Schellekens_BLongum_Obesity_Stress_EBIOM_103176 (003)

EBioMedicine xxx (xxxx) 103176

Contents lists available at ScienceDirect

EBioMedicine journal homepage:

Research paper Bifidobacterium longum counters the effects of obesity: Partial successful translation from rodent to human Harriët Schellekens a , b , # , £ , ⁎ , Cristina Torres-Fuentes a , # , Marcel van de Wouw a , # , Caitriona M. Long-Smith a , Avery Mitchell c , Conall Strain c , Kirsten Berding a , Thomaz F.S. Bastiaanssen a , b , Kieran Rea a , Anna V. Golubeva a , b , Silvia Arboleya a , c , Mathieu Verpaalen a , b , Matteo M. Pusceddu a , Amy Murphy a , c , Fiona Fouhy a , c , Kiera Murphy a , c , Paul Ross a , d , Bernard L. Roy e , Catherine Stanton a , c ,

Timothy G. Dinan a , f , John F. Cryan a , b , £ a APC Microbiome Ireland, University College Cork, Cork, Ireland b Department of Anatomy and Neuroscience, University College Cork, Cork, Ireland c Teagasc Food Research Centre, Moorepark, Fermoy, Cork, Ireland d College of Science Engineering & Food Science, University College Cork, Cork, Ireland e Cremo SA, Villars-sur-Glâne, Fribourg, Switzerland f Dept of Psychiatry and Behavioural Neuroscience, University College Cork, Cork, Ireland

ARTICLE INFO Article history: Received 6 May 2020 Received in revised form 13 October 2020 Accepted 3 December 2020 Available online xxx

ABSTRACT Background: The human gut microbiota has emerged as a key factor in the development of obesity. Certain pro- biotic strains have shown anti-obesity effects. The objective of this study was to investigate whether Bifidobac- terium longum APC1472 has anti-obesity effects in high-fat diet (HFD)-induced obese mice and whether B. longum APC1472 supplementation reduces body-mass index (BMI) in healthy overweight/obese individuals as the pri- mary outcome. B. longum APC1472 effects on waist-to-hip ratio (W/H ratio) and on obesity-associated plasma biomarkers were analysed as secondary outcomes. Methods: B. longum APC1472 was administered to HFD-fed C57BL/6 mice in drinking water for 16 weeks. In the human intervention trial, participants received B. longum APC1472 or placebo supplementation for 12 weeks, during which primary and secondary outcomes were measured at the beginning and end of the intervention. Findings: B. longum APC1472 supplementation was associated with decreased bodyweight, fat depots accumula- tion and increased glucose tolerance in HFD-fed mice. While, in healthy overweight/obese adults, the supplemen- tation of B. longum APC1472 strain did not change primary outcomes of BMI (0.03, 95% CI [-0.4, 0.3]) or W/ H ratio (0.003, 95% CI [-0.01, 0.01]), a positive effect on the secondary outcome of fasting blood glucose levels was found (-0.299, 95% CI [-0.44, -0.09]). Interpretation: This study shows a positive translational effect of B. longum APC1472 on fasting blood glucose from a preclinical mouse model of obesity to a human intervention study in otherwise healthy overweight and obese individuals. This highlights the promising potential of B. longum APC1472 to be developed as a valuable supple- ment in reducing specific markers of obesity. Funding: This research was funded in part by Science Foundation Ireland in the form of a Research Centre grant (SFI/12/RC/2273) to APC Microbiome Ireland and by a research grant from Cremo S.A. ©2020

Keywords Obesity Translational Fasting blood glucose Ghrelin Cortisol Gut microbiota Probiotic Bifidobacterium longum

For submission to EBioMedicine

⁎ Corresponding author. E-mail address: (H. Schellekens) # Equal contributing authors £ joint senior authors 2352-3964/© 2020.


H. Schellekens et al. / EBioMedicine xxx (xxxx) 103176

ciated with a persistent intestinal microbiome signature after successful dieting in obese mice [18]. Nonetheless, the exact mechanisms of how diet-induced changes in gut microbiota affect gut-brain signalling, including host metabolism, appetite regulation and brain health, are currently still lacking [19,20]. Interestingly, the obese-associated microbiota has been shown to have an increased capability to harvest energy from food and contributes to host insulin resistance, gut permeability, low-grade inflammation, and fat deposition [21,22]. Intestinal microbiota-derived metabolites have also been shown to impact the central regulation of appetite [9,23,24]. For example, certain bacterial strains modify gut peptides secretion, such as glucagon-like peptide (GLP) − 1, thus contributing to hypothal- amic appetite and satiety signalling via afferent nerve fibres of the va- gus nerve as well as by direct secretion into the circulatory system [24,25]. Furthermore, germ-free mice display marked decreases in ex- pression of intestinal satiety peptides, including cholecystokinin (CCK), peptide tyrosine-tyrosine (PYY) and GLP-1 and also lower circulating levels of leptin and ghrelin [26]. In addition, serum ghrelin levels are negatively correlated with the abundance of certain bacterial taxa, in- cluding Bifidobacterium and Lactobacillus species [27]. Moreover, intake of the prebiotic oligofructose, which promotes the growth of Bifidobac- terium and Lactobacillus , decreases the secretion of ghrelin in obese hu- mans [28]. Taken together, modulation of the gut microbiota is emerg- ing as a promising strategy for the management of obesity and obe- sity-related disorders such as type-2 diabetes and cardiovascular disease [4,7-9,29]. Several probiotic strains with different anti-obesity effects in humans have been identified [4,30-36]. The bacterial strain B. longum APC1472 has recently been shown to modulate ghrelinergic signalling in vitro [37], highlighting the therapeutic potential for host metabolism, ap- petite and obesity modulation. The ghrelin receptor (GHS-R1a) is ac- tivated by the endogenous hormone ghrelin, the first and only known peripheral orexigenic peptide, which regulates peripheral metabolism and energy expenditure as well as centrally regulated homeostatic ap- petite and food-motivated reward signalling, governing eating behav- iour and food intake [38-42]. Interestingly, obese individuals have at- tenuated postprandial suppression of ghrelin and a blunted nocturnal plasma ghrelin increase, reinforcing aberrant ghrelinergic signalling in obesity [43,44]. While the precise site of action of ghrelin is some- what controversial [45-47], the high prevalence of the ghrelin receptor throughout the small and large intestine, make it a likely target for in- teraction with the gut microbiota and thus may hold potential as a local therapeutic target [48]. As such, we investigated B. longum APC1472 for its ability to ame- liorate high-fat diet (HFD)-induced obesity in mice and observed signif- icant beneficial beneficial effects on adiposity and metabolism. Based on these promising effects of B. longum in the preclinical model, we subsequently investigated whether it could improve obesity symptoma- tology in healthy overweight/obese adults. The primary objective of the human intervention study was to determine whether a 12-week daily supplementation of B. longum APC1472 decreases body-mass index (BMI), while the secondary objective was to investigate the effects on waist-to-hip ration (W/H ratio), and biomarkers associated with obesity, such as glucose, insulin, HbA1c and ghrelin levels. The exploratory ob- jectives were to investigate the impact of B. longum APC1472 on the gut microbiota composition and diversity, peripheral inflammatory profile, stress hormone profile, self-reported perceived stress, anxiety and sati- ety. 2.Methods – animal study 2.1. Animals, diets and ethical approval Five-week-old male C57BL/6 mice (Harlan Laboratories, UK) (40 mice, n =8 – 10 per group) were housed in groups of 2 mice per cage in

