Purification of replicating pancreatic β-cells for gene expression studies

β-cell proliferation is a rare event in adult pancreatic islets. To study the replication-related β-cell biology we designed a replicating β-cells sorting system for gene expression experiments. Replicating β-cells were identified by EdU incorporation and purified by flow cytometry. For β-cell separation islet cells were sorted by size, granularity and Newport Green fluorescence emission that was combined with emitted fluorescence for EdU-labelled replicating cells sorting. The purity of the resulting sorted populations was evaluated by insulin staining and EdU for β-cell identification and for replicating cells, respectively. Total RNA was isolated from purified cell-sorted populations for gene expression analysis. Cell sorting of dispersed islet cells resulted in 96.2% purity for insulin positivity in the collected β-cell fraction and 100% efficiency of the EdU-based cell separation. RNA integrity was similar between FACS-sorted replicating and quiescent β-cells. Global transcriptome analysis of replicating vs quiescent β-cells showed the expected enrichment of categories related to cell division and DNA replication. Indeed, key genes in the spindle check-point were the most upregulated genes in replicating β-cells. This work provides a method that allows for the isolation of replicating β-cells, a very scarce population in adult pancreatic islets.

Pancreatic β-cells are the unique cell type in the body that produces and secretes insulin in response to small variations in blood glucose levels to tightly control systemic glucose homeostasis. Β-cell mass reduction is a central event in the development of type 1 and type 2 diabetes, and β-cell regeneration is a potential curative treatment of the disease.
Β-cell replication is the primary mechanism of β-cell mass expansion in adult individuals, and represents a target for diabetes treatment to increase the functional β-cell mass. However, β-cell proliferation is a rare event in adult pancreatic islets [1][2][3][4][5] and varies within a range of 0.4% to 6% of β-cells per day depending on age 5 . This low proportion of replicating β-cells inside the islet precludes molecular analysis of β-cell replication-related pathways in entire islets, since it may be masked by the highly abundant post-replicative islet cells. Recently, Klochendler and colleagues have developed a transgenic mouse model where cycling cells throughout the body become GFP-labelled allowing cell sorting of live replicating cells 6 . Despite the advantages provided by this system, its use is restricted to the study of β-cell replication in mouse islets.
On the other hand, nucleoside analogues have been extensively used for the identification of replicating cells. They are incorporated into the replicating DNA strands during the S phase of the cell cycle, and the most extensively used are BrdU, CldU and IdU. In order to visualize the incorporated nucleosides, all of them require the use of DNA denaturation facilitating sterical access of antibodies to the nucleosides. Alternatively, 5-ethynyl-20-deoxyuridine (EdU) is structurally similar to the natural nucleoside in which a terminal alkyne group replaces the methyl group in the 5th position. EdU detection is based on a copper-catalyzed covalent reaction between a dye-conjugated azide and the alkyne group of the EdU, known as Click chemistry 7 . The small sized dye-azide complex allows for efficient EdU detection avoiding harsh conditions that degrade the structure of the cells 8 .
Here, we sought to develop a method for replicating β-cells sorting based on EdU incorporation, suitable for global gene expression analysis and applicable to most of experimental designs and animal species.

Methods
Islet isolation and culture. FELASA guidelines and recommendations for the use of laboratory animals were followed (European and local government guidelines) and animal procedures were reviewed and approved by the Animal Research Committee of the University of Barcelona (identification number: DAAM 7082).
