Effect of Erythropoietin, Iron Deficiency and Iron Overload on Liver Matriptase-2 (TMPRSS6) Protein Content in Mice and Rats

Matriptase-2 (TMPRSS6) is an important negative regulator of hepcidin expression; however, the effects of iron overload or accelerated erythropoiesis on liver TMPRSS6 protein content in vivo are largely unknown. We determined TMPRSS6 protein content in plasma membrane-enriched fractions of liver homogenates by immunoblotting, using a commercial antibody raised against the catalytic domain of TMPRSS6. Plasma membrane-enriched fractions were obtained by centrifugation at 3000 g and washing. TMPRSS6 was detected in the 3000 g fraction as a 120 kDa full-length protein in both mice and rats. Feeding of iron-deficient diet as well as erythropoietin treatment increased TMPRSS6 protein content in rats and mice by a posttranscriptional mechanism; the increase in TMPRSS6 protein by erythropoietin was also observed in Bmp6-mutant mice. Administration of high doses of iron to mice (200, 350 and 700 mg/kg) decreased TMPRSS6 protein content. Hemojuvelin was detected in the plasma membrane-enriched fractions of control animals as a full length protein of approximately 52 kDa; in iron deficient animals, the full length protein was partially cleaved at the N-terminus, resulting in an additional weak band of approximately 47 kDa. In livers from hemojuvelin-mutant mice, TMPRSS6 protein content was strongly decreased, suggesting that intact hemojuvelin is necessary for stable TMPRSS6 expression in the membrane. Overall, the results demonstrate posttranscriptional regulation of liver TMPRSS6 protein by iron status and erythropoietin administration, and provide support for the interaction of TMPRSS6 and hemojuvelin proteins in vivo.


Introduction
Iron deficiency is the most common cause of anemia worldwide. Although most cases of iron deficiency anemia respond to oral iron supplementation, some patients suffer from iron-iron overload experiments, iron was administered as iron dextran (Sigma Aldrich) to male adult C57BL/6 mice at 200, 350 and 750 mg/kg by single intraperitoneal injection; animals were sacrificed one week after treatment. For short-term iron overload experiments, animals were administered a single intraperitoneal injection of iron dextran at 1000 mg/kg and were sacrificed 24 h later [17].
Female mice with disruption of the Hfe2 gene coding for hemojuvelin (Hfe2-/-mice), age 3-4 months, were originally a generous gift from Prof. Silvia Arber, Basel, Switzerland [18]. Female Bmp6-null mice were a generous gift from Prof. Roel Goldschmeding, University Medical Center Utrecht, the Netherlands [19]. The EPO administration scheme for Hfe2-and Bmp6-null mice was identical to C57BL/6 mice. Liver samples from mask mice [2], lacking the proteolytic domain of TMPRSS6, were a generous gift from Dr. Pauline Lee and Dr. Xin Du, Scripps Research Institute, La Jolla, CA, USA.

Sample Preparation for Immunoblotting
For all experiments, a plasma membrane-enriched fraction obtained by centrifugation at 3000 g was used. Samples of liver (approx. 250 mg) were homogenized with a 6 mm Ultra Turrax homogenizer (3 x 10 s at maximum speed) in 2 ml of 10 mM HEPES, pH 7.4, containing protease inhibitors and 2 mM EDTA. The homogenate was centrifuged for 10 min at 400 g, and the supernatant was centrifuged at 3000 g for 15 min. The pellet was resuspended (Ultra Turrax, 10 s) in 10 mM HEPES containing 2 M NaCl to deplete the sample of co-precipitated nonmembrane proteins associated by ionic bonds [20] and recentrifuged at 3000 g for 15 min. Subsequently, the pellet was resuspended in 0.1 M sodium carbonate, agitated for 1 h at 4°C and recentrifuged. This step is postulated to remove soluble proteins trapped in vesicles formed during homogenization [20]. Next, the pellet was resuspended in 10 mM HEPES containing 4 M urea and 100 mM NaCl to wash out residual non-membrane proteins. Finally, the pellet was washed with 10 mM HEPES and resuspended by sonication in 125 μl of 2% SDS containing 25 mM of ammonium bicarbonate. Insoluble material was pelleted by centrifugation at 16 000 g for 10 min and the supernatant was aliquoted and stored at -80°C. Typical yield for the washed 3000 g fraction was about 1.5 mg of protein per 1 g of liver. Immunoblotting was performed under reducing conditions as previously described [21]. Primary antibodies were: Rabbit anti-TMPRSS6, Ab56182, Abcam, 1:750; goat anti-HFE2, AF 3634, R&D Systems, 1:1500; goat antineogenin, AF1079, R&D Systems, 1:1000, and rabbit anti-Na+/K+ATPase α, SC-28800 (ATP1A), Santa Cruz, 1:10 000. Secondary antibodies (anti-rabbit, 711-036-152, 1:40 000 and anti-goat, 705-036-147, 1:40 000) were from Jackson Immunoresearch.

