Mapping early fate determination in Lgr5+ crypt stem cells using a novel Ki67-RFP allele

Authors


Abstract

Cycling Lgr5+ stem cells fuel the rapid turnover of the adult intestinal epithelium. The existence of quiescent Lgr5+ cells has been reported, while an alternative quiescent stem cell population is believed to reside at crypt position +4. Here, we generated a novel Ki67RFP knock-in allele that identifies dividing cells. Using Lgr5-GFP;Ki67RFP mice, we isolated crypt stem and progenitor cells with distinct Wnt signaling levels and cell cycle features and generated their molecular signature using microarrays. Stem cell potential of these populations was further characterized using the intestinal organoid culture. We found that Lgr5high stem cells are continuously in cell cycle, while a fraction of Lgr5low progenitors that reside predominantly at +4 position exit the cell cycle. Unlike fast dividing CBCs, Lgr5low Ki67 cells have lost their ability to initiate organoid cultures, are enriched in secretory differentiation factors, and resemble the Dll1 secretory precursors and the label-retaining cells of Winton and colleagues. Our findings support the cycling stem cell hypothesis and highlight the cell cycle heterogeneity of early progenitors during lineage commitment.

Synopsis

image

A Ki67-RFP knock-in allele highlights the heterogeneity of Lgr5+ stem- and progenitor cells based on in vivo cell cycle profiles.

  • Ki67-RFP allele allows identification and visualization of proliferating cells in vivo.
  • Lgr5high intestinal stem cells are continuously in the cell cycle.
  • Quiescent Lgr5low “+4” cells are secretory precursors.

Introduction

The intestinal epithelium is continuously replenished by stem cells residing at the highly proliferative crypts (Clevers, 2013). Wnt signaling constitutes the major regulator of intestinal homeostasis and its target gene Lgr5 identifies crypt base columnar cells (CBCs) as stem cells (Barker et al, 2007). The crypt bottom provides a unique niche that controls stem cell behavior and numbers. Upon loss of contact to their niche, stem cells differentiate into transit-amplifying cells that migrate upward to generate absorptive enterocyte and the secretory goblet, enteroendocrine and tuft cells. Paneth cells localize to the crypt bottom and secrete bactericidal products. They also serve as niche cells that support the neighboring CBCs (Sato et al, 2011). Both Paneth cells and the surrounding mesenchyme provide Wnt ligands that—together with Notch signaling—promote stem cell self-renewal (Farin et al, 2012). Notch signaling plays a key role during differentiation by inhibiting the secretory fate (van Es et al, 2005). BMP signaling forms an inverse gradient with Wnt activity and promotes differentiation of cells that lose contact with their niche (Haramis et al, 2004), while EGF/ErbB signaling is a major inducer of proliferation in the crypts (Wong et al, 2012).

Lgr5+ CBCs are highly proliferative, yet DNA label retention experiments suggest the presence of slow dividing crypt progenitors as well (Potten et al, 1978, 2002; Li & Clevers, 2010). A quiescent multipotent population is suggested to reside at the “+4” position just above the CBCs and express several stem cell markers including Bmi1, Tert and Hopx but has diminished Lgr5 expression (Sangiorgi & Capecchi, 2008; Montgomery et al, 2011; Takeda et al, 2011; Tian et al, 2011; Yan et al, 2012). In several studies, Lgr5+ CBC stem cells have been found to share these markers (Itzkovitz et al, 2012; Munoz et al, 2012; Powell et al, 2012; Wong et al, 2012; Wang et al, 2013). The use of a Histone label retention-based lineage tracing provided evidence for an alternative population, which resides in the crypt, expresses Lgr5 but exclusively generates secretory cells (Buczacki et al, 2013). In support, this study shows that some of the Lgr5-expressing CBCs lack cell cycle gene expression. A functionally equivalent population resides prominently at the “+5” position and is enriched in expression of the Notch ligand Delta-like 1 (Dll1) (van Es et al, 2012). In the traditional view, differentiation occurs upon exit from the Paneth cell zone, at the so-called “origin of differentiation” around position “= 5” where cells choose between the secretory and absorptive fates (Bjerknes & Cheng, 1981). Inhibition of Notch signaling leads to secretory differentiation of transit-amplifying cells and is coupled to cell cycle exit (van Es et al, 2012). However, label-retaining cells (LRCs) are reported within the Paneth cell zone and generate secretory cell types (Buczacki et al, 2013). Recent studies in ES cells point out the importance of the cell cycle and lineage commitment, suggesting that lineage commitment might precede differentiation of stem cells (Pauklin & Vallier, 2013).

The intestinal epithelium displays significant plasticity, as the quiescent +4 cells are reported to replenish the CBC pool when Lgr5+ cells are selectively killed (Tian et al, 2011). Similarly, Dll1+ and label-retaining cells can revert to a stem cell fate upon radiation-mediated loss of proliferating CBCs (van Es et al, 2012; Buczacki et al, 2013). These results suggest that intestinal crypt populations can be interconverted both during homeostasis and regeneration. To gain insight into the cell cycle dynamics of Lgr5+ CBCs, we have established a system to visualize actively proliferating cells using the proliferation-specific expression of the mKi67 gene and characterized the early fate choices of intestinal stem cells.

