The History of Studying Bacteria
Microbial communities have been studied for millenia by civilizations across the world
before they were described by modern science. In particular, beneficial techniques
utilizing consortia such as food fermentation and the use of manure on crops date
back to 6,000 BCE. Even though these early societies were not unaware of the molecular
or mechanistic foundations generations of processes were built on, we now know these
methods to be the result of complex, living ecosystems called microbiomes.
Before modern medicine, the gastrointestinal (GI) tract was frequently studied in
the context of digestion complications or disease. Societies would use microbial community-derived
products in the form of medicinal remedies and ancestral foods to “guard the stomach,”
treat ailments, and promote general well-being and health. There is evidence from
the Romans that disease transmission through waste material was understood with the
construction of aqueducts to introduce clean water and remove waste. However, we did
not learn until the mid-1600s that healthy human GI tracts were home to microbes,
and it was not until the 20th and 21st century we learned they were essential to human
health.
In 1676, Antony van Leeuwenhoek reported the first visual observation of bacteria.
With a single lens microscope of his own design, Leeuwenhoek studied the rod-shaped
cells within a mouth scraping sample and he called the organisms little animals or
“animalcules,” a term he had coined a few years earlier. In his examination of these
animalcules, Leeuwenhoek would go on to describe and draw their movement through liquid
(i.e., cellular motility), detail his belief that some could survive without air,
and he would accurately approximate their length (3 µm). From these descriptions and
what is now known about the residents of the mouth microbiome, he was most likely
describing a member of the genera, Selenomonas. Leeuwenhoek would also be the first to report a microbe, the parasite Giardia, from his own fecal material. While the discovery of unicellular organisms was fundamental
to the fields of microscopy and microbiology, Leeuwenhoek would not share his microscopic
methods and it would take approximately 200 years for scientists to validate his findings.
Discovering How to Culture and Define Bacteria
Following Louis Pasteur's report in 1860 on the first culture medium recipe that could
support organism growth, scientists would quickly develop tools and methods for the
selective isolation of bacteria from ecosystems. These critical studies included specific
growth medias, development of solid agar media, and design of sterile growth containers
(i.e., Petri dish). Additionally, microbes started to receive general classifications
based on characteristics such as gas production on specific sugars, motility, and
structure of their cell wall composition with the invention of the Gram stain by Hans
Christian Gram in 1884. Despite the fact bacteria's discovery originated from an oral
community and microbes were discovered within a fecal sample, the field would predominantly
focus on the isolation and study of pure cultures for the next 100 years.
During this new era of bacterial study, the discovery of gut-derived microbes and
anaerobic communities had begun with fundamental work by Pasteur in the 1860s. Pasteur
discovered the first pathogenic anaerobe, Clostridium septicum, and in 1863 he coined the terms “aérobis” and “anaérobis” depending on the microbes'
requirement for oxygen. During this time, additional methods for culturing gut-derived
microbes were reported. For example, in 1905, Alfred MacConkey added bile salts to
liquid culture to promote the growth of lactose fermenting bacteria from feces.[1] Pasteur and others also began experimenting with anaerobic growth conditions.[2]
[3] Studies at the end of the 19th century included the isolation of Clostridium tetani in pure culture with successful anti-toxin development against the anaerobe developed
only 1 year later in 1890.[4]
From Studying Microbial Community Natural Product Production to Shotgun Sequencing
The interest in microbial community functions and interspecies interactions was ignited
following the discovery of penicillin G from the fungus Penicillium notatum by Alexander Fleming in 1928. Fleming's foundational studies reminded the field that
for millions of years microbes have interacted with each other and an understanding
of this unknown world could provide significant breakthroughs for treatments and therapeutics.
A significant natural product isolation effort from environmental source materials
resulted in the discovery of several new chemical subclasses, many of which went on
to be the building blocks of large scale, high-throughput screening studies by the
pharmaceutical industry in the late 1900s (e.g., aminoglycosides, polyketides, cephalosporins,
macrolides, and tetracyclines).
