Fucose modifies short chain fatty acid and H2S formation through alterations of microbial cross-feeding activities

Abstract Algae are a rich but unexplored source of fibers with the potential to contribute to the next generation of prebiotics. The sulfated brown algae polysaccharide, fucoidan, is mainly composed of the deoxy-hexose L-fucose, which can be metabolized to 1,2-propanediol (1,2-PD) or lactate by gut microbes as precursors of propionate and butyrate. It was the aim of this study to investigate the impact of fucoidan on the fermentation capacity of the fecal microbiota and to compare to fucose. In batch fermentations of fecal microbiota collected from 17 donor samples, fucose promoted the production of propionate while no consistent effect was observed for commercial fucoidan and Fucus vesiculosus extract prepared in this study containing laminarin and fucoidan. H2S production was detected under all tested conditions, and levels were significantly lower in the presence of fucose in a dose-dependent manner. The addition of high fucose levels led to higher relative abundance of microbial 1,2-PD and lactate cross-feeders. Our results highlight that fucose and not fucoidan addition impacted fermentation capacity and increased the proportions of propionate and butyrate, which allows for precise modulation of intestinal microbiota activity.


Introduction
In comparison to societies from the Amazon, Malawi, or P a pua Ne w Guinea, gut micr obiomes fr om Western cultur es possess a lo w er abundance of fibre-fermenting bacterial species, which is associated with shifts in fermentation activity and the production of detrimental metabolites (Yatsunenko et al. 2012, Martinez et al. 2015. The observed link between health, diet, dietary fibre, and gut microbiome has led to a renaissance of prebiotics to addr ess a ga p in nutrition caused by contemporary dietary habits. The International Scientific Association for Probiotics and Prebiotics (ISAPP) defines a prebiotic as "a substrate that is selectively utilized by host micr oor ganisms conferring a health benefit" (Gibson et al. 2017 ). Most dietary prebiotics are non-digestible, soluble carbohydrates that reach the colon without being hydrolyzed by pancreatic and intestinal enzymes (Gibson et al. 2017 ). The colon has the highest microbial cell density in the gastrointestinal tract (up to 10 11 cells per gram of gut content, Derrien andVan Hylckama Vlieg 2015 , Vandeputte et al. 2017 ) with Bacillota (former Firmicutes ), Bacteroidota (former Bacteroidetes ), Pseudomonadota (former Proteabacteria ), Actinomycetota (former Actinobacteria ) and V errucomicrobiota (former V errucomicrobia ) as the most abundant phyla. Colonic microbes use the non-digestible carbohydrates as substrate for fermentation processes.
The pr obabl y most studied pr ebiotics ar e inulin, fructo-and galactooligosaccharides (FOS and GOS) together with resistant starc h. These fibr es ar e isolated fr om terr estrial plants or pr oduced biotec hnologicall y, and ar e c har acterized by simple composition and re petiti ve structures. FOS, GOS and resistant starch are composed of few monosaccharides, namely fructose, galactose , and glucose , whic h ar e connected by a limited number of gl ycosidic linka ge types. Curr ent concepts r elated to pr ecise microbiome engineering suggest the use of structur all y and compositionall y differ ent carbohydr ate pol ymers as nov el pr ebiotics (Deehan et al. 2020(Deehan et al. , 2022. The marine ecosystem, for example algae, is a rich, underexplored natural resource of fibres that differ in composition and structur e fr om their terrestrial counterparts (Gotteland et al. 2020 ). Brown algae contain the pol ysacc harides alginate (mannuronic and guluronic monomers), laminarin ( β-linked glucose monomers) and fucoidan, which is a unique natur al pol ymer mainl y composed of fucose in L-configuration and high degree of sulfation (SO 4 2 − , Gotteland et al. 2020 ). Fucose can be metabolized to 1,2-PD as a precursor of propionate (Bunesova et al. 2016, Schwab et al. 2017, and to lactate (Becerra et al. 2015 ). Gl ycosulfatases catal yse the r elease of sulfate (Corfield et al. 1992 ), which can be reduced stepwise to sulfite (SO 3 2 − ) and hydrogen sulfide (H 2 S) by sulfate reducing bacteria (SRB) like Desulfovibrionaceae , which harbour the enzyme dissimilatory sulfite reductase (DSR) enzyme (Christophersen et al. 2011 ). While the potential of fucoidan to act as a source of propionate can be beneficial within the gut ecosystem, there is the risk of excessive formation of H 2 S from sulfate. High levels of colonic H 2 S have been linked to the inhibition of the mitoc hondrial r espir atory c hain, lo w er m ucosal integrity thr ough genotoxicity reduction of mucosal disulfide bonds and inhibition of colonocyte butyrate oxidation through cytochrome-c inhibition (Blachier et al. 2021 ). T herefore , it is of high importance to determine whether microbial gut fermentation of fucoidan pr omotes H 2 S pr oduction. T he o v er all objectiv e of this study was to systematicall y inv estigate the fermentation of fucoidan pol ysacc harides fr om marine biomass by fecal micr obiota to establish the potential of fucoidan as next generation prebiotic. Two fucoidan pol ysacc harides wer e examined in fecal microbiota batch fermentations and compared to L-fucose: a commer-cial fucoidan and an extract from Fucus vesiculosus obtained in this study.

