Mitchell and Mendes (2013). A computational model of liver iron metabolism.
June 2014, model of the month by Martina Fröhlich
Original model: BIOMD0000000498
Iron is an essential nutrient required for many processes, such as oxygen transport, respiration, DNA synthesis and detoxification. It is crucial that its uptake and export are regulated properly. Iron deficiency can result in anemia, whereas, in excess, for example due to excessive intestinal absorption of iron as a consequence of a mutation in the haemochromatosis gene, it can lead to highly toxic free radicals and to a disease.
The main elements of the core iron regulatory cycle known from text book knowledge are:
- ferritin (FT), which is used to store iron within the cells and shield them from the surrounding environment,
- transferrin (Tf), to transport iron trough the body via the blood stream,
- the transferrin receptor (TfR1 and TfR2), to import iron into the cells,
- ferroportin (Fpn), to export iron from the cells, and
- iron response proteins (IRP), which signal the amount of iron in the cells and dictate whether more FT for iron storage (in case of iron overload) or more TfR (in case of iron iron deficiency) has to be produced. More information about the post-transcriptional control of FT and TfR by the IRP aconitase can be found in .
Mitchell and Mendes [2, BIOMD0000000498] propose a mathematical model (see Figure 1) consisting of ordinary differential equations (ODEs) that contains the above mentioned regulatory proteins. In addition, the model includes the hormone hepcidin, the haemochromatosis protein (HFE), haeme oxigenase as well as intercellular and extracellular haeme. The model consists of two compartments, which are the plasma and hepatocyte. The model's initial conditions were taken from the literature (see Table 2 in ) or set to zero (if generated through complex binding).
Complex formation reactions were modelled by using mass action kinetics, enzymatic or transport reactions using either Henri-Michaelis-Menten or Hill kinetics.The kinetic parameters were taken mostly from in vitro biochemical data and can be found in Table 1 of . All simulations and analyses included in the publication were performed with COPASI.
Local sensitivity analysis showed that the liable iron pool is mostly influenced by the synthesis and degradation of TfR2, TfR2 binding, as well as the iron export by Fpn. Within the global sensitivity analysis the effect of TfR2 degradation and its binding to iron might still be high, but this depends on the choice of the other model parameters. Instead, parameters such as hepcidin degradation or of the binding of the second iron to TfR2 could be of higher importance.
Sensitivity analysis of the effect of changes on the iron export showed similar results than the effect on the local iron pool regarding the iron import machinery, but a change in proteins and reactions affecting iron export have a reversed and minor effect.
Furthermore, the concentration of hepcidin is mostly influenced by the expression and degradation of hepcidin itself, as well as HFE expression and HFE-TfR2 degradation. Overall, changes in reactions affecting TfR2 concentration seem to have a higher effect than changes in reactions affecting TfR1 concentration.
Finally, by calculating the total receptor response for TfR1 and TfR2, it was shown that TfR1 is a poor sensor for high levels of intercellular iron. Although the linearity of its response is increased with increased receptor turnover, it is only linear over a very restricted Tf-Fe range. TfR2 on the other hand is more linear over the whole range analysed and thus can sensor changes even at high intercellular iron (see Figure 4 below as well as Figure 5 and 7 in .
The model was able to reproduce qualitative and quantitative biological findings. It was used to study the effect of iron overload on healthy and diseased cells. The effect of changes in the reaction rates on intracellular iron pools and iron export were analysed, suggesting the role of TfR2 (rather than TfR1) and HFE as relevant drug targets. Furthermore, it identified the importance of HFE and TfR2 as iron sensors.
Figure 1SBGN progress diagram of the human liver iron metabolism model. Figure taken from .
The first analyses presented have their emphasis on the model's steady state behaviour, which showed reasonable similarity with experimental data, even though the model was not forced to reproduce the data (see Table 3 of ).
Time course simulations were performed, in which an increase of the serum transferrin-bound iron led to an increased steady state value of the 2HFE-TfR2 complex, which directly regulates hepcidin. This results in an increase in the steady state hepcidin values. HFE-TfR on the other hand was decreasing (see Figure 2a and b). The authors conclude that this supports a role of HFE and TfR2 as sensors for cellular iron regulation.
Thereupon, an analysis of the Haemochromatosis phenotype was performed. In the model, this phenotype was represented via a 100-fold reduction in the HFE synthesis rate constant. The qualitative changes in Fpn expression between the wild type and the mutant were similar in the model and the experiments(see Figure 3). A drawback was here that the model does not include the systemic effects of a change in the hepcidin levels, as the intercellular concentration of transferrin-bound iron was fixed. Therefore, quantitatively accurate steady state levels of intracellular iron could not be reproduced.
Furthermore, a time course simulation was performed comparing the models for wild type and mutant. The effect of an oral dose of iron was analysed which resulted in a peak in hepcidin concentration after 4-8 hours. The levels of hepcidin during the time course were lower in the mutant than in the wild type (see Figure 2 of ).
Local and global sensitivity analysis was used to analyse the effect of a change in the reaction rates on the labile iron pool, the hepcidin concentration and the flux of iron out of the liver. Within the global sensitivity analysis optimization procedures were performed that maximized and minimized the control coefficients while the model parameters were changed within 10% of their value. This resulted in slightly different results, as some control coefficients could be high in some parameter settings, but low in others (the details about the method are described here ).
Figure 3Simulated Fpn expression rates at steady state for the WT and the HH mutant (reproduced with the model provided in the curation section in BIOMD0000000498, by addition of parameter scan for HFE expression between 2.3469e-13 and 2.3469e-11). (reproducing Figure 4 in ).
Figure 4Integral TfR binding. Shown are scans of the integral TfR1 binding (scaling factor 0 and 1 for low and high turnover, respectively) and integral TfR2 binding. The values show the integral from 0 to 500 h over a range of 0-20 µm Tf-Fe and are based on selected curves from Figure 5 and 7 in  using the respective file provided in the curation section from BIOMD0000000498.
- Alberts B, Johnson A, Lewis J, et al. Molecular Biology of the Cell. (2002). 4th edition. New York: Garland Science.
- Mitchell, S, Mendes, P. A computational model of Liver Iron Metabolism PLoS Comput Biol. 2013. Nov;9(11):e1003299.
- Sahle, S, Mendes, P, Hoops, S, Kummer, U. A new strategy for assessing sensitivities in biochemical models. Phil Trans R Soc A. 2008. 366: 3619–3631.