Cytokines are integral to the development of anaemia of chronic inflammation. Cytokines modulate hepcidin expression and iron sequestration by the reticuloendothelial system but their direct effects on small bowel iron transport are not well characterized. The aim of the present study was to examine the local effects of TNFα (tumour necrosis factor α) on small bowel iron transport and on iron transporter expression in the absence of hepcidin. The effects of TNFα on iron transport were determined using radiolabelled iron in an established Caco-2 cell model. The effect of TNFα on the expression and localization of the enterocyte iron transporters DMT-1 (divalent metal transporter 1), IREG-1 (iron-regulated transporter 1) and ferritin was determined utilizing Caco-2 cells and in a human ex vivo small bowel culture system. TNFα mediated an early induction in both iron import and iron export, which were associated with increased DMT-1 and IREG-1 mRNA and protein expression (P<0.05). However, by 24 h, both iron import and iron export were significantly inhibited, coinciding with an induction of ferritin heavy chain (P<0.05) and a decrease in DMT-1 and IREG-1 to baseline levels. In addition, there was a relocalization of IREG-1 away from the basolateral cell border and increased iron deposition in villous enterocytes. In conclusion, TNFα has a direct effect on small bowel iron transporter expression and function, leading to an inhibition of iron transport.
Iron is an essential dietary element and its dysregulation is common to a number of pathological conditions . As there is no physiological excretion of iron, its absorption from the small bowel must be tightly regulated, and the small bowel is therefore a vital site in the regulation of iron metabolism.
Ferrous iron is taken up into the enterocyte by the multi-ion DMT-1 (divalent metal transporter 1; also known as Nramp2 (natural-resistance-associated macrophage protein 2)  and DCT-1 (divalent cation transporter 1) ), which is located on the apical villous membrane of the duodenum and proximal jejunum . Once in the enterocyte, iron has two possible fates: it may be stored bound to ferritin, or is exported out of the enterocyte by the basal IREG-1 (iron-regulated transporter) , also known as ferroportin  and MTP-1 .
Anaemia of chronic inflammation represents a reorganization of the internal iron turnover, probably as an adaptive response to chronic inflammation or infection, iron being essential for the metabolism of pathogens . This condition is characterized by the sequestration of iron in reticuloendothelial cells with resultant hypoferraemia and reduced erythropoiesis in the face of adequate body iron stores .
Although the pathway of iron absorption has been described, its regulation is not well understood. Recently, the hepatic antimicrobial peptide hepcidin has been identified as the putative central regulator of iron absorption [9,10]. The regulation of hepcidin expression in response to the body's iron requirements is becoming well established . An intriguing feature of hepcidin expression is its modulation by cytokines, especially IL-6 (interleukin-6) . In a recent elegant study, Nemeth et al.  demonstrated that, in humans, IL-6 is able to induce hepcidin expression and significantly lower serum iron levels. This occurs probably by a combination of decreased small bowel iron absorption and increased sequestration of iron into the reticuloendothelial system. In addition, the proinflammatory cytokine interferon γ has been shown to increase the uptake and decrease the export of iron in a monocytic cell line . This establishes two likely roles for cytokines in the anaemia of chronic disease: up-regulation of hepcidin and direct effects of cytokines on the reticuloendothelial system.
Although there is evidence suggesting a role for cytokines in modulating hepatic hepcidin expression and iron transporter expression in monocytes, little is known regarding possible local effects of cytokines on small bowel iron transport.
In the present study, we have used two small bowel cell models, namely the cell line Caco-2 and a human small bowel ex vivo culture system, to investigate the direct effects of the proinflammatory cytokine TNFα (tumour necrosis factor α) on the expression of the main proteins involved in enterocyte iron transport: DMT-1, IREG-1 and ferritin. The Caco-2 cell line was selected since the cells have an enterocyte-like phenotype, expressing a number of small intestinal brush border enzymes and nutrient transporters [15,16]. In addition, this cell line expresses the key iron transport proteins, including DMT-1 , IREG-1  and ferritin , as well as being the most commonly used model to study iron transport.
We have clearly shown that TNFα can dynamically regulate the expression and localization of iron transporters, ultimately having an impact on iron transport, and have provided evidence for a novel mechanism whereby cytokines can contribute to the anaemia of chronic inflammation.