Real-Time PCR, PNGase F Treatment, and Iron Determinations
Real-time PCR was performed on a Biorad IQ5 instrument using SYBR Green protocol, primers are listed in S1 Table. For the removal of N-linked oligosaccharides, PNGase F (New England Biolabs) was used according to manufacturers instructions. Iron was determined according to Torrance and Bothwell [22].

Statistical Analysis
For densitometric quantification, TMPRSS6 band densities obtained on a Biorad GS-800 densitometer were normalized to their respective loading control band densities; treated vs. control sample values were then compared using paired t-test. Graphed band density data, generally calculated from several blots, are expressed in % of control, control values were set as 100%. For small data sets (n = 3), values obtained from single 15-well gels were analyzed by unpaired t-test. PCR data were evaluated by the Mann-Whitney test. Data are expressed as means ± SD.

Results
HFE2 and TMPRSS6 are Present as Full Length Proteins in the Liver 3000 g Fraction HFE2 has been clearly identified as a TMPRSS6 substrate in in vitro studies [9]. In order to examine the interaction between TMPRSS6 and HFE2 in vivo, it was necessary to select a cellular fraction which would allow reproducible quantification of both proteins. In our previous experiments, we found that HFE2 is easier to detect in microsomal fraction than in whole homogenate [23]. However, published studies on HFE2 processing indicate that HFE2 undergoes complex retrograde trafficking to the Golgi compartment [24]. As liver microsomes also contain parts of Golgi membranes, while the proposed cleavage of HFE2 occurs at the plasma membrane, liver microsomes do not appear to be ideally suited for the investigation of the TMPRSS6/HFE2 interaction. Therefore, a partially purified membrane fraction obtained by centrifugation at 3000 g was used for the current studies. The detection of TMPRSS6, and particularly the detection of HFE2, was significantly improved in this fraction, as was the detection of ATP1A, a marker of the plasma membrane (Fig 1). Whereas TMPRSS6 was detected as a 120 kDa band in all samples, the size of the HFE2 band was influenced by the method of Immunoblot of TMPRSS6, HFE2 and ATP1A in the 3000 g fraction. Detection of TMPRSS6, HFE2 and ATP1A is enhanced in the 3000g fraction of mouse liver homogenates (3000g) in comparison with whole liver homogenates prepared in 1% NP40 (NP40) or microsomes (Micros). Samples were prepared from C57BL/6 mice (C), mask mice (m), Hfe2+/+ mice (WT) and Hfe2-/-mice (KO). Arrows denote the TMPRSS6-, HFE2-and ATP1A-specific bands. In contrast to the NP40 whole liver homogenate or liver microsomes, HFE2 is present in the 3000 g fraction solely as a full-length protein. sample preparation and the detergent used. In our previous studies using NP-40-containing liver homogenates, as well as in detergent-free microsomes, HFE2 was detected as a heterodimer composed of two bands of 35 and 18 kDa [21,23]; in the present study using the 3000 g fraction HFE2 was present as a strong band of approximately 52 kDa (Fig 1). This size corresponds to full-length, glycosylated HFE2 protein. These data suggest that whereas the major part of cellular HFE2 protein exists as a heterodimer, a small amount of the protein is present at the plasma membrane as the uncleaved full-length 52 kDa chain.
In vivo studies on TMPRSS6 protein have so far used mostly non-commercial antibodies [13,25]. In an attempt to identify a suitable commercially available antibody, we used liver samples from C57BL/6 and mask mice. The Ab56182 antibody, raised against the catalytic domain of TMPRSS6, enabled detection of a TMPRSS6-specific band of about 120 kDa which was not present in samples from mask mice (Fig 1). The observed size corresponds to reported TMPRSS6 size in rats and mice [13,25], and very probably represents the full-length mouse protein (811 amino acids, predicted molecular weight 91 kDa plus 7 glycosylations). On mouse liver blots, an additional lower band of approximately 110 kDa was occasionally seen; however, since this band was also weakly detectable in some mask mice samples, only the 120 kDa band was used for quantification.
TMPRSS6 is reported to undergo proteolytic activation which is proposed to produce a non-glycosylated catalytic domain fragment of about 26 kDa, connected via disulfide bridges to the membrane-bound part of the protein [26,27]. In agreement with published data [13,28], no TMPRSS6-specific band of this size was observed on immunoblots (Fig 1), suggesting that the amount of activated TMPRSS6 is small compared to the full-length protein.

