Droplet Barcoding-Based Single Cell Transcriptomics of Adult Mammalian Tissues

All protocols described here are in accordance with and approved by the University of Calgary’s Animal Care Committee.

1. Dissociating Tissue (Day 1)

1. Euthanize mice with an overdose of sodium pentobarbital (i.p., 50 mg/kg) or as appropriate according to animal ethics protocol. Then remove unwanted hair from the back and legs of the mouse, and ethanol-sterilize the region of dissection.
2. Dissect the tissue or microenvironment of interest. For this protocol, we use skin and nerve tissues to demonstrate the generalizability of droplet barcoding-based single cell transcriptomics following adult tissue dissociation.
   1. For the sciatic nerve, use the detailed protocol found at Stratton et al.⁹. Briefly, cut the skin away from the hind region of the back/legs of mouse. Make an incision along the length of the thigh with a sterile scalpel blade. Use fine forceps and scissors to expose and remove the sciatic nerve.
   2. For the back skin, use the detailed protocol found at Biernaskie et al.¹⁰. Briefly, dissect dorsal back skin by making incisions from shoulder to shoulder, across the rump and down the back using fine forceps and scissors. Cut the skin into thin slices (0.5 cm thickness) using a sterile scalpel blade.
3. Wash tissue 2 times with ice-cold HBSS, and remove unwanted connective tissue, fat deposits or debris under a dissecting microscope.
   1. For the skin dermis only, float the slices in dispase (5 mg/mL, 5 U/mL) in HBSS for 30 – 40 min at 37 °C. Surgically separate the epidermis from the dermis. Discard the epidermis or further dissociate using trypsin if of interest.
4. Mince sample into 1-2 mm pieces using a sterile scalpel blades and put into freshly-thawed 2 mg/mL cold collagenase-IV enzyme (2 mg/mL, 125 CDU/mg, in F12 media).
   1. For the nerve, use ~500 µL per 2x sciatic nerves. For the skin, use ~8 mL per 1x mouse back skin.
      NOTE: The tissues should be fully submerged in the collagenase-IV solution. It is critical that any digestion enzymes are handled, stored, and prepared appropriately. If enzymes are left at room temperature for long periods of time, single-cell isolation will require excessive mechanical trituration and reduce cell viability. Collagenase-IV can also be made up in cell culture media where cell viability is most optimal. However, this could alter enzyme activity or transcriptional signature so should be optimized by the user.
5. Incubate the sample in the enzyme in a 37 °C bath for 30 min with gentle shaking every 10 min. A shaker placed at 37 °C is also an appropriate alternative.
6. Triturate with a P1000 pipettor 20-30 times at 30 min post-enzyme addition.
7. Repeat trituration every 30 min until solution appears cloudy and chunks of tissue are largely dissociated.
   NOTE: Ensure full release of cells (Figure 2b, 2c). To confirm full release, plate cells with Nuc Blue (2 drops per 1 mL) and after 20 min, check under the microscope to ensure all nuclei are associated with single cells rather than debris. It is critical to check the degree of cell release within a given experiment for each tissue type or condition. In fibrotic tissue (i.e., chronic injury) or uninjured adult tissue, the release of cells will vary dramatically from acute injury or embryonic tissue. This is especially important because certain cell types are less likely to release from the tissue than others, thus preferentially excluding those cells from downstream analysis.
   1. For the nerve, dissociate tissue for 0.5-1.5 hours total. For the skin, dissociate tissue for 2 hours total (in the last hour of incubation, add DNase (1 mg/mL) to the skin sample).
8. Filter twice with a 40 µm filter. Rinse the filter with ice-cold 1% BSA/HBSS.
9. Centrifuge at 260 x g for 8 min. Then remove supernatant.
   1. Resuspend the cell pellet in HBSS containing 1% BSA using a wide-bore tip, and place on ice. The resuspension volume is based on tissue volume (800 mg wet weight for skin = 800 µL volume; 10 mg wet weight for nerves = 100 µL volume).
   2. Optionally, start with a low resuspension volume and then adjust as necessary based on flow rate (events per second) on FACS sorter. The most efficient sort density (maximize number of cells collected while minimalizing time) for collections is 3,000-7,000 events/second.
10. If using viability dye, take out a subaliquot for an unstained control. Then add 1:15,000 viability dye (stock: 20,000 nM/µL) to sample (1.3 nM/µL final concentration) using a wide-bore tip to reduce shearing.
    NOTE: It is critical to check for the degree of cell death within a given experiment for each tissue type or condition. Some cell types within a sample are more likely to die than others, thus being preferentially excluded from downstream analysis.
    1. Incubate sample with viability dye for 5-10 min on ice in the dark. Then add 4 mL of ice-cold 1% BSA/HBSS to sample. Centrifuge at 260 x g for 8 min to remove excess viability dye. Treat the subaliquot with no viability dye unstained control in the same manner.

