Infant Diet Promotes Bifidobacterium Community Cooperation Within A

bioRxiv preprint doi: https://doi.org/10.1101/711234; this version posted July 23, 2019. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1 Infant diet promotes Bifidobacterium community cooperation within a

2 single ecosystem.

3 Melissa AE Lawson1*, Ian J O’Neill1*, Magdalena Kujawska1, Anisha Wijeyesekera2, Zak

† 4 Flegg1, Lisa Chalklen1, Lindsay J Hall1 .

6 Affiliations

7 1Gut Microbes & Health, Quadram Institute Bioscience, Norwich Research Park, UK

8 2Department of Food & Nutritional Sciences, University of Reading, UK

9 *These authors contributed equally

† 10 Correspondence: [email protected]

12 Abstract

13 Diet-microbe interactions play an important role in modulating the early life microbiota, with

14 Bifidobacterium strains and species dominating the gut microbiota of breast-fed infants. Here,

15 we sought to explore how infant diet drives distinct bifidobacterial community composition

16 and dynamics within individual infant ecosystems. Genomic characterisation of 19 strains

17 isolated from breast-fed infants revealed a diverse genomic architecture enriched in

18 carbohydrate metabolism genes, which was distinct to each strain, but collectively formed a

19 pangenome across infants. Presence of gene clusters implicated in digestion of human milk

20 oligosaccharides (HMOs) varied between species, with growth studies indicating within

21 infant differences in the ability to utilise 2’FL and LNnT HMOs between strains. We also

22 performed cross-feeding experiments using metabolic products from growth on 2’FL or

23 LNnT for non-HMO degrading isolates, these compounds were identified to include fucose,

24 galactose, acetate and N-acetylglucosamine. These data highlight the cooperative nature of

25 individual bifidobacterial ‘founder’ strains within an infant ecosystem, and how sharing

26 resources maximises nutrient consumption from the diet. We propose that this social

27 behaviour contributes to the diversity and dominance of Bifidobacterium in early life and

28 suggests avenues for development of new diet and microbiota based therapies to promote

29 infant health.

31 Introduction

32 The early life developmental window represents a critical time for microbe-host interactions

33 as this is when foundations for future health and wellbeing are established. Colonisation of

34 pioneer microbes shortly after birth represents a key first step in this mutualistic relationship;

35 shaping the developing microbial community, and in turn impacting numerous host

36 physiological processes (1,2). Although the microbiota of adults is complex in nature, the

37 gastrointestinal (GI) tract of full term healthy infants is relatively simplistic, with upwards of

38 80% of the total microbiota being comprised of the genus Bifidobacterium (1). Loss of

39 Bifidobacterium, or indeed gain of other species during this critical window of opportunity,

40 may significantly alter the entire community with negative consequences for host health (3).

41 In infants, Bifidobacterium can be considered a foundation microbiota member which,

42 due to their abundance, strongly influence the environment, the structure of burgeoning

43 microbial communities and host development (4,5). Infant diet is suggested to be one of the

44 key factors that shapes the early life microbiota, and recently the WHO (6) and the Scientific

45 Advisory Committee on Nutrition (UK) (7) released new guidelines regarding the optimal

46 time to start breast feeding, and highlighted the health benefits associated with solely breast-

47 feeding infants. In breast-fed infants, the early life microbial community is dominated by a

48 single bacterial genus Bifidobacterium; that is, until the diet changes to include either infant

49 formula and/or simple solid foods. It is known that breast-fed and formula-fed infants differ

50 in microbial composition (8), including significant differences in bifidobacterial populations,

51 which has also been linked to differential health outcomes, both in the short- and longer-term

52 i.e. induction of allergies, asthma and obesity in formula fed infants (8,9).

53 Establishment of this bifidobacterial dominant community in response to diet

54 (specifically breast milk) suggests a cooperative relationship between different strains and

55 species, rather than competition, which is often described for other bacterial species (10).

56 Indeed, it appears that closely related Bifidobacterium strains coexist in a single infant GI

57 tract, rather than one strain dominating and competitively excluding other strains. However, a

58 cooperative balance between bifidobacterial strains in the early life microbiota may further

59 enhance their dominance in breast fed infants by enabling a genus-specific exploitative

60 competition i.e. depleting the GI tract of breast milk-derived nutrients, thereby preventing

61 colonisation of other microbes, including pathobionts. To investigate these key community

62 dynamic questions, we have probed the genomic and phenotypic similarities between

63 bifidobacteria strains that coexist in the same individual, including their responses to specific

64 early life diet components, namely human milk oligosaccharides (HMOs). By examining

65 microbial interactions on a strain-level we provide important insights into how multiple

66 species of Bifidobacterium co-exist within a single infant in early life, which may have

67 implications for design of diet- and microbial-based early life therapies.

69 Results

70 Bifidobacterium dominates the infant microbiota. To investigate bifidobacterial

71 community interactions in early life, the faecal microbial community profiles from three full

72 term, healthy infants (referred to as infant V1, V2 and V3) was subjected to metataxonomic

73 profiling using 16S rRNA gene sequencing (Fig 1A and S1A). At the time of sample

74 collection, all infants were similar in age (mean = 145 ± 38 d), born vaginally and exclusively

75 breast-fed (with the exception of two isolations coming from infant V3 at an earlier time

76 point, Fig S1B and Table S1). In agreement with previous studies, we observed a high

77 prevalence of Bifidobacterium in each infant faecal sample (mean = 82.53±12.36%). Further

78 analysis indicated a dynamic bifidobacterial community comprised of different strains and

79 species present in each infant, thus we isolated multiple Bifidobacterium strains to explore

80 bifidobacterial community dynamics. A total of 19 strains were isolated and their genetic

81 relatedness was inferred based on core genome phylogeny (Fig 1B), and Average Nucleotide

82 Identity (ANI) values, and compared to 83 publicly available bifidobacterial genomes (Fig

83 S2). The generated phylogenetic tree indicates the clustering of all analysed strains into three

84 main phylogenetic groups denoted as the Bifidobacterium longum (encompassing the

85 members of the longum and the infantis subspecies), Bifidobacterium breve and

86 Bifidobacterium pseudocatenulatum groups. Strains isolated from infant V1 were classified

87 as either B. longum subspecies longum (hereafter referred to as B. longum) or B.

88 pseudocatenulatum; V2 strains classified as B. breve or B. longum subspecies infantis

89 (hereafter referred to as B. infantis), and V3 strains were classified as either B.

90 pseudocatenulatum, B. infantis, or B. longum (Fig 1B). Genome sizes ranged from 2.25Mb

91 (B. pseudocatenulatum LH13) to 2.75Mb (B. longum subsp. infantis LH23), with an average

92 of 2.38Mb, consistent with previously published data (11). The G+C content ranged from

93 56.50% for B. pseudocatenulatum LH9 to 60.04% for B. longum subsp. longum LH277,

94 while the number of predicted ORFs was lowest in B. pseudocatenulatum LH11 (1,888), and

95 highest in B. longum subsp. infantis LH23 (2,521) (Table S1).

97 Analysis of bifidobacterial genomes reveals distinct and diverse functional profiles

98 within and between infant isolates. Previous studies have highlighted that diet may act as

99 an evolutionary driver in Bifidobacterium, as functionally, 13.7% of genes in the

100 bifidobacterial pan genome are involved in carbohydrate metabolism, thereby reflecting their

101 ability to grow within the glycan rich environment of the colon (12,13). We functionally

102 characterised genomes and determined that, apart from genes of unknown function, genes

103 classified as carbohydrate transport and metabolism were the most abundant in all genomes,

104 indicating the saccharolytic lifestyle of Bifidobacterium spp. (Fig 1C and Table S3). Strains

105 from infant V1 had the highest percentage of carbohydrate metabolism and transport genes of

106 all infants (10.32%), followed by strains from infant V2 (10.06%), and then infant V3

107 (9.53%). Based on functional genome analysis there appears to be interspecies differences in

108 genes associated with carbohydrate transport and metabolism, for example, 10.67± 0.11% of

109 the total genome of B. pseudocatenulatum strains from infant V1 were dedicated to this

110 functional group, whereas V3 B. pseudocatenulatum strains had 1% less (9.60±0.18%) of

111 their genome annotated for carbohydrate use. These data suggest the same species in different

112 ecological niches (i.e. infant) may have altered metabolic function.

113 Due to the high proportion of B. pseudocatenulatum strains isolated from infant

114 V1 and V3, and apparent differences in functional annotation, we further investigated

115 differences in the core and accessory genes (Fig. S3A, Tables S6A & S6B). B.

116 pseudocatenulatum strains from the same infant shared a large core genome, but there were

117 minor differences in accessory genes. Strains from infant V1 had more unique genes

118 compared to strains from infant V3 (Fig S3). Further characterisation of these unique genes

119 indicated that infant V1 strains LH13 and LH14 possess unique putative carbohydrate

120 utilisation gene clusters (LH_13_00067-LH_13_00071 and LH_14_01835-LH_14_01839)

121 that are not present in other strains and may contribute to altered carbohydrate digestion. The

122 LH14 genome also contains genes encoding fimbriae and sortase genes (LH_14_01074-

123 LH_14_01081), suggesting presence of sortase dependent fimbriae. We have also identified a

124 partial prophage gene cluster in LH13 (LH_13_01645-LH_13_01657) that encodes phage tail

125 and head proteins and a tyrosine recombinase, however, due to the size of the cluster it is

126 likely to be an unsuccessful phage insertion event. Interestingly, two large partial prophage

127 clusters were also identified in genomes of LH656 and LH658 from infant V3 (Table S7A &

128 S7B).

129 Bifidobacterium possess a large repertoire of glycoside hydrolases (GH) that facilitate

130 digestion in the glycan rich gut environment, therefore, we determined the number of GH

131 present in each infant faecal ecosystem (Fig 2 and S3B). Genes from a total of 39 different

132 GH families were represented in at least one genome with B. pseudocatenulatum strains have

133 the highest number of GH genes (mean = 62 per genome), followed by B. longum (mean = 48

134 per genome), B. breve (mean = 46 per genome), and B. infantis (mean = 42 per genome). In

135 all infants, the most represented family were members of GH13 representing enzymes

136 involved in the hydrolysis of alpha-glucosidic linkages found in plant di-, oligo- and

137 polysaccharides. The second most abundant GH family in infants V1 and V3 (but not V2)

138 was GH43 which contains beta-xylosidases involved in the digestion of the plant

139 polysaccharide xylan. Other highly represented families were GH3, containing beta-

140 glucosidases that hydrolyse a wide range of glycans present in plant cell walls, GH2 and

141 GH42, containing beta-galactosidases that are active on galactooligosaccharides and

142 galactans found in plant cell walls, and also active on lactose that is present in human breast

143 milk.

