Nature | Vol 585 | 24 September 2020 | 513which counteracted a dysbiosis characterized by potentially harmful
mucus-degrading Akkermansia species^98. Notably, mice deficient in
mucin 2—which lack an intact mucus barrier—spontaneously develop
inflammation-driven CRC^99. Low-fibre diets have also been shown to
promote the expansion of mucus-degrading bacteria that can cause
erosion of the intestinal barrier^100.
IL-18 prevents spontaneous colitis in Il10−/− mice, at least in part by
inhibiting the colonization of Akkermansia muciniphila^101. In support
of these data, Il18−/− mice display both enhanced dextran sodium sul-
fate (DSS)-induced colitis^102 and exacerbated inflammation-induced
CRC^79 ,^103. Notably, co-housing IL-18-deficient and wild-type mice
to enable microbiota transfer led to an increased incidence of
inflammation-induced CRC in the wild-type mice in comparison with
those that were single-housed^79 ; further work will be required to deter-
mine the duration of this effect and the severity of the resulting disease.
By contrast, other studies report that IL-18 can promote DSS-induced
colitis, possibly by inhibiting the maturation of mucus-producing gob-
let cells^102. Whether these seemingly contradictory data arise as a result
of taking measurements at different time points, or from differences
in doses or housing facilities, is unknown.
‘Oncomicrobes’ alter immune composition
Tumours create a permissive tumour microenvironment by
recruiting certain immune cells^22 , the density and composition of
which can be skewed by the microbiota. For example, levels of the
lymphocyte-attracting chemokines CCL5, CCL20 and CXCL11 corre-
late with members of the Bacteroidetes and Firmicutes phyla and can
be induced in vitro by F. nucleatum and E. coli^104. However, whether
these chemoattractants recruit protective or pathogenic T cell sub-
sets is unclear. Such decisions may be affected by the genetic land-
scape, because a distinct microbiota correlated with cytotoxic CD8+
T cell and CD4+ TH1 responses specifically in mice that lacked the
autophagy-regulating gene Atg7 in IECs^105. Altering the microbiota
to skew towards a ‘hot’ tumour microenvironment, which is often
characterized by infiltrating CD8+ T cells and beneficial patient out-
comes, is therefore an intriguing therapeutic avenue. However, care
must be taken, because a recent study has shown that CD8+ T cells are
not always anti-tumorigenic. A distinct mouse microbiota, character-
ized by an increased abundance of Prevotellaceae, correlated with a
higher tumour burden and a skewing towards exhausted instead of
IFNγ-producing intratumoral CD8+ T cells^106. Although correlative,
this study highlights the importance of characterizing the activation
status of cells alongside their abundance, as dysfunctional T cells may
aid tumour evasion^107.
ETBF was found to induce a pro-carcinogenic TH17 response driven by
STAT3 activation in ApcMin/+ mice^80. This was accompanied by the recruit-
ment and differentiation of iNoshigh monocytic-like myeloid-derived sup-
pressor cells (M-MDSCs), with Nos2 upregulation driven by intratumoral
IL-17A binding to IL-17R+ myeloid cells. These M-MDSCs supressed the
activity of cytotoxic CD8+ T cells while inducing the expression of genes
involved in tumour growth (Mmp9) and angiogenesis (Veg fa)^108. Notably,
ETBF caused MDSC accumulation specifically in the distal colon, where
locally restricted NF-κb signalling in the epithelium triggered confined
production of the myeloid-recruiting chemokines CXCL1 and CXCL2^109.
That study is one of few taking into consideration tumour location, an
important variable given evidence for distinct microbial and immune
niches throughout the colon^110 and location-specific effects of the micro-
biota in preclinical CRC models^43. In humans, ETBF similarly induced the
expression and secretion of the CXCL1-orthologue IL-8 from epithelial
cells in an NF-κb-dependent manner^111 ,^112.
