Invertible promoters mediate bacterial phase variation, antibiotic resistance, and host adaptation in the gut

Abstract

Switching ON resistance
Clonal bacterial colonies will often grow dissimilar patches, similar to a tortoiseshell pattern. These differing phenotypes arise by reversible mechanisms called phase variation. Jiang et al. developed an algorithm to survey bacterial genomes for invertible promoters that cause phase variation. Inverted repeats signal the presence of these promoters, which can flip between ON and OFF states catalyzed by phage integrase analogs called invertases. Invertible promoters linked to antibiotic resistance genes were widespread among vertebrate gut–associated organisms, including Bacteroidetes, Spirochaetes, and Verrucomicrobia. These bacteria are thus equipped and prepared for sudden environmental stress, including antibiotic exposure.
Science , this issue p. 181

Full text

AN TI BI OT IC R ESISTA NC E
Invertible promoters mediate bacterial
phase variation, antibiotic resistance,
and host adaptation in the gut
Xiaofang Jiang
1,2
*, A. Brantley Hall
2,3
*, Timothy D. Arthur
2
, Damian R. Plichta
2,3
,
Christian T. Covington
2,3
, Mathilde Poyet
1,2,4
, Jessica Crothers
5
, Peter L. Moses
6
,
Andrew C. Tolonen
2,3
, Hera Vlamakis
2,3
, Eric J. Alm
1,2,4
†, Ramnik J. Xavier
1,2,3,7
†
Phase variation, the reversible alternation between genetic states, enables infection by
pathogens and colonization by commensals. However, the diversity of phase variation remains
underexplored.We developed the PhaseFinder algorithm to quantify DNA inversion–mediated
phase variation. A systematic search of 54,875 bacterial genomes identified 4686 intergenic
invertible DNA regions (invertons), revealinganenrichmentinhost-associatedbacteria.
Invertons containing promoters often regulate extracellular products, underscoring the
importance of surface diversity for gut colonization. We found invertons containing promoters
regulating antibiotic resistance genes that shift to the ON orientation after antibiotic treatment
in human metagenomic data and in vitro, thereby mitigating the cost of antibiotic resistance.
We ob served tha t th e orien tatio ns o f some in ver to ns diverge a fter fe cal mic robio ta tran splant,
potentially as a result of individual-specific selective forces.
Phase variation is a process that bacteria
use to generate frequent and reversible
changes within specific hypermutable
loci, introducing phenotypic diversity
into clonal populations. Such phenotypic
diversity plays an important role in mediating
preemptive adaptation to abrupt and severe se-
lective events and is often crucial for infection
by pathogens and colonization by commensals
(1–5). In bacteria, phase variation often manifests
through regions of DNA that invert between two
states in a predictable, reversible manner (6).
The mechanism of inversion involves enzymes
called invertases, which recognize a set of inverted
repeats flanking the invertible DNA region and
catalyze its inversion in a reversible manner (7).
Invertible regions commonly contain promoters
oriented such that in the ON orientation, the pro-
moter is poised to activate transcription of an
operon (7). In the opposite OFF orientation, the
promoter is oriented away from the operon, which
is therefore not transcribed (7). Additional types
of regulatory elements, such as terminators, may
also be contained within these invertible DNA re-
gions (8). Invertases catalyze frequent inversions—
for example, one inversion in every 100 to 1000
Escherichia coli cells, a rate at least three orders
of magnitude higher than the rate of point muta-
tions (7,9,10). Thus, invertible promoters generate
genetic diversity in populations, enabling rapid
and reversible adaptation. Studies in specific
pathogens and commensals have reported in-
vertible promoters that regulate genes involved
in virulence or colonization, such as those that
encode fimbriae, flagella, and capsular polysac-
charides (1,2,4,7,11–17).
Phase variation mediated by DNA inversion
is an underexplored mechanism with broad con-
sequences for adaptation to abrupt and severe
selective events. Here, we sought to systemat-
ically identify invertons, which we define as
single intergenic invertible DNA regions flanked
by inverted repeats likely recognized and in-
verted by invertase proteins in a reversible man-
ner. The term inverton encompasses invertible
promoters and intergenic invertible DNA re-
gions containing alternate types of regulatory
regions. Through our systematic search for in-
vertons, we aimed to address three long-standing
questions regarding this mechanism of regula-
tion: (i) How prevalent are invertons? (ii) What
are the functions of genes regulated by inver-
tons? (iii) In the context of a host, do individual-
specific selective pressures modulate inverton
orientation? We found that invertons are widely
distributedacross bacteria, yielding fundamental
insights into bacterial infection and colonization.
We confirmed and expanded upon previous ob-
servations that the orientations of some invertons
regulating capsular polysaccharide biosynthesis
operons of human gut bacteria are stable within
individuals and divergent between individuals
(17). Using a fecal microbiota transplant (FMT)
study to observe the orientation of invertons
from the same strain in multipleindividuals, we
observed divergences in orientation between
donor and patient. We also identified invertible
promoters regulating antibiotic resistance genes.
