Sialic acid catabolism drives intestinal inflammation and microbial dysbiosis in mice

Abstract

Rapid shifts in microbial composition frequently occur during intestinal inflammation, but the mechanisms underlying such changes remain elusive. Here we demonstrate that an increased caecal sialidase activity is critical in conferring a growth advantage for some bacteria including Escherichia coli (E. coli) during intestinal inflammation in mice. This sialidase activity originates among others from Bacteroides vulgatus, whose intestinal levels expand after dextran sulphate sodium administration. Increased sialidase activity mediates the release of sialic acid from intestinal tissue, which promotes the outgrowth of E. coli during inflammation. The outburst of E. coli likely exacerbates the inflammatory response by stimulating the production of pro-inflammatory cytokines by intestinal dendritic cells. Oral administration of a sialidase inhibitor and low levels of intestinal α2,3-linked sialic acid decrease E. coli outgrowth and the severity of colitis in mice. Regulation of sialic acid catabolism opens new perspectives for the treatment of intestinal inflammation as manifested by E. coli dysbiosis.

ARTICLE

Received 6 Feb 2015 | Accepted 22 Jul 2015 | Published 25 Aug 2015

Rapid shifts in microbial composition frequently occur during intestinal inammation, but the mechanisms underlying such changes remain elusive. Here we demonstrate that an increased caecal sialidase activity is critical in conferring a growth advantage for some bacteria including Escherichia coli (E. coli) during intestinal inammation in mice. This sialidase activity originates among others from Bacteroides vulgatus, whose intestinal levels expand after dextran sulphate sodium administration. Increased sialidase activity mediates the release of sialic acid from intestinal tissue, which promotes the outgrowth of E. coli during inammation. The outburst of E. coli likely exacerbates the inammatory response by stimulating the production of pro-inammatory cytokines by intestinal dendritic cells. Oral administration of a sialidase inhibitor and low levels of intestinal a2,3-linked sialic acid decrease E. coli outgrowth and the severity of colitis in mice. Regulation of sialic acid catabolism opens new perspectives for the treatment of intestinal inammation as manifested by E. coli dysbiosis.

DOI: 10.1038/ncomms9141 OPEN

Sialic acid catabolism drives intestinal inammation and microbial dysbiosis in mice

Yen-Lin Huang1, Christophe Chassard2,w, Martin Hausmann3, Mark von Itzstein4 & Thierry Hennet1

1 Institute of Physiology and Zurich Center of Integrative Human Physiology, University of Zurich, Zurich CH-8057, Switzerland. 2 Laboratory of Food Biotechnology, Institute of Food, Nutrition and Health, ETH Zurich, Zurich CH-8092, Switzerland. 3 Division of Gastroenterology and Hepatology, University Hospital of Zurich, Zurich CH-8006, Switzerland. 4 Institute for Glycomics, Grifth University, Gold Coast Campus, Gold Coast, Queensland 4222, Australia.

w Present address: Institut National de la Recherche Agronomique, UR 545 URF, 15000 Aurillac, France. Correspondence and requests for materials should be addressed to T.H. (email: mailto:[email protected]

Web End [email protected] ).

NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 1

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141

The intestinal microbiota has emerged as a key player in the regulation of physiological pathways and in the development of diseases. Along with intestinal diseases, such as

necrotizing enterocolitis1 and inammatory bowel disease2, gut microbiota contribute among others to the aetiology of diabetes3, asthma4, autoimmunity5 and cancer6. Accordingly, much effort has been dedicated in understanding the factors inuencing the composition of the intestinal microbiota to maintain or restore health in the host organism.

Carbohydrates are a major class of food products that profoundly affect the gut microbiota. Whereas most monosaccharides are absorbed by the small intestine, oligo- and polysaccharides are not digested in the upper gastrointestinal tract and reach the colon intact. The impact of complex carbohydrates on microbial composition is based on the expression of specic hydrolases7, which enable some bacterial species to process and utilize breakdown products as nutrients, thereby conferring a proliferative advantage over bacteria that cannot process complex carbohydrates8. The rst carbohydrates ingested just after birth are provided by breast milk, which is a rich source of lactose and oligosaccharides9. The uptake of milk oligosaccharides coincides with the microbial colonization of the gut and favors the proliferation of bacteria equipped with carbohydrate-processing enzymes, such as Bidobacterium and Bacteroides spp. that are enriched in breast-fed infants10.

In addition to food carbohydrates, several intestinal bacteria can process host-derived carbohydrates, which are prominent constituents of mucosal layers. Besides providing carbon sources for bacterial growth, released host-derived carbohydrates inuence gene expression in the microbiota, thereby affecting the virulence of pathogenic bacteria as demonstrated by the regulation of virulence factors in enterohaemorrhagic Escherichia coli by fucose11. Other host-derived carbohydrates, such as sialic acids, are taken up by bacteria lacking de novo biosynthetic pathways for these sugars, and incorporated into bacterial capsule and lipooligosaccharides12. The decoration of bacterial glycoconjugates with sialic acid protects microbes from recognition by the host immune system13 and regulates the host immune response through interactions with sialic acid-binding lectins14. Finally, the interplay between intestinal microbiota and host glycosylation is not limited to the utilization of host glycans by bacteria. Sialic acids as terminal residues on intestinal glycoconjugates are a prime target for bacterial adhesins and toxins from Vibrio cholerae, Helicobacter pylori and E. coli15,16.

The structural complexity of carbohydrates, either ingested in the form of milk oligosaccharides or expressed as host-derived glycans, hampers the elucidation of their impact on the gut microbiota. Accordingly, little is known about the relevance of specic carbohydrates on microbiota composition and on intestinal physiology. The application of mice decient for glycosyltransferases enables the investigation of interactions between dened carbohydrates, intestinal microbes and the host immunity. For example, a study of a1,2-fucosyltransferase

Fut2 knockout mice has recently demonstrated the interplay between fucosylated glycans and diet polysaccharides on shaping the gut microbiota17. The study of a2,3 sialyltransferase St3gal4 knockout (ST) mice, which mediates a2,3-sialyllactose (3SL) synthesis in mammary gland, has established the role of the milk oligosaccharide on the gut microbiota and thereby on the susceptibility of mice in dextran sulphate sodium (DSS)-induced acute18 and chronic colitis19. Through the investigation of DSS-mediated colitis in ST mice and the modulation of the intestinal microbiota by selective antibiotic treatment, the present study reveals the critical role of a2,3-linked sialic acid in establishing a niche for intestinal E. coli after lactation and during intestinal inammation.

ResultsGut microbiota change during DSS-induced colitis. To unravel the relationship between a2,3-linked sialic acid and the intestinal microbiota, and to identify the mechanisms of a2,3-linked sialic acid effects on colitis development, we have rst addressed the impact of intestinal bacterial groups on colitis by treating mice with a panel of antibiotics. Correlations between the resulting changes in microbial composition and susceptibility to DSS-mediated colitis pointed to specic bacterial families possibly regulating the severity of colitis in wild-type (WT) and ST mice. In fact, the composition of colonic bacteria in WT and ST mice differed at the adult stage. The most abundant bacterial family in WT mice was Ruminococcaceae, reaching 44% of total bacteria. By contrast, Ruminococcaceae only represented 10% of colonic bacteria in ST mice, whereas Porphyromonadaceae dominated by reaching 37% (Fig. 1a). The bacterial composition of mice undergoing intestinal inammation induced by DSS changed markedly, as seen by a strong expansion of Bacteroidaceae and Enterobacteriaceae in WT mice. ST mice, which were less susceptible to DSS than WT mice18, also showed increased Bacteroidaceae levels during DSS challenge, whereas

a b

Relative abundance of bacterial family (%)

100

90 80 70 60 50 40 30 20 10

Alcaligenaceae Anaeroplasmataceae Bacteroidaceae Enterobacteriaceae Erysipelotrichaceae Helicobacteraceae Lachnospiraceae Lactobacillaceae Porphyromonadaceae Prevotellaceae Rikenellaceae Ruminococcaceae

Relative abundance of bacterial genus (%)

100

90 80 70 60 50 40 30 20 10

0

Alistipes Anaeroplasma Bacteroides BarnesiellaBlautia Coprobacillus Escherichia/shigella Helicobacter Lactobacillus Mucispirillum Oscillibacter Parabacteroides Parasutterella Prevotella

0 WT WTDSS STDSS

ST WT WTDSS STDSS

ST

Figure 1 | Bacterial composition in mice with DSS-induced colitis. (a) 16S rRNA pyrosequencing analysis of faecal microbial taxa families in control and DSS-treated WT and ST mice (at day 8 after DSS addition). (b) Pyrosequencing analysis of faecal microbial taxa at the genus level. Data show the average percentage of total identied sequences obtained from a pool of eight mice per group. Only the bacterial taxa representing at least 1% of total identied sequences are presented.

