOI: http://dx.doi.org/10.7554/eLife.19804.014

DOI: http://dx.doi.org/10.7554/eLife.19804.015

Total mRNA was extracted from the liver and heart of an adult a Slc40a1 C326Y mouse, reverse transcribed using primers for exon 7 of the Slc40a1 mRNA transcript. Products were sequenced to confirm successful heterozygous expression of the C326Y transcript in the heart but not in the liver.

DOI: http://dx.doi.org/10.7554/eLife.19804.016

Cre+ mice, Slc40a1 C326Y mice and their respective controls at three months and 6 months of age. n = 4 per group. Data are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.017

Figure 3—figure supplement 1.

Characterisation of Slc40a1 C326Y;Myh6.Cre+ mice.

Total mRNA was extracted from the liver and heart of an adult a Slc40a1 C326Y mouse, reverse transcribed using primers for exon 7 of the Slc40a1 mRNA transcript. Products were sequenced to confirm successful heterozygous expression of the C326Y transcript in the heart but not in the liver.

DOI: http://dx.doi.org/10.7554/eLife.19804.016

Table 3.

Characterisation of Slc40a1 C326Y mice.

Indices of iron status in Slc40a1 C326Y and Slc40a1 C326Y mice at six months of age (n = 4 per group). Values are shown as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.018

	Slc40a1 C326Yfl/fl	Slc40a1 C326Y fl/fl; Myh6.Cre+
liver total elemental iron (ng/mg tissue)	92.77 ± 21.30	84.00 ± 26.00
liver ferritin (µg/mg total protein)	0.87 ± 0.06	0.92 ± 0.05
serum iron (µmol/L)	27.30 ± 5.20	29.60 ± 7.20
serum ferritin (mg/L)	2.10 ± 0.04	2.20 ± 0.15
hemoglobin (g/L)	125.70 ± 8.80	126.00 ± 12.30
serum hepcidin (µg/L)	25.90 ± 11.60	27.50 ± 8.40

Source data file for Figure 3JK and L.

Slc40a1 C326Y

Slc40a1 C326Y

Like Hamp mice, Slc40a1 C326Y mice also had increased mortality relative to their littermate controls (Figure 3C). We then determined whether Slc40a1 C326Y mice developed the same phenotype of cardiac dysfunction as Hamp mice. Histologically, Slc40a1 C326Y hearts from six month old mice also showed LV enlargement (Figure 3D), hypertrophied cardiomyocytes (Figure 3E–F), upregulation of hypertrophy markers Myh7 and Nppb (Figure 3G) and a greater degree of apoptosis compared to Slc40a1 C326Y controls (Figure 3H–I). When we measured cardiac performance by cine MRI, we found that Slc40a1 C326Y mice also developed LV dysfunction by 6 months of age, with a reduction in LVEF from 73% to 55% (Figure 3J–L).

The similarity between Hamp and Slc40a1 C326Y mice in terms of the nature and time course of cardiac dysfunction suggests a common mechanism of cardiac dysfunction involving upregulation of cardiomyocyte FPN. Therefore, we tested whether this upregulation of FPN resulted in increased iron efflux from cardiomyocytes. Iron Fe55 efflux was indeed significantly greater in cardiomyocytes isolated from Hamp and Slc40a1 C326Y hearts than in cardiomyocytes isolated from their respective controls (Figure 3M). Addition of exogenous mouse HAMP in the efflux medium inhibited the increase in Fe55 efflux from Hamp cardiomyocytes but not from Slc40a1 C326Y cardiomyocytes, consistent with the HAMP-resistant mutation in Slc40a1 C326Y cardiomyocytes. We hypothesised that upregulation of cardiac FPN and iron export in Hamp and Slc40a1 C326Y hearts caused cardiomyocyte iron depletion. To test this hypothesis, we quantified iron levels both in total hearts and in the isolated cardiomyocyte fractions at 3 months and 6 months of age. While iron levels in total hearts were not significantly different between any of the genotypes (Figure 3—figure supplement 2), the iron content of the cardiomyocyte fraction was significantly lower in Hamp and in Slc40a1 C326Y mice than in their respective controls (Figure 3N). Furthermore, expression of TfR1 mRNA was upregulated (Figure 3O) and Slc40a1 mRNA was downregulated (Figure 3P) in Hamp and Slc40a1 C326Y hearts relative to their respective controls, consistent with a transcriptional response to intracellular iron deficiency (Rouault and Klausner, 1997; Rouault, 2006; Ward and Kaplan, 1823). Together these results demonstrate that loss of either cardiac hepcidin or hepcidin responsiveness in the heart results in upregulation of cardiomyocyte FPN, and that cardiomyocytes of Hamp and Slc40a1 C326Y hearts are iron deficient as a result of upregulation of FPN-mediated iron export.

Figure 3—figure supplement 2.

Elemental iron levels in total hearts of Hamp

Cre+ mice, Slc40a1 C326Y mice and their respective controls at three months and 6 months of age. n = 4 per group. Data are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.017

The role of cardiomyocyte iron deficiency and metabolic derangement in cardiac dysfunction

As cardiomyocyte iron deficiency preceded the development of cardiac dysfunction, we hypothesised it is the cause of the cardiac phenotype in Hamp mice. To test this hypothesis, we treated Hamp mice and Hamp controls with fortnightly intravenous injections of ferric carboxymaltose solution containing 0.5 mg iron from three months of age, and confirmed the effects of this treatment on cardiac and systemic iron indices at 6 months of age (Table 4). At this timepoint, we performed cine MRI and found that the LV enlargement and the reduced LVEF, seen in untreated Hamp mice, were prevented in iron-treated Hamp mice (Figure 4A–C). The transcriptional response to intracellular iron deficiency in untreated Hamp mice (upregulation of TfR1 mRNA and downregulation of Slc40a1 mRNA relative to Hamp controls), was absent in iron-treated Hamp mice, consistent with correction of cardiomyocyte iron deficiency (Figure 4D–E). Prevention of cardiac dysfunction in Hamp mice by intravenous iron treatment confirms the causal relationship between cardiomyocyte iron deficiency and cardiac dysfunction in this setting.

Table 4.

Effect of intravenous iron treatment on iron indices.

Total cardiac and liver elemental iron, serum iron and circulating HAMP in 6-month old untreated and I.V iron-treated Hamp mice and Hamp littermate controls. Treated mice were injected with 0.5 mg iron fortnighly from the age of 3 months. Tissues and serum were harvested 12 hr after the final injection. n = 5 per group. *p<0.05 relative to untreated Hamp mice. †p < 0.05 relative to untreated Hamp. Values are shown as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.021

	Hampfl/fl		Hamp fl/fl;Myh6.Cre+
	untreated	treated with I.V iron	untreated	treated with I.V iron
cardiac total elemental iron (ng/mg tissue)	82.2 ± 16.9	331.3 ± 21.5*	74.9 ± 7	399.8 ± 68.5†
liver total elemental iron (ng/mg tissue)	100.4 ± 11	2527.6 ± 27.63*	96.3 ± 14	2258.2 ± 239.9†
serum iron (µmol/L)	30.09 ± 6.37	74.48 ± 17.96*	31.5 ± 6.9	80.12 ± 24.9†
serum hepcidin (µg/L)	27.41 ± 6.7	237.3 ± 16.7*	28.9 ± 9.4	209.8 ± 38.8†

Figure 4.

