Hamp1 mRNA and plasma hepcidin levels are influenced by sex and strain but do not predict tissue iron levels in inbred mice

Both sex and strain have a significant effect on both hepcidin mRNA (primarily a sex effect) and plasma hepcidin levels (primarily a strain effect). Liver iron and diferric transferrin levels are not predictors of Hamp1 mRNA levels in mice, nor are the Hamp1 mRNA and plasma hepcidin levels good predictors of tissue iron levels, at least in males.

Abstract

Iron homeostasis is tightly regulated, and the peptide hormone hepcidin is considered to be a principal regulator of iron metabolism. Previous studies in a limited number of mouse strains found equivocal sex- and strain-dependent differences in mRNA and serum levels of hepcidin and reported conflicting data on the relationship between hepcidin (Hamp1) mRNA levels and iron status. Our aim was to clarify the relationships between strain, sex, and hepcidin expression by examining multiple tissues and the effects of different dietary conditions in multiple inbred strains. Two studies were done: first, Hamp1 mRNA, liver iron, and plasma diferric transferrin levels were measured in 14 inbred strains on a control diet; and second, Hamp1 mRNA and plasma hepcidin levels in both sexes and iron levels in the heart, kidneys, liver, pancreas, and spleen in males were measured in nine inbred/recombinant inbred strains raised on an iron-sufficient or high-iron diet. Both sex and strain have a significant effect on both hepcidin mRNA (primarily a sex effect) and plasma hepcidin levels (primarily a strain effect). However, liver iron and diferric transferrin levels are not predictors of Hamp1 mRNA levels in mice fed iron-sufficient or high-iron diets, nor are the Hamp1 mRNA and plasma hepcidin levels good predictors of tissue iron levels, at least in males. We also measured plasma erythroferrone, performed RNA-sequencing analysis of liver samples from six inbred strains fed the iron-sufficient, low-iron, or high-iron diets, and explored differences in gene expression between the strains with the highest and lowest hepcidin levels.

NEW & NOTEWORTHY Both sex and strain have a significant effect on both hepcidin mRNA (primarily a sex effect) and plasma hepcidin levels (primarily a strain effect). Liver iron and diferric transferrin levels are not predictors of Hamp1 mRNA levels in mice, nor are the Hamp1 mRNA and plasma hepcidin levels good predictors of tissue iron levels, at least in males.

iron is an essential element required for many biological processes, including the transport of oxygen. Body iron homeostasis is tightly regulated. Poor bioavailability from the diet and lack of a regulated excretion mechanism for iron required the development of highly efficient mechanisms for iron acquisition and conservation, especially because iron can catalyze the formation of highly reactive free radicals that are damaging to cellular components (3). The iron-regulatory hormone hepcidin, mainly produced in hepatocytes, negatively regulates iron absorption in the duodenum and the release of iron from macrophages and hepatocytes (11, 43). It does so by binding to the iron exporter ferroportin and inducing its internalization and degradation. More hepcidin is produced when iron is abundant, suppressing further iron absorption and release, and less or no hepcidin is produced in iron deficiency, allowing more iron to enter the plasma (29).

Previous studies in mice found conflicting results on the relationship between strain, sex, and hepcidin levels. Unlike humans, who have only one hepcidin gene, mice have two: hepcidin-1 (Hamp1) and hepcidin-2 (Hamp2), which share 68% of similarity at the protein level but serve distinct functions (15, 30, 34, 39). Only Hamp1 has a proven role in iron metabolism (20, 22). Two studies examined the role of strain and sex in the expression of hepcidin genes and showed that female mice have higher hepcidin levels than male mice of the same strain and that the differences between the strains are mainly due to the differences in Hamp2 expression (8, 18). They also showed that females had higher levels of liver, spleen, and serum iron, but no differences in transferrin saturation, than males of the same strain, and there was a significant difference between the strains in liver and spleen iron levels in males (8, 18). However, these two studies were focused on two inbred strains only: C57BL/6(N) and DBA/2(N). A study looking at three strains, C57BL/6, BALB/c, and DBA/2, confirmed the impact of genetic background on systemic iron metabolism parameters (5). This study also found no statistically significant relationship between Hamp1 mRNA levels and plasma iron, transferrin saturation, and tissue iron concentrations (5). A study of serum and urine hepcidin concluded that serum hepcidin (Hep-1) significantly correlates with liver Hamp1 mRNA levels, as well as liver and spleen iron levels in different mouse models of hereditary hemochromatosis (39). Lu et al. reported altered iron homeostasis in all examined tissues (liver, duodenum, spleen, kidney, heart, brain, and serum) in both liver-specific and global Hamp1 knockout mice and Hamp1/2 double-knockout mice, with the changes in tissue and serum iron content of double-knockout mice not statistically different from those in Hamp1 single-knockout mice (22). They have also reported age dependency in kidney, heart, and brain in all knockout mice, with iron levels higher in older mice and higher levels of spleen iron in female Hamp1 and Hamp1/2 floxed mice compared with males (22). Together, these studies suggest a potentially complex relationship between sex, strain, and iron status that requires further clarification.

Our aim was to expand the current knowledge of hepcidin and tissue iron regulation by examining multiple tissues and the effects of different dietary conditions on a greater number of inbred strains. In the first study, we measured Hamp1 expression in both male and female mice from 14 inbred strains on a defined iron-sufficient diet (35-ppm iron) and examined the correlation between Hamp1 mRNA levels and liver iron and plasma diferric transferrin levels. We further examined this issue in a more detailed study, focused on six inbred and three recombinant inbred strains raised on either a low-iron (12-ppm iron), a control (50-ppm iron), or a high-iron (20,000-ppm iron) diet. Hamp1 mRNA and plasma hepcidin levels in both sexes, as well as iron levels in the heart, kidneys, liver, pancreas, and spleen in male mice, were measured. Plasma erythroferrone (ERFE) was measured in a subset of samples, and we also performed RNA-sequencing (RNA-Seq) analysis of liver samples in an attempt to shed light on possible mechanisms of hepcidin expression regulation.

