Single cell lineage tracing reveals a role for TgfβR2 in intestinal stem cell dynamics and differentiation

Significance

Although Tgfβ signaling is important in intestinal development and cancer, little is known about the consequences of sporadic transforming growth factor β receptor 2 (TgfβR2) mutation in intestinal stem cells (ISCs). By labeling single, TgfβR2-mutant ISCs, we measured the effects of TgfβR2 loss on competition-driven clonal dynamics and differentiation. Specifically, we found that stochastic loss of TgfβR2 increases clonal survival while paradoxically decreasing clonal expansion and crypt fission, further elucidating mechanisms responsible for the role of Tgfβ signaling in ISCs on tumor initiation and tissue regeneration. In addition, we found that Tgfβ signaling modulates the generation of secretory cell precursors, revealing a role for Tgfβ signaling in altering ISC differentiation with implications for cancer, tissue regeneration, and inflammation.

Abstract

Intestinal stem cells (ISCs) are maintained by a niche mechanism, in which multiple ISCs undergo differential fates where a single ISC clone ultimately occupies the niche. Importantly, mutations continually accumulate within ISCs creating a potential competitive niche environment. Here we use single cell lineage tracing following stochastic transforming growth factor β receptor 2 (TgfβR2) mutation to show cell autonomous effects of TgfβR2 loss on ISC clonal dynamics and differentiation. Specifically, TgfβR2 mutation in ISCs increased clone survival while lengthening times to monoclonality, suggesting that Tgfβ signaling controls both ISC clone extinction and expansion, independent of proliferation. In addition, TgfβR2 loss in vivo reduced crypt fission, irradiation-induced crypt regeneration, and differentiation toward Paneth cells. Finally, altered Tgfβ signaling in cultured mouse and human enteroids supports further the in vivo data and reveals a critical role for Tgfβ signaling in generating precursor secretory cells. Overall, our data reveal a key role for Tgfβ signaling in regulating ISCs clonal dynamics and differentiation, with implications for cancer, tissue regeneration, and inflammation.

The intestinal epithelium is constantly renewed by proliferating, multipotent, and self-renewing intestinal stem cells (ISCs) (1). There are two main populations of ISCs: (i) a proliferating ISC population that is important for homeostasis of the niche residing below the +4 position and expressing a set of markers [e.g., leucine-rich repeat-containing G-protein coupled receptor 5 (Lgr5) and Olfm4] and (ii) a quiescent ISC population residing near the +4 position and expressing a different set of markers (e.g., Bmi1 and Hopx) (2). Proliferating ISCs are the workhorses during normal homeostasis and are maintained within the niche by a close relationship with Paneth cells (3) and the stroma (4). The proliferating ISC population can be further divided into a smaller number (4–8) of functional ISCs (5, 6), which are located at the bottom of the crypt and are biased toward survival within the niche (7). The proliferating ISCs use a population niche mechanism called neutral drift that combines differential ISC clone fates (8, 9). In neutral drift, a constant number of proliferative ISCs are maintained by a balance of ISC clone extinction with ISC clone expansion. Thus, the ISC niche must have signaling mechanisms that maintain and balance the different states of ISCs.

Much is known about the effects of WNT, BMP, and Notch signaling within the ISC niche (10), whereas little is known about the role that Tgfβ signaling through transforming growth factor β receptor 2 (TgfβR2) has on the ISC niche. Tgfβ signaling is known to play important roles in differentiation, cell motility, cell cycle, apoptosis, and inflammation (11), is critical during several phases of mammalian development (12–14), and is altered in cancer (15, 16). Tgfβ signaling involves Tgfβ ligands (Tgfβ1, 2, or 3) that bind to and activate Tgfβ receptors on the cell surface. The receptors, TgfβR1 and TgfβR2, form a heterodimer on ligand binding to create an active complex. The activated Tgfβ receptor complex phosphorylates and activates Smad2 and Smad3 (pSmad2/3), which in turn bind to Smad4, forming a transcriptional complex, which translocates to the nucleus and regulates target genes. Given the basic role of Tgfβ signaling within a cell, it seems likely that Tgfβ signaling will play a role within ISCs.

Previous investigations of TgfβR2 mutation in the intestine using epithelium-wide deletion did not detect any obvious phenotypes (17–19). However, the design of these studies would not have detected phenotypes resulting from competition between Tgfβ-positive and -negative cells within the crypt. For example, there is evidence from the hematopoietic system that competition between cells with and without Tgfβ signaling resulted in a different phenotype compared with an environment with no competition (20). ISCs are constantly dividing and therefore continually accumulating diverse mutations, which can potentially result in competition-driven drift between ISCs. Recent studies have demonstrated that isolated single ISCs with mutations in Kras and Apc are more prone to clonal expansion relative to surrounding WT ISCs (21, 22). Here we examine the effects of stochastic loss of TgfβR2 on competition between mutant and WT ISCs.

Results

Continuous and Pulse Labeling of ISCs Reveal Altered Clonal Dynamics Following TgfβR2 Mutation.

We used the stochastic Pms2^cre system to determine the consequences of sporadic, low-frequency, single cell TgfβR2 disruption in isolated crypts within the mouse small intestine (23–25). In our system, the Pms2^cre allele is comprised of a revertible out-of-frame cre gene that is targeted to Pms2, a DNA mismatch repair gene expressed in multiple cell types, including ISCs. By using a stochastic process (spontaneous frame-shift mutation), activation of Cre recombinase occurs at a defined rate resulting in continuous labeling similar to another system (5) (Fig. S1A). Lineage labeling in the intestine will only be retained when Cre activation occurs in a long-lived progenitor cell (i.e., stem cell), thus making the Pms2^cre mouse system ideal for continuous clonal labeling (Fig. 1A). When this system is combined with conditional TgfβR2 alleles (TgfβR2^fx), we can monitor the fate of isolated ISCs in a niche with neutral drift (i.e., labeled WT ISC surrounded by unlabeled, WT ISCs) or in a niche with competition-driven drift (i.e., labeled TgfβR2-mutant ISC surrounded by unlabeled, WT ISCs).

Fig. S1.