The role of cardiomyocyte iron deficiency and metabolic derangement in cardiac dysfunction.

(A–C) Cine MRI measurements of LV lumen, at end-systole (LVES, *p=0.048), end-diastole (LVED, *p=0.031), and of ejection fraction (LVEF, *p=0.004) in 6-month old untreated (-iron) and I.V iron-treated (+iron) Hamp mice and Hamp littermate controls. n = 5 per group. (D–E) Relative TfR1 (*p=0.001) and Fpn (*p=0.002) expression in hearts of 6-month old untreated (-iron) and I.V iron-treated (+iron) Hamp mice and Hamp littermate controls. n = 4 per group. (F–H) Enzymatic activities of Aconitase I (*p=0.035, 0.041), Complex I (*p=0.004, 0.030) and Complex IV (*p=0.003, 0.026) in untreated (-iron) 3-month and 6-month old and in I.V iron-treated (+iron) 6-month old Hamp mice and Hamp littermate controls. n = 4 per group. (I) Representative EM micrographs of hearts from untreated (-iron) 3-month and 6-month old and I.V iron-treated (+iron) 6-month old Hamp mice and Hamp littermate controls. Scale bar = 2 µm. (J–L) Relative Hk2 (*p=0.010, 0.002), Eno (*p=0.014, 0.021) and Ldha (*p=0.003, 0.001) expression levels in hearts of untreated (-iron) 3-month and 6-month old and in I.V iron-treated (+iron) 6-month old Hamp mice and Hamp littermate controls. n = 4 per group. NS=not significant. All values are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.019

DOI: http://dx.doi.org/10.7554/eLife.19804.020

Source data file for Figure 4A,B and C.

Having confirmed a causal relationship between cardiomyocyte iron deficiency and cardiac dysfunction, we aimed to understand the mechanisms linking the two. Iron is a cofactor for several enzymes involved in metabolism (Meyer, 2008; Sono et al., 1996; Solomon et al., 2003), and metabolic derangement is a well-recognized precursor to cardiac dysfunction (Belke et al., 2000; Kakinuma et al., 2000; Ashrafian et al., 2007). Iron deficiency has been reported to reduce the levels and/or activities of key metabolic iron-containing enzymes, in cell lines, in the hearts of mice with impaired cardiac iron uptake (cardiac-specific TfR1 knockouts) and in the hearts of mice fed an iron-deficient diet (Dallman, 1986; Dhur et al., 1989; Xu et al., 2015; Oexle et al., 1999). Based on those studies, we postulated that iron deficiency in the cardiomyocytes of Hamp mice would also result in reduction in the activities of key metabolic iron-containing enzymes. To test this hypothesis, we measured the activities of the iron-sulphur containing enzyme Aconitase I as well as electron transport chain (ETC) complexes in cardiac lysates from Hamp hearts at 3 months and 6 months of age. We found that Aconitase I, Complex I and Complex IV activities were significantly reduced in Hamp hearts compared to Hamp controls at 3 months and 6 months of age, and that this reduction in activity was prevented in iron-treated 6-month old mice (Figure 4F–H). As ETC activity is essential to mitochondrial function, we examined whether Hamp hearts had signs of mitochondrial failure, and whether such mitochondrial failure was prevented by intravenous iron supplementation. By electron microscopy (EM), dilation of mitochondrial cristae was seen in Hamp hearts as early as 3 months of age and progressed further at 6 months of age. However, this was prevented in iron-treated Hamp mice, which had healthy-looking mitochondria at 6 months of age (Figure 4I). Similar changes in mitochondrial morphology were seen in the hearts of 3 month old Slc40a1 C326Y mice (data not shown).

Impairment of electron transport is known to drive glycolysis, as an alternative route of ATP production (Schönekess et al., 1997; Kato et al., 2010). Therefore, we examined the expression of a number of genes encoding glycolytic enzymes. We found that expression of genes encoding Hexokinase 2 (Hk2) which catalyses the first step of glycolysis, Enolase (Eno), which catalyses the penultimate step of glycolysis and of Lactate dehydrogenase A (Ldha) which catalyses the ultimate step of glycolysis were all significantly increased at 3 months and 6 months of age in hearts of untreated Hamp mice. This upregulation of glycolytic genes was not seen in iron-treated Hamp mice (Figure 4J–L). These results demonstrate that reduction in the activities of key iron-containing metabolic enzymes, mitochondrial failure and upregulated glycolysis precede the development of cardiac dysfunction in Hamp hearts, and are prevented by intravenous iron treatment.

Discussion

The major finding of this study is that cardiomyocyte hepcidin is required for autonomous cellular iron homeostasis. Loss of hepcidin responsiveness specifically in cardiomyocytes engendered the same effects as loss of cardiac hepcidin, demonstrating that cardiac hepcidin operates in an autocrine fashion by regulating cardiomyocyte FPN. A role in cellular iron homeostasis, and an autocrine mode of action for the HAMP/FPN axis have not been described previously in any other tissue. Indeed, hepcidin and FPN are better known to interact in an endocrine fashion, at the level of the gut, spleen and liver, to regulate systemic iron homeostasis.

A second important finding is that this cardiac HAMP/FPN axis is essential for normal heart function, and that its disruption leads to ultimately fatal cardiac metabolic and contractile dysfunction, even against a background of intact systemic iron homeostasis. Metabolic and contractile dysfunction are preceded by cardiomyocyte iron deficiency and prevented by intravenous iron supplementation, indicating a causal relationship between cardiomyocyte iron deficiency and cardiac dysfunction. This causal relationship has previously been demonstrated in studies using mice lacking cardiac TfR1, in which cardiac iron deficiency also affects the heart against a background of otherwise intact systemic iron homeostasis (Xu et al., 2015). Furthermore, the importance of metabolic derangement seen in our mouse models is also consistent with the findings in mice lacking the cardiacTfR1, and in mice and rats fed iron deficient diets (Dhur et al., 1989; Xu et al., 2015; Tanne et al., 1994; Walter et al., 2002). Correction of metabolic and contractile dysfunction by intravenous iron treatment likely involves not only increased iron availability for uptake into cardiomyocytes, but also the effects of increased circulating HAMP on cardiomyocyte FPN.

Surprisingly, we found that, both in vitro and in vivo, cardiac HAMP protein responded differently from its transcript to changes in iron levels. The current understanding of hepcidin regulation is based on studies of hepatic hepcidin. In that setting, release of the active mature HAMP peptide is dependent on cleavage of the propeptide by Furin (Valore and Ganz, 2008). We found that Furin inhibition increased iron export from Hamp but not from Hamp cardiomyocytes (Appendix 1—figure 1), demonstrating that cardiomyocytes secrete an active HAMP peptide in a Furin-dependent manner. Furthermore, we found that increased HAMP release from iron-deficient cardiomyocytes depended on Furin, and that cardiac Furin itself is upregulated by iron deficiency both in vitro and in vivo. The latter finding is consistent with the reported regulation of Furin by iron deficiency through Hypoxia-Inducible Factors HIFs (Silvestri et al., 2008). These data suggest that differential regulation of Furin by iron may explain the divergent effects of iron on Hamp transcript and HAMP protein. Comprehensive studies using cardiac-specific knockouts of putative regulators will be required to explore formally the regulation or otherwise, of cardiac hepcidin by pathways known to regulate hepatic hepcidin.

Appendix 1—figure 1.