Experimental set-up
We first extracted and characterized the major polysaccharides from F. vesiculosus ( Suppl. Fig. S1 ). Next, two in vitro batch fermentation experiments were conducted that used MacFarlane as base medium (Fig. 2 ). In experiment 1, we investigated the impact of the addition of 0.4 g L −1 fucose (FUS0.4) or commercially available fucoidan (FUC, Merck), or our extract from Fucus vesiculosus (EXT). Batch fermentations inoculated with slurries pr epar ed fr om nine donors wer e compar ed to contr ols (CON) using basic MacFarlane medium. In experiment 2, fucose was added at 0.8 (FUS0.8) and 1.6 (FUS1.6) g L −1 , whereas commercial fucoidan was supplied at 0.4 g L −1 (FUC) (Fig. 2 ). We determined the concentrations of SCFA at 0, 24, and 48 h of incubation and anal ysed H 2 S le v els at 0 and 48 h. Microbial composition and the abundance of selected bacterial groups was determined with 16S rRNA gene sequencing and qPCR using fecal samples and biomass from CON and FUS1.6 fermentation collected in experiment 2.

Extraction of fucoidan from F. vesiculosus
F. vesiculosus was chosen for extraction as this species was the source of the commercial fucoidan (Merc k, Denmark). Fr esh F. vesiculosus was collected in September 2021 at the Aarhus Bay coast in the proximity of Aarhus (Denmark). T he sea w eed w as washed with distilled water to r emov e sand and impurities and lyophilized (Christ, Gamma, 1-16, LSC). T he dried sea w eed w as ground into powder (0.5 mm mesh, Retsch, ZM200) and was vacuum packed and stored at −20 • C until further processing. Seaweed pol ysacc harides wer e extr acted as specified by Ptak et al. ( 2019 ) with modifications ( Suppl. Fig. S1 ). Briefly, seaweed powder (60 g) was resuspended in 600 ml EtOH (95% v/v) and was stirred for 4 h at room temperature . T he suspension was cen- trifuged, 4600 r/m, 30 min, 20 • C (Heraeus, MULTIFUGE 3 S-R). The pellet w as w ashed with 100 ml acetone 99.5 (v/v%), centrifuged and dried in an oven (Memmert, UF55) at 35 • C o vernight. T he dried material was vacuum packed and stored at −20 • C before extr action. Befor e extr action, dried pellets (1.5 g) were mixed with 45 ml 0.1 M HCl, and micro w ave extracted (Anton Paar Multiwave 3000) with a po w er-gr adient: r amp fr om 0 to 300 W for 5 min, hold at 300 W for 15 min (fan 1), follo w ed b y 0 W for 5 min (fan 3). The maxim um temper atur e was set to 90 • C, if 90 • C was r eac hed, power turned off automatically. After micro w ave assisted extraction, all suspensions were mixed and centrifuged at 4600 r/m and 20 • C for 30 min. The pellet w as discar ded and the supernatant was k e pt at 4 • C until purification. To precipitate alginate, CaCl 2 was added to a final concentration of 2 (% w/v) and the mixture was stored at 4 • C o vernight. T he mixture was centrifuged at 4600 r/m 20 • C for 30 min and the pellet was discar ded. The suspension w as dial yzed in dial ysis ba gs (Spectra/Por 3, MWCO: 3500 Da) for 48 h in Mili Q water with water changes every 12 h. To precipitate polymers, EtOH was added to a final concentration of (72% v/v) and precipitation occurred overnight at 4 • C. Fucoidan extract was harvested by centrifugation at 4600 r/m and 20 • C for 30 min. The pellet was freeze-dried (Christ, Gamma, 1-16, LSC) and the Fucus extract (EXT) was stored at −20 • C.

Determination of polymer monosaccharide composition
To analyse the monosaccharide composition of fucoidan, Fucus extract, samples (1 mg) were hydrolyzed with 450 μL (2.5 M) trifluoroacetic acid at 121 • C for 2 h (Memmert, UF55) and dried with a N-EVAP 112 nitrogen evaporator (Organomation) at 50 • C. Using the sample hydr ol ysis pr ocedur e, we determined the fucose content within the dry MacFarlane medium components (outlined below) including carbohydrate and peptide sources. Dried samples wer e r e-suspended in 1 ml MilliQ w ater, v ortexed and passed through a 0.45 μm syringe filter with a nylon membrane. L-fucose (Mer ck) w as processed the same w ay for comparison.
Neutr al monosacc harides wer e anal ysed on a Dionex ICS-6000 high-performance anion exc hange c hr omatogr a phy coupled with a pulse amperometric detector (HP AEC-P AD) (all Thermo Scientific). The system was equipped with a Single Pass/Double Pass pump working as single pass, an AS-AP autosampler, a 10 μL injection loop and an ICS 6000 electr oc hemical detector operated with a gold electrode and an AgCl reference electrode. Neutral monosacc harides wer e separ ated at 25 • C using a Dionex CarboP ac PA-1 (2 ×250 mm) column attached to a Dionex CarboPac PA-1 guard column (2 ×50 mm) at a flow rate of 0.25 ml min −1 . The mobile phase consisted of nitrogen degassed solvent A (MiliQ water), and 9% solvent B (200 mM NaOH) run in isocratic mode . T he system was controlled by Chromeleon 7.2 SR4 (Thermo Scientific). Monomers were identified with external standards and all samples were analysed in triplicates.