EXPERIMENTAL PROCEDURES

Study 1

Mouse husbandry and the measurements of liver iron (µg/g) and plasma diferric transferrin (%) levels have been described in detail previously (24). Briefly, animals (6 per sex unless otherwise noted) were obtained at 4 wk of age from the Jackson Laboratory (Bar Harbor, ME) from 18 inbred strains: 129S1/SvImJ, A/HeJ, A/J, AKR/J, B10.D2-Hc0H2dH2-T18c/oSnJ, BALB/cByJ, BALB/cJ, C3H/HeJ (n_female = 5), C57BL/6J (n_female = 10), CAST/EiJ (n_male = 7, n_female = 11), DBA/2J (n_male/female = 12), LG/J, LP/J, MRL/MpJ, NZB/BlNJ (n_male = 10), NZW/LacJ (only female animals), SM/J (n_female = 4), and SPRET/EiJ (n_male = 4). Mice were fed ad libitum a purified diet containing ~35-ppm iron (AIN-93G, no. 510017; Dyets) until 8 wk of age. At the time of death, mice were euthanized by carbon dioxide gas without fasting. Blood was collected by cardiac puncture, and at least 30 min later it was centrifuged at 1,000 g for 15 min at room temperature. Serum was aliquoted into microcentrifuge tubes, snap frozen in liquid nitrogen, and stored at −70°C. Liver tissue was collected, snap frozen in liquid nitrogen, and stored at −70°C. All experimental procedures were approved by the Office of Laboratory Animal Care at the University of California, Berkeley. Three male and female mice from each strain were used for quantitative RT-PCR, except for DBA/2J, where six mice of each sex were used. Two micrograms of total RNA isolated from liver using TRIzol reagent (Invitrogen) were DNaseI treated and reverse transcribed for 1 h with priming by oligo(dT) and random primers using SuperscriptIII reverse transcriptase (Gibco-BRL). The resulting complementary DNA (cDNA) was diluted and subjected to quantitative PCR (qPCR). The qPCR was performed with the ABI Prism 7700 and SYBR Green PCR Master Mix (Applied Biosystems). Relative cDNA abundance was calculated running the standard dilution curve for each assay, based on automatically calculated cycle threshold values by ABI. The geometric mean of GAPDH, β-actin, histone H2A member Z, and 18S ribosomal RNA was used to normalize the data. The following primer sets were used: hepcidin (Hamp1), forward 5′-GGCAGACATTGCGATACCAA-3′ and reverse 5′-TGGCTCTAGGCTATGTTTTGCA-3′; GAPDH, forward 5′-TCTCCCTCACAATTTCCATCCCAG-3′ and reverse 5′-GGGTGCAGCGAACTTTATTGATGG-3′; β-actin, forward 5′-GGCTGTATTCCCCTCCATCG-3′ and reverse 5′-CCAGTTGGTAACAATGCCATGT-3′; histone H2A member Z, forward 5′-AGTTGGCAGGAAATGCGTCAA-3′ and reverse 5′-CGATCAGCGATTTGTGGATGT-3′; 18S ribosomal RNA, forward 5′-GTAACCCGTTGAACCCCATT-3′ and reverse 5′-CCATCCAATCGGTAGTAGCG-3′. Finally, only 14 strains were used for further statistical analyses: NZB/BlNJ, NZW/LacJ, SM/J, and SPRET/EiJ were removed as information from only one sex was available.

Study 2

Mouse husbandry.

Animals (6 per sex unless otherwise noted) from six inbred strains and three recombinant inbred strains [A/J, AKR/J (n_female/high = 5), BALB/cJ, C3H/HeJ, C57BL/6J (n_female/high = 5), DBA/2J, AXB6/PgnJ (n_control = 0, n_male/high = 5, n_female/high = 3), BXA-13/PgnJ (n_{both sex/control} = 2, n_male/high = 2, n_female/high = 3), and CXB9/HiAJ (n_{both sex/control} = 5, n_{both sex/high} = 4)] were obtained from the Jackson Laboratory and maintained on a 12-h light-dark cycle. Four-week-old mice were fed either a low-iron diet (12-ppm iron, no. 115111; Dyets), a high-iron diet (20,000-ppm iron as carbonyl iron, no. 115122; Dyets), or a sufficient-iron diet, i.e., control (50-ppm iron, no. 515005; Dyets). Please note that for low-iron diet, we had six male mice for the six above-mentioned inbred strains only. After 6 wk on the diets, mice were killed after a 4-h fast. Blood was taken from the retroorbital plexus under isoflurane anesthesia using a heparin-coated capillary tube, with one aliquot used immediately for blood cell profiling and the remaining blood placed in a lithium heparin collection tube (cat. no. 365965; BD Biosciences) for plasma collection. Complete blood cell profiling was performed using the Heska (Loveland, CO) HemaTrue Veterinary Hematology Analyzer. Mice were then perfused via the heart with PBS to flush remaining blood from the tissues. Tissues were collected and frozen in liquid nitrogen and stored at −80°C until analysis. All animal procedures were approved by the Institutional Care and Use Committee at University of California, Los Angeles.

Inductively coupled plasma measurements.

Tissue samples of male mice were transferred to polypropylene tubes and dried overnight at 60°C. The desiccated tissue samples were then dissolved in OmniTrace 70% HNO₃ overnight at 60°C with 250-rpm orbital shaking, diluted to 5% HNO₃ with OmniTrace water, and analyzed by inductively coupled plasma-atomic emission spectroscopy (Vista Pro; Varian). Elemental content was normalized to the dry weight of the tissue. Quality control and analytical procedures were performed as previously described (33). Further dilutions of the samples were used where necessary to obtain results for highly abundant elements.

Mouse hepcidin ELISA.

Mouse hepcidin-1 monoclonal antibodies, Ab2B10 (capture) and Ab2H4-horseradish peroxidase (Ab2H4-HRP; detection), as well as synthetic mouse hepcidin-25, were a generous gift from Amgen (Thousand Oaks, CA). High-binding 96-well enzyme immunoassay plates (no. 3590; Corning Costar, Tewksbury MA) were coated overnight at room temperature with 50 μl/well of 3.6 μg/ml Ab2B10 in 0.2 M carbonate-bicarbonate buffer, pH 9.4 (Pierce-Thermo Scientific, Rockford, IL). Plates were washed with wash buffer (PBS and 0.5% Tween 20) and then incubated 45 min in blocking buffer (PBS, 1% BSA, 1% normal goat serum, and 0.5% Tween 20). Plasma samples (diluted between 250 and 10,000 times in blocking buffer) and standards were then placed in the wells in duplicate. A standard curve was generated by diluting the stock mouse hepcidin peptide (50 ng/µl) to a final concentration of 200 pg/ml for the highest standard followed by twofold dilutions in blocking buffer, thereby generating a seven-point standard curve. After a 1-h incubation period at room temperature (with mixing), the wells were washed and then incubated for an hour with 50 μl/well of 130 ng/ml Ab2H4-HRP, then developed with 100 μl/well Ultra-TMB substrate (Thermo Scientific) for 15–30 min in the dark at room temperature. The reaction was stopped by adding 50 μl of 2 M sulfuric acid, and the absorbance was measured at 450 nm using a 96-well plate reader (Molecular Devices, Sunnyvale, CA).

Liver RNA purification.

Total RNA was extracted from mouse liver using the TRIzol method (Invitrogen, Grand Island, NY) according to the manufacturer’s instructions. In brief, 1 ml TRIzol was added to a small piece of tissue (~50 mg) and homogenized using a Bio-Gen Pro200 homogenizer (Pro Scientific, Oxford, CT). Chloroform was added and the extract was vigorously shaken and then centrifuged at 13,000 g to phase separate the organic and aqueous phases. RNA in the aqueous phase was precipitated with isopropanol and then dissolved in nuclease-free water. Using the Bio-Rad iScript cDNA kit (Hercules, CA), cDNA was made from total RNA according to the manufacturer’s instructions and using ~250 ng of total RNA with the following conditions: 25°C for 5 min, 42°C for 30 min, and 85°C for 5 min.

qRT-PCR.

