Defining Stem Cell Dynamics in Models of Intestinal Tumor Initiation

See allHide authors and affiliations

Science  22 Nov 2013:
Vol. 342, Issue 6161, pp. 995-998
DOI: 10.1126/science.1243148

Limiting Tumor Initiation

What is the competitive advantage of cells with frequently occurring mutations during tumor development? Vermeulen et al. (p. 995; see the Perspective by Bozic and Nowak) quantified the advantages of Apc-loss, Kras activation, and P53 mutation during tumor initiation in the mouse intestine. The mutations conferred only a limited clonal advantage. Indeed, many mutated stem cells were stochastically replaced by wild-type stem cells, helping to limit tumor initiation.


Cancer is a disease in which cells accumulate genetic aberrations that are believed to confer a clonal advantage over cells in the surrounding tissue. However, the quantitative benefit of frequently occurring mutations during tumor development remains unknown. We quantified the competitive advantage of Apc loss, Kras activation, and P53 mutations in the mouse intestine. Our findings indicate that the fate conferred by these mutations is not deterministic, and many mutated stem cells are replaced by wild-type stem cells after biased, but still stochastic events. Furthermore, P53 mutations display a condition-dependent advantage, and especially in colitis-affected intestines, clones harboring mutations in this gene are favored. Our work confirms the previously theoretical notion that the tissue architecture of the intestine suppresses the accumulation of mutated lineages.

Cancer development involves competition between normal and deviant cell lineages, ultimately resulting in disruptive tissue overgrowth (1). The cellular effects of mutations can increase proliferation or impair response to cell death–inducing signals (2). However, only very limited quantitative data exist on the net effects of oncogenic alterations at the cell population level. In the intestine, cancer arises from an initial transformation event occurring primarily, but not exclusively, in the stem cell compartment (3, 4). Because normal intestinal crypt homeostasis is characterized by competition between equipotent stem cells that continuously replace each other in a random fashion (Fig. 1A) (5, 6), oncogenic mutations may confer an advantage on the clone in which they arise by acting on these dynamics. We confirmed and used this assumption to quantify the competitive benefit of mutations frequently occurring in colorectal cancer (CRC).

Fig. 1 Quantifying the clonal benefit of KrasG12D.

(A) Intestinal stem cells are equipotent and continuously replace each other in a stochastic fashion. (B) Confocal images of SI crypt bottoms of AhCreER/tdTom–/fl mice (WT) and AhCreER/tdTom–/fl/Kras-G12Dfl (KrasG12D) at the indicated time points after clone induction. Clone sizes are indicated as fractions (in eighths) of the crypt circumference. Blue, nuclear stain (DAPI); red, tdTom expression; scale bars represent 30 μm. (C) Heat maps depict the relative frequency of clones of the indicated size (columns) at various time points (rows) for both WT and KrasG12D. (D and E) Graph displays the average size (D) or the percentage of fixed clones (E) of WT and KrasG12D clones at different time points after induction. Error bars indicate the SEM. [(C) to (E)] N = 5 mice for each group; for each time point and condition, >200 clones were analyzed. (F and G) Bayesian inference of the biased drift parameter PR for WT versus WT (F) and KrasG12D versus WT (G) clones. The vertical axis indicates the probability density; the horizontal axis indicates the PR values. The higher the probability density value, the more likely this PR value underlies the observed biased drift. PR = 0.5 indicates neutral drift. The cartoons summarize the replacement properties (with color coding as in the graphs).

To track the fate of wild-type (WT) and mutated cell lineages, we induced low-level intestinal recombination either specifically in the crypt base using Lgr5-EGFP-CreER mice or more generally in AhCreER mice, both crossed to the R26-Lox-STOP-Lox-tdTomato (tdTomfl/fl) reporter strain. Clones were visualized and quantified at the bottom of the crypt, allowing robust clone size quantification around the circumference of the crypt (fig. S1). We observed that on average, clones expand, and the number of fixed clones (i.e., crypts within which the whole epithelium is tdTom+) increases with time (Fig. 1, B to E). By quantitative analysis of the clone size distributions using the stochastic master equation and Bayesian inference, we confirmed earlier studies that continuous, one-dimensional neutral replacements govern intestinal stem cell dynamics (figs. S2 and S3) (5, 6). This process is fully defined by only two parameters: the number of functional stem cells per crypt (N) and the rate at which these replace each other (λ). Because of the high-quality clone size distribution data we obtained, we can directly infer both N and λ with considerable precision (fig. S2). For the proximal small intestine (SI), we find that N = 5 and λ = 0.1 replacements per stem cell per day (figs. S2 and S3). This inferred number of stem cells is considerably lower than the number of Lgr5+ cells (~16) per crypt (5) but agrees with a functional marker-free estimate of stem cell numbers (7). In fact, we directly confirmed a previous suggestion that a sizable fraction of Lgr5+ cells are actually more committed progenitor cells and do not function as stem cells in homeostasis (8) (fig. S4).

