A robust metatranscriptomic technology for population-scale studies of diet, gut microbiome, and human health Andrew Hatch, James Horne, Ryan Toma, Brittany L. Twibell, Kalie M. Somerville, Benjamin Pelle, Kinga P. Canfield, Guruduth Banavar, Ally Perlina, Helen Messier, Niels Klitgord and Momchilo Vuyisich* Viome, Inc, Los Alamos, NM 87544, United States. *Correspondence should be addressed to Momchilo Vuyisich; [email protected] ​ Abstract The gut microbiome plays a major role in many chronic diseases. It is the only “organ” in our body whose function can easily be manipulated by our diet and lifestyle. Using personalized nutrition should enable precise control of the gut microbiome’s functions that support human health and prevent chronic diseases. Without a functional readout of the gut microbiome, however, personalized nutrition cannot be realized. Stool metatranscriptomic analysis offers a comprehensive functional view of the gut microbiome. Despite its usefulness, metatranscriptomics has seen very limited use in clinical studies due to its complexity, cost, and bioinformatic challenges associated with both microbial taxonomy and functions. This method has also received some criticism due to potential intra-sample variability, rapid changes, and RNA degradation. Here, we describe a robust and automated stool metatranscriptomic method, Viomega. This method was specifically developed for population-scale studies on the effects of gut microbiome on human health and disease, with the goal to develop personalized nutrition algorithms. Viomega includes sample collection, ambient temperature sample preservation, total RNA extraction, physical removal of ribosomal RNAs (rRNAs), preparation of directional Illumina libraries, Illumina sequencing, taxonomic classification based on a database of >110,000 microbial genomes, and quantitative microbial gene expression analysis using a database of ~100 million microbial genes. In this report, we demonstrate the robustness of Viomega technology. We also applied the method to 10,000 human stool samples and report the taxonomic and functional data. Finally, we performed several small clinical studies to demonstrate the connections between diet and the gut metatranscriptome. 1 Introduction The human gut contains a vast number of commensal microorganisms performing a wide variety of metabolic functions. Metabolites produced by these microorganisms can have profound effects on human physiology, with direct links to our health and disease [1-4]. Gut dysbiosis likely contributes to the development and progression of many diseases and disorders, such as cardiovascular disease, hypertension, obesity, diabetes, and autoimmune diseases [5-9]. There is also strong evidence that the gut microorganisms directly interact with the nervous system, establishing the gut-brain axis [10]. The gut-brain axis has been shown to modulate the development of neurodegenerative diseases such as Alzheimer’s, Autism Spectrum Disorder (ASD) and Parkinson’s [11-14]. The gut microbiome plays a critical role in physiological homeostasis, and there has been increasing scientific investigation into the extent of the role of the human gut microbiome in human health and disease. Several next generation sequencing-based methods are used for analyzing the gut microbiome, each with their own advantages and drawbacks (Table 1). The simplest, least expensive, and most common is 16S gene sequencing [15], which sequences a small portion of the highly conserved prokaryotic 16S ribosomal RNA gene [16]. The method provides low taxonomic resolution, limited to genus level at best [17,18], and does not measure the biochemical functions of the microorganisms [16]. In addition, it misses some bacteria, most archaea, and all eukaryotic organisms and viruses [19]. Metagenomic (shotgun DNA) sequencing provides strain-level resolution of all DNA-based microorganisms [16] (it does not detect RNA viruses or RNA bacteriophages). However, it only identifies the potential biochemical functions of the microbiome, and can ​ ​ neither identify nor quantify the active biochemical pathways. This is a disadvantage for studying dysbiosis related to disease states, such as inflammatory bowel disease (IBD), which has been shown to have a disparity between metagenomic potential pathways and actual biochemical pathways expressed in disease and control populations [20]. Humans have co-evolved with the microbiome and become dependent on its biochemical output, such as certain vitamins and short chain fatty acids [21,22]. In addition, the gut microbiome can produce harmful biochemicals which have been implicated in various disease states [21]. To fully understand the relationships between the gut microbiome and human health and disease, biochemical