144 Mucin present in the infant gut contains sialic acid residues that can be utilised by

145 many members of the microbiota including Bifidobacterium (14). B. infantis and B. breve

146 strains from infants V2 and V3 contains exo-sialidases from GH33 potentially allowing these

147 strains to utilise host sialic acids (Fig. 2). Interestingly B. pseudocatenulatum and B. longum

148 from infant V3 do not contain these genes, however, crossfeeding of sialic acids between

149 Bifidobacterium species has previously been reported (15).

150 Ultimately, the presence of multiple strains within the Bifidobacterium community

151 results in a larger repertoire of GHs, potentially indicating an ability to digest a wider range

152 of oligosaccharides within the infant gut.

153

154 Infant bifidobacterial isolates have enhanced survival against exposure to intestinal

155 environmental stressors. To colonise and maintain a successful bifidobacterial ecosystem,

156 strains must first be able to traverse the length of the GI tract, to their preferred niche i.e. the

157 colon. When exposed to acid shock at pH2 for four hours (representing stomach pH and

158 transit time), all strains (except B. breve LH24) were able to re-establish growth in MRS

159 media (Fig 3A). As expected all strains exhibited poor growth in aerobic, compared to

160 anaerobic conditions, although there was modest growth in some strains, which might

161 suggest an ability for vertical and horizontal transmission (Fig 3B). Growth in 0.3% bovine

162 bile salt resulted in statistically significantly less growth of all strains, except for B. infantis

163 LH23 from infant V2 (Fig 3C). All strains contained homologs to known bile salt hydrolase

164 genes (Table S3), and therefore we tested the ability of each strain to hydrolyse specific

165 individual bile salts including taurocholic acid, taurodeoxycholic acid and sodium

166 glycodeoxycholate bile acids (Table S4). We found differences between strains from the

167 same individual, such as B. pseudocatenulatum LH11 was capable of hydrolysing taurocholic

168 acid, taurodeoxycholic acid and sodium glycodeoxycholate bile acids, while B. longum

169 isolates (LH206 and LH277) were unable to hydrolyse any of the bile salt tested, and only the

170 B. longum strain from infant V1 (LH12) could precipitate taurodeoxycholic bile acid. Within

171 a single infant bifidobacterial community, it appears there are those strains that have

172 enhanced intestinal stressor survival, compared to highly related intra-infant strains,

173 indicating adaptation that may enhance colonisation potential of these strains, thus impacting

174 community succession.

175

176 The presence of HMO gene clusters is species and strain specific With respect to an

177 infant-specific diet, Bifidobacterium genomes encode a large repertoire of glycosyl

178 hydrolases (GH) that allow for digestion of human milk oligosaccharides (HMO) (Katayama,

179 2016), which are a key component of maternal breast milk. To date, there have been over 200

180 different HMOs identified in maternal milk, which are unconjugated glycans comprised of a

181 lactose core with varying chain length of 3 to 15 carbohydrates containing either glucose,

182 galactose, fucose, N-acetylglucosamine (GlcNAc), and N-acetylneuraminic acid (NeuAc) or

183 sialic acid (16,17). In addition, the lactose backbone of HMOs can be either fucosylated or

184 sialylated to form trisaccharide HMO structures, namely 2’ or 3’-fucosyllactose (FL) and 3’

185 or 6’-sialyllactose (SL), respectively. Gene clusters involved in the digestion of HMOs have

186 also been described for B. infantis, B. breve, B. longum and B. pseduocatenulatum, however,

187 the genetic organisation and size of these HMO clusters varies from species to species (18–

188 22). Thus, we analysed the genomes in this study for the presence of the clusters previously

189 described for B. infantis, B. breve, B. longum and B. pseduocatenulatum.

190 Comparison with the large, 45kb, B. infantis ATCC 15697 HMO cluster

191 (BLON_RS12070-BLON_RS12215) identified similar clusters in B. infantis LH23, LH664,

192 LH665 and LH666 (Fig 4A), however, the genetic organisation of this cluster varied in our

193 infant isolates. Importantly, homologues to all three GH contained within the cluster are

194 found in similar clusters in our genomes, however, instead of a large continuous cluster,

195 homologues of genes from ATCC 15697 are found in two separate clusters within LH23,

196 which may reflect the presence of transposon sequences flanking each cluster. B. infantis

197 LH665 lacks four transport-reported permease genes present in the ATCC 15697 cluster,

198 however, other transport binding proteins and permeases encoded within the LH665 HMO

199 cluster may perform a similar function.

200 Digestion of 2’FL in B. longum SC569 is linked to the presence of genes encoding an

201 alpha1-3/4-fucosidase (GH29) and/or alpha1-2 fucosides (GH95), which are present in a gene

202 cluster with carbohydrate utilisation genes (BLNG_01254-BLNG_01264). Interestingly,

203 neither this cluster nor any alpha-fucosidades were identified in any of the B. longum strains

204 from any of the infants. Homologues to these GHs were identified in all B. infantis (LH23,

205 LH664, LH665 and LH666) strains from infant V2 and V3 (data not shown). A highly similar

206 gene cluster is found in B. pseudocatenulatum DSM 20438, but only contains one alpha-

207 fucosidase (GH95) and all B. pseudocatenulatum strains in infant V1, LH9, LH11, LH13 and

208 LH14, encode homologs of these clusters, including the key GH95 gene (Fig 4A).

209 B. breve UCC2003 encodes four separate gene clusters; lnt cluster (BBR_RS13080-

210 BBR_RS13100); lac cluster (BBR_RS18470-BBR_RS18480), the nah cluster

211 (BBR_RS18490-BBR_RS18520) and lnp/glt cluster (BBR_RS18650-BBR_RS18675) that

212 are important for the digestion of LNT and LNnT (22,23). Homologs to the lac, nah and

213 lnp/lgt clusters were identified in B. breve LH21 and LH24. The lnt cluster is upregulated

214 during growth on LNT and LNnT, but this cluster was not present in either LH21 or LH24,

215 however, it was identified in B. infantis strains LH206 and LH277, and in B. longum strain

216 LH12. Unsurprisingly, the lnp/glt cluster was identified in B. longum strains LH12, LH206,

217 LH277 as this cluster has also been described for B. longum JCM1217 . B. infantis strains

218 LH664, LH665 and LH666 contained a partial lnp/glt cluster with highly similar enzymatic

219 genes, but no transporter genes in close genomic proximity, suggesting that these strains may

220 degrade LNB, but may transport this HMO using other genes in the genome.

221 Sialidases aid breakdown of sialidated HMOs and B. bifidum ATCC 15696 has an

222 extracellular sialidase (SiaBb1) that can digest these HMOs without importing them into the

223 cytosol (24,25). Our analysis identified sialidases in B. infantis strains LH23, LH664, LH665

224 and LH666, however, the required transmembrane domains were absent suggesting that these

225 strains are intracellular sialidase utilisers (Table S8). Thus, it appears that infant-specific

226 strains may maximise utilisation of a wide assortment of HMOs within a community via

227 differential genetic adaptation to a breast-milk diet and early life niche.

228

229 Phenotypic characterisation of HMO usage indicates known and unknown enzymatic

230 gene clusters. Next to lactose, HMOs are the second most abundant carbohydrate in breast

231 milk (5-15 g/L in breast milk) (26), and are a key metabolic source for Bifidobacterium

232 (27,28). Although we identified a wide range of putative HMO genomic clusters in our

233 strains, we next sought to correlate these with an ability to metabolise HMOs, in particular

234 2’FL and LNnT (Fig. 4B-D). We were able to identify strains capable of using either 2’FL or

235 LNnT in each microbial ecosystem. All B. pseudocatenulatum strains, but not the B. longum

236 LH12 strain, from infant V1 could degrade the HMO 2’FL, most likely due to the presence of

237 a known fucosylated HMO utilisation gene cluster (Fig 4A). Only B. pseudocatenulatum

238 LH11 was capable of using the more complex HMO, LNnT, despite lacking known

239 enzymatic clusters, suggesting an alternative or novel mechanism of degradation.

240 Interestingly, only strain B. infantis LH21 isolated from infant V2 could degrade 2’FL for

241 growth (neither B. breve strains LH23 or LH24 could), and yet all three strains grew well

242 with LNnT as a sole carbon source. None of the B. pseudocatenulatum strains from infant V3

243 could use either of the HMOs tested, possibly due to the absence of appropriate HMO

244 clusters in their genomes, while all the B. longum and B. infantis strains appeared to

245 metabolise both HMOs for growth; apart from B. infantis LH666 which only utilised LNnT

246 for growth. This data is supported by bioinformatic analysis identifying a large HMO

247 degradation gene cluster and the presence of key alpha-fucosidases in all B. infantis strains.

248 Collectively these data demonstrate that HMO utilisation is dependent on the type of HMO

249 and the strain tested; supported by the fact that type strains for each species also had

250 differential HMO degradation (Fig. S4).

251

252 A syntrophic network for HMO utilisation exists within Bifidobacterium species. It is

253 clear that multiple stains and species exist as a community within a single ecosystem (i.e.

