ABSTRACT
During late mitosis and early G1, replication origins are licensed for replication by binding to MCM2-7 double hexamers. This signals proliferative fate commitment. Here, we investigate how licensing and proliferative commitment are coupled in the small-intestinal epithelium. We developed a method for identifying cells in intestinal crypts that contain DNA-bound MCM2-7 and are licensed for replication. Interphase cells at the top of the transit amplifying zone did not contain DNA-bound MCM2-7, but still expressed MCM2-7 protein. This suggests licensing is inhibited immediately at terminal differentiation, after a final mitosis. Strikingly, we found that at the crypt base the majority of Lgr5(+) intestinal stem cells reside in an unlicensed state, despite expressing MCM2-7 protein and the Ki67 proliferation marker. This state, which we call ‘shallow-G0’, might allow stem cells to be easily activated to re-enter the cell cycle. We demonstrate that the dynamics of the licensing system provides a novel means to assess the unique cell-cycle of intestinal epithelial cells.
INTRODUCTION
Cell division is necessary for adult tissue homeostasis. It allows for the replacement of aged or damaged cells and the provision of specialised cells critical for tissue function. The decision to proliferate is crucial, especially for stem cells which produce daughter cells that either maintain a stem cell fate or differentiate to produce specialised cells. The rapidly-renewing intestinal epithelium replenishes its cellular content every 4-5 days. This high turnover rate is maintained primarily by Lgr5(+) intestinal stem cells in the crypt base, thought to be continually proliferative (Basak et al., 2014) as confirmed by proteomic and transcriptomic analysis (Munoz et al., 2012). There is also a quiescent stem cell-population that can re-engage with the cell-cycle to repopulate the Lgr5(+) cell population if it becomes depleted. These quiescent stem cells reside at the +4 position and constitute a subset of Lgr5(+) cells and are immature secretory lineage precursors (Buczacki et al., 2013). Lgr5(+) stem cells can divide to form transit-amplifying (TA) cells, which undergo several rounds of cell division before differentiating and losing proliferative competency (Potten and Loeffler, 1990).
How proliferative fate decisions are governed in stem cells and transit-amplifying cells is not understood. Lineage tracing studies suggest that in homeostatic intestinal tissue only 5-7 intestinal stem cells are ‘active’ out of the 12-16 Lgr5(+) cells present in the crypt base (Baker et al., 2014, Kozar et al., 2013). Interestingly, Lgr5(+) cells have a significantly longer cell-cycle than transit-amplifying cells (Schepers et al., 2011). The functional significance of the prolonged cell-cycle time of Lgr5(+) stem cells is currently unknown, but suggests active regulation of cell-cycle progression and proliferative fate commitment.
Proliferative fate decisions are typically visualised by detecting markers that are present in all cell-cycle phases, and only distinguish proliferative from quiescent cells. Visualising the incorporation of labelled nucleosides such as BrdU or EdU marks cells in S-phase. The limitation of these methods is that they cannot discriminate early proliferative fate decisions made during the preceding mitosis, or in the early stages of G1. DNA replication in S phase depends on origins having been licensed, which involves the regulated loading of minichromosome maintenance 2-7 (MCM2-7) complexes onto origins of DNA replication (reviewed in (Champeris Tsaniras et al., 2014)). During S phase, DNA-bound MCM2-7 hexamers are activated to form the catalytic core of the DNA helicase as part of the CMG (Cdc45, MCM2-7, GINS) complex (Moyer et al., 2006, Ilves et al., 2010, Makarova et al., 2012). Replication licensing is thought to occur from late mitosis throughout G1 until passage through the restriction point (Dimitrova et al., 2002, Haland et al., 2015, Namdar and Kearsey, 2006, Symeonidou et al., 2013). Correspondingly, insufficient origin licensing directly limits the ability to progress past the restriction point causing cell cycle arrest (Alver et al., 2014, Liu et al., 2009, Shreeram et al., 2002).
When cells enter G0, MCM2-7 proteins are transiently downregulated and degraded, primarily via E2F-mediated transcriptional control of MCM2-7, Cdc6 and Cdt1 (Leone et al., 1998, Ohtani et al., 1999, Williams et al., 1998). This prevents terminally differentiated cells from re-entering the cell cycle. In mammalian cells, artificial induction of quiescence through contact inhibition leads to gradual downregulation of Cdc6 and MCM2-7 over several days (Kingsbury et al., 2005). These features have led to the suggestion that quiescence can be defined by being an unlicensed state (Blow and Hodgson, 2002). Equally, the licensing status can define a different restriction point that signals proliferative fate commitment at the end of mitosis and in early G1, independently of the Rb/E2F restriction point.
The dynamics of replication licensing in the intricate cellular hierarchy of a complex, rapidly renewing adult tissue, is not understood. We therefore investigated the licensing system in the intestinal epithelium, aiming to understand dynamics of early cell-cycle commitment in stem and transit-amplifying cells and during terminal differentiation.
MATERIALS AND METHODS
Mice
All experiments were performed under UK home office guidelines. CL57BL/6 (Wild-type), R26-rtTA Col1A1-H2B-GFP (H2B-GFP), Lgr5-EGFP-IRES-creERT2 (Lgr5GFP/+) and ApcMin/+ mice were sacrificed by cervical dislocation or CO2 asphyxiation.