We observed that antibiotic treatment results in
ashiftfromtheOFFtoONorientationofthese
invertons in humans, and confirmed that anti-
biotics cause the orientation shift in vitro, which
could mitigate the fitness cost of maintaining anti-
biotic resistance genes inthe absence ofantibiotics.
We developed the PhaseFinder algorithm to
computationally identify invertons and quantify
their orientations with genomic or metagenomic
sequencing reads by identifying regions flanked
by inverted repeats, mimicking their inversion in
silico, and identifying regions where sequencing
reads support both orientations (figs. S1 and S2).
Simulations to benchmark the performance of
PhaseFinder reveal that given enough coverage,
PhaseFinder can identify most invertons without
a substantial rate of false positives. We reasoned
that if inversion rates were high, both orienta-
tions of invertons would coexist, allowing for the
identification of invertons in bacterial popula-
tions used for genome sequencing. Therefore, we
used PhaseFinder to search for invertons in all
available NCBI genomes from RefSeq that were
sequenced using Illumina paired-end sequencing
with data deposited in the NCBI Short Read
Archive. In total, 54,875 bacterial genomes span-
ning the breadth of cultured bacterial diversity
were searched, leading to the discovery of 4686
putative invertons in 2414 genomes (tables S1 and
S2). Invertons were found in 10 of 19 bacterial
phyla. Five phyla harbored invertons in at least
20% of their genomes (Table 1). The lack of a
systematic method to identify invertons was the
impetus for our study; however, the limited scope
of known inverton examples may lead to biases in
the PhaseFinder algorithmagainstinvertonswith
features divergent from known invertible pro-
moters. Additionally, the identification of inver-
tons with PhaseFinder relies on the presence of
both orientations of the inverton in sequenced
samples. Therefore, applying the PhaseFinder
algorithm to additional bacterial genomes de-
rived from diverse conditions and sequenced at
higher coverage or with longer reads will likely
lead to the discovery of many more invertons.
To explore how invertons are distributed across
environmental niches, we used information from
ProGenomes and the Joint Genome Institute
to categorize species into aquatic, terrestrial, and
host-associated habitats (18,19). The prevalence
of invertons was higher in host-associated species
[Fisher exact test, host versus aquatic FDR (false
discovery rate) P=3.5×10
–5
,oddsratio=6.4;host
versus terrestrial FDR P=0.0053,oddsratio=
4.8) (Fig. 1A and table S3). This overall enrichment
in the prevalence of invertons is due to phylum-
level enrichment in Bacteroidetes (FDR P=2.35×
10
–15
)andProteobacteria(FDRP=8.51×10
–5
)
and the fact that all Spirochaetes and Verrucomi-
crobia found with invertons were associated with
vertebrate hosts. Additionally, we observed an
increase in the number of invertons per genome
RESEARCH
Jiang et al., Science 363, 181–187 (2019) 11 January 2019 1 of 7
1
Center for Microbiome Informatics and Therapeutics,
Massachusetts Institute of Technology, Cambridge, MA
02139, USA.
2
Broad Institute of MIT and Harvard,
Cambridge, MA 02142, USA.
3
Center for Computational and
Integrative Biology, Massachusetts General Hospital and
Harvard Medical School, Boston, MA 02114, USA.
4
Department of Biological Engineering, Massachusetts
Institute of Technology, Cambridge, MA 02142, USA.
5
Department of Pathology and Laboratory Medicine,
University of Vermont Medical Center, Burlington, VT 05401,
USA.
6
Division of Gastroenterology and Hepatology,
University of Vermont, Burlington, VT 05401, USA.
7
Gastrointestinal Unit and Center for the Study of Inflammatory
Bowel Disease, Massachusetts General Hospital and Harvard
Medical School, Boston, MA 02114, USA.
*These authors contributed equally to this work.
†Corresponding author. Email: xavier@molbio.mgh.harvard.edu
(R.J.X.); ejalm@mit.edu (E.J.A.)
on January 15, 2019 http://science.sciencemag.org/Downloaded from

in host-associated species (Wilcoxon rank sum
test, host versus aquatic FDR P= 0.00023, W =
361,190; host versus terrestrial FDR P= 0.007,
W = 180,680) (Fig. 1B). The enrichment of in-
vertons in host-associated species is not due to
habitat-specific differences in coverage (fig. S3).
Overall, our results suggest that diverse species
likely use invertons to increase their fitness in
host-associated niches.
We acquired detailed information on the
niches inhabited by species from the phylum
Bacteroidetes (table S4). In Bacteroidetes, the
prevalence of invertons was higher in host gut–
associated species (Fisher exact test, host-gut
versus aquatic FDR P= 3.3 ×10
–35
, odds ratio =
328.4; host-gut versus terrestrial FDR P= 2.3 ×
10
–34
,oddsratio=220.8;host-gutversushost-
other, FDR P=3.9×10
–17
,oddsratio=35.2)(Fig.
