Science - USA (2022-05-27)

Farnesoid X receptor (FXR), to inhibit prosta-
glandin E2 synthesis and promote colonic crypt
regeneration after intestinal mucosal wounding
( 52 ).Secondarybileacidscanmediateresponses
to alphavirus infection by promoting systemic
type I interferon responses ( 53 )andinducepro-
liferation and DNA damage in intestinal can-
cer stem cells through FXR ( 54 ). A recent study
showed that microbial enzymes abundant in
the feces of patients with UC are involved in
driving colitis severity upon transplant into
germ-free mice ( 55 ). Together these studies
indicate that microbial products can extensively
modulate host physiology locally and systemi-
cally, making them attractive targets for ther-
apeutic interventions in metabolic disorders.

Technical limitations and new avenues

Genomics approaches that infer differential
abundances of microbes by analyzing bacterial
16 S rRNA genes or internal transcribed spacers
(ITS) of fungi can achieve genus-level sensi-
tivity, but these approaches do not allow for
differentiation between closely related spe-
cies or strains. 16S rRNA gene and ITS se-
quencing analyses are useful for detecting
high-abundance microbes, but differentially
abundant microbes may prove to be only pas-
sengers, not the drivers, of a given host pheno-
type. Functional, low-abundance microbes that
might drive a phenotype could be missed. It is
therefore essential that sequencing studies are
supported by experimental functional analy-
ses in animal models and microbial isolates

from hosts with phenotypes of interest. One

study showed that variousBacteroides ovatus

isolates from human stool induced the pro-

duction of strain-specific SIgA in monocolo-

nized mice, highlighting that variable mucosal

immune responses might be caused by varia-

ble microbial strains ( 56 ). A new study has also

revealed genetic diversity among opportunis-

tic C. albicans strains isolated from patients

with UC. Strains that secrete the cell-damaging

toxin candidalysin trigger intestinal inflam-

mation when colonized in mice, reflecting dis-

ease features of individual UC patients and

underscoring microbial strain–specific effects on

immune-mediated inflammatory diseases ( 57 ).

One way to avoid selective vision in under-

standing the proximal causes of microbiota-

mediated disease is to use shotgun metagenomics

of entire microbiomes. This technique is cul-

ture independent, retrieves all the genomes

from a microbial community, and can be used

not only for taxonomic purposes but also for

discovering previously uncharacterized genes,

enzymes, and metabolites. Catalogs of both

mouse and human gut metagenomes are now

available for mining ( 58 , 59 ). However, despite

technical improvements, many challenges re-

main, especially for the assignment of genes to

function or to specific microbes.

Pure culture of microorganisms remains the

standard method for testing functional roles

in experimental models and for testing ther-

apeutic possibilities. In recent years, the num-

ber of microbes that can now be cultured from

both mouse and human sources has greatly

expanded ( 60 ). Integration of data from cul-

ture collections with genomes, transcriptomes,

and metabolomes has now made it possible to

characterize the behavior and natural products

of cultured and uncultured microbiota to a

high level of resolution (Fig. 3) ( 61 ).

A long-standing problem in understanding

the full contribution of the microbiota to host

physiology is the current focus on bacteria,

although increasingly more attention is being

focused on fungi. Viruses are challenging to

identify if they are not persistent, and most

viral studies have focused on bacteriophages.

Macroparasites have been largely ignored, par-

tially owing to a lack of tools, although several

studies show that hostphysiology and immu-

nitycanbeprofoundlyaffectedbyprotozoan

and helminth parasites in several mammalian

body sites.

To date, most experiments in functional mi-

crobiome research tend to be gain-of-function

studies. Loss-of-function studies at the strain

or gene level can precisely define the role of a

microbe or a gene in a disease model. The ac-

tivity of bacteriophages shows how reducing

the load of a specific pathogenic bacterium in

thegutcanbeeffectiveinattenuatingdisease

( 43 ), but the possibilities for therapeutic ex-

ploitation of phages is complicated by evolu-

tionary and ecological considerations. Studies

have shown that CRISPR-Cas9–based genome

editing can be used on a complex microbiome

for site-specific gene insertions and deletions

of targeted members in a microbial community

( 62 , 63 ). These studies show exiting potential

for experimental manipulation of microbial

function within a community without per-

turbing the whole population and without

any need to culture organisms.

The laboratory mouse is the most popular

animal model for microbiome research, and

much of what we understand at present about

the mammalian microbiota has emerged from

these models. Nevertheless, microbiome studies

in mice can fail in translation to humans ( 64 ).

Alternative models are being developed; for

example, the wormCaenorhabditis elegans,

fruit flyDrosophila melanogaster, zebrafish

Danio rerio,hydraHydra vulgaris,mosquito

Anopheles gambiae( 65 ), and wild pigSus

scrofa( 66 ) have all been used for addressing

specific questions (Fig. 3). Despite these ad-

vances, a substantial challenge in microbiome

research remains the heterogeneity of human

microbiota, both in terms of density and of

composition among individuals. Interperso-

nal heterogeneity may underpin variation in

a range of attributes, from responses to anti-

biotics and metabolic profile to susceptibil-

ity and resistance to infection ( 67 ), as well as

outcomes of cancer immunotherapy ( 68 ). In

the future, the integration of multiple ani-

mal models together with in vitro approaches

Lu et al., Science 376 , 950–955 (2022) 27 May 2022 5of6

In vivo studies

Human studies

Hypothesis generation

New hypothesis

New hypothesis

Multiomics of the human microbiome

In vitro studies

Fig. 3. Mechanistic microbiome studies with integrated approaches to find disease targets.
Approaches of microbiome studies have been established by integrating experimental and computational
methodologies. The hypotheses generated from multi-omics (culturomics, metagenomics, proteomics,
metabolomics, etc.) analyses of the human microbiome are tested in both in vitro and in vivo models, such as
organoids, air-liquid interface cultures, and multiple animal models. New hypotheses can be generated
again from in vitro or in vivo studies. In the next phase of microbiome research, we propose that there should
be a major focus on mechanism-based studies to predict microbe-based therapeutic interventions.

THE SYSTEMIC MICROBIOME

ILLUSTRATION: KELLIE HOLOSKI/

SCIENCE