For the next several decades, investigators would continue to rely on experimental,
in vitro evidence to determine the function of a microbe or a consortium. However,
with the application of 16S rRNA-based sequencing and whole genome sequencing (WGS)
in the 1970s and 1990s, the functional potential of a microbe and a population could
be assessed without culturing. In 1995, back-to-back Science publications reported
the first two WGS from bacteria: Haemophilus influenzae (1,830,137 bp) and Mycoplasma genitalium (580,070 bp).[5]
[6] Although WGS provides an organism's complete genome, the technique requires the
growth and isolation of individual organisms which is a significant barrier for community-based
studies. With the rapid progression of sequencing technologies and decreased run costs,
16S rRNA-based sequencing became a high-throughput technique that could be used to
survey and define the taxa of consortia. From these data, it was clear that in vitro
culturing techniques were insufficient for culturing all organisms in a sample, which
further pushed the need of culture-independent methods for assessing microbial communities
or isolates.
Jo Handelsman was the first to coin the term metagenomics.[7] From her work on assessing biosynthetic gene clusters and their resulting natural
products, she was one of the first to suggest DNA analysis from an entire sample was
useful for predicting if the community contained new bioactive small molecules. Two
major questions remained: Can you and how do you reassemble every genome from a whole
sample? In 2004, foundational work by Tyson and colleagues used deep shotgun sequencing
with relaxed reassembly requirements to assemble and bin more contigs allowing for
the near-complete reconstruction of two genomes and partial reconstruction for three
other genomes from a single biofilm sample.[8] There are accepted best practices for sample processing,[9]
[10] however, there is not a standard method for data analysis pipelines although there
has been some effort to create standard protocols despite the inherent challenges.[11]
[12]
[13] Essential work in genome curation[14] and data visualization strategies such as Anvi'o[15] are ongoing and because the field and its techniques are evolving, data analysis
is not trivial and requires an investigator familiar with complex datasets to lead.
Studies can now reconstruct individual strain genomes from consortia through shotgun
sequencing to track biological function within communities (i.e., gene and bacterial
fitness,[16] niche partitioning[17]), identify single nucleotide variants, identify single amino acid variants, and
explore and compare genomes across all branches of life.
With the rapid increase in our ability to resolve whole genomes within consortia,
the application of shotgun sequencing to human study has resulted in the generation
of enormous datasets. In one publication by Pasolli and colleagues, 46 datasets were
analyzed, and more than 150,000 genomes were reported from approximately 10,000 human
metagenomes.[18] There have been recent cancer studies associating taxa with disease phenotypes such
as Flavonifractor plautii, Bacteroides vulgatus, and Parabacteroides spp CT06 in early-onset colorectal cancer[19] and Streptococcus spp, Veillonella spp, and Actinomyces spp in three independent cohorts of patients with pancreatic ductal carcinoma.[20]
Complementary to metagenomics, metatranscriptomics is the analysis of gene expression
within a microbial community. Pathways are identified and can be mapped back to organisms
through paired metagenomic data. RNA-based studies are inherently challenging due
to the labile nature of the single-stranded biomolecule and processing requires specific
preservation methods to prevent degradation.[21]
[22] One study that resulted from the Human Microbiome Project 2 (HMP2) reported species-specific
transcriptional activity in their inflammatory bowel disease (IBD) patients compared
with non-diseased controls. Two organisms, Alistipes putredinis and Bacteroides vulgatus, were correlated with disease severity and responsible for the expression of the
methylerythritol phosphate pathway in IBD. A. putredinis was negatively correlated while B. vulgatus was positively correlated.[23] These data were of high interest because the mechanisms for gut dysbiosis and chronic
diseases such as IBD remain unknown.
A New Era of Protein and Metabolite Analysis
Alongside the genomic revolution, metabolomics and proteomics have been experiencing
a wave of advancement with the introduction of instrumentation into laboratories and
more robust data analysis pipelines. Nuclear magnetic resonance (NMR) and a series
of mass spectrometers such as the liquid chromatography-quadrupole time-of-flight-mass
spectrometer (LC-QTOF-MS), liquid chromatography-orbitrap-mass spectrometer (LC-Orbitrap-MS),
and electron impact/chemical ionization gas chromatography mass spectrometer (EI/CI-GC-MS)
are most widely used in studies today. For one metabolomics project, several instruments
can be used to detect a wide range of compounds with varying physicochemical properties.
Although the technology for many instruments was developed decades before their widespread
application to the multi-omics studies, systems have become more widely available
to academic institutions since the early 2000s and these instruments have launched
proteomics and metabolomics into a new era of discovery.