MATERIALS AND METHODS
Caco-2 cells were cultured in Dulbecco's modified Eagle's medium (Invitrogen-Gibco, Paisley, U.K.) with 10% (v/v) foetal calf serum supplemented with 100 units/ml penicillin and 0.1 mg/ml streptomycin, and grown to 14 days post-confluency before experimentation. A dose–response curve for the effect of TNFα on DMT-1 mRNA expression in Caco-2 cells was generated to identify the optimal dose for stimulation experiments (5 ng/ml; results not shown). Cells were challenged with or without TNFα for 1–24 h and then processed for the determination of iron transporter expression levels or iron uptake studies.
Ex vivo small bowel cultures
Endoscopic duodenal biopsies were collected from adult patients undergoing upper gastrointestinal endoscopy for dyspeptic symptoms. All biopsies were confirmed to be histologically normal and, in addition, all patients had normal serum iron levels (14.3–35.8 μM). Endoscopic duodenal biopsies (n=60) were cultured ex vivo for up to 6 h as described previously . Each biopsy was divided into two and cultured in Medium 199 (Gibco), supplemented with 10% foetal calf serum, 1 μg/ml insulin, 500 units/ml streptomycin and 250 units/ml penicillin in a 95% O2/5% CO2 atmosphere at 37 °C for up to 6 h in the presence or absence of TNFα (0.5 ng/ml, optimum dose; results not shown). At two time points (1 and 6 h), biopsies were processed to determine iron transporter expression levels and localization by real-time PCR, Western blotting and immunohistochemistry respectively. Viability of biopsies was assessed by the lactate dehydrogenase assay before its inclusion in the present study as described previously .
Real-time PCR was performed as described previously . Briefly, all reactions were performed using 18 S rRNA as an internal standard (PE Biosystems, Foster City, CA, U.S.A; Roche, Indianapolis, IN, U.S.A.), and contained one of the sets of probes and primers listed in Table 1, in the presence of 1× Mastermix (PE Biosystems), 50 nM 18 S 5′- and 3′-primers and 200 nM 18 S probe [5′-VIC®, 3′-TAMRA (tetramethylrhodamine) labelled] and 0.25 μl of cDNA (equivalent to 12.5 ng of reverse transcribed RNA) in a 25 μl reaction mixture. Reactions without cDNA were included as negative controls. All reactions were performed in triplicate. Gene expression was normalized to the 18 S and represented as ΔCt values. For each sample, the mean for three ΔCt values was calculated. Comparison of gene expression between control and treated samples was derived by subtracting control ΔCt values from treatment ΔCt values to give a ΔΔCt value, and relative gene expression was calculated as 2−ΔΔCt. Relative gene expression is normalized to 1.0 (100%) of controls. Each experiment was performed in triplicate.
|Probe (5′-FAM 3′-TAMRA)||Forward primer (5′–3′)||Reverse primer (5′–3′)|
|DMT-1 IRE +ve||CTCTATCAGGCTTAGGA||CCATATGAAATATAAA||CCCCTCTTAACTTCCAC|
|DMT-1 exon 1A||AGGCAGCTCCACACTGT||CATGAGGAAGAAGCAG||TGTTCAGGACCCAGCAC|
|DMT-1 exon 1B||TTCTAAGAACTCAGCCA||TGCGGAGCTGGTAAGA||TGTTCAGGACCCAGCAC|
|Probe (5′-FAM 3′-TAMRA)||Forward primer (5′–3′)||Reverse primer (5′–3′)|
|DMT-1 IRE +ve||CTCTATCAGGCTTAGGA||CCATATGAAATATAAA||CCCCTCTTAACTTCCAC|
|DMT-1 exon 1A||AGGCAGCTCCACACTGT||CATGAGGAAGAAGCAG||TGTTCAGGACCCAGCAC|
|DMT-1 exon 1B||TTCTAAGAACTCAGCCA||TGCGGAGCTGGTAAGA||TGTTCAGGACCCAGCAC|
Caco-2 cells after subjecting to stimulation were processed for Western blotting as described previously  with monoclonal antibodies raised against either DMT-1 [5 μg/ml NRAMP24-A (recognizing IRE +ve/−ve isoforms); Alpha Diagnostic International (ADI), San Antonio, TX, U.S.A.; IRE stands for ironresponsive element], IREG-1 (5 μg/ml MTP11-A; ADI), ferritin H (ferritin heavy chain; 1:500; Abcam, Cambridge, U.K.) or Cytokeratin 19 (1:2.000; Oncogene Research Products, Cambridge, MA, U.S.A.), the latter being employed for normalization of epithelial protein loading. Immunoreactive bands were then subjected to densitometry using the software NIH Image 1.62. Before reprobing with the different antibodies, the membranes were incubated in stripping buffer containing 2% (w/v) SDS and 100 mM 2-mercaptoethanol in 62.5 mM Tris/HCl (pH 6.8) for 30 min at 70 °C.
