A reserve stem cell population in small intestine renders Lgr5-positive cells dispensable

Two types of stem cells have been described in the small intestine based on location and cycling dynamics^1–4. Fast-cycling stem cells express markers including Lgr5, CD133 and Sox9^{1, 5, 6} and are present throughout the intestine. Also known as crypt base columnar cells (CBCs), these slender cells populate the crypt and villi within 3 days, and are interspersed among the Paneth cells that support them^{7, 8}. Slower-cycling stem cells, marked by enriched expression of Bmi1 or mTERT, represent a rarer cell population^{2, 9}. These cells form a descending gradient from proximal to distal regions of the intestine, such that they are more prevalent in the duodenum than in the colon. Despite their rarity, Bmi1-expressing stem cells are crucial for crypt maintenance².

To study the function of Lgr5-expressing cells, we replaced the first ATG codon of Lgr5 with two distinct cassettes. The first consisted of a dsRED-ires-creERT2 sequence to enable genetic lineage tracing studies by tamoxifen (TAM)-inducible expression of Cre in Lgr5-expressing cells (Supplementary Fig. 1a, Lgr5^CreER allele). The second cassette contained EGFP linked in frame to a human DTR cDNA (Supplementary Fig. 1a, Lgr5^DTR allele), producing a fusion protein. Consistent with previous reports, one injection of TAM in Lgr5^CreER;R26R mice marked Lgr5-expressing stem cells in a mosaic fashion and led to generation of labeled progeny for more than 60 days (Supplementary Fig. 1b). Expression of DTR-EGFP in Lgr5^DTR mice functioned as a reporter for Lgr5 expression (Fig. 1a) and also conferred diphtheria toxin (DT) sensitivity on CBCs. Expression of EGFP in mice carrying the Lgr5^DTR allele was detected at the membrane of cycling CBCs in every crypt (Supplementary Fig. 1c-e, CBCs are marked by asterisks).

We next set out to test the effects of eliminating Lgr5-expressing cells by administering DT to Lgr5^DTR mice. Twenty-four hours after DT administration, all EGFP-positive cells were depleted, including CBCs (Fig. 1a, 1b, 1j, 1k, 1p, 1q). Loss of Lgr5-expressing cells was further confirmed by the absence of Lgr5 mRNA (Fig. 1d, 1e) and was accompanied by extensive apoptosis at the base of the crypts, with shedding of dead cells into the lumen (Fig. 1n).

After sustained DT exposure for 10 days, both the EGFP reporter and Lgr5 mRNA were completely absent from the base of the crypts (Fig. 1c, 1f, and Supplementary Fig. 2), but strikingly, crypt architecture was comparable to controls (Fig. 1g, 1i, 1j, 1l). Proliferating CBCs were absent from the crypt (Fig. 1l, 1r), such that the crypt base was occupied mostly or entirely by Paneth cells (Supplementary Fig. 3a, 3b). The extensive apoptosis detected 24 hours after DT treatment had significantly decreased by day 10 (compare Fig. 1n with Fig. 1o) but was still detectable. No increase in crypt fission after DT treatment was observed by H&E staining at any time point (Fig 1g-i).

Because Lgr5-expressing cells have been proposed to play a critical role in renewal of the intestine, it was surprising that the architecture of the intestinal epithelium was essentially intact after ablation of Lgr5-expressing CBCs (Fig 1g-i). Within the villi, very little change in the total number of endocrine cells was observed (Supplementary Fig. 3c, 3d), and goblet cells were abundant in the crypts and villi (Supplementary Fig. 3g, 3h, 3j). Upon CBC ablation, Paneth cells were found at the bottom of the crypts and in some cases were mislocalized to the villi (Supplementary Fig. 3a, 3b and data not shown); additionally, migration of cells as assessed by BrdU pulse-chase labeling was normal (Supplementary Fig. 4). The only major difference from controls that we observed was in the secretory cell lineage; the number of chromogranin A positive enteroendocrine cells in the crypts doubled 10 days post DT (Supplementary Fig. 3e, 3f, 3i).

We did not detect any Lgr5-expressing CBCs using either the EGFP reporter or in situ hybridization after 10 days of DT (Fig. 1c, 1f and Supplementary Fig. 2), but it was still possible that a few CBCs could have escaped ablation and repopulated the epithelium, as a similar scenario was reported in c-Myc and Ascl2 conditional null animals^{10, 11}. To directly address this possibility, we visualized Lgr5-expressing cell activity during DT selection by producing Lgr5^DTR/CreER;R26R mice. These mutant mice carried two null alleles at the Lgr5 locus, one of which enabled ablation of Lgr5-expressing cells and the other enabled lineage tracing of any possibly remaining Lgr5-expressing cells. These mice died at postnatal day (P) 1, consistent with previous reports that Lgr5 null mice are not viable¹². To analyze the postnatal gut, we grew pieces of small intestine from embryonic day (E) 15 Lgr5^DTR/CreER;R26R embryos under the kidney capsule of immunocompromised mice for three weeks, at which point they formed crypts comparable to P17 intestine (Fig. 2a-e)¹³. Following 10 days of TAM treatment, columns of blue cells emanated from the crypt base, and progeny of Lgr5-expressing cells differentiated into all four major cell types of the intestinal epithelium (Fig. 2a-e). Concomitant administration of DT and TAM for 10 days eradicated all EGFP-positive CBCs (Fig. 2g), and no cells descended from Lgr5-expressing cells were observed (Fig. 2f), confirming that the Lgr5^DTR allele leads to complete elimination of these cells. Importantly, no abnormalities in graft morphology, differentiation or proliferation were observed in these mice compared to controls (Fig. 2a-j).