SsoAdvanced Universal SYBR Green Supermix (Bio-Rad) was used for qRT-PCR reactions with a 2-step program set at 95°C for 10 s and 60°C for 30 s for 35 cycles. β-Actin was used to normalize the data. The following primer sets were used: hepcidin (Hamp1), forward 5′-AAGCAGGGCAGACATTGCGAT-3′ and reverse 5′-CAGGATGTGGCTCTAGGCTAT-3′; actin, forward 5′-ACCCACACTGTGCCCATCTA-3′ and reverse 5′-CACGCTCGGTCAGGATCTTC-3′.

Erythroferrone ELISA.

Mouse ERFE monoclonal antibodies and recombinant mouse ERFE standard were produced by Silarus Therapeutics. High-binding 96-well enzyme immunoassay plates (Corning) were coated overnight at 4°C with 100 μl/well of 1.0 μg/ml capture antibody in 50 mM sodium carbonate buffer (pH 9.6). Plates were washed (Tris-buffered saline and 0.5% Tween 20) and blocked for 45 min with 200 μl/well SuperBlock T20 blocking buffer (Pierce). Samples and standards diluted in SuperBlock T20 were incubated for 2 h at room temperature. Plates were washed and incubated for 1 h with 100 μl/well of 1.0 μg/ml biotinylated detection antibody and then washed and incubated for 45 min with streptavidin-HRP conjugate (Invitrogen). Plates were developed with 100 μl/well Supersensitive 3,3′,5,5′-tetramethylbenzidine substrate (Sigma) in the dark at room temperature, the reaction was stopped by adding 100 μl of 0.5 M sulfuric acid, and the absorbance was measured at 450 nm. Erfe^−/− and Th3^/+ mouse sera were used as negative and positive control, respectively, and recombinant mouse ERFE was used to determine the limit of detection (100 pg/ml).

Library preparation and next-generation sequencing.

The integrity of liver total RNA was analyzed by using a Nanodrop (Schwerte, Germany) and an Agilent RNA 6000 Nano kit (Böblingen, Germany) according to the manufacturers’ instructions, and all RNA had an RNA integrity number of 8.9 or higher. Libraries were prepared by using the NEBNext Poly(A) mRNA Magnetic Isolation Module, the NEBNext Ultra RNA Library Prep Kit for Illumina, and the NEBNext Multiplex Oligos for Illumina (Ipswich, MA) according to the manufacturer’s instructions. In brief, 500 ng of liver total RNA were mixed with prewashed NEBNext Oligo d(T)25 beads, placed on the magnetic rack, and eluted with 0.1× Invitrogen Tris-EDTA buffer (Carlsbad, CA) to isolate poly-A mRNA. Purified mRNA was used to perform first-strand cDNA synthesis and second-strand cDNA synthesis reactions. Double-stranded cDNA was purified with Agencourt AMPure XP Beads (Brea, CA) followed by adaptor ligation. The cDNA libraries were amplified using the appropriate index primers in the following PCR cycling conditions: 98°C for 30 s, 98°C for 10 s, and 65°C for 75 s for 12 cycles and 65°C for 5 min. Library quality was assessed using the Agilent High Sensitivity DNA kit according to the manufacturer’s instructions. The libraries were then pooled and sequenced on an Illumina HiSeq 4000 system (San Diego, CA).

Statistical Methods

Summary descriptive statistics and box plots were examined for outliers by strain, sex, and diet for both plasma hepcidin and hepcidin mRNA levels. Shapiro-Wilk’s test and Leven’s test were used to test for normality and homogeneity of variance, respectively. We examined the effect of strain and sex on each trait separately for a given diet. The heteroscedasticity of the samples in both studies could not be corrected by data transformation alone. Therefore we applied the aligned rank transform to a factorial model to conduct nonparametric analyses of variance, as described by Wobbrock et al. (42) and implemented in R package ARTool version 0.9.5. Significant interaction effects, measured using nonparametric two-way ANOVA, were followed up by a test of simple effects. Bonferroni correction was used to correct for multiple comparisons. Hierarchical multiple linear regression was employed to examine whether hepcidin mRNA levels were predictive of plasma hepcidin levels in samples on control or high-iron diets. Regression models were built first using sex and strain as predictors, then adding hepcidin mRNA levels. Hierarchical multiple linear regression was also used to determine whether either hepcidin mRNA levels, plasma hepcidin levels, or both were predictive of any of the four tissue iron levels in male mice only. In this instance, models were first built using strain as the predictor, then adding hepcidin mRNA levels in the second model and plasma hepcidin in the third model. Vice versa, we looked at whether liver iron levels predict hepcidin mRNA and plasma hepcidin levels in both sexes using strain as the first predictor followed by the liver iron levels. At each step, predictors were entered into a model simultaneously. Correlations (Pearson’s r) between predictors were used to test for multicollinearity, and Durbin-Watson statistics were used to check whether the residuals in the model were independent. Apart from nonparametric analysis of variance, which was performed as described above, all other analyses were performed using IBM SPSS Statistics version 22.

To analyze sequencing data, we uploaded raw data (compressed FASTQ files) into Galaxy (1), and read quality was checked using the FASTQC tool. For mapping against mouse genome (mm10) and for quantifying counts, we applied the HISAT2/htseq-count pipeline (2, 17). The appropriate gene transfer format (gtf) file was acquired through the University of California, Santa Cruz, browser. The overall alignment rate was 93.0% on average (with 77.0% on average aligned exactly 1 time). The htseq-count was applied with default settings considering a gene as a union of its exons. A table with raw counts was used as input for the DESeq2 and edgeR packages (21, 23, 35) installed with Bioconductor in the R environment (version 3.4.0) with default settings to quantify candidate genes. Experimental design was based on one for paired samples to compensate for batch (here: lane) effect, if any. We applied functions goana() and kegga() from the package limma for gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis to interpret the differential expression results in a biological context (6).

RESULTS

Study 1: Hamp1 mRNA Levels in 14 Strains on Iron-Sufficient Diet

We analyzed liver Hamp1 expression in 14 different mouse strains in both sexes at 8 wk of age to assess whether there were significant differences in hepcidin mRNA levels between strains and sex. One extreme value for hepcidin mRNA levels was removed as an outlier for a female BALB/cJ with a value 20 times larger than the second-highest value in the sample. Analyses were performed on the 14 strains where information on both sexes was available. Both sex and strain effects were highly significant (P < 0.0001), and the sex-by-strain interaction effect was suggestive (Table 1). Simple effects analysis showed that sex influenced hepcidin mRNA levels within five strains (AKR/J, BALB/cByJ, DBA/2J, LP/J, and MRL/MpJ; Table 2, part A). The reverse simple effects analysis of strains showed highly significant results within both sexes (Table 2, part B). Mean values of Hamp1 mRNA levels in females were equal to or higher than those in males, except for 129S1/SvImJ (Fig. 1).