Effect of Furin inhibitor on iron export in cardiomyocytes.

Fe55 efflux measured in primary adult cardiomyocytes from Hamp fl/fl and Hamp fl/fl;Myh6.Cre+ mice following culture in control medium (-CMK) or medium containing Furin inhibitor (+CMK) for 2 hr. n = 3. values are plotted as mean ± SEM. *p=0.027. †p = 0.024.

DOI: http://dx.doi.org/10.7554/eLife.19804.023

The upregulation of cardiac HAMP in mice fed an iron-deficient diet raises the possibility that it may be involved in protecting the heart in the setting of systemic iron deficiency. This hypothesis is supported by the finding that, when provided an iron-deficient diet, Hamp mice exhibited a greater cardiac hypertrophic response than their Hamp littermate controls (Appendix 1—figure 2).

Appendix 1—figure 2.

Effect of genotype on hypertrophic response to iron-deficient diet.

Heart/body weight ratio in Hampl;Myh6.Cre+ mice and littermate Hampl controls provided a control diet (Fe 200 ppm) or an iron-deficient diet (Fe 2–5 ppm) from weaning for six weeks. n = 5 per group, NS=not significant. *p=0.013, †p = 0.04. Values are shown as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.024

Intracellular iron levels are dependent both on cellular homeostatic pathways and on systemic iron availability in plasma. Therefore, the interplay between the cardiac and the systemic HAMP/FPN axes is important in determining cardiomyocyte iron levels. Some insight into this interplay is gained from comparing systemic and cardiac mouse models of disrupted iron homeostasis. It is interesting that ubiquitous Hamp knockout (Lakhal-Littleton et al., 2015) and ubiquitous Slc40a1 C326Y knock-in mice (Appendix 1—figure 3), both models of systemic iron overload, do not develop the cardiac dysfunction seen in cardiomyocyte-specific Hamp knockout and cardiomyocyte-specific Slc40a1 C326Y knock-in mice. This suggests that, while upregulation of cardiomyocyte FPN under conditions of normal iron availability (cardiac-specific models described in this study) is detrimental to cardiac function, it is protective under conditions of increased systemic iron availability (systemic models). Previously, we also showed that deletion of cardiomyocyte FPN resulted in fatal cardiomyocyte iron overload, preventable by dietary iron restriction (Lakhal-Littleton et al., 2015). Together, our studies demonstrate that iron levels within cardiomyocytes are a balance between cellular iron efflux which is regulated by the cardiac HAMP/FPN axis, and systemic iron availability which is regulated by the systemic HAMP/FPN axis (Figure 5).

Appendix 1—figure 3.

Normal cardiac function in ubiquitous Slc40a1 C326Y KI mice.

(A–C) Cine MRI measurements of LV lumen, at end-systole (LVES), end-diastole (LVED), and of ejection fraction (LVEF) in Slc40a1 C326YKI/+ mice and +/+ littermate controls at three months (n = 7 per group), six months (n = 7 per group) and nine months (n = 7 per group) of age. Values are plotted as mean ± SEM.

DOI: http://dx.doi.org/10.7554/eLife.19804.025

Figure 5.

Interplay between systemic and cardiac iron HAMP/FPN axes.

Cardiomyocyte iron content is determined by both systemic iron availability, which is regulated by liver HAMP, and by the cardiac HAMP/FPN axis, which regulates cardiomyocyte iron efflux. In the wild type heart, cardiac HAMP regulates the levels of cardiac FPN and iron release from cardiomyocytes. In this study, we have demonstrated that loss of cardiac HAMP (cardiac Hamp KO) or loss of cardiac HAMP responsiveness (cardiac Slc40a1 C326Y KI) result in cardiomyocyte iron deficiency due to increased cardiomyocyte FPN and iron release. Previously, we also demonstrated that loss of cardiomyocyte FPN caused cardiomyocyte iron overload. In these two sets of conditions, cardiomyocyte iron deficiency and cardiomyocyte iron overload cause cardiac dysfunction. We have also shown that upregulation of cardiac FPN occurs as a result of loss of either systemic HAMP or systemic HAMP responsiveness, and is protective against the otherwise detrimental effects of systemic iron overload.

DOI: http://dx.doi.org/10.7554/eLife.19804.022

Iron overload is detrimental to cardiac health, as demonstrated by iron overload cardiomyopathy in hemochromatosis and thalassemia major patients (Gulati et al., 2014). Our model of cardiac iron homeostasis implies that the cardiac HAMP/FPN axis may have a modifying effect on the severity of iron-overload cardiomyopathy. Thus, it would be interesting to explore whether differences in the levels of cardiac FPN and HAMP, possibly due to different local stimulatory and suppressive signals (e.g local inflammation, local ischemia), explain the reported lack of concordance between the degrees of cardiac iron overload and liver iron overload in a significant proportion of hemochromatosis and thalassemia major patients (Anderson et al., 2001; Noetzli et al., 2008).

Iron deficiency is also detrimental to cardiac health. Indeed, systemic iron deficiency correlates with functional and molecular markers of disease severity in patients with chronic heart failure (CHF) (Klip et al., 2013; Comín-Colet et al., 2013),and also appears to contribute to the risk of death after an episode of acute heart failure (Jankowska et al., 2014). Given the high prevalence of iron deficiency in patients with CHF, ranging between 30–50% (Klip et al., 2013; Nanas et al., 2006; Ezekowitz et al., 2003), the European Society of Cardiology recently recommended the assessment of iron deficiency as a comorbidity in CHF. Furthermore, several clinical trials have now established the benefits of intravenous iron supplementation in CHF patients, with or without anaemia (Ponikowski et al., 2015; Anker et al., 2009; Avni et al., 2012). The mechanisms underlying the anaemia-independent effects of iron deficiency and the benefits of intravenous iron in CHF patients are not fully understood. In light of the direct effect of cardiomyocyte iron deficiency on heart function, demonstrated in this and other studies, it would interesting to explore whether systemic iron deficiency in CHF patients is accompanied by cardiomyocyte iron deficiency, and whether correction of the later underlies the benefits of intravenous iron supplementation in non-anaemic patients.

Another open question is whether disruption of the cardiac HAMP/FPN axis contributes to the pathophysiology of heart disease. It has been shown, in the rat model of myocardial infarction (MI), that Hamp mRNA and HAMP protein are elevated in the ischemic myocardium during the acute phase (Simonis et al., 2010). In humans, circulating HAMP was shown to be elevated in the serum within 4 hr of MI, although the tissue source of this hepcidin was not identified (Suzuki et al., 2009). In addition, decreased cardiac HAMP expression has been reported in a transgenic mouse model of dilated cardiomyopathy, where the phenotype was ameliorated following transgenic overexpression of cardiac hepcidin (Zhang et al., 2012). Thus, further studies are warranted in humans to explore formally the role of the cardiac HAMP/FPN axis in the aetiology of heart disease.

Currently, there is considerable interest in targeting the HAMP/FPN axis for the treatment of iron overload and iron deficiency. Our studies suggest that such strategies may also impinge on cardiac iron homeostasis and function. Other than the heart, both FPN and hepcidin are also found in the brain, kidney and placenta (McCarthy and Kosman, 2014; Kulaksiz, 2005; Evans et al., 2011; Rouault, 2013; Moos and Rosengren Nielsen, 2006; Bastin et al., 2006; Wolff et al., 2011). It would be important to establish whether our findings in the heart extend to these tissues.