Although Lgr5-expressing cells were completely depleted within 24 hours of DT treatment, persistence of apoptotic bodies at the crypt base throughout the 10 day DT treatment suggested that Lgr5-expressing CBCs were continuously generated and eliminated during the treatment (Fig. 1n, 1o). This notion was supported by the quick recovery of Lgr5-expressing cells between 48 to 96 hours after the final dose of DT (Fig. 1s-v). To follow the fate of the newly generated Lgr5-expressing cells, mice implanted with Lgr5^DTR/CreER;R26R embryonic intestine fragments in the kidney capsule were allowed to recover for 6 days following 6 days of DT treatment. A row of blue cells emanated from the crypt base (Supplementary Fig. 5a), indicating that the newly formed Lgr5-expressing stem cells (Supplementary Fig. 5b, GFP-positive cells) gave rise to progeny that migrated out of the crypt. When the converse experiment was performed by injecting TAM for 6 days and then dosing with DT from days 6 to 12, blue cells were only present in the upper region of the villi (Supplementary Fig. 5c), indicating that progeny of Lgr5-expressing cells marked between day 1 and 6 migrated out of the crypts into the villus, but that during DT treatment between days 6 and 12, Lgr5-expressing stem cells were no longer available (Supplementary Fig. 5d, absence of GFP signal) to supply labeled (blue) progeny to replenish the epithelium.

To study the long-term effects of CBC ablation, we isolated crypts from Lgr5^DTR/+ mice to perform in vitro crypt organoid zcultures¹⁴. Crypts depleted of Lgr5-expressing CBCs by treatment for 10 days with DT, as indicated by absence of GFP expression, gave rise to organoids with similar efficiency as wild-type controls (Supplementary Fig. 6a, 6b). These could be passaged in vitro in DT for up to 2 months without losing their ability to expand and proliferate. No Lgr5-expressing (GFP-positive) cells were detected in organoid epithelium as long as the organoids were maintained in medium containing DT (Supplementary Fig. 6d). However, when DT was removed from the culture medium, Lgr5-expressing cells reappeared at the bottom of crypt-like structures within 3 days (Supplementary Fig. 6c, GFP-positive cells).

Because we found that Lgr5-expressing CBCs were dispensable for crypt maintenance, we next asked whether Bmi1-expressing stem cells were mobilized to compensate for the loss of the Lgr5-expressing stem cells. BMI1 regulates self-renewal of hematopoietic and neuronal stem cells¹⁵. We employed a GFP knock-in allele (Bmi1^GFP/+) to monitor Bmi1 gene expression¹⁶. Bmi1 expressing GFP-positive cells were most commonly observed at positions 3 to 6 from the crypt base (Fig. 3a), consistent with the Bmi1 mRNA expression pattern in the small intestine². Upon depletion of Lgr5-expressing CBCs in Lgr5^DTR/+;Bmi1^GFP/+ animals after 9 days of DT treatment, the number of GFP-positive cells per crypt increased three fold (Fig. 3a-d, and Supplementary Fig. 7a), and the proportion of crypts containing either single or multiple GFP-expressing cells increased by 40 percent compared to control animals (Supplementary Fig. 7b). Of note, 55 percent of the total number of GFP-positive crypts in Lgr5^DTR/+;Bmi1^GFP/+ animals now contained multiple GFP-positive cells (Fig. 3d and Supplementary Fig. 7b), compared with only 22 percent in Bmi1^GFP/+ control animals.

To trace the fate of cells descended from Bmi1-expressing cells after elimination of Lgr5-expressing CBCs, we generated a Bmi1CreER BAC transgenic allele (Supplementary Fig. 8). Labeling kinetics using the Bmi1CreER transgenic line crossed with the R26R reporter were identical to previously reported results using the Bmi1^CreER knock-in allele² (Fig. 3f). Bmi1CreER;R26R;Lgr5^DTR/+ animals were treated with alternating doses of DT and TAM per day for 7 days (Fig. 3e). Because Bmi1-expressing cells are most abundant in the first 5 cm of the duodenum, we focused our analysis on this region. Consistent with the increased number of Bmi1-expressing cells (Supplementary Fig. 7a), the proportion of LacZ-positive crypts (either partially or fully labeled) also increased 34 percent upon loss of Lgr5-expressing CBCs (Supplementary Fig. 7c). The most dramatic difference was in the number of fully labeled crypts. Only 2.3 percent of crypts were fully labeled in Bmi1CreER;R26R control animals during a 6 day lineage tracing period, which was comparable with previous studies using a Bmi1^CreER knock-in allele². Upon loss of Lgr5-expressing CBCs, the number of fully labeled crypts increased approximately 15-fold (Fig. 3h, 3i and Supplementary Fig. 7c.). These results indicate that in the absence of Lgr5-expressing cells, Bmi1-expressing cells are capable of directly giving rise to all intestinal cell types without going through Lgr5-positive intermediate cells. However, Bmi1-expressing stem cells did not give rise to an increased number of labeled crypts in more distal regions of small intestine and colon upon loss of Lgr5-expressing CBCs (Fig 3f, 3g), indicating that different stem cell pools must compensate for the loss of Lgr5-expressing stem cells in distal regions of the gut.