Table 1.

Analysis of variance results of 14 inbred mouse strains on the control diet in study 1

	Factorial Hepcidin mRNA Levels
Diet	Sex	Strain	Sex × Strain
Control	3.34 × 10−6*	4.6 × 10−7*	0.06

Values are P values.

^*P ≤ 0.0001.

Table 2.

Simple effects analysis of sex within strains (A) and strain within sex (B) for hepcidin mRNA levels on the control diet in study 1

	Control
Strain	F(1,61)	P
A: sex within strains
129S1/SvImJ	0.44	0.51
A/HeJ	0.23	0.63
A/J	1.75	0.19
AKR/J	5.03	0.03*
B10.D2-Hc0H2dH2-T18c/oSnJ	1.96	0.17
BALB/cByJ	3.97	0.05*
BALB/cJ	9.09 × 10−5	0.99
C3H/HeJ	0.002	0.96
C57BL/6J	0.05	0.82
CAST/EiJ	0.07	0.80
DBA/2J	8.61	0.005*
LG/J	1.57	0.22
LP/J	9.27	0.003*
MRL/MpJ	16.19	1.61 × 10−4*

	Control
Sex	F(13,61)	P
B: strain within sex
Male	4.62	1.87 × 10−5*
Female	4.96	7.50 × 10−6*

^*Bonferroni-adjusted P values above the significance threshold (P = 0.05).

Fig. 1.

Interaction of sex and strain on control diet in study 1, showing mean Hamp1 mRNA levels for a given strain and sex.

Multiple linear regression was used to examine whether liver iron levels and plasma diferric transferrin levels were predictive of hepcidin mRNA levels (Fig. 2). Hepcidin mRNA levels were not correlated with either liver iron or plasma diferric transferrin (0.096 and 0.088, respectively). Therefore neither liver iron nor diferric transferrin levels could predict Hamp1 mRNA levels (Table 3). Strain was negatively correlated with hepcidin mRNA levels and liver iron (−0.182 and −0.141, respectively) and not correlated with plasma diferric transferrin (−0.031). Sex was moderately positively correlated with hepcidin mRNA levels (0.345) but not correlated with liver iron (0.093). Sex was again moderately correlated with hepcidin mRNA levels (0.370) but poorly correlated with plasma diferric transferrin (0.178).

Fig. 2.

Hamp1 mRNA levels and liver iron (µg/g; A) and plasma diferric transferrin (Fe₂TF; %; B) levels in 14 inbred strains of mice on control diet in study 1. Mean values (±SD for Hamp1 mRNA levels) were calculated for a given strain and sex. Strains are ordered my increasing values of mean Hamp1 levels in females.

Table 3.

Results of multiple linear regression models predicting the effect of sex, strain, and liver iron or diferric transferrin levels on Hamp1 mRNA levels in 14 inbred mouse strains on the control diet in study 1

	Control Hamp1 mRNA Level
	B	SE	β	P
Analysis including liver iron levels
Model 1
Sex	0.07	0.02	0.35	0.001*
Strain	−0.005	0.003	−0.18	0.07
Model 2
Sex	0.07	0.02	0.34	0.001*
Strain	−0.004	0.003	−0.18	0.08
Liver iron levels	2.77 × 10−5	7.20 × 10−5	0.04	0.70
Analysis including diferric transferrin levels
Model 1
Sex	0.07	0.02	0.37	0.0003*
Strain	−0.005	0.002	−0.18	0.07
Model 2
Sex	0.07	0.02	0.37	0.0005*
Strain	−0.005	0.002	−0.18	0.07
Diferric transferrin	0.0001	0.001	0.02	0.86

Liver iron levels measured in micrograms per gram and diferric transferrin levels measured in percent were used.

^*P ≤ 0.001.

Study 2 (Part 1): Strain- and Sex-Specific Response to Changes in Iron Levels

Prior to analysis, three values in total were removed as outliers from three female mice on the control diet (for having extreme values, several times larger than the second-highest value in the sample): plasma hepcidin values for an A/J and a C57BL/6J mouse, and the Hamp1 mRNA value for a CXB9/HiAJ mouse. In addition, samples from the low-iron diet group were not included in these analyses since their hepcidin values were at or below the detection limit (0.25 ng/ml) and available for one sex only.

The effect of sex on hepcidin mRNA levels and sex-by-strain interaction effect were both highly significant on both diets (P < 0.0001; Fig. 3). The effect of strain on hepcidin mRNA levels was significant to a lesser extent (P = 0.0007 and P = 0.003 on control and high-iron diet, respectively; Table 4). Conversely, strain had a highly significant effect on plasma hepcidin levels (P < 0.0001), with sex having no effect for mice on the control diet (P = 0.49) but having a significant effect for those on the high-iron diet (P = 0.001); both interactions were significant to P = 0.01. The sex-by-strain interaction was analyzed using a simple effects analysis. Sex influenced hepcidin mRNA levels within all six inbred strains but not in the recombinant inbred strains on the control diet and within all six inbred strains and AXB6/PgnJ on the high-iron diet (Table 5, part A). However, sex only influenced plasma hepcidin levels within A/J and C57BL/6J on the control diet and BALB/cJ, C3H/HeJ, and C57BL/6J on the high-iron diet (Table 5, part A). The reverse simple effects analysis of strains showed significant results within both sexes on both diets and for both traits (Table 5, part B).

Fig. 3.

Interaction of sex and strain on control and high-iron (high Fe) diet in study 2, showing mean Hamp1 mRNA (A) and mean plasma hepcidin levels (ng/ml; B) for a given strain, sex, and diet.

Table 4.

Analysis of variance results of eight strains of mice on control diet and nine strains on high-iron diet for Hamp1 mRNA and plasma hepcidin levels in study 2

	Factorial Hepcidin mRNA Levels			Factorial Plasma Hepcidin Levels
Diet	Sex	Strain	Sex × Strain	Sex	Strain	Sex × Strain
Control	1.4 × 10−15*	0.0007†	4.0 × 10−5*	0.49	1.2 × 10−15*	0.005‡
High Fe	7.2 × 10−16*	0.003‡	2.92 × 10−7*	0.001†	9.9 × 10−11*	0.01‡

Values are P values. Plasma hepcidin levels measured in nanograms per milliliter were used. Control, eight strains of mice (6 inbred and 2 recombinant inbred) on control diet. High Fe, nine strains on high-iron diet (6 inbred and 3 recombinant inbred).

^*P ≤ 0.0001,

^†P ≤ 0.001,

^‡P ≤ 0.01.

Table 5.

