Multiplexed single-nucleus RNA sequencing using lipid-oligo barcodes

Abstract

This protocol describes a robust pipeline for simultaneously analyzing multiple samples by single nucleus (sn)RNA-seq. cDNA obtained from each single sample are labelled with the same lipid-coupled oligonucleotide barcode (10X Genomics). Nuclei from as many as twelve individual samples can be pooled together and simultaneously processed for cDNA library construction and subsequent DNA sequencing. While previous protocols using lipid-coupled oligonucleotide barcodes were optimized for analysis of samples consisting of viable cells, this protocol is optimized for analyses of quick-frozen cell samples. The protocol ensures efficient recovery of nuclei both by incorporating high sucrose buffered solutions and by including a tracking dye (trypan blue) during nuclei isolation. The protocol also describes a procedure for removing single nuclei ‘artifacts’ by removing cell debris prior to single nuclear fractionation. This protocol informs the use of computational tools for filtering poorly labelled nuclei and assigning sample identity using barcode unique molecular identifier (UMI) read counts percentages. The computational pipeline is applicable to either cultured or primary, fresh or frozen cells, regardless of their cell types and species. Overall, this protocol reduces batch effects and experimental costs while enhancing sample comparison.

Basic Protocol 1:

Labeling cells with lipid oligo barcodes and generating multiplexed single nucleus RNA-seq (snRNAseq) libraries

Basic Protocol 2:

Bioinformatic deconvolution of the multiplexed snRNAseq libraries

INTRODUCTION:

The advent of transcription profiling at single-cell resolution has become a powerful approach to reveal diverse cell populations and states within biological systems (Hwang et al., 2018). Advances in droplet-microfluidic technology, which encapsulates uniquely bead-barcoded cells with reagents for cDNA synthesis, have increased throughput from less than 100 cells to thousands of cells per run (Klein et al., 2015, Zheng et al., 2017). Droplet technology can profile tens of thousands of cells in parallel and is widely used for single cell transcriptome studies with both cultured cells and tissues. However, there are remaining challenges of high cost in library preparation and sequencing, and batch effects when multiple samples are processed in different experimental conditions (Lafzi et al., 2018).

Until recently, several approaches for single cell multiplexing have been developed, which labels cells of the same sample using a specific barcode before pooling and encapsulation. Barcodes currently used to identify samples include natural genetic variations (Kang et al., 2018), DNA-tagged antibodies targeting cell surface proteins (Stoeckius et al., 2018) or the nuclear pore complex (Gaublomme et al., 2019), and transient or heritable genetic barcodes transfected by delivery systems such as Lipofectamine 3000 or lentivirus (Guo et al., 2019; Shin et al., 2019). Each method described has its limitations. For example, the natural genetic variations technique relies on prior knowledge of genetic background and robust platforms to assign cells, which is less scalable nor accurate (Heaton et al., 2019). Transfecting genetic barcodes may perturb gene expression. Barcodes tagged with antibodies are not feasible with cell types that do not express ubiquitous surface proteins. In principle, such limitations can be overcome with the use of lipid tagged barcodes (McGinnis et al., 2019) and commercially available lipid cell multiplexing oligos from 10X Genomics. Lipid tagged oligos are rapidly incorporated into the membrane. This type of barcode has the potential to label cells or nuclei regardless of cell types or species and requires minimal effort in procedure.

Both whole cells and nuclei are widely used for single cell genomics, but they require differences in sample preparation and data processing. One major drawback of single cell RNA sequencing (scRNA-seq) is that it requires fresh tissue or cells, which often demands a strong collaboration between tissue provider and processor. Methods such as droplet-based scRNA-seq is only suitable for cells with a size of smaller than 50 μm (Del-Aguila et al., 2019). In general, scRNA-seq generates a more biased expression profile due to the loss of vulnerable cell types during tissue dissociation (Elmentaite et al., 2022). An advantage of using single nucleus RNA sequencing (snRNA-seq) is the ability to process frozen tissues with no limit for cell size and gather substantial information using small amounts of mRNA (Bakken et al., 2018; Denisenko et al., 2020; Nadelmann et al., 2021; Slyper et al., 2020).

Current applications of nuclei multiplexing have been hampered by two major issues. The first issue is feasibility. Methods for multiplexing are designed for whole cells demonstrated within very limited cell types (Guo et al., 2019; Shin et al., 2019; Stoeckius et al., 2018). Compared to using whole cells, barcoding nuclei is more difficult because of fewer labeling options and the nuclei’s fragile property. Practical considerations include isolating nuclei of high quality, reducing nuclei loss, monitoring nuclei pellet, and testing whether the barcodes are reliable. Although a lipid barcode provides a promising way to tag nuclei, this method has yet to be optimized for large-scale experiments. Accurate assignment of barcodes to their expected cell types is critical for data analysis. Several factors can affect the accuracy: the stability of the barcode bound to the nucleus, the removal of unconjugated barcodes to reduce cross-labeling, and a robust deconvolution strategy to capture the most enriched barcode from other off-target ones. To date, several sample classification and barcodes assignment strategies have been reported (Adamson et al., 2016; Gaublomme et al., 2019; McGinnis et al., 2019; Stoeckius et al., 2018). These strategies have different criteria for deconvolution, and some of them require high levels of computational expertise.