Results

Heterogeneous cell cycle dynamics of small intestinal CBCs

In order to understand the cell cycle dynamics of adult intestinal stem cells, we analyzed proliferation of CBCs on intestinal sections of Lgr5-GFP mice using double immunofluorescence analysis (Barker et al, 2007; Yan et al, 2012). KI67 is a nucleolar protein specifically expressed in G1-S-G2-M phases of the cell cycle, but is absent in the G0 phase, making it an excellent marker of proliferation (Hutchins et al, 2010). As reported, the vast majority of the Lgr5+ CBCs was proliferating and expressed the cell cycle marker KI67 (88.0 ± 2.0%; Fig 1A and B). However, we observed a small but significant proportion lacking detectable levels of the KI67 protein (Fig 1A). A quiescent stem cell population is proposed to reside at the “+4” position. Lgr5+ KI67 cells were located throughout the crypt bottom (Fig 1C and D). We quantified the number of Lgr5-, KI67-, and chromogranin A (CHGA)-expressing cells along the crypt axis (see 'Materials and Methods'). Lgr5+ CBCs expressing the enteroendocrine and label-retaining cell marker CHGA have previously been reported as secretory precursors (Sei et al, 2011; Buczacki et al, 2013). We found that CHGA+ CBCs are concentrated at +3/+4/+5 positions (2.45 ± 0.01), while KI67 CBCs residing at +1/+4 are CHGA (Fig 1C and D). However, the majority of the CHGA+ crypt cells expresses relatively low levels of (if any) Lgr5 and is distributed along the crypt with a peak at +5 position (15.0 ± 5.7% among CHGA+ crypt cells). Lysozyme (LYZ)-expressing Paneth cells were post-mitotic and did not express KI67 (0 ± 0% among LYZ+).

Figure 1.

Figure 1. Characterization of the cell cycle of Lgr5-expressing cells

  • A, BMost of the Lgr5+ (Lgr5-GFP, green) cells have high levels of KI67 (red; white arrowheads) as seen on transverse sections. A small fraction of Lgr5+ stem cells do not express the KI67 protein (yellow arrowheads). Lysozyme (LYZ, blue)-expressing Paneth cells mark the Paneth cell zone and are negative for Lgr5 and KI67. Scale bar, 20 μm. Quantification of the results is shown in (B).
  • CA cross section of an intestinal crypt showing Lgr5+ stem cells (Green), KI67 (Red), chromogranin A (CHGA, white) and the nucleus (DAPI, blue). Scale bar, 10 μm.
  • DQuantification of the position of Lgr5+ and CHGA+ cells seen in (C). pos: position, total: sum of Lgr5+ (left) and CHGA+ (right) cells in each position. Each square shows the percentage of cells with a given antigen expression at the respective position among Lgr5+ (left) and CHGA (right)-expressing cells. Lgr5 expression is visualized using the endogenous Lgr5GFP florescence.

Ki67RFP marks proliferating cells

In order to visualize actively cell cycling in live cells, we generated a Ki67RFP knock-in allele by introducing a TagRFP red fluorescent protein in frame at the C-terminus of the Ki67 coding sequence (Fig 2A). As a result, fluorescence is directly linked to the KI67 protein and hence to cell cycle activity. The Ki67RFP allele was transmitted at the expected Mendelian ratios, and homozygous mice were viable and fertile.

Figure 2.

Figure 2. Generation and characterization of the Ki67RFP knock-in mouse

  • AThe Ki67RFP targeting construct.
  • B, CKi67RFP expression can be visualized on live section from the small at low (B) and high (C) magnification.
  • DRFP-expressing cells from Ki67RFP-expressing small intestines can be identified and sorted using FACS.
  • E, FGraph showing Hoechst 34580 staining on dissociated live intestinal crypt cells. The x-axis shows the dye intensity, which is directly correlated to the DNA content of the cells. Red and blue lines indicate RFP+ and RFP fractions, respectively. Quantification of results is shown in (F). All indicates the profile of total live crypt cells.
  • GQuantitative PCR analysis of sorted Ki67RFP+ and Ki67RFP− population. Levels in sorted RFP+ cells is compared to RFP cells from the same mouse to indicate the fold change (n = 3). Results are normalized to expression levels in crypt, which is 1.
  • H, ISorted single cells were stained using anti-KI67 antibody. RFP+ cells express KI67 while most RFP cells lack KI67 expression. Quantification of the results is shown in (I).
  • J, KSorted RFP+ cells can initiate organoid cultures while RFP cells are less efficient. Quantification of the results is shown in (K).

Data information: Error bars indicate the standard deviation. In (D), FL1 channel is used as a blank channel to visualize autofluorescence. Scale bars: 100 μm (B), 20 μm (C), 10 μm (H), 200 μm (J). In (B) and (C), the gamma settings were adjusted for better visualization of the dim RFP fluorescence.

