We describe a method for the isolation of endocrine cells from embryonic, neonatal and postnatal pancreases followed by single-cell RNA sequencing. This method allows analyses of pancreatic endocrine lineage development, cell heterogeneity and transcriptomic dynamics.
Pancreatic endocrine cells, which are clustered in islets, regulate blood glucose stability and energy metabolism. The distinct cell types in islets, including insulin-secreting β cells, are differentiated from common endocrine progenitors during the embryonic stage. Immature endocrine cells expand via cell proliferation and mature during a long postnatal developmental period. However, the mechanisms underlying these processes are not clearly defined. Single-cell RNA-sequencing is a promising approach for the characterization of distinct cell populations and tracing cell lineage differentiation pathways. Here, we describe a method for the single-cell RNA-sequencing of isolated pancreatic β cells from embryonic, neonatal and postnatal pancreases.
The pancreas is a vital metabolic organ in mammals. The pancreas is comprised of endocrine and exocrine compartments. Pancreatic endocrine cells, including insulin-producing β cells and glucagon-producing α cells, cluster together in the islets of Langerhans and coordinately regulate systemic glucose homeostasis. Dysfunction of the endocrine cells results in diabetes mellitus, which has become a major public health issue worldwide.
Pancreatic endocrine cells are derived from Ngn3+ progenitors during embryogenesis1. Later, during the perinatal period, the endocrine cells proliferate to form immature islets. These immature cells continue to develop and gradually become mature islets, which become richly vascularized to regulate blood glucose homeostasis in adults2.
Although a group of transcriptional factors has been identified that regulate β cell differentiation, the precise maturation pathway of β cells is still unclear. Moreover, the β cell maturation process also involves the regulation of cell number expansion3,4 and the generation of cellular heterogeneity5,6. However, the regulatory mechanisms of these processes have not been well studied.
Single-cell RNA-sequencing is a powerful approach that can profile cell subpopulations and trace cell lineage developmental pathways7. Taking advantage of this technology, the key events that occur during pancreatic islet development can be deciphered at the single-cell level8. Among the single-cell RNA-sequencing protocols, Smart-seq2 allows the generation of full-length cDNA with improved sensitivity and accuracy, and the use of standard reagents at lower cost9. Smart-seq2 takes approximately two days to construct a cDNA library for sequencing10.
Here, we propose a method for the isolation of fluorescence-labeled β cells from the pancreases of fetal to adult Ins1-RFP transgenic mice11, using fluorescence-activated cell sorting (FACS), and the performance of transcriptomic analyses at the single-cell level, using Smart-seq2 technology (Figure 1).This protocol can be extended to analyze the transcriptomes of all pancreatic endocrine cell types in normal, pathological and aging states.
All methods described here have been approved by the Institutional Animal Care and Use Committee (IACUC) of Peking University.
1. Pancreas Isolation
2. Collagenase Digestion and Islet Isolation
3. Trypsin Digestion of Pancreatic Tissue or Islets
4. Single-cell Lysis
5. Single-cell cDNA Amplification
6. cDNA Library Construction
7. DNA Sequencing
8. Bioinformatics Analyses
Pancreases were dissected from embryonic, neonatal and postnatal mice (Figure 2A and 2B). For mice older than postnatal day 18, the digestive effect depends on the degree of perfusion; therefore, the injection is the most important step for islet isolation (Figure 2C-2E and Table 6). As much collagenase was injected as was possible to fill the pancreas during this step. The fully inflated pancreas is shown in Figure 2D. If the perfusion is not successful (Figure 2E), but the sample is precious, the pancreas can be torn into small pieces for sufficient digestion later.
After perfusion, the pancreatic tissue was digested into small pieces to release islets (Figure 2F). To shorten the FACS sorting time, we enriched the endocrine cells by picking the islets in advance. The size of the islets may vary depending on the mouse age and the digestion intensity. Occasionally, the islets are not round in shape. Islets should be selected according to color and compact state (Figure 2G and Table 6). If the transgenic mouse strain has a reporter gene, such as GFP or RFP for endocrine cells, the islets could also be picked under a fluorescence microscope.
Ins1-RFP+ cells were purified by FACS sorting (Figure 3A-3C), and then single cells were picked using a capillary pipette for single-cell RNA-seq (Figure 3D). Successfully amplified cDNA should have a full length above 500 bp and be enriched from 1.5 kb to 2 kb. Moreover, there is usually a 500-600 bp enrichment of cDNA observed in the Ins1-RFP+ cells, which may represent the insulin transcripts (Figure 4A). However, there have been some abnormal situations observed10 (Table 6). For example, the cDNA fragments near 100 bp are primer dimers (Figure 4B), which are usually caused by the excessive primers, that must be removed by repeating the DNA purification step. The existence of primer dimers may influence the calculations of total cDNA yields, which are used for the following library construction. The cDNA fragments between 100 bp and 500 bp usually represent degraded cDNA (Figure 4C), which is caused by bad cell status or reagent problems, such as RNase contamination. Under this condition, we should identify the cause of DNA degradation and exclude the disturbance. For instance, to ensure good cell status, tissues should be digested and cells should be sorted quickly and gently, and operations should be performed cautiously to avoid any type of pollutants.