Here, we compare RNA expression in multiple previously frozen samples derived from different cultured cells and describe a pipeline for snRNAseq that allows for analyzing multiplexed nuclei labelled with lipid oligo barcodes (10X Genomics). Basic Protocol 1 describes labeling nuclei and generating mulitplexed single nuclei RNAseq library. This protocol was optimized to increase the yield of labeled nuclei by a) employing buffers containing sucrose, salts and bovine serum albumin and b) using trypan blue to track nuclei, which are designed to reduce the loss of nuclei from degradation and adherence to plasticwares. Partially purified nuclei are separated from cell debris by fractionation on a florescence-activated cell sorter (FACS). Basic Protocol 2 describes a computational deconvolution strategy that can assign nuclei rapidly and accurately to cell populations within multiplexed samples. In summary, this protocol provides a highly reproducible pipeline for lipid oligo multiplexed snRNA-seq from sample processing to data analysis. The protocol is recommended for the analysis of multiple cultured cell samples by snRNAseq.

BASIC PROTOCOL 1

Labeling cells with lipid oligo barcodes and generating multiplexed single nucleus RNA-seq libraries

The schematic workflow of the procedure describes steps to multiplex multiple samples, obtain barcoded nuclei, subject these to snRNA-seq and deconvolute the resulting fastq files (Fig.1). Beginning with as many as 12 different cell cultures, the protocol comprises four sections: 1) harvest cultured cells into pellets; 2) isolate nuclei from cell pellets; 3) barcode the freshly isolated nuclei with cell multiplexing oligos (CMOs) and pool equal numbers of tagged nuclei; 4) stain the pooled nuclei with fluorescence and FACS sort to remove unwanted debris or clumps, and load samples onto 10X Genomics chips followed by a standard sequencing workflow.

Fig. 1. Schematic workflow of multiplexed snRNA-seq procedure.

Cells from different samples are collected and lysed to isolate nuclei. Nuclei from individual samples are labeled with distinctive barcodes and then pooled together. The mixed nuclei are sorted on FACS machine to remove subcellular debris and then loaded onto 10X Genomics lanes to encapsulate with gel beads. After library preparation and subsequent sequencing, cell identity is assigned using the deconvolution strategy introduced in this protocol. UMAP: Uniform manifold approximation and projection.

Materials:

Primary or cultured cells or tissues of interest. (Samples used in this article: twelve iPSC-derived samples: endothelial cells, cardiomyocytes, and neural crest cells)

BD Matrigel™ hESC-qualified Matrix (BD Biosciences, cat. no. 354277)

RPMI medium (Invitrogen, cat. no. 11875–119)

B27 supplement (Invitrogen, cat. no. 17504–044)

Fetal bovine serum (FBS, Thermo Fisher Scientific, cat. no. A3160402)

ROCK inhibitor Y-27632(R&D Systems, cat. no. 125410)

Dimethyl Sulfoxide (DMSO, Sigma-Aldrich, cat. no. D2650–5X10ML)

Phosphate-buffered saline (1×) (PBS, Invitrogen, cat. no. 10010–049)

TrypLE™ select enzyme (10×), no phenol red (Thermo Fisher Scientific, cat. no. A1217701)

Trypan blue stain, 0.4% (Invitrogen, cat. no. T8154–100M)

Homogenization buffer (see recipe in Reagents and Solutions)

Multiplex wash buffer (see recipe in Reagents and Solutions)

3’ CellPlex kit set A (10X Genomics, cat. no. 1000261)

NucBlue™ live ReadyProbes™ reagent (Hoechst 33342; Thermo Fisher Scientific, cat. no. R37605)

Storage buffer (see recipe in Reagents and Solutions)

Chromium next GEM single cell 3’ kit v3.1(10X Genomics, cat. no. 1000268)

Costar 6-well plate (Fisher Scientific, cat. no. 07–200-83)

50 mL 0.22 μM disposable vacuum filtration system (Millipore, cat. no. SE1M179M6)

5 mL Falcon disposable aspirating pipet (VWF, cat. no. 414004–266)

5 mL serological pipette (Westnet, cat. no. 229005B)

10 mL sterile TC pipette (VWF, cat. no. 89130–898)

50 mL Conical tubes (Fisher Scientific, cat. no. 352098)

Cell strainer 70 μM (BD Biosciences, cat. no. 352350)

Centrifuge (Beckman, cat. no. 8G089)

Centrifuge (Eppendorf, cat. no. 5417C)

Countess II automatic cell counter (Thermo Fisher Scientific, cat. no. AMQAX1000)

Cell counting chamber slides (Invitrogen, cat. no. C10228)

Fisherband 1.5mL microcentrifuge tube (Fisher, cat. no. 02–681-331)

Pluristrainer mini 40 μM (pluriSelect, cat. no. 43–10040-50)

Swinging bucket adaptors from TissueLyser II (Qiagen, cat. no. 69989)

5 mL polystyrene round-bottom tube with cell-strainer cap (Corning, cat. no. 352235)

Florescence-activated cell sorting (FACS) Aria II sorter (BD Biosciences, cat. no. 650033)

Protocol steps:

Section 1: Harvest cultured cells

In this section, we describe cell harvesting procedures that are suitable for adherent cells. This protocol is suitable for suspension cells after pelleting with centrifugation. Adherent cells should be detached from the culture dish using routine protocols (Phelan & May, 2015). As an example, we describe harvesting adherent induced pluripotent stem cell derived cardiomyocytes (iPSC-CM).

