Dysregulated hepcidin response to dietary iron in male mice with reduced Gnpat expression

Abstract Exome sequencing has identified the glyceronephosphate O-acyltransferase (GNPAT) gene as a genetic modifier of iron overload in hereditary hemochromatosis (HH). Subjects with HFE (Homeostatic Iron Regulator) p.C282Y mutations and the GNPAT p.D519G variant had more iron loading compared with subjects without the GNPAT variant. In response to an oral iron challenge, women with GNPAT polymorphisms loaded more iron as compared with women without polymorphisms, reinforcing a role for GNPAT in iron homeostasis. The aim of the present study was to develop and characterize an animal model of disease to further our understanding of genetic modifiers, and in particular the role of GNPAT in iron homeostasis. We generated an Hfe/Gnpat mouse model reminiscent of the patients previously studied and studied these mice for up to 26 weeks. We also examined the effect of dietary iron loading on mice with reduced Gnpat expression. Gnpat heterozygosity in Hfe knockout mice does not play a role in systemic iron homeostasis; Gnpat+/− mice fed a high-iron diet, however, had lower hepatic hepcidin (HAMP) mRNA expression, whereas they have significantly higher serum iron levels and transferrin saturation compared with wildtype (WT) littermates on a similar diet. These results reinforce an independent role of GNPAT in systemic iron homeostasis, reproducing in an animal model, the observations in women with GNPAT polymorphisms subjected to an iron tolerance test.


Introduction
Iron homeostasis is a tightly regulated process involving several genes and alternate pathways [1][2][3][4]. Many of these genes were identified using classical genetic approaches in subjects with dysregulated iron metabolism. These studies linked mutations in HFE (Homeostatic Iron Regulator) [5], transferrin receptor 2 (TFR2) [6], hemojuvelin (HJV) [7], hepcidin (HAMP) [8] and ferroportin (FPN/SLC40A1) [9,10] to different forms of the genetic iron overload disorder, hereditary hemochromatosis (HH). The most common among these mutations is HFE p.C282Y, a substitution of tyrosine for cysteine in the HFE protein at position 282 (p.Cys 282 Tyr; p.C282Y) [5]. Up to 1 in 200 people of Caucasian origin are homozygous for this mutation [11], however a variability in phenotypic and clinical presentation of HH and accumulation of iron loading has been reported [12]. These observations suggest that other genetic, epigenetic or environmental factors play a role in the development of disease [13].
Using exome sequencing, we previously identified a variant (refSNP rs11558492) in the glyceronephosphate O-acyltransferase gene (GNPAT p.D519G) [14], which was associated with increased iron overload in patients homozygous for the HFE p.C282Y mutation. The GNPAT p.D519G variant has been identified as a cause of the peroxisomal disorder rhizomelic chondrodysplasia punctata (RCDP) 2 [15], where it was shown to lead to a 70% decrease in enzymatic activity of GNPAT [15]. Subsequently, a number of studies either supporting [16,17] or opposing [18][19][20] the role of GNPAT as a modifier of HFE hemochromatosis have been reported. Interestingly, it was also reported that in response to an oral iron tolerance test, subjects with GNPAT polymorphisms had significantly higher serum iron levels, transferrin saturation [21,22] and loaded more iron compared with subjects who did not have this polymorphism, even in the absence of the HFE mutations. In addition, the p.D519G variant of GNPAT was shown to be associated with lower serum hepcidin levels in response to increased transferrin saturation [22]. These studies have suggested an independent role for GNPAT in iron homeostasis where it may be involved directly or indirectly in regulating hepcidin in response to body iron levels.
Gnpat +/− mice fed a high fat, high cholesterol diet are more susceptible to lipid accumulation and were less responsive to fluvastatin treatment [23]. This suggests that plasmalogen biosynthesis may have a role to play in liver function. GNPAT is required in plasmalogen synthesis and is the first enzyme in this pathway [24]. Defects in GN-PAT lead to disruption of membrane homeostasis due to the important role played by plasmalogens in membrane fluidity and lipid-lipid interactions [15]. Mutations in GNPAT lead to a type of rare genetic disorder, RCDP type 2. RCDP is a type of peroxisomal disorder which impairs the normal development of many parts of the body, leading to cranial abnormalities, hepatic dysfunction, visual and/or hearing loss, seizures and neural dysfunction [25] and there is evidence of disrupted iron homeostasis in patients suffering from defects in peroxisomes [26].
Based on our exome sequencing study [14] and recent mice studies [23,27] we examined the consequences of Gnpat heterozygosity in Hfe knockout mice at different ages and also in relation to dietary iron overload. While our study was being undertaken, a study reported that mice with total or hepatocyte-specific loss of Gnpat had normal systemic iron homeostasis [27] in the absence of Hfe, suggesting that Gnpat may not be a genetic modifier of iron homeostasis in mice.

