Aspirin’s effect on kinetic parameters of cells contributes to its role in reducing incidence of advanced colorectal adenomas, shown by a multiscale computational study

  1. Yifan Wang
  2. C Richard Boland
  3. Ajay Goel
  4. Dominik Wodarz
  5. Natalia L Komarova  Is a corresponding author
  1. Department of Mathematics, University of California Irvine, United States
  2. Department of Medicine, University of California San Diego School of Medicine, United States
  3. Department of Molecular Diagnostics and Experimental Therapeutics, Beckman Research Institute of City of Hope Comprehensive Cancer Center, United States
  4. Department of Population Health and Disease Prevention, University of California Irvine, United States
22 figures, 4 tables and 4 additional files

Figures

A schematic illustrating the mathematical model.

The six cell types are denoted by circles, the mutation rates that give rise to different types are marked by the straight arrows. Type 6 (late adenoma) is marked in red. Crypt conversion rates are …

Fitting the nonlinear model to late adenoma incidence data.

(a) The fitting error as a function of r1, for the best fitting pairs (γ3 and γ4) with KA = KR = 1000; see Figure 2—source data 1. (b) The fitting error as a function of r1, for the best fitting …

Pathways to adenoma.

(a) A schematic representation of the two pathways. (b) For the two groups of fits in Figure 2(b), the pairs (KA,KR) are shown. The two groups are characterized by KA >KR and KA <KR, respectively. …

The effect of aspirin on the advanced adenoma incidence curve.

(a) Modeling assumptions and references to figures that present the results. (b) A schematic showing the timing of the model with aspirin treatment. (c) Predicted advanced adenoma incidence in the …

Appendix 1—figure 1
Three pathways to adenoma.
Appendix 1—figure 2
Model fitting to the incidence data.

(a) The best fitting curves corresponding to increasing SC division rates, r1, are plotted together with the epidemiological data. (b) The best fitting parameters γ3 and γ4 are shown for each value …

Appendix 1—figure 3
Fitting the linear model.

Each panel shows fitting results for a particular parameter combination, (RAPC+/-,RKRAS); the values of these two fitness parameters are indicated, and FAPC+/+=2FAPC+/-. The green lines show the log10(fitting error) as a …

Appendix 1—figure 4
Fitting the linear system: for each parameter combination, (RAPC+/-,RKRAS), of Appendix 1—figure 3, the error landscape is shown that corresponds to the best fitting r1 (see the minima of the green lines in Appendix 1—figure 3).

The contour plot represents log10(fitting error) as a function of γ3 and γ4 (darker colors correspond to lower values). The rest of the parameters are as in Table Appendix 1—table 2.

Appendix 1—figure 5
Nonlinear model fitting to the incidence data (fixing the crypt carrying capacity, varying crypt fission rates).

(a) The best fitting curves corresponding to increasing SC division rates, r1, are plotted together with the epidemiological data; inset: the best fitting curves corresponding to the values of r1

Appendix 1—figure 6
Nonlinear model fitting to the incidence data (fixing crypt fission rates, varying the crypt carrying capacity).

Compared are the cases in the presence of a non-zero crypt death rate, δ=0.05 yrs-1 (a–c), and in the absence of crypt death rate, δ=0.0 yrs-1 (d-f). (a,d) The heatplot of the fitting error, where all …

Appendix 1—figure 7
Comparing deterministic solutions (system (LABEL:n1non-23, 12), solid lines) with the stochastic Gillespie simulations (symbols of the same color).

The three lines correspond to the different stopping conditions: (0) 1 crypt of type 6; (1) N1=102 crypts of type 6, (2) N2=105 crypts of type 6. The rest of the parameters are as in Appendix 1—figure 6a-c. …

Appendix 1—figure 8
Fitting the adenoma incidence curve to the nonlinear model with an explicit expansion phase.

Under a fixed KA=KR=1000, for each value of r1 the best fitting values of γ3 and γ4 were found. (a) The fitting error is shown as a function of r1 for the original model (blue line, same as in Appendix …

Appendix 1—figure 9
Fitting the adenoma incidence curve to the nonlinear model with an explicit expansion phase.