1C). The number of invertons per genome was also
higher in host gut–associated isolates (Wilcoxon
rank sum test, host-gut versus aquatic FDR P=
1.8 ×10
–26
,W=6911;host-gutversusterrestrial
FDR P=2.3×10
–26
,W=6967;host-gutversus
host-other FDR P=4.9×10
–13
,W=3956)(Fig.1D).
Because of the observed enrichment for in-
vertons in gut species, we performed an in-depth
analysis and curation of invertons in a non-
redundant selection of 49 representative species
from human stool using longitudinal metage-
nomic data instead of the reads used to assemble
reference genomes (table S5). We identified 459
putative invertons (table S6), 87.6% of which
were from species in the phylum Bacteroidetes,
which had an average of 19 invertons per ge-
nome. We also identified invertons in addi-
tional phyla. We found 53 invertons from two
Akkermansia species (phylum Verrucomicro-
bia), two invertons from a Eubacterium species
(phylum Firmicutes), and one inverton from a
Bifidobacterium species (phylum Actinobacteria)
(fig. S4).
We categorized the invertons according to
their flanking inverted repeats (IR) and identi-
fied four canonical motifs in Bacteroidetes: three
corresponding to known IR motifs in B. fragilis
and one uncharacterized motif (Fig. 2A and tables
S7 and S8) (14). We also identified a distinctive
motif class with tandem repeats within each
Jiang et al., Science 363, 181–187 (2019) 11 January 2019 2 of 7
Table 1. Invertons per phylum identified in a systematic search of bacterial genomes with PhaseFinder.
Phylum Genomes searched Genomes with invertons Percentage of genomes with invertons Total invertons
Acidobacteria 6 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Actinobacteria 5,262 67 1.3 200
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Armatimonadetes 2 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Bacteroidetes 491 160 32.6 1,254
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Chlamydiae 45 3 6.7 5
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Chloroflexi 1 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Cyanobacteria 20 4 20 7
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Deinococcus-Thermus 14 1 7.1 1
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Fibrobacteres 16 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Firmicutes 17,010 986 5.8 1,422
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Fusobacteria 10 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Nitrospirae 1 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Proteobacteria 14,872 1,133 7. 6 1,628
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Spirochaetes 138 57 41.3 140
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Synergistetes 8 2 25 2
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Tenericutes 21 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Thermodesulfobacteria 3 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Thermotogae 7 0 0 0
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Verrucomicrobia 5 1 20 27
.................................... .......................................................... .......................................................... ....................................................... .......................................................... .......................................................... .........
Fig. 1. Prevalence and number of invertons per genome are enriched
in host-associated species. (Aand B) The percentage of genomes
identified with invertons (A) and the number of invertons per genome (B)
from aquatic, terrestrial, and host-associated isolates. (Cand D) In
the phylum Bacteroidetes, the percentage of genomes identified with
invertons (C) and the number of invertons per genome (D) from aquatic,
terrestrial, host sites other than gut, and host gut–associated isolates.
**P≤0.01, ***P≤0.001.
RESEARCH |REPORT
on January 15, 2019 http://science.sciencemag.org/Downloaded from

inverted repeat, which we call motif 0 (fig. S5).
We found conserved promoter consensus mo-
tifs in 98% (231/235) of invertons with IR mo-
tifs 1 to 4 (Fig. 2B) (20). In contrast, promoter
motifs were not observed in any of the inver-
tible regions of IR motif 0–containing sequences
(table S9). Because of the lack of promoter mo-
tifs combined with their location downstream
of operons, invertons containing motif 0 may
function as another type of regulatory element,
such as a phase-variable terminator (8). In gut
Bacteroidetes, we identified a total of 255 in-
vertible promoters. On the basis of the orienta-
tion of the promoter consensus motif in relation
to surrounding genes, we determined which
genes or operons were regulated by invertible
promoters and whether the promoter was in
the ON or OFF orientation with respect to the
downstream gene (tables S7 and S8). In both
species of Akkermansia,allinvertonsareflanked
by inverted repeats with the same moti f (Fig.
2C), contain a promoter motif (Fig. 2D), and
lack upstream invertases, which suggests that
they are all invertible promoters co-regulated
by a single, master invertase (EAJ16_05345 in
Akkermansia muciniphila;MK095134inAk-
kermansia sp. aa_0143).
Through functional annotation, we found that
73% (228/312) of invertible promoters regulate
genes involved in the biosynthesis of polysac-
charides, fimbriae, outer membrane proteins,
and autotransporters; genes involved in the
utilization of polysaccharides; or genes encoding
PEP-CTERM domain–containing proteins (Fig.
2A, tables S7 and S8, and fig. S4). All of these
functional categories except polysaccharide uti-
lization are enriched in genes regulated by in-
vertible promoters (tables S10 and S11). Genes
regulated by invertible promoters are enriched
for cell surface products (e.g., GO:0016020, mem-
brane, P= 1.93 ×10
–14
) (table S11). The most
enriched functional class is capsular polysac-
charide biosynthesis loci, for which we found
at least one example regulated by an inverton in
four of the five major gut phyla: Bacteroidetes,
Actinobacteria, Verrucomicrobia, and Firmicutes
(table S11). Previous studies show that a reper-
toire of phase-variable capsular polysaccharides
is necessary for competitive gut colonization by
Bacteroides species (2,16,21). Our data show
that phase-variable capsular polysaccharide
biosynthesis loci are not just a peculiarity of
Bacteroides species but are likely a convergent
response to a strong selective mechanism in the
vertebrate gut that possibly originates from the
immune response or phages (2,21).