• 1Culture iPSC-CMs on 6-well plates with pre-coated Matrigel until day 30 of differentiation.

  We recommend harvesting a culture with a minimum of 5×10⁵cells as isolation of small numbers of nuclei can be difficult.

• 2
  On the day of harvest, prepare 50mL fresh dissociation medium following the table below. Sterilize by filtering it through a 50mL 0.22μM filtration system. Different dissociation media are used for other types of cells.

  Components	Volume/sample	Final Concentration
  RPMI medium	44.1 mL	-
  B27 supplement	0.9 mL	-
  FBS	5 mL	10 %(v/v)
  10 mM ROCK inhibitor	50 μL	10 μM
  total	50 mL

  To make 10 mM ROCK inhibitor, dissolve it with an appropriate volume of DMSO based on the molecular weight. Divide it into 200 μL aliquots and store at −20°C freezer for up to one year.

• 3Aspirate spent medium and add 2 mL PBS into each well to wash for one time.
• 4Remove PBS and add 1mL TrypLE (10×) into each well for enzymic dissociation.
• 5Incubate at 37°C. During incubation, vigorously swirl the plate every 3 minutes to help lift up the cells. Continue this process until all cells are detached from the surface of the plate. This step usually takes 10 minutes.

  Do not pipet cells while in TrypLE alone. The time for incubation varies for different cell types or digestion enzymes.

• 6Meanwhile, prepare a 50 mL Falcon tube with a 70-μM cell strainer sitting on top of it. Rinse the strainer with 1mL dissociation medium from step 2 using a P1000 pipettor.
• 7When the cells are no longer adhering to the plate, add 2 mL dissociation medium to each well to neutralize TrypLE. Gently disperse the cells with a P1000 pipettor and transfer the cell suspension into the 50 mL Falcon tube through a strainer from step 6.
• 8Wash each well with 1 mL dissociation medium and transfer the solution into the 50mL Falcon tube through a strainer.
• 9Centrifuge at 1000 rpm for 5 minutes at room temperature.
• 10Aspirate the supernatant carefully.
• 11Resuspend the cell pellet with 1mL PBS using a P1000 pipettor. Pipet up and down for several times to break the cell pellet into single cells.

  The recommended range of cell concentration for Cell Counter is 1×10⁴-1×10⁷ cells/mL. Therefore, adjust the resuspension volume if needed.

• 12Take 6 μL of the cell suspension and mix thoroughly with 6 μL trypan blue. Load 10 μL mixture into one chamber of the counting slide.
• 13Insert the slide into the Countess II automatic cell counter device to count viable cells. To improve the accuracy of counting, count at least two different regions for each sample and then average the values.
• 14Aliquot 1×10⁶ cells from the suspension into a 1.5mL microcentrifuge tube based on the concentration measured.

  The number of cells collected for each sample can range from 5×10⁵ – 2×10⁶. We found 1–2 million cells work the best for the following procedure.

• 15Centrifuge at 1000 rpm for 5min at room temperature.
• 16Aspirate the supernatant carefully.
• 17Use fresh pellets directly or freeze them at −80°C for long-term storage.

Section 2: Isolate nuclei

Nuclei are isolated from cell pellets from Section 1 by a combination of homogenization and cell membrane lysis (Nadelmann et al., 2021).

• 18Use fresh or frozen cell pellets collected in Section 1. The frozen cell pellets need to be completely thawed on wet ice (approximately 15 minutes).
• 19While samples are sitting on ice, prepare fresh homogenization buffer (HB; see Reagents and Solutions). Store the buffer on ice.
• 20Prepare fresh multiplex wash buffer (MWB; see Reagents and Solutions part). Store the buffer on ice.
• 21Prepare a new set of 1.5 mL microtubes. First, add 1 ml MWB to each tube and invert several times. (This procedure coats the tube surface and reduces the sticking of nuclei to the tube). Remove the MWB and put a Pluristrainer on top of the tube.
• 22Resuspend fresh pellet or thawed pellet of each sample in 500 μL HB using a P1000 pipettor. Pipet up and down to break up the pellet and lyse the cells.
• 23Incubate on ice for 5 minutes.
• 24Transfer the sample into new microtubes passing through the strainer prepared in step 21.
• 25Add 10 μL 0.4% trypan blue stain to each tube to dye the nuclei, which helps track the nuclei in subsequent steps. Mix the trypan blue with nuclei suspension by tapping the tube a few times.
• 26Centrifuge at 800 g for 5 minutes, 4°C, using swinging bucket rotors.