Materials and methods

Mice

All animal procedures were compliant with and approved under the UK Home Office Animals (Scientific Procedures) Act 1986. Both males and females were used in experiments, with the respective littermate control being of a matching sex.

The strategy for generating cardiac Hamp knockout mice is outlined in Appendix 1—figure 4. Briefly, a targeting vector was designed to introduce a floxed Hamp allele into C57BL/6N mouse ES cells (JM8F6) with exons 2 and 3, which encode the majority of the peptide, flanked by LoxP sites and a line of floxed mice was generated by blastocyst injection of targeted ES cells, as previous described (Lakhal-Littleton et al., 2015). Further breeding with a C57BL/6 Flp recombinase deleter mouse allowed removal of the Neomycin resistance cassette. Cardiac Hamp knockouts were then generated by by crossing Hampfl/fl animals with mice transgenic for Cre recombinase, under the control of cardiomyocyte-specific Alpha Myosin Heavy Chain (Myh6) promoter (B6.FVB-Tg(Myh6-cre)2182Mds/J).The subsequent breeding strategy was designed to produce cardiac Hamp knockouts and homozygous floxed controls (Hamp and Hamp respectively) in the same litter.

Appendix 1—figure 4.

Strategy for generation of Hamp;Myh6.Cre+ mice.

A targeting vector was designed to introduce a floxed Hamp allele into mouse ES cells, with exons 2 and 3, which encode the majority of the peptide, flanked by LoxP sites. Further breeding with a C57BL/6 Flp recombinase deleter mouse allowed removal of the Neo cassette. Cardiac Hamp knockouts were then generated by crossing homozygous Hamp animals with mice transgenic for Myh6-Cre recombinase, which is under the control of cardiomyocyte-specific Myosin Alpha Heavy chain six promoter.

DOI: http://dx.doi.org/10.7554/eLife.19804.026

The strategy for generating Cardiac Slc40a1 C326Y knock-in mice is outlined in Appendix 1—figure 5, and further details are provided in the Appendix.

Appendix 1—figure 5.

Strategy for generation of Slc40a1 C326Y;Myh6.Cre+ mice.

A targeting vector was designed to introduce a floxed Slc40a1 C326Y allele into mouse ES cells, containing mutant exon seven and wild type and to delete simultaneously endogenous exons 7 and 8. Further breeding with a C57BL/6 Flp recombinase deleter mouse allowed removal of the Neo cassette. Cardiac Slc40a1 C326Y knock-ins were then generated by crossing homozygous Slc40a1 C326Y animals with mice transgenic for Myh6-Cre recombinase, which is under the control of cardiomyocyte-specific Myosin Alpha Heavy chain six promoter.

DOI: http://dx.doi.org/10.7554/eLife.19804.027

Cine MRI

Mice were anaesthetized with 2% isofluorane in O2 and positioned supine in a purpose-built cradle. ECG electrodes were inserted into the forepaws, a respiration loop was taped across the chest and heart and respiration signals were monitored using a custom-built physiological motion gating device. The cradle was lowered into a vertical-bore, 11.7 T MR system with a 40 mm birdcage coil (Rapid Biomedical, Würzburg, Germany) and visualised using a Bruker console running Paravision 2.1.1. A stack of contiguous 1 mm thick true short-axis ECG and respiration-gated cine-FLASH images were acquired. The entire in vivo imaging protocol was performed in approximately 60 min. Image analysis was performed using ImageJ (NIH Image, Bethesda, MD). Left ventricular volumes and ejection fractions were calculated from the stack of cine images as described (Lakhal-Littleton et al., 2015).

Dietary iron content

Unless otherwise stated, animals were provided with a standard rodent chow diet containing 200 ppm iron. In iron manipulation experiments, mice were given an iron-deficient diet (2–5 ppm iron; Teklad TD.99397; Harlan Laboratories), or an iron-loaded diet (5000 ppm iron; Teklad TD.140464) or a matched control diet (200 ppm iron; Teklad TD.08713) from weaning for six weeks.

Isolation of primary adult mouse cardiomyocytes and in vitro treatment

Adult primary cardiomyocytes were isolated from eight week old C57BL/6 mice. Hearts were cannulated and mounted on a langendorff apparatus, then perfused using a liberase solution for 10 min. After filtration through a 400 µm gauze, cells were cultured in MEM medium containing Hanks salts, L-glutamine and antibiotics. Within 2 hr of cardiomyocyte culture, supernatants were replaced with fresh medium containing 10% Fetal calf serum, with 0.5 mmol/L ferric citrate (FAC) (F3388, Sigma Aldrich) or 100 µmol/L desferroxamine (D9533, Sigma Aldrich) for 8 hr. The Furin inhibitor decanoyl-Arg-Val-Lys-Arg-chloromethylketone (CMK) (N1505, Bachem) was added at a concentration of 50 µmol/L for the duration of DFO and FAC treatment.

Quantitative PCR

Total RNA extraction and cDNA synthesis were carried out as previously described (Lakhal-Littleton et al., 2015). Gene expression was measured using Applied Biosystems Taqman gene expression assay probes for Slc40a1, Hamp, TfR1, Myh7, Nppb, Ldha, Hk2, Eno and house-keeping gene β-Actin (Life Technologies, Carlsbad, CA). The CT value for the gene of interest was first normalised by deducting CT value for β-Actin to obtain a delta CT value. Delta CT values of test samples were further normalised to the average of the delta CT values for control samples to obtain delta delta CT values. Relative gene expression levels were then calculated as 2-delta deltaCT.

Immunohistochemistry

Tissues were prepared as described previously (Lakhal-Littleton et al., 2015) and stained with rabbit polyclonal anti-mouse HAMP antibody (ab30760, Abcam, RRID:AB_2115844) at 1/40 dilution, or rabbit polyclonal anti-mouse FPN antibody (MTP11-A, Alpha Diagnostics, RRID:AB_1619475) at 1/200 dilution. Results of control experiments confirming the specificity of the HAMP antibody are shown in Figure 1—figure supplement 5.

Figure 1—figure supplement 5.

Confirmation of HAMP antibody specificity.

(A) Staining with HAMP antibody (Abcam ab30760) in liver and heart of C57BL/6 mice is completely abrogated by co-incubation with hepcidin-25 blocking peptide (Abcam ab31875). Scale bar = 20 µm. (B) Loss of HAMP staining in Hamp;;Myh6.Cre+ hearts (Figure 1F) is consistent with the antibody detecting HAMP1 and not HAMP2, because Hamp2 mRNA expression is not altered in Hamp mice relative to Hamp controls, either with control diet or iron-deficient diet (six weeks from weaning). *p=0.007 and 0.047 relative to Hamp control under respective diet. n = 3 per group. Values are plotted as mean ± SEM. N.S=not significant.

DOI: http://dx.doi.org/10.7554/eLife.19804.008

HAMP enzyme-linked immunosorbent assay (ELISA)

HAMP was measured in mouse sera and in cardiomyocyte supernatants using a HAMP ELISA kit (E91979Mu, USCN) according to the manufacturer’s instructions. Results of control experiments confirming that in vitro treatments did not affect HAMP peptide detection by this ELISA kit are shown in Figure 1—figure supplement 6.

Figure 1—figure supplement 6.

HAMP detection by ELISA unaffected by FAC and DFO.

