Microbiota functional activity biosensors for characterizing nutrient utilization in vivo

Methods for measuring gut microbiota biochemical activities in vivo are needed to characterize its functional states in health and disease. To illustrate one approach, an arabinan-containing polysaccharide was purified from pea fiber, its structure defined, and forward genetic and proteomic analyses used to compare its effects, versus unfractionated pea fiber and sugar beet arabinan, on a human gut bacterial strain consortium in gnotobiotic mice. We produced ‘Microbiota Functional Activity Biosensors’ (MFABs) consisting of glycans covalently-linked to the surface of fluorescent paramagnetic microscopic glass beads. Three MFABs, each containing a unique glycan/fluorophore combination, were simultaneously orally gavaged into gnotobiotic mice, recovered from their intestines, and analyzed to directly quantify bacterial metabolism of structurally distinct arabinans in different human diet contexts. Colocalizing pea-fiber arabinan and another polysaccharide (glucomannan) on the bead surface enhanced in vivo metabolism of glucomannan. MFABs represent a potentially versatile platform for developing new prebiotics and more nutritious foods.

beads. First, the surfaces of 10 µm-diameter glass beads were sialyated by reaction with an amine-and/or 250 phosphonate-organosilane (Step 1 in the Figure 3A). This approach provided us with control over the 251 stoichiometry and properties of surface functional groups (amine and phosphonate) to be used for further 252 derivatization with a fluorophore and ligand immobilization. We found that coating with a 1:1 mol ratio 253 of (2-aminopropyl)triethoxysilane (APTS) and 3-(trihydroxysilyl)propyl methylphosphonate (THPMP) to 254 install both amine and phosphonate functional groups on the bead surface provided a nucleophilic handle 255 and decreased nonspecific ligand binding and bead aggregation (Bagwe et al., 2006). Surface sialyation 256 and the amount of reactive surface amine functional groups were monitored by measuring the zeta 257 potential of beads following organosilane derivatization with or without amine acetylation (Figure 3 ¾  258   figure supplement 1A). Surface amine functional groups were subsequently quantified using a ninhydrin-259 based colorimetric assay (Soto -Cantu et al., 2012). The results revealed that amine plus phosphonate 260 functionalized beads contain 2.18 x10 10 ± 3.49 x10 9 (mean ± SD) reactive amines per bead versus 1.20 261 x10 9 ± 1.92 x10 8 reactive amines after surface acetylation (Figure 3 ¾ figure supplement 1B). By 262 comparison, streptavidin-coated beads used in our previously published procedure (Patnode et al., 2019) 263 possess 23-fold fewer potential binding sites (9.21 x10 8 ± 1.13 x10 8 molecules of biotin per bead). 264 Second, we attached unique fluorogenic tags directly to the bead surface so that multiple bead types with 265 different immobilized polysaccharides could be analyzed simultaneously within a given gnotobiotic 266 animal. To do so, surface-modified beads were reacted with an N-hydroxysuccinimide (NHS) ester-267 activated fluorophore (Step 2 in Figure 3A). Fluorophore coupling was specific to beads with surface 268 amines (Figure 3 ¾ figure supplement 1C). Bead fluorescence could be modulated over four orders of 269 magnitude simply by titration of the reactant fluorophore (Figure 3 ¾ figure supplement 1D). Low 270 levels of fluorophore immobilization on beads not coated with APTS, or on acetylated beads likely 271 reflects incomplete acetylation with acetic anhydride or nonspecific fluorophore adsorption. Third, 272 polysaccharide was activated by reaction with 1-cyano-4-dimethylaminopyridinium tetrafluoroborate 273 (CDAP) to generate an electrophilic cyanate-ester intermediate ( Figure 3B); activated polysaccharide 274 reacts with amines on the surface of the amine plus phosphonate bead. Lastly, a hydride reduction was 275 performed to reduce any Schiff base formed with the polysaccharide reducing end and likely the resultant 276 isourea bond (see  (Figure 3C), likely through reductive amination of the polysaccharide reducing end. Lastly, using 291 maltodextrin as a model, we tested whether oligosaccharides could also be immobilized using CDAP 292 activation and capture onto amine plus phosphonate beads. Maltodextrin (dextrose equivalent 13-17; 293 estimated Mn ~ 1,300 Da) was activated with CDAP (0.2 mg/mg oligosaccharide) prior to attachment to 294 the bead surface. Conjugation resulted in 4.43 ± 1.03 ng of immobilized glucose per 1000 beads, or 2.05 x 295 10 9 molecules of maltodextrin per bead (Figure 3 ¾ figure supplement 3C); this represents 9.4 % of all 296 available surface reactive amines. 