Lastly, we tested whether Bmi1-expressing cells give rise to Lgr5-expressing cells under normal conditions. Since Bmi1- and Lgr5-expressing cells represent distinct though overlapping cell populations, we carried out a series of short term pulse-chase experiments using Bmi1CreER;R26R;Lgr5^DTR/+ animals. Twenty-four hours after TAM administration, most of the β-galactosidase (β-gal)-positive cells appeared as individuals, reflecting the normal pattern of Bmi1 expression (Fig. 4a) in the initially labeled cells. Bmi1-expressing cells (β-gal positive) overlapped with Lgr5-expressing cells (GFP positive) between positions 1 to 6 at the crypt base; the double-positive cells peaked at positions 3 and 4 (Fig. 4a-c). This observation is consistent with a previous report stating that Bmi1 mRNA expression (via qPCR analysis) was readily detectable in Lgr5-positive cells¹¹. Later, between 48–72 hours, clonal expansion from Bmi1-expressing cells was evident, as β-gal/GFP double-positive cells now appeared as doublets or triplets (Fig.4d-i). We scored a total of 500 crypts at each time point and found that while a few cells were β-gal/GFP double positive (i.e., expressing both Bmi1 and Lgr5) at 24 hours after TAM induction, this number doubled at 48 hours (Fig. 4j,k). Similarly, lineage tracing from Bmi1-expressing cells carried out in mice treated for 6 days with DT and allowed to recover for 72 hours demonstrated that newly formed Lgr5-positive cells at the bottom of the crypts arose from Bmi1-expressing cells (Fig 4m-o). Together, these data show that Bmi1-expressing cells can give rise to Lgr5-expressing cells both under normal physiological conditions and following insults that deplete CBCs. Similar to our observation, mTERT-expressing stem cells could also give rise to Lgr5-positive cells over a 5 day lineage tracing period⁹.

Our data support the existence of two stem cell pools in the epithelium of the small intestine: an actively proliferating stem cell compartment responsible for the daily maintenance of the intestine epithelium that is characterized by the expression of Lgr5, Ascl2, and Olfm4^{1, 11, 17} and a distinct pool of stem cells expressing Bmi1. Our results lend support to the two-stem-cell pool model that is based on computational approaches¹⁸, and provide experimental evidence for recent models predicting that the intestine could fully recover after complete elimination of cellular subpopulations deemed to be functional stem cells¹⁹. Our data do not support the recent proposal that Bmi1-expressing cells are exclusively a subset of Lgr5-expressing cells¹¹; rather they indicate that under normal circumstances, Bmi1-positive stem cells are upstream of rapidly cycling, Lgr5-expressing stem cells and replenish the pool of active stem cells, either to avoid exhaustion of actively cycling stem cells or to prevent the accumulation of damaged cells that may lead to the development of tumors. Importantly, we also demonstrate that when the Lgr5-expressing cell compartment is eliminated by DT treatment, the Bmi1-expressing cells can increase in number, presumably as a compensatory mechanism. Under these conditions, Bmi1-expressing cells contribute directly to the generation of all cell types of the intestinal epithelium to produce a functional organ until the rapidly-cycling stem cell compartment is able to recover. While it has been suggested that Bmi1-expressing stem cells are quiescent², this remains to be conclusively demonstrated.

Distinct stem cell pools with differing cycling dynamics have previously been observed in the hair follicle and in blood, organs that, like the intestine, undergo regular bouts of proliferation and regeneration^20–22. The factors that regulate the interplay between discrete populations of stem cells, and the precise hierarchical relationships among such populations, remain to be characterized. While we have found that loss of Lgr5-positive cells is sustainable under short-term conditions in vivo, it remains to be determined if such a scenario can persist for longer periods of time in the animal. Interestingly, depletion of Paneth cells, which are thought to be important for the maintenance of CBCs⁷, can be tolerated by mice for over 6 months without significant structural defects of the epithelium^{23, 24}, supporting the idea that the intestine can function normally in the absence of CBCs. It will be important to determine how different stem cell populations sense the activity of other populations, whether rapidly cycling cells can repopulate more quiescent stem cell populations, and whether additional sub-populations of stem cells exist.