Identification of functional chemical moieties by Fourier transform infrared (FT-IR) spectroscopy
A NICOLET Summit infr ar ed spectr oscopy (Thermo Scientific) equipped with a diamond crystal ATR Ev er est pr obe and under control of Omnic Paradigm software was used to perform FT-IR spectr oscopy anal ysis of dried fucoidan, Fucus extract and for comparison, fucose. Spectra were collected between 400 and 4000 cm −1 at a resolution of 2 cm −1 av er a ging 32 scans.

CHNS element analysis
Carbon, hydrogen, nitrogen and sulfur (CHNS) composition of the extr act was anal ysed using a VarioMacr ocube (Elementar) calibrated on a sulfanilamide standard after drying the samples in foil at 100 • C overnight. Samples were prepared by weighing 23, 22, 21.4, and 7 mg of dry mucin, fucose, fucoidan, and Fucus extr act, r espectiv el y, into tin foil capsules, along with one pinch of vanadium pentoxide that was required to combust the sulfur.

Donor recruitment and fecal sample processing
Fresh fecal samples were collected from healthy donors over a period of five months in Aarhus, Denmark. Anonymous sample collection and further processing is exempt from ethic approval according to the National Scientific Committee (National Videnskabsetisks K omite , NVK, Denmark). T he donors w ere betw een 20-49 years and had regular eating patterns and bo w el mo vements . Donors did not take any food supplements containing prebiotics or probiotics nor used any medication affecting the gut transit and digestion during the last three months preceding the sample donation. All donors provided written consent.
Samples were obtained over two sampling campaigns (experiment 1 and 2), and donors were not discouraged from donating in both experiments. In total, 17 fresh fecal samples were collected. Each fecal sample was immediately transferred to a sealed bag containing an anaerobic gas pack (BD) and was processed within 4 h of defecation. To pr epar e fecal slurries, all work was conducted in an anaerobic bench (Baker Ruskinn) in order to k ee p a strict anaer obic envir onment. Appr oximatel y 1 g of fr esh fecal sample was added to a tube containing four sterile glass beads for breaking down the structure. Peptone water was added to obtain a 10% (m/v) solution. The samples were vortexed for 5 min in order to create a homogenous mixture and were left for 5 min for the larger particles to sediment before inoculation. Aliquots were collected from fecal samples and stored at −20 • C for DNA isolation.

In vitro batch fermentations
In vitro in batch fermentations were conducted with freshly prepared fecal slurries as inoculum to investigate the effect of fucoidan as a prebiotic. Media were inoculated with 1% fecal slurry. CON, FUS0.4 and FUS0.8 fermentations were run in triplicates, EXT was run in duplicates in experiment 1. In experiment 2, CON and FUS were always fermented in triplicates, while FUC and FUS1.6 were run in duplicates for three donors . T here was no FUC incubation for donor sample 17.
Flasks were placed in a shaking incubator at 140 r/m and 37 • C. Samples were collected after 0, 24, and 48 h of fermentation for further analyses. For H 2 S analysis, fresh fermentation broth was used as outlined below. Additional broth was collected, supernatant and biomass were separated by centrifugation and stored at −20 • C for analysis of fermentation metabolites and DNA isolation, r espectiv el y. The pH was determined after 48 h of fermentation in experiment 1 using a FiveEasy pH meter F20 (Mettler Toledo).

DN A extr action fr om feces and in vitr o batch fermentations
DN A w ere both extracted from fecal samples and pellets collected from fermented FUS1.6 samples after 48 h (1 ml) with the FastDNA Spin Kit for soil (MP Biomedicals), which includes a bead beating step, and was eluted in 50 μL elution buffer. The extracted DNA was diluted 10-fold for all further assa ys .

