Where bacteria and eukaryotes meet

Led by Thorsten Mascher, a microbiologist and a speaker of the Deutsche Forschungsgemeinschaft-funded Priority Program “Emergent Functions of Bacterial Multicellularity,” and Pascal Silberzan, a physicist and a well-known expert in the field of wound healing in eukaryotic model systems, this discussion group highlighted key points in the analogy between bacterial and eukaryotic tissues. At the same time, it also exposed many open questions that are yet to be addressed to understand biofilm formation from a tissue-like perspective.

One of the first questions raised was “why do we need to compare bacteria with eukaryotes”? Most of the discussion panelists agreed that the value of this analogy lies in our ability to compare the knowledge about the processes that govern the formation of tissues in both study systems, allowing each system individually to then be reconsidered in light of what is known about the other. The discussion then revolved around the following three questions.

The discussion in this group was on bacterial extracellular matrix led by experts Matthew Chapman, Meytal Landau, and Regine Hengge. Their expertise covers both extracellular matrix proteins (Matthew, Meytal, and Regine) as well as polysaccharide components (Regine) from complementing directions, including structural, mechanistic, and regulatory aspects. This discussion centered around two major topics.

This discussion session was led by Nicolas Biais, an experimentalist specializing in mechanomicrobiology, and Stephan Klumpp, a theoretical physicist studying a broad set of topics, including modeling cell motility and transport by molecular motors. The discussion started with the realization of the need to specify the definition of mechanosensing (which is relevant for both eukaryotic and bacterial worlds). On the triggering side, there is a signal or a stimulus, but are these terms the same? A signal is often related to an intention to sense, but sensing may be non-specific, and it was therefore agreed that the term “stimulus” rather than “signal” may be more appropriate for the trigger of a sensation. A stimulus may be sensed actively or passively. Taking examples from humans, a passive reaction to a force stimulus is a reflex (e.g., when you hit a certain area on your knee and it jumps unwillingly). An active reaction to a force is related to intention, for example, the probing touch of a fruit before we decide to put it in our shopping bag. These examples from our everyday lives lead to the question: do bacteria have intentions to actively probe their environment or is mechanosensing by bacteria solely passive? An example of passive mechanosensing is magnetotactic bacteria that use a self-formed intracellular magnet to navigate by the magnetic field of the Earth, as was discussed in the talk by Stefan Klumpp (45). However, the adhesion of bacterial cells to each other or to other cells is probabilistic, and therefore, it may be claimed that mechanosensing by bacteria is not entirely passive, as was discussed in the talk by Vernita Gordon (46).

Mechanosensing thus is triggered by a mechanical stimulus, and, in order for this stimulus to lead to a response, it needs to be transmitted. Mechanosensing therefore involves mechanotransmission, which then leads to a certain activity, such as, for example, altered gene expression levels. Mechanosensation can also be distinguished based on what is sensed: interactions with the environment, or other cells, or changes occurring to the cell itself. Additionally, sensation can be performed by a single cell or by a group of cells. In the context of a single bacterial cell, the envelope of bacterial cells is composed of peptidoglycans and a membrane, which are the possible objects on which a stimulus can act. It is currently unclear how flexible the bacterial envelope is and to which extent it can respond to external stimuli, particularly compared to the eukaryotic cell. Here, the previously discussed analogy with eukaryotes (see the first two discussion group topics above) is more challenging: eukaryotic cells possess more advanced mechanosensing mechanisms both because mechanical stimuli may be sensed by the cytoskeleton in addition to the cell envelope and also due to specific mechanosensitive protein complexes and channels in the cell and nuclear membranes (47). That said, while bacterial and eukaryotic cells may not be all that similar in their mechanosensing potential and mechanisms, borrowing ideas for mechanosensing from eukaryotes may provide the basis for the terminology to describe mechanosensing in bacterial systems.

The question of immediate practical relevance is how to experimentally test mechanosensing in bacteria. A few plausible candidate processes to focus on are the triggers of turgor pressure and tension and their relationship to sensing elements of mechanosensitive channels (48), pili (49), and flagella (50). It would also be interesting to address whether bacteria respond to stimuli in a digital or analog manner. In a digital-type stimulus response, the sensor reacts above a certain threshold, whereas in an analog-type stimulus response, the reaction of the sensor is proportional to (and scales with the magnitude of) the stimulus. Similarly, it would be interesting to understand whether a bacterium can distinguish the “sign” of the stimulus, for example, discriminating “pushing” from “pulling” forces.