Tissue preparation: Whole small intestine
Dissected pieces of adult mouse small-intestine were washed briefly in PBS and then fixed in 4% PFA for 3 hours, 4°C. Intestines were cut into 2x2 cm2 pieces and fixed overnight in 4% PFA, 4°C. Tissue was embedded in 3% low melting temperature agarose and cut into 200 μm sections using a Vibratome (Leica). Sections were washed in PBS, permeabilised with 2% Triton-X100 for 2 hours and incubated with Blocking Buffer (1% BSA, 3% Normal Goat serum, 0.2% Triton-X100 in PBS) for 2 hours, 4°C. Tissue was incubated in Working Buffer (0.1% BSA, 0.3% Normal Goat Serum, 0.2% Triton-X100 in PBS) containing primary antibody, Mcm2 (Cell Signalling, 1:500), for 48 hours, 4°C. Sections were washed 5x with Working Buffer prior to 48 hour incubation with secondary antibodies diluted in Working Buffer: Alexafluor™ conjugated goat anti-rabbit (1:500, Molecular Probes) plus 5 μg/ml Hoechst 33342 and Alexafluor™ conjugated Phalloidin (1:150, Molecular Probes). Sections were mounted on coverslips in Prolong Gold between 2x120 μm spacers.
Tissue preparation: Isolated Crypts
Small intestines were dissected, washed in PBS and opened longitudinally. Villi were removed by sequential scraping of the luminal surface with a coverslip. Tissue was washed in PBS, incubated in 30 mM EDTA (25 minutes, 4°C) and crypts isolated by vigorous shaking in PBS. Crypt suspensions were centrifuged (fixed rotor, 88 RCF, 4°C) and the pellet washed twice in cold PBS. Crypts were fixed in 4% PFA (30min, room temperature), permeabilized in 1% Triton-X100 (1 hour, room temperature) and blocked in Blocking Buffer (2 hours, 4°C). Crypts were incubated with primary antibodies diluted in Working Buffer: Mcm2 (Cell Signalling, 1:500), phospo-HistoneH3 (Abcam, 1:500), Ki67 (Abcam ab15580, 1:250), αGFP (Abcam, 1:500), washed 5x with Working Buffer before overnight incubation with secondary antibodies diluted in Working buffer: Alexafluor™ conjugated goat anti-mouse or anti-rabbit (1:500, Molecular Probes) or stains: Rhodamine labelled Ulex Europaeus Agglutinin I (UEA, 1:500), 5 μg/ml Hoechst 33342 or Alexafluor™ conjugated Phalloidin (1:150), at 4°C. Crypts were mounted directly on slides in Prolong Gold, overnight.
CSK extraction
Soluble proteins were extracted from the epithelial cells in isolated crypts by incubation with CSK extraction buffer (10 mM HEPES, 100 mM NaCl, 3 mM MgCl2, 1 mM EGTA, 300 mM sucrose, 0.2% TritonX-100, 1 mM DTT, 2% BSA) supplemented with protease inhibitors (PMSF, Pepstatin, Leupeptin, Cystatin, Na3VO4, NaF, aprotinin) for 20 minutes on ice prior to fixation. Crypts were then fixed with 4% PFA and processed further.
H2B-GFP label retention
H2B-GFP expression in transgenic R26-rtTA Col1A1-H2B-GFP mice was induced by replacing normal drinking water with 5% sucrose water supplemented with 2 mg/ml doxycycline. After 7 days, doxycycline water was replaced with normal drinking water. Subsequently, mice were sacrificed after 7 days.
EdU incorporation and detection
Mice were injected intraperitoneally with 100 μg EdU (Invitrogen) prepared in 200 μl sterile PBS. Mice were sacrificed 1 hour or 17 hours post induction. For organoids, 10 μM EdU was included in crypt media for 1 hour before harvesting. EdU was detected by Click-it chemistry, by incubation in EdU working buffer (1.875 μM Alexafluor 488 azide (Invitrogen), 2 mM CuSO4, 10 mM Ascorbic acid), overnight at 4°C, prior to processing for immunofluorescence staining.
Organoid Culture
Isolated crypts were dissociated to single cells with TripLE express (Life Technologies) at 37°C, 5 minutes. Dissociated cells were filtered through a 40 μm cell strainer (Greiner) and suspended in growth factor reduced Matrigel (BD Biosciences). Organoids were grown in crypt media (ADF supplemented with 10 mM HEPES, 2 mM Glutamax, 1 mM N-Acetylcysteine, N2 (Gemini), B27 (Life technologies), Penicillin/Streptomycin (Sigma) supplemented with growth factors – ENR media (EGF (50 ng/ml, Invitrogen), Noggin (100 ng/ml, eBioscience) and RSpondin conditioned media produced from stably transfected L-cells (1:4). Chiron99021 (3 μM), Valproic acid (1 mM, Invitrogen) and Y27632 (10 μM, eBiosciences) were added to the culture for the first 48 hours. Organoids were passaged every 3-5 days by mechanically disrupting Matrigel and by washing and pipetting in ADF. Dissociated crypts were re-suspended in fresh Matrigel and grown in crypt media supplemented with growth factors.