Metaproteomics is the study of expressed protein content within a microbial population.
While genomic strategies can define the functional potential of a community, metaproteomics
aim to characterize the active functional state of a population based on the detected
peptide and protein profiles. Generally, proteins are extracted from samples and enzymatically
cleaved by trypsin into peptides for analysis.[24] Because peptides are synthesized with known building blocks and trypsin specifically
cleaves the C-terminal side of lysine and arginine, protein sequences can be reconstructed
from peptide fragmentation spectra and mapped back to paired metagenomic data for
the same sample. In a study by Tanca and coworkers, almost 30,000 microbial peptides
detected from a human colonic luminal content cohort[25] were reported. The same group also reports distinguishing microbial peptides between
three tumor clinicopathological features (294 distinguishing peptides for stage, 94
for grade, and 568 for tumor infiltrating lymphocytes, TILs) with distinguishing capabilities
of 95% accuracy by stage, 100% accuracy by grade, and 100% accuracy based on the presence
of TILs. In a separate study by Long and coworkers, 91,902 peptides and 30,062 protein
groups from the fecal samples of colorectal cancer and healthy donor cohorts[26] are described. From this, 341 peptide groups were identified to be significantly
altered in abundance between colorectal cancer and healthy donors. The groups were
associated with functions such as iron uptake, oxidative stress, and DNA recombination,
repair and replication. The clinical impact of these data will require follow-up,
longitudinal studies, and paired in vivo work but the annotation of microbial community
proteomes continues to grow as a powerful complement to other multi-omic strategies.
Metabolomics has long been a primary focus of microbial community studies because
of the biosynthetic potential of consortia to produce therapeutically relevant compounds
as previously described in this work. Additionally, microbe metabolite levels in the
GI tract can impact host immune function, nutrient uptake, mental health disorders,
and organ function. Metabolomics aims to detail the current functional state of the
microbiome through the analysis of compounds approximately less than 2,000 Da in size.
However, because microbes and humans produce many of the same compounds and the bioactivity
of a single compound in the context of thousands is not well understood, interpreting
metabolomics results remains a significant challenge. To address these issues, metabolic
modeling studies and pipelines such as KBase have great promise to link assembled
genomes to metabolic flux in a sample.[27] However, this analysis is currently limited to a few input genomes and not microbial
communities in complex environments.
Biospecimens contain thousands of compounds within a single sample. Metabolites range
in size, hydrophobicity, charge and other properties and a study will often require
multiple instruments to describe several compound classes. Inherently, metabolomic
studies have a significant bottleneck in compound identification and validation that
is not experienced by genomics or proteomics. Metabolite structures are not composed
of characterized, repeatable subunits like genes (nucleotides) or proteins (amino
acids) and modifications (e.g., dehydration, decarboxylation, reduction, oxidation)
can occur through known and unknown mechanisms by the host, other microbes, and the
environment. Although in silico fragmentation modeling tools are becoming more advanced (SIRIUS[28]), feature validation requires authentic standard comparison on the same instrument
as the sample was run to be considered the highest level of MS confirmation. Because
of this, many studies begin with targeted methods to evaluate known compound classes
such as bile acids, short chain fatty acids, amino acids, mono-/di-/tri-saccharides,
fatty acids, tryptophan catabolites, indoles, and other small organic acids. Many
of these compounds are routinely quantified and/or reported as normalized relative
abundance by academic and commercial laboratories.
In addition to known features, biospecimens are composed of a substantial percentage
of unknown features that are of significant interest due to unknown etiology of many
diseases. NMR and MSn techniques are frequently used to evaluate the unknown metabolite space of a sample.
In untargeted MS, compound fragmentation profiles can be compared with known compounds
to putatively assign m/z values to a previously described subclass. The open-source pipeline Global Natural
Products Social Molecular Networking (GNPS) compares fragmentation profiles to databases
for identification of structural similarity to known subclasses.[29] In a study by Quinn and coworkers, GNPS uncovered new bile acid variants with amino
acid modifications from murine material.[30] In humans, a recent application of the untargeted MS approach was applied by Gumpenberger
and colleagues to a cohort of 88 colorectal cancer patients, 200 high-risk adenoma
patients, and 200 low-risk adenoma patients.[31] They report 442 statistically significant molecular features from plasma discriminating
between colorectal cancer and adenoma diagnosis. Similar to other strategies, metabolomics
findings are often correlative and used as a starting point for studies and their
findings require more rigorous experimentation to validate the significance of an
altered metabolome.