To ensure that there was no inter-patient variability between treated and non-treated samples, each duodenal endoscopic biopsy was initially divided into two and cultured for 6 h in the presence or absence of TNFα. Biopsies were then fixed in formalin, embedded into paraffin and 7 μm sections were processed for immunohistochemistry. Sections were dewaxed and then incubated in 0.1% Tween 20 containing 1 mM EDTA (pH 8.0) for 16 h at 65 °C. Sections were then incubated for 1 h with either DMT-1 [10 μg/ml NRAMP24-A (recognizing IRE +ve/−ve isoforms); ADI], IREG-1 (10 μg/ml MTP11-A; ADI.) or ferritin H (1:500; Abcam). Immunoreactivity was detected by the avidin-biotinylated secondary antibody method (Dako ABC kit; Dako, Carpinteria, CA, U.S.A.) and visualized with diaminobenzidine reagent. Sections were counterstained with haemotoxylin. Omission of primary antibody was employed as a negative control. Stained sections were scored independently by three observers (N.S., C.T. and T.I.) with regard to cellular localization. Images were visualized from paraffin sections using a Nikon Eclipse E600 microscope and digital image was taken using a Nikon DXM1200F camera (Nikon, Surrey, U.K.). The Nikon ACT-1 version 2.62 software was used for image acquisition.
Cells were fixed in methanol/acetone, blocked [20% (v/v) goat serum in 1% BSA/PBS] and incubated with either monoclonal DMT-1 [100 μg/ml NRAMP24-A (recognizing IRE +ve/−ve isoforms); ADI], IREG-1 (100 μg/ml MTP11-A; ADI) or ferritin H (1:50; Abcam) antibodies for 1 h before labelling with FITC goat anti-mouse (1:500; Jackson Immunoresearch Laboratories, West Grove, PA, U.S.A.). Cells were washed and incubated in DAPI (4,6-diamidino-2-phenylindole; 1:10000) for 1 min before visualization. Omission of primary antibody was employed as a negative control. Images were visualized using an Olympus BX40 microscope and digital image was taken using a Sensys Photometrics camera (Middlesex, U.K.). Desksoft Smart-Capture 2 software was used for image acquisition (Desksoft, www.desksoft.com).
Prussian Blue staining of ex vivo small bowel biopsies
Paraffin sections were dewaxed and incubated in 1% HCl containing 1% ferrous cyanate for 20 min. Sections were washed and counterstained with neutral Red before visualization. Images were visualized from paraffin sections using a Nikon Eclipse E600 microscope and digital image was taken using a Nikon DXM1200F camera. Nikon ACT-1 version 2.62 software was used for image acquisition.
Iron uptake studies
Caco-2 cells (1×104) were seeded into the apical chamber of a transwell plate (Costar, High Wycombe, Bucks., U.K.) and cultured in 1.0 ml of growth media. Upon reaching 14 days post-confluency, the medium from the apical chamber was removed and cells were briefly washed in PBS several times. Cells were then cultured in 1.0 ml of growth media with or without TNFα (5 ng/ml) for up to 24 h. In all experiments, the basal chamber contained 1.5 ml of growth medium alone. The medium was then removed and the cells were washed with PBS followed by culture in 1 ml of growth medium containing 1 μM iron as 55FeCl3 (37 kBq/ml; Polatom Radioisotope Centre, Otwock-Swierk, Poland) and 56FeSO4 in a 1:10 molar ratio, with a 100-fold molar excess of sodium ascorbate, and 20 mM 2-(N-morpholino) ethanesulphonic acid-buffered salt solution (pH 5.5). After 1 h culture, aliquots were removed from the basal and apical chambers. In addition, the cell monolayer was lysed in 1% Triton X-100/0.1 M NaOH. All aliquots were then subject to scintillation counting. Iron import was calculated as the total counts in both the cell layer and basal chamber. Iron export was calculated as the counts within the basal chamber alone. Each experiment was performed in triplicate at each time point.