Micr obiota pr ofiling with 16S rRNA gene sequencing and data analysis
For library pre paration, a two-ste p PCR approach was used according to Illumina's 16S Metagenomic Sequencing Library Preparation guide. Briefly, the V3-V4 hypervariable region of the 16S rRNA gene was amplified using Bac341F and Bac805R with adapters ( Table S1 ) and a master mix containing 12.5 μL 2xKAPA HiFi Hot-Start readyMix (Roche) 0.5 μl forw ar d and reverse primer, 1 μl DNA and 10.5 μl nuclease free water. PCR conditions were 25 cycles of denaturation at 95 • C for 30 s, annealing at 55 • C for 30 s and extension at 72 • C for 30 s follo w ed b y 72 • C for 5 min. The second PCR emplo y ed primers with bar code and amplification for 8 c ycles. Samples w er e purified using Ampur e XP beads (Bec kman Coulter, Denmark) before sequencing. A pooled library comprising the amplicons of all samples was used for sequencing on a MiSeq sequencer (Illumina) at the Section of Microbiology at Aarhus University according to standard Illumina protocols. All samples including a negative control starting from a mock DNA isolation procedure and 61 samples from feces and fermentations were analysed in the same run and served as input for further bioinformatics processing.
Primer sequences ( Table S1 ) wer e r emov ed using cut adapt (v4.2; -O 12 -discard-untrimmed -g CCTA CGGGNGGCWGCA G -G GA CTA CHV GGGTATCTAATCC -pair-ada pters -minim um-length 75) (Martin 2011 ) and only inserts that contained both primers and were at least 75 bases were k e pt for downstream analysis. Reads wer e quality filter ed using the filterAndTrim function of the dada2 pac ka ge (maxEE = 2, truncQ = 3, minLen = 150, trimRight = 40, trimLeft = 40). The learnErrors and dada functions (Callahan et al., 2016 ) were used to calculate sample inference using pool = pseudo as parameter. Reads were merged using the mergePairs function and bimeras were removed with remo veBimeraDeno vo (method = pooled). Remaining Amplicon Sequence Variants (ASV) wer e taxonomicall y annotated using the IDTAXA classifier (Mur ali et al. 2018 ) in combination with the Silva v138 database (Quast et al. 2013 ). The median number of reads per processed sample was 26.904 (range 16.228-33.789 reads), the negative control yielded 563 reads. One fermentation sample was removed from further analysis as it failed sequencing.

Quantification of specific functional bacterial groups
We used quantitative PCR (qPCR) to analyse counts of total bacteria and the abundance of specific micr obial gr oups. To quantify total bacteria, the 16S rRNA gene was used as target ( Table S1 ). For sulfate-reducing Desulfovibrionaceae , both the 16S rRNA gene and the gene dsrA encoding subunit A of the dissimilatory sulfate r eductase wer e employed ( Table S1 ). Abundance of A. hallii was determined using pduC , the gene encoding the major subunit of the pr opanediol/diol dehydr atase as described (Ramir ez-   ( Table S1 ). A CFX Connect Real-Time PCR System (Bio-Rad) was used. The qPCR master mix contained 5 μL iTa g Univ ersal 2x SYBR Green Supermix (Bio-Rad), 1 μL forw ar d and r e v erse primer ( Table S1 ), 2 μL nuclease free water and 1 μL of diluted DNA in a total volume of 10 μL in 96-well (low profile clear/clear) PCR plate, sealed with Microseal 'B' Sealing film (both Bio-Rad).
Each sample was run in duplicates, each run included a standard and a negative control (nuclease free water). The standard curv es wer e made with linearized plasmids or purified PCR products. Reactions were run with the following temper atur e pr ofiles: one cycle of hot-start activation at 95 • C (3 min), denaturation at 95 • C for 10 s, annealing at 60 • C for 30 s for 40 cycles followed by melting curve analysis. Absolute cell abundance was calculated based on standard curv es, corr ection factors wer e used to account for multiple 16S rRNA gene copies in gut microbes (Stoddard et al. 2015 , Table S1 ) .

Metabolite analysis by high performance liquid chroma togr aphy coupled to a refractive index detector (HPLC-RI)
The concentrations of main SCFAs acetate, propionate and butyrate along with fucose, and lactate in feces and fermentation samples were determined by HPLC-RI using a 1260 Infinity II LC System with RID (all Agilent). Fermentation metabolites were sep-arated using a Hi-Plex H column (300 ×7.7 mm) attached to a guard (50 ×7.7 mm) column.
To extract fermentation metabolites from feces, 200-300 mg material was mixed with 5 mM H 2 SO4, vortex and centrifuged 10 000 r/m, 10 min. Supernatants fr om batc h fermentations were collected by centrifugation at 10 000 r/m for 4 min. All samples wer e filter ed thr ough a 0.45 μm n ylon membr ane filter befor e analysis . T he samples (10 μL injection volume) were eluted with 5 mM H 2 SO 4 at a flow rate of 0.6 ml min −1 at 40 • C. Fermentation metabolites were quantified using external standards.

Photometric determination of H 2 S in water
The H 2 S concentration was measured in fermentation samples collected at 0 and 48 h. The assay was ada pted fr om Cline ( 1969 ) that is based on the reaction of H 2 S with N, N-dimethyl-1,4phen ylendiamin that pr oduces methylene blue om the presence of iron (III) (F e3 + ). T he absorbance of the complex at 670 nm is proportional to the H 2 S concentration. Briefly, fermentation samples (0.5 ml) wer e r ecov er ed fr om closed serum flasks and were immediately dispersed in 0.5 ml 5% (w/v) zinc acetate to entr a p the H 2 S. Samples were vortexed and stored at −20 • C until further processing. After thawing, 20 μL was mixed with 980 μL MilliQ water. Diamine r ea gent (80 μL) w as added, samples w ere shaken and placed in the dark at r oom temper atur e for 30 min. After, 300 μL wer e tr ansferr ed to a 96 well micr otiter plate together with a standard curve prepared using zinc sulfide, and absorbance was measured in a FLUOstar Omega plate reader (BMG Labtech) at 670 nm.

