Death ligand mediated apoptotic activation is a mode of cell death that is widely used in cellular and physiological situations. Interest in studying death ligand induced apoptosis has increased due to the promising role of recombinant soluble forms of death ligands (mainly recombinant TRAIL) in anti-cancer therapy. A clear elucidation of how death ligands activate the type 1 and type 2 apoptotic pathways in healthy and cancer cells may help develop better chemotherapeutic strategies. In this work, we use kinetic Monte Carlo simulations to address the problem of type 1/ type 2 choice in death ligand mediated apoptosis of cancer cells. Our study provides insights into the activation of membrane proximal death module that results from complex interplay between death and decoy receptors. Relative abundance of death and decoy receptors was shown to be a key parameter for activation of the initiator caspases in the membrane module. Increased concentration of death ligands frequently increased the type 1 activation fraction in cancer cells, and, in certain cases changed the signaling phenotype from type 2 to type 1. Results of this study also indicate that inherent differences between cancer and healthy cells, such as in the membrane module, may allow robust activation of cancer cell apoptosis by death ligand induction. At the same time, large cell-to-cell variability through the type 2 pathway was shown to provide protection for healthy cells. Such elucidation of selective activation of apoptosis in cancer cells addresses a key question in cancer biology and cancer therapy.
Death ligand induced apoptosis is a frequently used mode of cell death activation and is also important for its potential application in cancer therapy. Death ligands are known to engage their specific receptors (death receptors) and can activate type 1 and type 2 apoptotic pathways. Tumor necrosis factor related apoptosis inducing ligand (TRAIL), or recombinant forms of that ligand, has particularly attracted recent attention due to its role as a promising anti-cancer agent [
Death receptor activation, typically by death ligand engagement, is the mode of apoptotic activation in a wide variety of cellular and physiological situations. Death ligand induced apoptosis has been implicated in the death of strongly self-reactive immature lymphocytes (also for peripheral tolerance) [
In our previous studies, we addressed the problem of understanding systems level regulation of type 1/type 2 choice in apoptotic activation [
Experimental studies including genomic analysis of cancer cells have indicated that the genome and proteome of cancer cells are markedly different from that in healthy cells. Anti-apoptotic proteins are frequently over-ex- pressed in cancer cells making them particularly resistant to apoptotic activation. How it is then possible to activate the apoptotic pathways selectively in cancer cells in such a manner that healthy cells remain mostly protected, remains a challenging question that needs to be addressed before successful anti-cancer therapies can be developed based on apoptotic induction. In our earlier studies, we addressed the mechanisms for selective killing of cancer cells such as by inhibition of Bcl-2 family anti-apoptotic proteins [
In this work, we use Kinetic Monte Carlo simulations to elucidate the single cell mechanisms of the type 1/ type 2 choice in cancer cells in silico. Results obtained from this study indicate that over-expression of certain anti-apoptotic proteins may make cancer cells prone to either type 1 or type 2 activation. However, in most cases, increased death ligand induction leads to an increase in type 1 activation, as also observed in our studies for healthy cells [
We used Monte Carlo simulations to carry out computational study of death ligand induced apoptosis in cancer cells. A simplified network model of apoptotic cell death signaling, which incorporates key regulators in both type 1 and type 2 pathways (
In our computational model, induction of death ligands (such as FasL, TRAIL) activates the initiator caspases (caspase 8) in the membrane proximal death module and thus acts as a trigger for apoptotic activation (
tive free energy parameter E, which could capture the effective attraction (reduction in energy) when two receptors occupied neighboring nodes. For free (unbound) death receptors (and decoy receptors) the energy parameter was assumed to be small ~−KB∙T to prevent spontaneous pre-clustering of receptor molecules. However, pre- clustering in lipid raft domains has been observed for DCR1 due to their GPI anchor [