254 infant), and that these bifidobacterial communities encode a diverse array of metabolic genes,

255 which correlates with strain-specific breakdown of breast milk-associated HMOs. Thus, to

256 assess how these bifidobacterial communities interact with each other, we determined if

257 strains unable to degrade HMOs, could utilise the metabolic by-products generated by HMO

258 degraders. To address this experimentally, all strains that had previously been found to use

259 either 2’FL (Fig 4C) or LNnT (Fig 4D) were grown in their respective HMOs, and

260 conditioned media was filtered and used as a carbon source for all other non-HMO users

261 within the same infant community (Fig 5A). We found that 2’FL derived-substrates from all

262 B. pseudocatenulatum strains in infant V1 supported growth of B. longum LH12 (Fig 5B).

263 This demonstrates a syntrophic/cross-feeding network, and cooperation between different

264 species within the same ecological niche. Conversely, V2-associated B. breve strain

265 breakdown products did not support B. infantis LH23 growth, indicating that these strains

266 were not able to cross-fed when HMO 2’FL is the initial carbon source. Both B. longum and

267 two B. infantis isolates from infant V3 (LH206, LH277, LH664 and LH665 respectively)

268 grew on 2’FL, however only the conditioned media from isolate B. longum LH206 could

269 support growth of other isolates within the same infant. Interestingly, bioinformatic analysis

270 did not identify any alpha-fucosylase genes in LH206. Moreover, LH206 conditioned media

271 enhanced growth of all tested isolates (both B. infantis and B. pseudocatenulatum species),

272 which suggests the metabolism of 2’FL by LH206 may generate a wide variety of

273 components for neighbouring isolates to use for growth within this ecosystem (i.e. infant V3).

274 Overall fewer isolates could grow using the more complex HMO LNnT, and as a

275 result no strains from infant V1 were tested. In contrast to the 2’FL growth profile, B. infantis

276 LH23 strain utilised LNnT, but associated by-products did not support growth of either B.

277 breve isolates from infant V2 (Fig 5C). B. longum LH206 and LH277, and B. infantis LH664

278 and LH665 grew in the presence of LNnT, however only conditioned media from B. longum

279 LH206 generated LNnT-metabolic compounds to support growth of other non-HMO strains

280 in infant V3. We observed moderate growth for B. pseudocatenulatum LH656, LH657 and

281 LH663, indicating a cooperative LNnT cross-feeding network within the same microbial

282 ecosystem.

283

284 Metabolite analysis indicates specific metabolic components involved in syntrophic

285 network. We next sought to identify the metabolic compounds involved in these cross-

286 feeding networks for both 2’FL and LnNT HMOs. Using 1H-NMR we analysed the

287 metabolic compounds generated by B. longum LH206 after growth, using 2’FL as a sole

288 carbon source, and compared that to B. pseudocatenulatum LH659 after growth on cell-free

289 supernatant from LH206 (Fig 6A). B. longum LH206 produced formate, lactose, galactose,

290 glucose, fucose and acetate in response to growth on 2’FL, and these compounds were

291 reduced after B. pseudocatenulatum LH659 growth, suggesting active metabolism (Fig 6A,

292 Fig S5 and Table S9). In addition, LH659 when grown on LH206 conditioned media resulted

293 in production of the metabolite intermediate pyruvate. Cross-feeding experiments using

294 strains from infant V1, indicated that HMO degrader B. pseudocatenulatum LH13 and HMO

295 non-degrader B. longum LH12 produced similar results (Fig S6).

296 B. longum LH206 can also use LNnT for growth, therefore we also profiled cross-

297 feeding metabolites using the non-LNnT degrader, B. pseudocatenulatum LH663 from infant

298 V3 (Fig 6B, Fig S7 and Table S9). Similar to the 2’FL cross-feeding experiments, we

299 observed production and consumption of many metabolites; in particular LH206 generated

300 free galactose from LNnT digestion that was no longer detectable in the conditioned media

301 from LH663, indicating galactose had been consumed for growth. There was also an increase

302 in the energy-related compound formate and the end product of fermentation ethanol,

303 suggesting that LH663 was also producing these metabolites in response to growth. In

304 addition, we also examined the metabolic profile of strain B. longum LH206 cross-feeding to

305 B. pseudocatenulatum LH657 and found additional ethanol production from LH657; but also

306 noted that it consumed acetate, galactose, and N-acetyl glucosamine, by products of LnNT

307 degradation by B. longum LH206 (Fig S8). Notably, strain B. longum LH206, and not isolate

308 LH277, could digest all HMOs tested, and generated a variety of metabolic by-products that

309 could be consumed by other bifidobacterial strains. Although strain specific, metabolic by-

310 products produced by these HMO degraders, can be used as ‘public goods’ to be shared

311 amongst the Bifidobacterium community, thus demonstrating cooperative traits that likely

312 contributes to bifidobacterial survival in the early life infant gut (Fig 7).

313

314 Discussion

315 Bifidobacterium spp. are central players in the early life microbiota and healthy infant

316 development. We show that this genus is present at very high levels in breast fed infants, and

317 that distinct bifidobacterial communities exist within an individual infant ecosystem. Our

318 data indicates a diverse genomic landscape for these individual strains, which links to their

319 ability to thrive on breast milk-associated dietary components i.e. HMOs within a

320 ‘community’ setting. These data highlight the important role that communities of

321 bifidobacteria play in an early life environment, and suggests avenues for development of

322 dietary and microbial-based supplements and therapies to promote infant health.

323 From three healthy breast-fed infants we isolated a significant number of diverse

324 bifidobacterial strains and species including members commonly associated with the infant

325 microbiota i.e. B. infantis, B. longum, B. breve and B. pseudocatenulatum (29,30). One

326 proposed method that Bifidobacterium are acquired by infants is through vertical

327 transmission from their mother’s vaginal canal during birth as vaginally born infants tend to

328 have a higher proportion of Bifidobacterium than those delivered by C-section (31,32).

329 However, horizontal transmission may also occur as the microbiota profile of C-section and

330 vaginally born infants converge after as little as six weeks (33). Both vertical and horizontal

331 transmission of Bifidobacterium to the infant gut require the ability to withstand oxygen

332 exposure and passage through the environmental stressors of GI tract i.e. acid and bile

333 exposure (34). In this study, 18 of 19 strains were showed resistance to all the above

334 environmental stressors and this indicates these strains have adapted to allow for effective

335 transmission to the host.

336 There are distinct genotypic and phenotypic differences between the strains within a

337 single infant bifidobacterial community; however, these differences enable a flexible and

338 cooperative relationship in a breast milk diet (i.e. HMOs) environment that may support

339 dominance of this genus in the wider early life microbiota. HMOs represent a key nutritional

340 component of breast milk, but these complex carbohydrates cannot be directly metabolised

341 by the infant (35). In infants specialised members of the resident gut microbiota allow the

342 breakdown of these compounds, in particular Bifidobacterium utilises HMOs and likely

343 contributes to its function as a foundation genus in early life (17,27). Multiple studies have

344 identified genomic clusters for the degradation of these oligosaccharides, including specific

345 clusters for utilisation of specific HMOs (18,21,22). We determined that the genomic

346 arrangement of these clusters exhibits interspecies variability and, consistent with other

347 studies, the presence of these clusters does not always result in a growth phenotype on

348 specified HMOs. For instance, both B. breve strains in this study possessed a key GH for

349 fucosylated HMO degradation, but did not grow on 2’FL which could potentially be due to

350 the lack of a second fucosidase (GH29) or appropriate transport genes (20). Furthermore, we

351 identified growth on HMOs in strains lacking known clusters, suggesting a wider variety of

352 novel gene clusters devoted to HMO degradation that could be explored further to provide

353 more mechanistic rationale for development of early life microbiota therapies

354 Cross-feeding of HMO metabolites between Bifidobacterium is a proposed

355 mechanism to allow for non-HMO degraders to survive in the HMO rich environment of the

356 breast-fed infant gut (12,25,36). Recently it has been shown that extracellular sialidases on B.

357 longum produce sialylated carbohydrates and free sialic acid to promote B. breve growth

358 (24,25); however, there is little evidence suggesting co-operation between non-extracellular

359 HMO degraders. We have identified that by-products of HMO degradation can be utilised by

360 non-degraders from the same infant suggesting sharing of resources within a single host. For

361 example, by products of 2’FL degradation from B. pseudocatenulatum strains can support the

362 growth of B. longum (Fig 7). Cross-feeding has not been described for B. pseudocatenulatum

363 to B. longum previously, however, 2’FL has been shown to allow cross-feeding between the

364 intracellular 2’FL degrader B. infantis and another microbiota member, Eubacterium halli,

365 driven by the 2’FL metabolic end product of 1,2, propanediol (37). Whilst we did not identify

366 this specific product in our system, we did see the consumption of other end products of

367 carbohydrate metabolism including fucose, acetate and pyruvate, which may be sustaining

368 growth in these cross-feeding events. This suggests a more generic sharing of resources

369 between Bifidobacterium strains that may also encourage growth of the wider infant gut

370 microbiota including minor, but important members such as Escherichia coli (38) and

371 Staphylococcus (39), with a mean proportion of 2.5±0.01% and 0.08±0.05% respectively,

372 within our infant microbiotas.

373 The impact of community sharing of metabolites suggests cooperation between

374 Bifidobacterium strains in the infant gut and eludes to the lack of resource competition that

375 has previously been described for identical species within a single ecosystem. It is plausible

376 that due to the abundance of Bifidobacterium in early life these cross-feeding events occur

377 specifically within the Bifidobacterium community. This advantage is likely key for

378 bifidobacteria survival and ultimate exclusion of potential invading bacteria into the early life

379 microbiota, and suggests Bifidobacterium are foundation species that may act as ecosystem

380 engineers, constructing and shaping the gut microbiota in early life. Our results demonstrate

381 that HMO utilisation is strain-specific, and Bifidobacterium ultimately have different and

382 complementary carbohydrate degradation mechanisms. Since we examined strains from

383 within a single ecological niche, it is likely that there is shared used of available resources

384 from the infant diet through metabolic cooperation and enables engraftment of

385 Bifidobacterium in the infant microbiota. Similar to our findings, recent work has also

386 proposed cooperation amongst multiple Bifidobacterium isolates from an SHIME model

387 colon system based on their ability to utilise and share the metabolic components of inulin-

388 type fructans and arabinoxylan oligosaccharides present in the adult diet (40). Despite the

389 concept that similar strains of a specific species tend to compete against each other for shared

390 resources and stable engraftment of new bacteria is enhanced in the absence of other

391 members of the same species (41), our data provides support that positive interactions, like

392 cooperation phenotypes, predominate in single microbial communities and environments

393 (42). In this work, for the first time we have identified altruistic behaviour in B. longum strain

394 LH206 to produce a variety of metabolites from HMO degradation that directly promotes the

395 growth of other Bifidobacterium strains within the same environment, clearly demonstrating

396 a form of social cooperation between microbes at an individual level. However, what remains

397 to be determined is whether or not there is also mutualistic cooperation occurring, i.e. that

398 LH206 receives a benefit from other strains within the community. Collectively, this research

399 provides new insight into the social behaviour of these proposed foundation species and how

400 they function collectively as a group to maximise nutrient utilisation from breast milk to

401 dominate the infant gut. Determining these interactions with respect to infant diet, will be

402 critical for development of optimal multiple strain/species microbiota therapies to promote

403 early life health.