2 NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141 ARTICLE

Enterobacteriaceae remained at low level. Sequence analysis at the genus level indicated that the Escherichia and Shigella accounted for the observed increase of Enterobacteriaceae in WT mice, and Bacteroides accounted for the increase of Bacteroidaceae in both WT and ST mice under DSS challenge (Fig. 1b).

Antibiotics effect on DSS-induced colitis. To determine whether a specic group of bacteria accounted for the different response to DSS, we treated WT mice with a panel of antibiotics before DSS challenge. Vancomycin, neomycin and penicillin exacerbated the severity of DSS-induced colitis as monitored by loss of body weight (Fig. 2a). Streptomycin was the only antibiotic that attenuated the loss of body weight during colitis, whereas chloramphenicol and metronidazole did not have much impact on the course of the inammatory response (Fig. 2a). To exclude any damaging effect caused by the use of antibiotics on intestinal barrier function, we assessed epithelial permeability by measuring the leakage of orally administered uorescein isothiocyanate

(FITC)dextran into the bloodstream. Permeability was only signicantly increased after DSS ingestion, but not after antibiotic treatment (Fig. 2b). We also tested the effect of a short-term treatment with antibiotics to exclude possible adaptations of the host mucosa to three weeks of altered microbiota composition. Focusing on vancomycin (Fig. 2c) and streptomycin (Fig. 2d), we could reproduce the worsening and improving effects of these two antibiotics by only administering vancomycin and streptomycin during DSS challenge. The protective effect of streptomycin towards DSS challenge was even more pronounced in a short-term treatment compared with a 3-week pre-treatment in both WT and ST mice. The extent of intestinal inammation was conrmed by measuring the length of the colon in treated mice. The shortening of colon length induced by DSS ingestion was aggravated by vancomycin and reduced by streptomycin (Fig. 2e). The impact of vancomycin and streptomycin treatment on the intestinal microbiota of WT mice was analysed by 16S rRNA pyrosequencing and compared with the changes observed

a b

WTDSS

15

10

5

WTDSS + Abx

*

Body weight

120 110 100

90 80 70

* * *

* * * * * *

(% of control)

FITCdextran

(g ml1 serum)

Van Van

Neo Pen Stp Stp

CAM Met

0 WT

WTDSS

c

d

Body weight (%)

110

105

100

95

90

85

80

75

110 100

90 80 70 60 50

70 0 1 2 3 4 5 6 7

WT

WTVan

STVan

ST

Body weight (%)

110

105

100

95

90

85

80

75

WT

WTStp

STStp

ST

70 0 1 2 3 4 5 6 7 8

Time (days)

Time (days)

e f

Colon length (%)

DSS

Antibiotic

Relative abundance of bacterial

family (%)

100

90 80 70 60 50 40 30 20 10

0

Alcaligenaceae Anaeroplasmataceae Bacteroidaceae Enterobacteriaceae Erysipelotrichaceae Helicobacteraceae Lachnospiraceae Lactobacillaceae Porphyromonadaceae Prevotellaceae Rikenellaceae Ruminococcaceae

+ + + + + +

Stp Stp

Van Van

WT WT

ST

Stp

Van

WTDSS

Figure 2 | Antibiotics effect on DSS-induced colitis. (a) Mice were treated with vancomycin (Van), neomycin (Neo), penicillin (Pen), streptomycin (Stp), chloramphenicol (CAM) and metronidazole (Met) for 3 weeks before DSS challenge for 5 days. Body weight was measured by day 8 after initiation of DSS challenge and given as percentage to the body weight of mice challenged with DSS without antibiotics. The data are represented as means.e.m. N 68,

*Po0.05 (two-tailed Students t-test). (b) Intestinal permeability was measured by FITCdextran levels in the serum from control, colitogenic mice on day 5 of DSS challenge, and 3 weeks of antibiotic pretreated mice. N 5, *Po0.05 (ANOVA, Bonferronis multiple comparison test). (c) Relative change in

body weight of WTand STmice treated with 0.5 g l 1 Van for 7 days and 3% DSS for 5 days; control mice received DSS without Van. (d) Relative change in body weight of WT and ST mice treated with 1 g l 1 Stp for 8 days and 3% DSS for 5 days. (e) Colon length was determined at the end point of DSS treatment. In (ce) the data are represented as means.e.m. from two independent experiments, N 6-8, *Po0.05 (ANOVA, Bonferronis multiple

comparison test). (f) 16S rRNA pyrosequencing analysis of faecal microbial taxa families in untreated WT mice, in Van, Stp-treated WT mice and DSS-challenged WT mice (WTDSS). The data are represented as the percentage of total identied sequences obtained from a pool of eight mice per group.

NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 3

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141

during DSS challenge. Vancomycin induced a strong increase of Enterobacteriaceae, which raised to 27% of total bacteria, whereas Enterobacteriaceae remained below 0.1% of total bacteria in streptomycin-treated mice (Fig. 2f). Under both antibiotics, Bacteroidaceae expanded to represent the major bacterial family, but the increase in Bacteroidaceae did not directly correlate with the severity of DSS-induced colitis. In contrast, the abundance of Enterobacteriaceae correlated with the magnitude of colitis. Overgrowth of Enterobacteriaceae and several Bacteroidaceae spp. during intestinal inammation is well documented, although the mechanisms underlying such expansions have not been identied in previous studies20,21.

Expansion of E. coli during DSS-induced colitis. To verify which species of Enterobacteriaceae expanded during intestinal inammation, we applied specic primers targeting the b-glucuronidase uidA gene22 and conrmed a signicant increase of E. coli in DSS-challenged mice and in vancomycin-treated mice (Fig. 3a). In contrast, a signicant reduction of E. coli was observed in both WT and ST mice treated with streptomycin. Overall, E. coli levels correlated with the severity of colitis in all genotypes and antibiotic treatments tested. Of note, the level ofE. coli in adult ST mice was 2 orders of magnitude lower than in WT mice, indicating that decreased exposure to a2,3-linked sialic acid, because of reduced sialylation of host glycans and the absence of 3SL in milk ingested during lactation, was accompanied by low level of intestinal E. coli. The importance of milk 3SL during lactation at promoting the low-level colonization of E. coli was also visible in WT mice that were fostered by ST mother during lactation. WT mice fed on 3SL-decient milk (WTXF) showed lower E. coli levels than littermates fed on normal milk. Similarly, ST mice fed on normal milk showed elevated E. coli levels compared with littermates fed on 3SL-decient milk (Fig. 3b). The relative abundance of intestinalE. coli in cross-fostered mice also reected the severity of DSS-induced colitis18. Overall, these data conrmed that neonatal exposure to milk 3SL contributed to establishing an intestinal niche for E. coli, thereby providing a ground for the subsequent expansion of E. coli during colitis induced by DSS.

Exposure to sialic acid promotes E. coli expansion. To address the impact of 3SL on E. coli proliferation, we isolated various strains of commensal E. coli from the colon of WT mice during DSS-induced colitis. The identity of the isolated bacteria withE. coli was conrmed by sequencing universal stress protein uspA gene23, gyrase gyrB gene24 and by biochemical testing using the

API-20E Enterobacteriaceae detection system. The culture of the isolated E. coli strain EHV2 in minimal medium containing unique monosaccharides as carbon source conrmed that N-acetylneuraminic acid was a preferential source of energy forE. coli (Fig. 4a). In contrast to free N-acetylneuraminic acid, the milk oligosaccharides 3SL and a2,6-sialyllactose (6SL) did not support E. coli growth in vitro (Fig. 4b). Proliferation could, however, be restored by adding sterile-ltered caecal uid from WT mice to the culture medium, whereas E. coli growth was more robust in 3SL than in 6SL containing minimal medium. BecauseE. coli do not produce sialidases, the restoration of bacterial growth pointed to the presence of an a2,3-preferential sialidase activity in caecal uid. Such a sialidase activity was conrmed in caecal uid and shown to increase signicantly during DSS-induced colitis in both WT and ST mice (Fig. 4c). The substrate specicity of this caecal uid sialidase was demonstrated by high-performance liquid chromatography (HPLC) analysis after incubation of 3SL with caecal uid (Supplementary Fig. 1). The sialidase activity was also increased in the caecal uid of WT mice treated with streptomycin or vancomycin. By contrast, the sialidase activity was strongly decreased in WT mice treated with a broad-spectrum antibiotic cocktail consisting of ampicillin, vancomycin, metronidazole and neomycin, which supported the bacterial origin of the sialidase activity in caecal uid (Fig. 4d). To identify the source of this sialidase activity, we focused on bacterial groups, such as commensal Bacteroides and Bidobacteria species that are known to secrete sialidases25. Host-derived sialidases were unlikely candidates since vertebrate sialidases are unstable as soluble proteins in the extracellular space26.

