Bacterial Metabolic Heterogeneity: Origins and Applications in Engineering and Infectious Disease

Abstract

Bacteria within an isoclonal population display significant heterogeneity in metabolism, even under tightly controlled environmental conditions. Metabolic heterogeneity enables influential functions not possible or measurable at the ensemble scale. Several molecular and cellular mechanisms are likely to give rise to metabolic heterogeneity including molecular noise in metabolic enzyme expression, positive feedback loops, and asymmetric partitioning of cellular components during cell division. Dissection of the mechanistic origins of metabolic heterogeneity has been enabled by recent developments in single-cell analytical tools. Finally, we provide a discussion of recent studies examining the importance of metabolic heterogeneity in applied settings such as infectious disease and metabolic engineering.

GRAPHICAL ABSTRACT

Introduction

Bacteria within isogenic populations consistently feature phenotypic variation, even under identical environmental conditions. Examining phenotypic heterogeneity at the single cell level often yields insights that are exceptionally relevant to the function of the overall population, yet not captured in bulk measurements[1]. For example, a given feature of a subpopulation may strongly predict which cells survive environmental changes, or bestow advantages to the population at the expense of the individual[2]. One important but largely unexplored type of phenotypic heterogeneity lies in cell-to-cell variations in metabolite levels and dynamics. Not only are these directly relevant to function, but they also strongly influence other cellular processes such as translation and growth. Thus, assessment of cellular metabolic heterogeneity and what causes it is indispensable towards fully understanding the behavior of a bacterial population. While methods to assess gene expression at the single cell level are well established, the diversity in chemical structure and rapid turnover dynamics of metabolites posits an ongoing challenge in assessing metabolic heterogeneity. Developments in single-cell instrumentation and molecular tools that allow quantitation of specific metabolites have just begun to enable investigations into both fundamental principles and several applied settings. Other focused reviews have covered related topics such as single-cell technologies for bacteria [3] and the relevance of heterogeneity in metabolic engineering [4,5]. Here, we provide an integrative overview of core concepts and recent progress across several topics in the field of bacterial metabolic heterogeneity. These include its biological origins and purposes, tools and innovative approaches to characterize it, and its applications in settings such as infectious disease and metabolic engineering. We focus mostly on findings from isoclonal populations of symmetric bacteria such as Escherichia coli, while metabolic heterogeneity caused by genetic variation or observed in asymmetric bacteria such as Caulobacter crescentus will not be discussed.

Fundamentals of bacterial metabolic heterogeneity

Variability in bacterial single cell metabolism has previously been attributed to microenvironmental differences, and these are known to have relevance in disease or bioproduction outcomes[6,7]. However, recent studies discovered significant metabolic heterogeneity even under the most idealized laboratory conditions, suggesting it is a fundamental feature of bacterial populations. Heterogeneity has been observed in many diverse nodes of bacterial metabolism to date, including central carbon metabolism[8,9], fatty acid biosynthesis [10], pyruvate-sensing networks[11], ATP[12], ribosome content[13-15], and amino acid pathways [16]. Depending on the given metabolite and system, cell metabolite distributions can range from narrow to broadly unimodal, or even multimodal [17]. Here, we discuss the basic biological relevance of metabolic heterogeneity and underlying molecular and cellular causes.

Much of the functional relevance of metabolic heterogeneity in natural settings can be conceptualized as “bet hedging”, where heterogeneity increases the likelihood that some subpopulation will survive a future stress. The phenomenon of arrested growth upon nutrient shifts serves as an excellent example. Here, homogenous responsive switching was once thought to explain resumed growth after a lag[8]. However, it is now known that many cells remain quiescent and are outperformed by expanding subpopulations better adapted to the new environment [8,9,18]. Klebsiella oxytoca, for example, exhibited a remarkable heterogeneity in nitrogen fixation when exogenous nitrogen (ammonia, NH₄⁺) is slightly limited. Under this condition, cells with high nitrogen fixation capability had a significant growth advantage during subsequent NH₄⁺ removal [19]. In contrast, cells equilibrated in either starved or high exogenous nitrogen conditions show expected high or low nitrogen fixation with minimal heterogeneity and are less adaptive to condition changes. Overall, these examples illustrate how metabolic heterogeneity ensures survival in changing environments.