After the purification of the cDNA libraries for sequencing, we obtained cDNA fragments of different sizes by following the instructions for adding different ratios of DNA purification beads to the sample. For example, we can obtain cDNA ranging from 250 bp to 450 bp by adding 0.7x and 0.15x DNA purification beads to the first and second rounds of purification, respectively (Figure 4D). This step has a high success rate. However, if there are fragments of approximately 100 bp, it is suggested to purify the libraries again with 1x DNA purification beads to remove dimers (Table 6). Unremoved dimers will skew the DNA quantification and will influence the sample pooling results, which results in the uneven data acquisition from each sample.
We analyzed the sequencing data with bioinformatics approaches (Figure 5A). The sequencing quality scores should be greater than 30 during sequencing quality evaluation (Figure 5B). After alignment, 80-90% of reads are expected to be mapped to the reference genome. To obtain high quality cells for downstream analyses, we excluded cells with fewer than 0.5 million mapped reads or with fewer than 4,000 detected genes (Figure 5C and 5D). After PCA and hierarchical clustering, we characterized different groups of cells and identified the genes that were heterogeneously expressed in different groups8.
Figure 1: Schematic overview for single-cell RNA-seq of mouse pancreatic endocrine cells. Please click here to view a larger version of this figure.
Figure 2: Pancreas dissection, perfusion and islet picking. (A) Dissection of embryonic pancreatic tissue (lower) from an E17.5 embryo (upper). The yellow dotted line demarcates the pancreatic tissue. Scale bar = 1,000 µm. (B) Dissection of postnatal pancreatic tissue (right) from a P10 mouse (left). Scale bar = 1,000 µm. (C-D) The pancreatic tissue of a P60 mouse before (C) and after (D) perfusion. The yellow dotted line demarcates the pancreatic tissue. White arrow indicates gallbladder. (E) The partially perfused pancreatic tissue of a P60 mouse. Red arrow indicates the well perfused area of the pancreas. Blue arrows indicates the poorly perfused areas of the pancreas. (F) The pancreatic tissue before (upper) and after (lower) collagenase digestion. (G) Arrows point to the islets that were released from collagenase digested pancreatic tissue. Scale bar = 200 µm. Please click here to view a larger version of this figure.
Figure 3: Manually picking cells with a 30–40 µm capillary pipette under a microscope. (A-C) Step-wise FACS gating for sorting Ins1-RFP+ cells. (D) The bright circles indicate cells and the tube is a capillary pipette. Pick the cells with better morphology (Arrow) and ignore the clustered cells (Arrowhead). Scale bar = 200 µm. Please click here to view a larger version of this figure.
Figure 4: Quality detection of cDNA and library size distribution of Ins1-RFP+ cells by a parallel capillary electrophoresis instrument. (A) The representative result of successfully pre-amplified cDNA. Normally, the cDNA profile should be above 500 bp with an ~1.5–2 kb peak. (B) The representative result of pre-amplified cDNA with a primer dimer peak. The primer dimer peak is approximately 100 bp. (C) The profile of pre-amplified cDNA with fragments between 100 bp and 500 bp indicates the possibility of cDNA degradation. (D) The size distribution of cDNA library ranging between 250 bp and 450 bp following the purification steps mentioned in this protocol. (E) The expression levels of Ins2 in cDNA libraries by qPCR (left) and relative sequencing data (right). The x-axes represent distinct 8 single cell samples. The y-axis (left) represents normalized ΔCt relative to Gapdh (CtGapdh– CtIns2 + 6) and the y-axis (right) represents normalized expression level of Ins2 relative to Gapdh (log2(TPMIns2 / TPMGapdh)). Please click here to view a larger version of this figure.
Figure 5: Bioinformatics analyses of single-cell transcriptome data. (A) Pipeline of bioinformatics analyses. (B) Sequencing quality scores across all bases of reads. (C) Distribution of mapped read count. (D) Distribution of detected gene count. Please click here to view a larger version of this figure.
Component | Volume (µL) |
Reverse transcriptase (200 U/µL) | 0.5 |
RNase inhibitor | 0.25 |
First-strand buffer (5x) | 2 |
DTT (100 mM) | 0.5 |
Betaine (5 M) | 2 |
MgCl2 (1 M) | 0.06 |
TSO (100 µM) | 0.1 |
Nuclease-free water | 0.29 |
Total | 5.7 |
Table 1: RT reagent mix components in a 5.7 µL reaction for each sample.
