Single cell suspension
T-47D single cell suspension was prepared by removal of growth medium and subsequent washing of cell layer using 10 ml PBS. Cells were disaggregated with 2 ml TrypLETM Express Enzyme (1X) (cat nr. 12,604,013, GibcoTM, USA) for approximately 2 min in an incubator at 37 C°. Reaction was stopped adding full growth media (RPMI-1640 + GlutaMAXTM (cat nr. 61,870–010, GibcoTM, UK) + 10% Fetal Bovine Serum (FBS) + 5% Penicillin/ Streptamycin) in double the amount of TrypLETM. Suspension was spun down 3 min 1200 rpm. Cells were washed once in PBS, resuspended in FACS buffer (PBS + 0.04% bovine serum albumin (BSA)), and filtered using a 100 μm strainer.
Cell surface marker staining
Cell suspension was stained with monoclonal antibodies toward human EpCAM (cat nr. 130- 111–116, lot nr. 5,190,125,111, Miltenyi Biotec, Germany) conjugated to R-phycoerythrin (PE) (Ex-Max 496 nm/ Em-Max 579 nm), and Human Integrin α6 (CD49f) (cat nr. 25–0495-80, lot nr. 4,319,156, eBioscienceTM, USA) conjugated to PE-cyanine 7 (PE/Cy7, Ex-Max 496 nm/ Em-Max 785 nm). 4’,6-diamidino-2-phenylindole (DAPI) (Ex-Max 358 nm/ Em-Max 461 nm)(cat nr. 10,236,276,001, Roche Diagnostics GmbH, Germany) was used to discriminate dead from live cells. Concentration was adjusted to 105–106 cells/ml, and antibody added at concentration of 1 μl per 105–106 cells. Cells were incubated 15 min at 4 C° in the dark. Following incubation, dye was diluted, adding 1 ml of FACS buffer. Suspension was subsequently washed once with 1 ml FACS buffer. Pellet was resuspended in 1 ml FACS buffer. 10 μl DAPI was added per 1 ml cell suspension, to a final concentration of 300 nM, 5 min prior to sorting.
Single cell sorting
Fluorescence activated cell sorting (FACS) was used for isolation of single T47D cells into lysis buffer containing 96-well semi-skirted PCR plates using instrument BD FACSAria™ III (BD bioscience, USA). Controls for calibrating instrument included unstained cells, cells stained with DAPI, and beads compensating for spectral overlap between fluorochromes PE and PE-CY7, using MACS Comp Bead Kits Anti-REA (cat nr. 130–104-693, lot nr. 5,181,012,289, Miltenyi Biotec, Germany) and Anti-RAT (cat nr. 130–107-755, lot nr. 5,181,015,526, Miltenyi Biotec, Germany). Cell suspension was gated to isolate double positive (EpCAM + /CD49f +) single T47D cells. A Multi-cell positive control (50 cells) and an empty-well negative control (0 cells), and at least one RNA-control diluted to 10 pg/ul were provided per plate. Following sorting cells were thoroughly vortexed, spinned down 1 min, flash- frozen in dry-ice and subsequently stored at -80 C°.
NEBNext® Single Cell/ Low Input RNA Library Prep Kit
Fourteen single cells were processed using NEBNext® Single Cell/ Low Input RNA Library Prep Kit (cat nr. E6420S, New England Biolabs (NEB), USA), and were sorted into 5 μl NEBNext Cell Lysis buffer (0.5 μl NEBNext Cell Lysis Buffer (10x), 0.25 μl Murine RNase Inhibitor, 4.25 μl H2O). Protocol was performed according to recommendation by manufacturer with minor changes—1 μl 1:106 dilution of ERCC spike ins (cat nr. 4,456,740, Invitrogen, Thermo Fischer Scientific, Lithuania) were added each single cell lysate prior to RT and PCR amplification was performed applying 22 cycles.
SMART-seq® HT Kit
Thirteen single cells were processed using SMART-seq® HT kit (cat nr. 634,862, Takara Bio Inc, USA), and were sorted into 12.5 μl FACS dispensing solution (0.95 μl 10xLysis buffer, 0.05 μl RNase Inhibitor,1 μl 3’SMART-Seq CDS Primer II A, 10.5 μl Nuclease-free H2O). Protocol was performed according to recommendation by manufacturer with minor changes—1 μl 1:106 dilution of ERCC spike ins (cat nr. 4,456,740, Invitrogen, Thermo Fischer Scientific, Lithuania) were added each single cell lysate prior to RT, and PCR amplification was performed applying 20 cycles.
G&T-seq
Thirteen single cells were processed by G&T-seq and were sorted into 2.5 μl RLT Plus buffer (cat nr. 1,053,393, Qiagen, Germany). ScRNA-seq was performed as described by Macaulay et al., 2016. Single cell DNA sequencing featured in this protocol was not conducted in this experiment, but DNA stored at -80 C°. Each single cell lysate were added 1 μl 1:106 dilution of ERCC spike ins (cat nr. 4,456,740, Invitrogen, Thermo Fischer Scientific, Lithuania) prior to RT. PCR amplification was performed applying 20 cycles.