Analysis of TagRFP (RFP) fluorescence on semi-thick vibratome sections from adult mice revealed expression in multiple proliferative tissues, including the spleen, thymus, brain, hair follicle, and colon. For the current study, we focused on the small intestine and characterized RFP expression by fluorescent microscopy and FACS (Fig 2). The fluorescent signal was localized to the crypt as visualized on vibratome sections (Fig 2B and C). In order to find out whether proliferating cells express RFP, we dissociated purified intestinal crypts (Supplementary Fig S1) from the Ki67RFP mice and performed FACS sorting. The DNA content of dissociated live crypt cells from the Ki67RFP mice was measured using Hoechst 34580 (Fig 2D–F) staining. On average, 12.3% (± 3.3) of the crypt cells were in S-M phases of the cell cycle. 52.9% (± 9.8) of the Ki67RFP+ cells were in S-M phase of the cell cycle, confirming that RFP expression correlates with cell cycle progression. In sharp contrast, only 4.8% (± 2.5) of the Ki67RFP− cells were in S-M phases of the cell cycle indicating a lack of cell cycle activity (Fig 2E and F). Quantitative PCR analysis revealed a striking enrichment of Ki67 (27.9-fold; 19.2 ± 7.1 versus 0.7 ± 1.0; P = 0.043) and Ccnb2 (3.5-fold; 3.6 ± 1.8 versus 1.0 ± 1.1; P = 0.003) expression in RFP+ fraction compared to the RFP cells, supporting proliferation-specific expression of the Ki67RFP allele (Fig 2G). The enteroendocrine and label-retaining cell marker ChgA (25.9–fold; 0.2 ± 0.1 versus 4.1 ± 1.6; P = 0.048) and the absorptive enterocyte marker Alpi (2.3-fold; 0.3 ± 0.3 versus 0.8 ± 0.3; P = 0.035) were enriched in the Ki67 fraction, while Lyz levels (2.6-fold; 0.3 ± 0.4 versus 0.8 ± 1.5; P = 0.5073) were not statistically different (Fig 2G). In addition, we failed to find any difference in the “+4” marker Bmi1 (1.2-fold; 0.9 ± 0.5 versus 1.0 ± 1.0; P = 0.842). Immunostaining analysis using antibodies against the KI67 antigen on sorted RFP populations confirmed RFP+ cells express KI67 (96.6 ± 5.6%), while the majority of RFP cells do not (14.4 ± 7.6%; Fig 2H and I). We concluded that Ki67RFP+ cells are actively in the cell cycle, and conversely, the vast majority of dividing cells express the Ki67RFP allele. Intestinal stem cells are capable of establishing organoid cultures that recapitulate the intestinal epithelium (Sato et al, 2009). Sorted RFP+ cells robustly initiated organoid cultures (0.43 ± 0.28%), while the RFP fraction had diminished capacity (0.05 ± 0.04%), suggesting that the majority of the stem cell activity resides within the proliferating epithelial fraction (Fig 2J and K).

Lgr5+ CBCs with distinct cell cycle features can be isolated using the Lgr5-GFP;Ki67RFP double knock-in mice

We generated Lgr5-GFP;Ki67RFP double knock-in mice to discriminate cycling and quiescent Lgr5+ CBCs. Both reporters were clearly visible on freshly isolated intestinal crypts (Fig 3A). We dissociated small intestinal crypts and performed FACS in an attempt to isolate the Ki67 putative quiescent CBCs. We observed that while most of the Lgr5+ cells are cycling, 10.2% (± 1.9%) along the GFP gradient lack Ki67RFP expression consistent with KI67 antigen expression in vivo (Fig 3B, K gates). We have previously identified stem cells and their progeny using GFP expression from the Lgr5 locus (Munoz et al, 2012). Here, we focused on stem cells (Lgr5high) and their immediate progeny (Lgr5low) and we sorted KI67+ and KI67 subpopulations. As a result, we describe four populations: Lgr5highKi67+ dividing stem cells, Lgr5highKi67 putative quiescent stem cells, Lgr5lowKi67+ dividing and Lgr5lowKi67 putative quiescent crypt progenitors.

Figure 3.

Figure 3. Identification of CBCs with distinct KI67 expression

  • AIntestinal crypts from control (WT), Ki67RFP, and Lgr5GFPKi67RFP mice used for the FACS sorting. Scale bar, 80 μm. The gamma settings were adjusted for better visualization of the dim RFP fluorescence.
  • BWe defined 4 populations of CBCs on FACS using different GFP and RFP expression levels: Lgr5GFP high Ki67RFP positive (LhighK+), Lgr5GFP high Ki67RFP negative (LhighK), Lgr5GFP low Ki67RFP positive (LlowK+), Lgr5GFP low Ki67RFP negative (LlowK).
  • C, DOrganoid cultures derived from sorted CBC populations (C) and their quantification (D). Error bars indicate the standard deviation. Scale bar, 80 μm.

We interrogated whether differences in KI67 levels reflect differences in stem cell potential. Lgr5highKi67+ and Lgr5highKi67 populations were similarly potent in initiating organoid cultures (single cell plating efficiency 1.23 ± 0.83% and 0.73 ± 0.68%, respectively; Fig 3C and D). Lgr5lowKi67+ cells initiated culture at lower levels compared to Lgr5high CBCs (0.12 ± 0.17, Fig 3C and D). In sharp contrast, Lgr5lowKi67 cells did not display clonogenic ability implying a loss of stemness.

Global gene expression analysis of CBC populations

To elucidate the molecular features of cycling and quiescent Lgr5+ crypt cells, we sorted samples from Lgr5high Ki67+ (n = 2), Lgr5high Ki67 (n = 2), Lgr5low Ki67+ (n = 4), and Lgr5low Ki67 (n = 2) and generated a global gene expression data set using Affymetrix chips. We compared the gene expression pattern of all 4 populations to document genes differentially expressed between populations using the R2 database (http://r2.amc.nl, AMC, P < 0.01, ANOVA test) and identified 1,151 differentially expressed genes (Fig 4A). We compared the molecular signatures of Lgr5high and Lgr5low populations (Fig 4B and C). Expression of Lgr5 (5.2-fold) as well as several genes previously identified as stem cell-specific were enriched more than twofold in GFPhigh populations (24 among Ki67+ and 19 among Ki67) compared to GFPlow. These included Tnfrsf19 (Fafilek et al, 2013; Stange et al, 2013), Fstl1, Nav1, and Cttnbp2 (Munoz et al, 2012) (Fig 4B and C). We employed the GSEA analysis to evaluate the distribution of the published Lgr5GFPiresCreER+ stem cell (Munoz et al, 2012) and Wnt (Sansom et al, 2004) signature genes between the Lgr5high and Lgr5low populations to confirm the efficiency of our Lgr5GFP;Ki67RFP double sorting paradigm to discriminate stem cells from their progeny (Fig 4D). Analysis revealed a high enrichment of both signatures in Lgr5highKi67+ (NES: 3.85 and 2.60) as well as Lgr5highKi67 populations (NES: 4.00 and 2.72), validating Lgr5-GFPhigh cells as stem cells.