Islets from young adult male Wistar rats (Janvier, Saint Berthevin, France) were isolated by collagenase (Collagenase P; Boehringer Mannheim Biochemicals, Mannheim, Germany) digestion of the pancreas as previously described 9 . Isolated islets were hand-picked under a stereomicroscope two or three times, until a population of pure islets was obtained. Islets were cultured in RPMI 1640 (Sigma Immunochemicals, St Louis, MO, USA) supplemented with 100 U/ml penicillin and 100 mg/ml streptomycin containing 10% heat-inactivated foetal bovine serum (FBS) at 37 °C in a humidified atmosphere with 5% CO2. Thymidine analogues, 5-Bromo-2′-deoxyuridine (BrdU, Sigma) and 5-Ethynyl-2′-deoxyuridine (EdU, Molecular probes, Life Technologies, Eugene, OR, USA) were added into the culture medium at final concentration of 10 µM. For the analysis of continuous labelling with EdU, islets were cultured in 5.5 mM or 22.2 mM glucose with or without EdU for 7 and 14 days. Culture medium was replaced on a daily basis and BrdU was added to the islets for the last 24 h of culture. For replicating β-cells sorting, islets were cultured in 5.5 mM glucose with EdU overnight starting on the day of isolation.
Cell dispersion and labelling. In order to strengthen β-cell identification, dispersed cells were labelled with Newport Green DCF, diacetate (NG, Invitrogen, Carlsbad, CA, USA). The diacetate form of NG requires esterase-mediated cleavage for subsequent binding to Zn 2+ . Once cleaved, NG becomes membrane-impermeant and is retained inside live β-cells 10 . For replicating cell identification, the DNA-incorporated EdU was detected by azide alkyne Huisgen cycloaddition reaction. The EdU contains an alkyne which can be reacted with an azide-containing detection reagent, to form a stable triazole ring (Click reaction), eliminating the use of antibodies for detection and subsequent major permeabilization-based protocols 7 .
Cultured islets were washed with Ca 2+ and Mg 2+ -free PBS prior trypsinization. Islets were disaggregated by gentle continuous pipetting in trypsin-EDTA 0.05% (Sigma) with DNase (1 mg/ml) (RQ1 RNases-Free DNase, Promega, Madison, WI, USA) for 6 min. in a water bath at 37 °C 11 . Dispersed islet cells were washed once with PBS and incubated with 25 µM NG at 37 °C for 30 min, protected from light. Following one 1% BSA-PBS wash, EdU labelling was performed by means of the Click-iT ® EdU Alexa Fluor ® 647 Flow cytometry Assay Kit (Invitrogen, Eugene, OR, USA). Briefly, cells were fixed in 2% paraformaldehyde-PBS at room temperature (RT) for 15 min., and washed once with 1% BSA-PBS before mild permeabilization with a saponin-based reagent for additional 15 min. at RT. The click reaction buffer was prepared following the manufacturer's instructions and added to the washed cells for 30 min. at RT. Prior to flow cytometry, labelled cells were washed twice with FACS buffer (PBS, 2 mM EDTA, 0.5%BSA) and transferred into 35 µm nylon mesh cell strainer caped-tubes (BD biosciences, Erembodegem, Belgium). 400 units of RNasin plus RNase inhibitor (Promega) were added to the buffer or solution used in each step from the cell fixation with PFA until cell sorting.
Cell sorting. Unstained cells and FMO (fluorescence minus one), treated in the same way, were used as negative control in each experiment and served as autofluorescence control. Sorting was done using a Beckman-Coulter MoFlo Astrios cell sorter equipped with a 100-μm flow tip and operated at a sheath pressure of 25 psi. The laser illumination power was set to 150 mW for NG excitation at 488 nm, and to 100 mW for Alexa 647 excitation at 640 nm. The barrier filters were 517/26 nm for NG, and 671/30 nm for EdU Alexa 647 fluorescence. Before cell sorting, cells were filtered through a 35 μm strainer to eliminate all remaining cell clumps. Initially, doublets are excluded by using pulse processing (FSC-H vs. FSC-A) followed by the exclusion of debris gating on a two physical parameter dot plot (FSC/SSC). Β-cells were gated as FCS hi /SSC hi /NG hi . Mitotic and post-mitotic cell subpopulations were selected as EdU+ or EdU-respectively. An average sorting rate of 200 events per second was maintained. Sorted cells were collected in 150 μl of PKD buffer (Qiagen).