The Foundational Microbiome Studies
The NIH HMP was a two phase, decade long, multi-institutional study that laid the
groundwork for exploring human microbiomes and health outcomes.[32] In the first phase (HMP1), researchers collected and analyzed samples from 242 healthy
adults across five major body sites (oral, skin, gut, airway, and vagina) at three
time points for 16S rRNA-based sequencing and shotgun sequencing.[33] From their initial report of 5,177 16S rRNA sequencing profiles and 681 shotgun
sequencing profiles, a major finding of this study was that the taxonomic profile
of a subject did not always correlate with host phenotype. In the next phase of the
project, the integrative HMP (iHMP or HMP2) aimed to expand sample collection and
scope to address the findings from HMP1. HMP2 included additional sample types (blood,
urine) from healthy, diseased (IBDs such as ulcerative colitis and Crohn's disease),
pregnancy and preterm birth, and a prediabetic cohort. These cohorts were analyzed
by several techniques including metagenomics, metatranscriptomics, metaproteomics,
metabolomics, virome profiling, antibody profiling, host genome profiling, epigenetic
profiling, and cytokine profiling. Many of these strategies such as antibody, cytokine,
and epigenetic profiling were added to explore the impact of microbes on the host
while strategies such as host genome profiling were added to elucidate if subjects
had a predisposition to a particular disease state based on host genomic variation.
The HMP laid a roadmap for large-scale human microbiome studies. In addition to their
biological findings, the study highlighted processes such as data repository development
as key to building knowledge within the microbiome community (HMP data portal, https://www.hmpdacc.org/). Public databases for platforms continue to grow with the collective goal of increasing
data availability for future data mining (e.g., MicrobiomeDB, Metabolomics Workbench,
GNPS, MetaboLights, NCBI, INSDC).
Clinical Considerations
Sample Collection for Microbiome Analysis
To evaluate the role of host microbes, patient material such as intestinal content,
urine, tissue, and blood (plasma, serum, blood spot cards) are analyzed by a growing
platform of techniques. In all studies, there are critical steps throughout the pipeline
that should be considered such as (1) Pre-sample collection planning, (2) Sample collection,
(3) Sample processing, and (4) Sample submission and data analysis ([Fig. 1]). Although this review cannot cover the best practices for acquisition, processing,
and storage of all sample types across all platforms, there are several reviews and
works covering these topics.[14]
[21]
[34]
[35]
[36] Additionally, there is a body of work for analyzing the impact of the GI microbiome
on the host that cannot be covered in this review such as murine studies, single cell
RNAseq, cytokine profiling, antibody profiling, virome profiling, tissue histology,
and organoid generation and testing.
Fig. 1
General pipeline for clinical human microbiome study. Within a human microbiome study, pre-sample collection steps such as (A) protocol approval, (B) completion of project methods, and (C) appointment of necessary laboratory roles will need to be finalized prior to sample
collection. (D) Sample collection will occur at home and in the clinic. (E) Once samples have been collected, they should be immediately split and preserved
according to each downstream analyses. (F) Samples will be submitted for a variety of analysis across different platforms such
as genome sequencing, mass spectrometry (MS), multistage fragmentation (MSn), and nuclear magnetic resonance (NMR). Data will be curated and visualized prior
to publication and deposit of data into data repositories (Created with BioRender.com).
Patient Consent, Sample Acquisition, and Data Analysis Considerations
At the start of a study, it is essential to plan for sample collection approval if
no prior institutional protocol has been approved for the work. Initially, investigators
are required to apply for Institutional Review Board or ethical committee approval
for all protocols ([Fig. 1A]). This process can take several months. General topics in the approval documentation
include: a description of the specimens needed (i.e., fecal, urine, blood), description
of the project in both lay and technical language, indication if these collections
impose additional risk to the subjects, report how many subjects will be needed to
complete study, report who will be recruited to your study and why, report what analyses
will be completed with the samples, and how will data be stored and protected.