297 To determine whether CDAP-298 immobilized polysaccharides are good proxies for hydrated plant cell wall fragments and polysaccharide 299 components (i.e. are anchored polysaccharides accessible to bacterial glycosidases that generate 300 oligosaccharide inducers of PULs), we first performed several in vitro experiments. We initially 301 incubated PFABN-coated beads with glycosyl hydrolyses and the fraction of arabinan remaining on the 302 bead surface was measured by GC-MS. Incubation with a combination of endo-arabinanase and a-L-303 arabinofuranosidase removed 88±1.2 % (mean±SD) of the immobilized PFABN after 30 minutes while 304 incubation with each enzyme individually removed greater than 80% of the glycan after 30 minutes and 305 more than 90 % after 20 hours. In contrast, an endo-inulinase with specificity toward b(1-2)-linked 306 fructose resides failed to degrade PFABN (Figure 4 ¾ figure supplement 1). These results indicate that a 307 substantial portion of the bead-immobilized glycan was susceptible to enzyme-catalyzed degradation. 308

Quantifying polysaccharide degradation with MFABs in vitro -
As a prelude to in vivo studies that would test the ability of MFABs to measure the saccharolytic 309 activity of the defined community in defined dietary contexts, we next incubated PFABN-coated MFABs 310 with B. thetaiotaomicron VPI-5482 or B. cellulosyliticus WH2 grown to mid-log phase in defined 311 minimal medium (McNulty et al., 2013). Saccharolytic activity was measured by using GC-MS to 312 quantify the mass of arabinose retained on beads as a function of incubation time and whether or not 313 soluble PFABN was included in the culture medium (reference control: beads recovered from control 314 incubations lacking bacteria). Under the conditions tested, added PFABN was required for degradation of 315 immobilized PFABN from beads; the amount of PFABN removed from the bead surface increased with 316 increasing incubation time, with maximum values of 64±9% (mean±SD) and 48±3% achieved with B. 317 thetaiotaomicron VPI-5482 and B. cellulosyliticus WH2, respectively (Figure 4 ¾ figure supplement 2). 318 The effect of supplemented PFABN in this in vitro MFAB-based degradation assay is consistent with our 319 in vivo metaproteomic analysis demonstrating its ability to induce PULs involved in its utilization 320 (Figure 2 ¾ figure supplement 1). Together, these results provided evidence of the accessibility of bead-321 bound polysaccharide to degradation by secreted or bacterial cell surface-associated glycoside hydrolases. 322

Quantifying polysaccharide degradation with MFABs in gnotobiotic mice -PFABN and 323
SBABN were immobilized onto amine plus phosphonate-derivatized beads. Beads acetylated with acetic 324 anhydride after fluorophore labeling were used as controls ( Figure 4A). Each of these three bead types 325 contained a unique fluorophore. The three bead types were pooled and the mixture was introduced by oral 326 gavage into four groups of mice 10 days after they received the 14-member consortium: one group of 327 recipient animals had been fed the unsupplemented HiSF-LoFV diet while the other groups had received 328 HiSF-LoFV containing unfractionated pea fiber, PFABN or SBABN (n=5 animals/group). Germ-free 329 mice fed HiSF-LoFV supplemented with PFABN served as controls (n=5; Experiment 1). The bead 330 mixtures were harvested using a magnet from the cecums of animals four hours after their introduction by 331 oral gavage; the individual bead types were then purified by fluorescence-activated cell sorting (FACS). 332 Polysaccharide degradation was quantified by gas chromatography-mass spectrometry (GC-MS) of 333 neutral monosaccharides released after acid hydrolysis of the purified beads. Results were referenced to 334 the masses of monosaccharides released from aliquots of each input bead type (i.e., the same bead 335 preparation but never introduced into mice). 