Figure 4.

Figure 4. Comparison of CBC populations

  1. Heatmap generated using the differentially expressed genes between CBC populations calculated as significantly different (ANOVA test, P < 0.01). Red indicates higher expression.
  2. Two-group plotters comparing Ki67+Lgr5high to Ki67+Lgr5low (left) and Ki67Lgr5high to Ki67Lgr5low (right).
  3. Quantification of the number of genes that are significantly different (ANOVA, P < 0.01) between groups and is expressed at least twofold (black bars) of 1.5-fold higher (gray bars) in one of the groups.
  4. Gene set enrichment analysis (GSEA) reveals high similarity between the top 200 stem cell genes (Munoz et al, 2012) and 58 genes upregulated upon deletion of APC in the small intestine (Sansom et al, 2004) to the genes enriched in Ki67+Lgr5high versus Ki67+Lgr5low and Ki67Lgr5high versus Ki67Lgr5low. Results are displayed in log2.

Lgr5high stem cells are continuously in cell cycle

To understand whether the differences in Ki67RFP expression reflects a difference in the cell cycle status of CBC populations, we compared the expression of genes under the GO term “cell cycle” that are statistically different (Fig 5A). 49 genes were differentially expressed including Ccnb1, Cnnb2, and AurkA. Even though RFP+ cells express the KI67 protein (Fig 2), we observed that the mKi67 RNA was not significantly enriched in any of the CBC populations. qPCR analysis confirmed that mKi67 and Ccnb2 genes were expressed at strikingly higher levels in all Lgr5 populations compared to the villus where most cells are terminally differentiated. Their expression was not significantly different between the two groups of Lgr5high stem cells but was significantly less in Lgr5lowKi67 cells compared to the Lgr5lowKi67+ cells consistent with the microarray data (Fig 5A and B). In agreement with their shared organoid-initiating ability, Lgr5highKi67+ and Lgr5highKi67 stem cells displayed a very high correlation in their gene expression pattern (Fig 5C and D). The low number of genes that are differentially expressed between Lgr5highKi67+ (0 gene > twofold and seven genes over 1.5-fold) and Lgr5highKi67 (1 gene > twofold and 17 genes > 1.5-fold) suggests that the populations are functionally identical (Fig 5C and D). Differences were much more pronounced between Lgr5lowKi67+ (four genes > twofold and 60 genes over 1.5-fold) and Lgr5lowKi67 (161 genes > twofold and 257 genes over 1.5-fold) populations (Fig 5C and D). Based on the overlap in their molecular signatures of Lgr5highKi67+ and Lgr5highKi67 populations, enrichment of stem cell genes and high levels of expression of cell cycle-related genes, we suggest that both classes of Lgr5high intestinal stem cells are continuously cycling. Lgr5lowKi67 cells display a distinct cell cycle pattern intermediate between other Lgr5 populations and differentiated cells.

Figure 5.

Figure 5. The cell cycle dynamics of CBC populations

  1. Heatmap displaying genes with the combined GO term “cell cycle” that are differentially expressed between CBC populations (ANOVA test, P < 0.01). Red indicates higher expression.
  2. qPCR analysis for the expression of cell cycle indicators mKi67 and Ccnb2 in CBC populations, crypt, and the villus. Results are normalized to the expression level in the crypt, which is displayed as 0. Error bars indicate the standard deviation.
  3. Quantification of the number of genes that are significantly different (ANOVA, P < 0.01) between groups and is expressed at least twofold (black bars) of 1.5-fold higher (gray bars) in one of the groups.
  4. Two-group plotters comparing Lgr5highKi67+ to Lgr5highKi67 (left) and Lgr5lowKi67+ to Lgr5lowKi67 (right).

Lgr5lowKi67 cells are early secretory precursors

To better understand the program controlling self-renewal and differentiation, we focused on the 61 genes annotated to have “transcription factor”, “transcription regulator activity”, or “transcriptional repressor activity” and are differentially expressed among Lgr5 populations (Fig 6A). Among those, 24 were enriched in Lgr5high cells (Zfp202, Maged1, Lbh, Cbfb, Ctnnb1, Smad5, Zscan2, Fem1b, Tcfap2, Tfdp2, Tcf7, Brca2, Sox4, Trim24, Ehf, Dach1, Gtf2i, Notch1, Atoh8, Mycl1, Atm, Bclaf1, Nr2e3, and Ascl2) including Ascl2, a major regulator of stem cell identity (Fig 6A) (van der Flier et al, 2009). We identified 20 transcription factors in the Lgr5lowKi67 population (Foxj2, Runx1, Prox1, Foxa3, Mxd4, Hmx2, Pou2f3, SpiB, Pax6, Nkx2-2, St18, Pax4, Neurod1, Isl1, Runx1t1, Creb3 l4, Fev, Spdef, Neurog3, and Atoh1) including Atoh1, a master regulator of secretory differentiation (Yang et al, 2001; van Es et al, 2005; Shroyer et al, 2007). Several factors implicated in secretory differentiation were in this group. Pax4, NeuroD1, Nkx2-2, and Isl1 are inducers of endocrine differentiation, while Spdef is required for both the Paneth and goblet cell differentiation (Naya et al, 1997; Larsson et al, 1998; Jenny et al, 2002; Desai et al, 2008; Gregorieff et al, 2009). SpiB is an important player in differentiation of M cells, which are derived from Lgr5+ stem cells, but are rare in the intact intestine (Kanaya et al, 2012; de Lau et al, 2012). As Lgr5lowKi67 cells express significantly higher levels of ChgA (7.0-fold more than Lgr5lowKi67+ and 6.2-fold more than Lgr5highKi67, Supplementary Fig S2) and likely represent the +4/5 cells, we analyzed the expression of reported quiescent stem cell markers. Tert, Hopx, and Bmi1 were not significantly higher in any of the populations while Lrig1 was expressed slightly higher in Lgr5high cells excluding them as specific markers of a slow-dividing CBC population (Fig 6B) (Montgomery et al, 2011; Yan et al, 2012; Wong et al, 2012). This confirms several recent reports on the shared expression of these markers between slow-cycling “+4” cells and Lgr5 CBC cells (Itzkovitz et al, 2012; Munoz et al, 2012; Powell et al, 2012; Wong et al, 2012; Wang et al, 2013).