Real-time PCR
RNA was isolated from mouse livers and spleens using TRIsure reagent (Bioline, Sydney, NSW, Australia). cDNA was prepared from 1 μg of total RNA using the SensiFAST cDNA synthesis kit (Bioline). For all investigated genes, real-time quantitative PCR (qPCR) was performed using the SensiFAST SYBR No-Rox kit (Bioline) under the following conditions: 5 min denaturation at 95 • C, followed by 45 cycles of 95 • C for 15 s, 60 • C for 10 s and 72 • C for 15

Hprt-F GGACTGATTATGGACAGGA
s. The relative expression of all target genes was calculated using the geometric mean of three reference genes: actin (Actb), hypoxanthine-guanine phosphoribosyl transferase (Hprt) and DNA directed RNA polymerase II subunit RPB1 (Polr2a). Primer sequences are presented in Table 1.

Western blotting
Liver tissue (100 mg) was homogenized using Precelllys Evolution tissue homogenizer (Bertin Instruments, Montigny-le-Bretonneux, France) in a buffer with phosphatase inhibitors [29]. Protein lysate (25 μg) was electrophoresed on a Bolt™ 4-12% Bis-Tris Plus gels (Thermo Fisher Scientific, Massachusetts, United States). After transferring to a nitrocellulose membrane, the proteins were blocked with 10% non-fat milk in Tris buffer with 0.1% Tween-20 (TBST) for 2 h at room temperature, and incubated with primary antibodies: anti-Tfr2 After washing with TBST and incubated with anti-rabbit IgG-horseradish peroxidase (Invitrogen, Life Technologies) 1:10000, diluted in 10% non-fat milk in TBST for 1 h at room temperature. Excess secondary antibodies were washed off and the blots were incubated with chemiluminescent substrate (Lumina Forte; Merck Millipore) for 5 min. Blots were exposed to X-ray film (Fujifilm, Brookvale, NSW, Australia) and developed using the Minolta film processor (Konica Minolta Medical and Graphic, Tokyo, Japan).

Histology
Tissues fixed with 10% formalin were processed, paraffin-embedded and sectioned by the Histology Laboratory, Central Analytical Research Facility, IHBI, QUT. Perls' staining was performed as described [32]. The slides were then scanned using the Leica SCN400 slide scanner (Leica Microsystems, Wetzlar, Germany). The analysis was performed using the Digital Image Hub, image analysis software (Leica).

Iron indices
Serum iron levels were determined using the commercial iron assay kit from Pointe Scientific (Canton, MI) [33]. Hepatic (HIC) and splenic (SIC) iron concentrations were measured in dried tissues using a colorimetric assay as previously described by Torrance and Bothwell [34].

Statistical analysis
Statistical analysis was performed using GraphPad Prism 6 software (GraphPad Software, San Diego, CA). Statistically significant differences were determined using two-way analysis of variance (ANOVA). Tukey's multiple comparison tests were performed to compare the differences between individual groups. P-values (<0.05) were considered to be statistically significant.

