Comparing Published Gut Microbiome Taxonomic Data Across Multinational Studies

Abstract

Background:

Nurse researchers are well poised to study the connection of the microbiome to health and disease. Evaluating published microbiome results can assist with study design and hypothesis generation.

Objectives:

This manuscript aims to present and define important analysis considerations in microbiome study planning and identify genera shared across studies despite methodological differences. This methods manuscript will highlight a workflow that the nurse scientist can use to combine and evaluate taxonomy tables for microbiome study or research proposal planning.

Methods:

We compiled taxonomy tables from 13 published gut microbiome studies that had used Ion Torrent sequencing technology. We searched for studies that had amplified multiple hypervariable (V) regions of the 16S rRNA gene when sequencing the bacteria from healthy gut samples.

Results:

We obtained 15 taxonomy tables from the 13 studies, comprised samples from four continents and eight V regions. Methodology among studies was highly variable, including differences in V regions amplified, geographic location, and population demographics. Nevertheless, of the 354 total genera identified from the 15 data sets, 25 were shared in all V regions and the four continents. When relative abundance differences across the V regions were compared, Dorea and Roseburia were statistically different. Taxonomy tables from Asian subjects had increased average abundances of Prevotella and lowered abundances of Bacteroides compared to the European, North American, and South American study subjects.

Discussion:

Evaluating taxonomy tables from previously published literature is essential for study planning. The genera found from different V regions and continents highlight geography and V region as important variables to consider in microbiome study design. The 25 shared genera across the various studies may represent genera commonly found in healthy gut microbiomes. Understanding the factors that may affect the results from a variety of microbiome studies will allow nurse scientists to plan research proposals in an informed manner. This work presents a valuable framework for future cross-study comparisons conducted across the globe.

The gut microbiome is the community of bacteria and other microbes that reside in and coexist with the intestinal tracts of mammals. Its homeostasis is essential for normal physiology and health (Ursell et al., 2012). Bacteria facilitate vital functions in the gut, including fermenting carbohydrates into neurotransmitters or short-chain fatty acids, which are important for colon and systemic health (Ferreira-Halder et al., 2017). Changes in the gut microbiome have been linked to health conditions in clinical and preclinical research including alcohol abuse (Ames et al., 2020), increased blood pressure (Maki et al., 2020), and colon cancer (Dai et al., 2018). Nurse scientists are well equipped to study the complex relationships between the gut microbiome and human physiology due to nurses’ clinical foundation in a holistic approach to understanding human health and disease (Maki et al., 2021). Incorporating associations between gut microbial communities and human health/disease in nursing science may facilitate mechanistic understanding of patient conditions and inform research aimed at improving clinical outcomes.

Nurses and nurse scientists are also involved in global health research and initiatives, forging partnerships worldwide to advance care delivery. They have potential to validate if microbiome-host interactions with health and disease are consistent across varying continents, economies, and lifestyles. The gut microbiome composition is unique and specific to the individual and can respond to changes in the individual’s health status and environment (David et al., 2014). Due to its intersection with diet, genetics, lifestyle, and ethnicity, geography has been shown to substantially affect which bacterial genera dominate the guts of specific populations (Deschasaux et al., 2018; Panek et al., 2018). One study found that the genus Prevotella was dominant in non-Western, rural populations with plant-based diets; the genus Bacteroides was dominant in Western populations that had higher meat and fat contents in their diets (Deschasaux et al., 2018).