For small molecule treatments, primary intestinal epithelial cells were cultured in ENR-CVY (ENR plus Chiron99021, Valproic acid and Y27632) for 3 Days, and then organoids were sub-cultured in ENR for two further days prior to the start of the experiment. Organoids were then treated with the stated small molecules for the indicated time periods. For induction of shallow-G0, organoids were treated with Gefitinib (5 μM) coupled with removal of EGF from the crypt media. For re-activation, the media was removed and fresh growth factors added. All growth factors and inhibitors were replenished every 2 days throughout the experiment.
Flow cytometry and cell sorting
Intestinal crypts were isolated and dissociated to single cells as described above. Isolated cells were filtered through 40 μm cell strainers. Cells were fixed in 0.5% PFA (pH7.40, 15 minutes, room temperature), washed once in working buffer and permeabilized with ice-cold 70% EtOH, 10 minutes. Cells were then washed in working buffer and re-suspended with primary antibodies (Mcm2, 1:500; GFP, 1:500; Ki67, 1:200) diluted in Working buffer (overnight, 4°C). Following two washes in working buffer, cells were re-suspended in secondary antibodies goat anti-mouse or anti-rabbit (Alexafluor647, 1:500 (Molecular Probes); Alexafluor488-Ki67, 1:400 (Clone SolA15, BD Biosciences), diluted in working buffer (1 hour, room temperature). After two washes in working buffer, cells were suspended in working buffer containing 15 μg/ml DAPI. Samples were analysed on a FACS Canto (BD Biosciences) and data analysed using FlowJo (Treestar).
For cell sorting, cells were isolated from Lgr5-GFP mice as described above by treatment with TripLE express for 15 minutes, 37°C followed by filtration through 40 μm filters (Greiner). Cells were sorted in ADF supplemented with 1% FCS and DAPI (15 μg/ml). Sorting was performed using an Influx™ Cell sorter (BD biosciences). Cells were checked post-sort to ensure sample purity by re-examining Lgr5 expression in the sorted gates.
Microscopy and Image analysis
Samples were imaged using a Zeiss LSM 710 microscope using a 40X LD Pan-Neofluar objective lens and immersion oil with a refractive index of 1.514. Z-stacks were acquired at optimal section intervals between 0.3 and 0.8 μm.
Image processing and analysis were performed using Imaris (Bitplane). Images of individual crypts were manually cropped to ensure that an individual crypt was the only region of interest. All nuclei were detected in individual crypts using automated thresholding in Imaris, set to detect nuclei at an estimated size of 3.5 μm. Missed or incorrectly assigned nuclei were manually identified. This function produced measurement points that segmented the specific region at the corresponding co-ordinate of the measurement point. Mean intensities for different channels were calculated per spot. This equates to the intensity at the centre region of each nucleus. A reference nucleus at the crypt base was used to define the crypt base position. The Euclidean distance to this point was measured and defined as the distance to the crypt base. Multiple images were analysed using the same workflow and the analysed files collated. For vibratome sections, a plane was manually defined running through to the muscle layer beneath the epithelium. The smallest distance to this surface was defined for segmented nuclei.
RESULTS
Mcm2 expression declines along the crypt-villus axis
Because of their abundance, strong conservation and association with the core DNA replication process, the presence of MCM2-7 proteins is frequently used to establish proliferative capacity in tissues, similar to Ki67 or PCNA (Gonzalez et al., 2005, Jurikova et al., 2016, Stoeber et al., 2001, Williams et al., 1998). Usually, terminally differentiated cells in mammalian tissues do not contain MCM2-7 (Stoeber et al., 2001, Todorov et al., 1998). To establish the overall MCM2-7 protein abundance along intestinal crypts, we first examined the expression of MCM2-7 proteins in adult murine small-intestinal epithelium using high-resolution immunofluorescence microscopy. We focused on Mcm2 as a surrogate for all the members of the MCM2-7 complex based on their similar function and localisation.
Consistent with previous reports in both murine and human intestinal epithelium, we observed that Mcm2 was highly expressed in intestinal crypts (Figure 1A) and declined gradually along the crypt-villus axis (Figure 1B), but persisted in a few cells in the villus compartment (S1 Figure A). MCM2-7 are highly abundant proteins and they have a relatively long (>24 hour) half-life (Musahl et al., 1998). This makes it likely that after cells differentiate, their MCM2-7 content declines at a slow rate. Mcm2 was nuclear in interphase cells but cytoplasmic during mitosis (Figure 1C). Although the majority of intestinal crypt cells expressed Mcm2, at the crypt base Mcm2(+) and Mcm2(-) cells were interspersed (Figure 1A, S1 Figure A), consistent with previous reports (Pruitt et al., 2010). This pattern is reminiscent of the alternating arrangement of Lgr5(+) stem cells and Paneth cells at the crypt base (Barker et al., 2007). Lgr5(+) stem cells express Ki67 and are continually proliferative whereas Paneth cells are fully differentiated and are Ki67(-) (Basak et al., 2014). As expected, Mcm2 was expressed in all Lgr5(+) stem cells and there was a strong correlation between Mcm2 and Lgr5 expression (Figure 1D). This is consistent with the idea that Lgr5Hi stem cells are the main proliferative stem cells in the intestinal crypt. Staining with Ulex Europaeus Agglutinin I (UEA), demonstrated that most of the Mcm2(-) cells in the crypt base were UEA(+) Paneth cells (Figure 1E).