Simple effects analysis of sex within strains (A) and strain within sex (B) for hepcidin mRNA levels and plasma hepcidin levels on both the control and high-iron diets in study 2

	Hepcidin mRNA Levels				Plasma Hepcidin Levels
	Control		High Fe		Control		High Fe
Strain	F(1,69)	P	F(1,72)	P	F(1,67)	P	F(1,72)	P
A: sex within strains
A/J	13.64	0.00044*	4.38	0.04*	5.24	0.03*	0.26	0.61
AKR/J	13.88	0.00039*	23.42	7.00 × 10−6*	1.55	0.22	0.44	0.51
BALB/cJ	34.04	1.59 × 10−7*	5.40	0.02*	0.37	0.55	3.65	0.06*
C3H/HeJ	17.18	9.50 × 10−5*	10.99	0.00144*	0.04	0.87	9.34	0.003*
C57BL/6J	5.75	0.02*	7.35	0.00838*	22.91	1.00 × 10−5*	3.32	0.07*
DBA/2J	42.15	1.09 × 10−8*	23.36	7.00 × 10−6*	3.27	0.08	0.01	0.95
AXB6/PgnJ	—	—	4.00	0.05*	—	—	0.05	0.83
BXA-13/PgnJ	1.27	0.26	0.90	0.35	0.35	0.56	0.03	0.86
CXB9/HiAJ	0.28	0.60	0.22	0.64	0.08	0.78	0.02	0.89

	Hepcidin mRNA Levels				Plasma Hepcidin Levels
	Control		High Fe		Control		High Fe
Sex	F(7,69)	P	F(8,72)	P	F(7,67)	P	F(8,72)	P
B: strain within sex
Male	3.37	0.004*	4.57	0.00016*	16.54	1.85 × 10−12*	6.12	5.00 × 10−6*
Female	7.25	2.00 × 10−6*	3.83	0.001*	6.07	1.70 × 10−5*	8.99	1.85 × 10−8*

Plasma hepcidin levels measured in nanograms per milliliter were used. High Fe, high-iron diet.

^*Bonferroni-adjusted P values above the significance threshold (P = 0.05).

Multiple linear regression was used to examine whether hepcidin mRNA levels predict plasma hepcidin levels (Fig. 4). Strain was negatively correlated with both hepcidin mRNA levels and plasma hepcidin levels; correlation was moderate on both the control and high-iron diets (−0.315 and −0.445, and −0.247 and −0.490, respectively). Sex was highly positively correlated with hepcidin mRNA levels (0.626) but not correlated with plasma hepcidin (0.034) on the control diet. A similar pattern was observed for the high-iron diet (0.482 and 0.095, respectively). Since no variable was correlated with another at r > 0.9, and after inspecting other collinearity statistics, we deduced that multicollinearity is unlikely to be a problem. Hepcidin mRNA levels and plasma hepcidin levels were moderately correlated on both diets (0.451 and 0.355 on control and high-iron diets, respectively). Table 6 summarizes the multiple linear regression analysis results. For the control diet, strain was a highly significant predictor of plasma hepcidin in model 1 (P = 3.00 × 10⁻⁵). However, hepcidin mRNA levels in model 2 were a highly significant predictor of plasma hepcidin levels, over and above sex and strain (P = 2.50 × 10⁻⁵). For the high-iron diet, strain was the significant predictor in both models (P = 1.00 × 10⁻⁶ in model 1 and P = 2.00 × 10⁻⁵ in model 2), and hepcidin mRNA levels were only marginally significant (P = 0.014). The proportion of variance explained by predictors (R²) in model 1 was 0.200 (P = 1.52 × 10⁻⁴), and the change in R² (ΔR²) in model 2 was 0.164 (P = 2.50 × 10⁻⁵) on the control diet and 0.248 (P = 1.00 × 10⁻⁶) and 0.052 (P = 0.014) on the high-iron diet, respectively.

Fig. 4.

Hamp1 mRNA levels and plasma hepcidin levels in eight strains of mice (6 inbred and 2 recombinant inbred) on control diet (A) and nine strains on high-iron diet (high Fe; 6 inbred and 3 recombinant inbred; B) in study 2. Mean values (±SD for Hamp1 mRNA levels) were calculated for a given strain and sex. Strains are ordered by increasing values of mean Hamp1 levels in females.

Table 6.

Results of multiple linear regression models predicting the effect of sex, strain, and hepcidin mRNA levels on plasma hepcidin levels in eight strains of mice on control diet and nine strains on high-iron diet in study 2

	Control				High Fe
	B	SE	β	P	B	SE	β	P
Model 1
Sex	4.87	14.15	0.04	0.73	49.81	52.40	0.09	0.34
Strain	−12.76	2.88	−0.45	3.00 × 10−5*	−56.44	10.73	−0.49	1.00 × 10−6*
Model 2
Sex	−45.12	16.90	−0.32	0.009†	−22.82	58.50	−0.04	0.70
Strain	−7.62	2.83	−0.27	0.009†	−48.93	10.84	−0.42	2.00 × 10−5*
GE	22.14	4.94	0.57	2.50 × 10−5*	63.05	25.08	0.27	0.014‡

Control, eight strains of mice (6 inbred and 2 recombinant inbred) on control diet. High Fe, nine strains on high-iron diet (6 inbred and 3 recombinant inbred). GE, gene expression.

^*P ≤ 0.0001,

^†P ≤ 0.01,

^‡P ≤ 0.05.

We also used multiple linear regression to determine whether hepcidin mRNA and plasma hepcidin levels can predict the levels of iron in four different tissues (heart, kidneys, pancreas, and spleen) as well as whether liver iron levels can predict hepcidin mRNA and plasma hepcidin levels in the male mice of six inbred strains (Tables 7 and 8). In the hearts of mice on the high-iron diet, strain is a nominally significant predictor of iron levels only in model 1 (P = 0.02), and hepcidin mRNA levels are nominally significant only in model 2 (P = 0.04), but both relationships are lost after introducing plasma hepcidin levels as a predictor in model 3 (Table 7). Strain is a significant predictor of kidney iron levels for mice on the high-iron diet (P = 0.004) but becomes less significant with the introduction of other predictors (Table 7). Surprisingly, plasma hepcidin significantly predicts pancreatic iron levels for mice on the high-iron diet (P = 0.000053), with ΔR² of 0.435 (P = 0.000053), and spleen iron levels on the control diet (P = 0.03), with ΔR² of 0.127 (P = 0.032; Table 7). In the liver, strain plays a significant role in predicting Hamp1 expression and plasma hepcidin levels for mice on the high-iron diet (P < 0.002; Table 8). However, liver iron levels are not a significant predictor of either Hamp1 expression or plasma hepcidin levels on either diet (Table 8).

Table 7.