Bacteroides vulgatus sialidase releases sialic acid. As Bacteroides species were strongly increased in the gut of mice challenged with DSS (Fig. 1b) as well as in mice treated with streptomycin or vancomycin (Fig. 2f), we searched for sialidase genes in the caecum of DSS-challenged mice using a series of PCR primers encompassing known Bacteroides sialidase sequences in the glycoside hydrolase family 33 of the CAZy database. This analysis revealed a 100-fold increase in copy number of the B. vulgatus BVU_4143 sialidase gene (gene ID: 5305102) in WT mice treated with DSS, whereas no change was detected in ST mice (Fig. 4e). The low levels of BVU_4143 sialidase gene in ST mice, however, shows that B. vulgatus is not the only source of sialidase activity in these mice. The abundance of B. vulgatus also increased accordingly in the colon of WT mice during DSS-induced colitis (Supplementary Fig. 2). Moreover, the abundance of the

a b

* * * * *

* *

RelativeE. coliDNA

to total bacteria (2Ct )

100

101 102 103 104 105 106 107 108

Relative E. coliDNA

to total bacteria (2Ct )

101 102 103 104 105 106 107 108

DSS

Antibiotic Stp Van Stp Van

+ + + + + +

WT WTXF ST STXF

WT ST

Figure 3 | Intestinal E. coli in mice. Relative abundance of E. coli was determined by real-time PCR using specic uidA primers and quantied using the 2 nCt method. (a) Relative abundance of E. coli in groups of 6-week-old WT and ST mice treated with the antibiotics vancomycin (Van) and streptomycin (Stp) and challenged with 3% DSS. Mice were treated for 8 days with or without antibiotics and DSS for the rst 5 days in drinking water.

Control mice received sterile drinking water. (b) Relative abundance of E. coli in 6-week-old WT, ST and respectively cross-fostered (XF) mice,which were fed by foster mothers of the other genotype during lactation. The data are represented as median values. Each point indicates a single mouse from two independent experiments. N 68, *Po0.05 (ANOVA, Bonferronis multiple comparison test).

4 NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141 ARTICLE

a

b

Cell density (OD 600)

1.5

1.0

0.5

0.0

Sia

Glc

Gal

GlcNAc

GalNAc

Fuc

Cell density (OD 600)

1.5

1.0

0.5

0.0

3SL+CF

6SL+CF

CF

3SL

6SL

0 5 10 15 20 25

Time (h)

0 5 10 15 20 25

Time (h)

c

d

Sialidase activity (4-MU-NeuNAc nmol ml1 )

* * *

*

*

1,000 800 600 400 200

Sialidase activity (4-MU-NeuNAc nmol ml1 )

1,000 800 600 400 200

*

0 WT WTDSS STDSS

ST

WT WTDSS STDSS

ST

0 C Stp Van AVMN

C Stp Van AVMN

e f

*

B. vulgatussialidase gene

(copy numbers per g fecal DNA)

108

107

106

105

104

B. vulgatussialidase gene

(copy numbers per g fecal DNA)

108

107

106

105

104

*

*

*

Figure 4 | Sialic acid processing and uptake by E. coli in vitro. (a) Growth of E. coli EHV2 in M9 minimal medium containing single monosaccharide at 10 mM. Sia, N-acetylneuraminic acid; Glc, glucose; GlcNAc, N-acetylglucosamine; GalNAc, N-acetylgalactosamine; Gal, galactose; Fuc, fucose. (b) Growth of E. coli in minimal medium containing 5 mM of 3SL and 6SL with and without supplementation of caecal uid (CF, 1.5%, v/v) derived from WT mice. (c) CF of conventional and DSS-challenged WT and ST mice were collected. Sialidase activity was determined by measuring uorescent 4-MU-NeuNAc at 440 nm. (d) Sialidase activity was also measured in the CF of WT and WT mice treated with either streptomycin (Stp 1 g l 1), vancomycin (Van0.5 g l 1) or antibiotic cocktail (AVMN: ampicillin, Van, metronidazole and neomycin). In (cd), the data are represented as means.e.m. from two independent experiments, N 58, *Po0.05 (two-tailed Students t-test). (e) The abundance of the B. vulgatus BVU_4143 sialidase gene was determined

by real-time PCR from caecum samples of WT and ST mice, and (f) caecum samples of antibiotic-treated WT mice. The data are represented as gene copy number per mg of faecal DNA. In (ef) each data point indicates a single mouse from two independent experiments, N 57, *Po0.05 (two-tailed

Students t-test).

BVU_4143 sialidase gene increased in WT mice treated with streptomycin and vancomycin, but decreased in mice treated with antibiotic cocktail (Fig. 4f).

The sialidase activity of BVU_4143 was conrmed after expression as a recombinant protein by demonstrating its ability to hydrolyse the aryl substrate 4-methylumbelliferyl N-acetylneuraminic acid (Supplementary Fig. 3a) and 2-O-(4-Nitrophenyl) N-acetylneuraminic acid (Supplementary Fig. 3b). The sialidase activity of recombinant BVU_4143 was inhibited by the sialidase inhibitor N-acetyl-2,3-didehydro-2-deoxyneuraminic acid (Neu5Ac2en) and lost by heat treatment. The addition of recombinant BVU_4143 to minimal media containing 3SL or 6SL enabled the growth of E. coli EHV2 on these oligosaccharides (Supplementary Fig. 3c), as observed for the restoration of E. coli EHV2 proliferation by addition of sterile-ltered caecal uid to culture medium containing 3SL and 6SL (Fig. 4b). Overall, these results indicate that the expansion of B. vulgatus and the concomitant increased sialidase activity during DSS-induced colitis enables the sialic acid-dependent outgrowth of E. coli during inammation.

Sialic acid is required for sustaining E. coli colonization. The dependence of E. coli on sialic acid in vivo was investigated by

deleting the sialic acid transporter nanT gene27. Disruption of nanT, the rst committed step in the sialic acid utilization pathway, abolished growth of the mutant E. coli strain in a sialic acid-containing minimal medium, but not growth in glucose-containing medium (Fig. 5a). By contrast, the disruption of mannose transporter ManX did not affect the growth of E. coli in both glucose and sialic acid-containing medium. To investigate the in vivo colonization efciency, nanT mutant and parentalE. coli were gavaged at equal amounts of each 106 cells to ampicillin-pretreated mice. Colonization efciency was determined over a period of 10 days after inoculation by counting E. coli isolated from freshly isolated faeces samples. Both strains were colonized at 108109 colony-forming units (c.f.u.) per gram faeces in WT mice by 2 days after inoculation. Parental E. coli remained stable over 10 days, but nanT mutants decreased markedly over the same period (Fig. 5b). The same experiment performed in ST mice showed that even parentalE. coli did not maintain their original levels in an environment decient of a2,3-linked sialic acid (Fig. 5c). The comparison of competitive index between parental and nanT E. coli in the intestines of WT and ST mice suggested that the growth advantage of parental E. coli was associated with the local availability of sialic acid (Fig. 5d). The levels of free Neu5Ac in the

NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 5

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141

a

b

Cell density (OD 600)

2.01.61.20.80.40.0

(log c.f.u. per gram faeces)

E. coli

E. coli nanT

E. colicounts

E. coli

E. coli nanT

Glucose Sialic acid

10

9

8

7

6 0 1 2 3 4 5 6 7 8 9 10

Days after colonization in WT mice

Days after colonization

c

d

(log c.f.u. per gram faeces)

10

9

8

7

WT

ST

*

*

E. colicounts

Competitive index

(parental/nanTmutant)

103

102

101

100

101

E. coli

E. coli nanT

E. coli ManX

1d 2d 5d 10d

6 0 1 2 3 4 5 6 7 8 9 10

Days after colonization in ST mice

*

e f

500

200

150

100

* * * *

*

400

Caecal Neu5Ac

concentration (M)

Caecal Neu5Ac

concentration (M)

300

200

50

100 0

0 WT WTDSS ST STDSS

C Stp Van

Figure 5 | Sialic acid catabolism is required for maintaining E. coli colonization. (a) Growth of parental E. coli, nanT mutant and ManX mutant in modied minimal media containing 10 mM of glucose or sialic acid. (b) Relative tness of parental E. coli and nanTmutant in colonization of WTmice and (c) STmice. Bacterial colonization tness was determined by counting c.f.u. of serially diluted faecal faeces collected at indicated time points. In (b,c), data are shown as medianinterquartile range. N 8, *Po0.05 (two-tailed Students t-test). (d) Competitive index was calculated as ratio of parental E. coli to nanT mutant

in WT and ST mice over 10 days after administration. The data are represented as median value, and each dot indicates an individual animal. N 8,