HAMP standard from mouse hepcidin ELISA (E91979Mu, Uscn) was diluted either in the kit’s own standard diluent, or in unconditioned growth medium alone, or containing 100 µmol/L DFO or 500 µmol/L FAC. ELISA was performed as per manufacturer’s instructions.

DOI: http://dx.doi.org/10.7554/eLife.19804.009

DAB-enhanced perls stain

Formalin-fixed paraffin-embedded tissue sections were deparaffinised using Xylene, then rehydrated in ethanol. Slides were then stained for 1 hr with 1% potassium ferricyanide in 0.1 mol/L HCl buffer. Endogenous peroxidase activity was quenched, then slides were stained with DAB chromogen substrate and counterstained with haematoxylin. They were visualised using a standard brightfield microscope.

Electron microscopy (EM)

Hearts were dissected and 0.5–1 mm3 slices were fixed by immersion for 2 hr in 2.5% glutaraldehyde in 0.1 mol/L cacodylate buffer and prepared for electron microscopy by standard methods. Briefly, cells were post-fixed in osmium tetroxide (1% w/v in 0.1 mol/L phosphate buffer), stained with uranyl acetate (2% w/v in distilled water), dehydrated through increasing concentrations of ethanol (70–100%) and acetone and embedded in TAAB resin (TAAB, Aldermaston, UK). Ultrathin sections (50–80 nm) were prepared using a Reichert ultracut S microtome and mounted on 200 mesh nickel grids. Sections were lightly counterstained with lead citrate and uranyl acetate and examined with a Jeol transmission electron microscope (JEM-1010, JEOL, Peabody MA).

Fe55 efflux in primary adult cardiomyocytes

Adult cardiomyocytes were isolated from mice of the desired genotype at 9 weeks of age as described above. Cardiomyocytes were then cultured in 24-well plates at equal densities for 16 hr before the efflux experiment was performed as described (McKie et al., 2000). Briefly, after washing with three times PBS, cells were incubated for 30 min in 200 µl uptake solution (98 mmol/L NaCl, 2.0 mmol/L KCl, 0.6 mmol/L CaCl2, 1.0 mmol/L MgCl2, 1.0 mmol/L ascorbic acid, 10 mmol/L HEPES [pH 6.0] with Tris base, 50 μmol/L Fe55 [NEN, Boston, MA]), then washed three times with PBS and incubated for 30 min with efflux solution (98 mmol/L NaCl, 2.0 mmol/L KCl, 0.6 mmol/L CaCl2, 1.0 mmol/L MgCl2, 10 mmol/L HEPES [pH 7.4] with Tris base, 300 U/ml bovine ceruloplasmin (cp) (Sigma) and 40 μg/ml human apotransferrin (tf) [Sigma]),in the absence or presence of 0.5 µmol/L mouse HAMP peptide (Peptides International). The efflux medium was then removed, the cells washed three times in ice-cold PBS and disrupted by incubation in 100 µl of 10% SDS solution for 10 min.The efflux solution and cell lysates were then transferred into scintillation vials for Fe55 counting. Where Furin inhibition was carried out, CMK was added to the culture medium at 50 µmol/L 2 hr before the efflux experiment was carried out.

Iron quantitation

Ferritin concentration in serum and in liver lysates was determined using the ferritin ELISA kit (ICL, Inc. Portland). Serum iron levels were determined using the ABX-Pentra system (Horiba Medical, CA). Determination of total elemental iron in the heart was carried out by inductively coupled plasma mass spectrometry (ICP-MS) as described previously (Lakhal-Littleton et al., 2015). Calibration was achieved using the process of standard additions, where spikes of 0 ng/g, 0,5 ng/g, 1 ng/g, 10 ng/g, 20 ng/g and 100 ng/g iron were added to replicates of a selected sample. An external iron standard (High Purity Standards ICP-MS-68-A solution) was diluted and measured to confirm the validity of the calibration. Rhodium was also spiked onto each blank, standard and sample as an internal standard at a concentration of 1 ng/g. Concentrations from ICP-MS were normalised to starting tissue weight.

Isolation of cardiomyocyte and non-cardiomyocyte fractions for iron quantitation

Following cardiac perfusion, hearts were dissected into small pieces in ice-cold Hanks buffer, and subject to collagenase P digestion at 37C for 1 hr (11213857001, Roche Diagnostics). Following lysis of red blood cells, cell suspensions were passed through a 70 µm sieve, before being labelled using cardiomyocyte isolation kit (130-100-825, Miltenyi Biotec). Separation of cardiomyocyte and non-cardiomyocyte fractions was carried out using MACS magnetic separation system according to the manufacturer’s instructions. Cardiac fractions were lysed immediately for ICP-MS analysis.

Activity assays for aconitase I and electron transport chain (ETC) complexes

Approximately 10 mg of frozen, crushed tissues were suspended in 200 µl of ice-cold KME buffer (100 mmol/L KCl, 50 mmol/L Mops, 0.5 mmol/L EGTA, pH 7.4), then homogenized by rupturing with a TissueRuptor (Qiagen, UK) over ice. In a plastic cuvette, the cardiac lysate is mixed with assay buffer and slotted into a spectrophotometer. Details of assay buffers and of reaction procedure for each enzyme are detailed in the Appendix.

Statistics

Values are shown as mean ± standard error of the mean (SEM). Comparison of iron indices, enzyme activities and parameters of cardiac function between groups was performed using Student’s T test. p values < 0.05 were deemed as indicating significant differences between groups. Where significant, exact p values for a figure panel are stated in the corresponding figure legend. No explicit power analysis was performed prior to the experiments to determine sample size, since we had no means to reliably estimate the size and variability of the effects of deleting hepcidin on parameters of cardiac function. For Cine MRI assessment of cardiac function, typically 5–11 animals of each genotype were used, with a matching number of littermate controls. For gene expression, iron quantitation and enzyme activity assays from mouse tissues, typically, 3–6 independent biological replicates and matching littermate controls were analysed. Since significant results were obtained from these set of experiments, no further animals were sacrificed. All 'n' values reported refer to independent biological replicates.

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "An essential cell-autonomous role for hepcidin in cardiac iron homeostasis" for consideration by eLife. Your article has been favorably evaluated by Naama Barkai (Senior Editor) and three reviewers, one of whom is a member of our Board of Reviewing Editors. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

Summary:

The reviewers found the manuscript to have merit and potentially to make an important contribution to understanding iron homeostasis at the tissue level. For the first time, the authors tested the conjecture that tissue hepcidin production may contribute to local tissue-level iron regulation.

However, a number of important issues were raised by the reviewers including the need to be stricter in the methodology used and its description. In particular, the reviewers were seeking further evidence that the data on hepcidin expression and regulation were robust and that bioactive hepcidin was being detected, with hepcidin KOs as controls. Reviewers were reluctant to accept immunohistochemical staining as quantification of bioactive secreted hepcidin. Furthermore, iron efflux experiments need to be described in more detail. Tissue heterogeneity including the possible presence of macrophages may be an issue to resolve. Any further evidence that can be provided on the regulation of hepcidin would strengthen the paper.