Normally, MCM2-7 expression is lost in terminally differentiated cells (Eward et al., 2004, Stoeber et al., 2001, Williams et al., 1998, Williams et al., 2004). The loss of expression has been suggested as a major contributor to the proliferation-differentiated switch in vivo. To test this idea, we measured the Mcm2 content of young and mature secretory cells in intestinal crypts and villi (Figure 1F, G). There was differential expression of Mcm2 in distinct secretory lineages. Many mature secretory cells including Paneth, Goblet and enteroendocrine cells were Mcm2(-), consistent with their differentiation status and long life-span in the epithelium (van der Flier and Clevers, 2009). We detected a number of UEA(+) Mcm2(+) and UEA(+) Mcm2 (-) cells in intestinal crypts (Figure 1F). Assuming that Mcm2 expression declines slowly after differentiation, Mcm2 content could reflect the maturity of a particular secretory cell. Consistently, Mcm2 expression in UEA(+) cells in crypts was significantly higher than in villi (Figure 1G), supporting the idea that Mcms are gradually lost upon terminal differentiation.
Visualisation of DNA replication licensing In vivo
MCM2-7 exist in three states: as hexamers free in the nucleoplasm, as double hexamers bound to DNA during late mitosis, G1 and S phase, or as CMG complexes at replication forks during S phase (Evrin et al., 2009, Gambus et al., 2011, Remus et al., 2009). To distinguish between DNA-bound and soluble forms, we developed a protocol involving a brief extraction of isolated crypts with non-ionic detergent to remove soluble MCM2-7. The remaining Mcm2 should mark cells whose origins are licensed for replication. Extraction did not visibly affect intestinal crypt integrity but made them more opaque compared to unextracted tissue (Figure 2A). The majority of cells in unextracted crypts were Mcm2(+) (Figure 2B) similar to tissue sections and mirrored the ubiquitous expression of Ki67 along the crypt axis. After extraction, the majority of the Mcm2 content in cells was lost (Figure 2B). The labelling index of isolated crypts revealed that only 10-30% of cells were licensed (Figure 2C). After extraction, Mcm2(+) was not present in mitotic cells expressing phosphorylated histone H3, confirming the extraction procedure successfully removed non-DNA bound MCM2-7 proteins (Figure 2D).
We used flow cytometry to measure MCM2-7 content more directly and further confirm the effectiveness of the extraction procedure. Whereas the majority of isolated epithelial cells expressed Mcm2 that persisted throughout the cell-cycle, extraction revealed a distinct profile of Mcm-containing cells in crypts (Figure 2E) consistent with what has been reported for other cells (Friedrich et al., 2005, Moreno et al., 2016). MCM2-7s are loaded onto DNA throughout G1, reach a maximum level before cells enter S-phase, and are subsequently displaced from DNA during replication. Interestingly, we noticed that isolated Intestinal epithelial cells had a large range of DNA-bound Mcm2 during G1. We propose that this represents intermediate stages of the transition between a fully quiescent G0 state, a dormant state at the G0/G1 boundary or in early G1 and a population fully committed to the cell-cycle, being fully licensed (Figure 2E’). Similar results were observed in cells isolated from intestinal organoids (Figure 2F, G)
Licensing status and cell-cycle progression along the crypt-villus axis
Cell-cycle dynamics of intestinal stem and progenitor cells are highly heterogeneous (Pruitt et al., 2010). The majority of Lgr5(+) stem cells are considered to be continually proliferative, but with a much longer cell-cycle than transit-amplifying progenitor cells, which are most commonly found in S-phase (Schepers et al., 2011). To investigate proliferative fate decisions of intestinal epithelial cells, we used our MCM2-7 extraction in crypts where S phase cells were labelled with the nucleoside analogue EdU. We then used image analysis software to correlate Mcm2 content with cell-cycle stage along the crypt-villus axis (S1 Figure B-H). This allowed quantification of licensing in relation to the cell-cycle and 3D spatial information.