Diagrams showing the design and outcomes of continuous and pulse labeling. (A and C) Mice contain three different alleles: Pms2^cre or Lgr5-CreER, TgfβR2^fx, and Rosa26. TgfβR2^fx is a conditional TgfβR2 allele with loxp sites surrounding exon 2. On activation of Cre, exon 2 is deleted and the TgfβR2 gene is nonfunctional. Rosa26 is a reporter allele that contains a floxed STOP cassette followed by the LacZ gene. On activation of Cre, the STOP cassette is removed and LacZ is activated. (A) Diagram for continuous labeling experiments. The Pms2^cre allele contains a mononucleotide repeat (A₁₂) putting cre out of frame. A stochastic, −1-bp frame-shift mutation results in functional Cre protein. (C) Diagram for pulse labeling. Lgr5-CreER allele contains the estrogen receptor fused to Cre targeted to the ISC marker, Lgr5. On tamoxifen addition, Cre is translocated from the cell membrane to the nucleus. (B and D) Schematic showing the differences between (B) continuous labeling (Pms2^cre) or (D) pulse labeling (pulse of tamoxifen with Lgr5-CreER mouse). Relevant data for continuous labeling are the number of fully and partially labeled crypts, whereas the relevant data for pulse labeling is the percent fully labeled (time to monoclonality) and percent of crypts with any label (crypt succession).

Fig. 1.

Continuous clonal labeling (Pms2^cre) following stochastic loss of TgfβR2. (A) Images of the small intestine from WT (281 d old) and TgfβR2-mutant (279 d old) mice showing β-gal⁺ crypts. (B) Images of partially (1/4, 1/2, 3/4) and fully (monoclonal) β-gal⁺ crypts. (C) Fully labeled crypts plotted with age. TgfβR2-mutant intestine have increased accumulation of fully β-gal⁺ crypts compared with WT. Dashed lines are linear regressions. (D) Partially labeled crypts plotted with age. TgfβR2-mutant intestine have more partially β-gal⁺ crypts compared with WT. Dashed lines are linear regressions. Each spot is an independent mouse and at least 9,000 total crypts were analyzed per mouse. See also Figs. S1–S3.

Using the stochastic system described above, we compared proximal small intestines of Pms2^cre/cre; TgfβR2^+/+; R26R (WT) and Pms2^cre/cre; TgfβR2^fx/fx; R26R (TgfβR2 mutant) mice. First, we determined the number of partial and fully labeled β-gal⁺ crypts at different ages. For simplicity, we divided the crypt into one-quarter fractions or clone sizes (Fig. 1B). For WT mice, we found a constant number of partially labeled crypts with age (average, 217 β-gal⁺ foci) (Fig. 1D) and, as expected, an increasing number of fully labeled crypts with age (∼4.1 β-gal⁺ foci/d) (Fig. 1C). Interestingly, for TgfβR2-mutant mice, we found greater numbers of both partially labeled (average, 630 β-gal⁺ foci; P < 0.001 for intercept) (Fig. 1D) and fully labeled crypts (∼8.3 β-gal⁺ foci/d; P = 0.001 for slope) compared with WT mice (Fig. 1C). The altered drift following TgfβR2 loss in ISCs was independent of cell proliferation, apoptosis, or the total cell number within the crypt (Fig. S2 A–C). In addition, Tgfβ-responsive cells, as measured by pSmad2/3 staining, were evenly distributed throughout the crypt bottom (Fig. S2D). These results are consistent with TgfβR2-mutant ISCs having greater competition-driven clonal survival (more fully labeled crypts with age) while also having an elongated time to full crypt occupancy (more partially labeled crypts) compared with labeled, WT ISCs.

Fig. S2.

Consequences of stochastic TgfβR2 loss in ISCs on proliferation, apoptosis, and cell number. (A) Pms2^cre mice were injected with a single dose of BrdU, and then killed 2 h later. No significant change in the number of BrdU⁺ cells per crypt bottom between WT β-gal^neg, WT β-gal⁺, TgfβR2^fx β-gal^neg, or TgfβR2^fx β-gal⁺ (n = 3 mice per genotype). Red asterisks mark BrdU⁺ cells in the crypt bottom. (B) ApopTag was used to measure the number of apoptotic cells within the bottom third of the crypt in WT β-gal^neg (n = 4 mice), WT β-gal⁺ (n = 4 mice), TgfβR2^fx β-gal^neg (n = 4 mice), and TgfβR2^fx β-gal⁺ (n = 4 mice). Red asterisk marks TUNEL⁺ cell in the mid-crypt, which was not scored because ISCs are not located in this region. (C) Number of cells per hemicrypt section (n = 80 crypts per phenotype). No difference between β-gal^neg and β-gal⁺ crypts in TgfβR2-mutant intestine (P = 0.59). Error bars are 1 SD. (D) Immunohistochemistry for pSmad2/3 was used to measure Tgfβ responsive cells within the crypt. WT crypts had an even distribution of pSmad2/3⁺ cells within the crypt bottom (from +1 to +10) (n = 44 pSmad2/3⁺ cells). Crypts from mice irradiated with 12 Gy of X-rays had a greater percentage of pSmad2/3⁺ cells near the base of the crypt (n = 71 pSmad2/3⁺ cells).

We verified Cre-mediated recombination of the TgfβR2 floxed allele by PCR assay on microdissected crypts and found that 92% (23/25) of β-gal⁺ foci were positive for TgfβR2 recombination, whereas only 10% (1/10) of β-gal^neg foci were positive for TgfβR2 recombination (P < 0.001; Fig. S3A). The PCR data strongly support efficient recombination of TgfβR2 in β-gal⁺ cells and the stochastic nature of the Pms2^cre system with a calculated β-gal⁺ activation rate of 0.0003 β-gal⁺ events per cell (Fig. S3C), making multiple, independent TgfβR2 mutations in a single crypt highly unlikely. In addition, we found pSmad2/3⁺ cells in 15% (151/1,028) of β-gal^neg crypts, but only in 4% (15/404) of β-gal⁺ crypts from TgfβR2^fx/fx mice, supporting loss of Tgfβ signaling in 75% [1 − (observed/expected)] of β-gal⁺ crypts. Therefore, the data for β-gal⁺ cells in the TgfβR2-mutant mice are representative of a single TgfβR2 mutant ISC arising within a crypt of WT ISCs.

Fig. S3.

Labeling and recombination efficiency in the continuous and pulse labeling mice. (A) PCR from microdissected crypts showing recombination of the TgfβR2^fx allele in both Pms2^cre and Lgr5-CreER mice. (B) Percentage of independent, β-gal⁺ events per crypt in Lgr5-CreER mice 3 d after tamoxifen treatment. Fifty-three percent of crypts revealed multiple independent, β-gal⁺ cells, suggesting that initial labeling of multiple cells per crypt was possible. (C) Image of a single β-gal⁺ cell within the crypt. Mutation rate in Pms2^cre/cre mice calculated from single β-gal⁺ cells.