We found that invertible promoters could
regulate antibiotic resistance genes, such as
IBP132, which is upstream of the macrolide
resistance gene ermG in B. stercoris. To in-
vestigate this mechanism in vivo, we searched
specifically for antibiotic resistance genes reg-
ulated by invertible promoters in a cohort of
39 Finnish children, 19 of whom had never been
exposed to antibiotics and 20 of whom had
been administered 9 to 15 antibiotic treat-
ments over a 3-year period (22). By coupling
PhaseFinder with metagenomic assembly anal-
ysis, we found three antibiotic resistance genes
regulated by invertible promoters: (i) the same
ermG macrolide resistance gene as IBP132, (ii) a
cmeABC multidrug resistance cassette conferring
resistance primarily to macrolides and cepha-
losporins, and (iii) pmrEL genes conferring re-
sistance to cationic antimicrobial peptides such
as polymixin B (Fig. 3A). At least one antibiotic
resistance gene regulated by an invertible pro-
moter was found in 38% (15/39) of individuals
from this Finnish cohort: 40% (8/20) of indi-
viduals who were administered antibiotics and
37% (7/19) of individuals who were untreated.
Invertons regulating antibiotic resistance genes
were also detected in metagenomic data from
healthy adults in the United States. All examples
of invertons regulating antibiotic resistance were
found in Bacteroides species, which are increas-
ingly associated with multi–drug-resistant infec-
tions (23). Surprisingly, all cmeABC and ermG
antibiotic resistance genes were regulated by
an identical invertible promoter. On the basis of
genomic context, the cmeABC/ermG invertible
promoter is likely located on an integrative
conjugative element homologous to CTNhyb,
an antibiotic resistance–transmitting mobile
element (fig. S6) (24).
We examined the orientation of the invertible
promoters regulating antibiotic resistance genes
in longitudinal metagenomic data from the
Finnish children. The mean orientation of the
cmeABC/ermG invertible promoter was 94%
OFF in untreated individuals and 84% OFF in
individuals administered antibiotics. We ob-
served an individual in which the cmeABC/ermG
invertible promoter was 99% OFF 7 days before
the macrolide azithromycin was administered
and 74% ON 27 days after treatment (Fig. 3B).
The cmeABC/ermG invertible promoter reverted
to 99% OFF within 5 months after azithromycin
administration (Fig. 3B).Asimilarphenomenon
was observed in a second individual (fig. S7). A
permutation test revealed that macrolide treat-
ment was positively associated with the ON
orientation of the cmeABC/ermG invertible pro-
moter [quantitative polymerase chain reaction
(qPCR) P= 0.0005, metagenomic P= 0.0465].
Thus, it appears that macrolides may select for
the ON orientation of the cmeABC/ermG in-
vertible promoter, and the orientation of the
invertible promoter drifts toward OFF after ces-
sation of antibiotic treatment.
To test whether antibiotics select for resistance
genes with invertible promoters in the ON
orientation, we first verified that the genes
regulated by the cmeABC/ermG invertible pro-
moter confer macrolide resistance. We cultivated
13 B. stercoris isolates with and one isolate
without the cmeABC/ermG invertible promoter
upstream of the macrolide resistance gene ermG
(table S12) (25). The invertible promoter was
primarily ON (>75%) in 10 isolates derived from
erythromycin-containing media and primarily
OFF (>97%) in three isolates derived from media
Jiang et al., Science 363, 181–187 (2019) 11 January 2019 3 of 7
Fig. 2. Motifs found in the
inverted repeats of Bacte-
roidetes and Akkermansia
invertons consist of five to
seven base pair palindromes
with two or three intervening
base pairs. (A)Functional
profiling of operons regulated
by invertons reveals specializa-
tions of each inverted repeat (IR)
motif. The heat map represents
the number of operons per func-
tional class regulated by invertons
with either global or local inver-
tases. CPS, capsular poly-
saccharide; Fimb, fimbriae; OmpA,
outer membrane protein A;
SusCD, starch utilization system
proteins C and D. Superscript I
indicates the presence of a local invertase. The absence of local invertases directly upstream suggests that IR motifs 2 and 4 are likely regulated by global
invertases. (B) Promoter motif identified from invertons from Bacteroidetes species. (C) The inverted repeat motif found in all identified invertons from
Akkermansia spp. (D)PromotermotifidentifiedfrominvertonsfromAkkermansia spp.