336 The quantities of neutral monosaccharides liberated by acid hydrolysis from the surfaces of beads 337 recovered from the cecums of germ-free mice were not significantly different from the amounts liberated 338 from the input bead preparations with one exception -a slight, albeit statistically significant, increase in 339 galactose from beads coated with PFABN and SBABN (Figure 4 ¾ figure supplement 3 In contrast to germ-free controls, the mass of arabinan was significantly decreased when  or SBABN-coated beads were recovered from colonized mice fed the unsupplemented HiSF-LoFV diet 345  available. Therefore, we reasoned that the MFAB platform could provide a way of testing whether 372 deliberately colocalizing distinct polysaccharides, akin to natural plant fiber particles, would result in 373 'synergistic' polysaccharide utilization by microbial community members. 374 To explore this notion, we turned to glucomannan, a hemicellulosic linear b(1-4) polysaccharide 375 composed of D-mannose and D-glucose. We found that among the pea fiber-responsive Bacteroides Based on these considerations, we hypothesized that supplementing the diet with pea fiber would 391 induce expression of PULs in community members so that they could readily utilize bead-associated 392 PFABN; moreover, those community members that could utilize PFABN and express b-mannanases 393 would be able to more efficiently access/metabolize glucomannan positioned on the same bead. To test 394 this hypothesis, we synthesized beads coated with PFABN alone, glucomannan alone, or both glycans 395 together, as well as control acetylated beads that lack a bound polysaccharide ( Figure 5B). These four 396 bead types, each labeled with a distinct fluorophore, were simultaneously introduced into two groups of 397 mice colonized with the 14-member community -one group was fed the unsupplemented HiSF-LoFV 398 diet while the other group received a pea fiber-supplemented diet (n=7-8 mice/group) (Supplementary 399 file 2). Beads were recovered from their cecums 4 hours after gavage; the different bead-types were then 400 isolated using FACS ( Figure 5C) and subjected to acid hydrolysis and neutral monosaccharide analysis 401 by GC-MS. We used the amount of mannose remaining on the bead as a proxy of glucomannan 402 degradation because it represents the bulk of monosaccharide present in glucomannan and is absent in 403 PFABN. The results revealed that glucomannan on beads coated with glucomannan alone was degraded 404 to a similar extent in mice receiving the unsupplemented or pea fiber-supplemented HiSF-LoFV diets 405 ( Figure 5D and Supplementary file 5; p=0.87, Mann-Whitney U test). However, when presented with 406 PFABN on the same bead, significantly more glucomannan was degraded by the microbiota of mice 407 receiving the pea fiber-supplemented diet as compared to the unsupplemented diet ( Figure 5D; p<0.05, 408 Mann-Whitney U test). The amount of arabinose remaining on beads coated with PFABN and 409 glucomannan, and PFABN alone, was also significantly reduced (degradation increased) with pea fiber 410 supplementation ( Figure 5D We have used gnotobiotic mice colonized with a defined model human gut microbial community 418 containing 58,537 known or predicted protein coding genes together with metaproteomic and forward 419 genetic analyses to demonstrate the sensitivity of a microbiota to structural differences in glycans 420 (arabinans) isolated from two distinct plant sources. The specific activity of each isolated arabinan (in this 421 case the increase in absolute abundance of the responsive Bacteroides per unit mass of diet ingredient 422 consumed per day) was superior to unfractionated pea fiber ( Figure 2C) was recovered by centrifugation (15,000 x g, 15 minutes). The precipitate and soluble material from the 499 dialysis, representing fractions one and two, respectively, were dried with lyophilization. 500 The pellet from the ammonium oxalate extraction was washed with 200 mL of water, centrifuged 501 (4,000 x g, 15 minutes), and the supernatant was discarded. The pellet was resuspended in 200 mL of 50 502 mM sodium carbonate (pH 10) containing 0.5 % (wt:wt) sodium borohydride and stirred at 23 °C for 20 503 hours. The suspension was centrifuged (6,000 x g, 15 minutes) and the supernatant was collected. 504 Borohydride was quenched by slowly adding glacial acetic acid. A stringy precipitate began to form as 505 the pH decreased. The suspension was concentrated (as above); the insoluble and soluble portions of the 506 resulting concentrated carbonate suspension were separated with centrifugation (15,000 x g, 15 minutes), 507 yielding fractions three and four, respectively. Fractions were dialyzed and dried with lyophilization. 