*Po0.05 (two-tailed Students t-test). (e) Levels of free Neu5Ac in the caecal uid of control and DSS-challenged WT mice were determined by HPLC analysis. (f) Levels of free Neu5Ac in the caecal uid of control (C), streptomycin- and vancomycin-treated mice measured 5 days post treatment. In (e,f), the data are represented as means.e.m. from two independent experiments. N 68, *Po0.05 (ANOVA, Bonferronis multiple comparison test).

caecal uid were indeed higher in WT mice than in ST mice (Fig. 5e), thereby correlating with the occurrence of intestinalE. coli in WT and ST mice (Fig. 3a). The concentrations of Neu5Ac measured in the caecum of WT mice treated with streptomycin and vancomycin (Fig. 5f) also matched the abundance of E. coli in these mice (Fig. 3a), but the low levels of Neu5Ac in streptomycin-treated mice also indicated that other bacteria consumed this carbohydrate when E. coli was suppressed. In fact streptomycin treatment increased the abundance of Bacteroidaceae and Porphyromonadaceae (Fig. 2f) that include several sialidase producers and sialic acid consumers, such as Bacteroides fragilis and Parabacteroides distasonis. In the late stage of DSS-induced colitis, the levels of free Neu5Ac decreased in WT mice and relatively increased in ST mice, which reected increased sialic acid usage by E. coli and increased sialidase activity (Fig. 4c) during intestinal inammation. Accordingly, these results were consistent with the hypothesis that E. coli outgrowth in the intestine depends on the release of sialic acid from host glycans.

Sialidase inhibition lowers E. coli expansion and colitis. On the basis of the requirement for sialidase activity to cleave a2,3-linked sialic acid and to promote E. coli proliferation during intestinal inammation, we hypothesized that sialidase inhibition would

decrease both the expansion of E. coli during DSS-induced colitis and the severity of the inammatory response. We rst conrmed the effectiveness of the sialidase inhibitor Neu5Ac2en at preventing E. coli growth in presence of 3SL and caecal uid in vitro (Fig. 6a). Next, we conrmed the effectiveness of Neu5Ac2en at reducing the release of sialic acid in vivo by showing decreased levels of free Neu5Ac in the caecum of WT mice treated with the sialidase inhibitor (Fig. 6b). Oral administration of Neu5Ac2en to WT mice during DSS challenge also prevented the outgrowth ofE. coli during inammation, as seen by a decrease of E. coli levels by 23 orders of magnitude (Fig. 6c), and decreased the severity of DSS-induced colitis as assessed by change in body weight (Fig. 6d) and colon length (Fig. 6e). Neu5Ac2en treatment also reduced the loss of colonic architecture and leukocyte inltration (Fig. 6f), although without reaching statistical signicance (Fig. 6g). Neu5Ac2en treatment was by contrast ineffective in ST mice challenged with DSS (Supplementary Fig. 4), which was expected considering the low levels of sialidase-producing Bacteroides spp. and low sialidase activity in the caecum of ST mice. Overall, these data demonstrated that inhibition of caecal sialidase activity signicantly reduced the outburst of E. coli and hence the severity of colitis in mice.

E. coli intensies dendritic cell activation. The question as to how E. coli proliferation affected intestinal inammation

6 NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141 ARTICLE

a

b c

* * *

* *

* *

Cell density

(% of untreated)

120 100

80 60 40 20

0 0 25 50 75 100

Caecal Neu5Ac

concentarion (M)

200

150

100

50

0

Relative E. coliDNA

to total bacteria (2 Ct )

101

102

103

104

105

106

107

Glc +CF

3SL +CF

Neu5Ac2en (M)

C

Neu5Ac2en

DSS

DSS/Neu5Ac2en

d e

g

*

P=0.092

Body weight (%)

110

105

100

95

90

85

80

75

Colon length (%)

Pathological scores

6

4

2

0

DSS/Neu5Ac2en

DSS

DSS

70 0 1 2 3 4 5 6 7 8

Time (days)

110 100

90 80 70 60 50

DSS/Neu5Ac2en DSS

DSS/Neu5Ac2en

f

Mock

DSS

DSS/Neu5Ac2en

Figure 6 | Reduced intestinal E. coli expansion by sialidase inhibition. (a) In vitro growth of E. coli EHV2 in minimal media containing 10 mM Glc or 3SL and the presence of caecal uid and inhibitor Neu5Ac2en (0, 20, 50 and 100 mM). (b) Levels of Neu5Ac in the caecal uid of control, and inhibitor-treated WTmice (2 h post 0.5 mg Neu5Ac2en administration) were measured by HPLC analysis. Control mice were administered with sterile PBS.

In (a,b), the data are represented as means.e.m. from two independent experiments, N 5, *Po0.05 (two-tailed Students t-test). (c) Relative

abundance of intestinal E. coli in control (PBS-treated) and Neu5Ac2en-treated WT mice (10 mg kg 1 per day) was determined at day 8 after initiation of DSS challenge. Data are represented as median values from two independent experiments, N 68, *Po0.05 (two-tailed Students t-test). (d) Relative

change in body weight of control and Neu5Ac2en-treated WT mice during DSS challenge for 5 days. Arrowheads indicate the time points of Neu5Ac2en administration. (e) Colon length was measured at day 8 after initiation of DSS challenge. In (d,e), the data are represented as means.e.m. from two independent experiments, N 68, *Po0.05 (two-tailed Students t-test). (f) Representative histological sections of colon tissues from untreated WTmice

(Mock), DSS-treated mice (DSS) and DSS-treated mice administered with sialidase inhibitor (DSS/Neu5Ac2en). Arrowheads indicate inltrating leukocytes. Scale bar, 100 mm. (g) Scoring of colitis severity by quantitative examination of tissue alteration and leukocytes inltration. Each dot indicates an individual animal from two independent experiments.

remained open. We therefore assessed the pro-inammatory potential of E. coli on intestinal CD11c dendritic cells (DCs).

Previous work has shown increased CD11c DC inltration to the colonic mucosa of WT mice compared with ST mice, suggesting a critical role of DCs during intestinal inammation19. We examined the stimulatory effect of E. coli EHV2 and of the Bacteroides thetaiotaomicron on mesenteric lymph node-derived CD11c DCs. B. thetaiotaomicron was chosen as reference because Bacteroides represents a major group of intestinal bacteria, which expanded in both WT and ST mice during DSS-mediated colitis (Fig. 1b). Stimulation of CD11c DCs with xed E. coli increased the expression of the activation markers major histocompatibility complex (MHC)-II, CD86 and CD40, whereas stimulation with xed B. thetaiotaomicron failed to

activate CD11c DCs (Fig. 7a). The pro-inammatory effect ofE. coli was not limited to mouse DCs as stimulation of the human monocytic cell line THP-1 also increased the expression of the activation marker CD54 (Fig. 7b). The effect of E. coli was even more pronounced when measuring the secretion of proinammatory cytokines from stimulated mouse CD11c DCs.

The levels of interleukin (IL)-6, tumour-necrosis factor (TNF)-a and IL-12p40 produced after E. coli stimulation exceeded those reached after stimulation with lipopolysaccharides (LPSs) at 500 ng ml 1. Under identical conditions, B. thetaiotaomicron did not increase cytokines production (Fig. 7c).

Overall, this study demonstrated that the expansion of commensal E. coli following the alteration of epithelial integrity caused by DSS uptake was mediated by increased exposure to

NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 7

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141

a b

8,000 6,000 4,000 2,000

* *

*

*

*

** *

Mean fluorescence

intensity

intensity (CD54-PE)

3,000

2,000

1,000

0 5:1 1:1

Bacteria : cell ratio

*

PBS

Bt Ec

Mean fluorescence

0 MHC II CD86 CD40

PBS

Bt

Ec

c

* * * * * * * *

*

IL-6 concentration

(pg ml1 medium)

150

100

50

0 PBS LPS Ec Bt PBS LPS Ec Bt PBS LPS Ec Bt

TNF-concentration

(pg ml1 medium)

5 4 3 2 1 0

IL-12p40 concentration

(pg ml1 medium)

500 400 300 200 100

0

Figure 7 | Stimulation of DCs and monocytes. (a) In stimulation assay, bacteria were xed with 0.5% paraformaldehyde in PBS, washed and co-cultured at a ratio of 100:1 with mouse mesenteric CD11c DCs. Ec, E. coli EHV2; Bt, B. thetaiotaomicron. Cell surface expression of CD86, CD40 and MHC-II was analysed by ow cytometry. (b) Human monocytic THP-1 cells were stimulated with xed bacteria at ratios of 5:1 and 1:1 to cells for 14 h at 37 C. Cell surface expression of CD54 was analysed by ow cytometry. In (a,b), data are shown as mean uorescence intensity, MFIs.e.m. from three independent experiments. N 6, *Po0.05 (ANOVA, Bonferronis multiple comparison test). (c) Cytokine expression in the culture supernatant of stimulated

mesenteric CD11c DCs stimulated for 14 h at 37 C with xed Ec, Bt and 500 ng ml 1 of LPS. PBS was used as negative stimulation control. Data are shown as means.e.m. from two independent experiments. N 45, *Po0.05 (ANOVA, Bonferronis multiple comparison test).

sialic acid, and that overgrowth of E. coli exacerbated intestinal inammation by stimulating the release of pro-inammatory cytokines from intestinal DCs.