Figure 6.

Figure 6. Lgr5lowKi67 population corresponds to the secretory precursors

  1. Heatmap displaying genes with the combined GO terms “transcription factor” and “transcriptional regulatory activity” that are differentially expressed between CBC populations (ANOVA test, P < 0.01). Red indicates higher expression.
  2. Plots displaying the distribution of the proposed quiescent stem cell markers to CBC populations. The numbers on the y-axis show the normalized expression values on microarrays in log2. Each dot represents a separate sample.
  3. Gene set enrichment analysis (GSEA) analysis. Genes significantly enriched in Lgr5lowKi67 population was used to generate a molecular signature, which was compared to published data sets including the Dll1+ (van Es et al, 2012), Paneth (PC; van Es et al, 2012), enteroendocrine (EEC; van Es et al, 2012) and label-retaining cell (LRC; Buczacki et al, 2013) expression profiles. NES: Normalized enrichment score.

Lgr5 subpopulations also displayed a distinct profile of signaling pathway components. Inhibition of Notch signaling results in an upregulation of Atoh1 expression and induces secretory differentiation. Consistently, Notch1 mRNA was high in stem cells and was virtually absent in Lgr5lowKi67 cells. Another Lgr5lowKi67-enriched gene, St18, is a tumor suppressor that during pancreatic development inhibits Notch signaling (Wang et al, 2007). The notion of secretory differentiation in Lgr5lowKi67 cells is supported by other factors highly enriched in this population. Enteroendocrine markers ChgA and ChgB, as well as hormones expressed by enteroendocrine subtypes (Stt, Gcg) and some of their regulators (Rfx6), are the most prominent factors (Supplementary Fig S2A). Confocal and qPCR analysis confirmed that some of the Lgr5+ cells express ChgA (Supplementary Fig S2B–D) as well as somatostatin (Stt) (Supplementary Fig S2D). Similarly, Kit and Muc2, Paneth and goblet cell factors, respectively, are also among the enriched genes (Supplementary Fig S2A). Taken together, the results imply that the Lgr5lowKi67 cells are in the secretory lineage.

Lgr5+ stem cells generate secretory cells through intermediate populations, that is the Dll1+ “+5” cells and Winton's slow dividing/label-retaining cells (van Es et al, 2012; Buczacki et al, 2013). To investigate the relationship of Lgr5lowKi67 cells with the proposed secretory precursors and related intestinal populations, we used the GSEA analysis (Fig 6C). We compared the genes significantly enriched in Lgr5lowKi67 cells to the ranked gene lists of Paneth cells, enteroendocrine cells, Dll1 cells (van Es et al, 2012), and label-retaining cells (Buczacki et al, 2013) (see 'Materials and Methods'). We employed the normalized enrichment score (NES) as a mean to evaluate their similarity to each population. Lgr5lowKi67 population was most similar to the Dll1+ population, followed by the Paneth cells, enteroendocrine cells, and the label-retaining cells (Fig 6C; NES: 4,20, 3,68, 3,20, respectively). These results imply that the Lgr5lowKi67 population represents the described secretory progenitor populations.

Lgr5 Ki67 crypt cells display a differentiated cell signature

Lgr5 is gradually down-regulated early during differentiation. In order to analyze Lgr5 crypt populations with distinct cell cycle features, we generated the Ki67RFP;Lgr5GFPDTR double knock-in mice, where GFP is expressed by every Lgr5+ cell (Tian et al, 2011). We isolated Lgr5 Ki67+ (Lgr5 K+) and Lgr5Ki67 (Lgr5K) populations and analyzed the differences in their gene expression pattern using qPCR (Supplementary Fig S3). Lgr5 mRNA expression was constantly enriched in Lgr5GFPDTR sorted cells on average by 32-fold (28.7 ± 14.1 versus 0.9 ± 0.3). As expected, proliferation markers Ki67 (16.8 ± 13.4) and Ccnb2 (2.50 ± 0.85) were highly enriched in Lgr5 K+ compared to Lgr5 K population. Transcription factors regulating secretory differentiation, Atoh1 (0.04 ± 0.04) and NeuroD1 (0.01 ± 0.00), were exclusively expressed in Lgr5 K cells, consistent with an inverse correlation between secretory differentiation and cell cycle progression. Paneth cell marker lysozyme (Lyz; 0.29 ± 0.15), the goblet cell marker Gob5 (0.09 ± 0.04), and the enteroendocrine/secretory progenitor marker chromagranin A (ChgA; 0.01 ± 0.00) were similarly enriched in non-dividing Lgr5 crypt fraction. We observed an enrichment of Fabp2 (0.46 ± 0.01) and Alpi (0.44 ± 0.12) in Lgr5 K cells compared to the Lgr5 K+ population. Both of these genes are upregulated in enterocytes upon differentiation and are expressed at lower levels in the crypt. Finally, even though we cannot exclude a rare quiescent population with high levels of Bmi1 expression, we failed to detect a correlation between cell cycle progression and Bmi1 expression, which is slightly higher in Lgr5 K cells (0.59 ± 0.02).