Next, we crossed AhCreER/tdTom-/fl mice with Kras-G12Dfl and Apcfl/fl mice. Recombination in these mice will result in activation of an oncogenic Kras variant or inactivation of one or both copies of the negative Wnt regulator Apc, in addition to tdTom expression. We confirmed that tdTom expression is tightly matched to oncogenic recombination events (fig. S5). Furthermore, our analysis was facilitated by the fact that the low-level clone induction of pre-neoplastic lineages does not substantially alter tissue morphology, as has been reported before (9, 10) (Fig. 1B and fig. S6). We found that activated KRAS confers an evident clonal advantage as the KrasG12D harboring clones expand and become fixed more rapidly than the WT lineages (Fig. 1, B to E). In order to express this effect quantitatively, we introduced an additional variable to the one-dimensional drift model to capture the biased drift properties: PR signifies the chance that a tdTom+ stem cell replaces its nonlabeled neighbor; conversely, 1-PR represents the chance that the nonlabeled stem cell replaces the tdTom+ neighboring stem cell. The neutral version of the model is described by unbiased stochastic replacements, and as expected, using the inference method in combination with the neutral clone size distribution data we find PR(WT versus WT) = 0.50 [0.48 to 0.52, 95% confidence interval (CI)] (Fig. 1F), indicating that two adjacent stem cells have equal probability to replace each other. Applying this inference strategy to the biased drift data of Kras mutant clones, we found a well-demarcated posterior probability peak: PR(KrasG12D versus WT) = 0.78 (0.75 to 0.81, 95% CI) (Fig. 1G). This implies that for replacement events occurring at the interface of a KrasG12D and a WT clone, the Kras mutant stem cell replaces the WT stem cell in ~80% of the cases, and conversely that the WT stem cell replaces the mutant stem cell in ~20% of the replacement events. Thus, interpreting the altered size distribution of clones in terms of individual stem cell fates provides an intuitive estimate of the potency of oncogenic mutations in the context of a stochastic model, by integrating the effects of proliferation rate, cell death frequencies, and self-renewal properties into a single parameter (PR).

Because human CRCs arise in the majority of cases after the inactivation of APC, we determined the clonal advantage of lineages harboring either heterozygous or homozygous inactivating mutations within this gene. Apc+/− lineages display only a limited, but significant, benefit over WT clones, whereas Apc−/− lineages have a more marked clonal benefit already at 4 days, making it unlikely that secondary effects of Apc loss [such as chromosomal instability (CIN)] are involved (Fig. 2, A to C). After applying the inference strategy to this data to determine PR, we found that PR(Apc+/− versus WT) = 0.62 (0.58 to 0.66, 95% CI) and PR(Apc−/− versus WT) = 0.79 (0.75 to 0.82, 95% CI) (Fig. 2, D and E). In reality, Apc−/− mutations usually occur in a clone of Apc+/− cells because of loss of heterozygosity (LOH) or an independent additional mutation (10); therefore, we also calculated the benefit of Apc−/− stem cells over Apc+/− stem cells: PR(Apc−/− versus Apc+/−) = 0.69 (0.65 to 0.72, 95% CI) (Fig. 2E). Both in the case of KrasG12D and in the case of the inactivation of Apc, the inferred values for PR result in clone size distributions in accordance with the measured data (fig. S7). A further confirmation of the validity of the model comes from the observation that tdTom+ clones induced in ApcMin mice, which are characterized by a heterozygous germline Apc mutation, display similar dynamics to neutral clones in a WT background [PR(ApcMin versus ApcMin) = 0.48 (0.46 to 0.51, 95% CI)], indicating neutral competition between Apc+/− lineages, although clones holding this mutation clearly have an advantage over WT clones (fig. S8).

Fig. 2 Quantifying the clonal benefit of Apc loss.

(A) Heat maps depict the relative frequency of clones of the indicated size (columns) at various time points (rows) for Apc+/− and Apc−/−. (B and C) Graph displays the average size (B) or the percentage of fixed clones (C) of WT, Apc+/−, and Apc−/− clones at different time points after induction. Error bars indicate the SEM. WT is as in Fig. 1. [(A) to (C)] N = 5 mice for each group; for each time point and condition, >200 clones were analyzed. (D and E) Bayesian inference of the biased drift parameter PR for Apc+/− versus WT (D) and Apc−/− versus WT (E).