Statistics
For α-div ersity anal ysis, ric hness and e v enness wer e calculated with the vegan package (Oksanen et al. 2023 ). Samples were normalized with r ar efaction based on the minimal sum of all reads in the sample (n = 16.228). For β-diversity analysis, coordination of samples were conducted based on Euclidean distance using nonmetric multidimensional scaling (NMDS), which is for datasets with multiple dimensions including distance . T he significance of coor dination b y differ ent tr eatments w as tested with P ermutational Multivariate Analysis of Variance (PERMANOVA).
The relationship of microbial composition in fecal samples and of major fermentation metabolites formed in fermentations was determined using Factor Analysis for Mixed Data (FAMD) using the R pac ka ges FactoMineR and Factoextra.
Statistical analysis of fermentation metabolites and microbial abundance data was performed using ANOVA and paired t-test implemented in PAST (Hammer et al. 2001 ) and SigmaPlot V15 (Alfasoft).

Fucus extract contained laminarin and fucoidan
While ther e hav e been r eports that fecal micr obiota, or selected gut microbes can, at least selectively, utilize algal polysaccharides (Cherry et al. 2019 ), there has been no systematic development and testing to position fucoidan as prebiotic, and no comparison to the major monomer L-fucose, possibl y partl y due to low commercial a vailability. T his led us to in vestigate isolation of fucoidan in this study. The dry matter of F. vesiculosus used for the extraction w as 19.0%. Emplo ying a micro w ave assisted extraction pr ocedur e, 142 mg of extr act was obtained with a yield of 0.2% (w/w% dry weight). The fucose content in the extract was analysed with HP AEC-P AD, together with the commercial fucoidan and fucose. Fucose contained mainly fucose (92.9 ±0.7%) whereas the commercial fucoidan had a fucose content of 78.6 ±2.2% and contained between 2.4% and 5.7% of galactose , mannose , rhamnose, glucose and xylose ( Table S2 ). The Fucus extract contained 11.1 ±2.9% fucose and mainly glucose (70.9 ±3.6%) ( Table S2 ).
T he sea weed extr action pr ocedur e used in this study was ada pted fr om Fletc her et al. ( 2017 ) and modified to the use of micro w ave assisted extraction according Ptak et al. ( 2019 ) albeit with modifications. Instead of 0.01 M H 2 SO 4 , 6% (w/v), at 120 • C for 30 min, 0.1 M HCl, 3.3% (w/v) at 90 • C for 15 min was applied together with lo w er temper atur e during micr o w ave assisted extr action, whic h can be one of the reason for the low fucoidan yield obtained. Furthermore Ptak et al. ( 2019 ) neutralized pH to 5-7 after micro w ave assisted extraction and removed laminaran with 40% (v/v) EtOH before the pr ecipitation, whic h was not done in this study. The high proportion of glucose in seaweed extract indicated that the Fucus extract was rich in laminarin and contained about 10% fucoidan.

Sulfur was present in fucoidan and the Fucus extract
To examine the element composition with a focus on sulfur content, CNHS element analysis was conducted of fucose, commercial fucoidan, Fucus extr act, and m ucin for comparison ( Table S3 ). Fucoidan contained 9.39% S, Fucus extr act 0.29%, m ucin 0.57% and fucose 0.04%. The minimal sulfur content of the medium, which could be derived from sulfate, was calculated based on the added le v els of m ucin, fucoidan and Fucus extr act as at least 22.7-22.8 mg L −1 (or 1.4 mM) for CON and FUS0.4-FUS1.6, 60.8 mg L −1 (or 3.8 mM) for FUC and 23.8 mg L −1 (or 1.5 mM) for EXT.
To further investigate the presence of sulfur and fucose in the Fucus extract and the nature of its chemical bond, we performed FT-IR spectroscopy ( Fig. S2A, B ). With reference to the fucoidan spectrum, we attributed the peak at 1224 cm −1 and its shoulder at 1250 cm −1 to S-O str etc hing of sulfate ester gr oups, and the peaks at 893 cm −1 and 830 cm −1 to C-O-S str etc hing ( Fig. S2B ) (Ptak et al. 2021, Almeida-Lima et al. 2010. Fucus extract sho w ed a clear peak at 1250 cm −1 . The peaks associated with C-O-S bonds were less evident. While the peak at 884 cm −1 can indicate the presence of C-O-S bonds typical of fucoidan (shifted with respect to 893 cm −1 ), it was pr obabl y associated to the anomeric structure of glucose (880-889 cm −1 ) (Rajauria et al. 2021 ). This latter inter pr etation, together with the presence of peaks near 1420 cm −1 and 1620 cm −1 is compatible with the presence of laminarin in the sample (Synytsya and Novak 2014 , Rajauria et al. 2021 ).
Similarly, the C-O-S peak at 830 cm −1 was not or only weakly pr esent. Giv en the complexity of the peaks in the fingerprint region observed by FT -IR spectroscopy , and the overlapping of bands that can be attributed to different bonds (e.g. peaks at 1129 cm −1 and 893-876 cm −1 ), it was not possible to undoubtedly substantiate the sulfur bonds in the Fucus extr act. In an y case, FT-IR analysis provided a strong indication that S-O (and C-O-S bonds) were present in fucoidan and to a lesser extent in Fucus extract in line with the findings derived from elemental analysis.
For fucose, the peak at 876 cm −1 was attributed to the unique c har acteristic vibr ational band of OH deformation ( Fig. S2B ) (Kossack et al. 2013 ). Other bands characteristic of OH stretching were visible at 3400 cm −1 , 3370 cm −1 , 3320 cm −1 , and 3240 cm −1 . The bands at 1170 cm −1 and 964 cm −1 that were attributed to the CH 3 str etc hing, and the band at 1129 cm −1 attributed to C-O-S or C-O-C str etc hing and deformation, were common to both fucose and fucoidan structures ( Fig. S2A ) (Kossack et al. 2013 ). The peak of OH deformation that is c har acteristic of fucose, was not present or shifted in the spectrum of the Fucus extr act, whic h can be indicative of either the absence or a low concentration of fucose in the sample.