To conduct a microbiome study, researchers must plan to collect samples, store those samples, extract genomic deoxyribonucleic acid (DNA), amplify the 16S ribosomal ribonucleic acid (rRNA) gene (for 16S rRNA gene amplicon sequencing), sequence the amplified DNA, and use bioinformatics processes to identify sequences and classify them to a reference taxonomy database (Barb et al., 2016; Maki et al., 2019; Sinha et al., 2015; see Supplemental Digital Content [SDC] 1 for a list of common microbiome analysis definitions). There is variation in methodology for each of these steps, so informed study and analysis planning are important for the nurse scientist undertaking microbiome research; these factors can lead to highly different results within the field of microbiome research (Maki et al., 2019). One step that can particularly bias results is the 16S rRNA gene amplification step and the primers used to target gene regions for amplification (D’Amore et al., 2016). The 16S rRNA gene is used in research to characterize bacteria in a microbiome sample. Nine regions of the 16S rRNA gene have varying sequences between different levels of taxonomy (i.e., genus or species level), and by targeting these hypervariable (V) regions, researchers can categorize bacteria by comparing the specific sequences to a reference database (Maki et al., 2019; see SDC2 for an overview of the 16S rRNA gene and primers). Most microbiome studies amplify only one or two regions of the gene (Callahan et al., 2016), frequently including the V3 or V4 regions (Deschasaux et al., 2018; Gorvitovskaia et al., 2016). However, each region may present bias in identifying accurate bacterial composition (Barb et al., 2016; Fouhy et al., 2016; Maki et al., 2019), so selecting the V region is an essential methodological choice. There are also different types of sequencing platform technologies, including Illumina, Ion Torrent (IT), and Pacific Biosciences. While Illumina is currently the most commonly used sequencing platform, Ion Torrent uses a different sequencing technology and has a sequencing kit that creates amplicons from several V regions, including V 2-4-8 and V3-6, 7-9 (see SDC 3 for an overview of Illumina vs. IT sequencing technology).

Although there is variability between different important study design variables such as sequencing platform and primers used, population sampled, and geographic location, previous research suggests there is a core microbiome, a subset of gut microbes that are present in most healthy individuals, regardless of geography and diet (Sekelja et al., 2011; Ursell et al., 2012). Multiple large-scale studies have characterized the healthy human core microbiome, including the Human Microbiome Project (HMP; Human Microbiome Project Consortium, 2012) and the American Gut Project (McDonald et al., 2018); these studies have provided an extensive repository of benchmark data sets that can be used for interrogating this idea. Where geography can affect abundances of taxa in specific populations, the core microbiome concept posits certain genera to be present in all healthy populations.

A feasible workflow to combine data from previously published studies is beneficial to nurse scientists for study, grant planning, and hypothesis generation. Synthesizing results from different studies would ideally have consistent methodology to reduce the forms of bias explained earlier. However, this is not always feasible due to lack of bioinformatics support, ability to obtain raw sequence data from published literature, and other limiting resource constraints. Because of these obstacles, it may be desirable for an investigator to compare taxonomy tables from previously published microbiome research at the genus level. Therefore, this paper aims to present and define important analysis considerations in microbiome study planning and identify genera shared across studies despite methodological differences. We highlight a workflow that nurse scientists can use to combine and evaluate taxonomy tables for microbiome study or research proposal planning.

Methods

Outline of Workflow

The framework for this manuscript workflow comprises a four-step process (Figure 1; SDC4). Step 1 is to obtain the data sets to be used for the analysis. Step 2 is to merge the data at the genus level to create one Combined Taxonomy Table representing the union across all combined data sets. Step 3 is to investigate the overlap of taxa by two differing study parameters: V region and Continent. Step 4 is to compare the Combined Taxonomy Table from this framework to a publicly available “gold standard” gut microbiome data set to investigate congruence.

Figure 1. Stepwise Workflow of Data Accrual and Manipulation Process.

Text in italics are results or output from example workflow. Abbreviations: IT = Ion Torrent; HMP = Human Microbiome Project; SD = Standard Deviation; V = hypervariable region. Figure created with Biorender.

Obtain the Data Sets: Microbiome Taxonomic Table Identification and Acquisition

Searches were carried out across two databases: National Center for Biotechnology Information PubMed and Google Scholar for manuscripts published between 2013 and 2019. The following search terms were entered into Google Scholar: “stool” OR “feces” OR “fecal” AND “microbiome” OR “microbiota” AND “Ion Torrent” AND “s5” OR “metagenomics kit” OR “v3” OR “v2” OR “v4” OR “v67” OR “v9” OR “v8” OR “v1” AND “human” OR “patient” OR “participant” OR “subject.”