RESEARCH |REPORT
on January 15, 2019 http://science.sciencemag.org/Downloaded from

without erythromycin (Fig. 3C). We established
that all B. stercoris isolates with ermG regu-
lated by the cmeABC/ermG invertible promoter
were resistant to erythromycin, whereas the
isolate without the ermG gene was susceptible
(fig. S8).
Next, we showed that erythromycin treatment
selects for the ON orientation of the cmeABC/
ermG invertible promoter. To quantify changes
in the relative abund ances of cells with the
cmeABC/ermG invertible promoter in ON and
OFF orientations, we performed qPCR compar-
ing relative amplification using a static primer
downstream of the invertible promoter paired
with either a primer that amplifies the ON ori-
entation or one that amplifies the OFF orienta-
tion (fig. S9). We transferred B. stercoris isolates
with the cmeABC/ermG invertible promoter
Jiang et al., Science 363, 181–187 (2019) 11 January 2019 4 of 7
Fig. 3. Invertible pro-
moters regulate anti-
biotic resistance
genes. (A) Three
classes of genes
conferring antibiotic
resistance are regu-
lated by invertible pro-
moters. The genomic
contexts of the loci are
shown with antibiotic
resistance genes
colored. Promoters are
designated by hooked
arrows; purple triangles
represent inverted
repeats. (B) An invert-
ible promoter regulat-
ing the cmeABC
multidrug efflux
cassette in individual
E011878 was 99% OFF
before antibiotic treat-
ment, shifted to 74%
ON 7 days afterward,
and then drifted back to
99% OFF over 5 months
as measured by both
qPCR and metagenomic
(MGX) data. (C)The
cmeABC/ermG invertible
promoter is oriented
ON in B. stercoris
isolated in medium
containing erythromycin
(Erm+) and OFF in
B. stercoris isolated in
medium without erythro-
mycin (Erm–). (D)The
orientation of the
cmeABC/ermG invertible
promoter under
antibiotic selection.
Isolate s11 grown in
Erm–medium remained
OFF, whereas the same
isolate transferred to
Erm+ medium shifted to
ON. Isolate s1 remained
ON in Erm–and Erm+
media. (E)Growthof
isolates under antibiotic
selection. Isolate s1 grew
in Erm+ medium without
a lag phase relative to Erm–medium, whereas isolate s11 grew in Erm+ medium after an extended lag phase, consistent with a lower number
of initially erythromycin-resistant cells. (F) Kinetics of the cmeABC/ermG invertible promoter. An isolate with the cmeABC/ermG invertible promoter
initially in the OFF orientation, s12, was exposed to erythromycin. Half of the culture was then propagated in the presence of erythromycin and the
other half in the absence of erythromycin for 24 1:1000 dilution transfers. In Erm+ medium, the promoter remained ON, whereas in Erm–medium, it
drifted toward OFF. Error bars in (D) and (E) denote SD.
RESEARCH |REPORT
on January 15, 2019 http://science.sciencemag.org/Downloaded from

Jiang et al., Science 363, 181–187 (2019) 11 January 2019 5 of 7
Fig. 4. The orientations of invertons are generally stable within individ-
uals but divergent between individuals. (A)Examplesofinvertonsfrom
donor “am”found in the forward (F: IBP106, IBP124) or reverse (R: IBP60)
orientation. (B)Examplesofinvertonsfromdonor“am”(IBP129, IBP212, and
IBP394) that are longitudinally stable. (C)Examplesofinvertonsfromdonor
“am”that are longitudinally unstable (IBP16, IBP199, and IBP353). (D)Selected
examples of invertible promoters with different orientations among four
individuals (“ae,”“am,”“an,”and “ao”). Numbers in each heat map cell represent
the average counts in the Fand R orientations (F:R); red and blue indicate Fand
Rorientation,respectively.(E,G,andI)Phylogenetictreesproducedby
StrainPhlAn from metagenomic data from FMT donor (green) and patient
(orange, pre-FMT; purple, post-FMT) as well as isolate genomes from the donor
(green) and unrelated reference genomes from the same species (black). The
phylogenetic trees demonstrate engraftment and persistence of B. fragilis (E),
B. ovatus (G), and B. vulgatus (I) strains from the donor to the patient.
Phylogenetic tree legends are the number of nucleotide substitutions per site.
(F,H,andJ)Examplesofdivergenceofinvertiblepromoterorientationsafter
engraftment in a patient. Black circles denote invertible promoters whose
orientation was significantly different between donor and patient after FMT.
White circles denote invertible promoters whose orientation is not significantly
different (Wilcoxon rank sum test, FDR P<0.05).Numbersineachheatmap
cell represent the counts in theFand R orientations (F:R). (F) After engraftment,
the orientations of IBP183 and IBP198 were reversed relative to the donor.
(H) A strain of B. ovatus was present in Patient 14 before FMT but was
outcompeted by the donor strain. After engraftment, the orientation of
IBP155 and IBP167 was reversed relative to the donor. (J) After B. vulgatus
engraftment, the orientation of IBP121 was only found in the F orientation,
but over the course of 135 days, the orientation reversed.