  Swinging bucket centrifugation, rather than fixed angle centrifugation, helps pellet the nuclei to the bottom of the tube rather than the sides of the tube.

• 27Using a P1000 pipettor, carefully remove the supernatant without disturbing the nuclei pellet. The pellet, containing the nuclei, should be blue from the trypan blue dye (Fig. 2A).

Fig. 2. Detecting nuclei during the multiplexing process.

(A) Pelleted nuclei, stained with trypan blue, are easily detected. (B) FACS separates individual nuclei from doublets and debris. In the top and middle panels, forward scatter (FSC-A: forward scatter area; FSC-W:forward scatter width) and sideward scatter (SSC) measurements identify single nuclei by size. In the bottom panel, Hoechst-A stained nuclei emit fluorescence (pseudo colored blue) allowing separation from unstained debris (grey). (C) Nuclei morphology is visualized by the automated cell counter. High quality nuclei (circled in red) appear uniform with average size of 13.86um.

Section 3: Tag nuclei and pool

In this section, freshly isolated nuclei are tagged with Cell Multiplexing Oligonucleotides (CMOs) and combined into a single pool consisting of equal numbers of nuclei from each sample.

• 28Wash the nuclei by adding 1mL chilled MWB. Pipet up and down for a few times to break up the pellet.

  Do not over pipet as this will lyse nuclei; complete resuspension is not required.

• 29Centrifuge in a swinging bucket centrifuge at 800 g for 5 minutes, 4°C.
• 30Use a P1000 pipettor to remove the supernatant, without disturbing the nuclear pellet.
• 31Add 100 μL CMO to each sample and mix about 10 times to resuspend.

  10X Genomics provides 12 different CMOs in the 3’ CellPlex kit set A (CMO301 to CMO312), which can label 12 individual samples. Use a different CMO for each of the samples to be pooled together. The oligos can be stored at −20°C and thawed at room temperature before use.

• 32Incubate for at least 10 minutes at room temperature.

  Start counting incubation time after the last sample if processing multiple samples.

• 33Add 900 μL MWB to each tube.
• 34Centrifuge at 800 g for 5 minutes, 4°C, using a swinging bucket centrifuge.
• 35Remove the supernatant, without disturbing the nuclear pellet.
• 36Resuspend with 1mL MWB per tube.

  If the blue color fades away, add another 10 μL 0.4% trypan blue and mix by tapping the tube.

• 37Centrifuge at 800 g for 5 minutes, 4°C, using swinging buckets.
• 38Remove the supernatant as much as possible, without disturbing the nuclei pellet.
• 39Resuspend the pellet with an appropriate volume of MWB to reach an ideal final concentration of 1×10⁴–1×10⁷ nuclei/mL.

  This number is determined by Automatic Cell Counter user’s guideline to get accurate counting results. Usually starting with 1 million cells, the final volume for resuspension is 50 μL.

• 40Determine the concentration of the nuclei using Countess device, following the steps described above (step 12–13) or other method.

  We typically observed ~25% recovery. Nuclei recovery rate was similar for multiple cell types and regardless of the initial number cells.

• 41Pool 100,000 nuclei per sample into a 1.5 mL microtube.

  If the smallest sample has more than 100,000 nuclei, we recommend pooling with a bigger number to maximize recovery.

• 42Transfer all pooled nuclei through the strainer of a 5 mL flowcytometry tube.

Section 4: FACS and snRNA sequencing

Pooled nuclei are stained with blue fluorescent dye and FACS sorted to remove debris and/or clumps. After being sorted, samples are counted and loaded onto 10X Genomics chips as described for the Chromium snRNA-seq workflow.

• 43Add two drops of Hoechst-A dye (NucBlue™ live ReadyProbes™ reagent). Mix by quick vortex.
• 44Stain the sample by keeping it on ice for 15 minutes.
• 45Meanwhile, prepare a new 1.5 mL microtube: add 150 μL storage buffer (SB; Reagents and Solutions) for sample collection during FACS sorting.
• 46Proceed to flowcytometry: Sort the pooled nuclei following the manufacturer’s instruction of the FACS machine. Collect Hoechst-A positive nuclei and discard debris (Fig. 2B). Record the number of sorted nuclei provided by the FACS measurement.
• 47Centrifuge the sorted sample at 800 g for 5 minutes, 4°C.
• 48Carefully remove the supernatant without disturbing the nuclei pellet.
• 49
  Resuspend in an appropriate volume of SB according to the number of sorted nuclei recorded from step 46. To maximize nuclei recovery, the ideal final suspension concentration should be 1.3–1.6 ×10⁶ nuclei/mL as recommended in the 10X Genomics 3’ library Manual. The table below describes the reference volume for resuspension, scale to other volumes accordingly.