To determine independently the consequences of eliminating Tgfβ signaling in proliferating ISCs, we used Lgr5-CreER mice with, or without, conditional TgfβR2 alleles. The Lgr5-CreER mouse contains Cre fused to the estrogen receptor and is expressed from the ISC-specific promoter, Lgr5 (26). Thus, after injection of tamoxifen (pulse labeling), Cre becomes active in ISCs and, when combined with a LacZ reporter allele (R26R), can label the ISC for lineage tracing (Fig. 2A). By labeling a small number of ISCs in any given crypt, we can follow the progression of a crypt from partially labeled to fully labeled (time to monoclonality) and the fraction of surviving β-gal⁺ crypts (crypt succession) (Fig. S1B). To study drift, we compared the clone size distribution or crypt succession with time in Lgr5-CreER; TgfβR2^+/+; R26R (WT) and Lgr5-CreER; TgfβR2^fx/fx; R26R (TgfβR2-mutant) mice, injected with tamoxifen at 2 mo of age. By using a single dose of tamoxifen (2 mg/mouse), we were able to induce mosaic recombination in a fraction of crypts (∼20%) while still obtaining recombination of the TgfβR2^fx alleles (Fig. S3A). The dose of tamoxifen used can result in multiple, recombination events per crypt (Fig. S3B); therefore, it is possible to have β-gal^neg and β-gal⁺; TgfβR2-mutant ISCs within a single crypt. For pulse labeling, clonal survival was increased in TgfβR2-mutant crypts compared with WT crypts (Fig. 2B) (P = 0.02 for log-transformed slope). In addition, the time to monoclonality was elongated in TgfβR2-mutant crypts compared with WT crypts (Fig. 2C) (P < 0.001 for logit-transformed slope). These results are again consistent with TgfβR2-mutant ISCs having greater competition-driven clonal survival (more labeled crypts) while also having an elongated time to full crypt occupancy (longer time to monoclonality) compared with β-gal⁺, WT ISCs.

Fig. 2.

Pulse labeling (Lgr5-CreER) following stochastic loss of TgfβR2. (A) Images of the small intestine from WT (49 d after induction) and TgfβR2-mutant (56 d after induction) mice showing β-gal⁺ crypts. (B) The number of remaining β-gal⁺ crypts was increased in TgfβR2-mutant intestine compared with WT. Dashed lines are exponential trend lines. (C) Time to monoclonality was elongated in TgfβR2-mutant crypts compared with WT. Each spot is an independent mouse and at least 9,000 total crypts were analyzed per mouse. See also Figs. S1 and S3.

Computational Simulations Reveal Both Decreased ISC Clone Expansion and Clone Extinction Following Stochastic Loss of TgfβR2.

To interpret how the changes in ISC clone survival and times to monoclonality affect ISC clonal dynamics, we simulated competition between WT and mutant (TgfβR2^−/−) ISCs. Our model has the following parameters: N (number of stem cells), m (mutation rate), time, λ (WT replacement rate), TgfβR2-λ (TgfβR2^−/− replacement rate), and F_R (TgfβR2^−/− replacement factor) (Fig. S4A). Our model is independent of cell proliferation; however, TgfβR2 loss did not have any appreciable effect on proliferation in cells located at the crypt base (Fig. S2A). Simplistically, the fate of an isolated single ISC depends on the relative balance between clone extinction vs. clone expansion, where λ reflects the rate of ISC clone extinction, whereas F_R reflects the rate of competitive ISC clone expansion.

Fig. S4.

Computational simulations following stochastic loss of TgfβR2. (A) Diagram of the computational model used to measure the competition between WT (white) and mutant (blue) stem cells. TgfβR2-mutant stem cells can have different rates of extinction (TgfβR2-λ) and/or competitive clonal expansion (F_R). (B and C) Continuous labeling. (B) TgfβR2-mutant data for fully labeled crypts can be simulated by either increasing F_R alone (green line; F_R = 2), decreasing TgfβR2-λ alone (purple line; TgfβR2-λ = 0.075), or decreasing both TgfβR2-λ and F_R (orange line; F_R = 0.3, TgfβR2-λ = 0.025). Dashed lines are linear regressions for WT (gray) or TgfβR2-mutant (red) data. (C) Using the same simulations for partially labeled crypts, TgfβR2-mutant data strongly support decreasing both TgfβR2-λ and F_R. (D–F) Pulse labeling. (D) TgfβR2-mutant data for clonal survival can be simulated by either decreasing TgfβR2-λ alone (purple line; TgfβR2-λ = 0.07), increasing F_R alone (light green line; F_R = 2), or decreasing both TgfβR2-λ and F_R (orange line; F_R = 0.25, TgfβR2-λ = 0.02). Dashed lines are exponential trend lines for WT (gray) or TgfβR2-mutant (red) data. (E) TgfβR2-mutant data for time to crypt dominance could only be fit by either decreasing F_R alone (dark green line; F_R = 0.1) or best fit by decreasing both TgfβR2-λ and F_R (orange line; F_R = 0.25, TgfβR2-λ = 0.02). Dashed lines are third-order polynomial trend lines. (F) Using the same simulations for partially labeled crypts, TgfβR2-mutant data strongly support decreasing both TgfβR2-λ and F_R. Dashed lines are third-order polynomial trend lines.

First, we calculated N and λ for WT crypts using our computational model with both the continuous and pulse labeling, which revealed that the best fit for WT crypts was N = 4 and λ = 0.14–0.15 (Table S1 and Fig. S5 A and C), similar to a previous estimate (5). Next, we calculated the effects of TgfβR2 mutation on ISC competition-driven drift. These simulations revealed an approximately sevenfold decreased clone extinction (TgfβR2-λ) and an approximately threefold decreased competitive clone expansion (F_R) (Fig. S4 B and C and Table S1). These simulations indicate an approximately twofold greater reduction in ISC clone extinction compared with clone expansion following loss of TgfβR2 when in competition with WT ISCs, leading to the slower but overall increased niche occupancy. The net result is that TgfβR2-mutant ISCs have increased clone survival compared with WT ISCs. Because ISC proliferation is not altered with TgfβR2 mutation, these results suggest that the increased clone survival of TgfβR2-mutant ISCs is through altered differentiation.

Table S1.

Values calculated for competitive clonal expansion and extinction following TgfβR2 mutation

Parameter	Range	Meaning
N	1–∞	Stem cell number
λ	0–∞	Replacement rate (extinction rate)
FR	0–∞ (1 = neutral)	Replacement factor (competitive clonal expansion)
	WT	TgfβR2 mutant
Continuous labeling
N	4	Assumed = WT
λ [fold Δ]	0.15	0.025 [−6]
FR [fold Δ]	1	0.3 [−3.3]
Pulse labeling
N	4	Assumed = WT
λ [fold Δ]	0.14	0.02 [−7]
FR [fold Δ]	1	0.25 [−4]

Fig. S5.

