Data anlalysis for the project hambiEvoEnvCoexist
Shane L Hogle 
Chuliang Song
Microbiology, Community assembly, Evolution, Antimicrobial resistance (AMR)
1 Manuscript:
1.1 Published record
TBD
1.2 Preprint
TBD
2 Introduction
Past work has shown that the composition of small bacterial communities can be predicted remarkably well from outcomes of competitions between small sets of species (Friedman, Higgins, and Gore 2017). Other work has questioned whether such simple qualitative assembly rules hold in more complex situations such as when the species pool shares an evolutionary history (Chang et al. 2023) (also see this blog post). If community assembly dynamics can truly be characterized by additive pairwise processes this would be a great advantage for modelers. It would mean that communities could be described sufficiently using paired species interactions (as in traditional Lotka Volterra). However, if higher-order interactions (Levine et al. 2017) are common and important (e.g., species C mediates the interaction between species A and B in a density independent way) then the number of parameters necessary to sufficiently model the system quickly blows up quickly making the entire problem intractable.
Understanding the fundamental processes in community assembly will be key to designing microbial communities with desirable properties for human use and to anticipate the response of communities in nature to change. Two theoretical camps have emerged in contemporary theory; one arguing for pairwise assembly rules and another arguing for the importance of networks of chained pairwise interactions and/or higher order interactions. Evidence exists for both processes occurring in nature. Thus, the question of when/how/why community assembly can be sufficiently described/predicted from additive pairwise processes is relevant and important for gaining a predictive understanding of microbiome dynamics. For example, when will the assembly process emerge as a property of the community, rather than being a straightforward additive process? In particular, do the previous evolutionary history and patterns of coexistence in the species pool determine emergent coexistence? What is the role for different growth conditions (substrate availability, antibacterial compounds, pH, predator presence)? Or does prior adaptive evolution to the environment change the outcomes? Do the “simple” rules governing community assembly fundamentally change as you add more layers of complexity?
The goal here is to shed some additional light on the question of the importance of pairwise competition/emergent coexistence in microbiomes. Our experiment assembles all possible 2, 3, and 4 species sub communities from a 4-species pool under a gradient of antibiotic concentrations and with different evolutionary histories of the four species. We assess the predictability of community assembly outcomes based on subsets of the reduced species assemblies and attempt to understand whether assembly operates in a predominantly pairwise way.
3 Experiment overview
3.1 Background:
In prior work from our lab we experimentally evolved all species from the HAMBI synthetic community to increasing concentrations of the aminoglycoside antibiotic streptomycin. Briefly, we would grow each species at the highest streptomycin concentration it could tolerate for a number of growth cycles, then we would transfer into a higher streptomycin. We did this until we reached a streptomycin concentration where we could no longer detect growth. Then these cultures were plated and a single clone was isolated.
These clones are the source “evolved” starting materials for this experiment, where their clonal ancestor is the “ancestral” starting material.
3.2 Traits and physiology of species:
The ancestral and evolved form of each species were grown for 48 hours at 30℃ in 100% R2A growth medium (Reasoner and Geldreich 1985) at concentrations of 0, 4, 8, 16, 32, 64, 128, 256, 512, 1024, 2048, and 4096 μg/ml streptomycin. Growth was followed using optical density at 600 nm in a Bioscreen plate reader. Growth rates and carrying capacities were determined from the high resolution growth curves using AMiGA.
3.3 Species interactions:
The ancestral and evolved forms of each species were grown for 48 hours at 30℃ in 100% R2A growth medium (Reasoner and Geldreich 1985). Cultures were filtered to remove bacterial biomass and the spent media was stored at 4℃ until use. Next the evolved and ancestral forms of each species were grown in the spent medium of each other evo/anc species for 48 hours at 30℃ and growth was followed using optical density at 600 nm in a Bioscreen plate reader.
For the Burkholder diffusion assay each species/evolutionary form was added to a soft agar on a petri dish. Then a small volume of each species was pipetted directly onto the agar and plates and checked for zones of inhibition.
3.4 Competition/coexistence outcomes (species proportions from communities):
Here we wish to ensure that any equilibrium coexistence or competitive exclusion between species is not multistable - meaning that the outcome is deterministic and truly based on the density-independent competitive abilities of the species and not dependent on their starting density. The multistability problem has been addressed in other studies by growing each species from each community at different starting densities. This is typically done by starting all of the rare species at a low density and one abundant species at a higher density (as measured by optical density). For more than 2 species you cannot reasonably achieve all possible permutations of different densities, but you can start each species from rare, while leaving one remaining species high.
Here we use starting conditions with 10%/90% for pairs, 80%/10%/10% for trios, and 70%/10%/10%/10% for quartets. For pairs, we alternated so that each species started from rare. For trios and quartets, we alternated so that each species was allowed to invade from rare but we did not use all possible staring proportions due to practical and cost limitations. Another important consideration is that we can not distinguish between the ancestral and evolved forms of the same species with amplicon sequencing (or with plate counts) so that our experimental arrangement cannot be fully factorial (e.g., we cannot compete ancestral 1977 against evolved 1977) and thus there are fewer combinations that one would expect from expected scaling of \(2^{(8-1)} = 128\)
For the coexistence/exclusion experiments experimental communities are inoculated in 500 μl of 100% R2A in deep 96 well plates at a 50 fold dilution at four different streptomycin levels: 0, 16, 64, and 256 μg/ml. Plates were incubated in the dark at 30℃ with shaking at 800 RPM. Serial transfers (50 fold dilution; 10 μl to 490 μl) were conducted every 48 hours for a total of eight growth cycles. For every growth cycle \(\mathrm{log_2(50)} \approx 5.6\) generations will have elapsed for a total of about 45 generations over eight cycles.
4 Measurements and data types
4.1 Traits and physiology of species:
- Growth rates and carrying capacity (in optical density units) of each species (both ancestral/evolved) on a gradient of streptomycin concentrations
4.2 Species interactions:
- Growth rates and carrying capacity (in optical density units) of each species (both ancestral/evolved) on the filtrate of all other species.
- Burkholder Diffusion Assays for species pairs to detect presence of interference competition (e.g., antibiotic or toxin production that kills competitors). These results show that no species/evolutionary form produced zones of inhibition for any other species/form. This suggests that there will not be significant interference competition in this experiment.
4.3 Competition/coexistence outcomes (species proportions from communities):
- Community DNA from pairs, trios, and quartets was sampled after eight 48 hour growth cycles
- 16S rRNA amplicon sequencing was performed on samples to get the proportion of each species
- Optical density was collected at every other growth cycle
5 Availability
Data and code in this GitHub repository (https://github.com/slhogle/hambiEvoEnvCoexist) is provided under GNU AGPL3. The rendered project site is available at https://slhogle.github.io/hambiEvoEnvCoexist/, which has been produced using Quarto notebooks. The content on the rendered site is released under the CC BY 4.0. This repository hosts all code and data for this project including the code necessary to fully recreate the rendered webpage.
An archived release of the code here is available from Zenodo:
Raw sequencing data using in the project is available from NCBI Bioproject .
6 Reproducibility
The project uses renv to create reproducible environment to execute the code in this project. See here for a brief overview on collaboration and reproduction of the entire project. To get up and running you can do: