Abstract
Neurog3HI endocrine-committing cells are generated from a population of Sox9+ mitotic progenitors with only a low level of Neurog3 transcriptional activity (Neurog3TA.LO). Low-level Neurog3 protein, in Neurog3TA.LO cells, is required to maintain their mitotic endocrine-lineage-primed status. Herein, we describe a Neurog3-driven FUCCI cell-cycle reporter (Neurog3P2A.FUCCI) derived from a Neurog3 BAC transgenic reporter that functions as a loxed cassette acceptor (LCA). In cycling Sox9+ Neurog3TA.LO progenitors, the majority of cells in S-G2-M phases have undetectable levels of Neurog3 with increased expression of endocrine progenitor markers, while those in G1 have low Neurog3 levels with increased expression of endocrine differentiation markers. These findings support a model in which variations in Neurog3 protein levels are coordinated with cell-cycle phase progression in Neurog3TA.LO progenitors with entrance into G1 triggering a concerted effort, beyond increasing Neurog3 levels, to maintain an endocrine-lineage-primed state by initiating expression of the downstream endocrine differentiation program prior to endocrine-commitment.
Introduction
Neurogenin3 (Neurog3) encodes a bHLH transcription factor essential for endocrine-lineage specification during mouse pancreas organogenesis (1). Neurog3 is also critical to human pancreatic endocrine-cell development, with null mutations causing neonatal diabetes, and blocking β-cell differentiation from hESC (2). During mouse pancreatic development, high-level Neurog3 expression (Neurog3HI) in Sox9+ pancreatic epithelial cells causes cell-cycle exit, endocrine commitment and epithelial delamination (3–6). We recently demonstrated, however, that low Neurog3 levels are necessary for maintaining a population of Sox9+ Neurog3- transcriptionally-active pancreatic epithelial cells in a mitotic endocrine-biased progenitor state (defined as Neurog3TA.LO), which pre-empts the transition to an endocrine-committed Neurog3HI state (6,7). Our findings presented a significant parallel to how a low level of Neurog2 promotes a neural-progenitor state while high levels cause neural differentiation and cell-cycle exit (8–11). In those studies, higher Cdk activity in rapidly cycling progenitors, which have a relatively short G1, keeps Neurog2 in a (hyper)-phosphorylated, unstable state that activates neural-progenitor target genes (10,12). When the cell cycle of neural progenitors lengthens, however, and G1 lengthens, Cdk activity decreases, resulting in accumulation of a more stable (hypo)-phosphorylated Neurog2 that preferentially activates neural-differentiation targets (10,12). Recently, we demonstrated that keeping Neurog3 levels low leads to an increased mitotic index of Neurog3TA.LO progenitors and expands their numbers within the pancreatic epithelium (6). Moreover, time-lapse observations show that the transition from the low level of Neurog3 observed in mitotic Neurog3TA.LO progenitors to the high level necessary for endocrine-commitment occurs ~3-6 hours after division of the parental Neurog3TA.LO cell, during G1 (6). These findings led to our proposal that the level and stability of Neurog3 in mitotic Sox9+ Neurog3TA.LO progenitors is regulated by the cell cycle and that G1 extension promotes Neurog3 stabilization, accumulation, and endocrine commitment (13). Two recent reports support this model, demonstrating that Neurog3 is targeted and destabilized by Cdks and that G1 lengthening, by reducing Cdk activity, causes the accumulation of a more stable un(der)phosphorylated form of Neurog3 (14,15).
We have been independently investigating if Neurog3 protein stability and progenitor maintenance vs. endocrine differentiation decisions are connected to cell-cycle progression in Neurog3TA.LO progenitors. To do so, we used recombinase-mediated cassette exchange (RMCE) to replace our previously described Neurog3RG BAC transgenic reporter – which was designed as a Loxed Cassette Acceptor (LCA) – with a Neurog3-driven single-transgene insert of the FUCCI (Fluorescence Ubiquitin Cell Cycle Indicator) reporter (Neurog3P2A.FUCCI). Our analysis of Neurog3P2A.FUCCI reporter activity showed that in cycling Sox9+ Neurog3TA.LO progenitors, Neurog3 protein levels are highest during G1 and lowest during S-G2-M. Moreover, Sox9+ Neurog3TA.LO progenitors in early G1 show increased expression of downstream Neurog3 targets usually associated with the forward passage into an endocrine commitment and progression program. We propose that these findings support a model in which the endocrine-differentiation program is already accessed, or preformed (albeit at a low or incomplete level), in mitotic Neurog3TA.LO progenitors prior to moving into endocrine-commitment. This work provides a new tool for investigating, under in vivo conditions, Neurog3 and cell-cycle connections in lineage-primed progenitors, and new insight on the role of Neurog3 in regulating progenitor maintenance vs endocrine-commitment decisions.
Results and Discussion
Generating a Neurog3-driven P2A-fused single transgene FUCCI reporter
The FUCCI reporter relies on cell-cycle-phase-dependent destruction of fluorescent proteins fused to “degradation boxes” from hGeminin and hCdt, specifically the regions hGem(1/110) and hCdt1(30/120) (16), allowing cell-cycle phase determination (Figure 1A). To investigate connections between Neurog3 protein levels and cell-cycle progression in Neurog3TA.LO progenitors we generated a single mKO2-hCdt1(30/120)-P2A-mVenus-hGem(1/110) FUCCI (P2A.FUCCI) cassette, enabling both FUCCI components to be expressed under the control of Neurog3 (Figure 1B). We selected the pairing of mKO2/mVenus because their fluorophores are spectrally separable from GFP and mCherry, allowing P2A.FUCCI visualization in cells carrying our previously described Neurog3- driven H2BmCherry-P2A-GFPGPI (Neurog3RG1 reporter) (6). As seen with the original FUCCI reporter (16), CMV-driven expression of P2A.FUCCI in HeLa cells resulted in mKO2-hCdt1(30/120) positivity during G1 and mVenus-hGem(1/110) positivity during S-G2-M, with a brief overlap of the two fusion proteins during the G1/S phase transition (Figure 1B).
To facilitate generating additional Neurog3 BAC transgenic reporters from the Neurog3RG reporter, we had flanked the Neurog3RG cassette with tandem lox71 and lox2272 sites, making a Loxed Cassette Acceptor (LCA) allele (see ref. 6; Figure 2A). This design was to allow Neurog3RG to be replaced with any lox66/lox2272-flanked cassette via RMCE in mESCs. To avoid potential issues with performing RMCE in cells carrying multiple LCA alleles, mESCs identified as having stably integrated the Neurog3RG BAC LCA transgene were screened for single-copy insertion by a qPCR-based assay (see methods and materials; Figure 2B and C) that accurately estimates transgene copy number (17). Using this assay, two mESC transgenic clones, referred to as Neurog3RG1 and Neurog3RG2, were identified as having copy numbers of 1.25 ±0.16 and 1.46 ±0.26 (Figure 2C). Derivation of Neurog3RG mESC lines and the subsequent Neurog3RG1 mouse line are described in (6). As with the Neurog3RG1 mouse (6), examination, in Neurog3RG2 mice, of gross tissue and islet architecture, ad libitum fed glucose levels, and proportions of Sox9+ Neurog3 protein-low (Neurog3pLO) versus Sox9- Neurog3 protein-high (Neurog3pHI) cells during pancreas development, revealed no abnormal phenotype (Figure 2-figure supplement 1A-D; data not shown). We next validated the LCA function of the Neurog3RG BAC transgene by using the Neurog3RG2 mESC line to derive a Neurog3P2A.FUCCI mESC line. A lox66/lox2272-flanked Neurog3P2A.FUCCI-PGK-hygroR cassette was generated with cassette placement mimicking that of Neurog3RG (Figure 2-figure supplement 2A). Following RMCE in Neurog3RG2 mESCs, PCR was performed to verify replacement of the lox71/lox2272-flanked Neurog3RG-PGK-PuroΔTK cassette with the lox66/lox2272 Neurog3P2A.FUCCI-PGK-hygroR cassette (Figure 2-figure supplement 2B). This derivative Neurog3P2A.FUCCI mESC line was then used to generate Neurog3P2A.FUCCI transgenic mice. Given that the genomic integration site is likely different in Neurog3RG2 vs. Neurog3RG1 mESC lines, we used Neurog3RG2 mESCs to derive Neurog3P2A.FUCCI mice to allow future breeding of Neurog3P2A.FUCCI to Neurog3RG1 mice to enable four-color reporting of cell-cycle phase and Neurog3 expression. Our proposal is that such visualization could facilitate experiments aimed at understanding if, like other progenitor populations, G1 length or overall cell-cycle length in mitotic Neurog3TA.LO progenitors plays a role in regulating progenitor maintenance vs. endocrine-commitment decisions, or even in determining whether one endocrine cell-type is produced over another at specific stages or locations within the developing pancreas.
Neurog3 levels and progenitor maintenance vs. endocrine-commitment decisions are coupled to the cell cycle
Recent studies showed that during S-G2-M, Neurog3 is kept in a hyperphosphorylated unstable state via Cdk phosphorylation, and that decreased Cdk activity associated with entrance into G1 result in stabilization and accumulation (14,15). These findings support our model that, in actively cycling Neurog3TA.LO progenitors, Neurog3 protein levels vary according to the cell-cycle phase (13). To address this issue, we used Neurog3P2A.FUCCI reporter expression in Neurog3TA.LO progenitors to track cell-cycle progression in relation to Neurog3 protein levels. Previously, Neurog3TA.LO progenitors were defined as a population of Sox9+ Neurog3-transciptionally active (low-level Neurog3RG1 reporter expression) progenitors comprising cells with either low (Neurog3TA.pLO) or immunologically undetectable Neurog3 (Neurog3TA.pUD) (6). Consistent with this definition, Neurog3TA.LO progenitors were herein defined as Sox9-positive and positive for either component of the Neurog3P2A.FUCCI reporter, with low or undetectable Neurog3 protein, whereas endocrine-committed Neurog3TA.HI cells should be Sox9-negative, Neurog3pHI and positive for mKO2-hCdt1(30/120) (Figure 3A). Unexpectedly, we detected significant residual cytoplasmic mVenus-hGem(1/110) fluorescence in post-mitotic, actively delaminating, endocrine-committed Neurog3TA.HI cells that showed the expected high mKO2-hCdt1(30/120) signal (Figure 3A). This observation was different from previous reports on the FUCCI reporter, where mVenus-hGem(1/110) was mostly degraded after M-phase, becoming absent by the time of mKO2-hCdt(30/120) detection in early G1 (16,18). This cytoplasmic mVenus-hGem(1/110) signal, however, was completely absent in islets (data not shown), which could suggest that high Neurog3P2A.FUCCI reporter expression in delaminating Neurog3TA.HI cells overwhelms the ubiquitin-mediated protein degradation pathway, extending the time necessary to fully degrade mVenus-hGem(1/110) after entering G1. We were therefore careful to score Neurog3P2A.FUCCI cells as only in S-G2-M if definitively nuclear mVenus signal was observed, with no indication of mKO2 (Figure 3A). By these criteria, the majority of Sox9+ Neurog3TA.LO cells were in S-G2-M and thus mitotic, while nearly all Neurog3TA.HI cells were in G1 (Figure 3B). To determine if Neurog3 protein levels vary through the cell cycle, we examined the cell-cycle status of Sox9-positive Neurog3TA.pLO versus Neurog3TA.pUD cells. Quantification revealed that 78% (± 8.1%) of Neurog3TA.pUD cells were in S-G2-M and 22% (± 8.1%) in G1 (Figure 3B). Although stabilization and accumulation of Neurog3 occurs during G1, previous work showed that Neurog3 protein is present during S phase, with rapid degradation occurring during G2-M (14). Corroborating that result, we show that while the majority (55% ± 3.6%) of Sox9+ Neurog3TA.pLO cells were in G1, 45% (± 3.6%) were in S-G2-M (Figure 3B). These findings show that the Neurog3 protein level in cycling Sox9+ Neurog3TA.LO progenitors is lowest during S-G2-M and highest during G1.