Sensing is also what happens when the immune cell encounters bacteria, whether at the molecular signaling or at the mechanical level. In the next discussion group, we contemplated biofilms in the context of host immune responses.

This focus session was chaired by Susanne Häußler (Head of the Department of Molecular Bacteriology at the Helmholtz Center for Infection Research), who is leading a unique global effort on DNA and RNA sequencing of clinical P. aeruginosa isolates, and Matthias Hannig, Head of the Department of Tooth Retention, Periodontology, and Preventive Dental Care of the Saarland University Hospital, whose research focuses on oral biofilms.

While much still remains to be learned about both the topics of biofilms and the host immune response, considering them in synergy might lead to important and practical discoveries that help treat microbial diseases. One emerging idea here is that the human-associated biofilms that exist in equilibrium with the immune system might be considered a constituent tissue or a part of the human organism. In the oral cavity, healthy, homeostatic, multi-species bacterial populations prevent the dominance of pathogenic (cariogenic) strains and serve as a protective barrier against infections (35); a similar situation occurs in the digestive system where gut bacteria not only work to exclude pathogens but also perform additional metabolic tasks (51). Overcoming the technical difficulty of separating bacterial and eukaryotic genomes and then performing DNA and RNA sequencing of both biofilms and host or patient tissue would lead to important correlative analyses of the preferred microbiome and possible crosstalk and interactions between hosts and their colonizing bacterial strains.

In the context of antimicrobial peptides and their relation to COVID-19 infection (discussed in a talk by Gerard Wong), the question of vaccination against caries was brought up. Indeed, anti-caries vaccines (in practice vaccination against Streptococcus mutans) were tested in animals as early as in the 1970s, together with local spray vaccines (as now also being tested for COVID); however, they were not deemed commercially interesting. A further challenge is that in a dysbiotic oral microbiome, often multiple bacterial strains are cariogenic; once again, this redirects us from thinking about the disease being caused by a single bacterium to instead recognizing that they may be multi-species and that neglecting apparent microbial bystanders might be a dangerous oversimplification. A related strategy to treat a disease is to impose homeostasis by introducing a certain probiotic strain of bacteria, which has also been implemented using non-pathogenic S. mutans mutants and other bacterial strains in the context of caries (52). For predicting how promising this replacement strategy is, it is critical to understand how bacteria “talk to each other” (see the key point below).

One very active effort in studying bacterial interactions with the immune system is in the area of anti-tumor bacterial therapies. This therapeutic approach builds on the natural abilities of bacteria to localize to certain tumors and uses genetically engineered bacterial strains that produce location-specific toxins. The novel twist is the usage of bacteria in the tumor environment to activate or modulate the immune system when the tumor cells are evading the immune response. Here, the importance of extracellular matrix modification induced by tumor cells (53) may have an interesting analogy with matrix production and regulation by bacterial biofilms (following the ideas of the second discussion topic above). Another blind spot of the immune system is the surfaces of medical implants, where it cannot reach a biofilm, emphasizing the importance of biofilm formation at biotic-abiotic interfaces. Several key technologies were mentioned during the discussion that address the aforementioned challenges, including experiments using germ-free (microbiome-free) mice that, together with rapidly evolving omics approaches, permit a high level of control on which bacterial species are present and provide the most natural environment for disease models. As was frequently highlighted, the interdisciplinary effort of integrating medical and basic research experts is critical in solving questions in this area of research.

The leaders of this session are both experts in bacterial interactions, with Michael Meijler (Ben-Gurion University of the Negev) representing the chemistry perspective and Rachel Neve (postdoctoral researcher in Elizabeth Shank’s group) highlighting the biological aspects of this topic.

The ability of cells to interact and communicate with each other is critical for the collective lifestyle of cells in bacterial biofilms that would benefit from the advantages offered by diversity and communal division of labor (9). Within these multi-species communities, bacteria interact through physical and chemical mechanisms. One of the first points brought up during our discussions was a concern about the connotation of the word “communicate” itself. Cells communicating implies intent, which clearly microbes are incapable of, leading participants to suggest the need for a more neutral word. This concern harkens back to longstanding semantic discussions in the bacterial cell-cell interaction field regarding the use of the term “signal” and efforts to disambiguate it from a “cue” when discussing microbial communication systems. Both “signals” and “cues” benefit the organisms that can receive or sense these molecular indicators (54). In contrast, the term “signal” has been codified (in evolutionary theory developed in animal systems) to indicate systems that have evolved due to their ability to alter the behavior of the organisms receiving these signals and where the receiving organisms have also evolved specific responses to those signals (54, 55). In spite of the terms “signal” and “cue” being used interchangeably in common parlance, the necessity for precise language describing these different scientific paradigms has implications for how we shape predictions about the outcomes of these interaction systems (54). Indeed, some participants argued that the language used is relevant in terms of altering the theoretical framework with which we interpret the results from interaction studies. Interestingly, a similar discussion on definitions was brought up in the discussion group on mechanosensing: distinguishing between a signal and a stimulus. Here, the use of a more neutral and less personified term to describe these phenomena would benefit the field, which is perhaps why “microbial interactions” are commonly evoked. This phrase more generically represents the potentially reciprocal—but also potentially one-sided—bacterial “communication” that may be occurring within natural systems. This ambiguity may be both useful and necessary during early studies, where the precise nature of the signal or cue is not yet known.