Active caspase 8 molecules engage both type 1 and type 2 pathways. In the type 1 pathway, caspase 8 directly activates the effector caspases (caspase 3/7). In the type 2 pathway, a second initiator caspase (caspase 9) is required for effector caspase activation and signal amplification. Initial routing of the activation signal through type 1/type 2 pathways is governed by relative affinity of caspase 8 for pro-caspase 3 (type 1 pathway) and Bid (type 2 pathway) [
Earlier studies indicated that the type 1/type 2 choice is death receptor specific [
Even though we simulate some major components of the apoptotic death regulatory pathway that are essential for addressing the type 1/type 2 choice in cancer cells, it has several simplifying assumptions. In our model, functionally similar molecules (pro- or anti-apoptotic) are coarse-grained by a representative protein. Bcl-2 (B cell lymphoma protein 2), for example, represents all the Bcl-2 family proteins (such as Bcl-2, Bcl-xL, Mcl-1) with similar anti-apoptotic properties. The present model does not consider the following: 1) synthesis and degradation of signaling molecules [
We developed a Hybrid Monte Carlo simulation that combines the following two approaches: 1) a probabilistic rate constant based (implicit free energy) Kinetic Monte Carlo simulation for various reaction moves such as diffusion, binding/unbinding and catalytic cleavage; 2) an explicit free-energy based Kinetic Monte Carlo Model that captures clustering of receptor molecules utilizing energy function based diffusion moves. At each Monte Carlo (MC) step, one molecule was sampled (on average) once to allow for either diffusion or a reaction move. Acceptance/Rejection of the moves was carried out based on either through the probability constants (in implicit energy MC) or by the Metropolis criteria (in explicit energy MC) and ensures detailed balance is satisfied.
A simulation volume of 1.2 ´ 1.2 ´ 1.2 mm3 (corresponding to a 60 ´ 60 ´ 60 lattice with lattice spacing Dx = 20 nm) is chosen in such a manner that the number of molecules (for each molecular species) is equal to the nanomolar concentration. Death and decoy receptors are placed on one surface of the simulation lattice where they engage death ligands placed on a surface parallel to death/decoy receptor surface. Cytochrome c/Smac is initially localized in an 18 ´ 18 ´ 18 sub-lattice within the intracellular volume. Each MC step (DT) is chosen to be 10−4 s based on known mobility of cytosolic molecules [
Each run of our Monte Carlo simulation corresponds to apoptotic activation in a single cell, thus, data obtained from Monte Carlo runs can capture cell-to-cell stochastic variability (including inherent variability) in signaling activation. Each run of the simulation corresponds to activation at a single cell level. Statistical analysis (average and standard deviation in cell death) was carried out using data obtained from many single cell runs (64 runs were used in current simulations). In some cases, additional statistical analysis was carried out to assess the probability of cell death (at the end of simulation runs) for both healthy and cancerous cells so that selectivity in cancer cell killing can be established. Samples of size 16 (sixteen single cell runs) were randomly chosen among 64 runs and 16 such samples were collected using bootstrapping [
Concentrations of apoptotic and anti-apoptotic molecules in the death signaling pathway are cell type specific. In cancer cells, such cell type specific variability is presumably much more pronounced as cancer cells found to exhibit genomic instability and aberrant expression of both pro- and anti-apoptotic molecules (in a cancer-type specific manner). Such variability may also exist among genetically heterogenous sub-populations within a given tumor cell [
When DL concentration is low (~2 molecules on 1.2 ´ 1.2 mm2 cell surface), very high XIAP expression (~3 fold or higher) suppresses type 2 apoptosis as XIAP strongly inhibits caspase 9 activation. Only weak type 1 activation was observed under such low DL concentrations. Increased DL concentration resulted in markedly increased apoptosis and also increased activation through the type 1 pathway (
ule. In contrast, when cancer cells had regular XIAP levels but decreased death adaptor proteins and increased cFLIP, death ligands induced weak type 2 activation (
