Splice variants or alternative isoforms play an important role in the understanding of human health and disease. It is estimated that nearly all protein-coding genes in the human genome are alternatively spliced, providing an essential source of protein diversity (Pan et al. 2008; Wang et al. 2008). As many cancer-associated genes are regulated by alternative splicing, tumor-specific splice variants have clear diagnostic value as biomarkers and may serve as potential drug targets in novel therapeutics (Zhang et al. 2021).

One of the toughest challenges in biology and medicine today is to map genotypes to phenotypes, which can be tackled by performing transcriptomics analysis via single-cell mRNA sequencing (Hwang et al. 2018). Full-transcriptome mRNA sequencing enables the discovery of rare biological events such as gene fusions, SNPs, and alternative splicing, which is an essential first step to new advancements in medicine.

Though there are plate-based methods for full-length scRNA-seq, they often lack integrated analysis tools. Our complete, automated solution for enhanced biomarker detection includes optimized SMART-Seq chemistry on the ICELL8 cx Single-Cell System and Cogent NGS bioinformatics tools. This streamlined workflow maintains the robust detection of genes and transcripts seen with plate-based kits like Takara Bio’s SMART-Seq v4 and surpasses common homebrew methods like Smart-seq2 (Figure 1). With the full-length coverage of our SMART-Seq Pro kit, you can identify clinically relevant novel biomarkers with confidence.

Figure 1. Comparison of genes detected from PBMCs using SMART-Seq Pro or Smart-seq2 (SS2) workflows.

Reference experiment

To demonstrate automation of in-depth scRNA-seq using the SMART-Seq Pro kit for the ICELL8 cx Single-Cell System, Takara Bio scientists recreated part of a transcript isoform study from Peking University (Li et al. 2020). This research focused on identifying and characterizing gene isoforms related to non-small cell lung cancer. The study required a full-length transcriptomics approach to explore these isoforms of interest and used plate-based RNA-seq to complement the limitations of 10x Chromium data.

Their study focused on the Protein Tyrosine Phosphatase Receptor Type C (PTPRC) gene, which encodes the transmembrane tyrosine phosphatase, CD45, and has several isoforms with variations across many exons. PTPRC isoforms are critical for T-cell receptor (TCR) signal transduction and are important for regulating tumor-infiltrating T cells. The authors further performed clonal analysis to show that isoform switching occurs during T-cell activation and differentiation. 

They observed that phenotypic naïve CD8+ T cells—cells that are mature but not yet activated—were represented dominantly in the PTPRC-209 isotype. Exhausted CD8+ T cells, cells with progressive loss of function due to prolonged antigen stimulation, were most represented in the PTPRC-201 isotype. Unique exon spanning profiles for PTPRC-201 and PTPRC-209 that were uncovered through the study are shown in Figure 2.

Figure 2. PTPRC-201 and PTPRC-209 isoforms show unique exon spanning profiles for the region shown. Data were generated using Ensembl tools, with the target region highlighted.

Characterization of PTPRC using the SMART-Seq Pro kit and ICELL8 cx system

We adapted the reference study’s approach for the ICELL8 cx Single-Cell System, Takara Bio’s automated platform for creating full-length scRNA-seq libraries. Our comparative study focused on the PTPRC gene to see if isoform patterns similar to the reference study would be found in PBMCs.

The experiment was carried out using the SMART-Seq Pro protocol for the ICELL8 cx system. Data were analyzed using Cogent NGS Analysis Pipeline (CogentAP) and Cogent NGS Discovery Software (CogentDS) tools, freely available for ICELL8 cx users. 

Our results demonstrated that the ICELL8 cx system successfully identified PTPRC-201 and PTPRC-209 isoform profiles, matching the abilities of plate-based techniques used in the reference study.

After demultiplexing and mapping the transcript data with CogentAP, data were analyzed using CogentDS to generate a UMAP showing various gene clusters (Figure 3, Panel A). Each of the clusters was characterized for the prevalence of the two isoforms of interest, PTPRC-201 and PTPRC-209, using the violin plots generated from publicly available IGViewer tools. PTPRC-201 was found in higher densities in Cluster 1 and lower densities in Cluster 5 while the reverse was true for PTPRC-209 (Figure 3, Panels B and C). The overlayed UMAPs generated from CogentDS software support these findings regarding the presence of PTPRC-201 and PTPRC-209 isoforms (Figure 3, Panels D and E). Transcript-level analysis helped to identify immune cell clusters with biological significance, similar to the findings of the reference study.

Figure 3. Data showing clustering of PTPRC-201 and PTPRC-209 isoforms in PBMCs. Panel A. UMAP plot showing cell clusters containing PTPRC-201 and PTPRC-209 isoforms using transcript-based counts (generated via CogentAP and UMAP clustering via CogentDS). Cluster 1 and Cluster 5 refer to the PTPRC-201- and PTPRC-209-rich clusters, respectively. Panels B and C. Violin plots show the overall distribution and density of transcript expression across cells in each UMAP cluster. PTPRC-201 showed a higher prevalence of transcripts in Cluster 1 and lower in Cluster 5, while PTPRC-209 showed an opposite trend. UMAP overlay features of the CogentDS highlight cells matching the profile of PTPRC-201 (Panel D) and PTPRC-209 isoforms (Panel E).