Prior to sample collection, the required downstream analyses for each sample type
should be established to ensure adequate mass or volume acquisition and proper storage
conditions are met ([Fig. 1B]). To avoid poor data quality, investigators should collaborate and consult with
the academic, clinical, or commercial laboratories that will be completing analyses
on all sample types prior to sample acquisition. To oversee study processes, a clinical
study coordinator, a sample processing lead, and a data analysis lead are recommended
to be appointed as these roles are essential ([Fig. 1C]). Microbiome experiments require sample splitting and analysis-dependent preservation
methods, and it is critical to have a team member familiar with all study needs. Although
many sites with routine collections have clinical coordinator teams that cover a project
as needed, those without teams will need to collaborate with academic laboratories
(if institution permits laboratory members to be trained for patient consent and sample
collection in clinic) or they will need to hire specifically for the project.
Often, a sample will be analyzed by several methods, and it should be split and preserved
for each method prior to storage ([Fig. 1E]). For example, one fecal sample that requires 16S rRNA-based sequencing and shotgun
sequencing from DNA, metatranscriptomics from RNA, metabolite quantitation, and cultivar
isolation should be split and stored with four different methods, each with their
own mass requirement. It is recommended that the collection and storage methods for
each specimen are consistent. Additionally, a sample should remain in the optimal
storage condition until analysis and freeze-thaw cycles should be avoided. If frozen
samples must be shipped, they should be stored on dry ice and shipped overnight to
the destination. Data curation and analysis are significant steps for all techniques
([Fig. 1F]). For many data types, quality control evaluation, data interrogation, visualization,
and presentation are not trivial and will require prior knowledge and training. Data
analysis is routinely included or offered for an additional cost from academic and
commercial laboratories. The sharing of curated data and subject metadata to public
data repositories is encouraged following publication. Integrating common efforts
such as the ability to mine previously acquired data for inter/intra-institutional
benefit will increase the quality, number, and frequency of foundational and field
progressing studies.
Outpatient Collection
Although many samples can be acquired while the subject is inpatient or onsite for
an appointment, some samples such as stool and metadata such as lifestyle surveys
and biometric information can be collected by the subject prior to the visit ([Fig. 1D]). Lifestyle surveys can capture valuable information such as diet and exercise levels
as well as biometric readouts (e.g., heart rate, sleeping patterns) from smart devices.
For example, it has been shown that vegans have overall lower levels of primary and
secondary bile acids.[37] For stool, at home collections the number of samples should be increased since not
all subjects can or prefer not to deposit at the time of their appointment. There
is not a standard kit for fecal collection, however, many institutions send home-made
kits that include an instructional pamphlet, toilet collection vessel and scoop, and
a sterilization product for cleaning hands after deposit. Depending on the analyses
needed, samples can either be frozen in the subject's freezer (approximately −20°C),
sent back to the institution by courier for highly sensitive material (i.e., RNA),
or stored at room temperature with preservation solutions (ex. ethanol, RNA-later,
Omni-met GUT or Omni-gene GUT kits).[38] If the subject stores their sample, they will bring it on the day of their visit
and upon arrival the collection team will immediately process.
Institutional Sample Bank Protocols
In addition to individual project protocols, programs with potential future microbiome
studies or those with multiple principal investigators consenting the same patient
population, the creation of a sample bank might be of interest. In this way, the protocol
can serve as an “umbrella” for the collection of medical records and biospecimens
for the purpose of current and future research. In a gastroenterology clinic, for
example, patients who are admitted for colonoscopies are consented and material such
as luminal aspirate, colonic biopsies, fecal material, urine, and blood can be collected
for the bank during their visit. Additionally, a bank of healthy or non-diseased donors
is a challenging cohort to capture for many studies. Sites have found success in consenting
and acquiring biospecimens from these non-diseased donors and making the material
available to other investigators at their site. Researchers can apply to analyze banked
samples based on their project goals. In this way, institutions can consolidate research
efforts, normalize collection and storage methods and provide a greater number of
patients to all studies, therefore increasing statistical significance and rate of
study completion. Outside of institutions, there are foundations such as the Crohn's
and Colitis Foundation who have generated their own intra-institutional bank (IBD
plexus) of samples that are available upon application approval.