Title screening excluded environmental studies, animal studies, mycobiome or virome studies, papers not written in English, review papers, and dissertations or theses. Inclusion criteria were: healthy human, nonpregnant adults (≥ 18 years of age); gut microbiome studies; and stool samples. For each project, a thorough methods review was done to verify the inclusion of healthy control subjects defined by various measures, including self-report, absence of any chronic medical conditions, antibiotic use restrictions, and body mass index thresholds. In practice, the nurse researcher should carefully evaluate the metadata of included microbiome studies to ensure the sampling represents the population and/or conditions of interest.

Of the 36 studies that passed the first round of review, five reported their unfiltered, complete taxonomy tables as supplementary materials and therefore were directly downloaded for this work (Bhute et al., 2016; Chávez-Carbajal et al., 2019; Hevia et al., 2016; Hoarau et al., 2016; Ribeiro et al., 2018). For the remaining 31 studies, emails were sent to the corresponding or first authors requesting the raw operational taxonomic unit (OTU) tables for their healthy control samples. Eight authors agreed to send taxonomy/OTU tables in either Microsoft Excel or text file format (Dadkhah et al., 2019; Hernández-Quiroz et al., 2020; Hidalgo-Cantabrana et al., 2019; Rajakaruna et al., 2019; Rodriguez et al., 2019; Senan et al., 2015; Suryavanshi et al., 2016; Zeber-Lubecka et al., 2016). Metadata for each project were collected and included in this analysis. All projects but one reported microbiome data in the form of OTUs (see SDC1 for a list of microbiome analysis definitions). The one project that did not report results as OTUs reported results as amplicon sequence variants (Hidalgo-Cantabrana et al., 2019). For the purposes of this work, we will refer to all grouped bacterial sequences as OTUs and bacterial abundance tables/biological observation matrices as taxonomy tables for consistency.

Creating One Master Data Set: Data Manipulation, Merging, and Filtering

Once individual taxonomy tables were acquired by authors or through direct download (Figure 1), data were imported into JMP^®, version 14 SAS Institute Inc., Cary, NC, 1989–2021, used for data discovery and manipulation merging, summarizing, and figure generation. Individual data sets came in one of two forms from each author/manuscript: (a) a taxonomy table in the form of raw OTU counts (either average counts across all healthy controls or counts for each healthy control included); or (b) a taxonomy table with relative abundance values calculated for each OTU/taxon (either average relative abundance over all healthy controls or a relative abundance for each healthy control included).

As 16S rRNA gene amplicon sequencing most reliably annotates to the taxonomic level of genus (Nguyen et al., 2016), all higher-resolution taxonomy (species- and OTU-level taxa) were summarized to the genus level (also classified as taxonomy level L6 in some data sets). Any OTU mapping to the same genus were added together (Figure 1). Genus-level counts were converted to relative abundances by dividing an individual genus count by the total counts over all genera in the sample. When relative abundance values were already calculated for a given data set (option two above), the average relative abundance was calculated to the same genus. The goal was to obtain an average genus-level taxon relative abundance for each individual data set/taxonomy table (Figure 1). Finally, data sets were merged using the genus as the unique identifier for joining (Figure 1 and SDC1, definition “merge”). We joined each taxonomy table in a stepwise process until the genera for all 15 individual taxonomy tables were together in one table, referred to as the Combined IT Taxonomy Table. To remove very low or sparsely assigned taxa, all genera must have had an average abundance of ≥ 0.01% in at least one of the 15 included data sets to be used for further investigation. Results are listed with mean and standard deviation, except for South America (which only had one study).

Investigate Shared Genera by Study Metadata: V Region and/or Continent

Overlap of classified and filtered genera was examined by V region or continent using Venn diagram analysis. In practice, the researcher should select metadata variables pertinent for the planned study (i.e., percent female, disease/disorder of interest, etc.). As genera are merged for each data set, metadata should be represented as one variable per data set/taxonomy table as opposed to the usual one metadata variable per sample.