  Nuclei numbers from FACS	Volume
  100,000	25 μL
  150,000	35 μL
  200,000	50 μL
  250,000	60 μL

• 50Determine the concentration of the nuclei either using a Countess device, following the steps described from above (step 12–13), or other suitable method for counting nuclei.
• 51Load the sample and reaction Master Mix onto a 10X Genomics chip to encapsulate with gel beads. Subsequently, amplify cDNA, construct library and sequencing. All procedures in this step follow the 10X Genomics Chromium Next GEM Single Cell 3’ reagent Kits User Guide.

BASIC PROTOCOL 2

Bioinformatic deconvolution of the multiplexed snRNAseq libraries

This protocol describes how to deconvolute sequence data to distinguish the sample identities. The binary base call (BCL) file generated by the Illumina sequencer is demultiplexed and fastq reads are aligned to the reference genome using Cell Ranger software suite (version 6.1 or higher). Barcodes are assigned to each nucleus and a Seurat object is generated for downstream analysis using R and Seurat package. (Satija et al., 2015)

Materials:

CellRanger software suite version 6.1 or higher (PMID: 28091601, https://support.10xgenomics.com/single-cell-gene-expression/software/overview/welcome)

Reference genome downloaded from: https://support.10xgenomics.com/single-cell-gene-expression/software/downloads/latest

R studio (version 4.0.1), installed with Seurat, tidyverse, and reshape2 package

Protocol steps:

1. Deconvolute the BCL sequence file generated by the Illumina sequencer, which generates per sample FASTQ files (text files of nucleotide sequences that passed quality control). Deconvolution is performed by the Cell Ranger software suite following the manufacture’s protocol and formatting for the sample sheets. Note: If no multiplexing was done, or if starting with appropriately deconvoluted FASTQ files, proceed to step 2.

   Cellranger mkfastq –id=SequencingRunID –run=/Path_To_Bcl_Files –samplesheet=SampleSheet.csv

2. Align nucleotide sequences from the FASTQ files to an appropriate species-specific reference genome using the Cell Ranger software suite (https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/using/count) following the manufacturer’s protocol. The flag “–include-introns” is added for snRNAseq samples to count reads that mapped to intronic regions of the genome.

   Cellranger count –id=Sample_ID –sample=Sample_name_from_samplesheet –project=Project_name_from_samplesheet --transcriptome=/refdata-cellranger-GRCh38–1.2.0 –fastqs=/Path_to_fastq_folder_generated_from_step_1 --include-introns

3. Steps 3 to 9 can be run as a batch script or interactively on R. If running interactively, open Rstudio and import the following packages by running:

   library(Seurat)

   library(tidyverse)

   library(reshape2)

4. Read in the count files generated from step 2 using Read10X command in R. This will return a list with two elements. The first element is gene expression matrix, and the second element is CMO read count matrix. Save them for the subsequent analysis.

   Example <- Read10X(data.dir = Feature_BC_Matrix_from_step_2)

   cmo <- as.data.frame(example[2])

   gene_expression <- example[1]

5. Read in the assignment_confidence_table.csv file generated from step 2. Create a Seurat object from the gene expression data from step 4. Filter for the accepted nuclei by matching the barcodes in the gene expression matrix with the list of barcodes in the assignment_confidence_table.csv. Save the R object for downstream analysis. Note: Assignment generated by assignment_confidence_table.csv can be used as alternative method for assigning CMO to nuclei.

   barcode <- read.csv(assignment_confidence_table.csv)

   pbmc <- CreateSeuratObject(counts = gene_expression$`Gene Expression`, min.cells = 3, min.features = 200)

   pbmc <- subset(pbmc, cells = barcode$Barcodes)

   save(pbmc, file= YOUR_OUTPUT_NAME)

6. Match the format of barcodes of the CMO read count table to the barcodes in the assignment_confidence_table.csv. Filter the CMO read count table using the barcodes in the assignment_confidence_table.csv.

   colnames(cmo) <- unlist(lapply(colnames(cmo), x <- function(st){return(paste(unlist(str_split(st, ‘\\.’))[3], ‘-’, unlist(str_split(st, ‘\\.’))[4], sep = “))}))

   cmo <- cmo[, colnames(cmo) %in% barcode$Barcodes]

7. Calculate the total read counts of CMOs per each nucleus and percentages of reads for each CMO.

   cmo <- t(cmo) %>% as.data.frame()

   cmo$Barcodes <- row.names(cmo)

   cmo <- melt(cmo, id.vars = ‘Barcodes’)

   colnames(cmo) <- c(‘Barcodes’, ‘CMO’, ‘read_count’)

   cmo %>% group_by(Barcodes) %>% summarise(tot = sum(read_count)) %>%

   right_join(cmo, by = ‘Barcodes’) %>% mutate(perc = read_count/tot*100) -> cmo

8. Using the percentages of CMO reads per nuclei, assign CMO for each nucleus. We found that assigning and filtering based on the highest CMO read percentage and the difference between the highest and the second highest read percentage work most accurately. For example, for a given nuclei, we find the CMO with highest read percentage, and if the percentage is greater than 40% and the difference between the read percentages of two CMOs with the highest and the second highest reads is greater than 30%, we assign the CMO to the nuclei.

   highest_cut = 40

   diff_cut = 30

   second_highest <- function(l) {return(sort(l, decreasing = T)[2])}

   df <- cmo %>% group_by(Barcodes) %>% summarise(first = max(perc, na.rm = T), second = second_highest(perc)) %>% mutate(diff = first – second)

   assignment <- filter(df, first >= highest_cut & diff >= diff_cut)

   write.table(assignment, paste(‘Example_table_assignment_high_’, highest_cut, ‘_diff_’, diff_cut, ‘.txt’, sep = “), sep = ‘\t’, col.names = T, row.names = F)

9. Add the assignment from step 8 as a metadata to the Seurat object generated in step 5. Proceed to analyze the gene expression data.