Simulations for continuous and pulse labeling. (A and B) Continuous labeling. (A) For WT intestine, simulations reveal a best fit model for the data as n = 4; λ = 0.15 (data from 200 d). (B) For TgfβR2-mutant intestine, simulations reveal a best fit model for the data as TgfβR2-λ = 0.025; F_R = 0.3 (data from 200 d). (C and D) Pulse labeling. (C) For WT intestine, simulations reveal a best fit model for the data as n = 4; λ = 0.14 (data from 50 d after induction). (D) For TgfβR2-mutant intestine, simulations reveal a best fit model for the data as TgfβR2-λ = 0.02; F_R = 0.25 (data from 50 d after induction).

Loss of TgfβR2 in ISCs in Vivo Reduces the Chance of Crypt Fission.

Our data revealed that loss of TgfβR2 could lengthen times to monoclonality, supporting reduced ISC clone expansion. Next, we examined the effects of stochastic mutation on crypt fission, the process by which a single crypt splits into two crypts presumably caused by an increased number of ISCs (i.e., ISC clone expansion). The stochastic nature of the Pms2^cre system make multiple, independent TgfβR2 mutations in neighboring crypts highly unlikely; thus, the distribution of crypt patch sizes reflects crypt fission events over time. Therefore, we determined the numbers of β-gal⁺ foci that contain multiple, neighboring β-gal⁺ crypts in similarly aged WT (mean, 192 d) or TgfβR2-mutant (mean, 192 d) mice (P = 0.99). Interestingly, there was an overall reduced size of β-gal⁺ foci in TgfβR2-mutant intestines (mean β-gal⁺ focus size = 1.5 crypts) compared with WT intestines (mean β-gal⁺ focus size = 1.9 crypts), suggesting reduced crypt fission in TgfβR2-mutant crypts (P = 0.003; Fig. 3A). These results are consistent with decreased crypt fission following stochastic TgfβR2 mutation.

Fig. 3.

Crypt fission and regeneration are reduced following loss of TgfβR2 in ISCs. (A) Using continuous clonal labeling, we observed a reduced fraction of larger β-gal⁺ foci (3+ neighboring β-gal⁺ crypts) in TgfβR2-mutant intestine (n = 5 mice) compared with WT intestine (n = 6 mice). (B) Image of a crypt undergoing fission. Graph showing reduced crypt fission index in intestines with intestine-wide deletion of TgfβR2 (n = 5 mice) compared with WT crypt fission index (n = 9 mice). (C) Images of stained intestine from WT or TgfβR2-mutant mice with pulse labeling and irradiation (12 Gy). Decreased number of repopulated β-gal⁺ clones following irradiation in TgfβR2-mutant intestine (n = 6 mice) compared with WT intestine (n = 4 mice). (D) Immunohistochemistry for pSmad2/3 in unirradiated and irradiated intestine. Irradiation (12 Gy of X-rays) (n = 4 mice) increased the fraction of pSmad2/3⁺ cells within the crypt, and specifically within Paneth cells, compared with unirradiated control intestine (n = 8 mice). No significant change in the number pSmad2/3⁺ stromal cells. Error bars are 1 SD.

Next, we combined the Villin-Cre allele with conditional TgfβR2^fx alleles (TgfβR2^IEKO) (17, 19, 27) to determine whether the effects of TgfβR2 loss on crypt fission were caused by competition with WT cells or by TgfβR2-mutant crypts inherently having a reduced level of crypt fission. The TgfβR2^IEKO mice exhibit loss of TgfβR2 throughout the intestinal epithelium; thus, almost all epithelial cells (and crypts) become TgfβR2^−/− without intra- or intercrypt competition. Previous reports on these mice did not report any obvious differences from the intestines of WT mice; however, there was not a close examination of crypt fission rates (17, 19, 27). Therefore, we examined the number of crypts undergoing fission (crypts in the process of budding) in adult TgfβR2^IEKO (mean, 16 mo) or WT (mean, 15.4 mo) mice (P = 0.76). Interestingly, we found a reduced percentage of crypts undergoing fission in TgfβR2^IEKO intestine compared with WT intestine (Fig. 3B). These results revealed that TgfβR2-mutant crypts inherently have reduced crypt fission whether in competition with neighboring crypts or not.

Loss of TgfβR2 in ISCs in Vivo Reduces the Chance of Regeneration Following Irradiation.

Next, we studied the effects of irradiation following sporadic loss of TgfβR2 in ISCs because ISC clonal expansion is necessary following crypt damage, and our data suggest that TgfβR2 loss retards ISC clonal expansion. A subset of Lgr5⁺ ISCs is necessary for crypt regeneration following irradiation (28), and intestine-wide loss of TgfβR2 resulted in a slower rate of crypt regeneration following irradiation (27). Therefore, we studied the effects of irradiation (12 Gy) on TgfβR2-mutant ISCs when competing with WT ISCs using the pulse labeling model (Lgr5-CreER mice). We define β-gal⁺ foci that extend from the crypt onto the body of the villus as a “repopulated clone” and β-gal⁺ foci that were only present on the body of the villus (not extending into the crypt) as an “extinguished clone.” Mice were given a single pulse of tamoxifen and then 3 wk later were exposed to 12 Gy of irradiation. Five to 7 d after irradiation, TgfβR2-mutant cell lineages in the proximal small intestine were scored as repopulated clones in only 35% of crypts compared with 66% in WT cell lineages (Fig. 3C). Finally, we found that irradiation altered the distribution of pSmad2/3⁺ cells toward the crypt base (Fig. S2D) and dramatically increased the pSmad2/3 staining intensity and number (∼20-fold) of pSmad2/3⁺ cells within the crypt, specifically in the Paneth cell lineage (Fig. 3D). However, there was minimal change in the number of pSmad2/3⁺ cells in the stroma after irradiation (Fig. 3D). These data support a key role for Tgfβ signaling through TgfβR2 within the ISC population in crypt regeneration after damage and a role for Tgfβ signaling in the Paneth cell lineage.

Cell Type Analysis in Vivo Reveals a Role for TgfβR2 in the Generation of Paneth Cells.

Because our data suggested a role for Tgfβ signaling in differentiation, we examined Pms2^cre mice for altered cell labeling after TgfβR2 deletion. We found that 21% of β-gal⁺ crypts from TgfβR2-mutant mice showed an absence of β-gal⁺ cells characteristic of Paneth cells compared with only 1% of β-gal⁺ crypts in WT mice (P = 0.03; Fig. 4A). We also analyzed Lgr5-CreER mice and noted that, at 4 wk following TgfβR2 loss, there was an average of 5.9 ± 0.4 β-gal⁺ Paneth cells in WT crypts, but only an average of 3.5 ± 0.4 β-gal⁺ Paneth cells in TgfβR2-mutant crypts (P = 0.002). These results suggest that TgfβR2-mutant ISCs are less likely to generate Paneth cells than WT ISCs.