What are the primary molecular mechanisms driving metabolic heterogeneity? First of all, basic principles governing noise in transcription and translation lead to variations in the abundance of metabolic enzymes, and in turn, affect levels of the metabolites they produce [20] (Figure 1A). Transcription is especially stochastic, occurs in bursting patterns, and leads to high variabilities in mRNA between single cells. This makes low-abundant proteins have particularly high variabilities when propagated to translation[21]. Even for high-abundant proteins, cell-to-cell variations in gene-extrinsic factors, such as ribosomes or RNA polymerases, contribute to heterogeneity in protein expression [21,22]. Interestingly, metabolic genes tend to be controlled by noisier promoters, while essential genes are often controlled by the least noisy promoters [23]. Such observations suggest that cells may explore a greater variation in enzyme expression to obtain large metabolic heterogeneity. Indeed, stochastic modeling using measured enzyme variability shows this alone can generate large metabolic heterogeneity within an isogenic population [20]. At the extreme end of possibility, stochastic expression can even lead to bimodal expression without bistability [24]. This is particularly consequential for transcription factors that regulate metabolic genes as variability can be readily transmitted to metabolic heterogeneity. In Pseudomonas Putida, expression of the transcription factor XylR follows this bimodal pattern and generate bimodal variability in xylene metabolism[25].

Figure 1: Sources of Metabolic Heterogeneity in Isogenic Bacterial Populations.

A. Transcription often occurs in stochastic bursts, leading to high variability between cells. Variability is transmitted to enzyme copy number and could lead to heterogeneity in metabolite concentration. B. Transcriptional positive feedback loop generates metabolic bistability. An example is shown where the lactose permease LacY facilitates lactose uptake. Lactose destabilizes the repressor LacI and prevents it from binding DNA. This allows further active transcription and translation of the lac operon including LacY, which then further accelerates lactose transport into the cell. Some aspects of the lac operon and regulation are omitted for simplicity. C. Positive feedback in enzymatic catalysis generates metabolic bistability. An example is shown where cells are trapped in a high-glucose-induced, growth-arrested state stabilized by an imbalance in glycolytic flux. High flux through upper glycolysis rapidly consumes ATP and assimilates generated phosphate (P_i) as fructose bisphosphate (FBP). Any P_i from other sources entering lower glycolysis is unable to rescue imbalance because generated ATP is strongly favored towards two further rounds of upper glycolysis, only further stabilizing this state. In other subpopulations and physiological conditions, regulatory mechanisms prevent this feedback from reaching stability. D. Asymmetric partitioning of cellular components upon cell division can generate heterogeneity in metabolic activity between daughter cells.

More commonly, multimodality is observed in multi-stable systems, where positive feedback regulatory architectures under molecular noise push cells to multiple stable states with distinctive phenotypes. One well-studied example is the E. coli lac operon, where lactose inactivates the repressor LacI and increases transcription of the permease lacY, which further transports more lactose inside the cell, forming a positive feedback loop and leading to two stable subpopulations/states with low and high intracellular lactose levels (Figure 1B) [26]. Additionally, metabolic multimodality can form through positive feedback of enzymatic catalysis or metabolic imbalance (Figure 1C). High-glucose induced imbalance between upper and lower glycolysis is one such case (Figure 1C). Rapid flux through upper glycolysis depletes ATP and results in fructose biphosphate accumulation. Cellular phosphate is consumed in lower glycolysis to regenerate ATP, but in the depleted state ATP is immediately channeled to upper glycolysis. Further, the 1:2 stoichiometry in ATP between lower and upper glycolysis only further ensures the accumulation of fructose biphosphate despite any further input of phosphate from other cellular sources. [27]. This results in stochastic formation of subpopulations with depleted ATP/Pi pools when this imbalance reached a critical threshold, and demonstrates how even tightly regulated pathways can reach multimodality[27]. Although this metabolic imbalance was observed in yeast, similar mechanisms may exist in bacteria[8].

Another source of metabolic heterogeneity is asymmetric partitioning of cellular components during cell division [28]. For any cellular component, the chance of asymmetric partitioning is high when it has low copy number and slow diffusion kinetics [29,30]. Most small-molecule metabolites have high copy numbers (in E. coli, 10³-10⁷ copies[31]) and rapid diffusion kinetics such that their direct asymmetric distribution is unlikely. However, daughters of the same mother cell still display different metabolic activity following division[32]. Several mechanisms could lead to different metabolic heterogeneity between daughter cells. First, asymmetric partitioning of low-abundant metabolic enzymes or extremely large enzyme complexes may cause immediate metabolic differences in daughter cells upon division [29]. Second, transcription factors that regulate metabolic pathways often have relatively low copy number, and their asymmetric partitioning could lead to different expression level of pathway enzymes. For an end product of a metabolic pathway, its concentration can takes as many as seven cell cycles to reach steady state upon transcriptional activation or repression[33]. Thus, metabolic variability may not be observed right after division, but may appear over the next a few generations after asymmetric partitioning of metabolite-regulating transcriptional factors [33]. Third, asymmetric segregation of low-copy- number subcellular components, such as inclusion bodies, may cause variation in daughter cell grow rate [34], which further affect other metabolic activities. Such asymmetric partitioning often confers faster growth to daughter cells with less aggregates, and broad stress-resistance to daughter cells inheriting more aggregates[34,35]. To summarize, metabolic heterogeneity can arise from a number of molecular and cellular mechanisms and often has important implications for population-level function.