Gnpat +/− heterozygosity does not affect iron homeostasis in Hfe knockout mice
Exome sequencing studies have suggested that the GNPAT polymorphism (p.D519G) may be acting as a genetic modifier in patients with HFE hemochromatosis [14]. Male patients heterozygous for this polymorphism had more severe iron overloading as compared with HFE subjects without these polymorphisms [14]. In order to mimic these polymorphisms in mice we used the heterozygous Gnpat +/− mouse model. We examined the effect of Gnpat heterozygosity in Hfe −/− male mice by generating Hfe −/− /Gnpat +/− mice and comparing them with Hfe −/− littermates. These mice were all on the same background thus reducing the contribution of background to basal iron levels. Since our exome-sequencing study involved male patients and to avoid any gender-related differences, we used male mice for all our studies. Mice were analyzed at 4 (Figure 1), 10 ( Figure 2) and 26 ( Figure 3) weeks of age to examine any age-dependent effect of the loss of Gnpat. It is also known that iron accumulates with age in humans and mice [35]. We measured the HIC, SIC and relative mRNA expression of Gnpat and Hamp in the livers of these mice. As expected, the relative mRNA levels of Gnpat were significantly lower in the livers  (Figures 1C-3C). The 4-week-old Hfe −/− /Gnpat +/− mice had significantly lower HIC as compared with the Hfe −/− mice ( Figure 1A) but there were no differences in the splenic iron levels or Hamp mRNA levels in the livers of these mice ( Figure 1B,D). There were no differences in HIC, SIC or relative mRNA levels of Hamp in the livers of the 10-and 26-week-old male mice (Figures 2 and 3). These results suggest that reduced Gnpat expression may not affect Hfe knockout mice in contrast with what might have expected from the patient studies, where the GNPAT p.D519G polymorphism was associated with more severe iron overload in HFE hemochromatosis patients [17].

Effect of dietary iron challenge on Gnpat +/− male mice
While we were generating Gnpat +/− mice, two studies were published where subjects with GNPAT p.D519G were challenged with oral iron doses (iron tolerance test) [21,22]. To examine this effect in our animal models, 4-week-old male Gnpat +/− mice and WT littermates were fed a diet containing 0.25% carbonyl iron for one week. We chose this concentration of iron diet as we have shown that a 0.25% carbonyl iron containing diet is sufficient to induce maximal hepcidin response [36]. The Gnpat +/− male mice had similar HIC levels as compared with their WT littermates ( Figure 4A). Mice fed a high-iron diet showed a significant increase in HIC, SIC, total serum iron and transferrin saturation as compared with mice fed a control diet with the same genotype. Interestingly, Gnpat +/− male mice fed a high-iron diet had significantly lower SIC as compared with WT mice on a similar diet ( Figure 4B). The Gnpat +/− male mice fed a high iron diet also had significantly higher serum iron and transferrin saturation ( Figure 4C,D) as compared with the control mice on a similar diet. These results indicate that there may be reduced iron deposition in the spleens of heterozygous Gnpat +/− male mice compared with WT littermates. Perls' staining of the liver and spleen sections from the WT and Gnpat +/− mice fed either control or high-iron diet did not show any differences in the localization of iron ( Figure 4F,G). We also examined the expression of Gnpat in the livers of these mice and as shown in Figure 4E, Gnpat mRNA expression does not change with increased iron.
The Bmp-Smad (sma and mothers and against decapentaplegic) pathway is known to be one of the most important signaling pathways involved in maintaining iron homeostasis. An increase in iron levels leads to an increase in bone morphogenetic protein 6 (Bmp6) which binds to its receptors and leads to activation of the Bmp-Smad pathway, resulting in increased expression of hepcidin [1,2]. Figure 5 shows that in response to an increase in dietary iron concentration, the levels of Hamp, Bmp6, inhibitor of DNA binding 1 (Id1) and SMAD family member 7 (Smad7) mRNA levels are increased in the livers of WT and Gnpat +/− mice, as expected. In response to increased iron levels, we saw a similar increase in Bmp6 levels in the livers of both the WT and Gnpat +/− mice ( Figure 5B). Interestingly, the relative expression of Hamp, Id1 and Smad7 ( Figure 5A,C,D) were significantly lower in the Gnpat +/− mice fed a high-iron diet as compared with their WT littermates. Both Id1 and Smad7 are downstream molecules in the Bmp-Smad pathway and a reduction in their expression levels suggests that reduction in GNPAT may interfere with signaling in this pathway. These results suggest a defect in the Bmp-Smad signaling pathway in the Gnpat +/− mice. The protein levels of phospho-Smad (pSmad) 1/5 ( Figure 5E,F) increased as the levels of iron increased in the diet. The pSmad levels in the livers of Gnpat +/− mice were similar to those of WT mice fed a high-iron diet. Similar to  the mRNA expression results, the pro-hepcidin levels were significantly lower in the livers of Gnpat +/− mice fed a high-iron diet as compared with WT mice fed the same diet. An increase in the levels of iron in the diet, lead to an increase in ferritin levels as expected and the Gnpat +/− mice on a control diet had slightly lower ferritin, reflecting the lower HIC in these mice ( Figure 4A).