Fig. 4.

Change in the formation of Paneth cells following deletion of TgβR2 in vivo. (A) Pms2^cre mice revealed a reduced rate of Paneth cell generation following TgfβR2 deletion in ISCs (n = 4 mice) compared with control, WT ISCs (n = 4 mice). Red asterisks mark unlabeled Paneth cells. (B) Intestinal epithelium wide deletion of TgfβR2 (IEKO) resulted in significantly fewer cells of the secretory lineage (Paneth and Goblet) in crypts (n = 5 mice) compared with control, WT crypts (n = 5 mice). (Scale bar, 25 μm.) Error bars are 1 SD.

To study further the effects of TgfβR2 loss on differentiation, we used the TgfβR2^IEKO mice (17, 19) to determine whether the effect of TgfβR2 loss on the formation of Paneth cells was caused by competition with WT cells or by an inherent defect of TgfβR2-mutant crypts. Interestingly, we observed 18% fewer Paneth cells per crypt section in TgfβR2^IEKO mice compared with TgfβR2^fx mice (P < 0.001; Fig. 4B). Overall, our results are consistent with TgfβR2-mutant ISCs possessing a reduced capacity to produce Paneth cells and thus a role for Tgfβ signaling in differentiation toward the secretory lineage.

Cultured Intestinal Enteroids Reveal a Role for Tgfβ Signaling in ISC Clonal Expansion.

To study further the effects of Tgfβ signaling modifications on the intestinal epithelium, we used cultured mouse proximal small intestinal enteroids, which are self-perpetuating and capable of producing each of the cell types characteristic of the intestinal epithelium (29). The enteroids allow us to study the early and rapid effects of both up- and down-regulating Tgfβ signaling by treating with either an inhibitor of TgfβR1/2 (30) or Tgfβ1 ligand. Intestinal enteroids treated with high levels of Tgfβ1 ligand (4 ng/mL) results in cell death as seen previously (31) and could be rescued by cotreatment with the inhibitor of TgfβR1/2 (Fig. S6A). Because initial treatment with 4 ng/mL of Tgfβ1 ligand resulted in rapid enteroid death, we treated enteroids with the highest dose that did not induce enteroid death (0.04 ng/mL Tgfβ1 ligand; Fig. S6A). Although Tgfβ responsive cells (pSmad2/3⁺) were rare (<3% of all cells) in both untreated and 0.04 ng/mL Tgfβ1-treated enteroids, Tgfβ1 treatment resulted in an approximately ninefold increased fraction of pSmad2/3⁺ cells compared with untreated (P = 0.003; Fig. S7A), suggesting this level of Tgfβ1 ligand increases Tgfβ signaling but only in a limited number of cells.

Fig. S6.

Tgfβ signaling in intestinal enteroids. (A) Images and survival curve showing that 4 ng/mL Tgfβ1 ligand (n = 3 wells) increases toxicity compared with control (n = 3 wells). The TgfβR1/2 inhibitor (n = 3 wells) is minimally toxic on its own and rescues the toxicity caused by Tgfβ1 ligand (n = 5 wells); 0.04 ng/mL of Tgfβ ligand is minimally toxic (n = 3 wells). (B) Proliferation of enteroids as measured by 1-h EdU labeling without treatment (n = 11 enteroids) or with Tgfβ inhibitor (n = 21 enteroids) or 0.04 ng/mL Tgfβ1 ligand (n = 21 enteroids). Treatment did not alter proliferation within the crypt bud (where ISCs are located), but did alter proliferation outside of the crypt bud. Crypt buds are outlined in yellow. Error bars are 1 SD.

Fig. S7.

Secretory cells in enteroids altered for Tgfβ signaling. (A) Immunohistochemistry images of enteroid sections stained for pSmad2/3 and counterstained with nuclear fast red. pSmad2/3⁺ cells are rare in untreated enteroids (n = 14) or not seen in Tgfβ inhibitor-treated enteroids (n = 10). The 0.04 ng/mL Tgfβ1 ligand treatment results in approximately ninefold increased, but still sporadic, pSmad2/3⁺ cells (n = 14 enteroids). (B) Images of sections from enteroids either untreated or treated with the Tgfβ inhibitor or ligand and then stained for UEA (green) and counterstained with DAPI (blue). The number of secretory precursor cells (UEA⁺) was decreased with Tgfβ inhibitor treatment (n = 30 crypt buds) and increased with Tgfβ1 treatment (n = 17 crypt buds) compared with control (n = 68 crypt buds). Error bars are 1 SD.

To identify global changes in enteroids after altering Tgfβ signaling, we used Gene Set Enrichment Analysis (GSEA) (32) to determine if our treatment groups had similar gene expression enrichment for stem cell genes by comparing Lgr5-GFP^high cells with Lgr5-GFP^low cells (33). GSEA on the microarray data showed that the Tgfβ inhibitor resulted in decreased expression of genes characteristic of stem cells, whereas Tgfβ ligand treatment showed increased expression of the same genes from stem cells (Fig. 5A). In support of a reduced stem cell signature with Tgfβ inhibition, we observed a reduced rate of crypt bud formation following treatment of enteroids with the TgfβR1/2 inhibitor (Fig. 5B). Treatment with either the TgfβR1/2 inhibitor or 0.04 ng/mL Tgfβ1 ligand had no detectable effect of proliferation within the crypt bud (where the ISCs are located), but dramatically decreased proliferation outside of the crypt bud (Fig. S6B). These results further support the in vivo data that Tgfβ signaling does not affect ISC division rates, but instead support that the effects Tgfβ signaling are on ISC dynamics and clonal expansion.

Fig. 5.

Tgfβ signaling is important for the generation of stem and secretory cell lineage in cultured enteroids. (A and C) GSEA for Tgfβ inhibitor or Tgfβ1 ligand vs. control treatment. Tgfβ inhibitor treatment decreases the enrichment score (ES) for (A) stem cell and (C) secretory precursor genes (P < 0.001 for each). Tgfβ1 treatment increases the ES for (A) stem cell and (C) secretory precursor genes (P < 0.001 for each). NES, normalized enrichment score. (B) Enteroids treated with the TgfβR1/2 inhibitor show a slower accumulation of new crypt-bud formation compared with control or Tgfβ1 ligand treatment (P = 0.02 for slope). (D) Lysozyme expression measured by qRT-PCR is reduced with Tgfβ inhibition and increased with Tgfβ1 ligand compared with control treatment. In addition, Notch inhibition (DAPT treatment) greatly increases lysozyme expression, but the addition of Tgfβ1 ligand exponentially increases lysozyme expression compared with the control and Tgfβ inhibitor. (n = 3 for all treatment groups). Error bars are 1 SD. See also Figs. S6–S9.