Immunofluorescence. Cultured islets were fixed overnight in 4% paraformaldehyde-PBS at 4 °C, embedded in paraffin, sectioned and immunostained after deparaffinization and rehydration.
For EdU detection in cultured islets the click chemistry-based Click-iT EdU assay (Click-iT ® EdU Alexa Fluor ® 594 Imaging Kit; Molecular Probes, Life Technologies) was used according to the manufacturer's instructions. Briefly, sections were permeabilized with saponin and incubated with the Click-iT reaction cocktail for 1 h at room temperature. This labelling method was combined with insulin immunofluorescence.
Β-cell purity in sorted populations was determined by insulin and c-peptide immunofluorescence as follows. Sorted populations were directly collected in 1% BSA-PBS, washed twice with PBS and incubated overnight with the rabbit anti-insulin antibody (1/100) or mouse-anti c-peptide antibody (1/50, Abcam) diluted in 1% BSA-PBS + 0.3% triton. Donkey anti-rabbit Alexafluor-488 labelled antibody (1/400) was used for β-cell visualization. Cells were loaded into the haemocytometer for counting. At least 200 cells were counted for each sorted population. The experiment was replicated 8 times with islets isolated from 8 different rats. or Complete Whole Transcriptome Amplification Kit (Sigma WTA2) following the manufacturer's protocol. RNA quality was assessed by using the Agilent RNA Nano kit in the Bioanalyzer 2100 (Agilent Technologies, Inc., Palo Alto, CA).
cDNA synthesis was performed from 200 ng of amplified RNA using the High Capacity cDNA Reverse transcription kit (Applied Biosystems, Life Technologies). qPCR was run in a 7900HT Fast Real-Time PCR system (Applied Biosystems). Gene expression data was analysed with the free-access software Gene Expression Suite v1.0.3 (Applied Biosystems). A full listing of assays, gene names and assay identification numbers is given in Table 1.
Global gene expression profiling. Total RNA was amplified using the TransPlex Complete Whole Transcriptome Amplification Kit (Sigma WTA2) to generate the cDNA. Cyanine-3 (Cy3) labelled cDNA was prepared from 500 ng of double stranded cDNA using the SureTag Complete DNA Labelling kit (Agilent 5190-4240) according to manufacturer's instructions. Dye incorporation and cDNA yield were checked with the NanoDrop ND-1000 Spectophotometer. 600 ng of Cy3 labelled cDNA was directly mixed with hybridization buffer and Agilent blocking agent incubated at 95 °C for 3 minutes and immediately transferred on ice. Then, it was hybridized to Agilent SurePrint G3 Rat Gene Expression v2 8 × 60 K Microarray (amadid ID: 074036) for 40 h at 65 °C in a rotating Agilent hybridization oven. After hybridization, microarrays were washed 1 minute at room temperature with GE Wash Buffer 1 (Agilent) and 1 minute with 37 °C GE Wash buffer 2 (Agilent), then dried immediately by brief centrifugation. Microarrays were scanned on an Agilent G2539A scanner at 3 µm resolution and 100%PMT. The intensity data of each individual hybridization were extracted and the quality was assessed with the Feature Extraction software 10.7 (Agilent). Samples were processed in triplicates: three quiescent β-cells samples, and three replicative β-cells samples. Raw data was corrected for background noise using the normexp method. Quantile normalization was applied to assure comparability between samples. Differential expression analysis was carried out on non control probes with an empirical Bayes approach on linear models. A paired test was applied, taking into account that the cells were separately sorted from 3 different islets