Addition of fucoidan and Fucus extract had minor o ver all impact on SCFA profiles
To determine the fermentability of fucoidan and Fucus extract, we conducted in vitro batch fermentations using taxonomically diverse human fecal microbiota ( Supplementary Results , Table S4 , Fig. S3 ) as inoculum. In experiment 1, batch fermentations were conducted with fecal microbiota of nine donor samples (D1-D9) and were incubated for 48 h in MacFarlane containing fucoidan (FUC) or Fucus extract (EXT), while in experiment 2 donor samples D10-D16 were tested with FUC medium. Controls (CON) were without additional supplementation and did not contain detectable le v els of fr ee fucose. We estimate the fucose content of Mac Farlane medium components at 0.05 ±0.02%. The production of SCFA was measured after 24 and 48 h of fermentation by HPLC-RI. The pH was determined after 48 h for CON and FUS0.4: 6.3 ±0.1, FUC: 6.4 ±0.1, and EXT 6.2 ±0.2 (experiment 1).
The release of fucose from fucoidan depends on the presence and activity of fucanases. Fucanases of gl ycosyl hydr olase (GH) families GH107 and GH168 cleave fucose from fucoidan polymers but were only described in environmental microbial isolates (Schultz-Johanson et al. 2018, Shen et al. 2020 and not in fecal micr obiota. In contr ast, GH29 and GH95 fucosidases that hydr ol yse shorter fuco-oligosaccharides or terminal fucose, are frequently harboured by gut microbes (Wu et al. 2023 ). As the addition of fucoidan did not change median levels or proportions of propionate compar ed to contr ols in experiments 1 and 2 (Fig. 3 B), our results suggest that fucose was not released from fucoidan at levels that would lead to detectable and consistent shifts of the SCFA profiles during the 48 h of incubation, which is in a gr eement with the gener all y low usability of fucoidan as prebiotic that was observed (Gotteland et al. 2020 ).
In a pr e vious in vitro study, laminarin was completely utilized during fermentation when supplied as sole carbohydrate source (Devillé et al. 2007 ). While our analyses did not allow to determine the fate of the supplied polymers, lo w er median le v els of acetate tr eatment-contr ol (FUC: -1.7 and -2.8 mM at 24 and 48 h, respectiv el y, EXT: -3.7 and -2.2 mM at 24 and 48 h, r espectiv el y) indicate a rather fermentation inhibitory than fermentation supportive effect by the addition of fucoidan (FUC) or Fucus extract (EXT) (Fig. 3 B).

