Schliemann et al., (2011). Heterogeneity Reduces Sensitivity of Cell Death for TNF-Stimuli.

July 2012, model of the month by Nick Juty
Original model: BIOMD0000000407


Apoptosis, the process of programmed cell death (PCD) in multicellular organisms, is an essential homeostatic mechanism; it's downregulation can result in uncontrolled growth resulting in cancers, while upregulation has been associated with neurodegenerative conditions such as Parkinson's and Alzheimer's diseases and atrophy. The regulatory network responsible for this control consists of intrinsic and extrinsic signaling pathways, counterbalanced by anti-apoptotic signals.

The network (Figure 1) considered in this work [1] is centred on Tumor Necrosis Factor (TNF), which activates the TNF receptor type 1 (TNF-R1). This induces cell proliferation through NF-kB, but also activates caspases to trigger PCD. The experimental component of this study was performed in KYM-1 cells, which were selected for their rapid development of apoptotic indicators.

To model the cell-to-cell variability that is often displayed in populations of cells, the modeling framework additionally used a 'cell ensemble' approach. This comprises individual cell models that accommodate variant parameter sets, which are then analyzed collectively.

Figure 2

Figure 2 TNF dose response of KYM-1 cells. KYM-1 cells were maintained in culture conditions with 1ng/ml TNF. Caspase-3 activity (apoptosis marker) was measured 4 to 8 hours following TNF addition (solid lines). Control cultures (magenta) were also measured alongside cultures containing a caspase inhibitor (dashed lines). (figure taken from [1]).

Figure 3

Figure 3 Simulation of single cell model reaction network. Simulations were performed to ascertain relative NF-kB nuclear translocation, and caspase-3 activation by TNF, over a range of concentrations from 0.001 to 100ng/ml. Pulsed stimulations were over 30min intervals. The relative activity for NF-kB is colour-coded, where blue represents low, yellow represents medium, and red represents high nuclear accumulation. For caspase-3, red trajectory represents early death, green indicates cell survival, and yellow indicates delayed death. (figure taken from [1]).

Since the experimental data represented pooled cells, an ensemble model was used to investigate further. The sensitivity to the time of death with respect to specific parameters, within individual cells, was studied. These parameters were the production rate of TNF receptor, IkBa mRNA expression, and XIAP production. A global sensitivity measure was defined using the slope of the linear approximation of time of death versus each parameter on a log-log plot (Figure 4).

Since TNFR displayed a negative slope, one concludes that it correlates with earlier cell death. XIAP, on the other hand, showed a positive correlation with delaying onset of cell death. The IkBa parameter showed no observable trend.

Figure 5

Figure 3 Global sensitivity analysis of single cell and ensemble models.Dots indicate cell population sensitivities to listed parameters, and stars indicate single cell model values of global sensitivity. The colour coding indicates slope value, where increasing 'redness' indicates sensitivity magnitude. Both continuous (A) and pulsed (B) simulations are shown. (figure taken from [1]).

Figure 1

Figure 1 Reaction network initiated by TNF-R1. The reaction network is composed of 3 components; the TNF-R1 receptor complex (green region), the NF-kB anti-apoptotic pathway (blue region), and the caspase activated apoptotic pathway (red region). (figure taken from [1]).

Experimental work identified that low levels of TNF (below 0.1ng/ml) did not induce apoptosis. High concentrations (above 30ng/ml) led to rapid and high levels of cell death. At intermediate levels (1-10ng/ml), a difference was observed between pulsing (30 mins exposure followed by cell washing to remove TNF) and continuous exposure; only continuous exposure resulted in apoptotic responses (Figure 2).

A single cell model was used to model this process by performing time course simulations over 16 hours following TNF addition (Figure 3), using pulsed and continuous stimulation conditions. Rapid NF-kB activation was seen in continuous conditions, converging to a higher activation level than observed in pulsed conditions. Under both conditions, the onset of caspase activation was observed to occur with a time delay of around 5 hours, being significantly later in pulsed conditions.

Figure 4

Figure 4 Ensemble model sensitivity analysis. Time of death for 2500 cells with lognormal distribution of production rates, plotted against normalised value for each rate. Cells surviving 100 hours post stimulus (TNF) were classified as survivors (magenta/cyan) in the “>100” domain. Results are shown for continuous stimulus conditions for all 3 parameters (A, C, E) using 10ng/ml TNF. The pulsed conditions are not shown here but were similar in trend). (figure taken from [1]).

Nineteen parameters were studied in this manner, each time comparing the effects on a single cell and the ensemble model. Significant differences were observed, particularly in pulsed experiments (Figure 5, TNFR, RIP, TRAF2, Procaspase-8 and BAR). These effects can be generalised into three groups; those where increasing the value of the parameter delayed cell death, whose members include caspase inhibitors, XIAP and BAR; those where increasing the parameter value resulted in earlier cell death, whose members include TNF receptor, RIP and TRAF2 and Procaspases-3 and -8. The final group is composed of those parameters whose value seems uncorrelated with time until cell death.

Global sensitivity analysis showed that BAR and XIAP had the greatest positive correlation with time of death, delaying its onset significantly. The most sensitive parameters were found to be TNFR, procaspases-8 and -3, TRAF2 and RIP, which advanced cell death.

Cell-to-cell variability within a clonal population is essential in many signaling processes, and is due to a particular cell's epigenetic state. As a consequence, the PCD of a total population can be staggered over a period of hours. Even then there may be some cells that are resistant to the stimulus, with the level of this resistance being dependent upon the stimulus duration and extent (dose). This heterogeneity and apparent resistance to TNF stimuli is well described by cell ensemble models, but not by single cell models. This highlights the need to perform computationally more complex sensitivity analyses on heterogeneous cell populations, rather than on single cell models.

Bibliographic References

  1. Schliemann et al. Heterogeneity reduces sensitivity of cell death for TNF-stimuli. BMC Syst Biol. 2011; Dec 28;5:204 [PMID:22204418]