Discussion

We generated a model where proliferating cells can be genetically identified, visualized and isolated for molecular and cellular analysis. The Ki67RFP allele is specifically expressed in dividing cells at all cell cycle phases and can be used to dissect specific populations into proliferating and quiescent sub-fractions. GFP expressing stem cell reporters are available for several organs where Ki67RFP is expressed (e.g. brain, hair follicles), making it a valuable tool for the stem cell community.

We performed the first analysis in one of the best-characterized somatic stem cell systems, the intestinal crypts, where proliferation must be tightly controlled to maintain stem cell numbers as well as the required output of differentiating daughter cells. The number of divisions that a fast cycling intestinal stem cell undergoes during mammalian life is high. For this reason, it has been assumed that an upstream/alternative stem cell population exists that has exited the cell cycle into quiescence. We isolated CBC populations using Lgr5 as an indicator of stemness and KI67 protein expression as an indicator of cell cycle progression. Our results indicate that Lgr5high stem cells display little variation in their cell cycle dynamics. Even though a KI67 population exists, its global expression pattern mimics that of the dividing stem cells. This is further supported by high expression of Ki67 and Ccnb2 in both Lgr5 populations. These levels are similar to the crypt average and much higher than that of the differentiated villus cells. Furthermore, both populations can efficiently initiate organoid cultures displaying comparable “stemness”. We conclude that Lgr5high CBC stem cells are continuously in the cell cycle. The difference in fluorescence levels of the Ki67RFP allele may reflect a Gaussian distribution in a uniformly cycling population. Alternatively, a transient exit from the cell cycle after mitosis could result in the absence of the KI67 protein in Lgr5+ cells that soon continue to proliferate. We have observed that a percentage of RFP cells express the KI67 protein supporting a gap period during the maturation of the TagRFP protein or its accumulation during the re-entry into the cell cycle.

The Lgr5lowKi67 cells, however, display an intermediate characteristic with significantly less cell cycle gene expression compared to both types of Lgr5high stem cells and Lgr5lowKi67+ cells. Lgr5lowKi67 cells reside in the differentiation zone and their expression pattern closely resembles that of the LRCs (Buczacki et al, 2013) and the Dll1+ cells (Stamataki et al, 2011; van Es et al, 2012). In agreement to their similarity to the LRCs, CHGA expressing Lgr5lowKi67 cells are distributed in the crypt with a peak at +4/5. It is likely that Lgr5lowKi67 population overlaps with both LRCs and Dll1+ progenitors (Fig 7). Lgr5low cells stall in the cell cycle and down-regulate Ki67 expression on transition from dividing stem cells into the slow dividing secretory progenitors. LRC cells can persist for days prior to differentiation into enteroendocrine or Paneth cells. Goblet cells and tuft cells rarely arise from LRCs and are likely direct descendants of Dll1+ cells (Stamataki et al, 2011; van Es et al, 2012; Buczacki et al, 2013). Lgr5lowKi67+ cells lose their stem cell signature, maintain Notch receptor expression and lack transcription factors implicated in secretory differentiation. Even though we cannot exclude the contribution of Lgr5lowKi67 cells, it is tempting to speculate that the Lgr5lowKi67+ are the main intermediate between the stem cells and enterocytes.

Figure 7.

Figure 7. Stem and progenitor cell compartmentalization in the small intestine

Paneth cells mark the stem cell zone where dividing Lgr5high stem cells reside. In the “zone of differentiation”, progenitors are specified to the absorptive or the secretory lineages (Bjerknes & Cheng, 1981). The “+4” markers are shared between both Lgr5high stem cells and Lgr5low progenitors independent of their cell cycle status. Lgr5lowKi67 intermediate secretory precursors arise from Lgr5high stem cells and overlap with Dll1+ cells, the common precursors of the Paneth, goblet, and the enteroendocrine cells (van Es et al, 2012). The label-retaining cells are the longest-lived secretory precursors that differentiate into Paneth and enteroendocrine cells (Buczacki et al, 2013). Both LRCs and Dll1+ can re-enter the cell cycle to generate multiple secretory cells, displaying limited proliferation potential (secretory precursors, red nucleus). Partially overlapping secretory precursors likely represent different maturational stages. Lgr5lowKi67+ cells rapidly proliferate and generate enterocytes through Lgr5Ki67+ and/or Lgr5Ki67 transient amplifying progenitors. Dark and light green indicate high and low Lgr5 expression, respectively. Red nucleus: mitotically active state; black nucleus: G0 state.

Our experiments indicate that the proliferating cells are much more efficient in organoid forming potential than quiescent cells. Lgr5high cells are highly proliferative and efficient in initiating organoid cultures. Albeit less efficiently, Lgr5low cells can also be induced to generate organoids. Both Dll1+ cells and LRCs retain some proliferative capacity and occasionally reenter the cell cycle becoming KI67+. This could potentially explain why unlike the label-retaining or Dll1+ secretory precursors, Lgr5lowKi67 cells do not readily generate organoids. A feasible alternative is the lack of regeneration cues that drives secretory cell proliferation upon injury in our organoid culture.