In corresponding healthy cells (when over-expressions of pro- and anti-apoptotic molecules are removed), low level of death ligand induction mainly activated the type 2 pathway but the activation was switched to type 1 dominant mode with increasing ligand concentrations (
Genome and proteome of cancer cells are markedly different from that in healthy cells and such differences may allow selective targeting and elimination of cancer cells. In many cases, the inherent state of the membrane mod- ule (including the death receptor expression) in cancer cells is more susceptible to apoptotic activation than that in healthy cells. Recent experiments indicated that expression of death receptors DR4 and/or DR5 (TRAIL receptors) can be high in certain cancer cells [
nant glioma cells [
We considered two possible strategies for selectively activating the death receptor induced pathway in cancer cells. The type 2 pathway can be activated in cells for which DR expression is not very high (DR/DCR ~1), Bid and Bax type proteins are over-expressed to some extent, apoptotic inhibitors Bcl-2 and XIAP are either not highly expressed or those can be inhibited (such as by embelin alone [
type 2 pathway remains strongly resistant to activation (such as due to low Bid-Bax levels and high Bcl-2 like inhibitors) but large concentration of death receptors is either present or can be induced, activating the type 1 pathway turns out to be a better strategy. In either case, we observe selective killing of cancer cells while activation in healthy cells remain low (as assessed by average cell death for 64 single cell runs). Increased probability for cancer cell killing could be further established by additional statistical data analysis (see Methods). We present quantitative estimations of type 1 and type 2 activations (based on population averaged fraction of type 1/type 2 activation over 64 single cell runs; see Methods). We can also assign a type 1/type 2 activation phenotype to each cell that has undergone sufficient effector caspase activation. For a single cell, if the type 1 activation exceeds 50% of total, then the cell is called a type 1 cell (see Methods).
In certain cancer cells the adaptor protein expression can be higher or it can be induced by genetic mechanisms (such as done for neural cancer cells [
The ratio DR/DCR emerges as a key parameter that determines the extent of activation in the membrane proximal module as well as the extent of cell-to-cell stochastic variability in caspase 8 activation. In certain cases the type 1/type 2 choice at the level of single cells may differ from that estimated based on average over many cells. In our simulations for cancer cells equipped with high level of Apaf (type 2 susceptibility), type 1/type 2 choice estimated at the level of single cells (
Type of cell | DCR1 | DCR2 | ||
---|---|---|---|---|
Cells undergone apoptosis | Type 1 cell | Cells undergone apoptosis | Type 1 cell | |
Healthy cells | 13 | 3 (23%) | 12 | 6 (50%) |
Cancer cells with high Apaf | 43 | 2 (5%) | 51 | 3 (6%) |
Cancer cells with high DR | 55 | 33 (60%) | 48 | 30 (63%) |
We also considered the scenario where negligible amount of decoy receptors were present on cancer cells, as indicated by some earlier experimental studies [
We next considered the case when engineered TRAIL affinity for death receptors DR4/DR5 is low (Kd ~10 mM). The affinity of the death ligand for decoy receptors was assumed to be similar to that for death receptors. Such low affinity death ligands might remain protective even for normal cells that have significant expression levels of both death and decoy receptors (as well as other membrane-proximal apoptotic molecules). In
In
tion data (for both high and low affinity ligands) should be relevant for apoptosis induction in certain neural cancers as recent experimental data indicates selective activation of intrinsic apoptosis in glioblastoma cells by a combined treatment of TRAIL and embelin [
A quantitative scoring approach can be developed for choosing optimal death ligand concentration needed for selective killing of cancer cells. In this scoring approach quantitative assessment can be done for various key factors such as: 1) number of cells undergoing apoptotic activation in a given simulation time (highest score of 10 for 100% apoptosis); 2) average time-to-death (highest score of 5 for Td = 0 to lowest 0 when Td equals the maximum simulation time); 3) cell-to-cell variability in Td (the case with cell-to-cell variability equals to half- maximal simulation time is assigned a score of 0 and no cell-to-cell variability is assigned a maximum score of 5); 4) number of apoptosis in corresponding healthy cells (highest score of 10 for 0% apoptosis). When we applied this quantitative method for cancer cells equipped with high death receptor (DR = 100), DL = 5 - 10 turned out to be the optimal DL level (