I have summarised the major issues below:

1) Subsection “Hamp expression and regulation in the heart”, first paragraph and Figure 1: The semiquantitation of hepcidin by immunohistochemical methods has serious liabilities because hepcidin is a secreted protein and because antibodies can detect functional and nonfunctional hepcidin peptide, including precursor and degraded forms. The conversion of propeptide to active peptide is a step that could be regulated by iron and this may not be detected by the method used. The Methods section indicates that it was a rabbit polyclonal anti-mouse antibody from Abcam, but it does not say which antibody from Abcam was used. Several listed on Abcam's website were prepared against human peptides but have anti-mouse reactivity; one was prepared against mouse antigen and recognizes both Hamp1 and Hamp2 in mice. Are the authors certain that the antibody is recognizing the hepcidin-25 peptide encoded by Hamp1, and not something else? If they are detecting either pro-hepcidin or Hamp2 the conclusions of the paper might be very different. Also, because hepcidin functions as an extracellular peptide, it is difficult to draw conclusions from intracellular antibody staining. Specifically: a) It is not clear which hepcidin forms are stained and where they are located. Does the antibody detect intracellular peptide which is mostly propeptide? Fluorescent confocal microscopy could help with localization. b) No evidence is provided that cardiac myocytes can generate the functional 25 amino acid hepcidin. The processing of peptide hormones to bioactive forms requires prohormone convertases which are expressed in a tissue-specific manner.

2) Subsection “Hamp expression and regulation in the heart”, first paragraph and Figure 1D: The detection of hepcidin secreted by cardiomyocytes in culture needs to be validated better because the concentrations are very low, and because the post-mRNA effect of DFO and FAC on hepcidin secretion is unprecedented. This experiment is crucial for understanding the function of local control by hepcidin-ferroportin. It is therefore necessary to be as careful as possible here. Controls should include hepcidin KO cardiomyocytes. At these low peptide concentrations, adsorption of peptide to plastic surfaces causes major losses and these could be affected by DFO and FAC treatment. Neither is it clear which forms of hepcidin are released into the media. Finally, primary hepatocytes within less than 24h in culture lose their ability to physiologically regulate hepcidin in response to iron, and this can result in delayed paradoxical suppression of hepcidin mRNA by FAC. If this happens with cardiomyocytes it could lead to artefacts. The regulation of hepcidin mRNA by FAC and DFO should therefore be checked as a function of time in culture.

3) To confirm the equivalence of Hamp KO and ferroportin C326Y the survival curves need to be compared. This is a hard endpoint that has produced a surprise previously (e.g. the marked difference in survival between global hepcidin KO and global ferroportin C326S). If data exists on the survival of Ferroportin C326Y transgenic mice, it should be presented.

4) It would be relatively straightforward for the authors to examine the regulation of cardiac Hamp using hemojuvelin floxed (Hfe2 fl/fl) mice (Chen W et al., Blood 2011) or other approaches, as regulation of hepatic Hamp expression is well understood. Hfe2 is expressed at high levels in the heart and might be relevant there as it is in the liver. The paper would be much stronger if there was some indication of the mechanism of hepcidin regulation in cardiomyocytes.

5) The authors refer to other animal models of cardiac iron deficiency, but they don't really distinguish between those caused by dietary iron deprivation, which would have the confounding existence of iron deficiency anemia, and a heart-specific Tfrc knockout model in which iron deficiency is restricted to the heart (Xu W et al., Cell Reports 2015, ref 31). The heart-specific Tfrc knockout was shown to have many characteristics that are similar to those described in this paper. One important difference, however, was in the levels of electron transport chain complexes found in untreated and iron treated control and cardiac Hamp-/- mice (Figure 4—figure supplement 2). Xu et al. reported decreases in the bands representing iron-containing ETC complexes using the same OXPHOS blotting approach, as would be expected in iron deficient cells. Those representative proteins are labile when the complexes do not form properly. How do the authors explain their unexpected results, particularly considering the decreased complex activity shown in Figure 4?

6) Several experiments were carried out using isolated adult mouse cardiomyocytes. How did the authors insure that the cell populations were not contaminated with other cell types, particularly macrophages, which might confound the results?

7) The iron efflux experiments shown in Figure 3L appear to be novel because there is no reference provided. Is this correct? For that reason it would be very helpful to know more about them, to assess their validity. Apparently, cells were loaded with non-transferrin bound 55Fe by a 30 minute incubation. How much iron was taken up under these conditions? Why was non-transferrin bound iron used instead of Fe-Tf, particularly considering the results described in Xu et al., 2015 indicating that the transferrin cycle is very important for cardiomyocyte iron uptake? What measures were undertaken to ensure that the PBS washes effectively removed non-assimilated radiolabeled iron? How many times were the experiments repeated, and how variable were the results? Why do the authors think the FpnC326Y mutant had higher iron efflux compared to Hamp-/- cardiomyocytes in the absence of Hamp?

8) Figure 1, panel A – this appears to be mRNA expression but this is not stated. If it is, how was it normalized? In panel B, as the authors pointed out, it is quite surprising that there appears to be most anti-Hamp staining under iron deficient conditions. If this is truly physiologically important staining (see point 2 above) then it becomes even more important to understand how Hamp expression is regulated (see point 3 above). Hamp is typically measured extracellularly, rather than intracellularly because it is a secreted protein; a defect in pre-protein processing might help explain the discrepancy between panels B and C in this figure. Panel D is particularly important – how many times was this experiment replicated? How reliable is the assay kit (the methods indicate that it is from "uscnk" but I don't know what that is – please specify), and is it specific for Hamp1 rather than Hamp2?

9) The heart sections in Figure 2B are suboptimal. It is particularly difficult to discern anatomy in the fl/fl section. In Figure 2D, the control cardiomyocytes were much smaller than those reported in Figure 1G of Xu et al., 2015, and the difference in cardiomyocyte surface area between controls and mutants was much larger. Do the authors have any idea what this difference might be due to?

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Thank you for resubmitting your article "An essential cell-autonomous role for hepcidin in cardiac iron homeostasis" for consideration by eLife. Your article has been favorably evaluated by Naama Barkai (Senior Editor) and a Reviewing Editor.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

The reviewers generally did not think that the authors use of tissue immunostaining to quantitate a secreted protein is bullet-proof but accept that the authors have enough additional overlapping evidence to give more confidence that the main conclusions are right. They have put together a strong case for hepcidin and ferroportin interacting in an autocrine/paracrine manner in the heart. The genetic argument from the equivalent pathology and similar survival of cardiomyocyte FpnC326Y (cardiomyocyte hepcidin resistance) and cardiomyocyte hepcidin KO is particularly appealing. The autocrine/paracrine function of hepcidin and ferroportin is an old conjecture but this is the first evidence for it.

However, the following is required for acceptance:

1) In the rebuttal there is a statement that the reviewers did not agree with. The cardiac-specific Tfr1 KO mice are not embryonic lethal – they live into the second postnatal week. The authors should be aware of this and modify the text if needed.

2) One reviewer did not believe that the authors had fully explained the outstanding issues with their OXPHOS blot – they thought it was poor quality with much more animal to animal variability than there should be. If possible, the authors should repeat this blot in order to show more consistency between replicates. Alternatively, the authors should delete this western.

[…]

1) Subsection “Hamp expression and regulation in the heart”, first paragraph and

Hepcidin is synthetized as an 84 amino acid prepropeptide, containing an N-terminal ER targeting signal sequence (aa 1-24), a proregion (pro) with a consensus Furin cleavage site (aa 25-59) and a C-terminal bioactive iron-regulatory mature peptide (aa 60-84). The antibody used for immunohistochemical detection of hepcidin in the heart is ab30760 from Abcam, which was raised against the human mature peptide (aa60-84). We now include this information in the Methods section of the revised manuscript.