REAGENTS AND SOLUTIONS:

Homogenization Buffer (HB)

The homogenization buffer is used to isolate nuclei (Nadelmann et al., 2021). The nuclei isolation buffer 1 (NIM1) can be prepared in advance and frozen for long-term storage in aliquots. The components of NIM1 buffer are listed below and can be stored at 4°C for up to 6 months. All components must be sterilized. Some components (e.g., 1.5 M Sucrose) can be sterilized throughout a 0.22 μM filter system. On the day of sample processing, prepare fresh HB by mixing NIM1 buffer with other reagents following the recipe from below. The HB needs to be chilled through the experiment procedure. Adjust the volume according to the number of samples if needed.

	Components	Volume	Final Concentration
NIM1 (For 10 mL)	1 M Tris-HCl PH 7.5	100 μL	10 mM
	2 M KCl	125 μL	25 mM
	1 M MgCl2	50 μL	5 mM
	1.5 M Sucrose	1.667 mL	250 mM
	Nuclease-Free Water	8.058 mL	-
	total	10 mL
HB (For one sample)	NIM1	474.5 μL	-
	1 mM DTT	0.5 μL	1 μM
	50× Protease Inhibitor	10 μL	1×
	40 U/μL RNase Inhibitor	5 μL	0.4 U/μL
	20 U/μL Superase Inhibitor	5 μL	0.2 U/μL
	10%(v/v) Triton X-100	5 μL	0.10 %(v/v)
	total	500 μL

Multiplex Wash Buffer (MWB)

MWB is used to wash nuclei through the oligo label procedure. Prepare fresh buffer following the recipe from below. This buffer needs to be kept on ice through the experiment procedure. Adjust the volume according to the number of samples if needed.

Components	Volume or weight	Final Concentration
1 M Tris-HCl PH 7.5	500 μL	10 mM
5 M NaCl	100 μL	10 mM
1 M MgCl2	250 μL	5 mM
1 M CaCl2	250 μL	5 mM
1.5 M Sucrose	8.33 mL	250 mM
BSA	2 g	4 % (w/v)
40 U/μL Protector RNase Inhibitor	50 μL	0.04 U/μL
Nuclease-Free Water	40.52 mL	-
total	50 mL

Storage Buffer (SB)

SB is used to collect nuclei after FACS. Nuclei will stay in this buffer before being loaded onto chips. Compared to MWB, SB has a lower viscosity. Prepare fresh buffer following the table below and store it on ice. Adjust the volume according to the number of samples if needed.

Components	Volume/sample	Final Concentration
PBS	995 μL	-
BSA	40 mg	4 % (w/v)
40 U/μL Protector RNase Inhibitor	5 μL	0.2 U/μL
total	1000 μL

COMMENTARY:

Background Information:

SnRNA-seq is widely applicable to differential single ‘cell’ expression studies with cultured cells or tissues. However, the cost per sample remains a significant constraint. Pooling samples with different barcodes, as described here, before snRNA-seq processing can significantly reduce per-sample cost and technical variations. Sample multiplexing requires modification of both nuclei processing and data interpretation protocols used for single cell/single sample analyses. The protocol described here introduces a scalable pipeline to multiplex nuclei with commercially available (e.g., 10X Genomics) lipid oligo barcodes. Compared to previously reported barcodes, lipid oligo barcodes do not require significant user input. The barcode contains a lipid tail that allows for a rapid conjugate into any membrane without complex procedures like electroporation. Although tagging the nuclei with lipid-linked oligonucleotide barcodes does not require complex processing, other factors impact the multiplexing procedure. Perhaps the most important question is how many nuclei are required per sample. This protocol is best for cultured cells, where approximately 5×10⁵ cells is sufficient to capture the diversity of a sample and is not particularly useful for tissue samples where 1,000–10,000 nuclei are required to even begin to capture cell composition. We describe here a deconvolution script to assign reliable barcodes to the nuclei. The script is simple to use and yields consistent results across large scale experiments.

Critical Parameters:

Sample preparation

Lipid CMOs incorporate into nuclear membrane of both cultured and primary cells, fresh and frozen cells and tissues, allowing versatility in sample preparation. Note that repeated freeze-thaw cycles damages nuclei in both quality and quantity and may cause poor recovery of nuclei. Nuclei isolation and the following oligonucleotide labelling step should be completed on the same day. We recommend collecting cells and keeping pellets at −80°C for long-term storage until ready for nuclei isolation, then using freshly isolated nuclei for multiplexing.