508 The pellet from the carbonate extraction was washed with water before resuspension in 200 mL 509 of 1 M potassium hydroxide containing 1 % wt:wt sodium borohydride and stirring for 20 hours at 23 °C. 510 The suspension was centrifuged (6,000 x g, 15 minutes) and the supernatant was removed. Five drops of 511 1-octanol were added to prevent foaming during borohydride quenching. A light precipitate began to form 512 in the solution as the pH decreased. The suspension was concentrated; the insoluble and soluble portions 513 of the concentrated 1 M hydroxide extract were separated by centrifugation (15,000 x g, 15 minutes), 514 yielding fractions five and six, respectively. The fractions were dialyzed and dried with lyophilization. 515 The pellet from the 1 M hydroxide extraction was washed with water before resuspension in 200 516 mL of 4 M potassium hydroxide containing 1 % wt:wt sodium borohydride. The mixture was stirred at 23 517 °C for 20 hours. The suspension was then centrifuged (6,000 x g, 15 minutes) and the supernatant was 518 removed. 1-Octanol were added to prevent foaming during borohydride quenching; during this process, a 519 precipitate formed, then dissolved, then reformed as the pH was lowered to 6.0. The resulting suspension 520 was concentrated; the insoluble and soluble portions of the concentrated 4 M hydroxide extract were 521 separated by centrifugation (15,000 x g, 15 min), yielding fractions seven and eight, respectively. These 522 fractions were dialyzed and dried with lyophilization. Note that after each extraction, sodium azide was 523 added to a final concentration of 0.05% prior to concentration and dialysis. 524 Characterization of pea fiber fractions -Each of the eight fractions was resuspended in water (1 525 mg/mL) by heating to 90 °C and sonication (Branson Sonifer). Insoluble material was removed by 526 centrifugation (18,000 x g, 5 minutes). The soluble material was assayed for protein content 527 (bicinchoninic acid assay) using bovine serum albumin as a standard, DNA content (UV-visible 528 absorbance spectroscopy, Denovix DS-11 spectrophotometer) and total carbohydrate content (phenol-529 sulfuric acid assay; (Masuko et al., 2005)) using D-glucose as a standard (Supplementary file 1). The 530 molecular size of each fraction was measured using an Agilent 1260 high performance liquid 531 chromatography (HPLC) system equipped with an evaporative light scattering detector. An Agilent Bio 532 Sec-5 column (Cat. No.: 5190-2526) and guard were used with water as the mobile phase. Unbranched 533 pullulan was employed as length standards (Shodex). The monosaccharide composition of each fraction 534 was measured using polysaccharide methanolysis followed by GC-MS (Doco et al., 2001). Five days prior to colonization, mice were switched to a HiSF-LoFV diet. This diet was produced 621 using human foods as described in a previous publication (Ridaura et al., 2013), freeze-dried and milled 622 Purified fecal DNA was processed as previously described (Wu et al., 2015). Genomic DNA was 678 digested with MmeI, size selected, ligated to sample-specific adapter primers, size selected, amplified by 679 PCR, and a specific 131 bp final product isolated from a 4% (wt:wt) MetaPhore (Lonza) DNA gel. 680 Purified DNA was sequenced, unidirectionally, on an Illumina HiSeq 2500 platform (50-nt reads) using a 681 custom primer that captures the species-specific barcode. Quantitation of each insertion mutant's 682 abundance (read counts) was determined using custom software (Wu et al., 2015). Count data were 683 normalized for library depth (within the same species), a pseudo count of 8 was added, and the data were 684 log2 transformed. Transformed count data from dpg 2 and dpg 6 were used to build linear models (limma  (Tanca et al., 2013). We have also documented an increase in total peptide 696 identifications, including membrane protein representation, after inclusion of SDS in the extraction 697 method which we have applied to a variety of microbial systems (Blakeley-Ruiz et al., 2019; Nickels et  698   al., 2020; Salvachua et al., 2020). While there may still be underrepresentation of some proteins, such as 699 membrane proteins with multiple transmembrane helices, if there is an underrepresentation of proteins 700 encoded by the PULs due to methodology-induced bias, this bias will be inherent and consistent across all 701 samples. As proteins encoded by the PULs are compared across treatment, we reasoned that 702 underrepresentation of these proteins would be the same across all samples and not impact our overall 703 findings. 704 Only data from peptides that uniquely map to a single protein were considered for analysis. 