Evidence before this study • Evidence has shown that the gut microbiota is an important component in the regulation of the host's physiology and metab- olism, modulating energy harvest, storage and expenditure and, therefore, represents a promising target in the treatment of obesity and obesity-related disorders. • Different probiotic strains have been shown to have differ- ent beneficial anti-obesity effects such as reduced body weight gain, improvements in insulin sensitivity and glucose uptake, and reduced fat depots accumulation in rodents. • The Bifidobacterium longum APC1472 strain was recently identified in our laboratory to modulate ghrelinergic signalling in vitro , which is an important signalling pathway modulating central appetite regulation and metabolism.

Added value of this study

• Research in context B. longum APC1472 demonstrated several significant beneficial effects in HFD-induced obese mice. • B. longum APC1472 reduced fasting glucose, cortisol awakening responses and increased active ghrelin in healthy obese adults. • Effectsof B. longum APC1472 partially translated from a preclin- ical mouse model to a human intervention study where this pro- biotic positively impacted markers of obesity.

Implications of available evidence

• B. longum APC1472 has promising potential to be developed as a valuable supplement in reducing specific markers of obe- sity and is poised to have significant relevance in conditions of heightened blood glucose, such as type 2 diabetes.

1. Introduction Obesity is one of the most pervasive, chronic diseases globally, in both developed and developing countries, contributing to at least 2.8 million deaths annually and significantly impacting the healthcare sys- tem [1]. The growing obesity epidemic is associated with increases in several comorbidities, such as cardiovascular disease, stroke, metabolic syndrome, type 2 diabetes and cancer [2,3]. Current available anti-obe- sity therapeutics are limited and associated with poor efficacy and ad- verse side effects [4,5]. Diet and exercise have been demonstrated to be the most potent in reducing obesity symptomatology [6]. In addition, natural compounds and their derivatives have been proposed as safer anti-obesity alternatives, either as functional foods or nutraceuticals [4]. The gut microbiota has emerged as a key component in the devel- opment of obesity and modulates the host's physiology and metabo- lism, including energy harvest, storage and expenditure [4,7-13]. Pre- clinical and clinical evidence demonstrating the critical role of the gas- trointestinal microbiota on host metabolism is steadily increasing. For example, germ-free mice are protected against obesity and are signifi- cantly leaner than normal control mice despite consuming more calories [14]. In addition, faecal transplantation from obese donors was shown to replicate the obese phenotype in lean germ-free mice independent of diet [15-17]. Moreover, accelerated post-dieting weight regain is asso

H. Schellekens et al. / EBioMedicine xxx (xxxx) 103176


standard holding cages with free access to food and water in the ani- mal care facility of University College Cork. The holding room temper- ature (21 ± 1 °C) and humidity (55 ± 10%) were controlled under a 12 h light/dark cycle (lights on 7.00 AM, lights off 7.00 PM). The mice were fed a low-fat diet (LFD) (10% fat (kcal/100 g), D12450B, Research Diet, USA) or a high-fat diet (HFD) (45% fat (kcal/100 g), D12451, Re- search Diet, USA) for 16 weeks. Food intake was recorded once per week and calculated on the basis of two mice per cage and five cages per group. The data were reported as cumulative food intake per mouse. Bodyweight was monitored weekly for 15 weeks. Experiments were con- ducted in accordance with the European Directive 86/609/EEC and the Recommendation 2007/526/65/EC and were approved by the Animal Experimentation Ethics Committee of University College Cork. 2.2. In vivo probiotic administration Bifidobacterium longum APC1472 was grown anaerobically in De Man, Rogosa and Sharpe (MRS) medium as previously described [37]. The bacterial cell pellet was washed and concentrated in sterile phos- phate buffered saline (PBS) containing 25% Glycerol (v/v) to an end concentration of ~7.5 × 10 9 CFU/mL, aliquoted and stored at − 80 °C. Aliquots were defrosted daily just prior to the start of the dark phase and diluted to ~2 × 10 8 CFU/mL in drinking water for administration to LFD-fed and HFD-fed mice for 16 weeks. Water intake was moni- tored throughout the experiment. Drinking water containing an equiv- alent end concentration of sterile PBS (2% v/v) and glycerol (0.5% v/ v) was administered to control mice. Water was replaced for probiotic/ vehicle-free water every morning. B. longum APC1472 survival in drink- ing water (distilled water) in ambient temperature and oxygen content was tested over 24 h prior to the start of the experiment. Bacteria counts (CFU/mL) did not decrease over 1 log unit for the first 12 h suggesting adequate viability of the strain upon the time of consumption ( Figure S1A) . No significant changes in water intake were observed within the Glucose tolerance was assessed after 15 weeks of treatment as pre- viously described [49], with minor modifications. Briefly, mice were fasted for 7 h during the light phase, with free access to water. Glucose levels were measured in tail vein blood using a glucometer (Bayer, UK) immediately before and 15, 30, 60, 90 and 120 min after intraperitoneal injection of glucose (1 g/kg of body weight in sterile saline). 2.4. Murine tissue sampling Mice were euthanized by decapitation. Trunk blood was collected in tubes containing 25 μ M dipeptidyl peptidase IV (DPP-IV) inhibitor, 2x protease inhibitor cocktail (Roche) (diluted in PBS) and 0.1% Na 2 EDTA for an expected blood volume of 400 µL, centrifuged at 3500 g for 15 min at 4 °C and placed on dry ice until storage at − 80 °C for further analysis. Adipose depots (epididymal, subcutaneous, mesenteric and retroperitoneal) were dissected and weighed. Whole-brains were collected and placed for 8 – 10 s into ice-cold isopentane. All tissues were frozen on dry ice and subsequently stored at − 80 °C for further analysis. 2.5. Murine biochemical analysis same diet groups ( Figure S1B ). 2.3. In vivo glucose tolerance test Plasma insulin and leptin levels were analysed by ELISA using the MILLIPLEX® MAP Mouse Metabolic Hormone Magnetic Bead Panel (Millipore, MMHMAG-44 K) accordingly to the manufacturer's instruc- tions. Plasma ghrelin levels were analysed using the Rat/Mouse Ghrelin