Discussion
GNPAT has been suggested to act as a genetic modifier of HH in patients with mutations in HFE [14,16,17]. Other reports using different cohorts failed to identify this similar correlation between HFE and GNPAT in patients with HH [18][19][20]. These varying results may be attributed to different study groups being used in the studies as well as other compounding factors, including but not limited to, alcohol consumption and the inclusion/exclusion criteria used to define the cohorts, which may lead to a low statistical power [37]. The original study used extreme phenotypes to differentiate between the two groups (patients with >1000 μg/ml of serum ferritin were considered as severe iron overload group and the mildly elevated group had serum ferritin values <300 μg/ml) [14]. Although this study had fewer numbers as compared with others [19,20,38], the other studies did not use extreme phenotypes as their case-control groups. This may also be one of the factors for differing results obtained in different cohorts as demonstrated by Besson-Fournier et al. [16], where they did not find any significant differences initially, but when they applied the same criteria as McLaren et al. [14] and examined the most severe males with serum ferritin levels >1000 μg/ml, the frequency of the rs11558492, allele G, was significantly enriched in this subset. Although controversial, the association of the rs11558492 allele G (p.D519G) with severe iron overload in HFE hemochromatosis patients cannot be completely ruled out.
In addition to the studies which analyzed the relationship between GNPAT and HFE in HH patients, a recent study also examined the relationship between iron and GNPAT polymorphisms in Taiwanese women [21]. Women with GNPAT polymorphisms had significantly higher fasting serum iron levels and after oral iron administration, both serum iron and transferrin saturation had increased significantly compared with women without GNPAT polymorphisms [21].
We fed 4-week-old male Gnpat +/− mice and their WT littermates a high-iron diet containing 0.25% carbonyl iron for 1 week. The Gnpat +/− male mice fed a high-iron diet had a significantly decreased SIC as compared with the WT littermates, suggesting that there may be a dysregulation in iron homeostasis in Gnpat +/− male mice. It has been shown that iron homeostasis in mice can be affected by strain-specific differences [39,40]. However, we minimized any effects of this by using WT littermate controls for these studies.
In our study, we did not observe any differences between the Gnpat +/− and WT mice in terms of Bmp6 levels, as also noted by An et al. [27]. Interestingly, mRNA expression of Id1and Smad7, molecules downstream in the pathway, were lower in the livers of Gnpat +/− mice fed a high-iron diet as compared with the WT littermates. The fact that there are no differences in the levels of upstream molecules (Bmp6) suggests that Gnpat may not be involved in sensing the levels of iron, but may be involved in the effective relaying of signaling either directly or indirectly, as suggested by the lower levels of mRNA expression of other downstream molecules involved in the Bmp-Smad pathway. An et al. [27] also observed a reduced hepcidin response to BMP6 treatments in primary murine hepatocytes where Gnpat had been targeted using siRNA.