RESEARCH |REPORT
on January 15, 2019 http://science.sciencemag.org/Downloaded from

oriented either primarily ON or OFF into media
with (Erm+) or without (Erm–) erythromycin
and quantified the percentage of cells in the
ON or OFF orientation after 24 hours (Fig. 3D).
The OFF cultures grew in Erm+ medium only
after an extended lag phase relative to growth in
Erm–medium, whereas ON cultures grew sim-
ilarly in Erm+ and Erm–media (Fig. 3E). OFF
isolates grown in Erm+ medium became pre-
dominantly ON, whereas OFF isolates grown in
Erm–medium remained OFF. ON isolates re-
mained predominantly ON in both Erm+ and
Erm–media (Fig. 3D).
Finally, during serial transfers in both Erm+
and Erm–media, we monitored the ON:OFF
ratio of the cmeABC/ermG invertible promoter
in a B. stercoris OFF isolate that had previously
been switched to ON in Erm+ medium. Over
the course of 24 transfers at 1:1000 dilution, the
orientation of the cmeABC/ermG invertible pro-
moter remained ON in Erm+ medium but grad-
ually drifted away from 100% ON in Erm–medium,
recapitulating the in vivo observations from
metagenomic data (Fig. 3F).
Although strongly favored in the presence of
antibiotics, high expression of antibiotic resist-
ance genes likely incurs a substantial fitness
cost, which could explain the reversion to the
OFF orientation after antibiotic treatment. Many
compensatory mechanisms to maintain anti-
biotic resistance in the absence of antibiotics
have been noted (26), but invertons are akin to
catastrophe insurance; a certain percentage of
the population is always prepared to resist fu-
ture antibiotic treatment and reintroduce het-
erogeneity after antibiotic selection.
Dense longitudinal metagenomic data allow
for a detailed view of the dynamics of invertons
over time in the human gut. We analyzed a
dataset of samples from 54 individuals, four of
whom (“ae,”“am,”“an,”“ao”) were sampled
densely over 5 to 18 months, and tracked the
orientations of invertons. The F orientation is the
same orientation of the inverton in the reference
genome, whereas the R orientation is the op-
posite orientation of the inverton in the ref-
erence genome. We identified 423 invertons with
sufficient coverage to track their temporal dy-
namics in these individuals. Of these, 322 were
predominantly found in one orientation within
an individual with little or no fluctuation (mean
> 95% and min > 75% for either the F or R ori-
entation) (Fig. 4A); the orientations of 59 were
relatively stable (max-min %R ≤50%) within an
individual (Fig. 4B), whereas the orientations of
42 were unstable (max-min %R > 50%) within
an individual (Fig. 4C and table S13).
Although the orientations of 90% of invertons
in the same individual were relatively stable over
time, the orientations between individuals varied
extensively (Fig. 4D and fig. S10). The mean %R
orientation of 214 out of 423 of the invertons
varied by more than 50% between individuals.
In 122 examples, averaging across time, the inver-
ton was predominantly (>95%) in the F orienta-
tion in at least one individual and predominantly
(>95%) in the R orientation in another. Addi-
tionally, 119 out of 238 invertons were signifi-
cantly different (Kruskal-Wallis H test, FDR P<
0.05) between the four individuals for whom we
had dense longitudinal metagenomic data. The
differences in the orientations of invertons be-
tween individuals could be explained by divergent
selective forces between individuals, different
optimal orientations between strains, or sto-
chastic variation in orientation.
Sculpted by the individual’s diet, lifestyle, im-
mune response, and genetics, the gut of every
individual is a distinctive environment for res-
ident bacteria. We observed the influence of
the individual on the orientations of invertons
in a cohort of patients with ulcerative colitis
who were the recipients of FMT from health y
donors. The source of the fecal microbiota was
donor “am,”whose longitudinal metagenomic
data were analyzed above. Therefore, we could
monitor the orientations of invertons from the
same strain for 18 months in a healthy donor
and up to 5 months in the patients.
First, we identified strains from the donor
that engrafted into the patient’smicrobiome.To
identify engraftment, we found cases where the
same strain was present in the donor and in at
least one patient after FMT, but absent in the
same patient(s) before FMT (Fig. 4, E, G, and I,
and fig. S11). We found three high-abundance
species with invertons from the donor that en-
grafted in patients: B. fragilis,B. vulgatus, and
B. ovatus.
Then, we compared the orientations of in-
vertons from the donor and patient after en-
graftment and found that the orientations of
42.8% of invertons (48/112) diverged from the
donor orientation (Wilcoxon rank sum test, FDR
P<0.05)(Fig.4,F,H,andJ,andfig.S12).In
B. fragilis,twoinvertons(IBP183andIBP198)
engrafted in the opposite orientation of the
donor strain and remained predominantly (87.2%
and 87.1%) in the R orientation, but drifted
toward F near the end ofthe sampling (Fig.4F).For
B. ovatus, a strain existed in Patient 014 b efore
FMT but was replaced by the donor B. ovatus
strain (Fig. 4G; compare orange to purple). IBP155
invertons from both the donor strain and pre-
FMT strain were predominantly (90.1% and
100%) in the R orientation, whereas the newly
engrafted strain was oriented entirely in the F
orientation and remained in the F orienta-
tion over the course of sampling (Fig. 4H). In
B. vulgatus, an inverton was initially present
completely in the F orientation but over the
course of 145 days completely reversed to the
R orientation (Fig. 4J, IBP121). In addition to
the invertons that reversed their orientations, we
also identified examples of invertons that main-
tained their orientations (Fig. 4F, IBP189; Fig. 4H,
IBP166; Fig. 4J, IBP125).