In our workflow, for the V region overlap, we assigned individual projects a specific V region based on that project’s primer used for amplicon sequencing (Table 1). A genus was assigned present by a particular V region if that targeted V region primer generated the data set. Venn diagram analyses investigating overlap were based on presence/absence of genera and not on abundance information. The two data sets from Zeber-Lubecka et al. (2016) were not used for the primer-specific V region Venn diagram analysis as they used the Ion 16S™ Metagenomics Kit (ThermoFisher Scientific, Waltham, MA) which amplified 7 out of 9 V regions (V2, V3, V4, V6-V7, V8, and V9). The data sets provided by Zeber-Lubecka et al. (2016) will be referred to in this workflow as the “IT Kit” when necessary. A Venn diagram across 4 V regions (V12, V23, V3, V4) was used to visualize unique and overlapping genera by V regions. Note: V12 and V23 indicate that the studies amplified the V1 and V2 or V2 and V3 regions, respectively.

Table 1.

Overview of Metadata from Included Studies

First author, year	Sample size (controls)	Continent	Country	Storage	Extraction	V region	Pipeline	Database	Demographics: Age Range, % Female
Bhute et al., 2016 ε	34	Asia	India	−80	QIAamp DNA Mini Kit	3	QIIME, UCLUST	Greengenes	Not provided in manuscript
Chávez-Carbajal et al., 2019 ε	25	North America	Mexico	aliquot, −78	ZR Fecal DNA Mini Prep	3	QIIME	Greengenes	18-51 years, 100%
Dadkhah et al., 2019 ©	53	North America	USA	−20, RNALater	FastDNA Spin Kit for Soil	12	UPARSE	RDP	62 (average), 40%
Hernández-Quiroz et al., 2020 ©	33	North America	Mexico	collected at home, 2 mL aliquot, −70	Favor prep stool kit	3	QIIME, PICRUST	Greengenes	21-53 years, 40%
Hevia et al., 2016 ε	22	Europe	Spain	Homogenized in RNALater, −80	QIAamp DNA Mini Kit	3	QIIME	RDP	39 (average), 68%
Hidalgo-Cantabrana et al., 2019 ©	20	Europe	Spain	−20	QIAamp DNA Mini Kit	23	QIIME	SILVA	43 (average), 75%
Hoarau et al., 2016 ε	28	Europe	France	Homogenized, −30, then −80	QIAMP DNA Mini Kit	4	QIIME, UCLUST	UNITE 5.8S	45 (average)
Rajakaruna et al., 2019 ©	20	North America	USA	OMR-200	ZR Fecal DNA MiniPrep kit	12	QIIME	RDP	29-66 years,
Rajakaruna et al., 2019 ©	20	North America	USA	OMR-200	ZR Fecal DNA MiniPrep kit	4	QIIME	RDP	29-66 years,
Ribeiro et al., 2018 ε	10	South America	Brazil	Tested 5 different kinds: −80, −20, RT, 4C, HCl-EDTA	Power soil DNA Isolation kit	4	QIIME, UCLUST, UPARSE	Greengenes	23-49 years, 50%
Rodriguez et al., 2019 ©	3	North America	USA	Collected at home, −80	QIAamp DNA Mini Kit	3	QIIME, UCLUST	Greengenes	29-65 years, 0%
Senan et al., 2015 ©	8	Asia	India	−20	QIAamp DNA Mini Kit	23	QIIME	BLASTN	64-74 years
Suryavanshi et al., 2016 ©	15	Asia	India	−80	QIAamp DNA Mini Kit	3	Mothur, QIIME	Greengenes	22-52 years, 0%
Zeber-Lubecka et al., 2016 ©	29	Europe	Poland	−20, aliquots	QIAamp DNA Mini Kit	2483679	Mothur	Greengenes	40 (average), 66%
Zeber-Lubecka et al., 2016 ©	29	Europe	Poland	−20, aliquots	QIAamp	2483679	Mothur	SILVA	40 (average), 66%

Note. The 13 data sets included in this manuscript organized by their first author (as cited in the paper) with the sample size and methodological differences including storage, DNA extraction method, pipeline, and database. BLASTN = Basic Local Alignment Search Tool Nucleotide; HCl-EDTA = Hydrochloric acid- Ethylenediaminetetraacetic acid; PICRUST = Phylogenetic Investigation of Communities by Reconstruction of Unobserved States; QIIME = Quantitative Insights into Microbial Ecology; RDP = Ribosomal Database Project; UNITE 5.8S = a database used more commonly for identifying eukaryotic species; this paper also studied the mycobiome in addition to bacterial genera.

