The transformative tools and technologies discussed below are areas where advancements are necessary and will have a significant impact across the technical themes in microbiome engineering.
Understanding microbe-microbe interactions is critical to predict and control microbiome function. These interactions can be incredibly rich and encompass physical linkages, metabolite/nutrient exchange, cell-to-cell signalling, electron/cofactor transfer, and secretion of antimicrobial compounds, over both short and long length scales. The ability to control these interactions will be fundamental to advancing microbiome engineering as they help direct when a cell grows or dies, regulate the metabolites or proteins a cell produces, and dictate where a cell colonizes in the environment. However, the nature and existence of these interactions are poorly understood for all but a few pairs of microbes. Fully engineering a microbiome will require controlling the interactions between cells of the same species, and between different microbial species, over long periods of time and distances.
Several broad aspects of engineering biology research will need to be advanced to achieve fully-programmable cellular interactions in an engineered microbiome. First, highly parallelizable or high throughput technologies that can tease out the positive and negative interactions within the thousands of taxons within a microbiome are needed to establish rules for community design. Emerging advances in microfluidics, microwell design, emulsion nano-scale reactions, high resolution imaging, and single cell -omics now promise to interrogate the rich network of interactions that drive community function in thousands of permutations of microbial consortia simultaneously.1Hansen, R. H., Timm, A. C., Timm, C. M., Bible, A. N., Morrell-Falvey, J. L., Pelletier, D. A., Simpson, M. L., Doktycz, M. J., & Retterer, S. T. (2016). Stochastic Assembly of Bacteria in Microwell Arrays Reveals the Importance of Confinement in Community Development. PLOS ONE, 11(5), e0155080. View Publication,2Kehe, J., Kulesa, A., Ortiz, A., Ackerman, C. M., Thakku, S. G., Sellers, D., Kuehn, S., Gore, J., Friedman, J., & Blainey, P. C. (2019). Massively parallel screening of synthetic microbial communities. Proceedings of the National Academy of Sciences, 116(26), 12804–12809. View Publication,3Lambert, B. S., Raina, J.-B., Fernandez, V. I., Rinke, C., Siboni, N., Rubino, F., Hugenholtz, P., Tyson, G. W., Seymour, J. R., & Stocker, R. (2017). A microfluidics-based in situ chemotaxis assay to study the behaviour of aquatic microbial communities. Nature Microbiology, 2(10), 1344–1349. View Publication,4Massalha, H., Korenblum, E., Malitsky, S., Shapiro, O. H., & Aharoni, A. (2017). Live imaging of root–bacteria interactions in a microfluidics setup. Proceedings of the National Academy of Sciences, 114(17), 4549–4554. View Publication,5Niepa, T. H. R., Hou, L., Jiang, H., Goulian, M., Koo, H., Stebe, K. J., & Lee, D. (2016). Microbial Nanoculture as an Artificial Microniche. Scientific Reports, 6(1), 30578. View Publication These systems must tease out not only binary pairwise interactions but the higher-order emergent interactions of n-member microbial communities required for community resilience and function. Quantification of higher-order interactions is a key challenge for microbiome engineering that is largely unresolved.
At their most basic, microbe-microbe interactions allow for information and energy to be transferred from one cell to another. This can be achieved through different means, such as a signaling molecule that is synthesized and secreted by the host organism, or one or more transmembrane proteins or polymers that interact with the same on another cell. However, additional interactions between a microbiome and its environment, competition for nutrients, and signal molecules acting at a distance are further factors that must be considered when engineering cell-to-cell signaling.6Fischbach, M. A., & Segre, J. A. (2016). Signaling in Host-Associated Microbial Communities. Cell, 164(6), 1288–1300. View Publication,7Foo, J. L., Ling, H., Lee, Y. S., & Chang, M. W. (2017). Microbiome engineering: Current applications and its future. Biotechnology Journal, 12(3), 1600099. View Publication While achieving functional biodiversity, spatiotemporal control, and distributed metabolism will require advances in cell-to-cell signaling technologies, cell-to-cell signaling plays a unique role in each one. Engineering functionally and taxonomically diverse microbiomes will require developing interspecies signaling networks that have multiple levels of redundancy to be robust over time. Achieving the necessary specificity and stability across species will likely require building novel orthogonal chemical and genetic signaling pathways due to the challenges that have arisen when engineering endogenous pathways.8Kapp, G. T., Liu, S., Stein, A., Wong, D. T., Remenyi, A., Yeh, B. J., Fraser, J. S., Taunton, J., Lim, W. A., & Kortemme, T. (2012). Control of protein signaling using a computationally designed GTPase/GEF orthogonal pair. Proceedings of the National Academy of Sciences, 109(14), 5277–5282. View Publication,9Mandell, D. J., & Kortemme, T. (2009). Computer-aided design of functional protein interactions. Nature Chemical Biology, 5(11), 797–807. View Publication,10Pai, A., Tanouchi, Y., Collins, C. H., & You, L. (2009). Engineering multicellular systems by cell–cell communication. Current Opinion in Biotechnology, 20(4), 461–470. View Publication,11Young, M., Dahoun, T., Sokrat, B., Arber, C., Chen, K. M., Bouvier, M., & Barth, P. (2018). Computational design of orthogonal membrane receptor-effector switches for rewiring signaling pathways. Proceedings of the National Academy of Sciences, 115(27), 7051–7056. View Publication Further work will be needed to understand the role of microbe-microbe communication over long distances, on the range of millimeter distances or more, and how those interactions change over time.