(Total) ELISA Kits (Millipore, EZGRA-88 K). Triglycerides levels were analysed with a Triglyceride Quantification Kit (Abcam Ltd, ab65336) following the to manufacturer's instructions. Corticosterone levels were assayed using ELISA kits (Enzo Life Sciences, ADI-900 – 097) according to the manufacturer's instructions. 2.6. Murine RNA isolation and quantitative real-time PCR Hypothalamus was dissected with a forceps (macropunch) from the frozen brain on dry ice and immediately processed for RNA extraction. Hypothalamus and epididymal adipose tissue total RNA were extracted using the mirVana ™ miRNA Isolation kit (Ambion/Life Technologies, AM1560) and RNeasy® Lipid Tissue Mini Kit (Qiagen, 74,804), re- spectively with DNase treatment using Turbo DNA-free (Ambion/life Technologies, AM1907) according to the manufacturer's recommenda- tions. Equal amounts of RNA were first reverse transcribed to cDNA using High Capacity cDNA Reverse Transcription Kit (Applied Biosys- tems, 4,368,814). Real-time PCR was performed using TaqMan Univer- sal Master Mix II, no Uracil-N glycoslyase (UNG) on a LightCycler®480 System (Roche). Mouse β -actin control mix Probe dye: VIC-MGB (Ap- plied Biosystems, 4352341E) was used as an endogenous control. Tar- get genes were amplified with probes designed by Integrated DNA Tech- nologies (Table S1). Cycle threshold (Ct) values were recorded, normal- ized to its endogenous control and transformed to relative gene expres- sion value using the 2 −ΔΔ Ct method [50]. Each sample was analysed in triplicate for both target gene and endogenous control. The gene expres- sion levels for each animal was calculated considering the mean from each of these triplicates. 3.Methods – human intervention study 3.1. Human intervention study outline This study has a parallel-controlled design. In total, 150 individu- als were screened, after which 124 were randomized into the treatment groups (Placebo: n = 50; Treatment: n = 74). The aim of the first visit of the participant was to assess participants for their eligibility to par- ticipate in the study and explain which procedures would be under- taken. Subjects were given an appointment for the next visit within a 3-week period. At the second visit, all baseline data and biologics were recorded, which was also done after 6 (visit 3) and 12 weeks (visit 4) of placebo or B. longum APC1472 treatment. Vital signs, anthropometric measurements and medical history were recorded. For women of child- bearing age, a urine sample was collected for a pregnancy test. Fasting blood samples (20 mL) were collected to assess glucose, insulin, HbA1c, lipid profiles, satiety/appetite hormone profiles, and inflammatory pro- files. Saliva samples were collected for the assessment of the cortisol awakening response, as well as a stool sample for the microbiota analy- sis and short-chain fatty acid (SCFA) quantification. Questionnaires were administered to assess self-reported perceived stress, anxiety, hunger/ satiety, exercise and diet. Participants were asked to take one capsule per day, providing a daily dose of 1 × 10 10 CFU. Subjects, study facilitators, nurses and re- search analysts were kept blind as to in which group they belonged. The randomisation of treatment schedules was carried out by a com- puter-generated program. The remaining study product was collected to check for compliance following visits 3 and 4 [51]. 3.2. Inclusion and exclusion criteria The inclusion criteria were as follows: subjects had to give written informed consent; had to be between 18 and 65 years of age; had a BMI between 28 and 34.9; had a W/H ratio ≥ 0.88 for males and ≥ 0.83 for females; had to be willing to consume the investigational product daily for the duration of the study. Subjects were excluded if they were


H. Schellekens et al. / EBioMedicine xxx (xxxx) 103176

pregnant, lactating, or female and wish to become a parent during the study; regularly took probiotics; were hypersensitive to any of the com- ponents of the test product; were severely immune-compromised (i.e. HIV positive, transplant patient, antirejection medications, on a steroid for >30 days, or underwent chemotherapy or radiotherapy within the last year); had Type 1 or Type 2 Diabetes Mellitus; had a history of bariatric surgery; had taken anti-obesity medication in the previous 12-weeks; were actively, or has within the last 3 months, participating in a weight loss program or incurred a weight change of more than 3 kg during the past 3 months; had a life-threatening illness; was on Met- formin, anti-psychotic drugs or any medication that the investigator de- termined could impact the results of the study; had commenced use of anti-hypertensive drugs, anti-depressive drugs, statins or any other med- ication that the investigator determined could impact the results of the study within 3-months of randomisation date; had a history of co-ex- isting gastrointestinal, and/or gynaecological, and/or urologic pathol- ogy (e.g. colon cancer, colitis, Crohn's Disease, celiac, Endometriosis, prostate cancer) or lactose intolerance; had a history of drug and/or al- cohol abuse; was currently, or planning, to participate in another study during the study period; had a history of non-compliance; had been on antibiotics in the 12-weeks prior to randomisation; or consumed vitamin D supplements (>5000 IU/d). 17.3% of all screened participants were excluded due to these exclusion criteria. Subjects were removed from the study if they independently elected to withdraw; he/she developed any condition which contravened the original criteria; or was considered at any point to be unsuitable to con- tinue the study, at the discretion of the investigator. 3.3. Study setting and ethical approval The study was conducted in accordance with the ethical principles set forth in the current version of the Declaration of Helsinki (seventh version, October 2013), the International Conference on Harmoniza- tion E6 (R2) Good Clinical Practice (ICH GCP, November 2016) and all applicable local regulatory requirements (i.e. Clinical Research Ethics