705 Summed peptide abundance data for each protein was log2-transformed. Missing data was imputed to 706 simulate 'instrument limit of detection' by calculating the mean and standard deviation of the protein 707 abundance distribution for each sample. Proteins included in this distribution were detected in more than 708 three mice within a given treatment group. Missing values were imputed as mean minus 2.2 times the 709 standard deviation with a width equal to 0.3 times the standard deviation. For species where greater than 710 100 proteins were quantified, data were normalized with cyclic loess normalization (limma package). 711 Loess-normalized protein abundance data were then used to build linear models (limma package) to 712 identify diet-supplement-responsive proteins (relative to levels in control mice receiving the 713 unsupplemented HiSF-LoFV diet) at dpg 6. P-values from the linear models were corrected for multiple 714 hypotheses. 715  Soto-Cantu et al., 2012). The reaction was allowed to proceed for 5 hours at 50 °C 724 with shaking and then terminated with three cycles of washing in water (using a magnet to recover the 725 beads after each wash cycle). Beads were stored at 4 °C in a sterile solution of 20 mM HEPES (pH 7.2) 726 and 100 mM NaCl. 727

Synthesis of amine plus phosphonate functionalized beads
Amine plus phosphonate bead acetylation -Beads were washed repeatedly with multiple 728 solvents with the goal of resuspending the beads in anhydrous methanol; to do so, beads were washed in 729 water, then methanol, then anhydrous methanol (1 volume equivalent; 5 x 10 6 beads /mL). Pyridine (0.5 730 volume equivalents) was then added as a base followed by acetic anhydride (0.5 volume equivalents). The 731 reaction was allowed to proceed for 3 hours at 22 °C and then terminated by repeated washing in water. 732 Beads were stored in 20 mM HEPES (pH 7.2) and 100 mM NaCl at 4 °C. Analytical characterization of beads -Bead Zeta potential was measured to characterize the 764 extent of modification of the bead surface; Zeta potential was determined for beads reacted with 765 organosilane reagents and beads subjected to surface amine acetylation. Zeta potential measurements 766 were made on a Malvern ZEN3600 instrument using disposable zeta potential cuvettes (Malvern). Beads 767 were resuspended to a concentration of 5 x 10 5 /mL in 10 mM HEPES (pH 7.2) passed through a 0.22 µm 768 filter (Millipore) and analyzed in triplicate. Measurements were obtained with the default settings of the 769 instrument, using the refractive index of SiO2 as the material and water as the dispersant. 770 Bead surface amines were quantified using a ninhydrin-based assay per the manufacturer's 771 instructions (Anaspec) with slight modification. Octylamine (Sigma Aldrich) in 60 % (vol:vol) 772 ethanol:water was used to generate a standard curve. Octylamine or beads resuspended in 60 % ethanol 773 were added to glass crimp top vials (1 volume equivalent; 100 µL). One volume equivalent of each of 774 three solutions used for detection of amines were added to the vials. Vials were then sealed and incubated 775 at 95 °C for 3 minutes. The reaction mixture was centrifuged (5,000 x g, 5 minutes) to remove the beads 776 and the absorbance at 570 nm was measured from the resulting supernatant (Biotek Eon). One million 777 amine plus phosphonate beads were analyzed, while three million acetylated amine plus phosphonate 778 beads were required for sufficient signal. Surface amines were calculated from linear regression of the 779 octylamine standard curve. was used to quantify the number of beads that had been transferred to that vial. The quantity of 793 monosaccharide released from a bead was determined from the linear fit of standards divided by the 794 number of beads transferred into the hydrolysis vial. For quantifying relative polysaccharide degradation, 795 the absolute amount of monosaccharide released from the bead surface was divided by the mass of that 796 monosaccharide quantified on input beads (with results expressed as a percentage). 797

Use of polysaccharide-coated MFABs in vitro -Individual glycosyl hydrolases, or combinations 798
of enzymes, were added to PFABN-coated beads (10 6 beads/mL in 400 µL of 100 mM sodium acetate 799 (pH 4) plus 0.5% bovine serum albumin). All reactions contained a total of 1 unit of enzyme (per the 800 manufacturer's documentation). Beads were incubated with rotation at 37 °C and aliquots were removed 801 after 30 minutes and 20 hours. Beads were then washed ≥3 times with 20 mM HEPES (7.2)/50 mM NaCl 802 on a magnetic tube stand, incubated at 80 °C for 10 minutes to inactivate any residual enzyme, and stored 803 in HNTB. The absolute mass of PFABN remaining on the bead surface was determined by using GC-MS 804 as described above. 805 Single bacterial colonies were picked and grown overnight in BMM (McNulty et al., 2013 was determined by using GC-MS as described above. 820