^ε= data obtained from supplement

^©data obtained directly from author

We repeated the same process of Venn diagram analysis, but instead, we used study continent of residence as the grouping variable (Table 1). A Venn diagram was created to determine unique and shared/overlapping genera by continent. All data sets were assigned to one of four continents represented across the 15 data sets. We cross-compared the union between genera shared across all V regions and genera shared across all continents and used this set of genera for further relative abundance bar plots.

Compare Combined IT Table to a Benchmark Publicly Available Data Set

To evaluate agreement with our Combined IT Taxonomy Table with an established benchmark microbiome data set from healthy gut samples, we compared relative abundances between the Combined IT Table and gut microbiome data from the HMP. In our workflow, data in the form of OTU taxonomy tables from the HMP targeting both the V13 and V35 regions from human gut samples were obtained from the data portal HMPdacc (https://www.hmpdacc.org/hmp/HMQCP/). The HMP consortium provided two sets of gut microbiome OTU tables (from regions V13 and V35); therefore, we were limited to interrogate a benchmark data set generated from these two V regions. All data used in the HMP were generated from healthy individuals from the North American continent (HMP, 2012). A total of 175 primary sample numbers were included in the V13 table, and 309 primary sample numbers were included in the V35 table. Genus level counts were obtained for both tables by summing over OTUs mapped to the same genus. The two HMP data sets were merged by genus, and the relative abundance was calculated for each sample. The average relative abundance over all samples in HMP was calculated and joined to the current IT table. Bivariate plots were created to visualize the level of agreement between the HMP data set and the data sets representing the four continents in the Combined IT Taxonomy table.

Statistical Analysis

Average relative abundances for each genus were compared between V region groups (V12, V23, V3, V4) and between continents (Asia, Europe, North America, and South America) using a Kruskal-Wallis test. Correlations between logarithm base 10 transformed average abundances of individual genera between the four continents, and the HMP data were assessed using Pearson’s correlation analysis.

Results

We present a workflow using taxonomy table data from 13 multinational gut microbiome studies published from 2013–2019 that analyzed stool samples from healthy individuals using IT next-generation sequencing technology (Table 1). Fifteen individual data sets from these 13 studies were collapsed at the taxonomic level of genus and filtered for downstream analysis. Before filtering, a total of 809 genera were included in the Combined IT Taxonomy Table.

Data Overview

The 15 data sets included in this framework used IT sequencing technology. Still, they differed in the following ways: (a) the V region primer used, (b) the number of gut samples in each data set, (c) the bioinformatics processing pipeline used, and (d) the geographic location from which the samples were comprised (Table 1). The combined IT table merging 15 data sets represented microbial communities from 300 healthy participants, but the number of participants from each study varied considerably and ranged from 3–53 subjects/study. Individual data sets represented participants from four continents—Asia, Europe, North America, and South America—and seven countries (Table 1); they covered eight of the nine V regions across the 16S rRNA gene (except V5). Six of the 15 data sets used the V3 primer, 3 of 15 data sets used the V4 primer, two data sets used the V12 primer, and two data sets used the V23 primer. One data set used a multi-V region amplification kit (the IT Kit) interrogating 7 and 9 V regions of the 16S rRNA gene. A total of 809 genera were included in the combined data set across all studies. After applying the 0.01% relative abundance filter, 354 genus-level taxa remained.

Comparison of Genera Across V Regions and Geographic Location

To explore differences across data sets by V region, we looked at unique and overlapping genera across studies that used different V region primers. There were 56 genera shared across the four groups of V regions (V12, V23, V3, and V4; SDC5A). This does not include the two studies which used the multi-V amplicon IT Kit (Figure 2A). Data sets from regions V12 and V4 had the highest number of shared genera (37 total; SDC5B), and V12 had the highest number of unique genera (62 total; SDC5C). To explore shared gut bacteria despite geographic differences, we investigated unique and overlapping genera by continent (Figure 2B). There were 26 genera shared across all four continents represented in this study sample (SDC5D).