Figure 3A shows tissue following the MCM2-7 extraction and after a short 1 hour EdU pulse to visualise S-phase. As expected, the majority of cells in the transit-amplifying compartment were labelled with EdU suggesting that most cells were in S-phase, consistent with early studies using BrdU and [3H]-thymidine labelling (Chwalinski and Potten, 1987). The patterns of replication foci were consistent with the reported S-phase replication timing programme (Rhind and Gilbert, 2013). Typically, all licensed cells had intense nuclear Mcm2 staining. Some cells completely lacked Mcm2 and EdU labelling, suggesting they could be in either G0, very early G1 or in G2. Some cells were labelled with both Mcm2 and EdU. These double-labelled cells typically showed patterns of EdU labelling consistent with early to mid S phase and Mcm2 labelling of DNA compartments expected to replicate later in S phase. This relationship has been observed in tissue culture cells (Krude et al., 1996) and is consistent with the idea that DNA-bound MCM2-7 are displaced from chromosomal domains as replication is completed. Cells with late S-phase patterns of EdU labelling had little or no detectable Mcm2, consistent with the displacement of the majority of MCM2-7 by the end of S phase. Quantification of the nuclear intensities of Mcm2 and EdU in these discrete populations (S1 Figure H) confirmed previous results using flow cytometry (Figure 2E) and allowed grouping of cells into 4 cell cycle groups: Unlicensed, G0/G1; G1 licensed; S-phase and Late-S/G2 (S1 Figure H). We also measured nuclear volume, which increases during S phase and G2. This showed that nuclear volume increased up to two-fold in cells classified as S-phase and Late-S/G2 by Mcm2 and EdU staining (Figure 3B). This confirms our cell-cycle assignment and also suggests that most Mcm2(-) cells are in G0 or G1, rather than in G2.
The combination of concurrently labelling DNA-bound Mcm2 and EdU showed a clear correlation between cell position and cell-cycle stage (Figure 3C, D, E). At the base of the crypt, G0/early-G1 cells predominate. At increasing distances from the crypt base there is a successive rise in licensed G1 cells, early/mid S phase cells and then late S/G2 cells. Further up the crypt, at the end of the TA compartment, these cell cycle stages decline in reverse order, until unlicensed G0 cells again predominate. This suggests that there is a co-ordinated progression through the cell division cycle as cells enter, then leave, the TA compartment. This was also observed as a field effect with many neighbouring cells showing similar replication patterns (S2 Figure A, B).
Terminal differentiation is associated with a binary licensing decision
At the terminal boundary of the transit-amplifying compartment, the majority of cells were unlicensed and had no DNA-bound Mcm2 (Figure 3C-E). Similarly, there were no licensed G1 cells beyond the TA compartment as defined by incorporation of EdU (Figure 3F). However, total Mcm2 expression extended significantly beyond the last cells with DNA-bound Mcm2 or incorporated EdU (Figure 3D, F). The distribution of total Mcm2 expression also corresponded to the zone where cells express Ki67 (S3 Figure). Although Mcm2 and Ki67 expression persists beyond the TA compartment, licensing does not occur in this area. This suggests that differentiation is not governed by a gradual reduction in total MCM2-7 levels, but is a binary decision and licensing is abolished immediately after the final mitosis preceding differentiation. To further examine this, we marked the terminally differentiated zone by a 1 hour EdU pulse, followed by a 16 hour chase (Figure 3G, H). After 16 hours, the majority of the distal end of the TA compartment became labelled with EdU. All labelled nuclei in this area were significantly smaller than EdU(+) cells at the proximal end of the TA compartment (data not shown), suggestive of their differentiation status. Importantly the EdU(+) differentiated cells at the distal end of the TA compartment lacked DNA-bound Mcm2, supportive of the model where licensing is inhibited immediately at terminal differentiation.
The majority of Intestinal stem cells reside in unlicensed shallow-G0’ state
The majority of cells in the crypt base expressed Mcm2, consistent with the finding that all Lgr5(+) cells express Mcm2 but mature secretory cells, such as Paneth cells, do not (Figure 1D, E). Surprisingly, extraction revealed that only 7-15% of cells were licensed in the crypt base (Figure 3C, D), with most cells in an unlicensed state despite expressing Mcm2. The abundance of licensed cells peaked 40-60 μm away from the crypt base, corresponding to just above the +4/+5 cell position (Figure 3D, E).
It is not possible to identify Lgr5 in these experiments, as it is extracted along with unbound Mcm2. We therefore identified Paneth cells by UEA staining and considered all UEA(-) cells in the crypt base to be intestinal stem cells (Figure 4A). >50% of the UEA(-) stem cells were in an unlicensed state and were not incorporating EdU (Figure 4B). Approximately 30-40% of all UEA(-) cells in the stem cell compartment were in an active phase of the cell cycle, (Figure 4B) corresponding to 5-6 stem cells out of the total 14 present (Snippert et al., 2010). This number is similar to the small number of proposed ‘working’ stem cells in the crypt base (Baker et al., 2014, Kozar et al., 2013). Unlicensed cells not incorporating EdU (i.e. unlabelled in this experiment) could theoretically be in either G0 or in G2. To distinguish between these possibilities we first isolated crypt cells from Lgr5-GFP mice and measured both GFP and DNA content. Both Lgr5(+) and Lgr5(-) cell populations had a similar cell-cycle profile with the majority of cells having 2N DNA content (S2 Figure C). We also examined the nuclear volume of cells at different positions along the crypt axis, after staining for EdU incorporation and DNA-bound Mcm2. The majority of unlicensed cells had a similar nuclear volume to fully licensed cells in G1 and not cells in Late-S/G2 phase (S2 Figure D). Together, these results suggest that, although they express abundant Mcm2, the majority of intestinal stem cells reside in an unlicensed state.