2. Isolating Viable and Healthy Cells (Day 1)

1. Ensure that the FACS facility follows appropriate fluorescence activated cell sorting (FACS) parameters.
   1. Prepare the FACS machine in advance to ensure that it is ready once the final centrifuge in Step 1 is complete, and ensure that the collection compartment is kept cold using ice blocks.
   2. Use the following parameters: Flow rate: 1.0 (corresponds roughly to 10 µL/min); Filter: 1.5 ND; Nozzle size: 100 µm; Forward scatter: 80 – 180 V (change as necessary in order to distinguish size of events); Side scatter: 150 – 220 V (change as necessary in order to distinguish granularity/shape of events); Laser: 100 – 400 V (change as necessary in order to distinguish viability dye positive vs negative events & check this against no viability dye control); Gates: Change as necessary to ensure all cells are collected. See Figure 2d-2g.
      NOTE: FACS parameters are highly dependent on the cell types and the sorter employed and therefore needs to be optimized by the user.
2. Prepare 15 mL narrow-bottom tubes with 8 mL of ice-cold 1% BSA/HBSS for sample collections. Static inside tube and surface tension can affect collection efficiency. Invert the tubes before collections to ensure interface between surface of liquid and the inside of tube is moist.
   NOTE: If working with very low cell numbers, adjust to small collection vessel as appropriate.
3. Once all cells are collected, centrifuge sample at 260 x g for 8 min.
   NOTE: Before centrifugation, add 1% BSA/HBSS to wash/push cells down from the side surface and invert/mix the tube immediately after FACS.
4. Resuspend cell pellet in 1% BSA/HBSS and keep on ice. The maximum volume per sample that is compatible with Step 3 processing is 33.8 µL, so ensure that final cell dilution/resuspension volume is appropriate to obtain ideal cell number in 33.8 µL. Other dilution media options for this step (and all previous dilutions in 1% BSA/HBSS) include DMEM, and up to 40 % serum, but avoid calcium, magnesium, or EDTA containing reagents.
   1. Leave cells on ice for a minimal amount of time. Ideally co-worker should prepare all equipment and reagents for the following step (Step 3) during final steps of Step 2. 
5. Cell preparation critical checks
   1. Confirm estimates of cell numbers obtained from FACS. Depending on tissue type and dissociation lengths, debris and cells can be very similar in size and shape. Thus, unless a fluorescent reporter is used, FACS cannot exclude all debris. It is recommended that a final cell count after FACS collection is performed to understand what percentage of events (according to FACS) are in fact cells for a given preparation (Figure 2g). Perform cell count using a hemocytometer or automated cell counter (repeat twice) and calculate the percentage of viable cells that is represented by the total events collected according to FACS machine.
   2. Validate cell preparation. Validate that no large particles (>100 µm) are present as they may clog equipment in downstream steps. Inadequate removal of debris may risk clogging the single-cell microfluidic chip. Plate remaining cells with Nuc Blue (as above) to ensure no large debris fragments are present. This will also allow for the confirmation that cells are singular (i.e., not sticking together) giving confidence that downstream single cell genetic analysis represents single cells rather than multiple cells.
   3. Decide on cell numbers to sequence: There is a large range of adult tissue-derived cell numbers per sample that can be loaded into the system with up to 8 samples that can be run at one time. Authors have loaded anywhere from 500 – 50,000 cells per sample and obtained good quality scRNA-Seq datasets. More discussion regarding the most appropriate cell numbers to load can be found in the Discussion section. The final output of sequenced cell numbers depends heavily on the quality of single-cells isolated. Loading 10,000 adult tissue-derived cells can return anywhere from 1,000 to 4,000 sequenced cells (10 – 40% return). If interested in sequencing high cell numbers (~10,000 cells, the maximum number recommended for this system), then loading 25,000 – 100,000 cells will be required.