Committee of the Cork Teaching Hospitals). This study was registered with (NCT04042181). The CONSORT diagram of this study is depicted in Fig. 1 , the study layout is depicted in Figure S2 . This study was run by Atlantia Food Clinical Trials (Cork, Ireland) (study reference: AFRCO-088). 3.4. Randomisation and blinding The investigational product arrived on site labelled with randomisa- tion number. A randomisation list was generated by an independent sta- tistician. Participants were assigned a randomisation number in chrono- logical order from this randomisation list. The study team, participants and researchers were unaware which randomisation numbers were ac- tive or placebo. Blinding was undone after all data had been analysed. 3.5. Study recruitment Subjects were recruited through the database of Atlantia Food Clini- cal Research Trials, general practitioners ’ offices and by posting adverts in local newspapers. Subjects underwent an initial phone screen. Eligi- ble subjects were scheduled for a screening visit. Subjects received € 300 upon completion of the study to cover costs and expenses incurred. 3.6. Product formulation and dosage Bifidobacterium longum has been granted Qualified Presumption of Safety (QPS) status by the European Food Safety Authority (EFSA). B. longum APC1472 grown culture and the corresponding placebo were freeze-dried (Sacco SRI, Italy) and provided as hydroxypropylmethylcel- lulose (HPMC) capsules in PE bottles (Nutrilinea, Italy). The freeze-dried powder of the strain was blended with standard food-grade excipients to achieve the target dose of 1 × 10 10 CFU, which was based on pre- vious publications [52-54]. The excipients consisted of corn starch,

Fig. 1. Consort diagram. Number of healthy overweight/obese participants that were assessed for eligibility and excluded or allocated to the trial, treated, followed, and analysed.

H. Schellekens et al. / EBioMedicine xxx (xxxx) 103176


magnesium stearate and silicon dioxide. The probiotic formulation con- sisted of B. longum APC1472, whereas the placebo contained maltodex- trin. The product was stored at − 20 °C until distributed to the study par- ticipant and the participant was instructed to keep the product refriger- ated. Participants returned any leftover product at their next visit, and the excess product was counted to check for compliance. 3.7. Collection and analysis of blood samples Fasting blood samples were taken into EDTA tubes, fasting defined as refraining from food overnight (at least 10 h), however drinking water was allowed throughout the duration of the fast. Samples for the analy- sis of active ghrelin were immediately treated with AEBSF (final concen- tration 1 mg/mL, Sigma, A8456), centrifuged and the resulting plasma was treated with HCl (final concentration 0.05 N). Blood samples for the analysis using the U-PLEX assays were treated with DPP-IV inhibitor (fi- nal concentration 1%, Sigma, DPP4) and centrifuged. Blood plasma sam- ples for other analyses did not undergo any additional processing, except for centrifugation. Centrifugation was performed at 1000 g for 10 min at 4 °C, after which samples were aliquoted and either processed or stored at − 80 °C for future analysis. Blood plasma from visit 1 was used to measure urea, creatinine, bilirubin, alanine aminotransferase, alkaline phosphatase, aspartate aminotransferase, gamma-glutamyl transferase, total protein, albumin, globulin, calcium, magnesium, phosphate, uric acid, cholesterol, HDL cholesterol, LDL cholesterol, total triglycerides, glucose, full blood count + 5-part diff. Safety blood, haematology and biochemistry para- meters were analysed by Biomnis-Eurofins Ireland. Blood from visits 2, 3 and 4 was used to measure total cholesterol, LDL, HDL, triglycerides HbA1c, glucose and insulin by Biomnis-Eurofins Ireland. Furthermore, blood plasma was assessed for active ghrelin lev- els using an ELISA (EMD Millipore, EZGRA-88BK) which was performed according to the manufacturer's instructions. Plates were read at 405 nm with a correction at 590 nm using the synergy HT plate reader (Biotek instruments). Blood plasma was also assessed for metabolic and inflam- matory biomarkers using custom U-PLEX assays (MSD, K151ACM-2), which were also performed according to the manufacturer's instructions. Blood plasma samples were diluted 1:3 for the U-PLEX assays. U-PLEX markers were linked as following; Plate 1: 1) Leptin, 2) PYY, 3) GLP-1 – total, 4) IFN γ , 5) Il-4, 7) TNF- α , 8) Il-10, 9) C-peptide, 10) Ghrelin – total; Plate 2: 1) GLP-1 – active. The working solution was supple- mented with DPP-IV inhibitor (final concentration 1%, Sigma, DPP4). Plates were read using the MESO QuickPlex SQ 120. Duplicates with ≥ 20% coefficient of variability were re-analysed. Samples did not undergo any additional freeze-thaw cycles. 3.8. Collection and analysis of cortisol awakening response samples To monitor the cortisol awakening response, saliva from visits 2 and 4 was collected in Salivette devices (Sarstedt, 51.1534.500) im- mediately upon awakening, and after 30, 45 and 60 min. Participants were instructed to keep samples in the fridge until delivery at the visit time, after which they were centrifuged at 1500 g for 5 min, the saliva was harvested and immediately stored at − 80 °C. Salivary cor- tisol concentrations were quantified using ELISA kits (Enzo life sci- ences, ADI-901 – 071), which were performed according to the manufac- turer's instructions. Saliva samples were diluted 1:2. Plates were read at 405 nm with a correction at 580 nm using the synergy HT plate reader (Biotek instruments). Duplicates with ≥ 20% coefficient of variability were re-analysed. Samples did not undergo any additional freeze-thaw cycles. Cortisol awakening response was calculated using the area under the curve increase (AUCi). Briefly, data from the 30-, 45- and 60-minute time-points were normalized (delta) to the samples taken immediately

upon awakening, after which the sum was taken of the 30-, 45- and 60-minute time-points [52].