Figure 2. Venn Diagram by V region and Continent.

A.) Four-way Venn Diagram of different V region combinations (i.e., V12, V23, V3, V4, and excluding the IT Kit). Overlap of any portion of the ovals indicates shared genera. Numbers within the union indicates how many genera were shared between V regions labeled, or the intersection indicates how many were unique to that region. Numbers within the Venn Diagram represent those 13 data sets that amplified regions V1-V4. The number outside of the Venn Diagram are those genera found unique to the ITKit. B.) Venn diagram showing the comparisons for each of the four continents surveyed in this study. C.) Venn diagram of the 25 genera found intersecting all V regions and all continents. The number outside of the Venn are the number genera not found on all V regions and Continents. Abbreviations: N. America=North America; S. America=South America.

In the 56 genera shared across V regions and the 26 genera shared across continents, there was an overlap of 25 genera shared irrespective of the V region or continent studied (Figure 2C, SDC5E). In these shared genera, across all samples Bacteroides, Prevotella, and Faecalibacterium had the highest average relative abundances at M = 14.7%, SD = 14.8, M = 8.5%, SD = 12.3, and M = 5.2%, SD = 6.1, respectively (SDC5F). Of the shared 25, Haemophilus, Fusobacterium, and Bilophila had the lowest average relative abundances, ranging from 0.11%–0.12%.

Relative Abundance Differences Across V Region or Continent

Across the four groups of V regions, Bacteroides showed the highest relative abundance overall (Figure 3A, SDC5G). When investigating average relative abundance differences across the four V region groups (V12, V23, V3, V4), Dorea (p = .05) and Roseburia (p = .05) were found to be significantly different.

Figure 3. Stacked Bar Charts of Bacteria Relative Abundance: V Region Analysis.

A.) Average relative abundance of 25 intersecting genera in all V regions and all continents, by specific V region (x-axis).

B.) Average relative abundance of 25 intersecting genera for each of the data sets, organized by continent.

Abbreviations: IT Kit = Ion Torrent Kit, amplifying V2,3,4,6/7,8,9; V = hypervariable

There were also differences in the average relative abundance percentages of genera in Asia as compared to Europe, North America, and South America (Figure 3B, SDC5H). The highest average relative abundance genus in Europe, North America, and South America (M = 22.3%, SD = 18.22, M = 12.7%, SD = 10.3, and M = 31.0%, respectively) was Bacteroides, compared to M = 0.3%, SD = 0.32 in the data from Asian subjects. Bacteroides was significantly different across the four continents (p = .04). Conversely, subjects from Asia had much greater average relative abundances of Prevotella (M = 22.6%, SD = 25.3), which was threefold to 4.5-fold higher compared to Europe, North America, and South America (M = 4.9%, SD = 3.4, M = 4.6%, SD = 2.8, M = 7.6%, respectively); however, there was high variability across studies (p = .54). Three other genera, Anaerostipes (p < .02), Desulfovibrio (p < .03), and Butyricimonas (p < .04), were found to be significantly different across continents.

Shared Bacterial Genera Between Combined IT Table and the HMP Table

The combined HMP data set comprised 466 total genera. Comparisons against a benchmark microbiome data set, like HMP, can assist the nurse researcher in anticipating microbiome community differences in specific disease populations when study planning. The HMP genera were merged to the Combined IT Table by genus, and 129 genera were found in common between the two (SDC5I). Average abundance of HMP was compared to the continent groupwise average genera abundance from the Combined IT taxonomy Table. The average abundance of overlapping genera between the studies from Asia and the HMP (80 total genera) showed the lowest correlation (r = .57, p < .0001). In contrast, those for the South American studies and HMP (24 in total genera) showed the highest correlation (r = .8, p < .0001). Comparisons between average abundances in HMP and North America (93 total genera), and HMP and Europe (92 total genera) showed similar correlations (r = .75, p < .0001; r = .76, p < .0001, respectively) for both (Figure 4).