Caco-2 monolayer permeability studies
Caco-2 cells (1×104) were cultured in the apical chamber of a transwell plate as described for the iron uptake studies above. Upon reaching 14 days post-confluency, the medium from the apical chamber was removed and cells were briefly washed in PBS several times. Cells were then cultured in 1.0 ml of growth media containing FITC-labelled dextran (200 μg/ml; Sigma–Aldrich) in the presence or absence of TNFα (5 ng/ml) for 24 h. The media from the apical and basal chambers were then assessed for fluorescence intensity (excitation wavelength=495 nm and emission wavelength=525 nm), using a fluorescence spectrophotometer F-4500. Control experiments included (i) a standard curve for fluorescence detection and (ii) studies to determine whether FITC-labelled dextran was able to cross the membrane of the transwell plate in the absence of cells. Each experiment was performed in triplicate.
All experimental errors are shown as 1 S.E.M. Statistical significance was calculated by unpaired Student's t test using SPSS version 10.0 (SPSS, San Rafael, CA, U.S.A.). Significance was accepted at P<0.05.
The present study has been carried out in accordance with the declaration of Helsinki (2000) of the World Medical Association. Ethical approval for this study was obtained from the Sandwell and West Birmingham LREC (Local Research Ethics Committee; no. 03/05/633). All patients have provided informed written consent.
Effects of TNFα on iron uptake in Caco-2 cells
To investigate the effect of TNFα on iron uptake, Caco-2 cells were challenged with TNFα for up to 24 h (Figure 1). Treatment with TNFα resulted in a significant early induction in both iron import (Figure 1A) and iron export (Figure 1B) at 6 h (P<0.05). However, by 24 h of culture with TNFα, a significant repression in both iron import and export was observed (P<0.05). At 24 h, the ratio of iron export to import is also significantly reduced in TNFα-treated cells, indicating that more iron is being incorporated within the cell, probably bound to ferritin (Figure 1C). To demonstrate that these changes in iron import/export were not a TNFα-mediated effect on monolayer integrity, a non-transported marker, FITC-labelled dextran, was utilized. In these experiments, it was shown that regardless of the presence of TNFα, less than 0.1% of the starting FITC–dextran was detected in the basal chamber and the remaining 99.9% remained in the apical chambers (control, apical chamber, 99.9±0.01%; basal chamber, 0.06±0.02%; TNFα-treated, apical chamber, 99.9±0.01% and basal chamber, 0.05±0.006%). There was no statistical difference in the level of fluorescence in the basal chambers of TNFα-treated or control (P>0.12). In addition, when FITC-labelled dextran was placed into the apical chamber of a transwell plate (in the absence of cells), it was demonstrated that the FITC–dextran was able to cross the membrane and, by 24 h, there was equilibration between the apical and basal chambers.
Iron import and export in Caco-2 cells in response to TNFα
The TNFα-mediated effects on iron transport in Caco-2 cells is hepcidin-independent
To determine if the TNFα-mediated effects on iron transport were a consequence of hepcidin, hepcidin mRNA expression was determined in both TNFα-treated and control Caco-2 cells (n=6) (Figure 2). Figure 2 demonstrates that, whereas human liver (n=6) has an abundance of hepcidin mRNA expression, Caco-2 cells, both control and TNFα-treated, failed to demonstrate any significant mRNA expression even after 40 cycles of amplification.
Hepcidin mRNA expression in human liver and Caco-2 cells
The effect of TNFα on DMT-1, IREG-1 and ferritin H expression in Caco-2 cells
To investigate whether the TNFα-induced changes in iron import/export were mediated by changes in the expression of iron transporter proteins, we determined the mRNA and protein levels of DMT-1, IREG-1 and ferritin H in Caco-2 cells.