Gavage and recovery of polysaccharide-coated MFABs from mice -Each bead type was 821
individually sterilized by washing in 70% ethanol (vol:vol) twice on a magnetic tube stand before 822 resuspension in HNTB. A pool of 10-15 x 10 6 beads (2.5-3.75 x 10 6 per bead type) in 400 µL of HNTB 823 was prepared for each mouse; a 350 µL aliquot of the pool was introduced by oral gavage; the remaining 824 50 µL was analyzed as the input beads (see above). Beads were isolated from the cecums of mice four 825 hours after gavage or from all fecal pellets that had been collected from a given animal during the 3-to 6-826 hour period following gavage. Beads were typically used in mice within 30 days after polysaccharide 827 conjugation. 828 Recovered beads were resuspended in 10 mL of HNTB by pipetting and subsequently by 829 vortexing. The resulting slurry was passed through a 100 µm nylon filter (Corning; Cat. No.: 352360). 830 Beads were isolated from the suspension by centrifugation (500 x g, 5 minutes) through Percoll Plus (GE 831 Healthcare) in a 50 mL conical tube. Beads were recovered from the bottom of the tube; recovered beads 832 from each animal were distributed into four 1.5 mL sterile tubes and washed at least three times with 833 HNTB on a magnetic tube stand until macroscopic particulate debris from intestinal contents were no 834 longer observed. The material from four tubes were subsequently recombined and beads were stored in 835 HNTB containing 0.01% (wt:wt) sodium azide at 4 °C. 836 Bead types were purified by fluorescence-activated sorting (FACSAriaIII; BD Biosciences). 837 Aliquots of input beads were sorted throughout the procedure to quantify and monitor sort yield and 838 purity. Bead purity typically exceeded 98%. Sorted beads were centrifuged (1,500 x g, 5 minutes), the 839 supernatant was aspirated, and beads were transferred into a 0.2 mL 96-well skirted PCR plate. Beads 840 were washed with HNTB using a magnetic plate holder and stored at 4 °C in HNTB plus 0.01% (wt:wt) 841 sodium azide until analysis. Beads were subjected to acid hydrolysis of the bound polysaccharide and the 842 amount of liberated neutral monosaccharides was determined by GC-MS. All samples of a given bead 843 type were analyzed in the same GC-MS run; however, the order of analysis of a given bead type 844 recovered from animals representing different treatment groups was randomized. If sufficient beads were 845 available, each bead type from each animal was analyzed up to three times. To identify PULs whose expression were altered significantly, we performed gene set enrichment 864 analysis (GSEA) on normalized fecal metaproteomic data, considering those PULs that had at least five 865 encoded proteins whose abundances were altered in the same direction (adjusted p value < 0.05, unpaired  (Figure 2 ¾ figure supplement 1). 872 Analyzing PULs as gene sets revealed responses that were shared between the three supplements 873 and responses that were unique. For B. thetaioatomicron VPI-5482, PUL7 was the most highly 874 upregulated PUL with unfractionated pea fiber or purified PFABN; it contains multiple glycoside 875 hydrolase (GH) family 43 and 51 members with reported arabinofuranosidase activity (Figure 2 ¾  876   figure supplement 1). During supplementation with SBABN, B. thetaioatomicron VPI-5482 PUL75 was 877 highly upregulated; it specifies multiple GH enzymes, polysaccharide lyases (PL) and carbohydrate 878 esterases (CE) involved in depolymerizing the repeating disaccharide RGI backbone (a-(1,2)-L-879 rhamnose-a-(1,4)-D-galacturonic acid) (Figure 2 ¾ figure supplement 1) (Luis et al., 2018) (Figure 2 ¾ figure supplement 1). B. ovatus PUL96 encodes 893 multiple genes for degrading homogalacturonan (PLs and carbohydrate esterases) and is transcriptionally 894 upregulated during in vitro growth on homogalacturonan (Martens et al., 2011). These observations 895 illustrate how human gut Bacteroides exhibit great sensitivity to diet supplement source and structure.  files 2-4). 951 The online version of this article includes the following figure supplement(s) for MFABs prior to and after their introduction into gnotobiotic mice (related to Figures 4 and 5).