Fully detailing the mechanisms of microbe-microbe interactions within a microbiome will require significant advances in experimental microbiome models. Current models for microbiome engineering are still in their infancy. As is the case with every model system, researchers are forced to choose between the system’s relative complexity and accuracy, versus the ease of use for addressing a scientific question.12Chevrette, M. G., Bratburd, J. R., Currie, C. R., & Stubbendieck, R. M. (2019). Experimental Microbiomes: Models Not to Scale. MSystems, 4(4), e00175-19, /msystems/4/4/mSys.00175-19.atom. View Publication Coculture systems (e.g., Bacillus subtilis and Streptomyces spp. to study secondary metabolites) are easy to create and manipulate, but offer relatively little insight into the actual environmental pressures these bacteria might naturally encounter. On the other side, microbiome enrichment methods may be more representative of a microbiome, but offer relatively little opportunity to perturb or manipulate the system.13Zegeye, E. K., Brislawn, C. J., Farris, Y., Fansler, S. J., Hofmockel, K. S., Jansson, J. K., Wright, A. T., Graham, E. B., Naylor, D., McClure, R. S., & Bernstein, H. C. (2019). Selection, Succession, and Stabilization of Soil Microbial Consortia. MSystems, 4(4), 13. View Publication Experimental model development should occur from both directions, such that models currently focusing on molecular mechanisms can increase in complexity to better reflect natural microbiomes, and ecologically complex models can be controlled more easily. Experimental models should also factor in the environmental surroundings of natural microbiomes as they are improved (e.g., replicate the immune system’s influence on human, animal, or plant microbiomes). Computational models of increasing accuracy will depend, and be supported by, more complex and controllable experimental models to feed data back into the model, driving future development.
Generally speaking, we need to improve our ability to model and determine the composition and stability of a microbial community over time in a 3D environment.14Wu, F., Lopatkin, A. J., Needs, D. A., Lee, C. T., Mukherjee, S., & You, L. (2019). A unifying framework for interpreting and predicting mutualistic systems. Nature Communications, 10(1), 242. View Publication Without a robust experimental and computational modeling effort, the prediction and control of temporal and spatial interactions in microbial communities will be limited. Several distinct areas of research will need to be advanced to enable improved microbiome model systems. Microbiome measurements and metrology must be improved, methods for manipulating microbes within a microbiome must be developed, and better predictive models must be designed. Improved models will have an important role for both capturing natural interactions and providing a means of designing engineered microbiomes for new functions. Particularly for complex or hard-to-access microbiomes, such as those in the soil or human gut, better models will capture more of their natural complexity, making it easier to engineer microbiomes with a reasonable expectation that they will work once added to a natural environment.
Advances in the three areas detailed below will be critical for realizing the full potential of engineered microbiomes. Increased speed and ease of experimental measurements or manipulations will be the starting point for improved computational models. Those computational models can in turn be used to perform better-informed experiments, and develop predictive frameworks that will be necessary for engineering spatiotemporal dynamics of a microbiome, distributing metabolism between species, and generating the biodiversity needed for engineered microbiomes to function in any desired environment.
Several existing technologies will need to be improved or further developed to facilitate experimental models. Not only will -omics techniques (e.g., genomics, transcriptomics, metabolomics, or proteomics) need to continue decreasing in cost to allow for increased temporal resolution, they will need to become more parallelizable. Significant advances in technologies will be needed, or orthogonal approaches will need to be developed, to measure microbiomes in their native environments, because these measurements are crucial for benchmarking the utility of experimental or computational models against natural microbiomes. Additional novel techniques such as stable isotope probing, non-canonical amino acid tagging, and various meta -omics methods will also contribute to a better understanding of microbiomes.15Gebreselassie, N. A., & Antoniewicz, M. R. (2015). 13C-metabolic flux analysis of co-cultures: A novel approach. Metabolic Engineering, 31, 132–139. View Publication,16Lawson, C. E., Harcombe, W. R., Hatzenpichler, R., Lindemann, S. R., Löffler, F. E., O’Malley, M. A., García Martín, H., Pfleger, B. F., Raskin, L., Venturelli, O. S., Weissbrodt, D. G., Noguera, D. R., & McMahon, K. D. (2019). Common principles and best practices for engineering microbiomes. Nature Reviews Microbiology, 17(12), 725–741. View Publication Simultaneous measurements of parameters such as transcription, translation, and metabolism will be critical for reducing experimental error and generating robust data sets used to build new experimental and computational models.