404

405 Methods

406 For detailed information see supplementary Materials and Methods.

407

408 Bacterial isolation and strains Sample collection was in accordance with protocols laid out

409 by the National Research Ethics Service (NRES) approved UEA/QIB Biorepository (Licence

410 no: 11208) and Quadram Institute Bioscience Ethics Committee. Infant faces were isolated

411 on RCM (Oxoid, Hampshire, UK) supplemented with mupirocin and L-cysteine (0.05

412 mg/mL each, Sigma-Aldrich, Dorset, UK). All Bifidobacterium and Lactobacillus strains

413 were grown at 37°C in either RCM, de Man Rogosa and Sharpe (MRS) media, or modified

414 MRS (mMRS) with specified carbohydrates in an anaerobic chamber (Don Whitley

415 Scientific, Bingley, UK), unless otherwise specified.

416

417 16S rRNA gene library preparation and bioinformatics analysis. DNA extraction was

418 performed using the FastDNA Spin Kit for Soil (MPBIO, California, USA) and the V1-V2

419 region of the 16S rRNA gene was amplified by PCR and run on the Illumina MiSeq platform

420 as previously described (43). Raw reads were initially processed through quality control

421 using FASTX-Toolkit53 maintaining a minimum quality threshold of 33 for at least 50% of

422 the bases. Passed read were then aligned against the SILVA database (44) using BLASTN55

423 (45) separately for both pairs. All output files were annotated using the paired-end protocol in

424 MEGAN (46).

425

426 Genomic DNA extraction. Bacterial pellets were lysed with lysozyme, Proteinase K, RNase

427 A (all from Roche Molecular Systems, West Sussex, UK), EDTA and Sarkosyl NL30

428 (Sigma-Aldrich). Samples were purified with three rounds of Phenol:Chloroform:Isoamyl

429 Alcohol (25:24:1) (Sigma-Aldrich) extraction followed by two rounds of extractions with

430 Chloroform:Isoamyl Alcohol (24:1) (Sigma-Aldrich). Genomic DNA pellets were

431 resuspended in 10mM Tris (pH 8.0) and quantified using Qubit dsDNA BR assay kit

432 (Invitrogen).

433

434 Whole genome sequencing. Isolated DNA was subject to multiplex Sanger Illumina library

435 preparation protocol followed by sequencing using Illumina HiSeq 2500 platform with read

436 length 2 × 125bp (paired-end reads) and an average sequencing coverage of 60×. Draft

437 genome assemblies were generated using previously described assembly and annotation

438 pipeline (47). Additionally, previously assembled publicly available sequences (n=64) were

439 retrieved online from NCBI Genomes database (48). All genomes were annotated using

440 Prokka v1.10 (49).

441

442 Phylogenetic analysis of whole genomes. General feature format files (GFF) of 83

443 Bifidobacterium strains were used as input for Roary pangenome pipeline v. 3.8.0 to obtain

444 core-genome data (50). The phylogeny was reconstructed from the core-genome alignment

445 generated using MAFFT v 7.305b (51) and subject to cleaning from poorly aligned positions

446 using Gblocks (52,53) and manual curation. Maximum likelihood analysis was performed in

447 Seaview v. 4.0 (54) using PhyML v. 3.1 with 100 bootstrap iterations (55). Python 3 module

448 pyANI with default BLASTN+ settings was employed to calculate the average nucleotide

449 identity (ANI) between the 83 Bifidobacterium genomes (56). A cut-off of 95% identity was

450 used for species delineation.

451

452 Functional annotation of genomes. For each genome, all ORFs were submitted to eggNOG-

453 mapper for annotation and classification (57,58). Prediction of HMO clusters was performed

454 by comparing known protein sequences to the draft genomes in this study using local BLAST

455 (45) (e-value<1e-50, percentage identify > 70%). HMO clusters were annotated ‘present’ if

456 over 90% of genes were homologous in the cluster. For prediction of GH, ORFs were

457 submitted to the dbCAN web server (59) and the number of GH were calculated. Prophage

458 presence was predicted using PHASTER (60,61).

459

460 Bile salt survival and hydrolysis. To determine Bifidobacterium survival in bile, isolates

461 were grown in MRS ± 0.3% unfractionated bovine bile salt (Sigma-Aldrich) as described by

462 (62). Bile salt hydrolyase activity was assessed on MRS agar supplemented with L-cysteine

463 and 0.5% w/v of either taurocholic acid, taurodeoxycholic acid, and sodium

464 glycodeoxycholate bile salt (Sigma-Aldrich). Bile salt precipitation was assessed after a

465 maximum of 96 hours of anaerobic incubation at 37°C.

466

467 Aerotolerance. Each strain was grown aerobically or anaerobically at 37°C and absorbance

468 at 600nm (OD600nm) was measured after 48 hours

469

470 HMO utilisation and cross-feeding. Bifidobacterium growth in mMRS + 2% (w/v) LNnT

471 or 2’FL (Glycom, Hørsholm, Denmark) was determined using a microplate

472 spectrophotometer. Cell free supernatants (CFS) were prepared by sterile filtration of cultures

473 grown in mMRS + 5% (w/v) LNnT or 2’FL. For cross-feeding, growth in CFS:mMRS (1:1)

474 was monitored every 15 min for 48 in a microplate spectrophotometer (Tecan Infinite F50).

475

476 1H-Nuclear Magnetic Resonance (NMR) Spectroscopy analysis. For functional

477 assessment of Bifidobacterium strains, media in which the bacterial cells had been grown,

478 were analysed using 1H-NMR Spectroscopy. Media samples were mixed (2:1) with 0.2M

479 sodium phosphate buffer solution (pH 7.4) made in 100% deuterium oxide and 0.01% of

480 sodium 3-(trimethylsilyl) [2,2,3,3,-2H4] propionate 3mM NaN3. The mixture was vortexed

481 and centrifuged and transferred to a 5mm outer diameter NMR tube (Wilmad). One-

482 dimensional spectroscopic data were acquired using a 500 MHz NMR spectrometer (Bruker

483 Biospin, Germany) operating at 300 K. A standard one-dimensional NMR pulse sequence

484 with water pre-saturation was applied to acquire spectroscopic data, using 4 dummy scans

485 followed by 64 scans and collected into 24 K data points. 1H NMR spectra were manually

486 corrected for phase and baseline distortions and referenced to the TSP signal at δ 0.0, using

487 the TopSpin 3.5 software package (Bruker Biospin, Germany). Spectra from the different

488 bacterial strains grown under different conditions were overlaid in TopSpin and compared for

489 differences. The integrate function was utilised to integrate peaks of interest. Spectral

490 compound libraries (e.g. Human Metabolome DataBase (HMDB), Biological Magnetic

491 Resonance Data Bank) published literature and in-house spectral reference libraries were

492 used to confirm metabolite assignments.

493

494 Data availability

495 All 16S and whole genome data is available at the European Nucleotide Archive, study

496 accession ID PRJEB28188.

497

498 Acknowledgements

499 We would like to thank the mothers and infants that donated their infant stool samples for

500 this project, and in particular Glycom A/S for the kind donation of purified HMOs: 2’FL and

501 LNnT. The authors would also like to thank Shabhonam Caim and Cristina Alcon-Giner for

502 their technical support with the 16S rRNA pipeline, and Jennifer Ketskemety for help with

503 strain isolations and sequencing prep. This work was funded by a Wellcome Trust

504 Investigator Award (100/974/C/13/Z), and the Biotechnology and Biological Sciences 21

505 Research Council (BBSRC); Institute Strategic Programme Gut Microbes and Health

506 BB/R012490/1, Institute Strategic Programme Gut Health and Food Safety BB/J004529/1 to

507 LJH, and support of the BBSRC Norwich Research Park Bioscience Doctoral Training Grant

508 (BB/M011216/1, supervisor LJH, student MK). MAEL was funded by a Marie Sklodowska-

509 Curie Individual Fellowship (Project 661594).

510

511 Conflict of interest

512 All authors declare no conflicts of interest

513

514 References

515 1. Bezirtzoglou E, Tsiotsias A, Welling GW, Fan W, Huo G, Li X, et al. Development of

516 the preterm infant gut microbiome: A research priority. World J Microbiol Biotechnol

517 [Internet]. 2014;2(6):2889–900. Available from:

518 https://digital.nhs.uk/catalogue/PUB30183

519 2. O’Neill I, Schofield Z, Hall LJ. Exploring the role of the microbiota member

520 Bifidobacterium in modulating immune-linked diseases. Emerg Top Life Sci

521 [Internet]. 2017;1(4):333–49. Available from:

522 http://www.emergtoplifesci.org/lookup/doi/10.1042/ETLS20170058

523 3. O’Neill I, Schofield Z, Hall LJ. Exploring the role of the microbiota member

524 Bifidobacterium in modulating immune-linked diseases. Emerg Top Life Sci.

525 2017;1(4):333–49.