DiscussionMultiple studies have documented that intestinal inammation is frequently accompanied by imbalanced microbiota. Such a dysbiosis is often characterized by a relative increase of facultative anaerobic Enterobacteriaceae28. Different factors such as nitrate29 and enterobactin30 promote Enterobacteriaceae expansion. The present study underlines the contribution of host glycosylation, specically of a2,3-linked sialic acids, in enabling the proliferation of Enterobacteriaceae during intestinal inammation in mice. Exposure to a2,3-linked sialic acids begins during lactation with the uptake of the milk oligosaccharide 3SL. After weaning, sialylated host glycans constitute the main source for the carbohydrate. Whereas, Enterobacteriaceae genomes encode various glycosidases, bacteria such as E. coli cannot digest sialylated oligosaccharides. Therefore, their growth relies on scavenging free monosaccharides released by glycosidases of other bacteria. Our comparative study of monosaccharides showed that the sialic acid yielded the fastest growth of E. coli among the main monosaccharides encountered in mammalian glycans. This nding is consistent with previous work showing that sialic acid catabolism conferred a growth advantage to intestinal E. coli31,32.

The growth advantage provided by sialic acid was dependent on a sialidase displaying a preference for a2,3-linked over a2,6-linked sialic acids, and which increased during intestinal inammation. Commensal E. coli do not express sialidases to liberate host sialylated glycans, therefore the access to bound sialic acids depends on secreted sialidases, such as the BVU_4143 sialidase identied in our study. We also detected other sialidase genes in colitogenic mice, such as sequences sharing similarity with sialidase genes encoded by B. fragilis and P. distasonis, although the abundance of these sialidase sequences did not vary between mouse genotypes and during DSS-mediated colitis.

The dependence of E. coli on sialidases secreted by Bacteroides spp. may explain the parallel increased abundance of Bacteroides spp. and E. coli observed in patients with colitis33. A recent study also demonstrated that the commensal sialidase-producingB. thetaiotaomicron was associated with proliferation of Salmonella enterica typhimurium and Clostridium difcile34. By contrast, colonization of mice with a sialidase-decient mutant reduced free sialic acid levels and thereby impaired the expansion of C. difcile.

The decreased severity of DSS-induced colitis in mice treated with sialidase inhibitor demonstrated the contribution of sialic acid in E. coli expansion and in the ensuing inammatory response. Administration of free sialic acid to mice before or during DSS challenge, however, failed to affect the levels of intestinal E. coli and the severity of colitis (Supplementary Fig. 5). Considering that monosaccharides are absorbed in the small intestine and only minute amounts reach the colon, oral supplementation with sialic acid is thus unlikely to inuence the outgrowth of E. coli in the colon. Therefore, the release of sialic acid from host glycans is critical in promoting the growth advantage of E. coli. The increase in sialylation of intestinal mucins during colitis35,36 likely facilitates the local release of free sialic acid during inammation.

Our ndings showed that commensal E. coli was a potent activator of a pro-inammatory response in intestinal DCs. The activation of DCs was likely induced by surface LPSs triggering Toll-like receptor-4 signalling. The strong pyrogenic effect of commensal E. coli over the one induced by B. thetaiotaomicron matches previous ndings showing that E. coli LPS was more active than Bacteroides spp.-derived LPS at inducing TNF-a production37. Similar differences in pyrogenicity were also noted in mouse models38,39. These results demonstrate that increasedE. coli levels likely exacerbate inammation through activation of mucosal immune cells such as DCs. Several intestinal bacteria regulate mucosal immunity, thereby affecting the occurrence of Th17 cells in the lamina propria in the case of segmented lamentous bacteria40,41, or the induction of Foxp3 Treg cells in the case of a group of Clostridium spp.42 and B. fragilis43.

8 NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141 ARTICLE

We cannot exclude that the disappearance of inammation-lessening bacteria also affects the severity of colitis alongside the expansion of pro-inammatory E. coli, but such contributions are unlikely since we failed to detect any differences in the distribution and amounts of mucosal immune cells between WT and ST mice before DSS-induced colitis18. Therefore, we conclude that the proliferation of E. coli supported by a2,3-linked sialic acids was the main factor regulating the magnitude of intestinal inammation triggered by DSS ingestion.

This study demonstrated the critical role of a2,3-linked sialic acids provided by milk-derived 3SL during lactation and by host mucosal glycans in establishing an intestinal niche for E. coli in mice. Expansion of E. coli during colitis directly depended on sialic acid release from host glycans after sialidase activity. This resulting overgrowth of E. coli leads to exacerbation of the pro-inammatory response by intestinal DCs. The benecial outcome of sialidase inhibition on the severity of DSS-induced colitis suggests that sialidase inhibitors should be investigated as agents able to reduce intestinal inammation by preventing dysbiosis manifested by Enterobacteriaceae expansion.

Methods

Bacterial DNA extraction and quantitative PCR. DNA was isolated from faecal samples using the QIAamp DNA stool mini kit (Qiagen) according to manufacturers instructions. Lysis temperature was increased to 95 C for 5 min to ensure complete cell lysis of Gram-positive cells. The proportion of bacterial family and genera in faecal samples were determined by real-time PCR using the EvaGreen qPCR Master Mix (Biotium). Cycling conditions were 40 cycles at 95 C for 10 s, 60 C for 10 s and 72 C for 25 s after an initial denaturation at 95 C for 3 min. Primer pairs specic for 16S rRNA of Bacteroidetes (Bac32F: 50-AACGCTAGCTA CAGGCTT-30, Bac303R: 50-CCAATGTGGGGGACCTTC-30); Enterobacteriaceae (Eco1457F: 50-CATTGACGTTACCCGCAGAAGAAGC-30, Eco1652R: 50-CTCTA CGAGACTCAAGCTTGC-30); and total bacteria (Eub338F: 50-ACTCCTACGGG AGGCAGCAG-30, Eub518R: 50-ATTACCGCGGCTGCTGG-30) were described previously44,45. Primers (UAL1939b: 50-ATGGAATTTCGCCGATTTTGC-30, UAL2105b: 50-ATTGTTTGCCTCCCTGCTGC-30) targeting b-glucuronidase uidA gene was used to evaluate the relative abundance of E. coli22. Quantication values were calculated by the 2 nCt method relative to total bacterial 16S rRNA amplicons46.

16S rRNA pyrosequencing. Faecal DNA was isolated from fresh stool samples of 7-week-old WT and ST male mice before and at day 8 after initiation of DSS treatment. The 16S rRNA V5V6 region was amplied from faecal DNA samples using primer 784F and 1061R47. Amplicons were sequenced using a Roche 454 GS-FLX system (DNAVision, Belgium). The QIIME software was used for taxonomic classication. Taxonomy was assigned using Ribosomal Database Project (RDP) classier and the Greengenes database48. Bacterial diversity was determined at the phylum, family and genus levels. The sequencing reads have been deposited in the NCBI sequence read archive SRA as Bioproject PRJNA289738.

Mouse models. WT and sialyltransferase St3gal4 / mice49 were of C57BL/6 background and derived from the same breed and maintained in light-cycled and climate-controlled facility. Animals were received regular laboratory chow diet (KLIBA extrudat no. 3436, Provimi Kliba, SA, Switzerland) and sterile water ad libitum. Synchronized matings were set up for WT and ST mice to allow the exchange of newborn mice for cross-fostering experiments. All experiments were performed in compliance with the Swiss Animal Protection Ordinance and approved by the Veterinary Ofce of the Canton of Zurich, Switzerland.

Antibiotic treatment and DSS-induced colitis. Six- to 7-week-old male mice were treated with 33.5% (w/v) DSS (molecular weight 3650 kDa; MP

Biomedicals) in drinking water for 5 days, followed by a supply of normal water until sacrice of the animals. Body weight and physical activity were monitored daily. For long-term antibiotic treatment, 3-week-old mice were provided with sterile drinking water supplemented with vancomycin (0.5 g l 1), streptomycin (1 g l 1), neomycin (1 g l 1), chloramphenicol (0.5 g l 1) or metronidazole(1 g l 1) plus aspartame (0.25%, w/v) for 3 weeks before the beginning of DSS treatment. For short-term antibiotic treatment, mice were administered sterile drinking water supplemented with vancomycin (0.5 g l 1) or streptomycin(1 g l 1) for 8 days, during which they also received 3% of DSS on the rst 5 days.