Results of multiple linear regression models predicting the effect of strain, hepcidin mRNA levels, and plasma hepcidin levels on tissue iron levels in male mice of six inbred strains on control and high-iron diet

	Control				High Fe
	B	SE	β	P	B	SE	β	P
Heart
Model 1
Strain	5.34	6.88	0.13	0.44	−20.05	8.08	−0.41	0.02‡
Model 2
Strain	7.29	6.97	0.18	0.30	−10.59	8.80	−0.22	0.24
GE	15.99	12.22	0.23	0.20	36.70	16.99	0.39	0.04‡
Model 3
Strain	7.06	7.22	0.17	0.34	−7.71	9.51	−0.16	0.42
GE	13.85	18.58	0.20	0.46	27.35	20.52	0.29	0.19
PH	0.03	0.001	0.04	0.89	0.06	0.08	0.19	0.42
Kidneys
Model 1
Strain	−3.90	6.80	−0.10	0.57	−19.51	6.33	−0.48	0.004†
Model 2
Strain	−2.70	7.00	−0.07	0.70	−14.44	7.21	−0.35	0.05‡
GE	9.79	12.27	0.14	0.43	19.89	14.16	0.25	0.17
Model 3
Strain	−0.47	7.00	−0.01	0.95	−10.89	7.21	−0.27	0.14
GE	30.77	17.99	0.44	0.10	4.03	16.15	0.05	0.80
PH	−0.34	0.22	−0.39	0.13	0.10	0.06	0.37	0.08
Pancreas
Model 1
Strain	−1.00	1.60	−0.11	0.54	5.73	7.87	0.13	0.47
Model 2
Strain	−1.21	1.65	−0.13	0.47	7.99	8.88	0.18	0.38
GE	−1.82	2.90	−0.11	0.53	10.34	18.12	0.12	0.57
Model 3
Strain	−0.95	1.69	−0.10	0.58	16.42	6.95	0.38	0.03‡
GE	0.78	4.40	0.05	0.86	−26.83	15.78	−0.30	0.10
PH	−0.04	0.05	−0.21	0.44	0.25	0.05	0.85	5.30 × 10-5*
Spleen
Model 1
Strain	93.24	54.40	0.29	0.10	113.28	182.53	0.11	0.54
Model 2
Strain	107.06	55.33	0.33	0.06	342.11	199.57	0.32	0.10
GE	107.03	90.80	0.20	0.25	894.01	394.85	0.43	0.03‡
Model 3
Strain	123.35	52.53	0.38	0.03‡	368.38	210.48	0.35	0.09
GE	322.97	128.45	0.61	0.02‡	781.19	472.86	0.37	0.11
PH	−3.49	1.55	−0.54	0.03‡	0.72	1.62	0.10	0.66

High Fe, high-iron diet. GE, gene expression; PH, plasma hepcidin level.

^*P ≤ 0.0001,

^†P ≤ 0.01,

^‡P ≤ 0.05.

Table 8.

Results of multiple linear regression models predicting the effect of strain, and liver iron levels on hepcidin mRNA and plasma hepcidin levels in male mice of six inbred strains on control and high-iron diet

	Control				High Fe
	B	SE	Β	P	B	SE	β	P
Hamp1 mRNA level
Model 1
Strain	−0.13	0.09	−0.24	0.18	−0.26	0.08	−0.51	0.002
Model 2
Strain	−0.22	0.11	−0.40	0.06	−0.30	0.08	−0.58	0.001
Liver iron levels	−0.001	0.001	−0.29	0.17	0.0001	0.0001	0.29	0.07
Plasma hepcidin level
Model 1
Strain	−2.01	7.51	−0.05	0.79	−84.67	20.47	−0.59	0.0002
Model 2
Strain	−5.10	9.09	−0.12	0.58	−91.59	20.71	−0.64	0.0001
Liver iron levels	−0.04	0.07	−0.13	0.55	0.02	0.02	0.21	0.16

Liver iron levels measured in micrograms per gram were used. High Fe, high-iron diet.

Study 2 (Part 2): Possible Mechanisms Affecting Hepcidin Expression

Plasma erythroferrone (ERFE) was measured in a subset of study 2 samples, mostly from male mice that were fed the low-iron diet. Most of the ERFE values were at or below the detection limit (0.04 ng/ml; see Supplemental Material, Supplemental Table S1; Supplemental Material for this article is available online at the Journal website). Mice fed the low-iron diet showed no signs of anemia as assessed by hemoglobin values (Supplemental Table S2); their results were comparable to those previously reported (36).

For the RNA-Seq analysis we used 9 liver samples per strain (3 on each diet), for a total of 54 samples. First, we compared the results for the two strains that had the highest (C57BL/6J) and lowest (DBA/2J) plasma hepcidin levels on both the high-iron and control diets (plasma hepcidin values in samples held on the low-iron diet were at or below the detection limit of 0.25 ng/ml). We identified 719 genes that were differentially expressed (false discovery rate-adjusted P < 0.05) between these 2 strains on the control diet, 797 genes on the low-iron diet, and 458 on the high-iron diet (Supplemental Table S3; “C57BL.vs.DBA”). Of these differentially expressed genes, 449 were in common between the mice on the control and low-iron diets, whereas 214 differentially expressed genes were shared between the control and high-iron diets. A total of 186 genes were differentially expressed between the strains on all 3 diets (Supplemental Table S3; “Overlap with CD” and “All three diets”). Of the genes with significantly different expression between the 2 strains, 241 genes were expressed only in liver samples from the control diet group, 285 genes were expressed only in samples from the low-iron diet group, and 181 genes were only expressed in samples from the high-iron diet group (Supplemental Table S3; “Only CD,” “Only ND,” and “Only HD”). Additionally, we used the set of genes that were in common for all three diets, and three set of genes that were unique to a given diet, to look for overrepresentation of gene ontology (GO) terms and KEGG pathways. For the genes in common for all three diets, top GO terms include “organic acids and oxoacid metabolic processes,” “oxidoreductase and monooxygenase activity,” and “metabolic processes,” whereas the top KEGG pathways are “metabolism of lipids and vitamins,” “chemical carcinogenesis,” and “drug metabolism.” For the genes unique to the control diet, top GO terms include “organelle,” “membrane-bounded organelle,” “cytoplasmic part,” “intracellular,” and “cell part,” whereas the top KEGG pathways are “infectious and immune diseases” and “immune system.” Genes unique to the low-iron diet have “small molecule metabolic,” “oxoacid metabolic,” “organic acid metabolic,” “single-organism metabolic,” and “carboxylic acid metabolic” processes as the top five GO terms, and the top pathways are “biosynthesis of unsaturated fatty acids” and “metabolism of amino acids.” Finally, the top GO terms for genes unique to the high-iron diet include “steroid,” “sterol,” and “lipid” biosynthesis and metabolism, and the top pathways are “glutathione metabolism” and “steroid biosynthesis.” All results are listed in Supplemental Table S4. Genes identified here need to be further assessed for their role in iron metabolism and possible connection to hepcidin.

Second, we looked specifically at iron-related genes to see which ones are differentially expressed between these two strains on the different diets. The list of iron-related genes comprised genes known to play a role in iron metabolism on the basis of previous publications, our human iron deficiency genome-wide association study (25, 26), our liver iron genome-wide association study in inbred mice (24), and functional annotation in Database for Annotation, Visualization and Integrated Discovery (DAVID) linking the gene to iron homeostasis/bone morphogenetic protein (BMP) pathway/homeostatic processes (13, 14). Using edgeR and a false discovery rate-adjusted P value <0.05 cutoff, we identified 11 genes differentially expressed between the 2 strains on the control diet (Arsb, Atp6v0c, Hfe, Hmox1, Sfxn, Fth1, Atp6v0d2, Sod2, Bmp4, Il6st, and Smad7), 7 genes on the low-iron diet (Arsb, Atp6v0b, B2m, Hfe, Sfxn2, Slc40a1, and Smad6), and 4 genes on the high-iron diet (Atp6v0c, Gstp1, Smad7, and Trf). All results are listed in Supplemental Table S5.