526 4. Stachowicz J. Mutualism, Facilitation, and the Structure of Ecological Communities:

527 Positive interactions play a critical, but underappreciated, role in ecological

528 communities by reducing physical or biotic stresses in existing habitats and by creating

529 new habitats on . Bioscience [Internet]. 2001;51(3):235–46. Available from:

530 https://academic.oup.com/bioscience/article/51/3/235/256224

531 5. Trosvik P, de Muinck EJ. Ecology of bacteria in the human gastrointestinal tract--

532 identification of keystone and foundation taxa. Microbiome [Internet]. 2015;3:44.

533 Available from: http://dx.doi.org/10.1186/s40168-015-0107-4

534 6. WHO. Capture the Moment. Early initiation of breastfeeding: The best start for every

535 newborn. http://www.who.int/nutrition/publications/infantfeeding/capture-moment-

536 early-initiation-bf/en/. 2018;41.

537 7. Scientific Advisory Committee on Nutrition (SACN).

538 //www.gov.uk/government/publications/sacn-report-on-feeding-in-the-first-year-of-

539 life. Feeding in the First Year of Life 2018 [Internet]. 2018. Available from:

540 https://www.gov.uk/government/publications/sacn-report-on-feeding-in-the-first-year-

541 of-life

542 8. Forbes JD, Azad MB, Vehling L, Tun HM, Konya TB, Guttman DS, et al. Association

543 of Exposure to Formula in the Hospital and Subsequent Infant Feeding Practices With

544 Gut Microbiota and Risk of Overweight in the First Year of Life. 2018;1–11.

545 9. Ly NP, Litonjua A, Gold DR, Celedón JC. Gut microbiota, probiotics, and vitamin D:

546 Interrelated exposures influencing allergy, asthma, and obesity? J Allergy Clin

547 Immunol [Internet]. 2011 May [cited 2017 Jun 22];127(5):1087–94. Available from:

548 http://www.ncbi.nlm.nih.gov/pubmed/21419479

549 10. Foster KR, Bell T. Competition, not cooperation, dominates interactions among

550 culturable microbial species. Curr Biol [Internet]. 2012;22(19):1845–50. Available

551 from: http://dx.doi.org/10.1016/j.cub.2012.08.005

552 11. Milani C, Lugli GA, Duranti S, Turroni F, Bottacini F, Mangifesta M, et al. Genomic

553 Encyclopedia of Type Strains of the Genus Bifidobacterium. Appl Environ Microbiol

554 [Internet]. 2014 Oct 15;80(20):6290–302. Available from:

555 http://aem.asm.org/cgi/doi/10.1128/AEM.02308-14

556 12. Milani C, Lugli GA, Duranti S, Turroni F, Mancabelli L, Ferrario C, et al.

557 Bifidobacteria exhibit social behavior through carbohydrate resource sharing in the

558 gut. Sci Rep [Internet]. 2015;5(1):15782. Available from:

559 http://www.nature.com/articles/srep15782

560 13. Lugli GA, Milani C, Turroni F, Duranti S, Mancabelli L, Mangifesta M, et al.

561 Comparative genomic and phylogenomic analyses of the Bifidobacteriaceae family.

562 BMC Genomics [Internet]. 2017 Dec 1;18(1):568. Available from:

563 http://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-017-3955-4

564 14. Juge N, Tailford L, Owen CD. Sialidases from gut bacteria: a mini-review. Biochem

565 Soc Trans [Internet]. 2016 Feb 15;44(1):166–75. Available from:

566 http://biochemsoctrans.org/cgi/doi/10.1042/BST20150226

567 15. Egan M, Motherway MOC, Ventura M, van Sinderen D. Metabolism of sialic acid by

568 Bifidobacterium breve UCC2003. Appl Environ Microbiol. 2014;80(14):4414–26.

569 16. Ninonuevo MR, Park Y, Yin H, Zhang J, Ward RE, Clowers BH, et al. A strategy for

570 annotating the human milk glycome. J Agric Food Chem. 2006;54(20):7471–80.

571 17. Thomson P, Medina DA, Garrido D. Human milk oligosaccharides and infant gut

572 bifidobacteria: Molecular strategies for their utilization. Food Microbiol [Internet].

573 2017;1–10. Available from: http://dx.doi.org/10.1016/j.fm.2017.09.001

574 18. Sela DA, Chapman J, Adeuya A, Kim JH, Chen F, Whitehead TR, et al. The genome

575 sequence of Bifidobacterium longum subsp. infantis reveals adaptations for milk

576 utilization within the infant microbiome. Proc Natl Acad Sci U S A [Internet].

577 2008;105(48):18964–9. Available from:

578 http://www.pnas.org/cgi/content/long/105/48/18964

579 19. Ashida H, Miyake A, Kiyohara M, Wada J, Yoshida E, Kumagai H, et al. Two distinct

580 -L-fucosidases from Bifidobacterium bifidum are essential for the utilization of

581 fucosylated milk oligosaccharides and glycoconjugates. Glycobiology [Internet]. 2009

582 Sep 1;19(9):1010–7. Available from: https://academic.oup.com/glycob/article-

583 lookup/doi/10.1093/glycob/cwp082

584 20. Ruiz-Moyano S, Totten SM, Garrido DA, Smilowitz JT, Bruce German J, Lebrilla CB,

585 et al. Variation in consumption of human milk oligosaccharides by infant gut-

586 associated strains of bifidobacterium breve. Appl Environ Microbiol.

587 2013;79(19):6040–9.

588 21. Garrido D, Ruiz-Moyano S, Kirmiz N, Davis JC, Totten SM, Lemay DG, et al. A

589 novel gene cluster allows preferential utilization of fucosylated milk oligosaccharides

590 in Bifidobacterium longum subsp. longum SC596. Sci Rep [Internet]. 2016 Dec

591 19;6(1):35045. Available from: http://www.nature.com/articles/srep35045

592 22. James K, Motherway MO, Bottacini F, van Sinderen D. Bifidobacterium breve

593 UCC2003 metabolises the human milk oligosaccharides lacto-N-tetraose and lacto-N-

594 neo-tetraose through overlapping, yet distinct pathways. Sci Rep [Internet].

595 2016;6(1):38560. Available from: http://www.nature.com/articles/srep38560

596 23. James K, Motherway MOC, Penno C, O’Brien RL, van Sinderen D. Bifidobacterium

597 breve UCC2003 employs multiple transcriptional regulators to control metabolism of

598 particular human milk oligosaccharides. Appl Environ Microbiol. 2018;84(9).

599 24. Nishiyama K, Yamamoto Y, Sugiyama M, Takaki T, Urashima T, Fukiya S, et al.

600 Bifidobacterium bifidum Extracellular Sialidase Enhances Adhesion to the Mucosal

601 Surface and Supports Carbohydrate Assimilation. Turner MS, editor. MBio [Internet].

602 2017 Nov 8;8(5):e00928-17. Available from:

603 http://linkinghub.elsevier.com/retrieve/pii/S1075996418300891

604 25. Nishiyama K, Nagai A, Uribayashi K, Yamamoto Y, Mukai T, Okada N. Two

605 extracellular sialidases from Bifidobacterium bifidum promote the degradation of

606 sialyl-oligosaccharides and support the growth of Bifidobacterium breve. Anaerobe

607 [Internet]. 2018; Available from:

608 http://linkinghub.elsevier.com/retrieve/pii/S1075996418300891

609 26. Smilowitz JT, O’Sullivan A, Barile D, German JB, Lonnerdal B, Slupsky CM. The

610 Human Milk Metabolome Reveals Diverse Oligosaccharide Profiles. J Nutr [Internet].

611 2013;143(11):1709–18. Available from:

612 http://jn.nutrition.org/cgi/doi/10.3945/jn.113.178772

613 27. Ward RE, Niñonuevo M, Mills DA, Lebrilla CB, German JB. In vitro fermentation of

614 breast milk oligosaccharides by Bifidobacterium infantis and Lactobacillus gasseri.

615 Appl Environ Microbiol. 2006;72(6):4497–9.

616 28. Matsuki T, Yahagi K, Mori H, Matsumoto H, Hara T, Tajima S, et al. A key genetic

617 factor for fucosyllactose utilization affects infant gut microbiota development. Nat

618 Commun [Internet]. 2016;7(May):11939. Available from:

619 http://www.nature.com/doifinder/10.1038/ncomms11939

620 29. Kato K, Odamaki T, Mitsuyama E, Sugahara H, Xiao J zhong, Osawa R. Age-Related

621 Changes in the Composition of Gut Bifidobacterium Species. Curr Microbiol.

622 2017;74(8):987–95.

623 30. Nagpal R, Kurakawa T, Tsuji H, Takahashi T, Kawashima K, Nagata S, et al.

624 Evolution of gut Bifidobacterium population in healthy Japanese infants over the first

625 three years of life: A quantitative assessment. Sci Rep. 2017;7(1):1–11.

626 31. Freitas AC, Hill JE. Bifidobacteria isolated from vaginal and gut microbiomes are

627 indistinguishable by comparative genomics. PLoS One. 2018;13(4):1–16.

628 32. Dominguez-Bello MG, De Jesus-Laboy KM, Shen N, Cox LM, Amir A, Gonzalez A,

629 et al. Partial restoration of the microbiota of cesarean-born infants via vaginal

630 microbial transfer. Nat Med [Internet]. 2016;22(3):250–3. Available from:

631 http://www.nature.com/doifinder/10.1038/nm.4039

632 33. Chu DM, Ma J, Prince AL, Antony KM, Seferovic MD, Aagaard KM. Maturation of

633 the infant microbiome community structure and function across multiple body sites

634 and in relation to mode of delivery. Nat Med [Internet]. 2017;23(3):314–326 13.

635 Available from: http://www.nature.com/doifinder/10.1038/nm.4272

636 34. Browne HP, Forster SC, Anonye BO, Kumar N, Neville BA, Stares MD, et al.