Furthermore, improved microbiome engineering will require these technologies to develop increased spatial resolution, as the three-dimensional structure of a microbiome (i.e., both the percentages of individual species and their distribution in space) is critical for a microbial community’s function. New imaging techniques are beginning to address this challenge by combining microscopy, imaging mass spectrometry, and other techniques, but there are still severe experimental bottlenecks that make it difficult to increase the number of measurements and scale-down measurements to micron-sized voxels.
As the quantity of data generated increases, further challenges arise to ensure that the data can be used to its fullest potential. The concept of FAIR (Findable, Accessible, Interoperable, and Reusable) data should be applied as much as possible, both to existing data sets and any future pieces of scientific infrastructure.17Wilkinson, M. D., Dumontier, M., Aalbersberg, Ij. J., Appleton, G., Axton, M., Baak, A., Blomberg, N., Boiten, J.-W., da Silva Santos, L. B., Bourne, P. E., Bouwman, J., Brookes, A. J., Clark, T., Crosas, M., Dillo, I., Dumon, O., Edmunds, S., Evelo, C. T., Finkers, R., … Mons, B. (2016). The FAIR Guiding Principles for scientific data management and stewardship. Scientific Data, 3(1), 160018. View Publication FAIR data will also facilitate faster development of software tools and application of machine learning techniques to advance computational models. In addition to the existing recommendations for FAIR data, it will be valuable to capture the number of samples taken at any given time and the intervals at which they are taken. In combination with further developments in analysis and modeling methods that can accept non-uniform timepoints, these data will allow for richer analysis of microbiomes at specific times when they may be more dynamic.
A key aspect of useful experimental models is how easily or quickly they can be manipulated. For microbiome engineering, many novel approaches will need to be developed to control parameters that stretch across scales. At an organism-level, new techniques must be developed to culture, and then genetically or transcriptionally manipulate, a single microbial species growing within a community.18Brophy, J. A. N., Triassi, A. J., Adams, B. L., Renberg, R. L., Stratis-Cullum, D. N., Grossman, A. D., & Voigt, C. A. (2018). Engineered integrative and conjugative elements for efficient and inducible DNA transfer to undomesticated bacteria. Nature Microbiology, 3(9), 1043–1053. View Publication,19Wang, G., Zhao, Z., Ke, J., Engel, Y., Shi, Y.-M., Robinson, D., Bingol, K., Zhang, Z., Bowen, B., Louie, K., Wang, B., Evans, R., Miyamoto, Y., Cheng, K., Kosina, S., De Raad, M., Silva, L., Luhrs, A., Lubbe, A., … Yoshikuni, Y. (2019). CRAGE enables rapid activation of biosynthetic gene clusters in undomesticated bacteria. Nature Microbiology, 4(12), 2498–2510. View Publication Alongside this effort is a need to better understand and annotate gene functions across microbiomes, as a functional understanding is key for targeted engineering approaches. At larger scales, advancements will be needed to control intra- and inter-species cell signaling (as discussed in “Engineering Microbe-Microbe Interactions”), leading to altered metabolism, growth, and spatial distributions.
This will also need to be paired with techniques to spatially control microbes within a microbiome. Many existing techniques either create a homogeneous mixture of microbes within the model system, or enrich existing communities with little to no ability to control the locations of the microbial cells that grow up. Ultimately, a priori knowledge of microbiome growth will mean that physical manipulation becomes unnecessary, but the process of generating that knowledge will be greatly assisted by improved manipulation techniques.
Computational modeling will be a central tool for engineering microbial interactions and predicting the necessary interaction networks within a community. Models will be helpful for determining how many new pathways will need to be added to achieve a desired outcome in a community. New computational advances will also be needed to determine the minimum network needed to execute a community’s function. Minimum networks will be crucial for reducing size of engineered pathways, which is itself important for long-term stability and maintenance of function. Generally, we have a poor understanding of how to represent microbial interactions and at what scale of interaction. Thus, the prediction capabilities of microbial models are limited20Wang, S., Fan, K., Luo, N., Cao, Y., Wu, F., Zhang, C., Heller, K. A., & You, L. (2019). Massive computational acceleration by using neural networks to emulate mechanism-based biological models. Nature Communications, 10(1), 4354. View Publication and under-informed by experimental data measurements. As experimental models advance, more data will become available to help generate better computational models with increased predictive power.