To confirm this conclusion, we flow sorted Lgr5-GFP(+) cells, extracted unbound MCM2-7 and stained them for Mcm2 and Ki67. Consistent with our previous results, most of the Lgr5(+) cells with a 2N DNA content had low levels of DNA-bound Mcm2, and were in an unlicensed state (Figure 4Ci, ii). Importantly, both the licensed and unlicensed cells were Ki67(+) indicating that they had not withdrawn from the cell-cycle long-term (Figure 4Cii).
This state – 2N DNA content with low levels of DNA-bound Mcm2 - could be explained by two, slightly different, scenarios. One possibility is that MCM2-7 are loaded on to DNA very slowly in Lgr5(+) cells, thereby extending G1 length (Schepers et al., 2011) (Dalton, 2015). In this case, the presence of unlicensed cells simply reflects the increased time required to fully license origins, and different levels of Mcm2 loading should be equally distributed between G1 cells. Alternatively, most Lgr5(+) cells are not in G1 and do not load MCM2-7 until an active decision is made to enter the cell cycle and activate the licensing system, at which time MCM2-7 proteins are rapidly loaded. In this case, there should be a discrete peak of unlicensed cells with a G1 DNA content representing cells that have withdrawn from the cell cycle, and a lower frequency of G1 cells that have loaded different amounts of MCM2-7. To distinguish between these two possibilities, we examined the frequency distribution of DNA-bound Mcm2 in Lgr5(+) cells with a 2N DNA content (Figure 4Ciii). The distribution of DNA-bound Mcm2 cells was most similar to the predicted distribution in the latter model and showed a discrete peak of unlicensed cells (Figure 4C). Since most of these unlicensed Lgr5(+) cells express abundant Mcm2 (Figure 1D), they are in a state that is distinct from that of previously described G0 cells, which typically do not express licensing proteins at all. We term this new intermediate state – where cells express abundant MCM2-7 proteins that are not bound to DNA – as ‘shallow-G0’. Because they do not need to synthesize more MCM2-7 proteins to enter the cell cycle, they are likely to be in a transient state of quiescence.
It has previously been reported that embryonic stem cells license more replication origins than neural stem/progenitor cells differentiated from them (Ge et al., 2015). To determine if stem and non-stem cells in intestinal crypts behave similarly, we compared the amount of DNA-bound Mcm2 in G1/G0 and early S phase Lgr5(+) cells with that in Lgr5(-) cells (Moreno et al., 2016). Although the majority of Lgr5(+) cells were unlicensed, when they entered S phase they had approximately twice as much DNA-bound Mcm2 compared to Lgr5(-) cells (Figure 4D). This is consistent with the idea that adult intestinal stem cells license more origins than TA cells. This may represent a mechanism to protect genomic integrity.
Intestinal label retaining cells are in a deep G0 state
Although the intestinal crypt base primarily consists of Lgr5(+) stem cells there is also a reserved pool of quiescent stem cells, often referred to as ‘+4 label retaining cells’ (LRCs) reflecting their position in the crypt base and their ability to retain nascent DNA labels (Potten et al., 2002). These cells are a rare subset of Lgr5(+) cells which are also secretory precursors (Buczacki et al., 2013). To further define the licensing status of these label-retaining intestinal stem cells, we identified UEA(-) LRCs by expressing H2B-GFP (which is incorporated into the chromatin of dividing cells) for 7 days and then chasing for a further 7 days (S4 Figure) (Buczacki et al., 2013, Roth et al., 2012). Labelled cells that did not divide during the 7-day chase period contain high levels of H2B-GFP, and cells that divided multiple times only have low levels. Strikingly, unlike the majority of the Lgr5(+) cells, quiescent LRCs with high GFP-H2B did not express Mcm2 (Figure 4E). Consistently, only non-LRC daughter cells with low levels of H2B-GFP had DNA-bound Mcm2 (Figure 4F, G). This shows that, as expected, the LRC stem cells are in deep G0, unable to license because they do not express MCM2-7. In contrast, the ‘active’ intestinal stem cells reside in a state of shallow G0, expressing MCM2-7, but remain unlicensed.
Stemness is associated with the unlicensed state
We wished to understand how intestinal stem cells were maintained in an unlicensed state and whether stemness was directly associated with the unlicensed shallow-G0 state. To investigate this relationship we designed a proof-of-concept assay using intestinal organoids. This allowed a preliminary assessment of licensing dynamics during entry and exit into quiescence. In contrast to intestinal crypts in vivo, intestinal organoids contained considerably more cells with DNA-bound Mcm2 in their crypt-like branches (Figure 5A). Cytometry-based quantification of cells with a 2N DNA content suggested that there were approximately twice as many fully licensed cells in organoids than in intestinal crypts (Figure 5Bi, ii). This suggests that organoids may represent intestinal epithelium in an accelerated state of self-renewal and do not fully recapitulate cell-cycle dynamics of intestinal epithelial cells in vivo.
