Nature | Vol 585 | 24 September 2020 | 511particularly evident in a study of ApcMin/+ mice that lacked the autophagy
gene Atg16l1 and thereby failed to recruit the DNA repair protein RAD51
after infection. As a result, DNA double-strand breaks accumulated
and tumour burden increased^45. Direct causality has been demon-
strated between this CRC-associated microorganism and intestinal
stem cell mutations by injecting pks+ E. coli into human organoids—
self-organizing, three-dimensional, in vitro culture systems of epithe-
lial cells. This induced a mutation signature that notably included 112
established CRC-driver mutations^46.
In mouse models, pks+ E. coli, ETBF and E. faecalis can similarly induce
DNA damage through the induction of inflammation and oxidative
stress^40 ,^47 –^49. In genetically susceptible Il10−/− mice, pks+ E. coli and ETBF
induced 8-oxoguanine DNA lesions that correlated with a higher inci-
dence of colonic tumours^48. More recently, tungstate administration
was found to reduce DNA damage and tumorigenesis in Il10−/− mice, at
least in part through the inhibition of metabolic pathways that are relied
on by the putatively pathogenic E. coli^50. The ETBF toxin also increased
spermine oxidase levels, which led to the generation of reactive oxygen
species and the induction of the DNA-damage marker γ-H2A.x^40.
The effect of microorganism-driven metabolism
Beyond the outgrowth and functions of specific strains, the collec-
tive activities of the microbiota—which is seemingly more stable than
taxonomic readouts^3 —deserve further attention. Several products
of bacterial metabolism have been implicated in CRC, many of which
are associated with dietary intake^14 or drug metabolism (for example,
aspirin^51 ). These include products of protein fermentation, second-
ary bile acids from high fat intake, and short-chain fatty acids (SCFAs)
metabolized from carbohydrates and phytochemicals^52.
Although the diet can be directly carcinogenic^53 , it can also alter
the ecosystem by skewing the abundance of specific species and
metabolites. To this end, a diet associated with increased levels of
high-sulfur-metabolizing bacteria correlates with increased risk of
distal colon and rectal cancer, an effect that is thought to be due to
genotoxic activities^54. Independently, a ‘Westernized’ high-fat diet
was found to correlate with CRC recurrence as well as with the colla-
genolytic activity of certain microorganisms^51 .To identify candidate
disease-associated metabolites, metaproteomic analysis has been
performed on stools from individuals with Lynch syndrome^35 , ade-
nomas and CRC^55 ,^56. These studies identified a heightened oxidative
metabolic microenvironment, which is thought to be due to increased
levels of reactive oxygen species and reactive nitrogen species in the
colon of patients with CRC^56 as well as to increased concentrations
of the DNA-damaging bile acid deoxycholic acid^55 ,^57 ,^58. Because these
observations are primarily correlative in nature, in-depth mechanistic
follow-up studies are required in order to establish causality.
More functionally, increased tumorigenesis is observed in mice that
lack free-fatty acid receptor 2 (FFAR2), a SCFA receptor. This effect is
thought to be driven by reduced IEC integrity, enabling increased bacte-
rial influx, overactivated dendritic cells and an exhausted CD8+ T cell
phenotype^59. Much work has focused on the putatively anti-tumorigenic
SCFA butyrate, which is fermented from dietary fibre^52 ,^60 , and on aryl
hydrocarbon receptor ligands such as indole-3-carbinol, which is
derived from vegetables^61. In mouse models of CRC, the latter reduced
IEC proliferation by restricting the accumulation of β-catenin^62.