Next, we looked at the presence of these isoforms across exons using sashimi plots that summarize splice junction reads as a visual representation (Figure 4). Similar to the reference publication, a clear profile was found between exons 3 through 7 to distinguish PTPRC-201 from PTPRC-209. We then compared our findings to data generated from the 10x Chromium instrument for the same cell types to illustrate the importance of full-length investigation.

For 10x Chromium data, where the transcript is only read at the 3’ end, a unique profile was not seen for PTPRC-201 and PTPRC-209. The low peaks seen for 10x Chromium in some of the exon regions were due to background noise and did not represent meaningful data.

Figure 4. Comparison of isoform data generated using the SMART-Seq Pro kit or 10x Chromium. The SMART-Seq Pro method using the ICELL8 cx system can identify junction reads across exons 3, 4, 5, 6, and 7 and shows distinct profiles for PTPRC-201 and PTPRC-209 isoforms. Data generated from 10x Chromium end-counting methods could not identify junctions in these same exon regions, with peaks representing noise from misaligned reads.

We successfully reproduced the reference study's PTPRC profile on our system, unlike droplet-based methods that lack the sequencing depth to identify biomarkers further within the transcriptome. The data illustrate how the SMART-Seq Pro chemistry can characterize clinically relevant isoforms and how the SMART-Seq Pro kit, together with the ICELL8 cx Single-Cell System, offers an efficient way to scale up studies for biomarker investigation.

The findings show how the ICELL8 cx system can automate the trusted performance of plate-based full-length scRNA-seq without compromising detection power, reproducibility, or resolution. With the included Cogent NGS analysis tools, the end-to-end solution of the system was demonstrated from sample to library preparation to data analysis for an isoform of clinical importance. 

Cell staining and preparation

Cryopreserved PBMCs (Cat. # HUMAN-PBMC-M-170046, lot # 00PB000450, BioIVT) were thawed in a 37°C water bath for 60 seconds and then topped off with pre-warmed RPMI-1640 Complete Medium (20% FBS). After centrifugation and subsequent removal of the supernatant, the concentration of recovered PBMCs was determined via a Moxi automated cell counter. PBMCs were then stained with Hoechst 33342 and propidium iodide as described in the SMART-Seq Pro application kit user manual.

Cell dispensation and imaging

The PBMCs and controls were then dispensed into 5,184 nanowells of the ICELL8 350v Chip using the ICELL8 cx system and ICELL8 cx CELLSTUDIO v2.5 Software. The nanowells were then imaged by the ICELL8 cx system with both blue and red wavelength filters. After imaging, the chip was frozen at –80°C while the ICELL8 cx CellSelect v2.5 Software was used to analyze the resulting images with the automated threshold detection. After selecting candidate wells, the software generated a filter file to use for all the remaining dispenses.

Library construction

The chip was thawed and returned to the ICELL8 cx system, where RT reagents were dispensed only into the nanowells defined as candidates by the CellSelect filter file. The chip was run through a program to perform first-strand cDNA synthesis on the ICELL8 cx Thermal Cycler initiated by SMART-Seq Pro CDS (an oligo-dT primer). Following first-strand cDNA synthesis, the SMART-Seq Pro Oligonucleotide was hybridized to the 3′ end of the full-length cDNA and mediated template switching, serving as a priming site for second-strand cDNA synthesis. Once synthesized, the second-strand cDNA was amplified through PCR, resulting in copies of unbiased, full-length cDNA. After amplification, the full-length cDNA was tagmented by Illumina Bead-Linked Transposome (BLT). The tagmented cDNA was then amplified using forward and reverse indexing primers, generating the final library construct.

The resulting libraries were extracted from the chip, purified, amplified, and purified again. After validation steps, the libraries were loaded on the Illumina NextSeq® 500 system. The resultant sequencing reads were downsampled to 100,000 reads per cell to see the number of genes detected.

Gene expression analysis

Cogent NGS Analysis Pipeline (CogentAP) v1.5 was used with the option for ICELL8 full-length chemistry. As part of the pipeline, steps including adapter trimming (using cutadapt tool), genome alignment (using STAR), gene read counting (using featureCounts), and transcript read counting (using RSEM) were performed. The resulting gene/transcript matrices were then input into Cogent NGS Discovery Software (CogentDS) v1.5 to perform clustering analysis and generate UMAP plots and clusters. For the 10x Chromium dataset, the same tools were used to generate the gene-matrix and bam files. The option for UMI-aware ICELL8 3’-DE chemistry in CogentAP was used for this analysis. The resulting bam files from both datasets were imported into the IGViewer tool to generate sashimi-plots.

Howe, K. L. et al. Ensembl 2021. Nucleic Acids Res. 49, 884–891 (2021).

Hwang, B. et al. Single-cell RNA sequencing technologies and bioinformatics pipelines. Exp. Mol. Med. 50, 1–14 (2018).

Li, J. et al. Landscape of transcript isoforms in single T cells infiltrating in non-small-cell lung cancer.  Genet. Genomics 47, 373–388 (2020).

Pan, Q. et al. Deep surveying of alternative splicing complexity in the human transcriptome by high-throughput sequencing. Nat. Genet. 40, 1413–1415 (2008).

Wang, E. et al. Alternative isoform regulation in human tissue transcriptomes. Nature 456, 470–476 (2008).

Zhang, Y. et al. Alternative splicing and cancer: a systematic review. Sig. Transduct. Target Ther. 6, 78 (2021).