Under fixed crypt fission rates, for each value of r1 the best fitting value of KA=KR was found. (a) The fitting error is shown as a function of r1 for the original model (blue line, same as in Appendix…

Appendix 1—figure 10
Pathways to adenoma: Fitting the incidence curve to the nonlinear model with an explicit expansion phase.

Under fixed crypt fission rates, for each value of r1 the best fitting pair (KA,KR) was found. (a) The fitting error is shown as a function of r1; 3 groups of parameter combinations are marked by blue, …

Appendix 1—figure 11
Same as Appendix 1—figure 10, except expansion to 105 crypts is assumed.
Appendix 1—figure 12
Compartment dynamics for the best fitting models with KA=KR.

(a) The expected crypt numbers (using equations (LABEL:n1non-23)) for the 5 compartments are shown for the 2nd parameter set in table Appendix 1—figure 9(c) (expansion to 102 crypts). (b) For the …

Appendix 1—figure 13
Same as Appendix 1—figure 12(b-c), but without the restriction KA=KR.

(a) Parameter set #2 from Table Appendix 1—table 3 was used. (b) Parameter set #4 from Table Appendix 1—table 3 was used.

Appendix 1—figure 14
extends the results of main text Figure 4(a) and shows that relative adenoma risk predicted for each decade of treatment is very similar for all parameter sets (#1-4, Table Appendix 1—table 3).
Appendix 1—figure 15
Relative incidence of advanced adenoma, where (a) parameter set #1 (Table Appendix 1—table 3) and (b) parameter sets #2 is used.

In each panel, a comparison is presented between the case where aspirin affects crypt fission/death rates (yellow) and where it does not affect crypt fission/death rates (blue). Aspirin affects …

Appendix 1—figure 16
Predicted advanced adenoma incidence in the absence of aspirin treatment (thin black lines are the fitted curves and black dots are incidence data); under aspirin treatment where the drug affects types 2–6 (yellow lines), under aspirin treatment where the drug affects type 6 only (blue lines).

Each panel corresponds to aspirin treatment administered during one decade (20-30 years, 30–40 years, etc). The treatment period is shaded light green. It was assumed that type 6 crypts grow to 105, …

Appendix 1—figure 17
A comparison of predicted relative advanced adenoma risk under the assumption that cell types 2–6 (yellow bars) and only type 6 cells (blue bars) are affected by aspirin.

Aspirin is administered during different decades of patients’ life (as marked under each pair of bars), and different panels correspond to different durations of the follow-up period. (a) Parameter …

Appendix 1—figure 18
A comparison of predicted relative advanced adenoma risk under the assumption that aspirin treatment lasts 10 years (yellow) or 5 years (blue).

(a) Predicted advanced adenoma incidence in the absence of aspirin treatment (thin black lines are the fitted curves and black dots are incidence data); the incidence curve fo patients treated for …

Tables

Table 1
Aspirin doses in xenograft experiments, the equivalent human dose and the resulting changes in kinetic rates.
Dose in xeno- grafts (mg/kg)Equivalent dose in humans (325 cm pills per week)Fold difference in division rate, FrFold difference Un death rate, FdFitness factor
151.80.91.50.86
506.10.751.750.70
10012.20.52.00.45
Appendix 1—table 1
Enumeration of the different genotypes.
Mutations in APCMutations in KRASType number
001
102
203
014
115
216
Appendix 1—table 2
Parameters, notations, and their values.
ParameterNotationValue/Range
Number of cryptsNcrypt107
Number of SCs per cryptK7
Rate of inactivation of APC (per cell division)u10-7
Rate of inactivation of APC (per cell division)μ10-9
Division rate of WT SCs (per year)r1(18,365)
Relative fitness of APC +/- cellsF2=FAPC+/-1.6
Relative fitness of APC-/- cellsF3=FAPC-/-3.76
Relative fitness of KRAS+ cellsF4=FKRAS3.54
Division rate of APC-/- crypts (per year)γ30.2
Division rate of KRAS+ crypts (per year)γ40.07
Appendix 1—table 3
Parameters obtained from the fitting procedures in Appendix 1—figure 10 and Appendix 1—figure 11.
Parameter set #Ni (crypts)ΔTi (yrs)r1(yrs-1)KAKRComment
11024.79204100017best fit
21024.7919810391712nd best fit
310511.9730060710best fit
410511.97288631682nd best fit

Additional files

Download links