Cultured Intestinal Enteroids Reveal a Role for Tgfβ Signaling in Differentiation Toward Secretory Cell Lineage Precursors.

To determine whether altered Tgfβ signaling had gene expression enrichment for secretory precursor cell genes, we again started by using GSEA (32) and comparing secretory progenitors with enterocytes (34). GSEA on the microarray data showed that the Tgfβ inhibitor resulted in decreased expression of genes characteristic of secretory precursor cells, whereas Tgfβ ligand treatment showed increased expression of the same genes (Fig. 5C). To confirm the expression array data, we stained for the lectin from Ulex europaeus (UEA), which is a marker for the secretory cell lineage: Paneth, enteroendocrine, and Goblet cells (35). We found that the low dose of Tgfβ1 ligand resulted in an increased number of UEA⁺ cells per bud (equivalent to the crypt), whereas treatment of the enteroids with the TgfβR1/2 inhibitor resulted in a decreased number of UEA⁺ cells per bud (Fig. S7B). These data in intestinal enteroids support the in vivo data that Tgfβ signaling is key for differentiation toward the secretory cell lineage.

To study further the role of Tgfβ signaling in differentiation toward the secretory lineage, we altered Tgfβ signaling in combination with inhibiting Notch signaling, which is known to be critical for formation of the secretory cell lineage (36). We pretreated enteroids with either the Tgfβ1 ligand or TgfβR1/2 inhibitor for 2 d and then cotreated with the γ-secretase inhibitor DAPT for 4 d. Enteroids were examined for expression of the secretory cell marker gene, lysozyme, and a control gene, GAPDH, via qRT-PCR. In agreement with the microarray data, lysozyme expression was increased in enteroids treated with Tgfβ1 ligand (1.7-fold) and decreased with Tgfβ inhibition (0.2-fold). As expected, we found that DAPT treatment had a dramatic increase (sevenfold) on lysozyme expression in control enteroids (Fig. 5D). Interestingly, the combination of Tgfβ1 and DAPT treatment exponentially increased the expression of lysozyme (40-fold) in enteroids compared with control enteroids (Fig. 5D). These data suggest that Tgfβ signaling is acting on a precursor secretory cell lineage or ISC, which facilitates rapid formation of secretory cells when Notch signaling is inhibited.

Cultured Human Intestinal Enteroids Reveal a Role for Tgfβ Signaling in ISC Dynamics and Differentiation Toward Secretory Cell Lineage.

To determine the relevance of these findings to human intestine, we cultured enteroids from a normal human duodenum. We found that increasing Tgfβ signaling increased GSEA for stem cell genes and secretory precursor cell genes (Fig. S8A) and increased the rate of new crypt bud formation (Fig. S8B). With time in culture, human enteroids progress from the budding phenotype to a more cyst-like phenotype. Notch inhibition (DAPT) alone dramatically increased the rate of invagination, which is an initial step in creating buds (Fig. S8C). Interestingly, the combination of DAPT and Tgfβ1 ligand treatment increased the number of invaginated enteroids compared with the DAPT-treated enteroids (Fig. S8C), supporting a role for Tgfβ signaling in conjunction with Notch inhibition in differentiation. These data are in agreement with our findings in mice and suggest that Tgfβ signaling is important in regulating stem cell dynamics and differentiation toward a precursor secretory cell lineage in human small intestinal epithelium.

Fig. S8.

Human duodenal enteroids treated with Tgfβ1 ligand. (A) GSEA on human enteroids treated with or without Tgfβ1 ligand. Stem cell and secretory precursor gene signatures were significantly increased with Tgfβ1 treatment. (B) Images of budding enteroids. Tgfβ1 ligand increased the number of buds per enteroid after 4 d of treatment compared with untreated (n = 3 wells for each treatment). (C) Images of cyst-like enteroids. DAPT treatment increased the number of enteroids with invaginations, and Tgfβ1 ligand increased the number of enteroids with invaginations after DAPT treatment over the control (n = 3 wells for each treatment). Error bars are 1 SD.

Cultured Intestinal Enteroids Reveal a Role for Tgfβ Signaling in Regeneration After Damage.

To study further the effects of altering Tgfβ signaling on crypt regeneration, we treated enteroids with either the Tgfβ inhibitor or ligand and the cytotoxic agent, FUDR (5-fluoro-2′-deoxyuridine), which kills proliferating cells (37). We pretreated enteroids for 3 d with the TgfβR1/2 inhibitor or Tgfβ1 ligand and then added FUDR for 1 d and allowed the enteroids to recover for 2 d. We found that altering the Tgfβ pathway alone had minimal impact on enteroid survival. However, when treated with FUDR, the TgfβR1/2 inhibitor pretreated enteroids showed decreased survival, whereas survival was increased in the enteroids pretreated with Tgfβ1 ligand (Fig. S9). These results again support the in vivo data that Tgfβ signaling has a role in crypt regeneration.

Fig. S9.

Percentage of surviving enteroids was not affected by either Tgfβ treatment alone (ANOVA, P = 0.13). However, the percentage of surviving enteroids following FUDR treatment was increased by pretreatment with Tgfβ1 ligand and decreased with Tgfβ inhibitor (ANOVA, P < 0.001) (n = 3 wells for each treatment group). Error bars are 1 SD.

Discussion

Multiple mammalian ISCs within each crypt are maintained by a population niche mechanism of ISC clone expansion and extinction, ultimately resulting in neutral drift (8, 9). These differentially fated ISCs provide flexibility because any ISC clone extinction is readily compensated by neighboring ISC clone expansion. Although neutral drift is normally random, mutations within ISCs can alter clonal dynamics and induce selection as shown previously in ISCs with sporadic Apc or Kras mutations (21, 22). Here, we demonstrate that ISC clonal dynamics can be modulated genetically by mutations in TgfβR2. Specifically, stochastic loss of TgfβR2 resulted in increased ISC clone survival compared with WT ISCs, but at the cost of clone expansion. In combination with the effects on ISC clonal dynamics, our data in vivo and in cultured enteroids strongly implicate Tgfβ signaling in the transition from ISC to a precursor secretory cell lineage. Thus, when an ISC receives Tgfβ signaling and transitions toward the secretory lineage, the end result for that ISC clone is extinction. Importantly, our data also suggest that Tgfβ signaling and thus precursor secretory cells are important in clone expansion, crypt fission, and ISC regeneration.