Stimulating Caco-2 cells with TNFα resulted in changes in DMT-1, IREG-1 and ferritin mRNA expression (Figure 3). Since four DMT-1 splice variants (exon 1A, exon 1B, IRE-containing and non-IRE-containing) have been described, we examined the effects of TNFα on these spliced variants initially (Figures 3A–3D). Our results demonstrate that there was a significant elevation at 1 h (P<0.05) in only the DMT-1 IRE-containing and DMT-1 exon 1A variants (Figures 3A and 3C) with no changes observed for either the exon 1B-containing or the non-IRE-containing variants (Figures 3B and 3D). However, these elevations were only short-lived and by 8 and 24 h the message levels had returned to baseline. Similarly, IREG-1 and ferritin mRNA were also elevated following TNFα stimulation with significant elevations at 1 and 8 h respectively, with both the transcripts returning to baseline by 24 h (Figures 3E and 3F).
Transcriptional modulation of DMT-1, IREG-1 and ferritin H by TNFα
To ensure specificity of TNFα treatment, we examined E-cadherin expression as previously reported  and observed a dramatic repression in E-cadherin mRNA over a 24 h time period (results not shown).
To confirm these observations, Western blotting was performed to examine protein levels (Figure 4). Exposure of Caco-2 cells to TNFα resulted in a significant increase in DMT-1 by 8 h (P<0.05), with protein levels returning to baseline by 24 h (Figure 4A). IREG-1 protein was significantly induced at 3 h (P<0.05), with levels returning to background at 8 and 24 h (Figure 4B). Additionally, ferritin H protein expression was significantly induced at 24 h only (P<0.05), supporting our mRNA observations (Figure 4C).
Modulation of DMT-1, IREG-1 and ferritin H protein expression by TNFα
The effect of TNFα on DMT-1, IREG-1 and ferritin H expression in ex vivo small bowel cultures
We sought to verify our results obtained in Caco-2 cells by using an ex vivo small bowel culture model system (Figure 5). Consistent with our Caco-2 cell findings, challenging small bowel cultures (n=40) with TNFα resulted in a significant elevation in the DMT-1 IRE-containing and the DMT-1 exon 1A variants (P<0.05), with no significant change in either DMT-1 IRE −ve or the DMT-1 exon 1B variants (Figures 5A–5D). Similarly, IREG-1 and ferritin H mRNA were also both elevated following 1 h culture with TNFα compared with untreated small bowel cultures (Figures 5E and 5F).
Modulation of DMT-1, IREG-1 and ferritin H mRNA by TNFα in ex vivo cultures
Modulation of DMT-1, IREG-1 and ferritin H protein by TNFα in ex vivo cultures
The effect of TNFα on DMT-1, IREG-1 and ferritin H localization in Caco-2 cells
Results from the iron uptake assay (Figures 1A and 1B) clearly demonstrate that at 24 h there was a marked repression in iron import and export. This is associated with a decrease in DMT-1 and IREG-1 expression to baseline levels (Figures 3 and 4). However, the repression of iron import and export below baseline levels suggests an additional aberration of function of these proteins. One possible mechanism for this aberration of function might be explained by a possible redistribution of these proteins. To investigate whether this was a possible explanation, immunolocalization for DMT-1, IREG-1 and ferritin H was performed on Caco-2 cells and ex vivo small bowel cell cultures. Challenging Caco-2 cells for 24 h with TNFα failed to cause any relocalization of DMT-1 (Figures 7A–7C). However, IREG-1 was lost from the cell borders with immunoreactivity being predominantly weak and cytoplasmic (Figures 7D–7F). Ferritin H was also strongly relocalized to the perinuclear compartment (Figures 7G–7I).
Localization of DMT-1, IREG-1 and ferritin H in Caco-2 cells stimulated with TNFα
The effect of TNFα on DMT-1, IREG-1 and ferritin H localization in ex vivo small bowel cultures
Immunolocalization studies were performed in ex vivo small bowel cultures (n=10) treated for 6 h with or without TNFα. Ex vivo small bowel cultures challenged with TNFα did not cause a relocalization in DMT-1 expression. DMT-1 remained localized in the apical surface of villous enterocytes (Figures 8A–8C). However, culturing in the presence of TNFα did result in a dramatic relocalization of IREG-1 (Figures 8D–8F). In untreated cultures, IREG-1 was localized predominantly on the basolateral border of crypt enterocytes. However, exposure to TNFα caused a loss of this discrete basolateral border immunoreactivity with only diffuse cytoplasmic immunoreactivity remaining. In untreated cultures, ferritin H was weakly localized within the cytoplasm, whereas culture with TNFα resulted in strong immunoreactivity in the apical cytoplasmic pole of the villous enterocytes (Figures 8G–8I). To examine whether the increase in ferritin H expression observed with TNFα treatment leads to an increase in enterocyte iron loading, we stained ex vivo small bowel sections with Prussian Blue (Figures 8J and 8K). In the untreated cultures, no Prussian Blue staining was discernible; however, in treated cultures, Prussian Blue staining was clearly observed in the apical poles of the villous enterocytes in agreement with the pattern of immunoreactivity observed for ferritin H.