3. GEM (Gel Bead in Emulsion) Generation and Barcoding (Day 1)

NOTE: Steps 3-6 of this protocol are designed to be used in conjunction with the most common microdroplet-based single-cell platform, manufactured by 10X Genomics. Detailed guidelines for Steps 3 and 4 are outlined in the manufacturer’s protocol (Refer to the Chromium Single Cell 3' protocol)^11,12 and must be followed in conjunction with this protocol. For best results, Step 3 must be completed immediately after dissociation (Step 1) and cell isolation (Step 2) steps on day 1 of this protocol.

1. Prepare chip according to the manufacturer’s protocol ^11,12. This microdroplet-based single-cell platform uses technology that samples ~750,000 barcodes to separately index each cell’s transcriptome. This is achieved by partitioning cells into Gel Bead in EMulsions (GEMs) where generated cDNA share a common barcode. During GEM generation, cells are delivered so that the majority (90 – 99%) of generated GEMs contain no cells, while the remainder, for the most part, contain a single cell.
   1. Place the chip in the chip holder.
   2. Prepare cell master mix on ice.
   3. Add 50% glycerol to unused wells and add 90 µL of the cell master mix to well 1, 90 µL of gel beads to well 2, and 270 µL of partitioning oil to well 3.
   4. Cover the chip with the gasket.
2. Load the chip and run in a single-cell controller.
   1. Eject the tray, place the chip in tray, retract the tray, and press Play. A single cell 3’ gel bead in a GEM includes primers containing a partial Illumina R1 sequence (read 1 sequencing primer), a 16 nucleotides (nt) 10x Barcode, a 10 nt Unique Molecular Identifier (UMI), and a poly-dT primer sequence. During the run, gel beads in the controller are released and mixed with cell lysate and master mix.
   2. Collect 100 µL of the sample and place in a PCR tube.
   3. Place PCR tubes in pre-set PCR machine and run the PCR according to the kit. Following incubation, GEMs will include full-length, barcoded cDNA from poly-adenylated mRNA.
3. Following the run, place at -20 °C overnight for up to 1 week before preceding to next step.

4. Clean-Up, Amplification, Library Construction and Library Quantification (Day 2 Onwards)

NOTE: Detailed guidelines for Steps 4 are outlined in the manufacturer’s protocol ^11,12, and must be followed in conjunction with this protocol.

1. Use silane magnetic beads to remove leftover biochemical reagents/primers from GEM reaction mixture.
2. Amplify full-length, barcoded cDNA to generate sufficient mass for library construction.
3. Assess DNA yield. Prior to library construction, assess DNA yield of the sample. This will determine how many cycles to use in downstream PCR step (Sample Index PCR during library construction). Depending on the RNA content of a given sample, which can vary depending on activation states (e.g., control vs injured, etc.), cell type, and cell yield, the recommended cycle number may change.
   1. For sequencing ~3,000 tissue-derived cells (irrelevant to activation states), the authors have found that 14 cycles (samples: ~10 – 100 ng DNA) is standard.
   2. Use a Bioanalyzer for DNA analysis. Refer to the User Guide¹³.
4. Fragment sample and select the size of DNA. Prior to library construction, use enzymatic fragmentation and size selection protocols to obtain appropriate cDNA amplicon size.
5. Prepare sample for library construction. While R1 (read 1 primer sequence) is added to the molecules during GEM incubation; P5, P7 (a sample index), and R2 (read 2 primer sequence) are added during library construction.
6. Assess DNA yield. Most sequencing facilities require submission of final libraries that include DNA yield and quality information. Thus, run the bioanalyzer following the completion of the entire protocol and before transporting to the sequencing facility.
7. Store samples at -80 °C for up to 2 months.
8. Before sequencing, quantify samples using a DNA quantification kit. This can be done at the sequencing facility.

5. Library Sequencing (Day 3 Onwards)

NOTE: The single-cell transcriptome barcoding platform used in this protocol generates Illumina-compatible paired-end libraries beginning and ending with P5 and P7 sequences. Although minimum depth needed to resolve cell-type identity can be as few as 10,000–50,000 reads/cell^15,16, ~100,000 reads/cell is recommended as an optimal cost-coverage trade-off for adult in vivo cells (keeping in mind some cell types or minimally activated cell states will reach saturation at 30,000-50,000 reads/cell).