Similar to ETBF, oral gavage with the CRC-associated bacterium
Streptococcus gallolyticus in the DSS-induced mouse model of CRC
led to an increased tumour burden, the selective recruitment of immu-
nosuppressive CD11b+ myeloid populations and increased levels of
myeloid-derived cytokines, including IL-6 and IL-8^113. A greater tumour
load in ApcMin/+ mice fed with F. nucleatum was also accompanied by
an increased number of immunosuppressive intratumoral myeloid
cells—characterized as mononuclear and granulocytic MDSCs,
tumour-associated macrophages, tumour-associated neutrophils and
dendritic cells. More functionally, the MDSCs and tumour-associated
macrophages isolated from mouse tumours suppressed the prolif-
eration of CD4+ T cells in ex vivo co-culture experiments^114. The
CRC-associated bacterium Peptostreptococcus anaerobius has also
been reported to trigger an expansion of pro-tumour myeloid popula-
tions in ApcMin/+ mice, by selectively adhering to colon cancer cells and
inducing NF-κb. A simultaneous increase of T regulatory cells, TH17 and
cytotoxic CD8+ T cell frequencies was also observed^115. F. nucleatum
can additionally act in a cytokine-independent manner and directly
inhibit natural killer (NK)-cell cytotoxicity, enabling the tumour to
evade immune destruction. Pre-treating CRC cell lines with F. nucleatum
reduced the activity of co-cultured NK cells via binding of the receptor
TIGIT to the F. nucleatum-derived protein Fap2^116.
The ability of distinct species to drive similar downstream effector
functions, including NF-κB activation, could help to explain heteroge-
neity in the disease-associated microbiome. Therefore, targeting the
consequences of dysbiosis, such as the recruitment of certain immu-
nosuppressive populations, might be more therapeutically efficacious
than inhibiting specific bacteria. However, caution must be taken given
the pleiotropy of immune-derived factors and the fact that the ultimate
outcome depends on the density, activation status and localization of
different cells.Technologies to investigate microbiome causality
Despite substantial progress in the field, there is a lack of evidence for
therapeutically tractable causal interactions between the altered micro-
biota and host. Compiling a complete CRC-associated microbiome
that includes the abundance and function of bacteria, fungi, viruses,
archaea and metabolites—as well as considering the multi-layer cross-
talk between these factors and the host—is a daunting task that is yet
to be tackled. This is in part due to limitations in sequencing depths, in
the ability to culture certain species, and of models to recapitulate the
human tumour microenvironment—issues that technological advances
are addressing (Fig. 3 ).
Organoids are three-dimensional culture systems of epithelial tissue
that enable the stable in vitro culture of, for example, patient-derived
tumours^117. By using this technique, the effect of various stimuli can
be tested, but poor accessibility to the enclosed apical surface of
organoids renders culture with microbiota components challeng-
ing^118. However, the development of intra-organoid microinjections
of bacteria^119 and reversing the epithelial polarity^118 have begun to
solve this issue. Indeed, injection of pks+ E. coli into human organoids
revealed the capacity of this bacterium to directly cause clinically rel-
evant CRC-driver mutations^46. Despite the value of such reductionist
approaches, it is important to know the capacity of oncomicrobes to
act on the genetic landscape and in the context of other cells from the
tumour microenvironment. To this end, organoid experiments can be
expanded to co-culture with immune and mesenchymal cells^117 , while
targeted mutation combinations are introduced using CRISPR–Cas9^117.
The effect of such manipulations on post-translational modifications
can also now be determined using mass cytometry^120.
To further account for the fluidic and mechanistic properties of an
organ, a microfluidic in vitro system termed ‘organ-on-a-chip’ was devel-
oped that utilizes specific channel permeabilities to integrate several
cellular elements^121. Such a microfluidic model, termed ‘HuMIX’, has
provided functional insight into CRC-related metabolites through the
observation that simultaneous co-culture of non-digestible nutrients,
live microorganisms and human epithelial cells was required to gener-
ate a distinct SCFA ratio that limited the self-renewal capacity of IECs^122.