Fucose addition increased propionate levels
In both experiment 1 and 2 we also performed in vitro incubations that were supplied with fucose. In experiment 1, the supplied fucose (FUS0.4, 2.8 mM) was depleted after 24 h and the fermentation intermediate 1,2-PD was not detected in any sample while the fermentation intermediate lactate was detected in three samples only (3.3-15.2 mM). Complete fucose utilization and the lack of detection of 1,2-PD suggested that all donor microbiota were capable of fucose based cross-feeding. Box plots show median, 25 th and 75 th percentiles, the whiskers indicate 5 th and 95 th percentiles. Dots r epr esent means of the individual donor fermentation that were run in triplicates or duplicates as stated in the methods. Differences between all fucose treatments (C) were determined using One-Way Anova with Holm-Sidak All Pairwise Multiple Comparison Procedures; differences between FUS0.8 and FUS1.6 were also tested with pair ed t-test. Tr eatments that differ significantl y ( P < 0.05) or with a tr end (0.5 < P < 0.1) ar e indicated in the gr a ph; * indicates the p-v alue deriv ed fr om the paired t-test.
As we observed in experiment 1 that the addition of fucose increased the formation of propionate in comparison to CON, we compared two higher concentrations of fucose (FUS0.8, 4.2 mM, and FUS1.6, 9.9 mM) to CON in experiment 2. Again, the supplied Samples were fermented in triplicates unless indicated otherwise in the methods, dots r epr esent means of the triplicates. Lines connect samples from the same donor. Box plants show median, 25 th and 75 th percentiles, the whiskers indicate 5 th and 95 th percentiles. Differences between all fucose treatments (C) were determined using One-Way Anova with Holm-Sidak All Pairwise Multiple Comparison Procedures; differences between FUS0.8 and FUS1.6 were also tested with paired t-test. Treatments that differ significantly ( P < 0.05) or with a trend (0.5 < P < 0.1) are indicated in the graph; * indicates the p-value derived from the paired t-test.
fucose was depleted after 24 h, and 1,2-PD was not detected in any sample; lactate was r ecov er ed fr om 9 samples (1.3-12.3 mM) after 24 h with no clear pattern related to treatment. While the addition of 0.8 g L −1 fucose did not lead to the expected ov er all higher pr opionate le v els compar ed to FUS0.4, pr opionate FUS0.8/FUS1.6-CON le v els wer e significantl y ( P < 0.05) higher in FUS1.6 compared to FUS0.4, and to FUS0.8 if paired t-test was used (Fig. 3 C).
Similar as reported by Ramirez et al. ( 2021 ), who supplemented intestinal microbiota with the fucose fermentation intermediate 1,2-PD, this study observed lo w er acetate le v els and a higher proportion of butyrate in addition to propionate when fucose was ad ded. Ad ditionally, this effect was dose dependent with a significantl y differ ence of acetate le v els with higher le v els of fucose ( P < 0.05, Fig. 3 C). As certain bacterial groups can produce butyrate from lactate and acetate (e.g. A. hallii ), our observations point out that at the addition of fucose might impact on cross-feeding activities not only with 1,2-PD but also with lactate as intermediate, and that the effect is dose-dependent.

Addition of fucose reduced H 2 S levels in a dose dependent manner
As fucoidan rich in sulfate could serve as a source for the formation of H 2 S, the concentration of H 2 S was determined for all fermentations after 48 h of incubation with a photometric assay. In the colon, inorganic sulfate and sulfite, and sulfo amino acids such as cysteine act as sources for the production of H 2 S by sulfate reducers including dissimilatory sulfate reducers, or by utilizers of amino acids (Yao et al. 2018 ). Basic MacFarlane medium contained porcine mucin, tryptone and yeast extract which provide substrates for the production of H 2 S, and H 2 S formation was observed during fermentation in all samples (Fig. 4 A). H 2 S concentrations of CON ranged from 0.92 to 2.03 mM (Fig. 4 A) with no difference between CON, FUC ( median -0.04 mM; Q1;3 -0.16;0.28 mM) and EXT ( median 0.05 mM; Q1;3 -0.13;0.29 mM) (Fig. 4 B). H 2 S produced in the presence of FUS0.4 was 0.01-0.95 mM lower than CON ( median -0.20 mM; Q1;3 -0.31;0.05 mM) and H 2 S le v els w ere lo w er for FUS1.6 ( median -0.27 mM, Q1;3 -0.52;-0.17 mM) while H 2 S FUS1.6-CON was significantly lower ( P < 0.05 mM) than H 2 S FUS0.8-CON (paired t-test; Fig. 4 C).
As the addition of the heavily sulfated fucoidan or the laminarin-rich Fucus extract did not enhance H 2 S formation, our results indicate that bound SO 4 2 − was not a major contributor to H 2 S formation in batc h cultur es. Yao et al. ( 2018 ) observed that the addition of cysteine increased H 2 S formation by fecal micr obiota mor e effectiv el y compar ed to sulfate supplementation and proposed that a major part of intestinal H 2 S is formed from amino acids . T he addition of fermentable FOS reduced H 2 S formation e v en when added together with cysteine as carbohydrates metabolism might be preferred over amino acids (Yao et al. 2018 ). Similarly, the addition of the fermentable deoxyhexose fucose might have contributed to the lo w er H 2 S formation observed in this study as a favorable substrate compared to amino acids.

