Differential expression (DE) analysis focuses on comparing the relative expression levels of different transcripts in the cell and is one of the primary analysis tools used to explore transcriptome variability. End-capture methods are appealing for DE analysis as they can decrease the number of reads necessary to determine differential expression between cells. By focusing the sequencing data on a portion of each transcript (in this case, the 3' end of each transcript), we can reduce the number of reads and hence the overall cost required to identify expressed genes. Additionally, samples can be pooled prior to sequencing, decreasing the work and resources required and increasing the multiplexing capabilities of each sequencing lane.

In this tech note, we demonstrate the use of SMART-Seq v4 chemistry combined with an end-capture method for low-input amounts down to the single-cell level. As we have previously reported, SMART-Seq v4 technology produces high-quality cDNA libraries from individual cells that closely represent the original in vivo mRNAs. It is extremely sensitive, works with transcripts of different lengths, and has excellent gene body coverage across a wide range of GC content (see References for details). In order to reduce labor and cost, we have adapted this robust technology for 3' end capture, and here we present this approach and validate it by sequencing single K562 cells.

This approach allows each pool of cDNA from 12 single-cell reactions to be tagged by one of the 96 Illumina HT barcode combinations, enabling up to 1,152 separate cell reads per run. Researchers using this kit can confidently determine differentially expressed transcripts while decreasing the cost and time required for discovery. This method utilizes a modified oligo(dT) primer including an in-line index which serves as a cell barcode and a portion of the Illumina read primer 2 sequence in order to accommodate a pooled library generation protocol (Figure 1). The in-line index is placed between the transcript and the Illumina read primer 2 (RP2), and enables pooled cell samples to be demultiplexed after sequencing.

Figure 1. Overview of the end-capture method for analyzing differential expression. cDNA (black lines) is synthesized with a blocked (black star) and modified oligo(dT) primer that adds sequences for subsequent amplification and analysis—a cell barcode (magenta), part of the Illumina read primer 2 sequence (RP2, yellow), and the SMART IIA sequence (green). The SMART IIA sequence is used as a priming site during cDNA amplification, the Illumina RP2 sequence is used as a priming site during library amplification, and the cell barcode is used for demultiplexing pooled samples during analysis. The process works as follows: first, the template for SMARTScribe reverse transcriptase switches from the mRNA (blue wavy line) to the SMART-Seq v4 Oligonucleotide (green). After reverse transcription, the full-length cDNA is amplified by PCR with blocked Primer IIA oligonucleotides. After cDNA amplification, the presence of the barcode (magenta) allows for pooling up to 12 samples. The pooled samples are tagmented and Illumina Nextera® read primer 1 and 2 sequences are added by the Nextera Tn5 transposon (TnRP1 and TnRP2, orange and purple respectively). The 3' ends of the original mRNA are captured by selective PCR with primers for the TnRP1 and RP2 sequences. Other products of the transposon-based reaction are not amplified, either because there are no primer sites for amplification or because of suppression PCR. Cluster generation (pink and dark purple) and indexing sequences (light blue and dark blue) are added during this PCR stage to generate a library ready for sequencing on an Illumina platform.

High sensitivity of the end capture method

We used this method to sequence 12 single K562 (human immortalized myelogenous leukemia) cells in one pooled sample. The total number of reads in the pool was 21.6 million, with reads from individual cells ranging from 0.2 to 5.0 million, as seen in Table 1. This variability is due to the different quantities of cDNA generated from each cell. These reads were then used to identify expressed genes in each cell and determine the correlation between expression profiles for each cell.

Gene expression analysis and cell-to-cell correlation

After obtaining the read data from the pooled cell samples, the data was analyzed to determine the number of identified genes using different cutoff values for the Counts Per Million mapped reads (CPM). As seen in Figure 2, a high number of genes were identified for all cells, with some differences due to the total number of reads varying between cells. We then determined the Pearson correlations of regularized log-transformed read-counts, shown below in Figure 3. As expected, we observed high correlations (>0.7) between all cells, with most cells demonstrating a correlation >0.9.

Figure 2. Number of genes identified from K562 single cells. Mapped libraries were analyzed with CPM) generated from STAR v.2.3.0.1 (Dobin et al. 2013). The number of genes identified with different cutoffs (0.1, 1.1, 2.1, 3.1, 4.1, 5.1, 6.1, and 7.1) for log-transformed CPM+1 are plotted. The amount of cDNA produced from each cell varies, leading to different read depths per cell. This affects the number of genes identified, as seen most obviously for sample 10 at lower expression cutoff values.

Figure 3. Pearson correlation heat map matrix of K562 single cells. The heat map represents the Pearson correlations of  expression levels for the 12 single-cell libraries. For all comparisons, the correlation (R) was >0.8, while the majority of single-cell libraries are highly correlated (>0.9).

Verification of the end-capture method

Finally, we verified that the end-capture method was functioning correctly by determining where the majority of reads were occurring along each transcript. As seen in Figure 4, the majority of reads mapped to the last 30% of the transcripts, as expected with our end-capture method.

Figure 4. Gene body coverage analysis. Once the reads were mapped to the human genome, gene body coverage analysis was performed to assess the ability of the methods to capture the 3' ends of the cDNA. The majority of reads across all transcripts mapped to the last 30% of the transcripts (normalized in length to 100%).

The SMART-Seq v4 3' DE Kit provides a more cost-effective method for identifying differentially expressed genes in single cells. This will enable DE experiments with higher sensitivity for scientists interested in quickly identifying differences between cells without sequencing the entire transcriptome. Our method demonstrates great sensitivity and delivers high-quality, robust, and reproducible transcriptomic data.

In the experiments described here, K562 cells (human leukemia cell line) were isolated, evaluated via optical microscopy, lysed, and processed for cDNA library construction. This was accomplished by diluting the cells in PBS to 1 cell/µl, then spotting 1 µl drops in each well of a 96-well plate and visually choosing single cells to pick and process. cDNA libraries were prepared from single cells according to the protocol provided with the SMART-Seq v4 3' DE Kit.

Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013). 

Heaton, N. S. et al. Long-term survival of influenza virus infected club cells drives immunopathology. J. Exp. Med. 211, 1707–14 (2014). 

Henley, B. M. et al. Transcriptional regulation by nicotine in dopaminergic neurons. Biochem. Pharmacol. 86, 1074–83 (2013).

Saliba, A.-E., Westermann, A. J., Gorski, S. A. & Vogel, J. Single-cell RNA-seq: advances and future challenges. Nucleic Acids Res. 42, 8845–60 (2014).

Additional reading about SMART technology:
SMART-Seq v4 sensitivity is discussed in our SMART-Seq v4 single-cell tech note and additional data on sensitivity across a wide range of GC content is discussed in our Fluidigm C1 single cell tech note.