1. Transport cDNA libraries on dry ice to a sequencing facility equipped with an appropriate Illumina sequencer.
2. Provide the following information to the sequencing facility:
   1. Provide sample details: Sample index IDs corresponding to each library; species; genomic database for primary assembly (i.e., GRCm38 for mouse); electropherogram showing fragment sizes from the bioanalyzer (between 200 and 9,000 bp); cDNA concentration (ng/µL) and total library concentration (total yields range from 200 – 1400 ng); volume (µL) of sample.
   2. Provide sequencing requests: Quantify samples using a DNA quantification kit; adapter/index type (TruSeq DNA); plate type (Eppendorf twin.tec, Full Skirt – recommended for DNA); sequencing technology/library type (10x, full sequencing instructions and cycle recommendations)¹⁷.
3. Run shallow sequencing (optional): Studies analyzing multiple biological samples will benefit from pooling samples (aggregation) to generate a single gene-barcode matrix containing data from all samples. To minimize batch effects between samples when pooling, the read depth between different libraries should be standardized. In order to do this, an accurate approximation of single cell numbers is necessary. The MiSeq sequencer will allow shallow sequencing and is a cost-effective, practical way to obtain accurate cell estimations.
   NOTE: One run using a MiSeq SR50 sequencer provides sufficient coverage to accurately estimate approximately 20,000 cells. This run will approximate the number of UMI recovered for each unique barcode. In Figure 3a, the header of an example (Sample 1.6) output (.csv) is shown, listing barcodes and its corresponding UMI counts as determined by confidently mapped reads.
   1. Consult with a bioinformatician to become familiar with the R programming language. Refer to DataCamp tutorials for more information¹⁸.
   2. Assess raw data obtained from the sequencer using the provided R script as a template¹⁹. Raw data refers to number of UMIs mapped to each unique cell barcode. The script reads a .csv file where the first column is a list of barcodes and the second column are its corresponding UMI counts. This script will provide a plot (Figure 3b) as well as the estimated number of barcoded cells in each sample. Adjust script to ensure that the inputted number of UMI counts for a given sample is at the one-third point of the first steep drop. In Figure 3b, this elbow falls around 225 UMIs corresponding to 3,480 barcoded cells.
   3. Comparable to full-depth sequencing using HiSeq (where 3,516 cells were successfully sequenced, Figure 3c), shallow sequencing estimates predicted 3,480 cells.
4. Either use cell recovery approximations (from step 5.3) or use the recovery chart found in the manufacturer’s protocol²⁰ to plan lane distribution for deeper sequencing. Each sample should receive comparable coverage, so if shallow sequencing reveals that there are differing numbers of cells in each sample (which is often the case) then lane distribution should be calculated accordingly. One HiSeq flow cell (which comprises 8 lanes) can sequence up to 2.4 billion custom paired end reads. Example flow cell set-up is presented in Figure 3d.

6. Processing Read Files

NOTE: Sequencing a single cell 3’ Library using this protocol generates raw data in binary base call (BCL) format. The Cell Ranger package is used to generate text-based FASTQ files from BCL files, perform genomic and transcriptomic alignments, gene counts, demultiplexing, and aggregation of samples. In this section, the key steps that enable users to download raw BCL data from a sequencing facility and generate filtered gene-barcode matrices ready for downstream bioinformatics is presented.