Our findings highlight the role of invertons
in host-microbe coexistence. Genes regulated
by invertons were highly enriched for products
located on the exterior of the cell where they are
exposed to the host immune system and phages,
indicating that they may be primary targets for
selection that are beneficial for gut commensals
to diversify their surface architectures or es-
sential processes, such as antibiotic resistance,
whose expression has a high fitness cost. The
high prevalence of antibiotic resistance genes
regulated by invertons containing promoters
suggests that this is an example of bet-hedging
(27). This could lead to longer persistence of
antibiotic resistance genes in a microbial com-
munity, further increasing the burden to com-
bat antibiotic resistance. Our results indicate
that in the human gut, invertons help bacterial
populations regain heterogeneity after bottle-
necks encountered during colonization of a
new host or severe perturbations. Overall, our
study provides insights into a mechanism al-
lowing adaptive tradeoffs in bacteria that have
evolved to successfully colonize host-associated
niches.
REFERENCES AND NOTES
1. M. J. Coyne, M. Chatzidaki-Livanis, L. C. Paoletti,
L. E. Comstock, Proc. Natl. Acad. Sci. U.S.A. 105, 13099–13104
(2008).
2. N. T. Porter, P. Canales, D. A. Peterson, E. C. Martens,
Cell Host Microbe 22, 494–506.e8 (2017).
3. J. C. Gauntlett, H.-O. Nilsson, A. Fulurija, B. J. Marshall,
M. Benghezal, Gut Pathog. 6, 35 (2014).
4. M. Taketani, M. S. Donia, A. N. Jacobson, J. D. Lambris,
M. A. Fischbach, mBio 6, e01339-15 (2015).
5. J. K. Lim et al., Infect. Immun. 66, 3303–3310 (1998).
6. I. R. Henderson, P. Owen, J. P. Nataro, Mol. Microbiol. 33,
919–932 (1999).
7. M. J. Coyne, K. G. Weinacht, C. M. Krinos, L. E. Comstock,
Proc. Natl. Acad. Sci. U.S.A. 100, 10446–10451 (2003).
8. P. Hinde, P. Deighan, C. J. Dorman, J. Bacteriol. 187,
8256–8266 (2005).
9. T. H. Kawula, P. E. Orndorff, J. Bacteriol. 173, 4116–4123
(1991).
10. C. D. Bayliss, FEMS Microbiol. Rev. 33, 504–520
(2009).
11. O. Sekulovic et al., PLOS Genet. 14, e1007332 (2018).
12. M. Silverman, M. Simon, Cell 19, 845–854 (1980).
13. J. M. Abraham, C. S. Freitag, J. R. Clements, B. I. Eisenstein,
Proc. Natl. Acad. Sci. U.S.A. 82, 5724–5727 (1985).
14. H. Nakayama-Imaohji et al., J. Bacteriol. 191, 6003–6011
(2009).
15. H. Nak ayama-Imao hji et al., PLOS ONE 11, e014888 7
(2016).
16. E. B. Troy, V. J. Carey, D. L. Kasper, L. E. Comstock,
J. Bacteriol. 192, 5832–5836 (2010).
17. N. L. Zitomersky, M. J. Coyne, L. E. Comstock, Infect. Immun.
79, 2012–2020 (2011).
18. D. R. Mende et al., Nucleic Acids Res. 45, D529–D534
(2017).
19. I. V. Grigoriev et al., Nucleic Acids Res. 40, D26–D32
(2012).
20. D. P. Bayley, E. R. Rocha, C. J. Smith, FEMS Microbiol. Lett.
193, 149–154 (2000).
21. C. H. Liu, S. M. Lee, J. M. Vanlare, D. L. Kasper,
S. K. Mazmanian, Proc. Natl. Acad. Sci. U.S.A. 105, 3951–3956
(2008).
22. M. Yassour et al., Sci. Transl. Med. 8, 343ra81 (2016).
23. G. N. Hartmeyer, J. Sóki, E. Nagy, U. S. Justesen,
J. Med. Microbiol. 61, 1784–1788 (2012).
24. F. Husa in et al., Mob. Genet. Elements 4, e29801
(2014).
25. M. Monod, S. Mohan, D. Dubnau, J. Bacteriol. 169, 340–350
(1987).
26. D. I. Andersson, D. Hughes, FEMS Microbiol. Rev. 35, 901–911
(2011).