DISCUSSION

It has been established that there are sex and strain differences in iron homeostasis between different inbred mouse strains, with underlying genetic factors playing a role (7, 9, 12, 16, 19, 24, 28). Equally so, a number of studies have confirmed that there are sex and strains differences in the expression of both mouse hepcidin genes (8, 18, 22). Additionally, Hep-1 and Hep-2 serum and urine levels were examined by mass spectrophotometry and found to differ between different inbred strains and in different experimental mouse models with regard to their localization and quantity (39). These studies were done on a limited number of strains/experimental models. Our aim was to expand these findings by, at first, examining hepcidin mRNA levels in 14 inbred strains. We focused on Hamp1 only, and not Hamp2, as the previous studies confirmed that only Hamp1 plays a central role in iron metabolism (20, 22, 30, 31). Our results showed that both sex and strain have significant effects on Hamp1 expression levels, confirming previous findings, whereas the sex-by-strain interaction was suggestively significant. Significant differences were noted between the sexes in five strains in simple effects analysis; conversely, strain played an important role within both sexes. In general, mean values for females were equal to or higher than those in males in all but one case. In the second study of six inbred and three recombinant inbred strains, with two different diets (a control and a high-iron diet), we measured Hamp1 mRNA and plasma hepcidin levels in all mice. Additionally, we measured iron levels in the heart, kidneys, liver, pancreas, and spleen of male mice from six inbred strains. Our results showed that once again, both sex and strain have a significant effect on hepcidin mRNA levels on the control diet and additionally on the high-iron diet. However, sex more significantly affected the difference between Hamp1 mRNA levels than strain. The opposite was true for plasma hepcidin: strain had a highly significant effect on both diets, and sex showed a significant effect only on the high-iron diet. The sex-by-strain interaction was significant too, meaning that the effect of both factors is not independent. Further analysis of simple effects, comparing both sexes within each strain, showed that sex has a more significant effect on hepcidin mRNA levels than on plasma hepcidin. Conversely, strain shows a significant effect within each sex on both diets and for both traits but is stronger for plasma hepcidin levels. Genetic and sex differences influence gene expression in mice, and a study of these differences in the liver of three inbred strains showed that mice of the same sex, but different strains, were more similar in terms of gene expression than mice in the same strain but of different sex (37). Evidence from this study suggests that this effect extends to alternative splicing, and the authors propose trans-acting factors on the sex chromosomes, epigenetic variations, and/or hormonal differences as possible mechanisms that give rise to gene expression and/or splicing variation (37). Alternatively, the efficiency of hepcidin stability in the circulation and hepcidin excretion are some of the factors that may be influenced by the strain but are not captured by the measurements. Not yet identified parameters, including the role of noniron metals, as suggested by Cavey et al. (5), could also play a role.

We have shown that hepatic hepcidin mRNA levels are a highly significant predictor of plasma hepcidin levels, over and above sex and strain, on the control diet. One other study that measured serum hepcidin, by Tjalsma et al., also found that serum Hep-1 levels correlated well with liver Hamp1 expression (39). However, strain is still the most significant predictor of plasma hepcidin levels on the high-iron diet and has a much stronger effect than hepcidin mRNA levels.

Courselaud et al. suggested that the relationship between hepcidin expression and iron stores exists on the basis of the combined mRNA levels of both hepcidin genes and liver and spleen iron content (8). Krijt et al. report that the strain difference in hepcidin mRNA content was mainly caused by differential expression of the Hamp2 gene, with DBA/2N having higher hepcidin and higher iron stores then C57BL/6N, while Hamp1 levels are relatively constant (18). However, these two studies based their conclusions only on the observation that certain strains/sex with higher liver iron also had higher hepcidin mRNA levels in two strains only. In contrast, Cavey et al. found no statistically significant relationship between levels of Hamp1 mRNA and plasma iron, transferrin saturation, and tissue iron concentrations (5). Our results confirmed the Cavey et al. findings and also expanded them to look at the relationship between plasma hepcidin and iron parameters. We found that in general, in healthy animals in baseline conditions, hepcidin mRNA and plasma hepcidin levels are not good predictors of tissue iron.

We have shown that liver iron levels in inbred mice on an iron-sufficient control diet are not a good predictor of Hamp1 mRNA levels in 14 inbred strains. Confirming the findings of our first study, we found that liver iron levels could not predict Hamp1 mRNA levels or plasma hepcidin levels on the control diet in the second study, and equally so on the high-iron diet. Factors other than iron stores, such as erythropoietic demand, may play a role in the production of hepcidin, which, in turn, regulates iron absorption and recycling (11, 29). Plasma diferric transferrin levels were not correlated with, and therefore could not predict, Hamp1 mRNA levels, although plasma diferric transferrin may be one of the factors that increases the synthesis of hepcidin (11). The existence of feedback control between hepcidin and plasma/tissue iron levels does not predict baseline hepcidin in healthy animals, as a balance between the hormone and substrate is reached in each strain. However, this relationship may be different in pathological conditions where hepcidin levels are altered because of mutations in hepcidin regulators or because of the presence of pathological regulators of hepcidin production. In these conditions, liver and spleen iron concentrations may significantly correlate with levels of hepcidin. Tjalsma et al. found that liver iron level significantly correlated with serum Hep-1 levels (39); however, this study utilized not only wild-type mice but also mouse models of hereditary hemochromatosis in which hepcidin is deficient because of mutations in its regulators (Hfe, Tfr2, or both), and in these models, hepcidin deficiency likely drives the iron loading of the liver (positive correlation) and the spleen (negative correlation). They also found that serum Hep-1 levels were an excellent determinant for spleen iron levels (39), and our study results corroborate this finding but not the one by Courselaud et al. suggesting a link between spleen iron and combined levels of both Hamp1 and Hamp2 mRNA (8). In our study, plasma hepcidin levels are significant predictors of spleen iron levels for male mice on the control diet, but the hepcidin mRNA levels are not.

We found that plasma hepcidin is a significant predictor of pancreas iron levels in male mice on the high-iron diet. A study of mouse models of hemochromatosis suggested a more gradual accumulation of iron in the heart and pancreas once the liver had reached its capacity and also suggested a common mechanism for iron loading of the pancreas and heart that differs from the one in liver (38). A possible mechanism for the loading of iron into the pancreas and heart may involve the uptake of non-transferrin-bound iron, when transferrin becomes fully saturated (38).