4.Methods – murine and human microbiota 4.1. Murine and human microbiota sequencing

Murine caecal DNA was isolated using the QIAamp Fast DNA Stool Mini kit (Qiagen) as previously described and kept at − 20 °C until further analysis [53]. Isolated DNA was quantified on a NanoDrop ND2000 spectrophotometer (Thermo Scientific, DE) and used for 16S ribosomal RNA sequencing by Illumina MiSeq System (Illumina Inc., USA) according to the manufacturer's instructions. Briefly, PCR ampli- cons (primers for V3-V4 hypervariable region of the 16S rRNA gene: F (5 ′ -TCGTCGGCAGCGTCAGATGTGTATAAGAGAC AGCCTACGGGNG- GCWGCAG-3 ′ ) and R (5 ′ -GTCTCGTGGGCTCGGAGATGTGTATA AGA- GACAGGACTACH VGGGTATCTAATCC-3 ′ ) were purified and libraries prepared as previously described [53]. Briefly, the 16S V3-V4 ampli- cons were generated using Kapa HiFi HS ready mix and purified using the Agencourt AMPure XP system (Beckman Coulter Genomics, Take- ley, UK). The Nextera XT Index Kit (Illumina Inc., USA) was used to barcode each sample. PCR products were cleaned using AMPure XP beads and a magnetic 96-well plate. Final barcoded amplicons were measured using the Qubit dsDNA High Sensitivity assay kit on the Qubit 3.0 fluorometer, diluted to 5 ng/µL and pooled. The PCR products from both PCR steps (Amplicon & Indexing) were visualised in agarose gels stained with SYBR Safe DNA gel stain (Invitrogen). Samples were se- quenced at Clinical-Microbiomics, Denmark on the Illumina MiSeq plat- form using a 2 × 300 bp kit. After sequencing, reads were assembled, processed and analysed as previously described [53]. In the microbiota composition analysis, LDA Effect Size (LEfSe: Linear Discriminant Analy- sis Effect Size) was used as an algorithm with default settings on the interface Galaxy ( [54] to identify taxa with differentiating abundances. The differentially abun- dant features are ranked by effect size after undergoing linear discrim- inant analysis (LDA), using an effect size threshold of 2 (log10 scale). In non-technical terms, LEfSe pre-selects features that are different be- tween groups and then tries to fit a model to see how well these features explain the groups. The score is an average between the effect size and how well the model fits, after which they are transformed to a value be- tween − 6 and 6. Principal coordinates Analysis (PCoA) was performed based on Bray-Curtis beta diversity distances using the Adonis function in the “ vegan ” (2.4 – 3) package for R (version 3.3.1). For the human intervention study, faecal sample collection and DNA extraction was performed as previously described (see supplementary material for details) [55]. The DNA samples were processed according to the Illumina 16S Metagenomic Sequencing Library Preparation in- structions as described above for the murine DNA samples. Final bar- coded amplicons were measured using the Qubit dsDNA High Sensitiv- ity assay kit on the Qubit 3.0 fluorometer, diluted to 8.3 ng/µL, pooled and sent for sequencing. Microbiome analysis was carried out in R (ver- sion 3.6.1) with Rstudio (version 1.2.1335). DADA2 was used to denoise and call amplicon sequence variants (ASVs). Taxonomy was assigned using the SILVA SSUREf database version 132. ASVs unknown on a genus level were excluded, as well as ASVs present in two or fewer sam- ples. The ALDEx2 library used to compute the centred log-ratio trans- formed values of the remaining taxa [56]. For principal components analysis (PCA), a pairwise implementation of the adonis() PERMANOVA function in the vegan library followed by the Bonferroni-Holm correc- tion was used to test for difference in β -diversity in terms of Aitchi- son distance (source: Oksanen, Jari, et al. "Package ‘ vegan ’ ." Commu- nity ecology package, version 2.9 (2013): 1 – 295). Differential abun- dance was assessed using a pairwise implementation of the aldex.test() function, followed by Benjamini-Hochberg correction. In all


H. Schellekens et al. / EBioMedicine xxx (xxxx) 103176

cases, a q-value < 0.1 was considered significant. α -diversity was com- puted using the iNEXT library [57]. 4.2. Faecal SCFA quantification Faecal samples were homogenised with acidified water (HCl pH 3) at a ratio of 1:7.5 w/v and analysed by gas chromatography flame ion- isation detection (GC-FID) using a Varian 3800 GC system, fitted with an Agilent DB-FFAP column (30 mL x 0.32 mm ID x 0.25 μ m df; Agi- lent) and a flame ionisation detector with a CP-8400 auto-sampler. He- lium was employed as the carrier gas at an initial flow rate of 1.3 mL/ min. The initial oven temperature was 50 °C, was maintained for 30 s, raised to 140 °C at 10 °C/min and held for 30 s, before being increased to 240 °C at 20 °C/min, and held for 5 min (total run time 20 min). The temperatures of the detector and the injection port were 300 °C and 240 °C, respectively. A split-less injection of 0.2 µL was carried out for each sample or standard using a 10 µL syringe (Agilent) installed to a CP-8400 auto-sampler (Varian). A 5 m guard column was installed be- tween the injector and analytical column (Restek). Peak integration was performed using Varian Star Chromatography Workstation version 6.0 software. Vials containing 1800 µL of water were run between each sam- ple duplicates as blanks to control for any potential carryover. Standards were included in each run to maintain the calibration. For further details on sample and standards preparation see supplementary information. 4.3. Questionnaires Using self-report scales, participants were assessed for perceived stress using Cohen's Perceived Stress Scale and anxiety and depression using the Hospital Anxiety and Depression Scale (HADS) at baseline, af- ter 6 and after 12 weeks, as previously described[58], [59]. In addi- tion, satiety/hunger was determined using a visual analogue Hunger/ Satiety scale, physical activity using the International Physical Activity Questionnaire (IPAQ) [60]. Nutrient intake was assessed using a Food Frequency Questionnaire (FFQ), as previously described [61]. 4.4. Statistical analysis Preclinical data were assessed for normality using the Shapiro-Wilk test. Normally distributed data were analysed using a two-way ANOVA, followed by Fisher's least significant difference (LSD) post hoc test. Non-parametric datasets were analysed using the Kruskal-Wallis test,