637 Culturing of ‘unculturable’ human microbiota reveals novel taxa and extensive

638 sporulation. Nature [Internet]. 2016;533(7604):in press. Available from:

639 http://www.nature.com/doifinder/10.1038/nature17645

640 35. Engfer MB, Stahl B, Finke B, Sawatzki G, Daniel H. Human milk oligosaccharides are

641 resistant to enzymatic hydrolysis in the upper gastrointestinal tract. Am J Clin Nutr

642 [Internet]. 2000;71(6):1589–96. Available from:

643 http://www.ncbi.nlm.nih.gov/pubmed/10837303

644 36. Turroni F, Milani C, Duranti S, Mahony J, van Sinderen D, Ventura M. Glycan

645 Utilization and Cross-Feeding Activities by Bifidobacteria. Trends Microbiol

646 [Internet]. 2017 Oct;xx:1–12. Available from:

647 http://linkinghub.elsevier.com/retrieve/pii/S0966842X17302287

648 37. Schwab C, Ruscheweyh HJ, Bunesova V, Pham VT, Beerenwinkel N, Lacroix C.

649 Trophic interactions of infant bifidobacteria and eubacterium hallii during L-fucose

650 and fucosyllactose degradation. Front Microbiol. 2017;

651 38. Wolfe AJ. The Acetate Switch. Microbiol Mol Biol Rev [Internet]. 2005;69(1):12–50.

652 Available from: http://mmbr.asm.org/cgi/doi/10.1128/MMBR.69.1.12-50.2005

653 39. Halsey CR, Lei S, Wax JK, Lehman MK, Nuxoll AS, Steinke L, et al. Amino Acid

654 Catabolism in Staphylococcus aureus and the Function of Carbon Catabolite

655 Repression. MbioAsmOrg. 2017;1–19.

656 40. Rivière A, Selak M, Geirnaert A, Van den Abbeele P, De Vuyst L. Complementary

657 degradation mechanisms of inulin-type fructans and arabinoxylan-oligosaccharides

658 among bifidobacterial strains suggest bacterial cooperation. Appl Environ Microbiol

659 [Internet]. 2018;84(9):AEM.02893-17. Available from:

660 http://aem.asm.org/lookup/doi/10.1128/AEM.02893-17

661 41. Maldonado-Gómez MX, Martínez I, Bottacini F, O’Callaghan A, Ventura M, van

662 Sinderen D, et al. Stable Engraftment of Bifidobacterium longum AH1206 in the

663 Human Gut Depends on Individualized Features of the Resident Microbiome. Cell

664 Host Microbe [Internet]. 2016 Oct;20(4):515–26. Available from:

665 http://linkinghub.elsevier.com/retrieve/pii/S193131281630378X

666 42. Nadell CD, Xavier JB, Foster KR. The sociobiology of biofilms. FEMS Microbiol

667 Rev. 2009;33(1):206–24.

668 43. Alcon-Giner C, Caim S, Mitra S, Ketskemety J, Wegmann U, Wain J, et al.

669 Optimisation of 16S rRNA gut microbiota profiling of extremely low birth weight

670 infants. BMC Genomics [Internet]. 2017 Dec 2;18(1):841. Available from:

671 http://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-017-4229-x

672 44. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA

673 ribosomal RNA gene database project: Improved data processing and web-based tools.

674 Nucleic Acids Res. 2013;41(D1):590–6.

675 45. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, et al. Gapped

676 BLAST and PSI-BLAST: A new generation of protein database search programs.

677 Nucleic Acids Res. 1997;25(17):3389–402.

678 46. Huson D, Mitra S, Ruscheweyh H. Integrative analysis of environmental sequences

679 using MEGAN4. Genome Res. 2011;21(9):1552–60.

680 47. Page AJ, De Silva N, Hunt M, Quail MA, Parkhill J, Harris SR, et al. Robust high-

681 throughput prokaryote de novo assembly and improvement pipeline for Illumina data.

682 Microb genomics [Internet]. 2016 Aug 25;2(8):e000083. Available from:

683 http://www.microbiologyresearch.org/content/journal/mgen/10.1099/mgen.0.000083

684 48. Geer LY, Marchler-Bauer A, Geer RC, Han L, He J, He S, et al. The NCBI

685 BioSystems database. Nucleic Acids Res [Internet]. 2010 Jan;38(suppl_1):D492–6.

686 Available from: https://academic.oup.com/nar/article-lookup/doi/10.1093/nar/gkp858

687 49. Seemann T. Prokka: rapid prokaryotic genome annotation. Bioinformatics [Internet].

688 2014 Jul 15;30(14):2068–9. Available from:

689 https://academic.oup.com/bioinformatics/article-

690 lookup/doi/10.1093/bioinformatics/btu153

691 50. Page AJ, Cummins CA, Hunt M, Wong VK, Reuter S, Holden MTG, et al. Roary:

692 Rapid large-scale prokaryote pan genome analysis. Bioinformatics. 2015;31(22):3691–

693 3.

694 51. Katoh K, Standley DM. MAFFT multiple sequence alignment software version 7:

695 Improvements in performance and usability. Mol Biol Evol. 2013;30(4):772–80.

696 52. Castresana J. Selection of conserved blocks from multiple alignments for their use in

697 phylogenetic analysis. Mol Biol Evol [Internet]. 2000 Apr;17(4):540–52. Available

698 from: http://www.ncbi.nlm.nih.gov/pubmed/10742046

699 53. Talavera G, Castresana J. Improvement of phylogenies after removing divergent and

700 ambiguously aligned blocks from protein sequence alignments. Syst Biol [Internet].

701 2007 Aug;56(4):564–77. Available from:

702 http://www.ncbi.nlm.nih.gov/pubmed/17654362

703 54. Gouy M, Guindon S, Gascuel O. Sea view version 4: A multiplatform graphical user

704 interface for sequence alignment and phylogenetic tree building. Mol Biol Evol.

705 2010;27(2):221–4.

706 55. Guindon S, Dufayard JF, Lefort V, Anisimova M, Hordijk W, Gascuel O. New

707 algorithms and methods to estimate maximum-likelihood phylogenies: Assessing the

708 performance of PhyML 3.0. Syst Biol. 2010;59(3):307–21.

709 56. Pritchard L, Glover RH, Humphris S, Elphinstone JG, Toth IK. Genomics and

710 taxonomy in diagnostics for food security: soft-rotting enterobacterial plant pathogens.

711 Anal Methods [Internet]. 2016;8(1):12–24. Available from:

712 http://xlink.rsc.org/?DOI=C5AY02550H

713 57. Huerta-Cepas J, Forslund K, Coelho LP, Szklarczyk D, Jensen LJ, Von Mering C, et

714 al. Fast genome-wide functional annotation through orthology assignment by

715 eggNOG-mapper. Mol Biol Evol. 2017;34(8):2115–22.

716 58. Huerta-Cepas J, Szklarczyk D, Forslund K, Cook H, Heller D, Walter MC, et al.

717 EGGNOG 4.5: A hierarchical orthology framework with improved functional

718 annotations for eukaryotic, prokaryotic and viral sequences. Nucleic Acids Res.

719 2016;44(D1):D286–93.

720 59. Yin Y, Mao X, Yang J, Chen X, Mao F, Xu Y. DbCAN: A web resource for

721 automated carbohydrate-active enzyme annotation. Nucleic Acids Res.

722 2012;40(W1):445–51.

723 60. Zhou Y, Liang Y, Lynch KH, Dennis JJ, Wishart DS. PHAST: a fast phage search

724 tool. Nucleic Acids Res. 2011 Jul;39(Web Server issue):W347-52.

725 61. Arndt D, Grant JR, Marcu A, Sajed T, Pon A, Liang Y, et al. PHASTER: a better,

726 faster version of the PHAST phage search tool. Nucleic Acids Res. 2016

727 Jul;44(W1):W16-21.

728 62. Fanning S, Hall LJ, Cronin M, Zomer A, MacSharry J, Goulding D, et al.

729 Bifidobacterial surface-exopolysaccharide facilitates commensal-host interaction

730 through immune modulation and pathogen protection. Proc Natl Acad Sci [Internet].

731 2012 Feb 7;109(6):2108–13. Available from:

732 http://www.scopus.com/inward/record.url?eid=2-s2.0-

733 84857128238&partnerID=tZOtx3y1

734

735

736 Figure 1: Diverse Bifidobacterium strains dominate the early life microbiota (A) Faecal

737 bacterial community profiles of three healthy, full-term infants as determined by 16S rRNA

738 gene sequencing. Paired-end reads were generated using the MiSeq Illumina platform, all

739 data sets were normalised and relative abundance of each bacterial taxa is represented in

740 percentages of number of total reads for the top 10 most prevalent genus in each infant. Bar

741 colours represent different genus taxa, and bar lengths signify the relative abundance of each

742 taxon. 16S rRNA bacterial profiles are named according to the sample used for bifidobacteria

743 isolation. V1 at 102 days of age, V2 at 174 days of age, and V3 at 159 days of age. The

744 number of reads obtained by 16S rRNA gene sequencing data for each sample can be found

745 in Table S2. (B) Core-genome phylogeny of 83 Bifidobacterium isolates, 19 of which are

746 novel strains identified in this study and denoted by arrows and the annotation of LHXX.

747 Isolates from infant V1 are denoted with a red star, V2 with a purple circle and V3 with a

748 blue triangle. Bootstrap values >70 are shown with a yellow square on each node. (C) ORFs

749 from each genome was submitted to eggNOG-mapper

750 (http://eggnogdb.embl.de/#/app/emapper) for functional classification. The proportion of

751 ORFs for each classification was calculated and is presented as a percentage of total ORFs in

752 each genome. The values indicated for each orthologous group represents the average

753 percentage of ORFs in that group for all genomes in that infant.

754

755 Figure 2: Functional classification of Bifidobacterium genomes. (A) Presence of genes

756 encoding glycosyl hydrolases was determined using the dbCAN server

757 (http://csbl.bmb.uga.edu/dbCAN/) which annotates genes based on HMMs of GH generated

758 from data in the CAZy database (http://www.cazy.org). The heatmap shows the number of

759 ORFs annotated as GH for each GH family (y-axis) for each genome (x-axis) (Enumeration

760 of GH ORFs in subfamilies of GH5, GH13 and GH43 is shown in Supplementary figure 3B).