preparations. Results were corrected for multiple testing according to the False Discovery Rate (FDR) method 13 . All pre-processing and statistical analyses were performed with the Bioconductor project in the R statistical environment, in particular the limma package 14,15 . For Gene Set Enrichment Analysis (GSEA), rattus norvegicus Gene Ontology gene sets were downloaded from the Gene Ontology website (www.geneontology.org/page/gene-associations/gene_association.rgd.gz) and transformed into a format suitable for the GSEA software. P-values were computed using a t-Test, using 1000 permutations ("gene_set" permutation type) and the default Signal2Noise metric. Nominal p-values were then corrected for multiple testing using the False Discovery Rate. For EnrichR analysis, probes were selected based on p-value lower than 0.001 and absolute linear fold change higher than 2. Median intensities are retrieved for a gene if it is represented by multiple probes. Those genes were fed into the EnrichR software to

Results
Continuous labelling with EdU ex vivo. Long term labelling of mice with halogenated thymidine analogues has been found to inhibit β-cell proliferation in vivo 17 . In order to evaluate the effects of EdU on β-cell replication and viability, rat pancreatic islets were cultured with or without 10 µM EdU at basal and high glucose concentrations. B-cell replication and apoptosis were unaffected by one week exposure to EdU at both glucose concentrations (Fig. 1A,C). After 14 days exposure to EdU, β-cell replication remained unaffected but β-cell apoptosis was significantly increased (Fig. 1B,D). Accordingly, the percentage of EdU labelled β-cells at the end of the 7 days-culture was approximately 7-fold over the 1 day labelling with BrdU (5.5 mM glc and 24 h-BrdU pulse: 0.51 ± 0.08% labelling; 5.5 mM glc and 7days-EdU pulse: 4.3 ± 0.48% labelling); and longer pulse of EdU did not result in further β-cell labelling (5.5 mM glc and 14days-EdU pulse: 2.45 ± 0.94% labelling) (Fig. 1E,F). Therefore, continuous exposure of cultured pancreatic islets to EdU for one week is suitable for β-cell replication experiments, whereas longer cultures should be avoided.  further improves β-cell identification 22 . EdU labelling requires cell fixation and mild permeabilization to allow the Alexa Fluor-labelled azide molecule reach the DNA-integrated thymidine analogue in replicating cells. Cell fixation and permeabilization resulted in loss of intracellular NG with subsequent poor green fluorescence in β-cells and reduced size (Supplementary Figure 1). Nevertheless, pure β-cells were obtained by sorting islet cells by size, granularity and green fluorescence emission ( Fig. 2A). Four cell fractions were collected: Replicating β-cells (SSC hi , FSC hi , NG + , EdU + ), Quiescent β-cells (SSC hi , FSC hi , NG + , EdU − ), Replicating non-β-cells (SSC low , FSC low , NG − , EdU + ) and Quiescent non-β-cells (SSC low , FSC low , NG − , EdU − ). One hundred per cent of the cells collected in the replicating β-cell fraction was positive for EdU and 96 ± 1% and 98.8 ± 1.2% stained for insulin and c-peptide, respectively (Fig. 2B,C). In the non-replicating β-cell population, 100% of cells were negative for EdU and 95 ± 1%, and 99.5 ± 0.07% stained positive for insulin and c-peptide, respectively (Fig. 2B,C). In non-β-cell fractions 7.8 ± 0.1% of cells stained positive for insulin. The number of cells collected per rat pancreas ranged from 760 to 1200 cells in the proliferative β-cell fraction and from 4 × 10 4 to 6 × 10 4 cells in the quiescent β-cell fraction.