We also studied death ligands have increased affinity for decoy receptors than death receptors to probe whether such ligands could selectively induce apoptosis in cancer cells. When expression of death receptors (DR4/DR5) is significant in healthy cells (cells that will be under the action of chemotherapeutic drugs), but still the DR/DCR ratio is larger in cancer cells compared to that in healthy cells, engineered ligands that target the
DL = 2 | DL = 5 | DL = 10 | |
---|---|---|---|
DCR1 | 17.9 | 20 | 20 |
DCR2 | 18.7 | 21.4 | 20.6 |
death receptors with lower than usual affinity (~nanomolar) might allow minimal cytotoxicity in normal cells. The affinity for decoy receptors remains in the usual nanomolar range so that most ligands bind to decoy receptors on normal cells instead of death receptors. Even TRAIL (the natural death ligand) might bind with DCR2 with slightly higher effective affinity (than that for DR4/DR5) due to elongated and rigid extracellular structure of DCR2 [
In many cancer cells, over-expression of pro- and anti-apoptotic molecules may make them particularly prone to either type 1 or type 2 activation. For example, high over-expression level of XIAP makes the type 2 pathway particularly resistant. In contrast, low expression level of death receptors and high cFLIP level may preferentially activate the type 2 pathway in a cancer cell. However, increased concentration of DL increases the type 1 activation in most cases. Increase in type 1 activation results from increased caspase 8 activation that compensates for the lower binding affinity of caspase 8-procaspase 3 (than that of caspase 8-Bid). For some types of cancer, increased DL concentration will presumably allow a type 2 to type 1 transition (as also observed in our earlier studies for healthy cells [
the mechanisms of two different types of decoy receptors. For the case of DCR1, the decoy receptors are pre- clustered (possibly in lipid raft domains [
How to selectively target the apoptotic pathways in cancer cells, that would spare normal cells, has been a key question in cancer biology and cancer therapy. Here, results obtained from our computational studies elucidate mechanisms of selective killing of cancer cells under TRAIL therapy. We show that inherent differences in the DR/DCR ratio (death receptor to decoy receptor ratio) between cancer and healthy cells may provide a mechanism for TRAIL induced selective killing of cancer cells. The DR/DCR ratio modulates cell-to-cell variability in apoptotic activation generating crucial differences in apoptotic activation between cancer and healthy cells. It might also be possible to activate other death ligand mediated pathways for certain cancer types (such as Fas- FasL for brain tumors [
Selective targeting of cancer cells might be possible in various possible manners. In cancer cells equipped with high level of death receptors (TRAIL receptors DR4/DR5) and low level decoy receptors activating the type 1 apoptotic pathway might be suitable for selective killing. In some other cancer cells, with slightly defective membrane proximal signaling module (such as due to low level of death adaptor proteins) selective apoptotic activation through the type 2 pathway might be possible by death ligand induction. If Bcl-2 is also over- expressed in such cancer cells then selective activation can be achieved under combined action of TRAIL and Bcl-2 inhibitor. In yet another scenario, if cancer cells equipped with high concentrations of anti-apoptotic Bcl-2 proteins also have strongly over-expressed Bid and Bax (pro-apoptotic) molecules, then removal of Bcl-2 inhibition by inhibitory small molecules could be sufficient for selective activation of the cancer cell type 2 pathway. In silico approaches may help develop the optimal strategy for targeting the apoptotic pathway in specific cancer cells. Information regarding clonal heterogeneity in a tumor population [
S. R. thanks the National Institute of Health (NIH, USA) for providing equipment (computing cluster).