Localization of DMT-1, IREG-1 and ferritin H in TNFα-challenged ex vivo cultures
Previous studies have suggested a role for TNFα in the modulation of iron transport proteins. Torti et al.  and Ludwiczek et al.  have demonstrated that ferritin and DMT-1 respectively can be induced by TNFα. Furthermore, murine studies have suggested a physiological role for TNFα in regulating small bowel iron absorption by the effect of dietary iron on small bowel IELs (intraepithelial lymphocytes) [25,26]. In the present study, we have demonstrated that, in a hepcidin-free environment, exposure to TNFα is associated with a rapid and dynamic modulation of iron transporter expression in both an established cell line model and an ex vivo small bowel culture system. Importantly, this modulation was associated with a coincident effect on iron transport.
Culturing Caco-2 cells in the presence of TNFα led to an initial induction in iron import and export, whereas prolonged exposure (24 h) led to a dramatic repression in both. The net effect of prolonged exposure was cellular iron accumulation. To explain these observations, we examined mRNA and protein levels of DMT-1, IREG-1 and ferritin H in both Caco-2 cells and in ex vivo small bowel cultures after TNFα treatment.
Since, in total, four DMT1 splice variants have been described so far , we initially examined what effect TNFα had on these DMT-1 splice variants in the Caco-2 cell line. Utilizing probes to the 5′ variants (exon 1A or 1B) and 3′-untranslated region variants (IRE +ve or IRE −ve), we have been able to demonstrate that responsiveness to TNFα was exhibited by the exon 1A (IRE +ve) splice variant of DMT-1. This variant is also believed to be the main iron-responsive form of DMT-1 in the small bowel . Little or no response to TNFα was observed with the exon 1B and IRE −ve splice variants.
In Caco-2 cells, the elevation in DMT-1 and IREG-1 mRNA was maximal at 3 h, whereas maximal ferritin H expression occurred at 8 h. Interestingly, TNFα-mediated increase in DMT-1 and IREG-1 mRNA in Caco-2 cells was short-lived, transcripts being restored to baseline after 8 h culture. In contrast, expression of ferritin H mRNA decreased to baseline at 24 h. These changes in mRNA level were accompanied by similar changes in protein expression for all three proteins.
In the ex vivo small bowel culture system, exposure to TNFα resulted in a similar pattern of mRNA expression. However, the time course was significantly different compared with the Caco-2 cell model with all the three transcripts [DMT-1 exon 1A (IRE +ve), IREG-1 and ferritin] significantly increased at 1 h. In addition, by 6 h, there was no change in DMT-1 or IREG-1 protein expression but there was a significant induction in ferritin H. This suggests that, at this later time point, the induction in DMT-1 and IREG-1 had occurred and returned to baseline, whereas ferritin H induction, which in the cell model was a later event, was up-regulated as anticipated. Alternatively, it could be that, in these ex vivo cultures, 6 h was too short a time period to observe changes in DMT-1 and IREG-1 protein expression. The variation in the time course of these events probably reflects inherent differences between the two model systems utilized in the present study. It appears likely that the ex vivo system is a more representative model of the in vivo scenario due to conservation of cell function and a villous architecture. Indeed, we believe we have uniquely demonstrated that this is a valid model for the future study of iron transporter regulation.
The biological relevance of the early induction in DMT-1 and IREG-1 is unclear. Both promoter sequences contain putative TNFα response elements and the rapidity of their induction suggests a direct effect on the promoter of these transporters. Alternatively, a TNFα-mediated inhibition of mRNA degradation might also be a possible cause of this early increase in mRNA expression. The later repression of both proteins may reflect a second competing pathway, which becomes dominant at later time points. This hypothesis is supported by results from a recent study that demonstrated a repression of DMT-1 in Caco-2 cells after 72 h TNFα exposure, an effect that appeared to be promoter-independent . However, in this study, the DMT-1 exon 1B variant was utilized, which we have demonstrated is not TNFα-responsive for shorter time periods.