Fucose addition enhanced the abundance of selected bacterial groups linked to fucose utilization and cross-feeding
Next, w e conducted 16S rRN A gene sequencing of biomass collected after 48 h batch fermentations of CON and FUS1.6 to de- Figure 5. Differences in microbiota composition between CON and FUS1.6 (experiment 2). Microbiota composition was determined using 16S rRNA gene sequencing targeting the V3-V4 region. (A) Beta-diversity plots of CON and FUS1.6 of individual donor samples (D10-17, shown as 10-11) after 48 h fermentation. (B) Relative abundance of major families present in fermented samples CON and FUS1.6 was determined after 48 h incubations using 16S rRNA gene sequencing targeting the V3/V4 r egion. Relativ e abundance of selected bacterial families (C) and genera (D) that differed between CON and FUS 1.6after 48 h fermentation. Samples were fermented in triplicates unless indicated otherwise in the methods, differences were calculated from means. Box plants show median, 25 th and 75 th percentiles, the whiskers indicate 5 th and 95 th percentiles. Dots r epr esent means of the individual donor fermentation that were run in triplicates or duplicates.
termine how fucose addition impact microbial composition. We additionally quantified selected microbial groups (i.e. the hydrogen and propionate producing A. hallii and sulfate-reducing Desulfovibrionaceae ) using qPCR to test whether the observed major differences in SCFA profiles and H 2 S formation were linked to differences in microbial abundance. Based on α-diversity analysis, the median number of observed species (170, range 111-223) was significantly ( P < 0.05, Kruskal-Wallis test with Dunn's posthoc test) lo w er for fermentation samples compared to feces ( Table S4 ). Similarly, both Shannon and Simpson indices were lower in fermented samples than in feces ( Table S4 ). When compared between treatments, the number of observed species was higher in CON samples than in FUS1.6 ( Table S4 ). For β-diversity analysis, the significance of coordination by different donors and sampling time (t = 0 h, fecal samples, and t = 48 fermentations) were tested with permutational multiv ariate anal ysis of v ariance (PERMANOVA). The distance was significantly coordinated by time and donors ( P = 0.001) ( Fig. S5 ) but not by treatment (Fig. 5 A).
Our analysis thus identified bacterial groups that have been pr e viousl y linked to fucose utilization and cross-feeding: Formation of propionate and propanol from fucose has been shown for gut bacteria belonging to the family Lachnospiraceae , including Roseburia species (Scott et al. 2006 ) or Lachnoclostridium (Petit et al. 2013 ). Bifidobacterium spp. and Lacticaseibacillus rhamnosus are able to degrade fucose to 1,2-PD, or to 1,2-PD and lactate (Becerra et al. 2015, Bunesova et al. 2016, which can serve as metabolite in 1,2-PD cr oss-feeders suc h as A. hallii, Blautia obeum, Ruminococcus gnavus, Flavonifractor plautii and Limosilactobacillus reuteri (Engels et al. 2016, Schwab et al. 2017, Zhang et al. 2019 or lactate utilizers including again A. hallii and Veillonella spp. whic h pr oduce butyr ate or pr opionate, r espectiv el y (Fig. 1 ) (Ng andHamilton 1971 , Duncan et al. 2004 ). We pr e viousl y identified A. hallii as a k e y taxon in the metabolism of 1,2-PD and gl ycer ol, whic h ar e both catal yzed by the enzyme gl ycer ol/diol dehydr atase (Ramir ez , and also observed an increase of abundance of the A. hallii group in this study. A. hallii can produce butyrate from lactate and acetate (Duncan et al. 2004 ), which might have contributed to the lo w er acetate le v els in FUC1.6 samples.
Using an in silico a ppr oac h, Br accia et al. ( 2021 ) r eported that cysteine degr aders ar e common within the human micr obiota and more abundant than sulfate-reducing bacteria, which is in agreement with the low abundance of Desulfovibrionaceae observed in this study. Enterobacteriaceae r epr esent a taxon that harbors a diversity of genes encoding enzymes involved in cysteine degradation and H 2 S production (Braccia et al. 2021 ). We observed a lo w er r elativ e abundance of Enterobacteriaceae in FUS1.6 fermentations, which could be due to a competitive disadv anta ge of utilizers of cysteine or other sulfo amino acids in the presence of carbohydrates in agreement with Yao et al. ( 2018 ). Veillonella spp., whose r elativ e abundance was higher in most FUC1.6 samples, can concurr entl y utilize lactate to form propionate, acetate and H 2 (Distler and Krönke 1981 ) and metabolise cysteine to produce H 2 S (Washio et al. 2014 ). The extent of cysteine metabolism depends on growth state (resting cells versus cell extract), pH (more H 2 S formed at pH7 than pH5) and lactate le v els (significant higher levels formed in the presence of 10 mM lactate) (Washio et al. 2014 ). T he o v er all lo w er H 2 S le v els in FUC1.6 samples indicate that the higher abundance of Veillonella was rather linked to lactate crossfeeding.

Conclusion
The addition of fucose had a major impact on fermentation metabolite cross-feeding via both 1,2-PD and lactate and was also linked to alterations of microbial community composition. We show for the first time that addition of fucose reduced H 2 S formation possibly due to pr eferr ed utilization of the provided carbohydr ate compar ed to amino acids. While it was pr e viousl y shown that A. hallii can benefit within a microbial community from the utilization of the pathway intermediate 1,2PD, we report here that the species also benefits through cross-feeding during fucose metabolism, which might be relevant in the gut ecosystem.
The effects of fucose addition can be consider ed pr ebiotic with the observed increase of propionate and butyrate formation. Howe v er, this study highlights an important consideration in prebiotic r esearc h addr essing whether human gut micr obiota harbours the necessary enzymatic functionality to degrade specific polysacc harides suc h as fucoidan. As dietary fucose might be absorbed during gastrointestinal transit, and the corresponding fucoidan polymer seemed to be little utilized under the tested conditions, biotec hnologicall y pr oduced fuco-oligosacc harides could be the solution for precise fucose-based microbiome engineering strategies.