Here, we show that TgfβR2 plays an important role in maintenance of the ISC clonal dynamics, which has important implications for cancer, tissue regeneration, and inflammation. First, the increased ISC clone survival comes at the cost of reduced ISC clone expansion and crypt fission, which will hinder tumor initiation and progression. In contrast, stochastic mutations in Apc or Kras increase crypt fission and clonal expansion (21, 23–25). Thus, it is likely that the reduced ISC clonal expansion and crypt fission following TgfβR2 mutation represents a key reason why TgfβR2 mutations are rare early mutational events in sporadic CRC (38). Second, the decreased ISC clone expansion and crypt fission following sporadic TgfβR2 mutations is detrimental to ISC and tissue regeneration following damage lending credence for why Tgfβ signaling is important for recovery after tissue damage (39, 40). Third, the correlation between Tgfβ signaling and formation of the secretory cell lineage has important implications in intestinal infection and inflammatory diseases. Specifically, Paneth cells maintain the homeostatic balance between the epithelium and the microbiota and are at the site of inflammation (41, 42). In conclusion, our data reveal the consequences of TgfβR2 loss on ISC clonal dynamics and differentiation with implications for how mutation of TgfβR2 can impact tissue homeostasis and alter tumorigenesis.

Methods

Complete materials and methods are reported in SI Methods and Table S2. All mouse experiments were approved by the Institutional Animal Care and Use Committee at Oregon Health and Science University. Intestines were stained for β-gal activity as previously described (25). Human duodenum was obtained with institutional review board approval at OHSU. Enteroids were treated with 20 μM LY2109761 (TgfβR1/2 inhibitor) (Adooq) (30) or 0.04 ng/mL (unless specified differently) Tgfβ1 ligand (R&D Biosystems). The computational model is a continuous-time, asynchronous model with a constant number of stem cells (N), modified from a previous model (22). To calculate statistical significance, we used univariate linear regressions in StatistiXL on both datasets (WT and mutant) and compared the slopes of each regression with ANOVA.

Table S2.

Primers used for genotyping and detecting recombination

Gene	Primer	Sequence	Annealing temperature (°C)
Pms2	A	TTCGGTGACAGATTTGTAAATG	60
	W	TCACCATAAAAATAGTTTCCCG
Cre	F	AACATTCTCCCACCGTCAGT	60
	R	CATTTGGGCCAGCTAAACAT
R26R	Mut	GCGAAGAGTTTGTCCTCAACC	60
	Common	AAAGTCGCTCTGAGTTGTTAT
	WT	GGAGCGGGAGAAATGGATATG
TgfβR2	Gen1	GCAGGCATCAGGACCTCAGTTTGATCC	64
	Gen2	AGAGTGAAGCCGTGGTAGGTGAGCTTG
TgfβR2	fx1	AGGGATGAATGGGCTTGCTT	58
Recombine	fx2	CTCACCTCAGAGCCTGATTA
TgfβR2	8w	TAAACAAGGTCCGGAGCCCA	58
Recombine	mSAr	AGAGTGAAGCCGTGGTAGGTGAGCTTG

SI METHODS

Mice.

Mice were housed in a specific pathogen-free high-efficiency particulate air (HEPA)-filtered room and were fed a diet of LabDiet PicoLab Rodent Diet 20. All experiments were approved by the IACUC committee at OHSU. Lgr5-CreER mice were generated as previously described (26). Pms2^cre mice were generated as previously described (23). TgfβR2^fx mice were generated as previously described (18). Villin-Cre; TgfβR2^fx were generated as previously described (17). The R26R (Rosa26 Reporter mice with lox-stop-lox LacZ) allele was used (43). The R26R and Lgr5-CreER mouse strains were received from the Jackson Laboratory. Mice were irradiated in a Rad Source RS 2000 X-ray irradiator at OHSU with approval from the IACUC committee. Mice were genotyped as previously described (25) with the primers and annealing temperatures listed in Table S2.

Scoring β-gal⁺ Crypts.

Intestines were stained for β-gal activity as previously described (25). Tamoxifen (Sigma) was prepared in sunflower seed oil (Sigma) at a concentration of 10 mg/mL; 200 μL of the tamoxifen solution was i.p. injected once into mice at 2 mo of age (final dose, 2 mg/mouse; ∼100 mg/kg). After staining the intestines for β-gal, the number of positive foci was scored from the serosal side (looking at the crypts). To distinguish between partially and wholly labeled crypts, the serous membrane and muscularis were removed. β-gal⁺ foci were examined under a Leica MZ6 dissecting microscope at 20× power, with a 25-mm² field of view. We scored at least 75 β-gal⁺ foci per mouse. For scoring of clone size, β-gal⁺ crypts were divided into one-quarter fractions. For determining the number of independent events following tamoxifen induction, we examined mice at 3 d after induction and counted the number of β-gal⁺ cells within the crypt that were separated by at least one β-gal^neg cell. Two milligrams of tamoxifen induced multiple events in 53% of crypts (Fig. S3B).

Sectioning and Staining.

Sections of intestinal tissue were prepared as previously described (25). For proliferation studies, mice were injected with 120 mg/kg of BrdU 2 h before death. Tissues were paraffin embedded, and sections were cut to 8 microns, deparaffinized, and rehydrated. BrdU immunohistochemistry was performed by incubating sections in 1 M HCl at 4 °C for 10 min, followed by 2 N HCl at 37 °C for 20 min. After neutralizing in borate buffer (0.5 M borate, pH 8.5) for 15 min, sections were blocked with BSA blocking buffer [2% (wt/vol) BSA in PBS] for 30 min, and then probed with mouse anti-BRDU (Roche) diluted 1:100 in BSA block overnight at 4 °C. Secondary detection was performed using DAKO Cytomation EnVision+ System-HRP, according to the manufacturer’s directions. BrdU sections were counterstained with nuclear fast red. Proliferation index was determined by counting the number of BrdU⁺ cells on one hemicrypt of complete crypts, as defined by having a horseshoe shape and containing Paneth cells. Apoptotic cells were visualized with the ApopTag kit (Chemicon) used according to the manufacturer’s directions. ApopTag-stained slides were counterstained with methyl green or nuclear fast red. Apoptosis was determined by examining crypts at 400× magnification. Total number of cells per crypt was determined by counting the column of cells in one hemicrypt, from base to neck, of a horseshoe-shaped crypt. Eighty hemicrypts were counted per crypt type and averaged. For the Villin-Cre experiments, crypt fission was scored on H&E-stained, Swiss-roll tissue sections. For pSmad2/3 staining, the phospho-Smad2/3 Rabbit mAB (Cell Signaling; #8828) was used at 1:100 and then detected with the Vectastain Elite ABC HRP kit (Vector).

Enteroid Culture.