Although, in our study, it appears that the IRE-containing form of DMT-1 is up-regulated in response to TNFα, the changes in expression are unlikely to be modulated by changes in IRP (iron-responsive protein) function. The IRE for DMT-1 is found on the 3′-end of its transcript, whereas the transcript of IREG-1 contains a 5′-IRE. As such, TNFα-modulated DMT-1 and IREG-1 expression would have been reciprocal if it were IRP/IRE-dependent. Clearly, further studies are required to delineate the mechanisms underlying the early mRNA increases and the opposing effects on the temporal expression of both DMT-1 and IREG-1.
The dynamic changes in protein expression correlate well with the effects seen on iron transport in Caco-2 cells. The early de novo induction in DMT-1 and IREG-1 resulted in early induction in iron import and export. However, the repression in iron import and export observed at 24 h was over and above that which could be explained by the decrease in DMT-1 and IREG-1 protein expression. It is likely that the relocalization of IREG-1 away from the basolateral membrane combined with increased ferritin H expression allows for decreased iron export and increased accumulation within enterocytes. Indeed Prussian Blue staining of ex vivo cultures after exposure to TNFα demonstrated iron deposition within the apical portion of the villous enterocytes in a similar pattern as that observed for ferritin H staining.
Thus we believe that, in our hepcidin-free system, enterocyte iron trapping and ultimate repression of iron transport is TNFα-driven and that TNFα is potentially mediating its effect partially through IREG-1 relocalization (rather than changes in IREG-1 protein expression). In this regard, it is with great interest that a recent elegant study by Nemeth et al.  has demonstrated using tissue culture models that hepcidin can bind IREG-1 and this causes internalization and subsequent degradation of IREG-1, leading to decreased export of cellular iron. This raises the possibility that cytokines and hepcidin can regulate enterocyte iron transport through a common protein, IREG-1.
Recent murine studies have suggested a possible physiological role for TNFα in regulating small bowel iron absorption. Small bowel IELs release TNFα in response to dietary iron . Blocking the effects of TNFα by utilizing a TNF receptor-2 knockout mouse led to splenic iron overload . In the same study, utilizing an HFE (mutated in hereditary haemochromatosis) knockout mouse led to hepatic iron overload and also removed the ability of IELs to produce TNFα on dietary iron challenge.
Moreover, there is a growing body of evidence suggesting a role for TNFα in the anaemia of chronic inflammation. In a murine study, recombinant TNFα induced serum iron deficiency by iron sequestration within the reticuloendothelial system . In addition, TNFα has been shown to directly inhibit erythropoeisis in patients with active rheumatoid arthritis, an effect that was reversed by anti-TNFα therapy [30,31]. Neither study characterized changes in small bowel iron transporter expression as a consequence of systemic inflammation.
We believe that, in the present study, we have been able to delineate the local effects of TNFα from the systemic effects of hepcidin. Our hypothesis is that, in normal physiology, local TNFα release, probably by IELs, is able to modulate a rapid induction in iron transport, which is then repressed with chronic exposure to TNFα, causing a block on iron absorption. However, why there should be this early induction still remains unexplained. In the anaemia of chronic inflammation, it is likely that there is a local and systemic TNFα effect in addition to hepcidin, both culminating in a block on iron absorption.
The effects of the pleiotropic agent TNFα on iron transport in vivo is likely to be complex with the interplay of multiple signal-transduction pathways and other effectors of iron metabolism. Clearly, it would be enlightening to extend this work to a whole organism to characterize the interplay between the small bowel, TNFα and these other known regulators of iron metabolism, namely hepcidin.
In summary, utilizing two model systems, we demonstrate that TNFα has a dynamic effect on iron transporter expression in enterocytes, resulting ultimately in a block in iron transport and increased iron storage. It is likely that, in chronic inflammation, an increase in systemic TNFα would cause a more sustained block on iron absorption, leading ultimately to anaemia. Thus it is interesting to speculate on the future of anti-TNFα therapy in the treatment of the anaemia of chronic inflammation.
This work was supported by the Biotechnology and Biological Sciences Research Council (project number BBS/B/10781) and City Hospital Research and Development Fund.
These authors have contributed equally to this work.