1. Use a centralized server for running the program. BCL files, FASTQ files and most of the downstream bioinformatics processing demands significant processing power.
2. Download all raw read files onto server (or FASTQ files if they are available).
   1. Consult with server administrator to set up an account on a centralized server or cluster, and to get familiar with Unix²¹.
   2. Use a fetch command appropriate for the server’s operating system to download all files from the sequencing facility’s server.
      1. Most sequencing facilities provide a command to download files from a secure path that can be run from the command line (see example below).
      2. Replace the "<username>" and "<password>" placeholders in command line with credentials provided.
         wget -O – "https://your_sequencing_facilitys_server.com/path_to_raw_read_files/ –no-cookies –no-check-certificate –post-data 'j_username=username&j_password=password' | wget –no-cookies –no-check-certificate –post-data 'j_username=username&j_password=password' -ci –
   3. If only an absolute path to files is provided (i.e. https://your_sequencing_facilitys_server.com/path_to_raw_read_files/), insert this path into a fetch command.
3. Unzip files: If downloaded files end with a ".gz" extension, it has been compressed using the "gzip" command. To unzip, run unzip command in the command line (see example below).
   gunzip raw_read_files.gz
4. Download the latest version of Cell Ranger onto the server as a self-contained .tar²².
   1. Critical: Prior to download, ensure the Linux system meets minimum requirements²³. Ensure a minimum of 8-core Intel processor with 64 GB RAM and 1 TB of free disk space.
      NOTE: Cell Ranger provides pre-built human and rodent reference transcriptomes. These can be modified using cellranger mkref command to detect genes like GFP²⁴.
5. Generate FASTQ files from sequencer's base call files (BCLs) using the cellranger mkfastq command.
   NOTE: The program will align raw reads (from FASTQ files) to a reference genome and generate gene-cell matrices for downstream analysis. It uses STAR aligner that performs splicing-aware alignment of reads to a reference genome. Only confidently mapped reads (i.e., reads compatible with a single gene annotation) are used for UMI counting.
   1. For example, use the cellranger mkfastq command:
      cellranger mkfastq –id=sample_name
                –run=/path/to/sample
                –csv=csv_file_containing_lane_sample_index.csv
6. Run cellranger count on FASTQ files generated using mkfastq to generate single-cell gene counts.
   1. For example, use the cellranger count command:
      cellranger count –id=sample_name
                –transcriptome=refdata-cellranger-mm10-1.2.0
                –fastqs=/absolute/path/to/fastq/files
                –sample=same_sample_name_supplied_to_cellranger_mkfastq
                –localcores=30
7. Multi-library aggregation (optional): To combine samples, pool cellranger count outputs using cellranger aggr. This results in a single gene-barcode matrix containing data pooled from multiple libraries. Example cellranger aggr command:
   cellranger aggr –id=sample_name
             –csv=csv_with_libraryID_&_path_to_molecule_h5.csv
             –normalize=mapped
   NOTE: Libraries can be aggregated using three normalization modes (mapped, raw, none). Mapped is recommended as it subsamples higher-depth libraries until all libraries have equal sequencing depth²⁵.
8. For immediate visualization/analysis of data, import the .cloupe output file (generated using cellranger count or cellranger aggr) into 10x Loupe Cell Browser²⁶.

7. Advanced Analysis of scRNA-Seq Datasets

NOTE: A complete scRNA-Seq tools database can be found at scRNA-tools^3,27. Below is a framework for unsupervised cell clustering using Seurat² and pseudotemporal ordering using Monocle⁶. Although much of this work can be done on a local computer, the following steps assume that computation will be completed using an institutional server.

1. Download the most recent version of Miniconda onto server account using the Linux platform²⁸.
2. Install the latest version of R using conda²⁹.
3. Plot data using the provided Seurat R script as a template³⁰.
   NOTE: Seurat is an R-based toolkit that enables quality control checks, clustering, differential gene expression analysis, marker gene identification, dimensionality reduction, and visualization of scRNA-Seq data. A comprehensive description of Seurat coding and tutorials can be found on the Satija Lab website³¹.
4. Plot data using the provided Monocle R script as a template³².
   NOTE: Monocle is another R-based toolkit that enables visualization of expression changes over pseudotime and identifies genes underlying cell fate decisions. A comprehensive description of Monocle coding and tutorials can be found on the Monocle website³³.
5. R-packages such as kBET can be employed to test and correct batch effects as a result of pooling datasets³⁴.

8. NCBI’s GEO and SRA Submissions

NOTE: Since easy access to raw sequencing files ensure reproducibility and reanalysis, accessioned submissions to online publicly available repositories are recommended or required prior to manuscript submission. National Center for Biotechnology Information’s (NCBI) Gene Expression Omnibus (GEO) and Sequence Read Archive (SRA) are publicly accessible data repositories for high-throughput sequencing data^35,36.

1. Register for NCBI’s GEO Submitter account³⁷.
2. Complete GEO submission which includes three components compiled into a directory/folder (titled as the GEO submitter’s username): 1) Metadata record (one spreadsheet per project submission); 2) Raw Data Files; 3) Processed Data Files.
   1. Download and complete the Metadata spreadsheet³⁸. The following public GEO submission can be used as a guide (GSE100320)³⁹. Place the spreadsheet in the directory.
   2. Place Raw Data Files generated from cellranger count script for all libraries into the directory.
   3. Place Processed Data Files (filtered barcodes.tsv, genes.tsv, and matrix.mtx files) generated from cellranger count script for all libraries into the directory.
3. Use GEO submitter’s FTP server credentials to transfer directory containing all three components. For Linux/Unix users: ncftp, lftp, ftp, sftp, and ncftpput can be used.
4. Notify GEO for all transfers³⁸.