SS3
Fourteen single cells were processed by SS3 protocol and sorted into 3 μl SS3 lysis buffer mix [35]. ScRNA-seq was performed as described by Sandberg et al. 2020. Each well was added a concentration of 1:106 ERCC spike ins (cat nr. 4,456,740, Invitrogen, Thermo Fischer Scientific, Lithuania). PCR mix was prepared using a working concentration of Kapa HiFi HotStart ReadyMix (1X) (Kapa, cat. no. KK2601), Fwrd. primer (0.5uM), Rev. primer (0.1uM). PCR amplification was performed applying 20 cycles.
Sequencing
T47D single cell cDNA libraries were paired-end sequenced in groups according to each library preparation protocol (13 or 14 single cells per run) on MiSeq Benchtop Sequencer (Illumina, USA), using MiSeq Reagent kit v2 300 cycles (cat nr. MS-102–2002, Illumina, USA). Prior to sequencing each single cell library was diluted to a concentration of 4 nM in EB buffer + 0.1% Tween 20. Prior to sequencing 3 μl of each 4 nM library was pooled in an Eppendorf tube. 5 μl 4 nM pool was mixed with 5 μl 0.2 nM NaOH and incubated 5 min at RT, for denaturing of double stranded cDNA. The denatured sample pool was diluted to a concentration of 20 pM by mixing 10 μl 2 nM sample pool with 990 cold Hybridization Buffer 1 (HT1). Finally, 20 pM sample pool was diluted to a concentration of 10 pM, by mixing 500 μl 20 pM sample pool with 500 μl cold HT1.
Alignment/ trimming
Illumina sequencing raw reads were converted to fastq files. Fastq files were processed on a bash shell. Raw reads were trimmed using Trim Galore v.0.4.0 [36] with default parameters, where two rounds of trimming were performed. The first trimming removed Nextera XT adaptors (”CTGTCTCTTATACACATCT”), and the second trimming removed cDNA amplification adaptors (”AAGCAGTGGTATCAACGCAGAGT”). Quality assessment of sequencing output was performed by two rounds of FastQC after each trim. Trimmed sequences were aligned using the splice-aware aligner STAR v2.5.2b [37] to Genome Reference Consortium Human Build 38 (GRCh38), with the following parameters: –genomeLoad NoSharedMemory –quantMode TranscriptomeSAM GeneCounts –readFilesCommand zcat –outSAMtype BAM SortedByCoordinate –limitBAMsortRAM 35,000,000,000. STAR output files comprising ReadsPerGene.out.tab for each sample, which were merged into a single tab delimited file, so called “count matrix”, for further analysis.
Data visualisation
Introductory figures Illustrated using Biorender (https://biorender.com/). ScRNA-seq data was imported into R studies as an expression matrix. The count matrix was transformed into a Single Cell Experiment Object (SCE-object) using Rstudios (v3.6.1 Opensource, https://rstudio.com/) package SingleCellExperiment (v1.6.0). Data graphs were generated using ggplot2 (v3.3.0). Plots are either Sinaplots [38] or box-plots where with outlier threshold at X and wiskers at Y, stars are denoted based on Wilcoxon test p-value; ns: p > 0.05, *: p < = 0.05, **: p < 0.01, ***: p < 0.001, ****: p < 0.0001.
Single cell Quality Control (QC)
Expression level of genes were quantified by CPM (counts per million). Genes with an average expression above zero (CPM > 0) across all cells were kept in the dataset. Cells not expressed in any cell (CPM = 0) were filtered away. Bad quality cells (or empty wells) were filtered away based on the following criteria: 1) cells that had less than 1000 uniquely expressed genes, 2) cells that had library sizes below 0.6e6 million reads, 3) cells that had more than 30% reads mapped to mitochondrial genes, and 4) cells that had more than 25% of reads mapped to ERCC spike-inn genes. Data analysis was performed with RStudio (version 3.6.1) using Bioconductor [https://www.bioconductor.org] packages (SingleCellExperiment, sinaplot [38], scater [39], ggplot2, GenomicFeatures [40], sincell [41], TxDb.Hsapiens.UCSC.hg19.knownGene, SummarizedExperiment, robCompositions, splatter [42], reshape2, ggforce, gdata, hrbrthemes, viridis, VennDiagram, DESeq2 [43], dplyr, tidyverse [44], gtable, gridExtra, hrbrthemes, ggforce, ggpubr) following guidelines from [https://scrnaseq-course.cog.sanger.ac.uk/website/index.html].
Gene body coverage
Read coverage over gene body was analysed using geneBody_coverage method in RSeQC package v4.0.0 [45]. The further visualisation was performed using ggplot2 (v3.3.0) in R.