followed by the Mann-Whitney U test with Bonferroni adjustment of p-values. Body weight changes and glucose levels in glucose toler- ance test were analysed with a two way repeated-measures ANOVA (with Diet and Probiotic as two independent factors and Time as a re- peated-measured factor), followed by LSD post hoc test at each time point. Statistical analysis was performed using SPSS software (IBM SPSS statistics 22). Preclinical data are represented as mean ± SEM. For the human intervention study, differences between the treatment and placebo groups at the last visit (i.e. visit 4) were analysed using an analysis of covariance (ANCOVA), correcting for baseline variance (i.e. visit 2) and sex. Comparisons between baseline measurements (visit 2) and post-intervention measurements (visit 4) were analysed using an unpaired student's T-test. Analyses were performed on the intention to treat populations. Statistical analysis was performed using SPSS soft- ware version 26 (IBM Corp). Data in table are presented as mean ± SEM or 95% CI. P-Values <0.05 were considered statistically signifi- cant. For significant associations, a Benjamini-Hochberg procedure was performed with a threshold of q <0.1. Partial eta-squared ( η 2 )was used to estimate effect size [62]. Effect sizes were interpreted as following: η 2 ≤ 0.06 was considered small, 0.06 > η 2 ≤ 0.14 was considered mod- erate, η 2 ≥ 0.14 was considered large. 5. Results 5.1. B. longum APC1472 decreases body weight gain and fat depots accumulation in obese mice B. longum APC1472 decreased body weight gain after 15 weeks of administration ( F (1, 33) = 4.751, p =0.037) ( Fig. 2A, 2B ). HFD feed- ing increased caloric intake (F (1, 15) = 9.229, p =0.008) ( Figure S1C ), body weight (F (1, 33) = 29.715, p < 0.001) ( Fig. 2A ) and fat depot accumulation (mesenteric (F (1, 33) = 61.328, p < 0.001), retroperitoneal (F (1, 32) = 128.409, p < 0.001), subcutaneous ( F (1, 31) = 124.091, p < 0.001) and epididymal ( F (1, 33) = 81.673, p < 0.001)) ( Fig. 2C, D, E, F ). Pairwise comparisons showed a significant decreased body weight effect of B. longum APC1472 in HFD-fed mice ( p= 0.047) ( Fig. 2B ), which was independent of caloric intake ( Fig- ure S1 ). Furthermore, the administration of B. longum APC1472 signif- icantly reduced fat depot accumulation (mesenteric ( F (1, 33) = 5.908, p = 0.021), and subcutaneous ( F (1, 33) = 4.270, p =0.047)) ( Fig. 2C, D, E, F ). Finally, pairwaise comparisons revealed a significant de- creased fat depot accumation effect of B. longum APC1472 administra- tion in HFD-fed mice (mesenteric p = 0.002, retroperitoneal p =0.05 and subcutaneous p =0.023).

Fig. 2. Effects of Bifidobacterium longum APC1472 on body weight and fat depots accumulation in mice. (A) Weekly body weight gain, (B) total body weight gain and (C) mesenteric, (D) retroperitoneal, (E) subcutaneous and (F) epididymal fat depots accumulation (% of total body weight) in control mice treated with drinking water containing sterile PBS (2% v/v) and glycerol (0.5% v/v) and fed a control low-fat diet (LFD) ( n = 10) or a high-fat diet (HFD) ( n = 9) and in mice treated with B. longum APC1472 in drinking water (2 × 10 8 CFU/mL) and fed a LFD ( n =9inA,B,C,EandF; n =8inD)oraHFD( n =9inA,B,C,D,andF; n = 8 in E) for 15 (A, and B) or 16 weeks (C, D, E and F). Data are shown as mean ± SEM.. Data are significant different ( p <0.05) accordingly to Repeated Measures ANOVA (A) or two-way ANOVA followed by LSD post-hoc test (B, C, D, E and F). * indicates significant diet treatment effect (* p <0.05, ** p <0.01, *** p <0.001) and # indicates significant B. longum APC1472 treatment effect ( # p <0.05, ## p <0.01).

H. Schellekens et al. / EBioMedicine xxx (xxxx) 103176


5.2. B. longum APC1472 administration improves glucose tolerance, circulating levels of leptin and corticosterone in obese mice Effects of HFD feeding ( F (5, 155) = 3.321, p = 0.007) and B. longum APC1472 supplementation ( F (5, 155) = 4.792, p < 0.001) were observed, as well as an interaction effect between these two fac- tors and time ( F (5, 155) = 3.307, p = 0.007) in the glucose toler- ance test. Supplementation with B. longum APC1472 normalized glucose levels after 15 mins of glucose administration in HFD-fed obese mice ( p = 0.006) and significantly decreased glucose after 90 ( p =0.019) and 120 min ( p = 0.018) respectively ( Fig. 3A ) as determined by 2 way ANOVAs at each individual timepoint. Moreover, HFD feeding ( F (1, 33) = 29.761, p <0.001), B. longum APC1472 ( F (1, 33) = 4.425, p = 0.043) and interaction effects between these two factors ( F (1, 33) = 5.337, p = 0.027) were also observed when analysing the area under the curve (AUC) for glucose levels ( Fig. 3B ), with B. longum APC1472 administration significantly reducing AUC in HFD-fed mice ( p = 0.003) as determined by post-hoc comparison ( Fig. 3B ). In ad- dition, both a HFD feeding (F (1, 33) = 9.167, p = 0.005) and a B. longum APC1472 effect ( F (1, 33) = 4.796, p = 0.036) were observed for non-fasting insulin levels ( Fig. 3C) . Interestingly, B. longum APC1472 reduced non-fasting insulin levels in LFD-fed mice ( p =0.054) but not in HFD-fed mice ( Fig. 3C) . However, for fasting glucose lev- els, only a HFD feeding effect was observed (F (1, 32) = 29.153, p < 0.001) ( Fig. 3D ). Moreover, both a HFD feeding (F (1, 31) = 30.926, p < 0.001) and a B. longum APC1472 effect ( F (1,

31) = 17.917, p < 0.001) were observed for epididymal insulin recep- tor substrate 1 ( IRS-1 ) expression ( Fig. 3E ). Post-hoc comparisons de- termined that B. longum APC1472 significantly reduced IRS -1 expres- sion in both LFD ( p = 0.002) and HFD-fed mice ( p =0.011) ( Fig. 3E ). Both a HFD feeding ( F (1, 33) = 38.023, p < 0.001) and a B. longum APC1472 ( F (1, 33) = 5.340, p = 0.027) effect as well as an interac- tion effect (F (1, 33) = 4.237, p = 0.048) were observed for fasting leptin levels ( Fig. 3F) . The effect of HFD on leptin levels was attenu- ated by B. longum APC1472 treatment ( p = 0.004). Finally, we found a significant B. longum APC1472 treatment effect ( F (1, 32) = 7.774, p = 0.009) for plasma corticosterone levels ( Fig. 3G ). Administration of B. longum APC1472 significantly decreased plasma corticosterone lev- els in HFD-fed mice ( p =0.011) ( Fig. 3G ), which may have contributed to its overall impact on glucose homeostasis [98]. 5.3. B. longum APC1472 induces changes of hypothalamic neuropeptide expression in mice Analysis of the gene expression levels of hypothalamic neuropep- tides involved in appetite modulation revealed a significant HFD effect on the gene expression of the orexigenic peptide agouti-related protein ( AgRP ) ( F (1, 33) = 10.412, p = 0.003) but a non-significant reduction in neuropeptide Y ( NPY ) expression ( Figure S3 ). Interestingly, both a B. longum APC1472 effect ( F (1, 33) = 7.820, p = 0.009) and an inter- action effect ( F (1, 33) = 5.881, p = 0.021) were observed for cocaine- and amphetamine-regulated transcript ( CART ) expression ( Figure S3 ). Indeed, B. longum APC1472 administration significantly reduced CART

