Applying single-cell long nanopore sequencing reads to identify the genetic basis underlying resistance in leukaemia treatment

The genetic conundrum underlying treatment resistance

Leukaemia kills 0.35 million people every year and the most common adult leukaemia in Western countries is chronic lymphocytic leukaemia (CLL). In 2016, a breakthrough therapy called venetoclax was approved by the FDA and revolutionised the treatment of CLL¹. While we all hoped that the BCL2 inhibitor venetoclax was the cure for CLL, most patients started to relapse after years on treatment. DNA bulk sequencing by ourselves and others demonstrated that a BCL2 mutation² or amplification of another pro-survival gene MCL1³ could confer resistance, however this is not the full story. These phenomena only happen at a subclonal level and not in all patients. What is happening in the cells that don’t have these genetic changes but still grow out under continuous venetoclax pressure? To answer this question, we applied single-cell RNA sequencing (scRNA-seq) to research samples from individuals with progressive CLL, obtained from the longest follow-up venetoclax trial in the world.

Method development for simultaneous analysis of expression and genotype in single cells

We captured 10,000 cells per sample with a high-throughput droplet-based method that indexed each transcript with a unique cell barcode to identify transcripts from the same cell. In standard scRNA-seq methods, short reads are used as the read-out, so most of the cDNA information is lost, including mutation status and isoform usage. To better understand how cell phenotype contributes to drug resistance, we combined single-cell short-read sequencing with long nanopore sequencing reads to simultaneously measure gene expression and full-length transcripts to determine genotype.

Performing whole transcriptome nanopore sequencing on 10,000 single cells is relatively expensive, so we developed two new methods to reduce sequencing costs (see Figure). These costs were reduced by less input in the first method, developed by Luyi Tian and Jafar Jabbari⁴. After we captured single cells in droplets, the droplets containing the indexed cDNA were split into 80% and 20% subsamples and processed as two independent samples for short-read library prep and sequencing. From the 20% subsample (~2,000 cells), full-length cDNA was further amplified and sequenced by the Oxford Nanopore platform. Using the FLAMES pipeline⁴, short reads were processed according to standard procedures and the cell barcodes from short reads were used to demultiplex the long nanopore reads.

In our recent publication⁵, we described the second method that Jafar and I developed. Instead of reducing the cell number, we reduced output by looking at full-length transcripts of specific genes. In our method called single-cell Rapid Capture Hybridization sequencing (RaCH-seq), we used biotinylated probes to target each exon of our 17 genes of interest for pulldown with streptavidin beads. We applied this method to the stored cDNA of the 80% subsample (~8,000 cells) and the captured full-length transcripts were sequenced on the Oxford Nanopore platform. Leveraging single-cell long nanopore reads with short-read sequencing allowed us to analyse the transcription profile of all 10,000 cells and overlay the unbiased genotype information of 2,000 cells and the targeted genotype information of 8,000 cells (see Figure).

Demonstrating the transcriptional changes potentially mediating relapse

Our single-cell multiomics analyses revealed that the emergence of venetoclax relapse in CLL is highly heterogeneous, with multiple co-existing changes present even in a single sample⁵. Surprisingly, besides the sub-clonal BCL2 mutation or MCL1 amplification, we did not discover many novel mutations or isoform usages contributing to venetoclax resistance. We demonstrated complexity of transcriptional changes that mediate venetoclax relapse CLL, even in the absence of obvious genetic changes.

The complexity of the observed changes might suggest that venetoclax resistance will prove impossible to treat clinically. However, in yet another unique aspect of our study, we provided compelling evidence that the changes were mainly sustained by ongoing venetoclax therapy. So, our study gives strong support and a rationale for limited duration venetoclax therapy in CLL, an approach now gaining favour in the clinic.

We are 'another step closer' to resolving the genetic complexities of cancer

This work was only possible because we took an innovative multiomics approach on matched samples. By integrating short-read with full-length nanopore sequencing of the same transcripts, we could link genotype, transcriptomics, cancer cell state and alternative splicing. Our approach in this disease provides a guide for future studies to use high-throughput single-cell applications (e.g., single-cell ATAC-sequencing) with full-length transcriptomic data to fully resolve tumour plasticity and heterogeneity underpinning resistance to other cancer therapeutics. RaCH-seq could be adapted to your genes of interest and applied to already stored indexed cDNA. It would be interesting to add another layer of information to single-cell spatial omics such as isoform usage and mutation status with the use of RaCH-seq. By integrating long nanopore sequencing reads to already established high-throughput single-cell technologies, we are another step closer to resolving cancer complexity and improving patient outcomes.

(Images: Left: Rachel with Stephen Wilcox, Jafar Jabbari, Angela Georgiou, Hongke Peng. Right: Rachel with Luyi Tian.)