The number of cells stored per pellet determines the number of nuclei available for analysis. According to the CMO manufacturer’s instruction, 100μL of the oligo reagent is supposed to label less than 2 million cells. Considering the inevitable nuclei loss through the procedure, we recommend collecting 5×10⁵ – 2×10⁶ cells for each sample. Samples with less than 5×10⁵ cells show excessive loss of nuclei during the labelling procedure, because the nuclei pellet is not visible, and are not recommended for use. There are twelve different CMOs in the current 10X Genomics kit, which allows up to twelve samples to be pooled.

Yield of nuclei

Nuclei yield is a critical parameter for the entire experiment. Nuclei are vulnerable due to the lack of protection from the cell membrane yet are subjected to repetitive rounds of washing and centrifugation during the process. Considerations to reduce nuclei loss and ensure a robust yield of the nuclei include:

Harvest cells in single cell suspension.

When collecting cells, make sure they are in single cell suspension and avoid cell clumps. Cell clumps cause nuclei loss in three ways: inaccurate cell count, resulting in less cells being harvested than expected; require increased pipetting, resulting in physical damage to nuclei; nondissociated nuclei are lost during filter purification.

Staining and tracking nuclei.

Nuclei are stained with trypan blue to allow visualization during isolation and removal of supernatants (Fig 2A). If color fades, we recommend repetitive staining after washing.

Wash buffer.

The ideal buffer stabilizes nuclei and avoids degradation. Buffers with BSA and sucrose prevent the nuclei sticking to tubes and tips. MWB is a Tris-HCl buffer consisting of salts, sucrose, BSA and RNase inhibitors. When visualized by Countess II automated cell counter, the nuclei processed in MWB buffer are uniform in size and morphology, indicating high quality (Fig. 2C). Rinsing new tubes or pipet tips with the MWB buffer can limit nuclei sticking to plasticwares. However, before loading the labeled nuclei onto chips, we recommend resuspending the nuclei in SB, which lacks sucrose (see Reagents and Solutions part).

Centrifugation.

Centrifugation speed and time need to be considered when handling nuclei. We found spinning with 600 – 800g for 5 minutes works well. Compared to fixed-angled spinning, swinging buckets yield better pellets at the bottom of the tubes. In this protocol, we assembled the adaptors from a TissueLyser II device into a swinging bucket to fit the Eppendorf tubes.

The use of flow cytometry

Flow cytometry sorting separates singlet nuclei from unwanted subcellular debris or clumps. Using forward and sideward scatter measurements, larger cell debris or multiplets can be separated. Smaller debris, similar in size to nuclei, is distinguished from nuclei by using Hoechst-A dye (NucBlue™ live ReadyProbes™ reagent), which stains the nuclei with a fluorescent dye, so that intact nuclei emit blue fluorescence and generates two populations: Hoechst-A positive nuclei and negative debris (Fig. 2B). Flow cytometry is an optional but highly recommended step. The sorting can be either done before or after sample pooling. To adjust the flow cytometry background fluorescence, prepare a negative control (nuclei without Hoechst-A stain) alongside the labeled samples.

Deconvolution strategy

CellRanger provides two outputs for assigning each nucleus to CMO. One is assignment probability, which is calculated by the CellRanger software suite (assignment_confidence_table.csv). The second is CMO read count, normalized by UMIs. We found using CMO read counts for deconvolution yields consistent results across multiple runs regardless of the quality of library or samples. To assign nuclei to a sample, first convert CMO read counts to percentage of reads per nuclei and assign based on the CMO with highest read percentage. Multi-barcoded nuclei (one nucleus with two lipid oligo barcodes) are identified by comparing the highest and second highest CMO read percentages. We assign a barcode to nuclei if the highest read percentage is greater than 40% of all barcodes, and the difference between the highest and the second highest read percentages is greater than 30%. Multi-barcoded nuclei usually comprise a small fraction of nuclei and are removed from subsequent analyses. Filtering thresholds can be adjusted depending on the final assigned nuclei count.

Troubleshooting:

See Table 1 for a list of problems, causes and solutions.

Table 1.