Our findings reveal that although KrasG12D mutations or the inactivation of Apc result in a marked competitive advantage of the respective cell lineage, this does not mean that the clone will become deterministically fixed. Mutated stem cells are commonly replaced by WT stem cells, thereby evading the accumulation of further mutations. By using the inferred parameters of the biased drift model (N, λ, PR) and the adapted stochastic master equations, we calculated the proportion of mutant clones that reach fixation at a particular clone age (Fig. 3A). This analysis reveals that the majority of stem cells acquiring an Apc+/− mutation will be stochastically replaced by WT stem cells, and these mutant lineages will disappear from the tissue. In addition only ~55% of Apc−/− mutant stem cells within an Apc+/− background will reach fixation. These findings reveal that the accumulation of mutations in the intestinal tissue is an inefficient process because of the particular tissue architecture and the continuous stochastic replacements of stem cells. To illustrate the inefficiency of the process, we determined the contribution of different mutational trajectories in the formation of aberrant crypt foci (ACF); i.e., a crypt with a fixed Apc−/− clone (Fig. 3B). Only 22% of the ACFs result from two subsequent Apc hits (or LOH), and a large proportion of ACFs arise in crypts in which multiple Apc hits have taken place, the majority of which have been eliminated through the stochastic loss of mutant stem cells (Fig. 3B and fig. S9). Incorporating the inefficiency of fixation of the earliest events in CRC into a simple model describing the population incidence of this disease (11) highlights how these values could be used in future studies to link them to cancer frequencies in the human population (fig. S9).

Fig. 3 Mutational trajectories.

(A) Graph depicts the likelihood of the indicated clone types reaching fixation in a crypt over time. Shaded areas indicate 95% CI. (B) Cartoon summarizes the most direct way to reach the stage of a fixed Apc−/− clone. The fixation probabilities are indicated in the figure [value at 100 days in (A)]. The Graph depicts the proportion of fixed Apc−/− clones (vertical axis) that arise after the total number of Apc hits, indicated at the horizontal axis.

Because TP53 mutations are often seen in patients with CRC (2), we next investigated the effect of the dominant-negative hotspot mutation P53R172H. However, no significant benefit of this mutation could be detected in the SI (fig. S8). Because TP53 is reported to be of particular importance in colitis-associated CRC formation (12), we also studied P53 mutations in a chronic colitis model (Fig. 4). Colitis was induced by feeding mice dextran sodium sulfate (DSS) starting a week before clone induction and was maintained until analysis of clone size distributions (Fig. 4, A and B, and fig. S10). Colitis induces a shift in stem cell dynamics in the colon, mainly by reducing the replacement rate (fig. S2), and although P53R172H did not confer a benefit to colon stem cells under normal conditions, in colitis, it significantly increases fitness: PR(P53R172H versus WT) = 0.48 (0.45 to 0.51, 95% CI) and PRcolits(P53R172H versus WT) = 0.58, (0.54 to 0.63, 95% CI), respectively (Fig. 4C). Therefore, P53 mutated clones have an increased likelihood to become fixed in a gut affected by inflammation (Fig. 4D). This presumably reflects the benefit of P53-mutated cells in dealing with colitis-associated reactive oxygen species formation (13). More generally, this result demonstrates that the competitive benefit of mutations during tumor initiation is dependent on the context in which they arise.

Fig. 4 P53 mutations confer a condition-dependent clonal advantage.

(A) Confocal images of colonic crypt bottoms of DSS-treated Lgr5-CreER/tdTom-/fl mice at the indicated time points after clone induction. Clone sizes are indicated in eighths of the crypt circumference. Green, Lgr5-GFP; red, tdTom; scale bars represent 30 μm. (B) Graph depicts the average size of WT or P53R172H clones at different time points after induction in a control (left) and colitis (right) setting. Error bars indicate the SEM, N = 5 mice for each group; for each time point and condition, >150 clones were analyzed. (C) Bayesian inference of the biased drift parameter PR for P53R172H versus WT in control animals and DSS-treated animals (colitis). (D) Graph depicts the likelihood of P53R172H clones to reach fixation in homeostasis and in colitis. Shaded areas indicate 95% CI.

Our work presents a quantification of the effects of relevant mutations on stem cell dynamics during the initiation of a solid malignancy. A limited competitive advantage of common mutations in CRC is revealed, and many mutations that occur are lost from the population because of stochastic, albeit biased, replacement by neighboring WT lineages. This finding supports the accepted but so far theoretical notion that the tissue architecture and the features of the “evolutionary graph” representing the intestinal stem cell population prevent deterministic fixation of mutated lineages (1416). The values now placed on the clonal advantages conferred by common mutations can be used in future modeling studies of tumor initiation in the intestine that so far have not used experimentally derived values for the benefits of individual mutations (17). This will help to assess other phenomena associated with CRC development, such as CIN (18). Furthermore, our work reveals that the potential competitive benefits of mutations are dependent on the environment in which they arise and provides a compelling explanation of why TP53 mutations are found frequently and already at early stages in colitis-associated CRC (12). The method used here provides a powerful tool to investigate therapeutic strategies to specifically eradicate (pre-)neoplastic stem cells while preserving their normal counterparts.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

References (1940)

References and Notes

  1. Acknowledgments: We thank the Cambridge Institute Biological Resource Unit for animal husbandry and F. De Sousa E Melo, J. P. Medema, and P. Calabrese for useful discussion. This work was supported by Cancer Research UK; L.V. received a fellowship from the Koningin Wilhelmina Fonds (KWF, Dutch Cancer Society). We declare no competing financial interests.
View Abstract

Navigate This Article