RNA isolation from quiescent and replicating sorted cells. RNA was isolated from replicating and quiescent β-cell fractions. RNA extraction from the quiescent β-cell fraction that consisted of approximately 2 × 10 4 cells, was up to 160 ng of RNA (~8 pg RNA/cell), whereas for the replicating fraction with ≈1000 cells, the collected RNA yielded ~5 ng of RNA (~5 pg of RNA/cell). Similar results were obtained in the RNA extraction from 1000 quiescent β-cells, excluding the possibility that replicating β-cells had lower content of RNAs. Thus, the efficiency of RNA extraction correlated with the number of cells in the fraction. mRNA integrity analysed by the Agilent 2100 Bioanalyzer was similar between the two collected fractions, and electropherograms displayed the expected RNA mobility curve shown for other cell types 23 . For gene expression analysis, equal amounts of retro-transcribed cRNA were loaded for replicating and quiescent β-cell samples. Expression of the housekeeping genes (18S, hpdt1, rpl30 and rps17) was similar between replicating and quiescent β-cells, indicating that the amount of loaded mRNA was similar in both groups (Fig. 3A).
In order to confirm the purity of the collected fractions, a panel of pancreatic cell types and proliferation markers were analysed in both fractions. Gene expression of amylase (amy), keratin 20 (krt20) and pecam1 were undetectable in our samples, indicating the absence of contaminating acinar, duct or endothelial cells (data not shown). However, mRNAs encoding for glucagon (gcn) and somatostatin (sst) were detected in 2 out of 11 samples from both fractions, and pancreatic polypeptide (ppy) in 6 out of 11. Finally, vimentin the major cytoskeletal component of mesenchymal cells was detectable in both β-cell fractions and levels of expression in the replicating β-cell fraction were doubled compared to the quiescent β-cell fraction. α-smooth muscle actin (acta2), a marker for the epithelial to mesenchymal transition (EMT) was undetectable.
ki67 and pcna mRNAs, two established markers of cell cycle progression 24 were significantly higher in replicating β-cells compared with the non-replicating β-cell population. Similarly, the Polycomb Repressive Complex (PRC) members, bmi1 polycomb ring finger oncogene (bmi1) and the enhancer of zeste homolog 2 (ezh2) were upregulated in replicating β-cells (Fig. 3B). Both repressors regulate the expression of the product of the Ink4a locus (cdkn2a) that in turn negatively regulates cell cycle progression by targeting cdk4-cycD2 25,26 . Histone Deacetylase 5 (HDAC5), a Class IIa HDAC that is associated with cell cycle progression in several cell types 27 was also upregulated in the replicating β-cell population. Finally, gene expression of the forkhead box protein O1 (Foxo1) and insulin (ins) was significantly downregulated in replicating β-cells. These results confirm the enrichment of replicating cells in the EdU+ β-cell fraction.
Genes up-regulated in replicating β-cells. We next analysed the global transcriptome of replicating and quiescent β-cells using gene expression arrays. A high variation within experimental groups did not allow to select genes of interest based on the adjusted p-value. However, different selection criteria (non-corrected p-value inferior or equal to 0.05 and absolute log2 ratio superior or equal to 1) allow to extract 5500 differentially expressed genes (2043 and 3457 down-and up-regulated in the replicating β-cells, respectively). Heat-map of most changing genes revealed good clustering of the samples within each experimental group (Fig. 4A). Gene Ontology analysis of the replicating β-cell signature showed the expected enrichment of categories related to cell division and DNA replication, such as Cell Division (GO:0051301), Centrosome (GO:0005813), DNA-binding (GO:0003677) or ATP-binding (GO:0005524) (Fig. 4B,C). Comparable results were obtained with the EnrichR and GSEA software indicating the consistency of the functional analysis.
The genes most significantly upregulated in replicating β-cells are mainly involved in the control of G2-M phases of the cell-cycle (Table 2 and Fig. 4D). Interestingly, the Slc16a1 gene that encodes for the Monocarboxilate transporter 1, which conveys pyruvate and lactate through the plasma membrane and its expression is negligible in mature β-cells, was significantly upregulated in replicating β-cells.

Discussion
This study reports a method for the purification of replicating and quiescent pancreatic β-cells for downstream gene expression analysis. This technique is based on the incorporation of EdU to the replicating DNA strands during the S phase of the cell cycle, and makes it compatible with diverse experimental designs and animal species.