Enteroids were cultured as previously described (24, 29). Human enteroids were cultured as previously described (44). For studies on the effects of altering Tgfβ signaling, enteroids were treated with 20 μM LY2109761 (TgfβR1/2 inhibitor) (Adooq) (30) or 0.04 ng/mL (unless specified differently) of Tgfβ1 ligand (R&D Biosystems). For proliferation experiments, 10 μM EdU (Invitrogen) was added to the culture media for 1 h. Enteroids were then washed in PBS and fixed for 5 min in 2% (wt/vol) formaldehyde and 0.5% gluteraldehyde. For the enteroid survival experiments, enteroids were treated with 0.04 ng/mL Tgfβ1 ligand (R&D Biosystems) or 20 μM LY2109761 (TgfβR1/2 inhibitor) (Adooq) and 1 μM of FUDR (5-fluoro-2′-deoxyuridine) (Sigma). For notch inhibition, enteroids were treated with 10 μM DAPT (Sigma).

Enteroids were prepared for sectioning by first fixing for 5 min in 2% (wt/vol) formaldehyde and 0.5% gluteraldehyde in the 24-well plate. The wells were washed in PBS. OCT was then added directly to the 24-well plate, covering the Matrigel (BD Biosciences). The 24-well plate was frozen at −80 °C. The embedded enteroids were removed from the 24-well plate and used for sectioning. Sections were stained for the lectin from Ulex europaeus (UEA) with a FITC- or TRITC-conjugated UEA (Sigma). The UEA was diluted 1:50 (final concentration of 20 μg/mL) in PBS and incubated for 1 h at room temperature. For pSmad2/3 staining, the phospho-Smad2/3 Rabbit mAB (Cell Signaling, #8828) was used at 1:100 and then detected with the Vectastain Elite ABC HRP kit (Vector).

For qRT-PCR expression analysis, RNA was isolated from enteroids with the RNeasy Micro Kit (Qiagen). cDNA was generated from 400 ng total RNA using oligo dT primer from the SuperScript III First-Strand Kit (Invitrogen). PCR was performed with Rotor-Gene SYBR Green PCR Kit (Qiagen) in a Rotor-Gene cycler (Qiagen). Lysozyme was amplified with the following primers: GCCAAGGTCTACAATCGTTGTGAGTTG and CAGTCAGCCAGCTTGACACCACG. GAPDH was amplified with the following primers: TCATCAACGGGAAGCCCATCAC and AGACTCCACGACATACTCAGCACCG.

Affymetrix Microarray.

Enteroids were treated with or without the TgfβR1/2 inhibitor or Tgfβ1 ligand for 5 d. Each treatment group was performed in triplicate. RNA was extracted using the miRNeasy Mini Kit (Qiagen). RNA samples (200 ng) were then amplified and labeled using the MessageAmp Premier (Ambion) for the 3′ in vitro transcription (IVT) assay. Each sample was then hybridized to an Affymetrix Mouse Genome 430 2.0 GeneChip array. Image processing and expression analysis were performed using Affymetrix GeneChip Command Console v. 3.1.1 and Affymetrix Expression Console v. 1.1 software. The data were analyzed in GeneSifter (www.geospiza.com). ANOVA test within GeneSifter was used for statistical comparison. The data discussed in this publication have been deposited in the National Center for Biotechnology Information's Gene Expression Omnibus (45) and are accessible through GEO Series accession number GSE58296 (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE58296) for mice and GEO Series accession number GSE83423 (www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE83423) for humans.

Computational Modeling.

We model the evolution of stem cells in a crypt similar to a Moran model (46). The computational model is a continuous-time, asynchronous model with a constant number of stem cells (N), modified from a previous model (22). For each time point, 100 simulations were run for 10,000 (continuous labeling) or 350 crypts (pulse labeling). For the pulse-chase case, initially one of the stem cells is labeled (blue), and the rest are unlabeled (white). For the continuous-label case, initially no stem cells are blue. Stem cells are replaced with a rate of λ, so 1/λ is the average time of replacement. Thus, a higher λ is replaced more often. When a stem cell is replaced (extinction), it is removed from the crypt and another stem cell is chosen randomly for clonal expansion so that the offspring of the replacement stem cell replaces the lost stem cell. For continuous clonal labeling, the mutation probability is per replacement event and mutation is only white to blue. The mutation rate (m) was calculated by counting the number of single, isolated β-gal⁺ cells in sections and dividing by the total number of cells counted. For these experiments, we used an m of 0.0003 for Pms2^cre/cre mice (calculated from four mice: 7 single β-gal⁺ cells/18,720 cells; 13/32,400; 3/10,260; and 2/9,000) (Fig. S3C).

In addition, for the competition experiments, we added parameters TgfβR2-λ and F_R. TgfβR2-λ is the replacement rate of labeled TgfβR2^−/− stem cells. F_R is a bias term affecting whether a labeled stem cell is the replacement cell (replacement factor). The replacement stem cell is chosen randomly from the other stem cells, as previously, except now chance of labeled = (number labeled × F_R)/[(number labeled × F_R) + number unlabeled]. Under neutral conditions, the replacement cell (clonal expansion) is randomly chosen from the remaining ISCs (F_R = 1). However, if the crypt contains competing WT and mutant ISCs, then a value of F_R > 1 increases the chance that the replacement cell is mutant, whereas a value of F_R < 1 decreases this chance. F_R is comparable to the probability of replacement (P_R) (22) in that both model survival bias, but F_R is a multiplier (range of values: 0–∞) instead of a probability (range of values: 0–1), and in our model, the replacement cell can be any of the cells instead of being one-dimensional where cells are only replaced by neighbors.

Statistics.

To calculate statistical significance, we used univariate linear regressions in StatistiXL on both datasets (WT and mutant) and compared the slopes of each regression with ANOVA. For the continuous labeling data, we performed the linear regression comparison on nontransformed data. For the pulse labeling crypt dominance data, we transformed the outcomes to logit(y) to fit the normality assumption and transformed the covariates to log(x) so that the relationship could be fit as linear. For the pulse labeling crypt succession data, we log transformed the outcomes so that the relationship could be fit as linear. For data from a single time point comparing WT to TgfβR2-mutant, the Student t test was used.

GSEA (32) was performed comparing data from our microarray (GSE58296 or GSE83423) with 1,000 gene set permutations on GSE51398 for secretory precursor cell gene expression by comparing genes in secretory progenitors with enterocytes greater than 2 logFC differentially expressed (176 genes) (34) and on GSE23672 for stem cell gene expression by comparing genes in lgr5-GFP^high cells with Lgr5-GFP^low cells greater than 1 logFC differentially expressed (195 genes) (33). Expression data from human enteroids was normalized with the Loess method across all samples.