Transepithelial permeability assay. Mice were gavaged with 600 mg kg 1 body weight of FITCdextran (MW 30005000, Sigma-Aldrich) and whole blood was collected by cardiac puncture 4 h after gavage. Blood serum was collected after centrifugation at 1500g for 10 min. Serum uorescence intensity was measured using a multi-detection microplate reader (Tecan Innite M200 Pro, Switzerland) with an excitation wavelength of 485 nm and an emission wavelength of 535 nm. FITC concentration (mg ml 1) was calculated from a standard curve using serial dilutions of FITCdextran50.

Bacterial strains and mutant construct. E. coli strain EHV2 was isolated from inamed colon surface of DSS-treated C57/BL6 mice. The isolate strain was conrmed by 16S rRNA sequencing, universal stress protein uspA23, gyrase gyrB24 sequencing and phenotypic analysis (API-20E Enterobacteriaceae identication kit; bioMerieux). E. coli strain HS996 was obtained from Genebridge and transformed with pET16b vector containing an ampicillin resistance gene. HS996-nanT knockout mutant was constructed by introducing a kanamycin-resistance cassette by homologous recombination into the nanT locus with followed by manufactures instruction (Genebridge, Germany). The genomic primers used for the targeting construct were designed according to the anking regions of the E. coli nanT gene: forward primer f182 was 50-ATACCAAAGCGTGTGGGCATCGCCCACCGCG

GGAGACTCACAATGAGTACAATTAACCCTCACTAAAGGGCG-30, and

reverse primer r1723 was 50-GCAACAGGATTAACTTTTGGTTTTGACTAAAT CGTTTTTGGCGCTGCCAATAATACGACTCACTATAGGGCTC-30. Null mutation at the nanT locus was conrmed by genome sequencing and growth phenotype. B. thetaiotaomicron (DSM 2079T) was obtained from the German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany).

Cells were grown in YCFA (yeast extract-casein hydrolysate-fatty acids) medium containing volatile fatty acids51 at 37 C anaerobically in rubber-sealed Hungate tubes. Cell density was measured by a spectrophotometer at 600 nm (S2100 Diode array, Biochrom WPA).

Carbohydrate metabolism assay. E. coli EHV2 (107 cells) were cultured in 3 ml of M9 minimal medium52 containing 10 mM of either glucose (Glc), galactose (Gal), N-acetylglucosamine (GlcNAc), N-acetylgalactosamine (GalNAc), fucose (Fuc) and N-acetylneuraminic acid (Sia) as single carbohydrate source at 37 C for 24 h. The milk oligosaccharides 3SL and 6SL were tested as 5 mM in 3 ml ofM9 minimal medium supplemented with PBS or sterile-ltered mouse caecal uid(1.5%, v/v). All neutral monosaccharides were purchased from Sigma-Aldrich except N-acetylneuraminic acid from Carbosynth (Berkshire, UK). The oligosaccharides 3SL and 6SL were obtained from Glycom A/S (Lyngby, Denmark). Cell density was determined by OD 600 values at 3, 6, 12 and 24 h. For determination of the growth phenotype in nanT mutant and ManX mutant, parental and mutant strains were cultured in modied M9 minimal medium (additional 0.003% L-histidine, 0.004% leucine and 0.01% yeast extract) containing 10 mM of glucose or N-acetylneuraminic acid.

Sialidase activity assay. Mouse caecal content was collected and centrifuged at 15,000g for 10 min at 4 C. The supernatant was ltered through a 0.45-mm membrane to yield caecal uid. The uorogenic substrate 20-(4-methylumbelliferyl)-a-D-N-acetylneuraminic acid sodium salt (4-MU-NeuNAc; Carbosynth) was used to determine sialidase activity. In brief, caecal uid (10%, v/v) was incubated with0.1 mM 4-MU-NeuNAc in 0.2 ml of 100 mM sodium acetate buffer (pH 7.4) at 37 C for 15 min. Assays were stopped by adding 0.8 ml of 0.5 M sodium carbonate buffer (pH 10.5) and further diluted 20-fold before uorescence measurement. Cleaved 4-methylumbelliferone (4-MU) was measured by uorescence detection in a multi-detection microplate reader at an excitation wavelength of 360 nm and an emission wavelength of 440 nm (ref. 53).

Quantitative PCR of bacterial sialidase genes. Sequences of sialidase (EC3.2.1.18) genes from Bacteroidaceae were retrieved from the GH33 sialidase family of the CAZy database. Primers encompassing conserved DNA stretches of sialidase genes from P. distasonis, B. vulgatus, B. thetaiotaomicron and B. fragilis were designed based on multiple alignment analysis. The lack of signicant sequence similarity of the selected primers with unrelated bacterial sequences was conrmed by BLAST analysis. The B. vulgatus sialidase primers used for quantitative PCR analysis were Bv-f266: 50-GGAGGGGAAAGACTTATTTTGC-30, Bv-r501: 50-TTC

CACCACTTCTGCCGAC-30; cycling conditions were 40 cycles at 95 C for 10 s, 60 C for 10 s and 72 C for 25 s after an initial denaturation at 95 C for 3 min. Quantication values are represented as gene copy numbers per mg of total faecal DNA.

Molecular cloning and purication of sialidase. The gene encodingB. vulgatus_4143 sialidase (Gene ID: 5305102) was amplied by PCR using the genomic DNA from caecum sample in WTDSS mouse as template. NdeI and BamHI sites were introduced in the forward Bvu_4143F 50-GGCCATATGAGA

AACCCTAGCTTATTA-30 and reverse primer Bvu_4143R: 50- GCGGGATCC TTATTTGGTCTTAATAAT-30, respectively. PCR conditions were thirty cycles of 30 s at 95 C, 30 s at 53 C, 3.5 min at 72 C. The PCR product was digested with

NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 9

& 2015 Macmillan Publishers Limited. All rights reserved.

ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141

NdeI and BamHI and subcloned into the pET16b expression vector (Novagen). The pET16b- BVU_4143 vector was transformed into E. coli BL21-star (DE3) cells (Invitrogen) cultured in LB broth and supplemented with ampicillin (100 mg ml 1)

at 37 C. On reaching an OD 600 value of 0.5, BVU_4143 expression was induced by adding 0.5 mM Isopropyl b-D-1-thiogalactopyranoside and incubating bacteria at 30 C for 5 h. Bacteria were then resuspended in 100 mM Tris-HCl plus 20 mM imidazole (pH 7.4) and disrupted by sonication. The resulting cell extract was incubated with Ni-sepharose (GE Healthcare Life Sciences) at 4 C overnight, washed with 100 mM Tris-HCl containing 40 mM imidazole and the His6-tagged

BVU_4143 sialidase was eluted with 500 mM imidazole.

Bacterial colonization tness assay. The relative tness of E. coli strainsfor colonization of the mouse intestine was monitored as described34. Briey, 6-week-old WT mice were given drinking water containing ampicillin (2 mg ml 1)

for 2 days before gavage with 106 c.f.u. of parental HS996-ampr and HS996-nanT mutant::Kanar E. coli. Fresh faecal samples were collected, serially diluted and plated on LB agar containing ampicillin (100 mg ml 1) or ampicillin plus kanamycin (30 mg ml 1) by days 1, 2, 5 and 10 after gavage. Competitive index was calculated as the ratio of parental to nanT mutant E. coli.

Quantication of caecal sialic acids. Mouse caecal content (B500 mg) was weighed out and centrifuged for 10 min at 16,000g at 4 C. The supernant was collected and ltered through 0.45 mm cronus HPLC membrane to get caecal uid and stored at 20 C before use. Caecal uid was derivatized with 1,2-diamino-

4,5-methylene-dioxybezene (DMB; Sigma-Aldrich) as described previously54.

In brief, 10 ml of caecal uid was incubated with 200 ml of the DMB solution at 50 C for 2.5 h in the dark. DMB solution was prepared by dissolving DMB dihydrochloride (7 mM) in 1.4 M acetic acid containing 0.75 M b-mercaptoethanol and 18 mM sodium hydrosulte. The reaction was stopped by adding 800 ml of ice-cold distilled water. The derivatized product was analysed by reverse-phase HPLC using a ODS Hypersil 150 3 mm column (Thermo scientic). The mobile

phase was acetonitrile/methanol/water (9:7:84, v/v) at a ow rate of 0.3 ml min 1. Florescence of the derivatized product was monitored at 373 nm (excitation) and 448 nm (emission). DMB-derivatized Neu5Ac was identied by comparison with authentic sialic acid standards.