Butyrate reportedly functions intracellularly as a histone deacety-
lase inhibitor to downregulate IL-6^63 and enhance the antimicrobial
functions of macrophages^64. Recently, mouse Faecalibaculum roden-
tium and its human counterpart Holdemanella biformis were iden-
tified as protective species in CRC, in particular due to their release
of butyrate and its subsequent downstream histone deacetylase
inhibitory activity that dampens tumour proliferation^65. By binding
to the receptor GPR109A, butyrate can promote the differentiation
of anti-inflammatory IL-10-expressing T cells^66 and, in the context of
a CRC model, induce apoptosis of IECs^67. In a mouse model of colitis,
butyrate was found to act through the transcription factor FOXO3 to
suppress ISC proliferation. Although this is thought to be protective
in established cancers, delaying proper wound repair could make it
detrimental at earlier stages of the disease^68. Alongside potential tem-
poral dependencies, the genetic landscape of the host could be critical
in determining whether butyrate has a pro- or an anti-tumorigenic
role in CRC. To this end, butyrate induced IEC hyperproliferation and
increased production of mitochondrial reactive oxygen species specifi-
cally in mice that were deficient in the mismatch-repair apparatus^48 ,^69.Influx of immune-stimulating microorganisms
Inflammation is a well-established driver of colorectal carcinogen-
esis^18 ,^25 , such that individuals with IBD have an increased risk of CRC^12.
Bacterially induced inflammation has also been shown to positively
correlate with tumour multiplicity^70. However, pathways through which
the microbiota shapes the immune environment of the tumour and,
in turn, how that modulates the surrounding microbiota, is the focus
of ongoing research (Fig. 2 ).
Disruption of the epithelial barrier enables an influx of previously
compartmentalized, and potentially harmful, microorganisms
into the tissue. Enhanced tumour multiplicity and IEC perme-
ability were reported in mice deficient in the pattern recognition
receptor-associated genes Nod1, Nod2 and Ripk2, highlighting the
importance of host defence and IEC function in preventing tumorigen-
esis^71 ,^72. Similarly, loss of the SCFA-sensing receptor FFAR2 in ApcMin/+
mice increased colonic permeability, which was associated with
reduced expression of the tight-junction protein E-cadherin^59.
Such loss of barrier integrity within tumours is reportedly accom-
panied by location-specific bacterial influx^73 or formation of invasive
bacterial aggregates^74. In humans, these aggregates of microorgan-
isms—termed ‘biofilms’—consist of a defined community that includes
B. fragilis, E. coli and F. nucleatum enclosed in an exo-polymeric
matrix^75 ,^76. Biofilms are also detected in benign polyps from patients
with familial adenomatous polyposis^76 , indicating a potential role in
the development of adenomas from polyps. Although understand-
ing of the precise function of biofilms is limited, inoculating ApcMin/+
mice with bacterial slurries from human biofilm-positive—but not
biofilm-negative—tissues was found to induce tumours. Notably,
tumorigenesis was observed irrespective of whether biofilm-positive
inoculants were derived from patients with CRC or from cancer-free
individuals, suggesting that these microorganism aggregates may be
an indicator of disease risk^77 ,^78. Several questions remain here, including
whether the carcinogenic properties of biofilms result from the pres-
ence of specific strains, the local microecology of the community, and/
or a maladapted response renders the host unable to react effectively
to a biofilm.
Bacterial translocation caused by deficiencies in the intestinal barrier
correlates with—and is thought to trigger—the production of several
cancer-related pro-inflammatory cytokines, including IL-1β, IL-23,
IL-22, IL-27 and—of particular importance—IL-17A and IL-6^59 ,^71 –^74 ,^76 ,^79 –^82.
In agreement with this, in mice and humans, the expression of IL-6,
IL-23 and IL-17A was found to increase within tumours that were infil-
trated by bacteria^73 ,^81. IL-23 and IL-6 are potent inducers of T helper
type 17 (TH17)-derived cytokines IL-17A and IL-22^83 ; in addition, IL-6
binds to its receptor on IECs to trigger aberrant proliferation^71 ,^74 ,^79. IL-6
and IL-17A therefore have intriguing translational potential, although
future studies are required to determine whether invasive microorgan-
isms unquestionably trigger increased expression at the protein level.
Specific cellular sources should also be investigated, because these
cytokines can be produced by—and act on—immune, epithelial and
mesenchymal cells. The importance of receptor expression patterns is
exemplified by a study showing that CD4+ T-cell-specific ablation of the
IL-1 receptor decreased tumour-elicited inflammation, whereas lack of