TMPRSS6 Protein is Increased in Iron-deficient Rats and Mice
It has previously been reported that TMPRSS6 increases following short-term administration of iron-deficient diet to weaned rats [13]. For our experiments, we used young mice and rats kept on an iron-deficient diet for several weeks. The treatments resulted in iron-deficiency anemia and a significant decrease of liver iron concentration ( Table 1).
The 120 kDa TMPRSS6-specific band was significantly increased in both mice and rats (Fig  2A and 2B). Liver Tmprss6 mRNA content tended to increase in iron-deficient rats, but the changes did not reach statistical significance ( Fig 2C). Rat liver samples probed with the Ab56182 antibody displayed, in addition to the 120 kDa TMPRSS6 band, a relatively strong band at 100 kDa (Fig 2A). Since this band did not react to any treatment, and was not seen in mouse samples, we regard it as a nonspecific band.

TMPRSS6 Protein is Increased in Erythropoietin-treated Rats and Mice
Hepcidin expression is strongly affected by the rate of erythropoiesis [7]. Therefore, it was of interest to determine the effect of EPO administration on TMPRSS6 protein expression. As expected, administration of high doses of EPO for four days resulted in dramatic downregulation of Hamp mRNA (less than 0.1% of control values, S1 Fig). TMPRSS6 protein content increased in both rats and mice (Fig 3A and 3B); the effect in rats was more pronounced. The effect of EPO on TMPRSS6 protein was posttranscriptional, as Tmprss6 mRNA content was not influenced by EPO treatment (Fig 3C).
Since the administration of EPO for four days resulted in a significant increase of blood hemoglobin and a decrease of liver iron (Table 1), it was not possible to determine whether the effect of EPO on TMPRSS6 protein content was primarily related to accelerated erythropoiesis, or to decreased iron availability. Therefore, we performed an additional experiment, in which rats were sacrificed 24 hours after EPO administration. In this experimental setting, liver Hamp mRNA content decreased only to approximately 25% of controls and liver iron content was not affected (S2 Fig). Liver TMPRSS6 protein content was not significantly changed (S2 Fig), indicating that EPO influences TMPRSS6 protein content only after prolonged administration, which also substantially affects iron homeostasis.

Mutant Mice
BMP6 protein is regarded as a key extracellular molecule which initiates iron-dependent signaling at the hepatocyte plasma membrane through the BMP6/HFE2/SMAD pathway, ultimately leading to an increase in hepcidin transcription [29,30]. It was therefore of interest to determine whether the observed regulation of TMPRSS6 protein by EPO remains functional in Bmp6-null mice. In accordance with results obtained in C57BL/6 mice, EPO tended to increase TMPRSS6 protein content in Bmp6-null mice (Fig 4A and 4B) by a posttranscriptional mechanism (Fig 4C). Although the results did not reach statistical significance, due to limited number of available mouse pairs (n = 3), they nevertheless indicate that the EPO-mediated regulation of TMPRSS6 does not require functional BMP6 protein. Similarly to C57BL/6 mice, EPO treatment dramatically decreased Hamp mRNA in Bmp6-null mice (Fig 4D), demonstrating that BMP6 is dispensable for the regulation of Hamp expression by accelerated erythropoiesis.