Sialidase inhibition. The sialidase inhibitor N-acetyl-2,3-didehydro-2-deoxyneuraminic acid (Neu5Ac2en) was prepared in house based on published procedures55. For in vitro inhibition, E. coli EHV2 was cultured for 24 h at 37 C in M9 minimal media containing 5 mM 3SL, caecal uid (1.5%, v/v) and varying Neu5Ac2en concentrations. For in vivo inhibition, mice were anesthetized by isourane inhalation, gavaged with 300 ml of Neu5Ac2en (10 mg kg 1 per day) in sterile saline at days 0, 1, 2 and 5 of DSS challenge. Control groups received sterile saline.

Histological staining of colonic tissue. Distal colons were removed, cut longitudinally, and xed in 10% neutral buffered formalin then embedded in parafn. Tissue samples were cut in serial 5-mm sections, which were stained with hematoxylin-eosin (Sigma-Aldrich). Histological sections were examined by using microscope Zeiss Axio Imager.Z2, objective Zeiss EC Plan Neouar 10 /0.3.

Image was acquired by Zeiss AxioCam HrC camera and analysed with Zeiss AxioVision software (AxioVs40V4.8.2.0). Sections were scored individually by an independent investigator blinded to the type of treatment. Morphological changes and leukocyte inltration in the colon were scored as previously described56. Histology was scored as follows: epithelium 0: normal morphology; 1: loss of goblet cells; 2: loss of goblet cells in large areas; 3: loss of crypts; and 4: loss of crypts in large areas. Inltration 0: no inltrate; 1: inltrate around crypt basis; 2: inltrate reaching to lamina muscularis mucosae; 3: extensive inltration reaching the lamina muscularis mucosae and thickening of the mucosa with abundant oedema; and 4: inltration of the lamina submucosa. The total score represents the sum of the epithelium and inltration score.

DC Isolation and stimulation. Mesenteric lymph nodes were isolated and incubated in 2.5 mg ml 1of collagenase type D (Roche) in RPMI 1640 containing 10% FCS for 10 min at 37 C. Tissues were gently homogenized by passing through an 18-gauge needle, then incubated for 30 min at 37 C. The resulting cell suspensions were ltered through 40-mm cell strainers and incubated with

Fc-blocker (anti-CD16/32; eBioscience) for 10 min. CD11c cells were isolated with anti-CD11c MicroBeads (Miltenyi Biotec) according to the manufacturers instructions. CD11c DCs (2 105 cells per ml) were culture in RPMI 1640

containing 10% FCS, and stimulated with xed bacteria or PBS for 14 h at 37 C.E. coli (EHV2) and B. thetaiotaomicron (DSMZ 2079T) were xed in 0.5% paraformaldehyde for 15 min at room temperature and washed with PBS before stimulation. Fixed bacteria were co-cultured with DCs in a ratio 100:1. After stimulation, DCs were stained with uorochrome-labeled anti-mouse antibodies: MHC-IIFITC (BD Biosciences), CD86-PE (eBioscience), CD40-PE-Cy7 (BioLegend) and CD11C-APC (BD Biosciences) for 30 min at 4 C. Cells were analysed by ow cytometry (FACSCanto II). For monocytes stimulation, human

THP-1 cells (1 106 cells per ml) were stimulated with bacteria or PBS for 14 h as

described above, but in bacteria to cells ratios of 5:1 and 1:1. After stimulation, THP-1 cells were stained with CD54-PE (BD Biosciences) for 30 min at 4 C and analysed by ow cytometry. The supernatants of stimulated DCs were collected and analysed for IL-6, IL-10, IL-12p40 and TNF-a cytokine production by multiplex bead array (Cytolab AG, Switzerland).

Statistical analysis. Results are presented as means.e.m. unless specied. Difference between groups was analysed by unpaired Students t-test (two-tailed) and one-way analysis of variance (ANOVA) with Bonferronis multiple comparison post-test using GraphPad Prism 5. P values below 0.05 were considered signicant.

References

1. Wang, Y. et al. 16S rRNA gene-based analysis of fecal microbiota from preterm infants with and without necrotizing enterocolitis. ISME J. 3, 944954 (2009).

2. Kamada, N., Seo, S. U., Chen, G. Y. & Nunez, G. Role of the gut microbiota in immunity and inammatory disease. Nat. Rev. Immunol. 13, 321335 (2013).

3. Zhao, L. The gut microbiota and obesity: from correlation to causality. Nat. Rev. Microbiol. 11, 639647 (2013).

4. Oertli, M. et al. Helicobacter pylori gamma-glutamyl transpeptidase and vacuolating cytotoxin promote gastric persistence and immune tolerance. Proc. Natl Acad. Sci. USA 110, 30473052 (2013).

5. Scher, J. U. & Abramson, S. B. The microbiome and rheumatoid arthritis. Nat. Rev. Rheumatol. 7, 569578 (2011).

6. Louis, P., Hold, G. L. & Flint, H. J. The gut microbiota, bacterial metabolites and colorectal cancer. Nat. Rev. Microbiol. 12, 661672 (2014).

7. El Kaoutari, A., Armougom, F., Gordon, J. I., Raoult, D. & Henrissat, B. The abundance and variety of carbohydrate-active enzymes in the human gut microbiota. Nat. Rev. Microbiol. 11, 497504 (2013).

8. Chassard, C. & Lacroix, C. Carbohydrates and the human gut microbiota. Curr. Opin. Clin. Nutr. Metab. Care. 16, 453460 (2013).

9. Zivkovic, A. M., German, J. B., Lebrilla, C. B. & Mills, D. A. Human milk glycobiome and its impact on the infant gastrointestinal microbiota. Proc. Natl Acad. Sci. USA 108, 46534658 (2011).

10. Marcobal, A. et al. Consumption of human milk oligosaccharides by gut-related microbes. J. Agric. Food. Chem. 58, 53345340 (2010).

11. Pacheco, A. R. et al. Fucose sensing regulates bacterial intestinal colonization. Nature 492, 113117 (2012).

12. Bouchet, V. et al. Host-derived sialic acid is incorporated into Haemophilus inuenzae lipopolysaccharide and is a major virulence factor in experimental otitis media. Proc. Natl Acad. Sci. USA 100, 88988903 (2003).

13. Harvey, H. A., Swords, W. E. & Apicella, M. A. The mimicry of human glycolipids and glycosphingolipids by the lipooligosaccharides of pathogenic neisseria and haemophilus. J. Autoimmun. 16, 257262 (2001).

14. Carlin, A. F., Lewis, A. L., Varki, A. & Nizet, V. Group B streptococcal capsular sialic acids interact with siglecs (immunoglobulin-like lectins) on human leukocytes. J. Bacteriol. 189, 12311237 (2007).

15. Aspholm, M. et al. SabA is the H. pylori hemagglutinin and is polymorphic in binding to sialylated glycans. PLoS Pathog. 2, e110, doi:10.1371/ journal.ppat.0020110 (2006).

16. Byres, E. et al. Incorporation of a non-human glycan mediates human susceptibility to a bacterial toxin. Nature 456, 648652 (2008).

17. Kashyap, P. C. et al. Genetically dictated change in host mucus carbohydrate landscape exerts a diet-dependent effect on the gut microbiota. Proc. Natl Acad. Sci. USA 110, 1705917064 (2013).

18. Fuhrer, A. et al. Milk sialyllactose inuences colitis in mice through selective intestinal bacterial colonization. J. Exp. Med. 207, 28432854 (2010).19. Kurakevich, E., Hennet, T., Hausmann, M., Rogler, G. & Borsig, L. Milk oligosaccharide sialyl(alpha2,3)lactose activates intestinal CD11c cells

through TLR4. Proc. Natl Acad Sci USA 110, 1744417449 (2013).20. Lupp, C. et al. Host-mediated inammation disrupts the intestinal microbiota and promotes the overgrowth of Enterobacteriaceae. Cell Host Microbe 2, 204 (2007).

21. Schuppler, M., Lotzsch, K., Waidmann, M. & Autenrieth, I. B. An abundance of Escherichia coli is harbored by the mucosa-associated bacterial ora of interleukin-2-decient mice. Infect. Immun. 72, 19831990 (2004).

22. Maheux, A. F. et al. Analytical comparison of nine PCR primer sets designed to detect the presence of Escherichia coli/Shigella in water samples. Water Res. 43, 30193028 (2009).

23. Chen, J. & Grifths, M. W. PCR differentiation of Escherichia coli from other gram-negative bacteria using primers derived from the nucleotide sequences anking the gene encoding the universal stress protein. Lett. Appl. Microbiol. 27, 369371 (1998).

