514 | Nature | Vol 585 | 24 September 2020
Review
Importantly, this technology has been extended to allow for modelling
of anaerobic and aerobic microorganisms while maintaining viable
IECs^123. Such advances will also facilitate the study of intra-microbe
competition for essential nutrients, which may underpin outgrowths
of oncomicrobes.
Genetically engineered mouse models (GEMMs) form the basis of
many of the studies discussed in this Review (for reviews of the different
models, see refs.^25 ,^124 ); however, several such models have a long latency
period and fail to recapitulate late-stage human disease. To reduce the
duration of models and thereby allow for antibiotic treatment without
the occurrence of bacterial resistance, tumorigenic organoids can be
locally injected into the rectum of experimental animals^117 ,^125. Ultimately,
however, gnotobiotic GEMMs will be indispensable for deciphering—in
a reductionist in vivo approach—the complexity of microbiota–host
interactions and proving causality in CRC. Humanizing mice through
faecal transplantation^27 , or using ‘wilding’ mice that have a more natural
microbiota^126 , will aid the study of inter-individual microbiota hetero-
geneity and help to tackle some issues with translatability^127. To study
late-stage CRC, metastatic models have been developed and provide
a much-needed tool to investigate potential roles for the microbiota in
modulating the epithelial–mesenchymal transition^128. Further GEMM
models are under development, and will hopefully help to resolve ques-
tions such as whether F. nucleatum—associated with a primary tumour and
identified in matched metastasis sites^129 —drives metastasis or is merely
a passenger, and whether this phenomenon applies to other species.
Experimental vigilance is essential to ensure the reproducibility of
findings, particularly because the microbiota can be affected by the
animals’ diet, the cages in which they are housed and the specific facil-
ity in which they are reared^130. Although antibiotics are a useful tool for
modulating the microbiota during disease progression, they often target
certain groups of microorganisms without adequate control or guaran-
tee of eradication, making data interpretation challenging. For example,
in the DSS–azoxymethane model of cancer, enhanced tumorigenesis
was observed in germ-free mice^131 but the opposite reported when the
microbiota was depleted by treatment with antibiotics^82. Such seemingly
contrasting data could be due to an outgrowth of uncharacterized and
antibiotic-resistant species in the colon, differences in the time points
at which the microorganisms were eliminated, or differences in—for
example—the immune system of germ-free mice^132.
Beyond bacteria and metabolites, there is emerging evidence of
fungal^133 , archaeal^134 and viral^135 changes in CRC, such as an increased
abundance of fungi from the genus Malassezia^30. Moreover, induc-
tion of IL-18 by fungal commensals was recently shown to inhibit
colitis-associated CRC in mice, thus highlighting a protective role^136.
Particular consideration should also be given to the specific strains
and culture techniques used, owing to variable intra-strain virulence
potential^137 ,^138 and the stark differences in cancer-related gene expres-
sion that are observed when comparing viable and heat-inactivated F.
nucleatum^139. Moving forwards, precise characterization of the numer-
ous species within a genus is needed to uncover the full functional
diversity of the microbiota. Such knowledge may help to explain appar-
ent contradictions, such as certain species of the putatively protective
Clostridia genus producing CRC-associated secondary bile acids^140.
For such integrated analysis pipelines, broad and systematic
‘multi-omic’ approaches will be paramount to building an interac-
tome^141 ,^142. These could also include spatial analysis using in situ imaging
alongside single-cell sequencing and in silico computation modelling
approaches that mathematically reconstruct the metabolome^143 –^145.
Another emerging area is the analysis of the epigenetic landscape in
the context of the microbiome^146 , such as Fusobacterium correlating
with distinctive methylation patterns in CRC tumours^147. Finally, given
extensive inter-individual heterogeneity and limitations within model
systems, large international patient cohorts that comprise integrated
data on the microbiota and genetic traits of the host will be crucial in
order to pinpoint clinically relevant findings.Future outlook
Rapidly accumulating studies over the past decade have detected dif-
ferences in the composition of the intestinal microbiota in individuals
with early-stage pre-cancerous lesions ranging to those with metastatic
CRC. Observations such as the increased abundance of F. nucleatum,
E. coli or B. fragilis in CRC have created a new dimension of oncology, and
such microorganisms could be incorporated as biomarkers alongsideMicrobiotaGEMMs
Gnotobiotic/antibiotics
‘Wilding’ of modelsSoluble
mediators
Immune cells,
stromal cellsLongitudinal international cohorts
(‘Multi-omic’ approaches, host
genetic traits, environmental factors)Cancer cells- Biomarkers
- Diagnostics
- Therapies
Organoids
(± CRISPR–CAS9)
Microinjections
(microorganisms, metabolites)
Co-culture systems
(cells, soluble mediators)
Organ-on-a-chipFig. 3 | Approaches to advance the translation of microbiome-based
therapeutics in CRC. Combining ‘multi-omic’ analysis of samples taken from
large, international patient cohorts (left) with improved mechanistic in vitro
(right) and in vivo (middle) approaches aims to distinguish between correlative
or causative observations. In vitro approaches that have shown recent
improvements include co-culturing patient-derived organoids with immune
and stromal cells, intra-organoid injections of microorganisms, reversing
organoid epithelial cell polarity and microf luidic ‘organs-on-a-chip’.
Patient-derived and CRISPR–Cas9-edited organoids can be injected into the
rectum of recipient mice to expedite in vivo tumour formation in the context
of the tumour microenvironment. These and other GEMMs of CRC enable
reductionist approaches to identify targetable host–microbe interactions
through longitudinal sequencing, antibiotic, gnotobiotic, faecal transplantation
and ‘wilding’ experiments.