Nucleotide analogues have been extensively used for the identification of replicating cells. However, long term exposure to BrdU and IdU has been associated with DNA instability, DNA stress and cell death 17 . In the present study we have determined the effects of long-term EdU exposure on β-cell replication and survival. We show that β-cell replication and viability are unaffected in cultured islets exposed to 10 μM of EdU for 7 days; whereas longer exposure has a negative impact on β-cell survival. EdU, in contrast to other thymidine analogues such as IdU, CldU or BrdU, does not require the use of HCL-based DNA denaturation for antigen retrieval and its visualization is free from secondary antibodies 7 , which results in the use of milder and quicker protocols for its detection and preserves RNA integrity. However, cell fixation with PFA, a mandatory step in the EdU labelling by Click reaction, significantly influenced RNA extraction yield. PFA fixation is assigned to its reactivity and ability to modify proteins and nucleic acid bases resulting in RNA-protein cross-linking 28 , which compromises RNA isolation from fixed cells. Accordingly, the inclusion of a protein digestion step in the RNA extraction protocol, with the addition of Proteinase K in the cell lysis buffer, significantly improved the quality and yield of the RNA.
Islet cell sorting according to cell size, granularity and NG labelling resulted in pure β-cell preparations. Notably, amylase and keratin 20 RNAs were undetectable in β-cell preparations. However, non β-cell pancreatic islet hormones (glucagon, somatostatin and pancreatic polypeptide) RNAs were detectable in some samples, indicating that contamination with endocrine non β-cells was not completely avoided.
Differential gene expression showed the expected upregulation of cell cycle -related genes in the replicating β-cell fraction, which confirms the replicative nature of the sorted cells. Additionally, Foxo1 was significantly downregulated in replicating β-cells. Nuclear accumulation of Foxo1 has been related to cell longevity, indeed, increased Foxo1 expression has been found in aged β-cells which are largely quiescent 29,30 . Therefore, the lower expression of foxo1 in replicating β-cells is in accordance with the inverse relationship between Foxo1 expression and β-cell replication. Finally, insulin 2 gene expression was downregulated in replicating β-cells. In accordance with our results, recent reports have shown a significant reduction on insulin synthesis prior to β-cell replication 31,32 .
Global transcriptome of replicating and quiescent β-cell fractions showed an enrichment in cell-cycle-related categories such as centrosome, DNA-replication and DNA-binding indicating that the EdU-based sorting method efficiently provided pure preparations of replicating β-cells. Key components of the G2 and M phases of the cell-cycle were identified in replicating β-cells. Of note, FoxM1, plk1 and cenp were upregulated in replicating β-cells. This axis has been shown to play a crucial role in insulin-stimulated β-cell proliferation [33][34][35] . Other regulators of the M phase of the cell-cycle such as pttg1 (Pituitary Tumor-Transforming 1, securin), ska3 (Spindle and Kinetochore Associated Complex Subunit 3) and spc25 (SPC25, NDC80 Kinetochore Complex Component) with unclear roles in β-cell expansion, were also upregulated in replicating β-cells. Another gene that was significantly induced in replicating β-cells was Slc16a1 that encodes for monocarboxilate transporter 1 (MCT1). Expression of MCT1 is specifically repressed in adult β-cells which prevents lactate and pyruvate permeability through the  plasma membrane, and allows for accurate glucose-induced insulin secretion 36,37 . Thus, replicating β-cells could incur in inappropriate insulin release in response to glucose levels. Further work will reveal whether the genetic program of cycling β-cells is completely restored in daughter cells after mitosis. In summary, we present a method that allows for the isolation of replicating β-cells, even in situations where they are very scarce. We anticipate that it will be particularly useful for the study of β-cell replication of human islets, in which replicating β-cells are infrequent.
Therefore, we expect this method to have a broad applicability in the study of β-cell proliferation in the context of diabetes and β-cell regeneration.