In addition to these linguistic issues, several technical issues were discussed on our ability to examine and study cellular interactions and communication within biofilms. Most natural biofilms are three-dimensional structures, many of which are “thicker” than can be examined using traditional optical microscopy. This poses issues for accurately probing the heterogeneous environments distributed throughout microbial biofilms and exploring how microbes are interacting within them. There are additional challenges with measuring the temporal dynamics of these interactions, which are critically important to understanding which cells are interacting over time within the biofilms. In addition, there remains the challenge of being able to not only identify and localize the relevant interacting bacterial cells but also simultaneously identify the signals produced by them to affect one another. While theoretically these barriers can be overcome using multimodal imaging techniques that integrate diverse spectroscopy approaches, their effective implementation can be difficult. The generation of tools that are more generally available to broad sets of researchers that enable these types of studies would greatly advance our understanding of microbial interactions within biofilms.

In terms of which model systems should be used to study microbial interactions, there remains—as has been true for decades—a tension between environmental relevance/complexity and experimental amenability/tractability (56). Not unexpectedly, some participants argued that using more native-like systems (perhaps via enrichments to simplify the complexity and diversity of the communities) would yield the most relevant data. The trade-off for this environmental relevance is that when more species are present, it becomes harder to detect and distinguish relevant signals due to the masking of the interactions through the system’s complexity. With more simplified systems, it becomes more straightforward to track down and identify potential molecular cues and signals, which can then be verified for activity in more natural systems. The caveat of the reductionist approach is that identifying which microbes best represent the critical functional attributes of native systems is highly challenging. For example, efforts to identify relevant microbes extracted from native settings often rely on bulk analyses, where the disconnection between the relevant spatial scales can lead us astray; the “bulk” analysis of microbes coming from a single grain of sand on a millimeter scale (an example brought up in Roberto Kolter’s talk) would indicate a single interconnected bacterial community, where, in reality, it can be a number of much smaller isolated microscopic patches: the sample size still remains orders of magnitude larger than that of individual microbial cells. In a way, this is equivalent to assuming that all people within an area a mile square are “interacting” with one another!

Understanding spatial relationships between bacteria and whether they are transient or conserved will similarly impact our understanding of the evolutionary conservation of these signals since many signals likely act over short spatial scales, and thus microbes may only “talk” to others in close physical proximity, as was discussed in the talk by Avigdor Eldar (57). In addition, these evolutionary relationships may affect whether there are any signaling principles that may be conserved across many microbes and their communication mechanisms, or whether each identified interaction needs to be considered as a unique entity. Overall, these considerations reinforce the need to experimentally validate any interactions we identify in the laboratory using simplistic systems back in their original environments, as well as not to overinterpret findings from laboratory systems.

One proposed approach to the dual challenges of both selecting study systems that replicate the diversity and function of natural communities and how to normalize results across research groups is offered by fabricated ecosystems (EcoFABs) (https://eco-fab.org) (58–60). In these laboratory study systems, physical chambers provide standardized growth environments to facilitate intra- and inter-laboratory reproducibility, facilitating the comparison of microbial responses to diverse variables and the utilization of more native-like communities. These tools require the consensus of the community in selecting which variables should be conserved across the EcoFABs but will then enable the generation of imaging, omics, and phenotyping datasets that build productively on other scientists’ work.

Overall, the take-home messages from the discussions in this group (which in fact are also applicable to the several previous key topics) were as follows. First, we need to work to expand the tools available and being utilized. Second, we need to normalize the types of data we can collect (i.e., multimodal dynamic imaging over time; using conserved study systems and microcosm environments) and thus enable all of our results to be more broadly applicable. It remains important to use modest experimental study systems and to then not overstate the ecological implications of our findings without first translating these studies into more realistic natural settings.