27. T. Philippi, J. Seger, Trends Ecol. Evol. 4, 41–44 (1989).
ACKNO WLEDG MENTS
We thank T. Poon, T. Vatanen, and M. Groussin for their assistance
in obtaining samples and data; T. Reimels for comments on the
manuscript; and the Microbial Omics Core of the Broad Institute
and the Broad Technology Labs for their assistance with DNA
Jiang et al., Science 363, 181–187 (2019) 11 January 2019 6 of 7
RESEARCH |REPORT
on January 15, 2019 http://science.sciencemag.org/Downloaded from

extraction, library preparation, and sequencing. The FMT research
was approved by the Committee on Human Research in the
Medical Sciences (CHRMS) (CHRMS 15-373). Funding: Supported
by NIH grants DK043351 and AT009708, the Cro hn’s and
Colitis Foundat ion of America, the Juven ile Diabetes Research
Foundation, and the Center for Mi crobiome Informatics an d
Therapeutics at MIT (R.J.X.). A .B.H. is a Merck Fellow of
the Helen Hay Whitn ey Foundation. Author con tributions:
Conceptualization, X.J., A.B.H., T.D.A., H .V., A.C.T., E.J.A., and
R.J.X.; methodol ogy, X.J., A.B.H., D.R.P., C.T.C. , and T.D.A.;
software, X.J. and A.B.H.; validation, X.J. and A.B.H.; formal analysis,
X.J., A.B.H., D.R.P., C.T.C., and T.D.A.; investigation, X.J., A.B.H.,
and T.D.A.; resources, M.P., P.L.M., and J.C. ; data cur ation, X.J.,
A.B.H., and T.D.A.; writing (original draft preparation), X.J. and
A.B.H., writing (review and editing), X.J., A.B.H., H.V., A.C.T.,
E.J.A., and R.J.X., visualization, X.J. and A.B.H.; supervision,
H.V., A.C.T., E.J.A., and R.J.X.; project administration and
funding acquisition, E.J.A. and R.J.X. Competing i nterests:
E.J.A. is a co-founder and shareholder of Finch Therapeutics.
J.C. consults for Finch Therapeutics. P.L.M. is currently
employed by Takeda Pharmaceuticals International. Data and
materials availability: Data used in the study are available
from the NCBI. Isolate genomes: PRJNA496358. Dense
longitudinal metagenomic data: PRJNA503484, FMT
metagenomic data: PRJ NA474024. PhaseFinder is available
from GitHub: https://github.com/XiaofangJ/PhaseFinder.
SUPPLEMENTARY MATERIALS
www.sciencemag.org/content/363/6423/181/suppl/DC1
Materials and Methods
Figs. S1 to S12
Tables S1 to S16
References (28–50)
19 June 2018; accepted 3 December 2018
10.1126/science.aau5238
Jiang et al., Science 363, 181–187 (2019) 11 January 2019 7 of 7
RESEARCH |REPORT
on January 15, 2019 http://science.sciencemag.org/Downloaded from

adaptation in the gut
Invertible promoters mediate bacterial phase variation, antibiotic resistance, and host
Crothers, Peter L. Moses, Andrew C. Tolonen, Hera Vlamakis, Eric J. Alm and Ramnik J. Xavier
Xiaofang Jiang, A. Brantley Hall, Timothy D. Arthur, Damian R. Plichta, Christian T. Covington, Mathilde Poyet, Jessica
DOI: 10.1126/science.aau5238
(6423), 181-187.363Science
, this issue p. 181Science
environmental stress, including antibiotic exposure.
Bacteroidetes, Spirochaetes, and Verrucomicrobia. These bacteria are thus equipped and prepared for sudden
associated organisms, including−promoters linked to antibiotic resistance genes were widespread among vertebrate gut
promoters, which can flip between ON and OFF states catalyzed by phage integrase analogs called invertases. Invertible
bacterial genomes for invertible promoters that cause phase variation. Inverted repeats signal the presence of these
developed an algorithm to survey et al.phenotypes arise by reversible mechanisms called phase variation. Jiang
Clonal bacterial colonies will often grow dissimilar patches, similar to a tortoiseshell pattern. These differing
Switching ON resistance
ARTICLE TOOLS http://science.sciencemag.org/content/363/6423/181
MATERIALS
SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2019/01/09/363.6423.181.DC1
CONTENT
RELATED
http://stm.sciencemag.org/content/scitransmed/5/204/204ra132.full
http://stm.sciencemag.org/content/scitransmed/8/327/327ra25.full
http://stm.sciencemag.org/content/scitransmed/8/343/343ra81.full
http://stm.sciencemag.org/content/scitransmed/7/297/297ra114.full
REFERENCES
http://science.sciencemag.org/content/363/6423/181#BIBL
This article cites 48 articles, 19 of which you can access for free
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Terms of ServiceUse of this article is subject to the
is a registered trademark of AAAS.Science
licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. The title
Science, 1200 New York Avenue NW, Washington, DC 20005. 2017 © The Authors, some rights reserved; exclusive
(print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience
on January 15, 2019 http://science.sciencemag.org/Downloaded from