Component | Volume (µL) |
First-strand reaction | 10 |
DNA Polymerase (2x ReadyMix) | 12.5 |
IS PCR primers (10 µM) | 0.25 |
Nuclease-free water | 2.25 |
Total | 25 |
Table 2: PCR pre-amplification reagent mix components in a 25 µL reaction for each sample
Component | Volume (µL) |
SYBR Green Master Mix | 5 |
Primers (5 µM) | 0.5 |
Nuclease-free water | 2.5 |
cDNA | 2 |
Total | 10 |
Table 3: qPCR reagent mix components in a 10 µL reaction using a standard 384-well plate.
Component | Volume (µL) |
5x TTBL | 1.6 |
2 ng cDNA | Variable |
ddH2O | Variable |
TTE Mix V5 | 2 |
Total | 8 |
Table 4: Tagmentation reagent mix components in an 8 µL reaction for each sample.
Component | Volume (µL) |
ddH2O | 1.6 |
Product of previous step | 10 |
5x TAB | 4 |
N6XX | 2 |
N8XX | 2 |
TAE | 0.4 |
Total | 20 |
Table 5: Enrichment PCR reagent mix components in a 20 µL reaction for each sample.
Step | Problem | Possibility | Solution |
1.3.4 | Slipping of the clamps | The outer surface of the intestine is wet caused by the existence of the small amount of interstitial fluid and collegenase leakness | Gently swap the duodenum wall and other organ walls in the abdominal cavity with cotton swabs |
1.3.6 | Intestine burst | Liquid pressure is too high in the intestine | Slow down the injection speed |
2.6 | Islets stick to exocrine tissue when picking islets | Insufficent digestion | Prolong the digestion duration and shaking strength |
5.3 | The cDNA yield is low after PCR amplification | The cells are in bad condition | Keep cells fresh and in good condition |
5.4 or 6.4 | Primer dimers can be seen | Excessive primers | Purify the cDNA one more time with DNA purification beads (0.8:1 or 1:1 ratio) |
5.4 | Degraded cDNA after PCR amplification | Poor quality of cells or contaminated reagents | Keep cells in good condition or change reagents |
6.3 | The amount of DNA is low after library construction | Too few PCR cycles or the quality of cDNA is not good | Increase the number of cycles or ensure cDNA quality before library construction |
Table 6: Troubleshooting.
In this protocol, we demonstrated an effective and easy-to-use method for studying the single-cell expression profiles of pancreatic β cells. This method could be used to isolate endocrine cells from embryonic, neonatal and postnatal pancreases and to perform single-cell transcriptomic analyses.
The most critical step is the isolation of single β cells in good condition. Fully perfused pancreases respond better to subsequent digestion. Insufficient perfusion, which usually occurs in the dorsal pancreas, will result in a low islet yield. After perfusion, the digestion time and shaking intensity require special attention. Over-digestion, resulting from long incubation times and vigorous shaking, can break the islets into pieces. Insufficient digestion will not completely separate the islets from adjacent tissues. After digestion of the pancreas, the islets were purified by hand picking rather than by density gradient centrifugation. Although density gradient centrifugation can enrich islets, many small islets or islets connected with acinar tissue are mistakenly discarded. Try to avoid picking acinar tissue, which may cause inefficient trypsin digestion and reduce the purity of the β cells.
Independent biological replicates are necessary for distinguishing biological variability and batch effects. If two biological replicates have a similar pattern in PCA and include similar sub-groups in clustering results, these samples might reveal reliable biological variability. Otherwise, additional biological replicates are required to confirm the results. For a metabolic organ such as the pancreas, the cell state associated with the circadian clock24 may account for batch differences. To solve this problem, we recommend collecting all samples at the same time of day.
The limitation of this approach is low throughput because we have to construct a library for each cell. Due to the excessive primers in the reaction, the cDNA before and after library construction generally should be purified twice to completely remove the primer dimers. These processes are time-consuming. Recently, a modified protocol25 was reported that could solve this problem to some extent. In this protocol, the cell-specific barcode labels cDNA fragments during the reverse transcription step. Therefore, the cDNA of each cell can be pooled together and then be purified, followed by library construction. This modification significantly increases the throughput of library construction. However, this modified method is less sensitive and detects fewer genes per cell. In addition, this method is a 3' counting method with reduced read coverage. Therefore, Smart-seq2 is still the most suitable method for pancreatic development.
The pancreatic preparation from other species may be different. For human pancreas, the islets can be isolated following previous protocols26,27. Then, the isolated islets can be dissociated into single cells28 and perform single-cell analyses using this method. This method can be widely applied to the research of mammalian pancreatic development, diseases and regeneration.
The authors have nothing to disclose.