To measure licensing dynamics in organoids during entry and exit into quiescence, we directly compared licensing states with the presence of Ki67. Most cells in organoids express Ki67 and it increased during cell-cycle progression (Figure 5Ci). The DNA-bound Mcm2 profile was similar to that in isolated crypts (compare Figure 5Cii and Figure 2E). Correlating Ki67 and DNA-bound Mcm2 produced a distinctive profile that is similar for isolated cells from organoids and intestinal crypts. This profile reveals a population of cells that appear to be losing Ki67 (dashed arrow in Figure 5 Ciii and iv) and which might represent cells losing proliferative capacity and transitioning towards differentiation (Figure 5C, S5 Figure A). This loss of proliferative capacity may begin in cells that express Ki67 but are unlicensed, i.e. cells in shallow G0. It also suggests that different stages of quiescence exist that are reflected by a spectrum of Ki67 and Mcm levels. To test this idea, we induced quiescence by inhibiting the EGF receptor (EGFR), which reduces MAP kinase activity and blocks DNA replication and cell division (Lynch et al., 2004). Strikingly, inhibition of EGFR for 24 hours induced arrest in shallow-G0, and caused the majority of cells to be unlicensed with a 2N DNA content, while expressing Mcm2 and Ki67 (Figure 5Dii). Prolonged EGFR inhibition caused a transition into an intermediate state between shallow-G0 and deep G0, with reduced Ki67 expression but with total Mcm2 levels maintained (Figure 5Diii). These shallow-G0 states were reversible by removal of EGFRi and re-addition of fresh growth factors (Figure 5Div).
We also used this assay to investigate how these shallow-G0 states and stemness might be related. Inhibiting EGFR also increases Lgr5 expression (Basak et al., 2017), suggesting that prolonged quiescence can be associated with ‘stemness’. The observed increase in shallow G0 cells after 24 hours EGFRi treatment is thus consistent with the idea that stem cells spend time in shallow G0. To test this idea more directly, we also used an alternative approach. Treatment of organoids with Chir99201 (a GS3K inhibitor) and Valproic acid (a Notch activator/histone deacetylase inhibitor) resulted in Lgr5 expression throughout the organoid epithelium (S5 Figure B) (Yin et al., 2014). This treatment also caused the appearance of a population of cells with low levels of Ki67 and intermediate levels of DNA-bound Mcm2 (Figure 5E) similar to the intermediate shallow G0 state caused by EGFRi. Surprisingly, we observed cells with low levels of Ki67 and intermediate levels of DNA-bound Mcm2, suggesting that re-licensing of these cells occurs before they express high levels of Ki67. Making measurements at intermediate times of CV treatment or after CV removal, confirmed the existence of the intermediate and shallow G0 states and also the ability to relicense before Ki67 re-expression (Figure 5F). Combining EGFRi and CV treatment, also suggested that cells can reactivated licensing from the intermediate G0 state directly (Figure 5G).
Treatment with Valproic acid alone, but not Chir99021, partially recapitulated this effect, suggesting that active Notch signalling is involved (S5 Figure C). Inhibiting Notch signalling with DAPT treatment, which induces terminal secretory cell differentiation (van Es et al., 2005), caused an induction of deep-G0, with reduced Ki67 and a loss of Mcm2 proteins (S5 Figure D). Together, this suggests that Notch signalling can affect the transition between deep and shallow-G0 states.
Discussion
The cell-cycle of intestinal stem and transit-amplifying cells is poorly understood. By comparing the total and DNA-bound Mcm2 in intact intestinal crypts we provide new insights into how licensing and cell-cycle commitment are coupled in this tissue. We provide evidence that after their final mitosis, transit amplifying cells do not license their replication origins and so immediately exit the cell cycle. We show that normally the majority of Lgr5(+) stem cells reside in an unlicensed state, despite expressing Mcm2 and Ki67. In this state of shallow-G0, stem cells might be poised to easily re-enter the cell division cycle.
Lgr5(+) stem cells have a cell-cycle length greater than transit-amplifying cells (Schepers et al., 2011). The biological relevance of this is currently unknown. The data presented here suggest a delay in origin licensing is responsible for the prolonged cell-cycle of Lgr5(+) cells. Although ∼80% of Lgr5(+) cells are thought to be continually proliferative and express high levels of both Ki67 (Basak et al., 2014) and Mcm2, we found that most Lgr5(+) cells reside in an unlicensed state, with 2N DNA content and Mcm2 not bound to DNA. Since the licensed state defines proliferative fate commitment, we suggest that these cells are in a prolonged state of shallow quiescence, which we term shallow G0, expressing proliferative makers such as Ki67 and Mcm2 without committing to proliferation since Mcm2 is not bound to DNA (Figure 6). The number of Lgr5(+) cells with DNA bound-Mcm2 was similar to the number of proposed ‘active’ stem cells determined in lineage tracing experiments (Baker et al., 2014, Kozar et al., 2013).