Gnpat +/− mice fed an iron-rich diet have significantly increased serum iron and transferrin saturation as compared with their control littermates, this is similar to what was observed in the female subjects in the Taiwanese study, where subjects with GNPAT polymorphisms had significantly higher serum iron and transferrin saturation when administered oral iron [21]. These results suggest that Gnpat may be involved in some way in regulation of iron transport. Indeed, previous studies have shown that the rate of endocytosis and the rate of labeled transferrin uptake was reduced in skin fibroblasts from patients with GNPAT mutations as compared with control cells [15], which suggests a reduced or slower uptake of iron. It was also shown that the rate of transferrin recycling was also slower in the fibroblasts from patients with GNPAT mutations [15]. These results are different from the observations made by An et al. [27] where dietary iron loading of Gnpat KO or Hfe/Gnpat double knock mice did not show any differences in systemic iron homeostasis suggesting Gnpat does not act as a genetic modifier in Hfe KO mice. The differences between the two studies could be due to the concentration of dietary iron used; An et al. used approximately 0.8% carbonyl carbon in their diet whereas in our study we used 0.25% iron. We have previously shown that 0.25% iron is sufficient to elicit a hepcidin response in the livers of WT mice and increased iron concentrations only lead to an accumulation of iron in the liver [36]. In this study we examined male mice, whereas An et al. examined the effect of total loss of Gnpat in female mice. Previous studies have shown that gender can have an effect on plasmalogen functions [41] and the effects of gender on systemic iron homeostasis are well known [42][43][44]. Taken together, these differences in the models may be able to explain the difference in the observations made in our study and An et al. [27].
Some of our data, specifically with the mice with Gnpat heterozygosity and Hfe suggests that Gnpat may not play a role as a genetic modifier in HFE hemochromatosis as also suggested by An et al. [27]. However, in our study the HIC was significantly lower in the Hfe KO mice with Gnpat heterozygosity at 4 weeks of age as compared with Hfe KO mice, which suggests that there may be some dysregulation in iron uptake in these mice. In our recent study we showed that defects in peroxisomal function can lead to defects in the expression of multiple membrane proteins involved in iron homeostasis [45], including transferrin receptor 1 (TFR1) the receptor involved in the uptake of transferrin bound iron, since Gnpat is an essential enzyme required for peroxisomal function, these observations combined with previous results showing defects in iron recycling in fibroblasts with GNPAT defects [15] may explain the lower iron levels in the livers of mice with Gnpat heterozygosity. The reduced response to BMP6 in primary hepatocytes lacking Gnpat, as observed by An et al. [27], and the observations from the dietary iron model in our study, suggest that Gnpat may play a role in iron homeostasis.
The significantly lower SIC, higher the transferrin saturation and serum iron levels in the Gnpat +/− mice is also similar to the hemochromatosis phenotype observed in Hfe and Tfr2 KO mice, where the KO mice have lower SIC as compared with their WT littermates suggesting increased release of iron from the macrophages as a result of reduced hepcidin levels.
The results from our study and An et al. [27] suggest that Gnpat may not be a genetic modifier in HFE-hemochromatosis, however, this study examined the effect of heterozygosity of Gnpat, and not the effect of the GNPAT p.D519G variant believed to be associated with iron overload in HFE HH patients, and the possibility of this variant having an effect on HFE function cannot be completely ruled out based on the present study.
In a recent study, we have also shown that defects in peroxisomes, the organelles where GNPAT is functional, affect iron homeostasis [45]. The biochemical defects due to the dysfunctional peroxisomes lead to several changes in the liver which include affecting the expression of membrane proteins involved in maintain iron homeostasis. These results taken together with the observations made in this study suggest a role for genes involved in plasmalogen synthesis in liver function including iron homeostasis.