We thank the National Center for Protein Sciences, Beijing (Peking University) and the Peking-Tsinghua Center for the Life Science Computing Platform. This work was supported by the Ministry of Science and Technology of China (2015CB942800), the National Natural Science Foundation of China (31521004, 31471358, and 31522036), and funding from Peking-Tsinghua Center for Life Sciences to C.-R.X.
Collagenase P | Roche | 11213873001 | |
Trypsin-EDTA (0.25 %), phenol red | Thermo Fisher Scientific | 25200114 | |
Fetal bovine serum (FBS) | Hyclone | SH30071.03 | |
Dumont #4 Forceps | Roboz | RS-4904 | |
Dumont #5 Forceps | Roboz | RS-5058 | |
30 G BD Needle 1/2" Length | BD | 305106 | |
Stereo Microscope | Zeiss | Stemi DV4 | |
Stereo Fluorescence microscope | Zeiss | Stereo Lumar V12 | |
Centrifuge | Eppendorf | 5810R | |
Centrifuge | Eppendorf | 5424R | |
Polystyrene Round-Bottom Tube with Cell-Strainer Cap | BD-Falcon | 352235 | |
96-Well PCR Microplate | Axygen | PCR-96-C | |
Silicone Sealing Mat | Axygen | AM-96-PCR-RD | |
Thin Well PCR Tube | Extragene | P-02X8-CF | |
Cell sorter | BD Biosciences | BD FACSAria | |
Capillary pipette | Sutter | B100-58-10 | |
RNaseZap | Ambion | AM9780 | |
ERCC RNA Spike-In Mix | Life Technologies | 4456740 | |
Distilled water | Gibco | 10977 | |
Triton X-100 | Sigma-Aldrich | T9284 | |
dNTP mix | New England Biolabs | N0447 | |
Recombinant RNase Inhibitor | Takara | 2313 | |
Superscript II reverse transcriptase | Invitrogen | 18064-014 | |
First-strand buffer (5x) | Invitrogen | 18064-014 | |
DTT | Invitrogen | 18064-014 | |
Betaine | Sigma-Aldrich | 107-43-7 | |
MgCl2 | Sigma-Aldrich | 7786-30-3 | |
Nuclease-free water | Invitrogen | AM9932 | |
KAPA HiFi HotStart ReadyMix (2x) | KAPA Biosystems | KK2601 | |
VAHTS DNA Clean Beads XP beads | Vazyme | N411-03 | |
Qubit dsDNA HS Assay Kit | Invitrogen | Q32854 | |
AceQ qPCR SYBR Green Master Mix | Vazyme | Q121-02 | |
TruePrep DNA Library Prep Kit V2 for Illumina | Vazyme | TD502 | Include 5x TTBL, 5x TTE, 5x TS, 5x TAB, TAE |
TruePrep Index Kit V3 for Illumina | Vazyme | TD203 | Include 16 N6XX and 24 N8XX |
High Sensitivity NGS Fragment Analysis Kit | Advanced Analytical Technologies | DNF-474 | |
1x HBSS without Ca2+ and Mg2+ | 138 mM NaCl; 5.34 mM KCl 4.17 mM NaHCO3; 0.34 mM Na2HPO4 0.44 mM KH2PO4 |
||
Isolation buffer | 1 × HBSS containing 10 mM HEPES, 1 mM MgCl2, 5 mM Glucose, pH 7.4 | ||
FACS buffer | 1 × HBSS containing 15 mM HEPES, 5.6 mM Glucose, 1% FBS, pH 7.4 | ||
NaCl | Sigma-Aldrich | S5886 | |
KCl | Sigma-Aldrich | P9541 | |
NaHCO3 | Sigma-Aldrich | S6297 | |
Na2HPO4 | Sigma-Aldrich | S5136 | |
KH2PO4 | Sigma-Aldrich | P5655 | |
D-(+)-Glucose | Sigma-Aldrich | G5767 | |
HEPES | Sigma-Aldrich | H4034 | |
MgCl2 | Sigma-Aldrich | M2393 | |
Oligo-dT30VN primer | 5'-AAGCAGTGGTATCAACGCAGAGTACT30VN-3' | ||
TSO | 5'-AAGCAGTGGTATCAACGCAGAGTACATrGrG+G-3' | ||
ISPCR primers | 5'-AAGCAGTGGTATCAACGCAGAGT-3' | ||
Gapdh Forward primer | 5'-ATGGTGAAGGTCGGTGTGAAC-3' | ||
Gapdh Reverse primer | 5'-GCCTTGACTGTGCCGTTGAAT-3' | ||
Ins2 Forward primer | 5'-TGGCTTCTTCTACACACCCA-3' | ||
Ins2 Reverse primer | 5'-TCTAGTTGCAGTAGTTCTCCA-3' |