Fig. 3. Bifidobacterium longum APC1472 improved glucose tolerance, leptin plasma levels and stress-induced corticosterone circulating levels in high-fat diet-induced obesity in mice. (A and B) Glucose tolerance test (GTT) glucose curve and area under the curve (AUC) after 1 g/kg glucose challenge, (C and D) non-fasting and fasting insulin plasma levels, (E) fasting leptin plasma levels, (F) epididymal fat insulin receptor substrate (IRS) − 1 mRNA expression and (G) fasting-induced corticosterone plasma in control mice treated with drinking water containing sterile PBS (2% v/v) and glycerol (0.5% v/v) and fed a control low-fat diet (LFD) ( n = 10 in A, B, C, E, F and G) or a high-fat diet (HFD) ( n =9inA,B,C,D,EandG; n =8 in F) and in mice treated with B. longum APC1472 in drinking water (2 × 10 8 CFU/mL) and fed a LFD ( n =9inA,B,C,D,EandF; n =8inG)oraHFD( n =9inA,B,C,D,EandF; n =8 in G) for 15 (A, B,C) or 16 weeks (D, E, F and G). Data are shown as mean ± SEM. Data are significant different ( p <0.05) accordingly to Repeated Measures ANOVA (A) or two-way ANOVA followed by LSD post-hoc test (B, C, D, E, F and G). * indicates significant diet treatment effect (* p <0.05, ** p <0.01, *** p <0.001) and # indicates significant B. longum APC1472 treatment effect ( # p <0.05, ## p <0.01).


H. Schellekens et al. / EBioMedicine xxx (xxxx) 103176

expression in HFD-fed mice ( p = 0.001). While a reduced expression was observed for the anorexigenic pro-opiomelanocortin ( POMC ) gene expression in HFD-fed animals treated with B. longum APC1472 com- pared to HFD-fed, this did not reach statistical significance. Finally, no significant change in leptin ( LEP-R ) nor ghrelin ( GHS-R1a ) receptor ex- pression was observed ( Figure S3 ). 5.4. Human intervention study population In the human study, no significant differences were observed in weight, BMI, W/H ratio, age, height, sex, ethnicity, mode of delivery, alcohol consumption, and medical/surgical history at baseline between B. longum APC1472 treatment and placebo groups, as well as compli- ance ( Table 1 ). We did observe an increased prevalence of concomi- tant medical or nutritional supplement consumption in the treatment group (48.6%) compared to the placebo group (33.3%). In addition, we also observed differences in the socioeconomic profile where there was a lower prevalence of employers and managers in the treatment group (2/74) compared to the placebo group (4/48). Similarly, we observed a lower prevalence of past smokers in the treatment group (28/74) com- pared to the placebo group (9/48). In conclusion, the baseline charac- teristics of our placebo group and B. longum APC1472 group are mostly the same.

Physical activity and food intake patterns were also assessed throughout the study using self-report questionnaires ( Table S2, S3 ). No differences in physical activity levels or calorie, macro- and micronutri- ent intake were observed over the 12-week treatment period or between the placebo and B. longum APC1472 group. 5.5. Adverse events There were seven adverse events (6 placebo participants and 1 treat- ment participant) that were possibly related to the investigational prod- uct. The adverse event of the treatment participant was constipation. The remaining 6 adverse events for placebo participants were; gastroin- testinal discomfort and increased appetite; bloating; increased flatu- lence; aches in joints and increased temperature; rash on knees, elbows, scalp and red blotches on chest & upper arm. 5.6. B. longum APC1472 does not affect BMI and W/H ratio in humans The primary outcome of this study was to investigate whether B. longum APC1472 supplementation could alter BMI, and a secondary out- come of change in W/H ratio was included to support the primary outcome. However, no differences were observed in BMI and W/H ra- tio over the 12-week treatment period, or between the placebo and B. longum APC1472 treatment groups ( Fig. 4 ). 5.7. B. longum APC1472 improves fasting glucose levels independent of other blood markers of energy metabolism and satiety in humans We subsequently measured markers associated with host energy me- tabolism and satiety as part of the secondary and exploratory outcome measures ( Fig. 5 and Table S4 for full statistical results). Here we ob- served that both the B. longum APC1472 and the placebo arm reduced fasting glucose levels over the 12-week treatment period ( Fig. 5A ). However, glucose levels were 0.266 mmol/L (95% CI [ − 0.44, − 0.09]) lower in the B. longum APC1472 group compared with the placebo group ( F (1112) = 9.073, p =0.003; q =0.075) ( Fig. 5B ). The effect size of the B. longum APC1472-induced decrease was moderate ( η 2 =0.075). We also observed that HbA1c levels decreased over the 12-week treat- ment period in both the placebo group ( t (62.372) = 4.277, p <0.001) and B. longum APC1472 treatment group ( t (85.983) = 5.787, p < 0.001) ( Fig. 5C ). However, there were no differences between the groups, indicating that the decrease in HbA1c levels is most likely ex- plained by the 12-week treatment period or placebo effect. No changes were observed in other biomarkers of host energy metabolism such as insulin, C-peptide, ghrelin (active and total), GLP-1 (active and total), PYY and leptin levels ( Fig. 5E-T ). 5.8. B. longum APC1472 does not influence human lipid and inflammatory profiles in humans It is well-known that obesity is associated with metabolic syndrome, hypertension and hyperlipidaemia [63]. B. longum APC1472 did not im- pact lipid profiles (i.e. cholesterol, triglycerides and LDL), and inflam- matory profiles (i.e. IL-10, TNF- α and IFN γ ) compared to the placebo group ( Table 2 ). In addition, vital signs remained unaltered through- out the study ( Table S5 ). These results reveal that B. longum APC1472 did not evoke any negative effects on vital signs or induced any in- flammation. Interestingly, even though no significant changes were ob- served in HDL levels over the 12-week treatment period, a small in- crease in HDL levels was observed in the B. longum APC1472 group ( F (1117) = 3.260, p = 0.074). The effect-size of the increase in HDL levels was small ( η 2 =0.027).