Iron Decreases Liver TMPRSS6 Protein Content in Rats and Mice
Injection of iron dextran at 200, 350 and 700 mg/kg decreased TMPRSS6 protein (Fig 5A). However, lower doses of 50 and 100 mg/kg did not produce an effect (S3 Fig). As expected [7,10], iron treatment increased Hamp and Bmp6 expression (S4 Fig). In contrast to the decrease of TMPRSS6 protein in mice treated with high doses of iron dextran (Fig 5A), Tmprss6 mRNA content was slightly increased (S4 Fig). In livers from iron-treated mice, a strong additional dose-dependent band was apparent at 75 kDa (Fig 5A). Since this band did not react to PNGase F treatment, and was also seen in liver samples from mask mice kept on an iron-enriched diet (S5 Fig), we regard it as a non-specific band, probably related to the high intracellular liver iron content (S2 Table). It has been reported that the response of Hamp gene to iron overload differs with different time intervals, and different sites of liver iron accumulation [17,[31][32][33]. It was therefore of interest to determine whether the observed iron-induced decrease in TMPRSS6 protein content also occurs early after iron administration. To this purpose, we used mice injected with 1000 mg/kg of iron dextran 24 h prior to sacrifice. In this experimental setting, iron is reportedly not upregulating liver Bmp6 expression [17], and the response of Hamp mRNA to iron remains limited (S4 Fig). TMPRSS6 protein content was again decreased (Fig 5B), suggesting that iron influences TMPRSS6 protein content and Bmp6 gene expression by different pathways.

Full Length HFE2 is Partially Cleaved in Iron-deficient Rats and Mice
In previous studies using liver microsomes, we found unchanged HFE2 protein content in samples from iron-deficient mice [34]. However, in the present study, the 3000 g samples from iron-deficient animals displayed, in addition to the main band of approximately 52 kDa, an additional weak truncated band of approximately 47 kDa (Fig 6A). Since TMPRSS6 protein was increased in the same samples from iron-deficient rats and mice (Fig 2A), it is theoretically possible that this shorter band represents a TMPRSS6-cleaved fragment of full-length HFE2. In contrast to the effect of iron deficiency, EPO treatment did not result in increased cleavage of HFE2, despite an increase in TMPRSS6 protein content (Fig 6B). These results possibly indicate that increased cleavage of HFE2 requires prolonged treatment and/or severe iron deficiency.
The modest shift of the cleaved band vs. the full-length protein (approximately 4 to 6 kDa) indicates a possible cleavage site at one of the several arginines at the N-terminus of HFE2. To determine whether the cleavage occurs in the glycosylated part of the protein, low-iron samples displaying distinct cleaved bands were treated with PNGase F, which removes N-linked oligosaccharides. This treatment reduced the apparent mass of the full length HFE2 protein by approximately 10 kDa (from 52 to 42 kDa); the size of the cleaved band present in iron-deficient samples was reduced by PNGase F treatment from 47 kDa to approximately 37 kDa ( Fig  6C). These data demonstrate that the iron deficiency-induced cleavage of HFE2 does not affect the number of N-linked oligosaccharides present in the membrane-bound part of the protein, indicating that the cleavage site is located N-terminally to the first N-linked oligosaccharide (Fig 6D).

TMPRSS6 Protein is Decreased in Hfe2-/-Mice
HFE2 is the proposed target for the protease activity of TMPRSS6 [9]. However, in our previous studies with whole liver homogenates and liver microsomes [21], we found decreased, rather than increased, HFE2 protein content in mask mice. A similar decrease in full-length HFE2 protein in mask mice was confirmed in the present study utilizing the 3000 g membrane fraction (Fig 7B). Vice versa, the amount of TMPRSS6 protein was strongly decreased in liver samples from Hfe2-/-mice (Fig 7A). In addition, Hfe2-/-mice displayed strongly decreased content of neogenin (Fig 7C), another protein reportedly participating in Hamp gene regulation [25,35].
Similar to mice injected with iron, the anti-TMPRSS6 antibody detected a strong band at 75 kDa in Hfe2-/-mice ( Fig 7A). As the position of this band did not change following PNGase F treatment (S5 Fig), it very probably represents a non-specific band.