761

762 Figure 3. Bifidobacterial survival after exposure to intestinal environmental stressors.

763 (A) Bacterial growth in MRS was measured for each strain after 30h growth in an aerobic

764 environment and compared to parallel samples grown anaerobically. Measurements shown

765 are representative of three independent experimental repeats. (B) Ability to withstand acid

766 shock for 4 h at a pH2. Following acid shock, strains were incubated for 48 h in an anaerobic

767 cabinet at 37˚C. Growth (optical density) was then measured. Data shown is mean ± standard

768 deviation for 3 independent experiments (C) Bacterial survival in MRS supplemented with

769 0.3% bile salts was measured after 48h growth and compared to parallel cultures grown in

770 MRS anaerobically at 37˚C. Data shown from five independent experiments; n.d. denotes

771 measurement not detectable. Blank/media only values are subtracted from all points in each

772 experiment, all negative values are represented as zero. Isolates from Baby V1 are

773 represented as black bars, V2 are grey bars, V3 are dotted bars, and the tested type and

774 control strains are shown as white bars. Error bars represent SEM.

775

776 Figure 4. HMOs function as key carbon sources for Bifidobacterium growth. (A) HMO

777 heatmap illustrating the presence of known HMO gene clusters in the 19 novel infant isolates

778 (see Methods for details); (B) chemical structure of HMO 2’FL and LNnT generated by

779 ChemDraw; (C-D) Growth kinetics of all 19 strains in mMRS with either HMO 2’FL (C) or

780 LNnT (D) as a sole carbon source; all strains from each infant are illustrated together. Data

781 shown is a representative graph of three independent experimental repeats, containing the

782 mean from duplicate/triplicate well measurements.

783

784 Figure 5. Cross-feeding networks exist within Bifidobacterium communities. (A)

785 Schematic illustrating experimental set-up and preparation of conditioned media for cross-

786 feeding experiments. (B) Using conditioned media from strains capable of using either HMO

787 2’FL (B) or LNnT (C) was combined with fresh mMRS (1:1) and used as growth media for

788 all HMO non-degrader isolates within the same ecological niche. Data shown is a

789 representative graph of three independent experimental repeats, containing the mean from

790 duplicate/triplicate well measurements. (D)1H NMR spectrum to identify metabolites

791 involved in cross-feeding between 2’FL-degrader B. longum LH206 and non-degrader B.

792 pseudocatenulatum LH659 and semi-quantification of identified metabolites. (E) 1H NMR

793 spectrum to identify metabolites involved in cross-feeding between LNnT-degrader B.

794 longum LH206 and non-degrader B. pseudocatenulatum LH657 and semi-quantification of

795 identified metabolites. n.d indicates metabolites were not detected.

796

797 Figure 6. Identification of metabolites involved within cross-feeding. (A)1H NMR

798 spectrum to identify metabolites involved in cross-feeding between 2’FL-degrader B. longum

799 LH206 and non-degrader B. pseudocatenulatum LH659 and semi-quantification of identified

800 metabolites. (E) 1H NMR spectrum to identify metabolites involved in cross-feeding between

801 LNnT-degrader B. longum LH206 and non-degrader B. pseudocatenulatum LH657 and semi-

802 quantification of identified metabolites. n.d indicates metabolites were not detected.

803

804 Figure 7: Schematic of the overview for the syntrophic cross-feeding network for HMO

805 2'FL. We have built a schematic combining the known identified metabolites generated and

806 consumed from 2’FL degradation by Bifidobacterium species. Known structures used to

807 import 2’FL into either Bifidobacterium isolates LH13 and LH206 (HMO degrader) are

808 denoted in green and blue, respectively. Secreted by-products from 2’FL metabolism are also 34

809 labelled depending on which strain produced them; LH13 (green), LH206 (blue), or both

810 (orange). Identified substrates generated after cross-feeding to non-HMO utilisers are shown

811 as orange for LH12, or navy for both LH12 and LH659. Additional unidentified mechanisms

812 and metabolites that may be involved in this event are denoted by a question mark (?).

813 Abbreviations: SBP, substrate binding protein; P, permease; GH95, glycoside hydrolase

814 family 95 protein, 2'FL 2'-Fucosyllactose.

815

B. longum A 100 B

B. longum subsp infantis TPY12-1 B. longum subsp infantis BIC1206122787

B. longum subsp infantis BIB1401242951 B. longumB. subsp longum infantis-IN-F29 subsp infantis EK3

B. longum NCC2705

B. longum subsp infantis BIC1307292462 B. longum DJO10A

bioRxiv preprint doi: https://doi.org/10.1101/711234; this version posted July 23, 2019.B. longum The subsp infantiscopyright BIC1401111250 holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved.B. longumNo subspreuse infantis BIC1401212621aallowed without permission.

LH 665 LH 664 B. longum subsp infantis BIC1401212621b LH 666

LH 23

50 B. longum subsp infantis DSM20088 B. longum subsp longum BBMN68 B. longum BG7 B. longum subsp infantis CCUG52486 B. longum subsp infantis CECT7210 B. longum 105A B. longum subsp longum NCIMB8809 B. longum subsp longum EK13 B. longum subsp longum DSM20219 B. longum 35624 B. longum subsp infantis BT1 B. longum subsp longum CCUG30698 B. longum subsp longum GT15 B. longum subsp infantis 1888B B. longum subsp longum LO-21 B. longum subsp longum AH1206 gene reads B. longum subsp infantis IN-07 B. longum subsp longum KACC91563 B. longum subsp infantis 157F LH 12 % of 16S rRNA B. longum subsp longum JDM301 LH 206 LH 277 B. longum BXY01 B. breve MCC0476 B. pseudocatenulatum CAC29 B. breve JCM7017 B. pseudocatenulatum 1896B 0 B. pseudocatenulatum CA05 B. breve DSM20213 B. pseudocatenulatum IPLA36007 LH 24 B. pseudocatenulatum 1E LH 21 LH 663 Infant V1 Infant V2 Infant V3 LH 662 LH 659 B. breve B.CECT7263 breve S27 LH 658 B. breve 31L LH 656 B. pseudocatenulatumLH 657 CAB29 B. pseudocatenulatum CECT7765 Bifidobacterium Tyzzerella Gemella Terrisporobacter B. pseudocatenulatum DSM20438 B. breve B. pseudocatenulatum 2789STDY5834840 B. breve 12L B. pseudocatenulatum CAK29a B. pseudocatenulatum CAD29 B. pseudocatenulatum C1 B. pseudocatenulatum C15 B. pseudocatenulatum C62 B. pseudocatenulatum C55 Streptococcus Erysipelatoclostridium Lachnoanaerobaculum Corynebacterium B. pseudocatenulatum C95 B. breve ACS-071-V-Sch8b B. breve NCFB2258 Escherichia Bilophila Rothia Enterobacter B. breve UCC2003 LH 14 LH 9 B. breve DRBB26 LH 11

B. breve 689b LH 13 B. breve BR3 Bacteroides Klebsiella Citrobacter Roseburia B. breve JCP7499

B. breve LMC520

Actinomyces Scardovia Clostridium Dorea B. breve JCM7019 Granulicatella Faecalitalea Dolosigranulum Intestinibacter

Enterococcus Citrobacter Parabacteroides Peptostreptococcus B. pseudocatenulatum Staphylococcus Lachnoanaerobaculum Blautia Cutibacterium

0.77% 2.86% 2.71% 2.83% 2.81% C 2.57% 3.02% 0.99%

8.62% 7.85%

2.69% 1.01% 15.66% 2.42% 0.94% 16.16% 3.40% 15.65% 3.27% 3.15% 0.95% 3.34% 8.36% 9.53%

10.06% Infant V1 10.32% Infant V2 Infant V3 0.30% 0.29% 4.75% 4.55% 0.22% 4.65% 3.96% 6.76% 1.57% 2.07% 2.63% 1.97% 1.94% 2.16% 1.81% 2.50% 3.76% 5.87% 5.62% 2.48% 3.96% 6.36% 7.04% 6.25% 7.66% 6.81% 5.40%

A : RNA processing and modification J : Translation, ribosomal structure Q : Secondary metabolites biosynthesis, and biogenesis transport, and catabolism C : Energy production and conversion K : Transcription S : Function unknown D : Cell cycle control, cell division, chromosome partitioning L : Replication, recombination and repair T : Signal transduction mechanisms E : Amino acid transport and metabolism M : Cell wall/membrane/envelope biogenesis U : Intracellular trafficking, secretion, and vesicular transport F : Nucleotide transport and metabolism N : Cell motility V : Defence mechanisms G : Carbohydrate transport and metabolism O : Post-translational modification, protein turnover, and chaperones W : Extracellular structures H : Coenzyme transport and metabolism P : Inorganic ion transport and metabolism Z : Cytoskeleton I : Lipid transport and metabolism Colour Key % of CDS 0 bioRxiv preprint 5 015 10 doi: certified bypeerreview)istheauthor/funder.Allrightsreserved.Noreuseallowedwithoutpermission. B. pseudocatenulatum LH9 https://doi.org/10.1101/711234 B. pseudocatenulatum

LH11 V1 B. longum LH12 B. pseudocatenulatum LH13 B. pseudocatenulatum LH14 ; B. breve LH21 this versionpostedJuly23,2019. V2 B. infantis LH23 B. breve LH24

B. longum LH206 B. longum LH277 B. pseudocatenulatum LH656 B. pseudocatenulatum LH657 The copyrightholderforthispreprint(whichwasnot

B. pseudocatenulatum LH658 V3 B. pseudocatenulatum LH659 B. pseudocatenulatum LH662 B. pseudocatenulatum LH663 B. infantis LH664 B. infantis LH665 B. infantis LH666 GH136 GH129 GH127 GH125 GH123 GH121 GH120 GH112 GH109 GH101 GH95 GH85 GH78 GH77 GH53 GH51 GH43 GH42 GH38 GH36 GH35 GH33 GH32 GH31 GH30 GH29 GH27 GH25 GH23 GH20 GH18 GH13 GH10 GH8 GH5 GH4 GH3 GH2 GH1 A

2.0 V1 V2 V3 bioRxiv preprint doi: https://doi.org/10.1101/711234; this version posted July 23, 2019. The copyright holder for this preprint (which was not 1.5 certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission.