Previous work shows that a low Neurog3 protein level maintains a mitotic, endocrine lineage-primed progenitor state (1,4–7). Given the cell-cycle-dependent variation of Neurog3 protein level, we hypothesized that the low-level accumulation of Neurog3 in Neurog3TA.LO progenitors in G1 could trigger gene expression changes that were consistent with endocrine lineage-priming, and involving genes other than solely Neurog3. Therefore, intraepithelial Neurog3P2A.FUCCI+ cells (Neurog3TA.LO progenitors) were isolated from E14.5 Neurog3P2A.FUCCI pancreatic explants by flow sorting of lumen-contacting (Muc1+) cells, then sorting cells in S-G2-M (mKO2- mVenus+) or G1 (mKO2+ mVenus-) (Figure 4A). As described above, actively delaminating Neurog3TA.HI cells display relatively bright, yet to be degraded, cytoplasmic mVenus and high nuclear mKO2 (Figure 3A). To exclude this population, the flow-cytometry gating was set so that mVenus/mKO2 co-positive cells were not collected (Figure 4A). Analysis via qRT-PCR showed that while in S-G2-M, Neurog3TA.LO progenitors are enriched for Sox9 and Hes1 (mitotic endocrine-progenitor markers) with low expression of Neurog3 and several markers indicating forward progression towards endocrine commitment and further differentiation (NeuroD1, Insm1, Glucagon, Insulin) (Figure 4B). Neurog3TA.LO progenitors in G1, however, showed significantly decreased Hes1 and increased Neurog3, NeuroD1, Insm1, Glucagon and Insulin (Figure 4B). Despite the increases in endocrine-commitment markers, entrance into G1 did not significantly alter Sox9 expression (Figure 4B), demonstrating that these cells are intraepithelial Neurog3TA.LO progenitors. The data are also consistent with the idea that cells in this mitotic progenitor state, when in G1, initiate expression of several genes representing the downstream endocrine differentiation program, at a stage prior to commitment. Given the role of Neurog3 in trans-activating NeuroD1 and Insm1 (19–21), we speculate that the low-level accumulation of Neurog3 specifically in G1 could be sufficient to induce low-level NeuroD1/Insm1 expression in lineage-primed progenitors. It is also possible that signals initiating the lineage-primed state activate low-level expression of other transcription-factor genes in a Neurog3-independent manner. It is plausible that the concerted expression of several trans-acting factors establishes a relatively weak or incomplete form of the GRN that is normally considered to work only in post-mitotic committed cells. It would be important to discover if entrance into G1 were also linked to alterations in chromatin architecture and DNA accessibility that allow low-level expression of GRN member genes contributing to lineage priming in Neurog3TA.LO progenitors. Understanding how cell-cycle progression regulates such gene expression programs could lead to understanding if the final hormone-secreting cell fate might become preconditioned in the mitotic lineage-biased stage, and possibly how to manipulate cells at this early phase of their lifespan to improve the generation of functional endocrine cells.
