Optimal Dual RNA-Seq Mapping for Accurate Pathogen Detection in Complex Eukaryotic Hosts

Background

Dual RNA sequencing is a powerful tool to precisely investigate the complete gene expression profile of two actively interacting organisms. When a pathogen infects a host, both develop adaptation mechanisms by revealing a series of changes at molecular levels. These biological changes are interrelated and can be effectively elucidated by analyzing the whole transcriptome data of the host–pathogen interaction. Until 2012, transcriptomics sequencing of infectious diseases was limited to separately capturing the cascade of biological events in either the host or the pathogen [1]. This allows us to study gene expression changes only from a linear perspective. Over the years, insightful research in transcriptomics analysis has encouraged researchers across the globe to devise the concept of extracting and sequencing whole RNA from the infection site. This idea will provide a broader scope for unraveling the role of smaller molecules and studying the biological changes in organisms at different stages of interaction. Compared to conventional RNA sequencing technology, dual RNA sequencing caters to a sufficient amount of host and pathogen RNA and other small RNA molecules (microRNA, long non-coding RNA, and other non-coding RNA) when pathogen-infected samples are subjected to the well-established and published methods of RNA extraction and sequencing procedures [2–5].

Dual sequencing data generated as reads or fragments comprises information on both the host and the pathogen. The infected host cells may not always reflect pathogen-related data in enormous quantities. Therefore, a detail-oriented analysis is necessary to capture the minimally found essential information on host–pathogen interplay. Several bioinformatics analysis methods and tools are available to explore the complex datasets involving eukaryotic and prokaryotic reads [6–8]. Most of these methods suggest a sequential mapping approach to extract host and pathogen reads. Some studies recommend a combined mapping approach where the reference genomes of the host and pathogen are concatenated, indexed, and used as a single reference [9,10]. As far as quality control and adapter removal are concerned, standard methods for read mapping have been receiving greater attention as pathogen data is found in lesser quantities. Espindula et al. [10] conducted studies on different infection models and showed that other alternative mapping approaches outperform the traditional host-first mapping approach. When trimmed reads are mapped first to the host reference, there are high chances of pathogen read mismapping due to its shorter read length. To avoid this, alternative mapping approaches were introduced, and this bio-protocol emphasizes one such mapping technique—the pathogen-first mapping approach—in detail. A comparison of mapping results proved that most of the pathogen reads have been restored in the pathogen-first mapping approach. The results obtained using this approach are now based on a higher confidence scale and can be used for further processing.

This protocol is presented with the aim of demonstrating a standardized bioinformatics procedure for productive mapping. In this protocol, human monocyte-derived macrophages (HMDMs) infected with Mycobacterium tuberculosis (Mtb) were used as a dual RNA-Seq model. Apart from humans infected with Mtb, other host and pathogen models can still be used in dual RNA-Seq experiments. For example, other dual RNA-Seq datasets are publicly available in Gene Expression Omnibus (GEO); one such dataset is Triticum aestivum infected with Fusarium graminearum (Accession: SRP439529; GEO: GSE233409). Dual RNA-Seq test dataset of human–Mtb was downloaded from Sequence Read Archive (SRA), National Centre for Biotechnology Information (Accession: SRP359986 [11]). This study primarily focuses on a compound that restricts Mycobacterium tuberculosis from catabolizing cholesterol by binding with iron. The data quality control before and after read trimming was performed using FastQC. The FastQC tool can be accessed online at https://www.bioinformatics.babraham.ac.uk/projects/fastqc/. For trimming low-quality bases and adapter removal, benchmarking was performed using the topmost adapter trimming software, namely fastp [12] and Trim-Galore (https://github.com/FelixKrueger/TrimGalore). Following this, the topmost splice-aware alignment tools STAR [13] and HISAT2 [14] were also benchmarked to obtain optimal results. The most important part of the analysis is to apply the best mapping method that serves the experimental purpose of dual RNA-Seq analysis. After mapping the reads to their respective genomes, the reads mapping to genes were quantified using featureCounts [15]. Other downstream analysis methods, which include read count normalization, differential expression analysis, gene ontology, and pathway enrichment, are not demonstrated here, as the key aim of this protocol is to highlight the crucial preparatory steps like quality control, adapter removal, and read mapping in a descriptive manner.