1.0

0.5

n.d. 0.0 after acid shock at pH2 600 OD B. breve LH21 B. breve LH24 B. infantis LH23 B. longum LH12 B. infantis LH664 B. infantis LH665 B. infantis LH666 B. longum LH206 B. longum LH277 B. pseudocatenulatum LH9 B. pseudocatenulatum LH11 B. pseudocatenulatum LH13 B. pseudocatenulatum LH14 B. pseudocatenulatum LH656 B. pseudocatenulatum LH657 B. pseudocatenulatum LH658 B. pseudocatenulatum LH659 B. pseudocatenulatum LH662 B. pseudocatenulatum LH663

B Anaerobic growth V1 V2 V3 Aerobic growth 2.0 1.5 1.0 0.5

0.10

0.05

in MRS after 48h growth 0.00 600 OD B. breve LH21 B. breve LH24 B. infantis LH23 B. longum LH12 B. infantis LH664 B. infantis LH665 B. infantis LH666 B. longum LH206 B. longum LH277 B. breve DSM 20213 B. infantis DSM 2021 9 L. casei NCIMB4114 B. longum DSM 20088 B. pseudocatenulatum LH9 B. pseudocatenulatum LH11 B. pseudocatenulatum LH13 B. pseudocatenulatum LH14 B. pseudocatenulatum LH656 B. pseudocatenulatum LH657 B. pseudocatenulatum LH658 B. pseudocatenulatum LH659 B. pseudocatenulatum LH662 B. pseudocatenulatum LH663 B. pseudocatenulatum DSM 20439 B. pseudocatenulatum DSM 20438 C V1 V2 V3 MRS alone 2.0 MRS + 0.3% bile salt 1.5 1.0 0.5 0.25 0.20 0.15 0.10 0.05

after 48h growth 0.00 in MRS ± 0.3% bile salt 600 OD B. breve LH21 B. breve LH24 B. infantis LH23 B. longum LH12 B. infantis LH664 B. infantis LH665 B. infantis LH666 B. longum LH206 B. longum LH277 B. breve DSM 20213 B. infantis DSM 2021 9 L. casei NCIMB4114 B. longum DSM 20088 B. pseudocatenulatum LH9 B. pseudocatenulatum LH11 B. pseudocatenulatum LH13 B. pseudocatenulatum LH14 B. pseudocatenulatum LH656 B. pseudocatenulatum LH657 B. pseudocatenulatum LH658 B. pseudocatenulatum LH659 B. pseudocatenulatum LH662 B. pseudocatenulatum LH663 B. pseudocatenulatum DSM 20438 B. pseudocatenulatum DSM 20439 2’Fucosyllactose (2’FL) Lacto-N-neotetraose (LNnT) A B O OH

HO HO HMO clusters HO OH OH HO OH O O O O bioRxiv preprint doi: https://doi.org/10.1101/711234; this version posted JulyO 23,O 2019. The copyright holder for this preprint (which was not HO certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. OH HO OH

HO OH HO OO O O O HO

N_RS12215 HO O NH

OH OH HO OH O

OH Complete cluster

12070-BLO Partial cluster C Infant V1 D Infant V1 B. pseudocatenulatum LH9 1.25 Absent cluster 1.25 B. pseudocatenulatum LH11 1.00 1.00 B. longum LH12

BLLJ_08355-BLLJ_08385 BBR_RS13080-BBR_RS13100 BBR_RS18470-BBR_RS18480 BBR_RS18490-BBR_RS18520 BBR_RS18650-BBR_RS18675 BLON_RS BBPC_RS08935-BBPC_RS08975 B. pseudocatenulatum LH13 0.75 0.75 B. pseudocatenulatum LH9 B. pseudocatenulatum LH14 B. pseudocatenulatum LH11 0.50 0.50 in mMRS in 2% w/v LNnT in 2% w/v

in 2% w/v 2’FL 2’FL in 2% w/v 0.25 B. longum LH12 in mMRS 0.25

V1 600 B. pseudocatenulatum LH13 600 0.00 0.00 20 40 20 40 OD B. pseudocatenulatum LH14 OD -0.25 Time (hr) -0.25 Time (hr)

B. breve LH21 B. infantis LH23 B. breve LH21 V2 1.25 Infant V2 1.25 Infant V2 B. breve LH24 B. infantis LH23 1.00 1.00 B. breve LH24 B. longum LH206 0.75 0.75

B. longum LH277 0.50 0.50 B. pseudocatenulatum LH656 in 2% w/v 2’FL 2’FL in 2% w/v

in mMRS 0.25 0.25 in mMRS in 2% w/v LNnT in 2% w/v 600 B. pseudocatenulatum LH657 0.00 600 0.00 20 40 20 40 OD

B. pseudocatenulatum LH658 OD V3 Time (hr) -0.25 Time (hr) -0.25 B. longum LH206 B. pseudocatenulatum LH659 B. pseudocatenulatum LH662 B. longum LH277 B. pseudocatenulatum LH663 B. pseudocatenulatum LH656 B. infantis LH664 B. pseudocatenulatum LH657 1.25 Infant V3 1.25 Infant V3 B. infantis LH665 B. pseudocatenulatum LH658 1.00 1.00 B. infantis LH666 B. pseudocatenulatum LH659 0.75 0.75 B. pseudocatenulatum LH662 0.50 0.50 B. pseudocatenulatum LH663 in mMRS in 2% w/v LNnT in 2% w/v 0.25 in 2% w/v 2’FL 2’FL in 2% w/v in mMRS 0.25 B. infantis LH664 600 600 0.00 0.00 20 40 20 40 B. infantis LH665 OD OD -0.25 Time (hr) -0.25 Time (hr) B. infantis LH666 A C 0.2 Infant V2

x Cross-feeding from een bioRxiv preprint doi: https://doi.org/10.1101/711234Spin down & Conditioned; this version media posted July 23, 2019. The copyright holder for thisB preprint. infantis (whichLH23 was not

O 0.1

OH O

HO O HO OH HO OH

OH O O O O HO HO HO

HO HO OH O HO OH O OH OH OH HO OH sterile filter

O O O HO O O O HO t0 and t

HO OH

O O O HO

HO OH HMOs O OH O

O OH

OH HMO

O O OH

O HO 600

HO 0.0 OH HO O

OH ce betw en ce HMO degrader O O

HO HO non-degrader 10 20 30 40 50

OD Time (h)

Di ff er -0.1 B Infant V1 Infant V2 Infant V3 Cross-feeding to Cross-feeding from 0.2 x

0.4 een 0.4 Supernatant from LH206 x x

een B. longum LH12 B. infantis LH23 een 0.3 0.3 0.1 0.2 t0 and t 0.2 t0 and t t0 and t 600 ce betw en ce 600 600

ce betw en ce 0.1 0.1 50

ce betw en ce 0.0

OD 10 20 30 40 OD OD

0.0 Di ff er 0.0 Time (h) 10 20 30 40 50

Di ff er 10 20 30 40 50 Di ff er -0.1 -0.1 Time (h) -0.1 Time (h) Supernatant source Isolates tested 0.2 Supernatant from LH277 x LH9 LH13 LH11 LH14 LH21 LH24 een 0.1

Infant V3 t0 and t 600 0.4 Supernatant from LH206 0.4 Supernatant from LH664 betw en ce 0.0 10 20 30 40 50 OD Time (h) een x x 0.3

0.3 een Di ff er -0.1 0.2 0.2 0.2 Supernatant from LH664 t0 and t t0 and t 0.1

0.1 x een ce betw en ce 600 600 ce betw en ce 0.0 0.0 0.1 10 20 30 40 50 OD OD 10 20 30 40 50 Di ff er

-0.1 Time (h) -0.1 Time (h) t0 and t Di ff er

600 0.0 ce betw en ce 10 20 30 40 50

Supernatant from LH277 Supernatant from LH665 OD Time (h) 0.4 0.4 een x x Di ff er

een -0.1 0.3 0.3 0.2 Supernatant from LH665 0.2 0.2 x t0 and t t0 and t een ce betw en ce 600 0.1 600 0.1

ce betw en ce 0.1 OD OD 0.0 0.0 t0 and t

Di ff er 10 20 30 40 50 10 20 30 40 50 Di ff er 600

-0.1 Time (h) -0.1 Time (h) betw en ce 0.0 10 20 30 40 50 Isolates tested OD Time (h) LH656 LH657 LH658 LH659 LH662 LH663 LH666 Di ff er -0.1 A 2'FL cross feeding metabolites n.d. 2’FL spectrum Pyruvate B. longum LH206 bioRxiv preprint doi: https://doi.org/10.1101/711234; this version posted July 23, 2019. The copyright holder for this preprint (which was not (HMO degr certified by peer review) is the author/funder. All rights reserved.Acetate No reuse allowed without permission. ader) Galactose Fucose Fucose Formate B. pseudocatenulatum LH659 Fucose Fucose Fucose Galactose Lactose Galactose (water) Lactose Glucose 0 1 2 3 20 40 60 Relative abundance (AU)

B LNnT cross feeding metabolites LNnT spectrum B. longum LH206 Ethylene glycol Ethanol (HMO degrader)

Acetate B. pseudocatenulatum LH663 N-acetylglucosamine N-acetylglucosamine Formate

Galactose Galactose n.d. Galactose (water) N-acetyl glucosamine 0 1 2 3 20 40 60 Relative abundance (AU) OH

HO OH O

O O O O bioRxiv preprint doi: https://doi.org/10.1101/711234; this version posted July 23, 2019. The copyright holder for this preprint (which was not HO OH certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. O O HO HO OH OH OH HO OH

O O 2' FL SBPO O HO OH O O HO HO OH OH

P P O OH HO O O OH HO OH OH HO OH HO OH OH

O O ? O O Lactose HO OH OH O Sugar metaboism O HO GH95 HO OH HO OH ? OH O ? Fucose HMO degrader Acetate Fucose Galactose Pyruvate Formate Ethanol Lactose Unidentified metabolites?

? ? Acetate Pyruvate ? Metabolism Ethanol

HMO non-degraders