Table 1 Baseline characteristics of subjects in the placebo and treatment arms at visit 1 (screening visit).

Placebo ( n = 48, mean ± STD)

B. longum APC1472 ( n =74, mean± STD)


Weight (kg) W/H ratio Age (years) Height (m) BMI

87.9±1.7 31.2±0.3 0.95±0.01 46.3±9.9 1.67±0.10 19 (39.6%) 29 (60.4%) 48 (100%) 0 (0%) 15 (31.3%) 14 (29.2%) 4 (8.3%) 4 (8.3%) 4 (8.3%) 3 (6.3%)

89.0±1.3 30.8±0.2 0.96±0.01 44.9±11.4 1.70±0.09 34 (45.9%) 40 (54.1%) 73 (98.6%) 1 (1.4%) 21 (28.4%) 19 (25.7%) 8 (10.8%) 8 (10.8%) 2 (2.7%) 7 (9.5%) 5 (6.8%) 2 (2.7%)

Sex (no. of subject (%)) Male


Race or ethnicity (no. of subject (%)) Caucasian


Socioeconomic status (no. of subject (%)) Non-manual

Lower Professional Manual skilled Semi-skilled Employers and managers Own account workers Higher Professional

3 (6.3%) 1 (2.1%)

All others gainfully occupied and unknown

Farmer Unskilled

0 (0%) 0 (0%)

1 (1.4%) 1 (1.4%)

Smoking status (no. of subject (%)) Non-smoker

22 (45.8%) 17 (35.4%) 9 (18.8%)

40 (54.1%) 28 (37.8%) 6 (8.1%)

Past smoker Current smoker

Alcohol consumption (mean ± SEM) Units per week

4.97±0.68 4.31±0.46 Currently on concomitant medical or nutritional supplements (no. of subject (%)) Yes 16 (33.3%) 36 (48.6%) No 32 (66.7%) 38 (51.4%) Compliance (% product consumed) Week 6 95.8±1.2 97.9±0.8 Week 12 94.0±2.0 97.2±1.2 Abbreviations: BMI = Body-mass index; W/H ratio = waist-to-hip ratio.

H. Schellekens et al. / EBioMedicine xxx (xxxx) 103176


Fig. 4. B. longum APC1472 supplementation does not impact BMI and W/H ratio in overweight and obese individuals. Body mass index (BMI) (A, B) and waist-to-hip ratio (W/H ratio) (C, D) were measured as the beginning of the study (pre), after 6 weeks (mid) and after 12 weeks (post) of treatment. All BMI and W/H ratio data are depicted of all 3 timepoints (A, C), as well as the change after 12 weeks compared to at the beginning of the study (B, D). Data are depicted as boxplot or scatter dot plot, where the dots depict individual datapoints, with

n = 48 for the placebo group and n =74 for the B. longum APC1472 treatment group. 5.9. B. longum APC1472 does not affect satiety, mood, perceived stress and cortisol awakening response in humans Considering that the gut microbiota has been implicated in the mod- ulation of host mood and food intake behaviour [10,64], we investi- gated whether B. longum APC1472 could improve levels of the stress hormone cortisol upon waking (i.e. cortisol awakening response), or self-reported measures of satiety, and self-reported measures of mood (i.e. perceived stress, anxiety and depression) ( Table 3 ). B. longum APC1472 did not impact cortisol awakening response, or self-reported satiety, perceived stress, anxiety and depression measures. 5.10. B. longum APC1472 improves fasting glucose levels, active ghrelin and cortisol awakening response in obese individuals Participants in this study were either overweight ( n =40; 28 ≥ BMI < 30) or obese ( n =82; 30 ≥ BMI < 35). It is possible that B. longum APC1472 may evoke a stronger effect in obese individuals as they have a stronger phenotype compared to overweight individuals. As such, we investigated whether any of the anthropomorphic measures, blood bio- markers and measures of mood were affected by B. longum APC1472 in the obese subpopulation only, compared to placebo ( Fig. 6 and Table S7 – 11 for population characteristics and full statistical results). Similar to the analysis on the entire study population, B. longum APC1472 and placebo reduced fasting glucose levels over the 12-week treatment pe- riod ( Fig. 6A ). However, glucose levels were 0.295 mmol/L (95% CI [ − 0.5, − 0.1]) lower in the B. longum APC1472 group compared to the

placebo group ( F (1,75) = 7.566, p = 0.007), in obese individuals, with a moderate effect size ( η 2 =0.092) ( Fig. 6B ). Furthermore, B. longum APC1472 increased active ghrelin levels ( F (1,74) = 4.903, p =0.030), with a moderate effect size ( η 2 = 0.062). Moreover, B. longum APC1472 also reduced cortisol awakening response ( F (1,51) = 4.415, p = 0.041), with a moderate effect size ( η 2 = 0.080), in the obese sub- population analysis. Overall, these results show beneficial effects of B. longum APC1472 on fasting plasma glucose levels, active ghrelin levels and cortisol awak- ening response in obese individuals. It is also important to note that the effect size in the obese subpopulation ( η 2 = 0.092) was bigger than the effect size in the overall study population ( η 2 = 0.075). This indicates that B. longum APC1472 has a more robust beneficial effect on fasting glucose levels in obese, rather than in overweight, individuals. 5.11. B. longum APC1472 does not induce major rearrangements on the microbiota composition but increases the abundance of Bifidobacterium We subsequently investigated whether the observed changes induced by the B. longum APC1472 strain were mediated in part through mod- ulation of the gut microbiota. Investigations into the caecal micro- biota in the preclinical experiment revealed that there was a signif- icant dissimilarity in beta diversity between LFD- and HFD-fed mice ( p < 0.01) ( Figure S5A ), with a decreased relative abundances of Bacteroidetes phylum and increased relative abundances of Firmicutes class Clostridia , respectively ( Figure S5B ), which is in line with previ- ous studies [65,66]. Different phylotypes were responsible for the cae- cal microbiota differences amongst the treatment groups ( Figure S5C ), showing increments on different Firmicutes members in HFD-fed mice treated with B. longum APC1472. Moreover, B. longum APC1472 par

Page 1 Page 2 Page 3 Page 4 Page 5 Page 6 Page 7 Page 8 Page 9 Page 10 Page 11 Page 12 Page 13 Page 14 Page 15 Page 16

Powered by