mRNA vaccine quality analysis using RNA sequencing

Abstract

The success of mRNA vaccines has been realised, in part, by advances in manufacturing that enabled billions of doses to be produced at sufficient quality and safety. However, mRNA vaccines must be rigorously analysed to measure their integrity and detect contaminants that reduce their effectiveness and induce side-effects. Currently, mRNA vaccines and therapies are analysed using a range of time-consuming and costly methods. Here we describe a streamlined method to analyse mRNA vaccines and therapies using long-read nanopore sequencing. Compared to other industry-standard techniques, VAX-seq can comprehensively measure key mRNA vaccine quality attributes, including sequence, length, integrity, and purity. We also show how direct RNA sequencing can analyse mRNA chemistry, including the detection of nucleoside modifications. To support this approach, we provide supporting software to automatically report on mRNA and plasmid template quality and integrity. Given these advantages, we anticipate that RNA sequencing methods, such as VAX-seq, will become central to the development and manufacture of mRNA drugs.

Fig. 1. mRNA vaccine production and VAX-seq workflow.

Schematic diagram illustrates the steps during mRNA manufacture (left panel), and the steps during VAX-seq analysis (right panel). This includes laboratory steps of long-read nanopore sequencing, followed by bioinformatic steps to analyse output data, including the supporting Mana software toolkit. mRNA vaccine quality features that can be analysed by VAX-seq are indicated (listed in red and green). In the bottom left corner is an IGV plot comparing Oxford Nanopore and Illumina sequencing of a plasmid DNA template. Coverage indicates the number of reads at each nucleotide position while the lower alignments grey bars indicate unique, individual alignments, with colouring indicating their similarity to the reference genome. Source data are provided as a Source data file.

Fig. 2. Analysis of reference eGFP mRNA vaccine using long-read Oxford Nanopore sequencing (PCS111).

a Genome-browser (IGV) view of long-read cDNA alignments to the reference plasmid sequence. Coverage indicates the number of reads at each nucleotide position while the lower alignments grey bars indicate unique, individual alignments, with colouring indicating their similarity to the reference genome. b Sequencing error and type profile across mRNA vaccine and poly(A) tail sequences. c Detail shows sequencing coverage and error at the poly(A) tail, showing a characteristic m-shaped deletion profile. d mRNA length as measured using ONT full-length sequencing shows full-length and fragmented mRNA vaccines. e poly(A) tail length measured using tailfindr for eGFP mRNA (green), compared to a cDNA library with variable poly(A) tail lengths. Source data are provided as a Source data file.

Fig. 3. Analysis of reference eGFP mRNA vaccine using short-read Illumina sequencing (TruSeq).

a Genome-browser (IGV) view of short-read cDNA sequenced libraries aligned to the reference plasmid sequence. Coverage indicates the alignment depth at each nucleotide position, while the lower alignments grey bars indicate unique, individual alignments, with colouring indicating their similarity to the reference genome. b Detail of transcription start site shows alignment of long- and short-read sequencing at 5’ end of mRNA vaccine. c Detail genome browser view of poly(A) tail shows uneven coverage and deletion of linker sequence due to short-read misalignment. Source data are provided as a Source data file.

Fig. 4. Direct RNA sequencing of reference eGFP mRNA vaccine.

a Genome-browser (IGV) view of direct RNA sequencing alignments to the reference sequence. Coverage indicates the number of reads at each nucleoside position while the lower alignments grey bars indicate unique, individual alignments, with colouring indicating their similarity to the reference genome. b Direct RNA sequencing shows error type and frequencies across the mRNA vaccine and poly(A) tail sequence. c Plot shows the length of mRNA vaccine as measured from direct RNA sequencing, showing additional smaller peak resulting from artefactual trimming of poly(A) tail from reads. d Poly(A) tail length measured using tailfindr across three technical replicate direct RNA sequencing libraries. e Schematic diagram shows the different mRNA species, fragment size and contaminations identified by the VAX-seq workflow. Source data are provided as a Source data file.

Fig. 5. Analysis of reference eGFP mRNA vaccine incorporating modified nucleosides using direct RNA sequencing.

a, b Genome-browser (IGV) view of long-read (ONT) alignment to the plasmid reference, sequenced using direct RNA sequencing for mRNA vaccines prepared with a Uridine and b N1 methylpseudouridine. Coverage indicates the number of reads at each nucleoside position while the lower alignments grey bars indicate unique, individual alignments, with colouring indicating their similarity to the reference genome. Heterogenous coverage is observed in the direct RNA sequencing, likely due to fragmentation of the modified mRNA vaccine. c mRNA length analyses demonstrate shorter length for modified mRNA vaccines due to fragmentation and enrichment in deletion sequencing errors (n = 4). d Direct RNA sequencing shows stereotypical errors (cytosine, blue; uridine; red) at N1-methylpseudouridine compared to uridine nucleosides. e Cumulative distribution plot shows per-nucleoside error profile for N1-methylpseudouridine (red/orange) compared to unmodified nucleosides (blue/green). Source data are provided as a Source data file.