Materials and Methods
Mice and transgene copy number analysis
Animal protocols were approved by the Vanderbilt University Institutional Animal Care and Use Committee. All animals were PCR genotyped. Sequences for genotyping primers are listed in Supplemental Table 1. Generation of the Neurog3RG BAC transgene and subsequent derivation of the Neurog3RG2 mESC line and Neurog3RG2 reporter mice was described previously (6). Although not previously reported in (6), mouse ES cells that stably integrated the Neurog3RG BAC LCA were analyzed by a qPCR-based assay that accurately estimates transgene copy number (17). Briefly, we generated primers specific for the puromycin-resistance gene (PuroR) in the PuroR-ΔTK-em7-NeoR (PuroΔTK) selection cassette in the Neurog3RG transgene (Figure 2A). Quantitative PCR was run on 2.5, 10, 20, 40 and 200 ng of genomic DNA from a TL1 mESC knock-in line, carrying one copy of the PuroΔTK cassette inserted via homologous recombination, to yield a ΔCT curve reflecting copy number (Figure 2B). Triplicate runs of exactly 20 ng of DNA from 23 candidate Neurog3RG mESC lines used the standard curve to define copy number.
Generation of P2A.FUCCI transgene and the Neurog3P2A.FUCCI reporter mouse line
To generate the mKO2-hCdt1(30/120)-P2A-mVenus-hGem(1/110) (P2A.FUCCI) cassette, mKO2-hCdt1(30/120) and mVenus-hGem(1/110) were PCR-amplified from plasmids provided by Dr. Atsushi Miyawaki (RIKEN Brain Science Institute) (Sakaue-Sawano et al. 2008). Amplification of mKO2-hCdt1(30/120) involved attaching a 40 bp Neurog3 homology region 5′ of the mKO2 start codon along with the first 25 base pairs of a P2A sequence 3′ of mKO2. Amplification of mVenus-hGem(1/110) involved attaching a 5′ BamHI site and a 3′ ApaI site. A third PCR was used to generate a P2A cassette with 25 base pairs of the 3′ end of mKO2-hCdt1(30/120) attached to its 5′ end and a BamH1 site at its 3′ end. The resulting mKO2-hCdt1(30/120) and P2A amplicons were then fused together by overlap extension PCR (22), using a forward primer specific for the mKO2-hCdt1(30/120) amplicon and a reverse primer specific for the P2A amplicon. The resulting mKO2-hCdt1(30/120)-P2A amplicon was attached to the mVenus-hGem(1/110) amplicon via the BamHI site and inserted into a pBS KS(-) vector. The resulting P2A.FUCCI cassette was removed from pBS KS (-) and inserted into a pCMV5 vector with a PGK-neomycin selection cassette for expression in HeLa cells (described below). The P2A.FUCCI cassette was also inserted in place of the RG cassette in the PL451-RG-FRT-PuroR-ΔTK-em7-NeoR-FRT-lox2272 vector described previously (6). Using BAC recombineering the resulting P2A.FUCCI-FRT-PuroR-ΔTK-em7-NeoR-FRT-lox2272 cassette was inserted immediately upstream of the Neurog3 start codon in the Neurog3-containing RPCI-23-121F10 BAC (6). Using BAC recombineering, the P2A.FUCCI-FRT-PuroR-ΔTK-em7-NeoR-FRT-lox2272 cassette was retrieved into a vector containing a lox66 site in a manner that ensured that placement of the lox66 site precisely mimicked that of its lox71 counterpart in the lox71/lox2272 flanked Neurog3RG BAC LCA. Subsequently, the FRT-flanked PuroR-ΔTK-em7-NeoR cassette was replaced with an FRT-flanked PGK-HygroR selection cassette. This final lox66/lox2272 flanked Neurog3P2A.FUCCI exchange plasmid was linearized and used to replace, via RMCE, the Neurog3RG BAC LCA in the Neurog3RG2 mESC line. Successful replacement with Neurog3P2A.FUCCI was verified by PCR (Figure Supplement 2B). A single, verified, Neurog3P2A.FUCCI mES cell line was expanded, karyotyped and injected into blastocyst-stage embryos to derive the Neurog3P2A.FUCCI reporter mouse strain. The LCA capability of the Neurog3RG1 mESC line was also tested and shown to allow efficient RMCE of lox66/lox2272 flanked cassettes (data not shown).