Prolonged arrest may eventually result in degradation of MCM2-7 proteins and lead to induction of a state of deep quiescence. Consistent with this idea, we observed that LRCs, thought to provide a reserve of quiescent stem cells, did not express Mcm2. The lack of Mcm2 expression may reflect that a significant period of time has passed since these cells divided. The delay in activating the licensing system may create a prolonged time-window for Lgr5(+) cells to receive and interpret environmental cues before deciding to commit to duplication, offering a means to control their number. It is likely that the majority of Lgr5(+) cells eventually re-enter the cell cycle, given their continual expression of proliferation markers (Basak et al., 2014). The identity and decisions of Lgr5(+) cells are likely governed by stochastic choices and the ability to enter a shallow G0 stage offers unique flexibility for stem cells to make these choices. Consistent with this idea, modulation of the stem cell niche by Chir99021 and Valproic acid induces stemness throughout the crypt-villus axis (Yin et al., 2014) and also significantly enriches a unique population of unlicensed cells with unique cell-cycle dynamics. The increase in Lgr5(+) cells in response to Valproic acid and Chir99021 suggests a corresponding increase in the number of Lgr5(+) reserve stem cells, which are in deep G0 (Buczacki et al., 2013). We propose that the reactivation of these cells by injury for instance, could proceed via the intermediate G0 state we describe. Initially, these cells must re-express Mcm proteins and then can directly commit to the cell-cycle from the intermediat-G0 state (S5 Figure E). Together, this demonstrates the unique cell-cycle characteristics of intestinal stem cells, which can be functionally defined by the licensed state.
Growing evidence suggests that intestinal stem cell fate is not governed by asymmetric segregation of fate determinants (Lopez-Garcia et al., 2010, Snippert et al., 2010, Steinhauser et al., 2012). Components of the stem cell niche, such as the combination of Wnt and Notch signalling can affect stem cell fate decisions and also reduce the cycle rate of intestinal stem cells (Hirata et al., 2013). This is consistent with the idea that as well as decreasing proliferation rate, increased G0/G1 length might underpin cell fate choices. Indeed, G1 elongation of mouse and human embryonic stem cells can drive differentiation (Calder et al., 2013, Coronado et al., 2013). Similarly, long G1 phases are associated with the production of fate-restricted progenitors during neurogenesis (Arai et al., 2011). This has been suggested to be facilitated by an extended time window in the cell-cycle to allow niche factors and/or fate determinants to (Calegari and Huttner, 2003). In the case of intestinal stem cells, holding cells in shallow G0 may allow an extended time for stem cell fate factors to act and maintain stem cell fate.
Like embryonic stem cells (Ge et al., 2015), intestinal stem cells appear to have licensed more origins than non-stem cells when they enter S phase. Intestinal stem cells may therefore more readily engage the licensing checkpoint that ensures that all origins are licensed before cells enter S phase (Alver et al., 2014, Liu et al., 2009, Shreeram et al., 2002). This additional demand for licensed origins in stem cells may explain why crypts hypomorphic for Mcm2 have a stem-cell deficiency (Pruitt et al., 2007).
Terminal differentiation in the intestinal epithelium is associated with disengagement from the proliferative niche and involves the gradual dilution of niche factors, which causes cells to exit the cell-cycle (Farin et al., 2016, Mariadason et al., 2005). Consistent with previous reports, we observed that Mcm2 expression gradually declined along the crypt-villus axis (Figure 1) (Stoeber et al., 2001). We also found that there is an abrupt loss of DNA-bound Mcm2 as cells exit the transit amplifying zone and undergo terminal differentiation. This suggests that the proliferation-differentiation switch is a binary decision made in the last cell cycle in the mid-upper transit-amplifying compartment.
It is unclear why licensing is rapidly inhibited at the top of the transit amplifying zone or in most of the Lgr5(+) stem cells. The simplest explanation is that licensing factors such as Cdt1 or Cdc6 are not readily available in new-born stem cells, and their synthesis has to be directed by an upstream signal for fate commitment. This occurs after prolonged quiescence which is accompanied by passive downregulation of such licensing factors (Coller, 2007). In contrast, in continually dividing cells their levels are maintained. Consistently, licensing factors such as Cdc6, along with many cyclin-CDK complexes, are down regulated beyond the end of the TA zone (Frey et al., 2000) (Smartt et al., 2007).
In summary, we demonstrate that the dynamics of the DNA replication licensing system provides a new way of measuring the proliferative fate of intestinal stem cells. We suggest a model for ‘working’ intestinal stem cells that reside in a state of shallow quiescence until a proliferative fate decision is made. Consistent exit from the cell-cycle in label retaining ‘+4’ cells leads to loss of proliferative capacity and loss of Mcm2 expression causing cells to enter a deeply quiescent state (Figure 6). We suggest that the shallow-G0 state serves stem cells in controlling their numbers by regulating the cell-cycle.
Author contributions
T.D.C, J.J.B and I.N conceived and designed the study; T.D.C collected the data and performed the analysis; I.P.N assisted with animal experiments; T.D.C wrote the manuscript with assistance from I.N and J.J.B.
Conflicts of Interest
The authors report no conflicts of interest.
Acknowledgements
We would like to thank members of the Näthke and Blow laboratories for general assistance and helpful discussions. We thank Dr Paul Appleton, Dr Graeme Ball and the Dundee Imaging and Tissue Imaging Facility for support with microscopy and image analysis. The imaging facility is funded by the Welcome Trust Technology Platform award (097945/B/11/Z) and Welcome Trust award (101468/Z/13/Z). We thank Dr Rosemary Clarke and the Dundee Flow Cytometry Facility for support with flow cytometry, cell sorting and analysis. This work was supported by a programme grant from Cancer Research UK to I.N (C430/A11243) and to J.J.B (C303/A14301), Wellcome Trust grant WT096598MA and an MRC studentship award to T.D.C.