The reviewers’ concerns over the specificity of this antibody are ones that we also had, and consequently we did carry out appropriate control experiments. We now include the results of those control experiments in the revised manuscript. First, we showed that staining with this antibody is absent in the heart of cardiac-specific hamp-1 knockouts (Figure 1F), despite Hamp-2 expression being unaltered in these hearts compared to controls (Figure 1—figure supplement 5). These data demonstrate that this antibody detects HAMP1 and not HAMP2. Second, we showed that staining in the heart of wild-type animals is completely abrogated by the use of the mature peptide (hepcidin-25, Abcam ab 31875) as a blocking peptide (Figure 1—figure supplement 5). In hepatocytes, it has been shown that the propeptide can also be secreted, and that this propeptide is detectable by antibodies raised against the mature peptide (Rouault and Klausner, 1997). Thus, in theory our antibody detects not only the mature secreted peptide but also the propeptide, and even demonstration of extracellular staining by immunohistochemistry would not constitute formal evidence for the presence of the active mature peptide. We do, however, provide direct evidence for the secretion of an active hepcidin by cardiomyocytes. First, in our in vitro experiments, where we detect hepcidin by ELISA in supernatants of primary cardiomyocytes from control animals (original data). We reproduced this finding in a second independent experiment (new Figure 1D), and further show that hepcidin is greatly reduced or undetectable in supernatants of cardiomyocytes from cardiac Hamp-1 knockout animals. We now include these data as Figure 1—figure supplement 4. Second, we also provide evidence that the secreted peptide is active, and this is discussed below.

The reviewer is right to point out that the release of the bioactive hepcidin mature peptide requires prohormone convertases, which may or may not be present in the heart. Furin is the prohormone convertase required for release of hepatic bioactive mature hepcidin peptide. In hepatocytes, it has been shown that Furin inhibition leads to decreased release of the active mature peptide and increased release of the larger inactive propetide, and that both of these protein species are detected by assays targeting the mature peptide (Rouault and Klausner, 1997). Furin is also highly expressed in the heart (Beaubien et al., 1995), raising the intriguing possibility that it may be involved in the production of the bioactive hepcidin mature peptide from cardiomyocytes. To test this hypothesis, we performed additional experiments using the Furin inhibitor Decanoyl-Arg-Val-Lys-Arg-chloromethylketone trifluoroacetate salt(CMK). These experiments showed that while Furin inhibition under iron replete conditions did not decrease secreted HAMP levels (consistent with the release of the propeptide as reported in hepatocytes) (Valore and Ganz, 2008), it nonetheless increased iron export from primary cardiomyocytes. Furthermore, the effect of Furin inhibitor on iron export was only seen in Hamp fl/fl and not in Hamp fl/fl Myh6.Cre+ cardiomyocytes. We include these data in Appendix 1—Figure 1. Together, these results provide evidence that cardiomyocytes release an active form of hepcidin in a Furin-dependent manner.

Furthermore, we found that increased HAMP release from DFO-treated cardiomyocytes depended on Furin. These data are added to Figure 1D. We also found that Furin itself was upregulated by iron deficiency in vitro and in vivo. We include these data as Figure 1—figure supplement 2. These findings are consistent with the previously reported regulation of Furin by iron deficiency through Hypoxia-inducible factors HIFs (Silvestri, Pagani and Camaschella, 2008). They suggest that differential regulation of Furin by iron may explain the divergent effects of iron on Hamp transcript and HAMP protein.

2) Subsection “Hamp expression and regulation in the heart”, first paragraph and

As suggested by the reviewer, we have performed additional experiments with primary cardiomyocytes from cardiac hamp1-ko cells, and found that levels of HAMP proteins were greatly reduced or undetectable in their supernatants. We include this in Figure 1—figure supplement 4. Furthermore, the concentrations detected in the supernatants are well within the detection range (4.94-400pg/mL) of the ELISA kit used (Uscn Life Science Inc. #E91979Mu). The reviewer was also concerned that in vitro treatments (DFO and FAC) may affected detection of the hepcidin peptide by this ELISA assay. To guard against the possibility of differential absorption resulting from the presence of DFO or FAC, we compared the standard curves obtained by dilution of the peptide standard in a) standard diluent provided in the kit, b) unconditioned cardiomyocyte growth medium, c) unconditioned cardiomyocyte growth medium with 100uM DFO, and d) unconditioned cardiomyocyte growth medium with 500uM FAC. We found no significant differences between the 4 standard curves. We include these data as Figure 1—figure supplement 6. This demonstrates that the presence of DFO or FAC does not affect the detection of the hepcidin peptide by this ELISA kit.

The reviewer was concerned that physiological control of hepcidin was lost in cardiomyocytes due to the length of culture. First, both in the original and in the additional experiments performed herein, cardiomyocytes were used within 2 hours of harvest. Additionally, as suggested by the reviewer, we carried out a timecourse of treatment with FAC and DFO, and now present these data in the manuscript (new Figure 1C for Hamp transcript and new Figure 1D for HAMP in supernatants). Overall, the direction of change in response to DFO and FAC, both for hepcidin transcript and protein are maintained over the 24 hour period and are consistent with the single-timepoint data presented in the original manuscript prior to revision.

3) To confirm the equivalence of Hamp KO and ferroportin C326Y the survival curves need to be compared. This is a hard endpoint that has produced a surprise previously (e.g. the marked difference in survival between global hepcidin KO and global ferroportin C326S). If data exists on the survival of Ferroportin C326Y transgenic mice, it should be presented.

We now provide those data as new Figure 3C. As for Hamp fl/fl;Myh6.Cre+ mice, Slc40a1 fl/fl;Myh6.Cre+ animals also have greater mortality compared to their littermate controls with mean survival of 32 weeks.

4) It would be relatively straightforward for the authors to examine the regulation of cardiac Hamp using hemojuvelin floxed (Hfe2 fl/fl) mice (Chen W et al., Blood 2011) or other approaches, as regulation of hepatic Hamp expression is well understood. Hfe2 is expressed at high levels in the heart and might be relevant there as it is in the liver. The paper would be much stronger if there was some indication of the mechanism of hepcidin regulation in cardiomyocytes.

HJV is one of many putative regulators of cardiac hepcidin. As it happens, muscle-specific HJV KO mice have increased rather than decreased cardiac hepcidin expression (Kostas Pantopoulos, personal communication). Thus, it appears that both with respect to HJV and to iron, cardiac hepcidin is regulated differently from liver hepcidin. Substantive additional studies are required to explore the role of known regulators of hepatic hepcidin in the regulation of cardiac hepcidin. We consider such studies beyond the scope of this revision as the work would constitute an independent manuscript.