Discussion
TMPRSS6 is a hepatocyte transmembrane serine protease whose mutations result in ironrefractory iron deficiency anemia [1]. The identification of TMPRSS6 as a potent negative regulator of hepcidin, the key iron-regulatory hormone, has raised the possibility that hepcidin expression could be regulated by modulation of TMPRSS6-dependent proteolytic activity. However, very little information is available on the in vivo regulation of TMPRSS6 protein levels.  Fig 2A. (B) Comparison of the effects of iron-deficient diet (ID) or EPO treatment (E) on TMPRSS6 protein content and HFE2 cleavage in rats. Both treatments increase TMPRSS6 protein content; cleaved HFE2 band (double arrow) is apparent only in iron-deficient animals. (C) Cleaved HFE2 protein has the same number of N-linked oligosaccharides as the full length protein. Samples from control (C) or iron-deficient (ID) rats and mice were incubated with PNGase F (PNG +) or water (-). PNGase F treatment reduced the size of full length HFE2 protein (arrow) by about 10 kDa; the same size reduction of approximately 10 kDa was observed for the cleaved HFE2 protein (double arrow) present in iron-deficient samples. (D) Schematic representation of full length mouse HFE2 protein. The 361 amino acid chain has Q33 as the N-terninal amino acid (UniProt entry Q7TQ32), and is bound with a GPI anchor to the plasma membrane (PM). Arrow denotes the approximate position of the observed cleavage site, which is located N-terminally of the three N-linked oligosaccharide (NGlc) chains.
doi:10.1371/journal.pone.0148540.g006 Hepcidin expression is known to respond to iron overload and iron deficiency, as well as to the rate of erythropoiesis [7]. Therefore, the primary aim of this study was to determine TMPRSS6 protein content in plasma membrane-enriched fractions of liver samples from experimental animals subjected to manipulations of iron homeostasis or EPO administration. Since HFE2 is the proposed target of TMPRSS6, HFE2 protein levels were determined as well.
Analysis of samples from iron-deficient animals confirmed the previously reported [13] increase of TMPRSS6 protein by iron deficiency. Tmprss6 mRNA did not significantly change in iron deficiency anemia, highlighting the importance of posttranscriptional regulation of TMPRSS6 [36]. Interestingly, immunoblotting of HFE2 protein indicated that in iron-deficient animals the full length HFE2 protein is partially cleaved. Since TMPRSS6 protein content was increased in the same samples, these results indirectly support the proposed concept that membrane HFE2 could be the physiological target of TMPRSS6 [9]. The modest band shift of about 3-6 kDa suggests that HFE2 is cleaved relatively close to the N-terminus, while treatment with PNGase F indicates that the resulting shortened membrane-bound fragment retains all three of its N-linked oligosaccharides (Fig 6D). In this respect, it is interesting to note that the ligand of HFE2, BMP6, was proposed to bind at the N-terminal part of HFE2 [37]. It is therefore possible that the partial cleavage of HFE2 at the N-terminus could diminish its ability to function as BMP6 coreceptor. In conditions of iron deficiency, increased TMPRSS6 protein content could thus result in decreased activity of the BMP6/HFE2 signaling pathway, contributing to the well-documented decrease of Hamp expression. In sufficiently exposed blots, the cleaved band could also be seen in control samples, demonstrating that HFE2 cleavage occurs to some degree in control livers as well. However, even under conditions of severe iron deficiency, the intensity of the truncated band was much weaker than the intensity of the full-length protein band, suggesting that only a minor part of the full-lengh plasma membrane HFE2 is actually cleaved.
Injections of iron dextran to mice decreased liver TMPRSS6 protein content by a posttranscriptional mechanism. This response of TMPRSS6 to iron again confirms the inverse correlation between hepatic Hamp mRNA content and TMPRSS6 protein content, indicating that the modulation of TMPRSS6 activity could be an important factor in the regulation of hepcidin expression.