Figure 4: Average Relative Abundance of HMP versus Continent of origin.

A.) Logarithm base 10 average relative abundance of HMP (y-axis) versus logarithm base 10 average relative abundance of North America (x-axis).

B.) Logarithm base 10 average relative abundance of HMP (y-axis) versus logarithm base 10 average relative abundance of Asia (x-axis).

C.) Logarithm base 10 average relative abundance of HMP (y-axis) versus logarithm base 10 average relative abundance of South America (x-axis).

D.) Logarithm base 10 average relative abundance of HMP (y-axis) versus logarithm base 10 average relative abundance of Europe (x-axis).

Discussion

We present here, for the first time, an easy, reproducible workflow on how to combine published taxonomy tables from published literature sources. The primary outcomes from this methodology are as follows: (a) It is feasible to combine taxonomy tables from published literature for preliminary analyses and study planning; (b) Methodological choices like continent and V region influence the presence and abundance of gut microbial genera; (c) Despite specific inherent study differences, some genera can be found in common across healthy gut microbiomes; and (d) Combining taxa from an investigator’s microbiome study with an established microbiome data set such as the HMP can further validate commonly found gut microbes.

To our knowledge, this is the first investigation using a simplified workflow for combining published microbiome data from healthy controls using various references databases, bioinformatics pipelines, and V regions. Nurse scientists are well-positioned to contribute to the field of microbiome research due to their clinical background and understanding of the human condition as an interconnected system in which multiple body systems, including the bacteria of the gut microbiome, communicate in healthy physiologic functioning. Although possessing the clinical knowledge needed to design research questions that can be quickly translatable, nurse scientists may not initially have the computational or bioinformatics training required to run a combined microbiome analysis from raw sequence data; therefore, this work provides a detailed workflow on how best to do a project like this. A foundational aspect of this work involves our use of published, processed data. Our approach did not reprocess these data using a single pipeline. Instead, we simply sought to examine the level of congruence across studies via OTU table analyses, which is a more pragmatic approach for nurse researchers who do not have the resources or time to reprocess data sets when planning research or conducting preliminary analyses. Although we present an exemplar workflow for using IT sequencing studies to compare healthy controls and merge data at the taxonomic level of genus, researchers using this method for study planning can make modifications to fit their specific question of interest. Other microbiome niches that have lower bacterial biomass (i.e., oral and vaginal microbiome) may more reliably annotate to the species level. Furthermore, nurse researchers may employ this workflow to evaluate bacteria that may be altered in identification or relative abundance in association with specific disease conditions.

Selection of the V region-specific primer for 16S sequencing technology is a crucial study consideration, as it can introduce its own specific taxon biases (Fouhy et al., 2016). Therefore, we investigated similarities/differences in annotated bacteria and relative abundances by V region. In our analysis, we determined the V4 region to have the most overlap/agreement with other V regions, so it would be reasonable for the researcher to choose a V4 region primer for their amplicon targeting to achieve a broad survey of common taxa. Region V12 had the highest number of unique species and therefore may help target new or unusual taxa. Importantly, these aforementioned differences further confirm the utility of collecting preliminary microbiome data to assist in rigorous experimental design and study planning.

There was a noticeable difference in the genera relative abundance across continents. More specifically, Prevotella was notably more abundant, and Bacteroides was significantly less abundant in the Asian data sets than in the data sets from Europe and North and South America. Given this finding, it is vital that nurse researchers pay attention to geographic location of samples when investigating gut microbiomes across the globe or even within a population. This observation agrees with another published microbiome study that reported these patterns when comparing Western countries and non-Western countries (Gorvitovskaia et al., 2016). Nevertheless, this finding should be confirmed in future research; this workflow is designed to assist study planning without a detailed analysis of confounders or individual variability in the gut microbiome. These findings demonstrate an essential study consideration in how diet and other lifestyle factors influence the gut microbiome, and their consideration/control is paramount to planning a rigorous microbiome study (David et al., 2014).