A central challenge will be understanding the relevant principles that govern interactions and growth within multispecies communities. This will require finding the appropriate level of abstraction to deal with the combinatorial complexity of multi-species interactions, using mechanistic-based modeling or machine learning approaches. It will also require improved models that can analyze chemical interactions and transport within a microbiome in four dimensions, and account for additional boundary layer effects where a microbiome interacts with the environment.
Additionally, mathematical frameworks and computational tools will need to be developed to capture the temporal and spatial dynamics of microbial communities at different scales for the purpose of designing their composition. This will require both top-down and bottom-up modeling efforts, because each approach individually suffers from distinct drawbacks. Top-down modeling efforts often capture thousands of species and a very simplified picture of their interactions (e.g., commensalism is captured only by one or two parameters that have little biological meaning since they are aggregates of many biological quantities). In contrast, bottom-up models21Gorochowski, T. E., Matyjaszkiewicz, A., Todd, T., Oak, N., Kowalska, K., Reid, S., Tsaneva-Atanasova, K. T., Savery, N. J., Grierson, C. S., & di Bernardo, M. (2012). BSim: An Agent-Based Tool for Modeling Bacterial Populations in Systems and Synthetic Biology. PLoS ONE, 7(8), e42790. View Publication,22Heirendt, L., Arreckx, S., Pfau, T., Mendoza, S. N., Richelle, A., Heinken, A., Haraldsdóttir, H. S., Wachowiak, J., Keating, S. M., Vlasov, V., Magnusdóttir, S., Ng, C. Y., Preciat, G., Žagare, A., Chan, S. H. J., Aurich, M. K., Clancy, C. M., Modamio, J., Sauls, J. T., … Fleming, R. M. T. (2019). Creation and analysis of biochemical constraint-based models using the COBRA Toolbox v.3.0. Nature Protocols, 14(3), 639–702. View Publication,23Jang, S. S., Oishi, K. T., Egbert, R. G., & Klavins, E. (2012). Specification and Simulation of Synthetic Multicelled Behaviors. ACS Synthetic Biology, 1(8), 365–374. View Publication,24Klavins, E. (2014). Lightening the load in synthetic biology. Nature Biotechnology, 32(12), 1198–1200. View Publication,25Matyjaszkiewicz, A., Fiore, G., Annunziata, F., Grierson, C. S., Savery, N. J., Marucci, L., & di Bernardo, M. (2017). BSim 2.0: An Advanced Agent-Based Cell Simulator. ACS Synthetic Biology, 6(10), 1969–1972. View Publication,26Oishi, K., & Klavins, E. (2014). Framework for Engineering Finite State Machines in Gene Regulatory Networks. ACS Synthetic Biology, 3(9), 652–665. View Publication are computationally expensive and the parameters used are rarely measured.27Green, L. N., Hu, C. Y., Ren, X., & Murray, R. M. (2019). Bacterial Controller Aided Wound Healing: A Case Study in Dynamical Population Controller Design [Preprint]. Synthetic Biology. View Publication,28Ren, X., & Murray, R. M. (2018). Role of interaction network topology in controlling microbial population in consortia. 2018 IEEE Conference on Decision and Control (CDC), 2691–2697. View Publication Regardless of approach, models must be capable of simulating microbiomes at time-scales that have practical value (e.g., hours to weeks) due to the temporal variations that microbiomes undergo,29Bogart, E., Creswell, R., & Gerber, G. K. (2019). MITRE: Inferring features from microbiota time-series data linked to host status. Genome Biology, 20(1), 186. View Publication,30Bucci, V., Tzen, B., Li, N., Simmons, M., Tanoue, T., Bogart, E., Deng, L., Yeliseyev, V., Delaney, M. L., Liu, Q., Olle, B., Stein, R. R., Honda, K., Bry, L., & Gerber, G. K. (2016). MDSINE: Microbial Dynamical Systems INference Engine for microbiome time-series analyses. Genome Biology, 17(1), 121. View Publication which may require increased computational resources or new mathematical frameworks to reduce computational requirements. They also need to capture ecological and evolutionary timescales, and the feedback mechanisms between evolution and ecology that dictate spatial distributions and functions.
As models are developed, it will be important to consider how an output value could be translated back to experimental procedures. Black box models that produce parameters that are unknown, cannot be measured experimentally, or do not correspond to a determinable biochemical quantity, make it challenging to replicate a model’s findings at the bench or in a bioreactor. Increased collaborations between computational biologists and experimental biologists, as well as investments in cross-disciplinary training, will help ensure that computational models are built on a foundational knowledge of biology.