Cell culture
HeLa cells were cultured on tissue culture grade plastic at 37° C in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% fetal bovine serum (FBS), and 100 U/mL penicillin-streptomycin. Cells were passaged by adding 0.05% trypsin-EDTA to a plate of semi-confluent (<90%) cells. To test expression of the P2A.FUCCI reporter, HeLa cells were transiently transfected with the CMVP2A.FUCCI expression plasmid using Lipofectamine 2000 (Thermo Fisher) according to manufacturer’s instructions. The same conclusions were obtained with stable clonal lines expressing CMVP2A.FUCCI selected for neomycin resistance over 14 days (not shown).
Immunodetection
E14.5 dorsal pancreata were fixed in 4% paraformaldehyde (4 hrs, 4°C) then equilibrated in 30% sucrose overnight at 4°C). A Leica CM3050S was used sucrose-equilibrated, OCT-embedded tissue (Tissue-Tek) into 10 μm tissue sections, sequentially placed on three separate sets of slides, each covering ~33% of the dorsal pancreas. Primary and Secondary antibodies are listed in Supplemental Table 2. All images are epifluorescence from a Zeiss ApoTome microscope with Zeiss Axiovision software.
Flow sorting and qRT-PCR analysis
Multiple E14.5 Neurog3P2A.FUCCI+ dorsal pancreata were pooled and dispersed into a single-cell suspension using Accumax (Sigma) (protocol available on request). Dispersed samples were washed and incubated on ice, first with Muc1 antibody for 1 hr, then anti-hamster Cy5 secondary antibody for an additional hour. DAPI was added to ensure sorting of viable cells. Flow sorting used a BD FACSAria III. cDNA was generated using iScript cDNA synthesis kit (BioRad) from RNA isolated from flow-sorted cells after TRIzol extraction. PCR was performed in a Bio-Rad CFX96 with SsoFast EvaGreen Supermix (Bio-Rad) using at least three technical replicates. Relative expression level (normalized to Gapdh) was calculated by first assessing the ΔCT between the gene of interest and Gapdh before converting the ΔCT to relative expression level (2ΔCT). The results in Figure 4 were independently repeated (biological replicate) with similar results. Primer sequences, except for Insm1 primers (Applied Biosystems), are listed in Supplemental Table 1.
Quantification and statistics
Cell counting and fluorescence intensity quantifications were done using NIH ImageJ software. For quantifications “n” indicates total cells counted, with “N” number of individual dorsal pancreata analyzed. As previously stated approximately 33% of an entire dorsal pancreas was analyzed for each dorsal pancreas. Previous reports indicate that only 2% of the total pancreas volume needs to be systematically sampled and analyzed to obtain a relative error of ≤ 10% (23). Error bars generated using standard error of the mean (SEM), with Student’s t-test (one-tailed) used to calculate p values. p values were deemed significant when ≤ 0.05.
Competing Interests
The authors declare that no competing interests exist.
Acknowledgements
We thank Atsushi Miyawaki (RIKEN Brain Science Institute) for the mKO2-hCdt1(30/120)/pCSII-EFMCS and mVenus-hGem(1/110)/pCSII-EF-MCS plasmids. This work utilized the Cell Imaging Shared Resource and Transgenic/ES Cell Shared Resource core facilities of the Vanderbilt Diabetes Research and Training Center funded by NIDDK grant 020593. Flow cytometry was performed in the VUMC Flow Cytometry Shared Resource supported by the Vanderbilt-Ingram Cancer Center (P30 CA68485) and the Vanderbilt Digestive Disease Research Center (DK0558404). Generation of Neurog3RG2 and Neurog3P2A.FUCCI mice was supported in part by the Beta Cell Biology Consortium Mouse ES Cell Core funded by the NIDDK (U01DK072473). We thank Anna Means, Guoqiang Gu, and members of the Wright/Gu labs for discussions. This study was supported by the NIH/NIDDK (U01DK089570) and an American Heart Association fellowship to MB (13POST14240011).