We found 26 genera that were shared between all continents and 56 genera that were shared across all V regions (not including taxonomy tables from the multi-V region IT kit). Of these two sets of genera, 25 were annotated across all continents and V regions and could represent commonly occurring genera found in healthy gut microbiomes (i.e., a possible core microbiome). These 25 genera serve important homeostatic functions within the gut, like carbohydrate metabolism and the production of short-chain fatty acids that aid the body in many important ways. The genera that were found in at least 14 of the 15 data sets include Blautia, Bifidobacterium, Ruminococcus, Dorea, Faecalibacterium, Bacteroides, Prevotella, Streptococcus, and Roseburia. Short-chain fatty acids have several vital roles in the maintenance of health and homeostasis, and Bifidobacterium, Blautia, Dorea, Faecalibacterium, Roseburia, and Ruminococcus are all putative short-chain fatty acid producers (Crost et al., 2018; Ferreira-Halder et al., 2017; O’Callaghan & van Sinderen, 2016; Shin et al., 2018; Tamanai-Shacoori et al., 2017; Taras et al., 2002; Valles-Colomer et al., 2019; Vitetta et al., 2019). Bacteroides and Prevotella also have important roles in human physiology and metabolize carbohydrates in the gut by fermentation (Ley, 2016; Wexler, 2007). These common bacteria may be altered as a result of disease. Although a detailed description of the effect disease may have on these important gut bacteria is beyond the scope of this review (see Ghaisas et al., 2016 and Honda & Littman, 2016, for comprehensive reviews), this workflow can help the nurse scientist understand which bacteria are important to a specific disease model of interest and assist in study planning.

The HMP is a well-established, free resource that any investigator, especially nurses, can easily download for comparisons with their own gut microbiome samples. The data are drawn from healthy individuals from North America (HMP, 2012) and could be used as a complementary set of data for interrogation and concordance testing. We presented a workflow to compare and contrast data from the HMP with an internal data set and showed that while methodological differences may be present between the two sets of data, there is promising potential for hypothesis testing and conclusion generation.

While this workflow describes how nurse researchers can compare published data across studies, some limitations should be mentioned. The main cause of bias is that we analyzed only 13 studies that used different methods and were not reprocessed, making conclusions based on those variations limited. We were not able to obtain taxonomy tables from all authors we contacted, which may further bias results. As researchers continue to compare and contrast published gut microbiome data across studies, it is evident that it is imperative researchers include complete taxonomy tables as supplemental files even when not required by the publishing journal. Furthermore, the metadata about the individual samples should also be included as part of a supplement so that reproducible analyses can be completed. Other limitations include differing bioinformatics pipelines across studies, differing lab processing techniques, and differing stool collection methods. All of these could pose specific biases into the taxa found for each study and should be considered when analyses of this type are reproduced. However, this work did not aim to make strong conclusions about the microbiome but to propose a method for nurse scientists and other microbiome researchers to use when study planning. Additionally, in this workflow, only IT studies were evaluated, and taxonomic similarities may not translate to microbiome studies using Illumina sequencing technology. Though, this workflow is designed to be used on taxonomy tables, irrespective of the platform used to generate the sequencing data and can be of interest to research planning microbiome research using other sequencing platforms.

Conclusion

This study presents a workflow to investigate genus-level taxon annotation and abundance consistency across published gut microbiome literature. We focused on similarities and differences in annotated taxa and relative abundance of genus-level bacteria, stratified by geography and V region to illustrate important factors for study planning. We found a shared bacterial community of 25 core genera present in most studies and each of the continents and V regions. These bacteria were composed of carbon fermenters and short-chain fatty acid producers that perform essential physiologic functions necessary for human health. Additional workflows like this will further confirm best practices when comparing published microbiome data to inform future study design evaluating human health and disease phenotypes.

Glossary

Abbreviation Used

Fully Written Out

DNA

Deoxyribonucleic acid

HMP

Human Microbiome Project

V (region)

Hypervariable (region)

IT

Ion Torrent

OUT

Operational Taxonomic Unit

rRNA

Ribosomal ribonucleic acid