The biological meaning behind the connection of cell cycle exit and secretory differentiation is intriguing. Enterocytes in the villus are more numerous compared to the secretory cells. The differences in the proliferation rate between secretory and enterocyte precursors would maintain the correct ratio of absorptive to secretory lineages. During genotoxic conditions, such as viral infection or inflammation, secretory precursors could constitute a reserve stem cells population that is long-lived and protected from accumulation of replicative stress and mutations. The regenerative capacity of the relatively quiescent secretory progenitors as seen by Buczacki et al (2013) and van Es et al (2012) suggests that reversion from relative quiescence into a fast cycling stem cell fate is a physiological phenomenon. A standing question is how this transition is controlled. As differences in Lgr5 levels reflect differences in Wnt signaling activity, high level of which is required for organoid formation, Wnt signaling is an attractive candidate (Sato et al, 2009; van Es et al, 2012). Myc is a major downstream effector of Wnt, which promotes proliferation during intestinal regeneration (Ireland et al, 2004; Muncan et al, 2006; Sansom et al, 2007). In support, constitutive activation of Wnt signaling and NF-kB signaling induces dedifferentiation in the intestine (Schwitalla et al, 2013). Constitutive activation of Ras signaling enhances intestinal stem cell proliferation, which in combination with Apc activation can induce dedifferentiation of intestinal villi (Schwitalla et al, 2013; Snippert et al, 2014). Inhibition of Notch signaling is a known inducer of secretory differentiation, which coincides with decreased crypt proliferation (van Es et al, 2005). Whether activation of Notch signaling in the secretory precursors is enough for their conversion or additional signals activated in response to injury would instruct their dedifferentiation remains to be explored.

Previous efforts to identify an alternative stem cell population to CBCs focused on label-retaining cells mainly residing directly above the Paneth cell zone. The high similarity in the global expression pattern of Dll1 cells, LRCs, and Lgr5lowKi67 cells strongly suggests that the GFPlow population in our analysis includes these +4/5 positions. The fact that the four “classical” markers for “+4” cells are expressed by these Lgr5lowKi67 cells implies that these cells most likely represent the proposed quiescent “+4” stem cells as identified in the original lineage tracing studies using Bmi1, Tert, Hopx, and Lrig1 (Sangiorgi & Capecchi, 2008; Montgomery et al, 2011; Takeda et al, 2011; Yan et al, 2012). Our observations are entirely consistent with those of Winton and colleagues and imply that a slowly cycling population of secretory precursor cells resides around position “+4” and expresses the markers Hopx, Bmi1, Tert, and Lrig1 in combination with intermediate levels of Lgr5. These precursors are normally destined to differentiate into Paneth and enteroendocrine cells, but have the capacity to revert to a CBC stem cell upon tissue damage. Thus, these cells—when taken at the individual level—do not represent long-lived stem cells. Yet, the pool of these quiescent cells that is constantly replenished functionally constitutes a “reserve” stem cell population, to be called into action upon damage. Our findings using a novel knock-in system to directly visualize KI67-expressing cycling cells provide support for the dividing stem cell hypothesis and suggest that the “+4” population represents slow dividing secretory precursors.

Materials and Methods

Generation of the Ki67RFP knock-in mouse and the animals used

Ki67RFP knock-in mouse was generated by homologous recombination in embryonic stem cells by targeting a TagRFP cassette at the stop codon of the mKi67 gene. The endogenous stop codon is deleted to generate a C-terminal fusion of the KI67 protein with the TagRFP red fluorescent protein (Merzlyak et al, 2007). Details of embryonic stem cell targeting, Lgr5GFPiresCreER (Barker et al, 2007) and Lgr5GFPDTR (Tian et al, 2011) mice were described elsewhere. Mice were maintained within the animal facilities at the Hubrecht Institute, and experiments were performed according to the national rules and regulations of the Netherlands.

FACS sorting

Freshly isolated small intestines of Ki67RFP and Lgr5GFP mice were incised along their length, and villi were removed by scraping. The tissue was then washed 5 times in PBS by vigorous shaking to remove the access mucus. After incubation in PBS/EDTA (2 mM) for 15 min, gentle shaking removed remaining villi and intestinal tissue was subsequently incubated in 5 mM PBS/EDTA for 30 min at 4°C. Vigorous shaking yielded free crypts that were filtered through a 100-μm mesh and incubated in PBS supplemented with Trypsin (10 mg/ml; Sigma) and DNase (0.8 mg/ml; Roche) for 15 min at 37°C. Subsequently, cells were spun down, resuspended in SMEM (Invitrogen) and filtered through a 40-μm mesh. GFP- and TagRFP-expressing cells were isolated using BD FACSAria II cell sorter (BD Biosciences). Approximately 100,000 cells from a combination of 5–8 mice were sorted per population for each experiment. All analyzed populations were collected simultaneously.

