Systems-level metabolism of the altered Schaedler flora, a complete gut microbiota

The altered Schaedler flora (ASF) is a model microbial community with both in vivo and in vitro relevance. Here we provide the first characterization of the ASF community in vitro, independent of a murine host. We compared the functional genetic content of the ASF to wild murine metagenomes and found that the ASF functionally represents wild microbiomes better than random consortia of similar taxonomic composition. We developed a chemically defined medium that supported growth of seven of the eight ASF members. To elucidate the metabolic capabilities of these ASF species—including potential for interactions such as cross-feeding—we performed a spent media screen and analyzed the results through dynamic growth measurements and non-targeted metabolic profiling. We found that cross-feeding is relatively rare (32 of 3570 possible cases), but is enriched between Clostridium ASF356 and Parabacteroides ASF519. We identified many cases of emergent metabolism (856 of 3570 possible cases). These data will inform efforts to understand ASF dynamics and spatial distribution in vivo, to design pre- and probiotics that modulate relative abundances of ASF members, and will be essential for validating computational models of ASF metabolism. Well-characterized, experimentally tractable microbial communities enable research that can translate into more effective microbiome-targeted therapies to improve human health.

colonies with a convex shape. A slight yellow pigmentation was observed in a few Parabacteroides ASF519 colonies. Mucispirillum ASF457 grown on supplemented BHI forms a translucent, uneven film. Here, Mucispirillum ASF457 was imaged using phase contrast to highlight colony edges. Scale bar represents 1 mm. Figure 7: Correlation of NOG distribution and metabolite consumption/production patterns in fresh media. The metabolites are represented in the columns, and the NOGs are represented in the rows (NOG identifiers are omitted due to space constraints). This heat map displays the Spearman correlation between NOG presence and metabolite production/consumption (see Supplemental Materials and Methods). Consumption and production were defined as ±2 standard deviations below or above the relative abundance in fresh media, respectively. The Spearman correlation was calculated between all pairs of 2 202 metabolic NOGs in the ASF with each of 73 metabolites. After Bonferroni correction, 11 079 of the 160 746 correlations (6.9%) were statistically significant (p-value < 3.1x10 -7 ). However, removing unique NOGs and metabolites which were uniquely consumed or produced by a single species reduced the significant correlations to 458 (0.2%). Red shading indicates a positive correlation (presence of a NOG coincides with production of a metabolite) while blue shading indicates a negative correlation (presence of a NOG coincides with consumption of a metabolite). Figure 8: Summary ASF interaction network. The interaction strength between each pair of species was calculated as fractional inhibition (see Figure 4A). For example, consider "AUC_alone" to be the area under the growth curve (AUC) of Lactobacillus 360 grown alone, and "AUC_356" to be the AUC for Lactobacillus 360 grown in spent356. The influence of Clostridium 356 on Lactobacillus 360 was calculated as (AUC_356 -AUC_alone) / AUC_alone. This results in a value close to -1 because Lactobacillus 360 is completely inhibited by spent media from Clostridium 356. Strong inhibitory relationships (i.e. -1) are shown by wide, faint, blue lines. Weaker relationships are shown by thin lines in darker color. Most relationships were very inhibitory under these experimental conditions, including all self-edges. Two relationships were weakly growth-promoting (thin, red lines). Table 1. NMR detection of expected compounds. Based on previous work defining the composition of LB, we composed a list of compounds expected to be present in the supplemented LB media. We gathered 1 H NMR spectra for fresh media samples and attempted to confirm the presence of these expected compounds. A subset of components expected to be in the media do not contain hydrogen bonds and are not detectable by 1 H NMR (such as Ca+, Cl-, Fe 2 +, etc.). The majority of expected compounds are clearly detectable (30/44), while several are on the border of the detectable limit (6/44). A small group of compounds were not detected (8/44) either because they were at a concentration below the detectable limit, or a reference spectrum was not available.

NMR Detection Adenosine
Detected AMP Very low / border of detectable limit CMP Detected Cytosine Very low / border of detectable limit Deoxyadenosine Very low / border of detectable limit Deoxycytidine Difficult to identify this component No standard or reference spectra Heme No standard or reference spectra Very low / border of detectable limit Not detected-below detectable limit Riboflavin Very low / border of detectable limit Shikimate Not detected-below detectable limit Thiamine phosphate No standard or reference spectra Thymidine Not detected-below detectable limit UMP Very low / border of detectable limit Uracil Detected Uridine Detected Supplemental Table 2. pH of ASF spent media. Each ASF member was grown in supplemented LB and the pH was measured using an Accumet pH meter.

Scanning Electron Microscopy
All ASF members excluding Mucispirillum ASF457 were cultured in BHI liquid medium in a 6-well plate containing an SEM stub coverslip. When growth was visible, cells were fixed for 30 min. with glutaraldehyde (2% by volume). Wells were rinsed 3 times for 5 min. each with 1xPBS. Samples were subjected to a graded ethanol dehydration, 10 min. each in 30, 50, 70, 80, 90, 100, 100% ethanol in water.
The coverslips were further dehydrated for 10 min. in hexamethyldisilazane (HDMS) (Sigma). Finally, the coverslips were stuck to SEM stubs using the Phenom starter kit (Ted Pella, Redding, CA, USA) and sputter coated with gold using a SCD005 sputter coater (Bal-tec, Los Angeles, CA, USA). The final samples were imaged using a Sigma VP HD Field-emission SEM (Zeiss, Pleasanton, CA, USA).