To shed some light on the possible mechanisms of hepcidin regulation, we measured plasma erythroferrone (ERFE) in a subset of samples, mostly from male mice fed the low-iron diet since this is the group where we would expect the highest ERFE. However, it seems that the low-iron diet provided enough iron for erythropoiesis and did not activate ERFE expression. We also sequenced liver mRNA samples and compared the strains that had the highest (C57BL/6J) and lowest (DBA/2J) plasma hepcidin levels on three different diets. There was a difference in the number of differentially expressed genes between the diets, with some in common for all three diets and some unique for a given diet. Overrepresented GO terms and KEGG pathways suggested a difference between the genes expressed in mice on the different diets. Genes identified here need to be further assessed for their role in iron metabolism and their possible connection to hepcidin.

Furthermore, we focused our attention specifically on iron-related genes that were differentially expressed between these two strains. Because of the way the list of iron-related genes was constructed, it contained genes with varying amounts of evidence for their role in iron homeostasis. Some genes with lesser-known roles in iron metabolism that were found to be differentially expressed included arylsulfatase B (Arsb), differentially expressed in mice on both the control and low-iron diets, and were associated previously with serum ferritin in an Australian population (4). Several vacuolar ATPases (Atp6v0b, Atp6v0c, and Atp6v0d2) were differentially expressed in samples on at least one diet. Interestingly, these genes are noted to be involved in insulin receptor recycling (10), and there is evidence that insulin directly regulates hepcidin in diabetic rats (41). Other differentially expressed genes include interleukin-6 signal transducer (Il6st), which activates Jak2 and mediates signals that regulate immune response (32), and superoxide dismutase 2 (Sod2), a member of the iron/manganese superoxide dismutase family, which were both differentially expressed in samples from mice on the control diet. Sideroflexin 1 (Sfxn), which differed between the strains on the control diet, and sideroflexin 2 (Sfxn2), which differed between those on the low-iron diet, are both part of the mitochondrial membrane and potential iron transporters. Finally, mice on the high-iron diet showed a difference in the expression of glutathione S-transferase pi 1 (Gstp1), a gene that potentially affects iron status in sickle-cell disease.

A number of iron-related genes with a known important role in iron metabolism were also differentially expressed between the two strains. For example, several members of BMP/Smad signaling pathway were differentially expressed on the control diet (Bmp4 and Smad7), the low-iron diet (Smad6), and the high-iron diet (Smad7). As expected, Bmp4 levels were higher in the C57BL/6J (high hepcidin) strain and Smad7 levels were higher in the DBA/2J (low hepcidin) strain on the control diet, and Smad7 levels were higher in DBA/2J on the high-iron diet. Surprisingly, Smad6 levels were higher in the high-hepcidin strain on the low-iron diet. The differences in expression of these genes may contribute to strain differences in hepcidin levels since Bmp4 is known to stimulate and Smad6 and Smad7 are known to inhibit hepcidin production (27, 40). Ferritin (Fth1; iron storage protein) is only differentially expressed on the control diet, as well as heme oxygenase 1 (Hmox1; essential for recycling of iron from heme). Hfe is differentially expressed only on the control and low-iron diets; however, on both diets it is higher in the low-hepcidin strain, implying that perhaps hepcidin production in conditions of iron overload is under the control of a different gene(s). Not surprisingly, solute carrier family 40 (iron-regulated transporter), member 1 (Slc40a1; i.e., ferroportin) is differentially expressed between the strains on low-iron diet and transferrin (Trf) on high-iron diet.

In summary, we have confirmed that both sex and strain have significant effects on Hamp1 mRNA levels and have established that the effects of these factors are not independent. We have also shown for the first time that strain has a stronger effect than sex on plasma hepcidin levels on both control and high-iron diets. We established that hepatic hepcidin mRNA level is a highly significant predictor of plasma hepcidin levels, over and above sex and strain, for mice on the control diet, but the strain is still the most significant predictor of plasma hepcidin levels for mice on the high-iron diet. We have confirmed that liver iron levels are not a predictor of hepcidin mRNA levels on the control diet. We showed for the first time that liver iron levels on a high-iron diet or diferric transferrin levels on a control diet could not predict Hamp1 mRNA levels and that liver iron levels could not predict plasma hepcidin levels on either diet. Finally, we have shown that in general, hepatic hepcidin mRNA and plasma hepcidin levels are not good predictors of tissue iron in male mice, with the exception of plasma hepcidin being a significant predictor of spleen iron levels on a control diet and of pancreas iron levels on a high-iron diet. A number of genes are differentially expressed between the strains with high vs. low hepcidin levels. Expression of iron-related genes varies with the diets, and we identified a few candidates beyond the BMP/Smad pathway that could potentially affect hepcidin levels. However, the exact factors that are driving the differences we observed between strains, sexes, and tissues will still have to be elucidated.

Conclusions

Study 1: Hamp1 mRNA levels in 14 strains on iron-sufficient diet.

1) Both sex and strain have a significant effect on hepcidin mRNA levels. 2) Simple effects analysis shows that sex has an effect in only some strains but strain has an effect on both sexes. 3) Liver iron or diferric transferrin levels are not a predictor of hepcidin mRNA levels in baseline conditions.

Study 2 (part 1): strain- and sex-specific response to changes in iron levels.

1) Both sex and strain have a significant effect on hepcidin mRNA levels on both diets; however, the effect is driven mostly by sex. Strain had a stronger effect on plasma hepcidin levels; it had a significant effect on plasma hepcidin levels on both diets whereas sex had a significant effect only on the high-iron diet. 2) Hepatic hepcidin mRNA level is a highly significant predictor of plasma hepcidin levels, over and above sex and strain, on the control diet, but the strain is still the most significant predictor of plasma hepcidin levels on high-iron diet. 3) Liver iron levels could not predict Hamp1 mRNA levels or plasma hepcidin levels on the control or high-iron diets. 4) In general, hepcidin mRNA and plasma hepcidin levels are not good predictors of tissue iron in male mice in baseline conditions. Nevertheless, plasma hepcidin is a significant predictor of pancreas iron levels on the high-iron diet and of spleen iron levels for mice on a control diet.

Study 2 (part 2): possible mechanisms affecting hepcidin expression.

1) A number of genes are differentially expressed between the strains with high vs. low hepcidin levels, but their relationship with hepcidin needs to be further assessed. 2) Expression of iron-related genes varies with diets, and we identified some candidates beyond the BMP/Smad pathway that could potentially affect hepcidin levels.

GRANTS

This work was funded by National Institute of General Medical Sciences Grant 83198. A. J. Lusis is supported by National Heart, Lung, and Blood Institute Grant 28481.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

C.D.V. conceived and designed research; K.E.P., S.-M.L., E.V., S.T.H., G.J., and J.Z. performed experiments; S.M. and A.J.L. analyzed data; S.M., T.G., and E.N. interpreted results of experiments; S.M. prepared figures; S.M. drafted manuscript; S.M., S.T.H., B.F., T.G., E.N., and C.D.V. edited and revised manuscript; S.M., K.E.P., S.-M.L., E.V., S.T.H., A.J.L., B.F., T.G., E.N., and C.D.V. approved final version of manuscript.