Cell culture analysis

FACS-sorted intestinal crypt populations were collected in the intestinal culture medium including Wnt3a (50% conditioned medium) supplemented with the ROCK inhibitor Y-27632 (Sigma Aldrich) as described (Sato et al, 2009). The basic culture medium (advanced Dulbecco's modified Eagle's medium/F12 supplemented with penicillin/streptomycin, 10 mM HEPES, Glutamax, B27 [Life Technologies, Carlsbad, CA] and 1 mM N-acetylcysteine [Sigma]) was supplemented with 50 ng/ml murine recombinant epidermal growth factor (EGF; Peprotech, Hamburg, Germany), R-spondin1 (conditioned medium, 5% final volume), and Noggin (conditioned medium, 5% final volume). Conditioned media were produced using HEK293T cells stably transfected with HA-mouse Rspo1-Fc (gift from Calvin Kuo, Stanford University) or after transient transfection with mouse Noggin-Fc expression vector. Advanced Dulbecco's modified Eagle's medium/F12 supplemented with penicillin/streptomycin, and Glutamax was conditioned for 1 week. Wnt3a conditioned medium was produced using stably transfected L cells after 1 week of conditioning in medium (as previously described Sato et al, 2009) containing 10% fetal bovine serum. Cells were plated in matrigel (BD Bioscience) and cultured for a week before the number of organoids was quantified. At least 10,000 cells for Ki67RFP and 1,000 cells for Lgr5GFPKi67RFP double sorts were used per population per experiment. Experiments were done in biological duplicates.

Tissue preparation and immunofluorescence analysis

Fresh isolated Ki67RFP small intestinal samples were embedded in 4% low melting agarose (Invitrogen) and cut by a vibrating microtome (HM650, Microm) in cold medium and visualized immediately after for immunofluorescence analysis. Due to dim fluorescence and increase in the background, we could not analyze the TagRFP expression after formaldehyde fixation. Lgr5GFP small intestinal samples were fixed in 4% paraformaldehyde overnight, processed in 30% sucrose overnight and frozen in tissue freezing medium (Jung) on dry ice. 100-μm-thick floating sections were prepared using a cryostat (Cryostar NX70, Thermo scientific). Sections were blocked with 2% normal donkey serum (Jackson ImunoResearch) for 1 h at RT, permeabilized in PBS supplemented with 0.5% Triton X-100 overnight and stained with the indicated primary antibody. Primary antibodies used were eFluor-660 conjugated rat anti-KI67 (1:1,000; eBioscience), rabbit anti-lysozyme (1:3,000; DAKO), goat anti-chromogranin A (1:500, Santa Cruz) and rabbit anti-somatostatin (1:500; Invitrogen). Subsequently, sections were incubated with the corresponding secondary antibodies Alexa568 conjugated anti-rabbit and anti-goat antibodies (1:1,000; Molecular Probes) in blocking buffer containing DAPI (1:1,000, Invitrogen) for 2 h at RT and embedded using Vectashield (Vector Labs). Sections were imaged with Sp5 and Sp8 confocal microscopes (Leica) and processed using Photoshop CS5 and ImageJ software. Isolated intestinal crypts were visualized using the EVOS microscope (Electron Microscopy Sciences). The position of Lgr5+ and ChgA+ cells were quantified on z-stacks spanning the entire depth of the quantified crypts. For Fig 1B, 221 Lgr5+ cells from biological duplicates were quantified. For Fig 1D, in total, 32 crypts from biological duplicates were quantified. The Lgr5+ cell at the bottom of the crypt was considered position +1.

Microarray analysis, bioinformatics, and quantitative PCR

The Affymetrix analysis was performed on a genome-wide mRNA expression platform (Mouse Gene ST 1.1 arrays). The expression data extracted from the raw files were MAS5-normalized with the RMA-sketch algorithm from Affymetrix Power Tools and log2-transformed. Data were analyzed using the R2 web application, which is freely available at http://r2.amc.nl. In total, 21,212 unique genes are represented on the array. For genes represented with multiple probes, the one with the highest average expression level across the arrays was used. Expression levels were averaged within each group of arrays (Lgr5highKi67+, Lgr5highKi67, Lgr5lowKi67+, and Lgr5lowKi67) and log2 ratios calculated. Genes differentially expressed among groups were calculated according to the level of significance (P < 0.01; ANOVA). The Gene Set Enrichment Analysis (GSEA) was performed using the freely available software (v.2.0, Broad Institute; Subramanian et al, 2005) using preranked lists of mean expression changes and input signatures that were derived from published microarray data (Sansom et al, 2004; van Es et al, 2012; Munoz et al, 2012; Heijmans et al, 2013). qPCR analysis was performed using the SYBR-Green and Bio-Rad systems as described (Munoz et al, 2012).

Accession numbers

Array data are available at Gene Expression Omnibus (GEO) under the accession number GSE52813 (http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=etodgcqynfmfzib&acc=GSE52813).

Acknowledgements

We would like to thank Richard Volckmann, Jan Koster, Richard van Voort, and Peter van Sluis for the excellent support with the microarrays and bioinformatics analysis, Drs. Henner Farin, Meritxell Huch and Bon-Kyoung Koo for the fruitful discussion. This work was supported in part by grants by the Cancer Genomics Center (CGCII) to O.B., KWF/PF-HUBR 2007-3956 to M.v.d.B. and EU/232814-StemCellMark to J.H.v.E.

Author contributions

OB and HC conceived, designed, and analyzed the experiments. OB constructed the Ki67RFP mouse, analyzed the microarray data, and performed the cell culture and qPCR experiments. SVDE and OB designed and performed the FACS experiments. JB was supervised by OB and performed the confocal microscopy analysis. MVDB has provided the technical assistance of the mouse experiments. JK has performed the ES cell injections. Data interpretation was aided by JHVE. Funding was provided by HC. The manuscript was written by OB and HC and commented on by all other authors.

Conflict of interest

The Hubrecht Institute holds several patents related to Lgr5.

Ancillary