Determining Colony Morphology
Agar plates were prepared using the supplemented LB medium described above (1.2% agar) except for MucispirillumASF457 which was grown on supplemented BHI (1.2% agar). ASF members were streaked to single colonies and grown for 70 hours (h) 37°C in the anaerobic chamber. Images were obtained using an EVOS XL digital inverted microscope (ThermoFisher Scientific, Waltham, MA, USA) with 2X objective.

Genomic Analysis and Comparison with Wild Murine Microbiota
Metagenomic data from the feces of 15 wild mice in a previous study (Wang et al. 2014) was used as a reference data set. Gene calls, performed using FragGeneScan (Rho et al. 2010), were downloaded from MG-RAST (Meyer et al. 2008) for the 15 mice specified at the W0 time point in Table S5 of (Wang et al. 2014) (MG-RAST project URL: http://metagenomics.anl.gov/?page=MetagenomeProject&project=5130).
Gene calls were annotated with HMMER Version 3.1b2 (Eddy 1998 metagenomic samples. NOGs containing functional annotations in more than one category were discarded during all portions of analysis (representing <1% of total annotations in any sample).
Media was equilibrated overnight in the anaerobic chamber before inoculation with ASF members. Vitamin K (0.5%): Mix 100µL into 19.9mL 95% ethanol. Store at -20°C. Add solution to fresh media before use -1 µL/10 mL base.
Sterilize media (20 minutes on liquid cycle) and put into anaerobic chamber when cool. Before experiments, make fresh complete media by adding hemin, Vitamin K, lactose, and Tween-20 in the anaerobic chamber.

Growth Measurements
Growth curves were obtained for ASF members in the anaerobic chamber using four small plate readers measuring optical density at 870 nm (Jensen et al. 2015). The plate readers measure OD at 870 nm because this wavelength is not absorbed by common bacterial pigments (Jensen et al. 2015). Furthermore, because each LED pair is independent, the overall growth curve needs to be normalized to a common standard. Overnight liquid cultures of 10 ml were prepared for each ASF member: The entire volume of the overnight cultures were centrifuged at 8 000 rpm for 2 min. and the resulting pellets were resuspended in fresh liquid medium to produce a dense suspension of 0.75 ml.

Determining Substrate Utilization and Byproduct Consumption with NMR Spectroscopy
Media (fresh or spent) samples of 2 ml were filter sterilized (0.22 μm pore size) and frozen at -80°C.
The loadings of pairwise principal component analysis models, comparing blank media with the spent media of each bacteria species, were used to identify metabolites generated or consumed in each experiment. The relevant regions of the spectra were integrated to calculate relative spectral intensities for each metabolite. Relative intensities in spent and double spent media were converted to z-scores with respect to metabolite abundances in fresh media. We defined significant abundance changes as those of magnitude greater than ±2 standard deviations from zero (zero being the metabolite abundance in fresh media). The peak integral data and associated R code for analysis and visualization is available in the online repository.

Classifying Instances of Emergent Metabolism
In this study, we define cases of "emergent metabolism" as any metabolic behavior which changes in the presence of another species. This broad definition can be divided into 6 categories: 1. ASF species only produced metabolite x when grown in fresh media 2. ASF species only consumed metabolite x when grown in spent media 3. ASF species switched from producing metabolite x (when grown in fresh media) to consuming metabolite x when grown in spent media 4. ASF species produced metabolite x only when grown in spent media 5. ASF species only consumed metabolite x when grown in fresh media 6. ASF species switched from consuming metabolite x (when grown in fresh media) to producing metabolite x when grown in spent media We define "rare emergent metabolites" for each ASF species as metabolites which display emergent behavior (any of the 6 categories) in a single condition.
We identified all cases of emergent metabolism by comparing the metabolomics data from each single spent media sample to the double spent media samples for the same species. For example, when Clostridium ASF356 is grown in fresh media, it does not produce or consume methionine. When Lactobacillus ASF360 is grown in fresh media, is also does not produce or consume methionine.
However, when Clostridium ASF356 is grown in the spent media from Lactobacillus ASF360, it does produce methionine. This would be an example of category 4, where Clostridium ASF356 produced methionine only when grown in the spent media from another species. We identified and categorized all cases of emergent metabolism using a custom R script, which is available in an open online repository.

Correlation of NOG Presence/Absence and Metabolite Consumption/Production
NOG presence was represented as a binary vector with a 1 or 0 indicating the presence or absence of that NOG in each of the seven ASF species which grew. Metabolite changes were represented as a vector with elements for each species indicating consumption (-1), no change (0), or production (1) when that species was grown in fresh media. Consumption and production were defined as -2 or +2 standard deviations from the relative abundance in fresh media, respectively. We excluded 60 metabolic NOGs which were present in all ASF members and 12 metabolites which held the same value in all spent media samples, because correlations are undefined if variance of either variable is zero. The Spearman correlation was calculated between all pairs of 2 202 metabolic NOGs in the ASF with each of 85 metabolites.

Code and Data Availability
Our data and analysis scripts are available at the following repository: http://mbi2gs.github.io/asf_characterization/.
Some large analysis output files and annotation files for metagenomic data are excluded due to file hosting size limitations, but are available upon request from the authors or can be generated using the indicated raw data, HMMer, associated eggNOG files, and scripts in the repository.

Supplemental Metabolomics Plots:
For all experimental conditions, we plotted the relative changes in NMR peak integral (z-score with respect fresh media) of each known metabolite in each individual replicate. Gray points with black boxes originate from the main data set. Red points and boxes originate from a second, independent set of biological replicates. The z-scores for the second data set were calculated using the standard deviations from the first data set, to facilitate comparisons between the two. Plots are labeled as "ASF-Species-in-Media-ID". There are eight plots for each species (one resulting from growth in fresh media, one from growth in its own spent media, and six resulting from growth in the spent media of the other ASF members).