functions of the microorganisms must be identified and quantified. Metatranscriptomic analysis (metatranscriptomics, RNA sequencing, RNAseq) offers a deep and accurate insight into the biochemical activities of the gut microbiome, while also providing strain-level taxonomic resolution [23,24]. Metatranscriptomic analyses of stool samples have been limited due to the cost and complexity of both laboratory and bioinformatic methods [25], though with the addition of protocols to remove less informative rRNA, more valuable transcriptome data can be generated with less sequencing depth [26,24], helping to ​ ​ reduce per sample sequencing costs. We have developed an automated technology for metatranscriptomic analysis of human clinical samples, called Viomega. 2 Table 1: Comparison of three next generation sequencing-based methods for microbiome analysis 16S Gene Sequencing Metagenomics Metatranscriptomics Identifies most Does not identify all bacteria. microorganisms to strain Identifies few archaea. Does not Which microbes level (cannot identify any identify any viruses (including Identifies all living can be RNA viruses/ RNA phages) or eukaryotes (including organisms. identified? bacteriophages). Cannot parasites). Cannot distinguish live distinguish live from dead from dead organisms. microorganisms. Low resolution (typically limited to High resolution (strain High resolution (strain Resolution genus level at best) level) level) Heavily influenced by amplification Sample lysis and depletion Method Biases method, but also sequencing quality, Sample lysis (if used) sample lysis, and bioinformatics Functional Gene Quantifies expression Does not measure expressed gene Does not measure expressed Expression levels of active microbial functions gene functions Analysis gene functions Allows assessment of pathway activities that can Actionable Does not allow to assess pathway Does not allow to assess lead to personalized health Pathway activities pathway activities insights and Analysis recommendations with molecular-level precision In this study we applied Viomega to 10,000 human stool samples to gain a better understanding of the strain-level taxonomies and functions. We also performed several small clinical studies to quantify the metatranscriptomic stability in the lower colon and measure the intra-sample variability of metatranscriptomic analyses. Materials and Methods Study participants, ethics, and sample collection and transportation For this study, Viome used data from 10,000 participants. All study participants consented to being in the study, and all study procedures were approved by an IRB. Participants were recruited from any age, gender, and ethnic group. 3 Stool samples were collected by the study participants at their own residences using the Viome Gut Intelligence kit. The kit included a sample collection tube with an integrated scoop, a proprietary RNA preservative, and sterile glass beads. A pea-sized stool sample was collected and placed inside the tube. The tube was vigorously shaken to homogenize the sample and expose it to the RNA preservative. The sample was then shipped at room temperature using a common courier to Viome labs for analysis. Shipping times ranged from 1-12 days. Besides stool samples, each participant filled out a questionnaire with general lifestyle and health information. Metatranscriptomic analysis of stool samples For metatranscriptomic analysis, Viome uses a proprietary sample-to-result automated platform called Viomega. Stool samples are lysed using bead-beating in a strong chemical denaturant. Following sample lysis, the samples are placed on an automated liquid handler, which performs all downstream laboratory methods. RNA is extracted using silica-coated beads and eluted in water. DNA is degraded using RNase-free DNase, which is then inactivated with EDTA and heat. RNA is quality controlled by quantifying with RiboGreen (ThermoFisher) and assessing its integrity with the Fragment Analyzer (Advanced Analytical). Majority of prokaryotic ribosomal RNAs (rRNAs: 16S and 23S) are removed using a custom subtractive hybridization method. Biotinylated DNA probes with sequences complementary to rRNAs are added to total RNA, the mixture is heated and cooled, and the probe-rRNA complexes are removed using magnetic streptavidin beads. The remaining RNAs are converted to directional sequencing libraries with unique dual-barcoded adapters and ultra-pure reagents. Ninety four human stool samples, a negative process control (NPC, water) and a positive process control (PPC, custom synthetic RNA) samples are processed on a 96-well microplate. Libraries are pooled and quality controlled with dsDNA Qubit (ThermoFisher) and Fragment Analyzer (Advanced Analytical). Library pools are sequenced on