24. Fukushima, M., Kakinuma, K. & Kawaguchi, R. Phylogenetic analysis of Salmonella, Shigella, and Escherichia coli strains on the basis of the gyrB gene sequence. J. Clin. Microbiol. 40, 27792785 (2002).

10 NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications

& 2015 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9141 ARTICLE

25. Kim, S., Oh, D. B., Kang, H. A. & Kwon, O. Features and applications of bacterial sialidases. Appl. Microbiol. Biotechnol. 91, 115 (2011).

26. Varki, A. & Gagneux, P. Multifarious roles of sialic acids in immunity. Ann. NY Acad. Sci. 1253, 1636 (2012).

27. Roy, S., Douglas, C. W. & Stafford, G. P. A novel sialic acid utilization and uptake system in the periodontal pathogen Tannerella forsythia. J. Bacteriol. 192, 22852293 (2010).

28. Nagalingam, N. A. & Lynch, S. V. Role of the microbiota in inammatory bowel diseases. Inamm. Bowel. Dis. 18, 968984 (2012).

29. Winter, S. E. et al. Host-derived nitrate boosts growth of E. coli in the inamed gut. Science 339, 708711 (2013).

30. Singh, V. et al. Interplay between enterobactin, myeloperoxidase and lipocalin 2 regulates E. coli survival in the inamed gut. Nat. Commun. 6, 7113 (2015).

31. Chang, D. E. et al. Carbon nutrition of Escherichia coli in the mouse intestine. Proc. Natl Acad. Sci. USA 101, 74277432 (2004).

32. Fabich, A. J. et al. Comparison of carbon nutrition for pathogenic and commensal Escherichia coli strains in the mouse intestine. Infect. Immun. 76, 11431152 (2008).

33. Gophna, U., Sommerfeld, K., Gophna, S., Doolittle, W. F. &Veldhuyzen van Zanten, S. J. Differences between tissue-associated intestinal microoras of patients with Crohns disease and ulcerative colitis. J. Clin. Microbiol. 44, 41364141 (2006).

34. Ng, K. M. et al. Microbiota-liberated host sugars facilitate post-antibiotic expansion of enteric pathogens. Nature 502, 9699 (2013).

35. Parker, N., Tsai, H. H., Ryder, S. D., Raouf, A. H. & Rhodes, J. M. Increased rate of sialylation of colonic mucin by cultured ulcerative colitis mucosal explants. Digestion 56, 5256 (1995).

36. Campbell, B. J., Yu, L. G. & Rhodes, J. M. Altered glycosylation in inammatory bowel disease: a possible role in cancer development. Glycoconj. J. 18, 851858 (2001).

37. Delahooke, D. M., Barclay, G. R. & Poxton, I. R. A re-appraisal of the biological activity of bacteroides LPS. J. Med. Microbiol. 42, 102112 (1995).

38. Takada, H. et al. Bacteroides lipopolysaccharides (LPS) induce anaphylactoid and lethal reactions in LPS-responsive and -nonresponsive mice primed with muramyl dipeptide. J. Infect. Dis. 162, 428434 (1990).

39. Poxton, I. R. & Edmond, D. M. Biological activity of Bacteroides lipopolysaccharidereappraisal. Clin. Infect. Dis. 20, S149S153 (1995).

40. Ivanov, I. I. et al. Induction of intestinal Th17 cells by segmented lamentous bacteria. Cell 139, 485498 (2009).

41. Gaboriau-Routhiau, V. et al. The key role of segmented lamentous bacteria in the coordinated maturation of gut helper T cell responses. Immunity 31, 677689 (2009).

42. Atarashi, K. et al. Induction of colonic regulatory T cells by indigenous Clostridium species. Science 331, 337341 (2011).

43. Mazmanian, S. K., Round, J. L. & Kasper, D. L. A microbial symbiosis factor prevents intestinal inammatory disease. Nature 453, 620625 (2008).44. Bernhard, A. E. & Field, K. G. Identication of nonpoint sources of fecal pollution in coastal waters by using host-specic 16S ribosomal DNA genetic markers from fecal anaerobes. Appl. Environ. Microbiol. 66, 15871594 (2000).

45. Fierer, N., Jackson, J. A., Vilgalys, R. & Jackson, R. B. Assessment of soil microbial community structure by use of taxon-specic quantitative PCR assays. Appl. Environ. Microbiol. 71, 41174120 (2005).

46. Schmittgen, T. D. & Livak, K. J. Analyzing real-time PCR data by the comparative C(T) method. Nat. Protoc. 3, 11011108 (2008).

47. Andersson, A. F. et al. Comparative analysis of human gut microbiota by barcoded pyrosequencing. PLoS ONE 3, e2836 (2008).

48. De Filippo, C. et al. Impact of diet in shaping gut microbiota revealed by a comparative study in children from Europe and rural Africa. Proc. Natl Acad. Sci. USA 107, 1469114696 (2010).

49. Ellies, L. G. et al. Sialyltransferase ST3Gal-IV operates as a dominant modier of hemostasis by concealing asialoglycoprotein receptor ligands. Proc. Natl Acad. Sci. USA 99, 1004210047 (2002).

50. Napolitano, L. M., Koruda, M. J., Meyer, A. A. & Baker, C. C. The impact of femur fracture with associated soft tissue injury on immune function and intestinal permeability. Shock 5, 202207 (1996).

51. Duncan, S. H., Hold, G. L., Harmsen, H. J., Stewart, C. S. & Flint, H. J. Growth requirements and fermentation products of Fusobacterium prausnitzii, and a proposal to reclassify it as Faecalibacterium prausnitzii gen. nov., comb. nov. Int. J. Syst. Evol. Microbiol. 52, 21412146 (2002).

52. Sambrook, J. & Russell, D. Molecular Cloning: A Laboratory Manual (Cold Spring Harbor Laboratory Press, 2001).

53. Thompson, H., Homer, K. A., Rao, S., Booth, V. & Hosie, A. H. F. An orthologue of Bacteroides fragilis NanH is the principal sialidase in Tannerella forsythia. J. Bacteriol. 191, 36233628 (2009).

54. Hara, S. et al. Determination of mono-O-acetylated N-acetylneuraminic acids in human and rat sera by uorometric high-performance liquid chromatography. Anal. Biochem. 179, 162166 (1989).

55. von Itzstein, M., Wu, W. Y. & Jin, B. The synthesis of 2,3-didehydro-2,4-dideoxy-4-guanidinyl-N-acetylneuraminic acid: a potent inuenza virus sialidase inhibitor. Carbohydr. Res. 259, 301305 (1994).

56. Hausmann, M. et al. In vivo treatment with the herbal phenylethanoid acteoside ameliorates intestinal inammation in dextran sulphate sodium-induced colitis. Clin. Exp. Immunol. 148, 373381 (2007).

Acknowledgements

We thank Glycom (Lyngby, Denmark) for providing 3SL and 6SL, Jesus Glaus Garzn

and Lubor Borsig for technical assistance with histology, and Marek Whitehead for

technical assistance with FACS analysis and Andreas Hlsmeier for proteomic analysis.

This work was supported by the Zurich Center for Integrative Human Physiology,

by Swiss National Foundation grant CRSII3_154488/1 and the National Health and

Medical Research Council, Australia grant 1047824.

Author contributions

T.H. and Y.-L.H. designed the study; Y.-L.H. performed the experiments; M.H.

performed histological evaluation; M.v.I. provided sialidase inhibitors; T.H., Y.-L.H.

and C.C. analysed data; all authors discussed results and wrote the manuscript.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/naturecommunications

Web End =http://www.nature.com/

http://www.nature.com/naturecommunications

Web End =naturecommunications Competing nancial interests: The authors declare no competing nancial interests. Reprints and permission information is available online at http://npg.nature.com/reprintsandpermissions/

Web End =http://npg.nature.com/

http://npg.nature.com/reprintsandpermissions/

Web End =reprintsandpermissions/

How to cite this article: Huang, Y. L. et al. Sialic acid catabolism drives

intestinal inammation and microbial dysbiosis in mice. Nat. Commun. 6:8141

doi: 10.1038/ncomms9141 (2015).

This work is licensed under a Creative Commons Attribution 4.0

International License. The images or other third party material in this

article are included in the articles Creative Commons license, unless indicated otherwise

in the credit line; if the material is not included under the Creative Commons license,

users will need to obtain permission from the license holder to reproduce the material.

To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/

Web End =http://creativecommons.org/licenses/by/4.0/

NATURE COMMUNICATIONS | 6:8141 | DOI: 10.1038/ncomms9141 | http://www.nature.com/naturecommunications

Web End =www.nature.com/naturecommunications 11

& 2015 Macmillan Publishers Limited. All rights reserved.