Troubleshooting Guide for Multiplexed snRNA-seq and Data Analysis

Step	Problem	Possible Cause	Solution
13	Poor cell viability detected by Cell Counter	Cells are not healthy	Obtain new cells, check the cells’ quality before collection, ensure no contamination
		Cells underwent enzymatic digestion for too long	Reduce digestion incubation time
		Pipet the cells with enzymatic digestion reagent alone, or dislodged for too many times	Neutralize enzyme with medium before pipetting. Don’t repeat pipetting for over 15 times
	Cells aggregate together	Dissociation reagent is not working well	Use fresh enzymic dissociation reagent, or warm it up at 37°C for half an hour before using
		Cells are not filtered through a strainer	Use desired strainer to filter the suspension
22	Cell pellets hard to dissociate	Cell pellets are not collected in single cells Old pellets frozen for too long Specific cell types, we find cultured cardiomyocytes are harder to break	Pipet more times, remove small unbreakable clumps with a filter. If none of these works, prepare new fresh sample, make sure it’s collected from single cell suspension
		Old pellets frozen for too long
		Specific cell types, we find cultured cardiomyocytes are harder to break
35,38	Nuclei pellets are not visible	Trypan Blue dye fades	Do not remove supernatant. Add another 10 μL Trypan Blue reagents to dye, mix and spin again
		Invalid centrifugation	Count nuclei in the supernatant, and if any, spin them down again
		Too much nuclei loss	Check if the wash buffer is correct and fresh. It is the most common reason for poor nuclei yield
40	Clumps of nuclei	Trypan Blue is old and clumpy	Use fresh Trypan Blue
		Insufficient resuspension	Resuspend again and recount
41	Samples have less than 100,000 nuclei prior to pooling	Low nuclei yield	This can be caused by many reasons: starting with too few cells, nuclei loss during isolation or washing, and inaccurate count results. In this case, pool everything. The smallest number we’ve tried to pool is 28,000 and managed to get a reasonable nuclei recovery for data analysis
9	Poor sample recovery by barcode assignment	Cross-labeling between samples	Ensure no mix-up between barcodes and sufficient washing to remove unconjugated barcodes
		Oligo barcodes are not working well: expired or go through too many freeze-thaw cycles	Use a new set of barcodes
		Inaccurate numbers of cells being pooled	When counting nuclei, ensure the concentration is within the recommended range. Nuclei will quickly pellet down without disturbing. Pipet up and down to make a uniform resuspension before pooling.
		Inappropriate threshold for assignment	Try less strict thresholds

Understanding Results:

An example of output from a twelve-sample experiment illustrates the sample multiplexing and deconvolution workflow. In this example, 12 samples comprised of four cultures for each of three differentiated cell types (endothelial cells, iPSC-ECs; cardiomyocytes, iPSC-CMs; and neural crest cells, iPSC-NCCs) were multiplexed. Each sample was lysed to isolate nuclei and then labeled with a distinctive lipid-oligo (CMO). Equivalent numbers of nuclei for each sample were pooled and loaded onto one snRNA-seq run. We deconvoluted the run and assigned a CMO to each nucleus when the highest CMO read percentage was greater than 40%, and the difference between the highest and the second highest read percentage was greater than 30%. Possible doublets were removed using Scrublet (score < 0.3) (Wolock et al., 2019). Using this threshold, approximately 1,000 nuclei were assigned to each sample (Table 2). Transcription profiles of the samples were then visualized by UMAP (Fig. 3). Nuclei that were assigned as endothelial cells, cardiomyocytes, or neural crest cells were highlighted with red, blue, or green respectively. We observed that the nuclei clustered by the assigned cell types. Marker gene expression for each cell type was also consistent with the sample assignment.

Table 2.

Numbers of CMO Assigned Nuclei for Each Sample

Cell types	Sample	No. of assigned nuclei
iPSC-EC	Endo 1	524
	Endo 2	968
	Endo 3	804
	Endo 4	1,515
iPSC-CM	CM 1	1,081
	CM 2	710
	CM 3	1,294
	CM 4	1,015
iPSC-NCC	NCC 1	1,274
	NCC 2	1,169
	NCC 3	1,168
	NCC 4	1,425

Fig. 3. The multiplexed snRNA-seq pipeline generates accurate cell identity assignment.

Transcription profiles of twelve samples are visualized by UMAPs. (A) Nuclei assigned to endothelial cells (left panel), cardiomyocytes (middle panel), and neural crest cells (right panel) are highlighted with red, blue, and green respectively. (B) Normalized expression of marker genes for each cell types are plotted: PECAM1 demarcates endothelial cells (Endo, red), TTN demarcates cardiomyocytes (CM, blue) and PAX3 demarcates neural crest cells (NCC, green).

Time Considerations:

Collecting cultured cell and storing cell pellets takes approximately 1 hour, although multiple cell cultures can be collected simultaneously. Each additional sample generally adds 5 minutes to processing time to allow for the aspiration and suspension.

Further nuclei processing also depends on the number of samples. Processing 12 samples simultaneously usually requires 30 minutes for reagent preparation and thawing frozen pellets. Subsequent cell lysis, incubation, filtration, and centrifugation is a 20-minute procedure. The nuclei labeling step takes approximately 1 hour, including a 10-minute barcodes incubation time and four rounds of washing and spinning. Nuclei concentrations of 12 samples can be determined using automated cell counter, in about 30 min. The FACS sorting to separate nuclei from debris typically takes 30 minutes. After sorting, nuclei should be spun down and resuspended to an ideal concentration before being loaded for encapsulation, which takes 30 minutes. Subsequent gel-beads encapsulation, cDNA amplification and library construction will take another 9 hours, although these steps can be processed on different days. Data processing usually requires a day to align the fastq files and another hour to deconvolute.

DATA AVAILABILITY STATEMENT:

The data that support the findings of this study are available from the corresponding author upon reasonable request.