5) The authors refer to other animal models of cardiac iron deficiency, but they don't really distinguish between those caused by dietary iron deprivation, which would have the confounding existence of iron deficiency anemia, and a heart-specific Tfrc knockout model in which iron deficiency is restricted to the heart (Xu W et al., Cell Reports 2015, ref 31). The heart-specific Tfrc knockout was shown to have many characteristics that are similar to those described in this paper. One important difference, however, was in the levels of electron transport chain complexes found in untreated and iron treated control and cardiac Hamp-/- mice (Figure 4—figure supplement 2). Xu et al. reported decreases in the bands representing iron-containing ETC complexes using the same OXPHOS blotting approach, as would be expected in iron deficient cells. Those representative proteins are labile when the complexes do not form properly. How do the authors explain their unexpected results, particularly considering the decreased complex activity shown in

With regards to the description of animal models of cardiac iron deficiency, we have now amended the text to distinguish between the various models, as suggested by the reviewer. With regards to the levels of ETC complexes detected by OXPHOS, we too were initially surprised that we did not detect a reduction as was seen in cardiac TfR knockout mice (Xu et al., 2015).However, this becomes less surprising when you consider the relative severity of cardiac iron deficiency, which is reflected in the fact that cardiac-specific TfR KO are embryonic lethal whereas cardiac-specific hepcidin knockouts have a mean survival of 28 weeks. In this context, it is important to consider the mechanisms through which iron deficiency impinges on the ETC. These mechanisms include not only limiting iron availability for complex assembly and for enzymatic activity of assembled complexes, but also reducing the levels of mitochondrially-encoded proteins. Indeed, the degree of iron deficiency (and iron excess) have been shown to correlate with damage to mitochondrial DNA (Walter et al., 2002), and many of the proteins in ETC complexes are encoded by mitochondrial genes. Thus, depending on the degree of iron deficiency in the cardiomyocytes, different mechanisms may come into play, meaning that complex assembly or activity could still be disrupted when levels of individual proteins in the complex are still maintained.

6) Several experiments were carried out using isolated adult mouse cardiomyocytes. How did the authors insure that the cell populations were not contaminated with other cell types, particularly macrophages, which might confound the results?

The reviewer is right to point out that macrophages are known to be present in the cardiac tissue. We cannot exclude the presence of macrophages in our cardiomyocyte culture. However, specifically with respect to hepcidin production in the supernatants, we can be confident that cardiomyocytes are the source of the hepcidin, because very little hepcidin was detected in supernatants of cardiac-specific hepcidin knockouts (where cardiomyocytes but not macrophages have deletion of the hepcidin gene). We include these data in Figure 1—figure supplement 4.

7) The iron efflux experiments shown in

The method for measuring iron efflux used in our experiments is not novel, but adapted from the method used by Mckie et al. 2000, which was the first publication to describe the function of FPN in iron efflux. We now include this reference in the text. This method does not enable measurement of absolute amounts of iron flux, but rather measurement of iron efflux as a percentage of the total iron taken up during incubation with the uptake solution; percentage efflux= 100 x Fe55 in efflux solution /(Fe55 in efflux+ Fe55 remaining in cells). The readouts are obtained using a scinitillation counter and are expressed as counts per minute. Count rate does not universally equate to dose rate, and there is no simple universal conversion factor. Any conversions are instrument-specific.

The reviewer also questions the use of non-transferrin bound Fe55 (as in Mckie et al., 2000) instead of transferrin-bound Fe55 for loading cells with iron. The rationale for this is to circumvent any confounding effects of differential iron uptake between cells of different genotypes due to differing levels of TfR1 expression. Indeed, we know that TfR1 expression is higher in cardiac-hepcidin knockout cells and in Slc40a1 C326Y knock-in cells than in their respective controls (Figure 3O). While TfR1 is the dominant route of iron uptake into cardiomyocytes, it is known that NTBI also enters cardiomyocytes via L and T-type calcium channels.

To minimise carryover of Fe55, cells were washed with PBS 3 times before addition of efflux medium. Efflux measurements were carried out in three independent experiments, with each experiment containing three biological replicates (3 wells of cardiomyocytes) for each genotype and condition. The range between values obtained from three biological replicates constituted between 0.07and 0.21 of the value of mean. Thus, the results were sufficiently accurate to make meaningful comparisons between genotypes and treatment conditions within each experiment, and the outcome of those comparisons was reproducible between experiments.

The reviewer raises an interesting question relating to the levels of iron export being higher in Slc40a1 C326Y knock-in cardiomyocytes than in Hamp knockout cardiomyocytes (p=0.014). We too were initially surprised at this result. One possible explanation is that FPN levels in Slc40a1 C326Y knock-in cardiomyocytes are higher than in Hamp KOs cardiomyocytes, due to the fact that FPN in the former setting is completely insensitive to any hepcidin, while FPN in the latter may still be subject to some regulation by hepcidin from non-cardiomyocyte sources.

8)

In Figure 1, panel A depicts mRNA expression of hepcidin. We have amended the corresponding figure legend to make this clear. Expression was measured using Taqman multiplex gene expression assays, where hepcidin and the house keeping gene (b-actin) were measured in the same sample. A δ CT (CT for hepcidin-CT for b-actin) value was generated for each sample. Subsequently δ CT values for each sample were normalised to the average δ CT of the control samples (hearts under control diet), to obtain a δ CT value. Relative expression was then calculated as 2-δ δ CT. The reviewer’s questions relating to the discrepancy between hepcidin transcript and protein, the possible role of protein processing and the measurement of hepcidin in supernatants are addressed in responses to points 1 and 2 above.

9) The heart sections in

We provide better heart sections in Figure 2bB In terms of cardiomyocyte size, differences in reported cardiomyocyte size between our data (range between 25 and 37 um2) and those of reported by Xu et al4 (range between 45 and 60 um2) in control animals may be due to any number of reasons including: a) the manual nature of quantitation of cardiomyocyte size, which requires drawing around individual WGA-stained cardiomyocytes, which is likely to vary between individual users, b) the method for tissue fixing and processing of the cardiac tissue, c)the area of the heart selected for histology and d) orientation of section plane relative to cardiomyocyte axis.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

The reviewers generally did not think that the authors use of tissue immunostaining to quantitate a secreted protein is bullet-proof but accept that the authors have enough additional overlapping evidence to give more confidence that the main conclusions are right. They have put together a strong case for hepcidin and ferroportin interacting in an autocrine/paracrine manner in the heart. The genetic argument from the equivalent pathology and similar survival of cardiomyocyte FpnC326Y (cardiomyocyte hepcidin resistance) and cardiomyocyte hepcidin KO is particularly appealing. The autocrine/paracrine function of hepcidin and ferroportin is an old conjecture but this is the first evidence for it.

However, the following is required for acceptance:

1) In the rebuttal there is a statement that the reviewers did not agree with. The cardiac-specific Tfr1 KO mice are not embryonic lethal – they live into the second postnatal week. The authors should be aware of this and modify the text if needed.

We thank the reviewers for drawing our attention to the erroneous statement that “cardiacTfR1 knockout are embryonic lethal”. In the original response letter, the phenotype of these mice was discussed in relation to our results showing the levels of ETC complexes as detected by the OXPHOS assay (Figure 4—figure supplement 1). As these data have now been removed from the manuscript, so has the erroneous statement in the revised response letter.

2) One reviewer did not believe that the authors had fully explained the outstanding issues with their OXPHOS blot – they thought it was poor quality with much more animal to animal variability than there should be. If possible, the authors should repeat this blot in order to show more consistency between replicates. Alternatively, the authors should delete this western.

In relation to the OXPHOS blot showing the levels of ETC complexes (Figure 4—figure supplement 1), we have now run the blot a second time. However, the variability between some of the samples persists. As such, we have opted to remove the blot from the manuscript, and have also removed all text relating to it in the Results, Discussion, Methods and figure legends section. Accordingly, we have also removed any text in our original response letter relating to this particular dataset.