Clinical as well as experimental studies indicate that TMPRSS6 is particularly important for the appropriate downregulation of hepcidin expression by signals related to accelerated erythropoiesis. It has for a long time been known that administration of EPO to experimental animals downregulates Hamp transcription [38]. This EPO-mediated decrease of Hamp expression occurs not only in rats and wild-type mice, but also in HFE2-deficient mice [39,40], TFR2-deficient mice [40], and BMP6-deficient mice (Fig 4D). In striking contrast to these animals, TMPRSS6-deficient mice fail to decrease Hamp mRNA content following EPO treatment [41,42]. In addition, it has been reported that EPO administration does not improve anemia in patients with TMPRSS6 mutations [43]. These reports strongly suggest that TMPRSS6 is a crucial component of the "erythroid regulator" which increases iron absorption in response to erythropoietic demand [44]. In agreement with this concept [45,46], we found increased TMPRSS6 protein content in the livers of EPO-treated mice and rats. Intriguingly, although liver TMPRSS6 protein increased both in response to EPO treatment or in response to iron deficiency, the partial cleavage of HFE2 protein was not observed in EPO-treated animals, possibly indicating that detectable cleavage occurs only in conditions of prolonged and severe iron depletion.
The exact pathway(s) which decrease Hamp expression following administration of EPO are at present under intensive investigation. Very probably, the dramatic downregulation of liver Hamp mRNA content observed in EPO-treated experimental animals results from a combination of two main events: The EPO-induced decrease in liver and plasma iron content, which causes decreased signaling through the iron-sensing signaling pathways, and the synthesis, in the activated bone marrow and spleen, of specific factor(s) mediating Hamp downregulation by an yet unknown mechanism [40,47]. It is not clear whether the observed EPO-induced increase in TMPRSS6 protein is related solely to the EPO-induced efflux of liver and plasma iron into the erythron, as suggested by the observed lack of effect of short-term EPO administration, or whether EPO also increases TMPRSS6 protein content by some other mechanism. In any case, the increase of TMPRSS6 protein content observed in mice and rats treated with EPO for several days could, by downregulating BMP6/HFE2 signaling, contribute to the profound downregulation of Hamp expression in these animals.
Although HFE2 is the most probable target of TMPRSS6, we previously paradoxically found decreased, rather than increased, amount of the autocatalytically cleaved 35 kDa HFE2 fragments in microsomes from TMPRSS6-mutated mice [21,23]. In the present study, we confirm decreased HFE2 protein content in the liver 3000 g fraction from mask mice. To further study the interaction between TMPRSS6 and HFE2, we now reciprocally determined TMPRSS6 protein levels in Hfe2-/-mice ( Fig 7A). Interestingly, disruption of the Hfe2 gene resulted in a dramatic decrease of TMPRSS6 protein content. A possible explanation for this observation is that TMPRSS6 and HFE2, and possibly also neogenin, function as a protein complex in the hepatocyte plasma membrane, as previously reported by other groups [9,25]. Disruption of one component of this protein complex could then lead to accelerated degradation of the other component(s). This would explain the decreased liver TMPRSS6 and neogenin content in Hfe2-/-mice, as well as the decreased HFE2 content in mask mice.
Overall, the in vivo results reported in this study do support the reported [9] interaction between TMPRSS6 and HFE2 proteins. However, this does not exclude the possibility that, in addition to HFE2, TMPRSS6 could have other important physiological targets as well. While available literature data indicate that the presence of functional TMPRSS6 protein represents a sine qua non condition for the EPO-induced decrease of liver Hamp mRNA [41][42][43]45] it has also very recently been reported that EPO-mediated downregulation of liver Hamp expression does not occur in mice lacking the newly identified erythroblast-secreted factor erythroferrone [40]. This could indicate that both TMPRSS6 and erythroferrone are important components of one signaling pathway [46]. In this respect, it should be noted that despite the profound decrease in hepatic TMPRSS6 protein in Hfe2-/-mice (Fig 7A), these mice are still able to downregulate Hamp expression following EPO administration [39,40], possibly indicating that even a limited amount of Tmprss6 protein enables efficient erythroferrone signaling. Obviously, further studies will be necessary to elucidate the exact role of TMPRSS6, and to identify its possible additional substrates.
In conclusion, our study demonstrated, in both mice and rats, a posttranscriptional increase of liver TMPRSS6 protein by erythropoietin. In addition, the observed partial cleavage of HFE2 in liver samples with high TMPRSS6 protein content, as well as the substantial decrease of TMPRSS6 protein in Hfe2-/-mouse liver samples, strongly suggests that TMPRSS6 and HFE2 proteins interact at the hepatocyte plasma membrane.