Approaches for measuring Bacterial Metabolic Heterogeneity

Recent advances in single cell instrumentation, imaging, and biochemical techniques have greatly accelerated our ability to assess metabolic heterogeneity. Two widely used class of tools for single cell assessment are chemical-based and genetically encoded metabolite biosensors, which couple concentrations of a specific metabolite to a quantitative fluorescent output [10,36]. Using such fluorescent output, metabolic heterogeneity can be measured by either flow cytometry or live cell imaging, and in certain cases fluorescence-activated cell sorting can be used to further separate cells based on their different phenotypes. Several strategies exist to construct genetically encoded metabolite biosensors. First, metabolite-responsive transcription factors can be used to turn on or off the expression of fluorescent reporters or RNA based fluorescent aptamers according to the concentration of the target metabolite [36-40]. RNA-based fluorescent aptamers carry the advantage of faster reporter kinetics versus fluorescent proteins, which require time to fold before fluorescing. Second, even faster reporting kinetics can be achieved with fluorescent resonance energy transfer (FRET) based biosensors. Here, metabolite binding induces a conformational shift that alters the proximity between a donor and acceptor fluorescent protein. Excitation of the donor fluorophore and measuring the relative emission spectra allows metabolite quantification. [36,41]. Third, metabolite-binding RNA aptamers can be used to stabilize the fluorescence of adjacent RNA-based fluorescent aptamers.This system carries similar kinetic advantages of FRET and can benefit from the large repertoire of metabolite-binding RNA aptamers or in vitro aptamer selection (also known as systematic evolution of ligands by exponential enrichment, SELEX)[37,42]. In general, most genetically-encoded metabolite biosensors are functional in living growing cells, and thus can provide dynamic information of a metabolite at a time scale longer than the sensor’s response time. However, these sensors are limited to cells that are amenable to genetic manipulation.

Mass spectrometry is the gold standard approach for versatile and quantitative metabolite assessment. Of several mass spectrometry techniques, nanoscale secondary ion mass spectrometry (NanoSIMS) is often used for studying bacterial metabolic heterogeneity due to its exceptional subcellular resolution (~50 nm) and ability to measure several metabolic species per scan. This can be performed on non-cultivable bacteria, and as a result has found extensive applications in environmental microbiology. For example, wastewater sludge bacteria were recently found to have extensive heterogeneity in oleic acid accumulation[43], and single cell N₂ and CO₂ fixation were observed to correlate in Chlorobium phaeobacteroides. NanoSIMS can also be combined with prior fluorescence-based secondary analyses to further link single-cell metabolic data to many other phenotypes[44,45]. However, it is destructive and generally incapable of time-resolved measurement in single cells.

Last, Raman spectroscopy is a nondestructive technique that probes intrinsic vibrational frequencies of biomolecules, and it can be combined with microscopy for single cell evaluation. A typical output features intensity as a function of wavenumber where intensity peaks represent specific chemical bonds encompassing virtually all major classes of biomolecules. Further, combining it with stable isotope probing produces predictable peak shifts and has allowed assessment of metabolic heterogeneity in pathways of interest including carboxymethylcellulose degradation[46] and carbon catabolite repression in naphthalene-degrading bacteria[47] . Though a fairly global assessment of metabolism is useful, establishing a robust connection between specific metabolites of interest and spectra can be difficult. High-dimensional analyses correlating features of Raman spectra with specific metabolites [48] or the transcriptome[49] may improve the insights gained. In conclusion, these technologies have served as the basis for uncovering metabolic heterogeneity, and further innovation will continue to expand the scope and breadth of what can be assessed.

Applications of Metabolic Heterogeneity in Infectious Disease and Metabolic Engineering

With a basic knowledge of the underlying principles and increasingly powerful tools to measure metabolic heterogeneity, there has never been a better time to study its relevance and uses in applied settings. Two particularly promising areas for application are infectious disease and metabolic engineering.

In infectious disease settings, heterogeneity in metabolism can be profoundly influential to bacterial survival, population adaptation, and infection outcomes. One example is bacterial persistence, a phenomena where small subsets of bacteria display increased phenotypic antibiotic tolerance compared to the rest of a population[50]. Post-antibiotic regrowth of persistent bacteria threatens patients and perhaps worse, is mechanistically linked to development of antibiotic resistance[51]. Metabolic heterogeneity seems to be a fundamental principle underlying persister formation, phenotypes, and regrowth (Figure 2A). First, in a manner analogous to bet-hedging, prior stress factors such as starvation [52] or nutrient shifts[8,53-55] induce metabolic heterogeneity in populations to increase persister formation. Though few specific aspects of persister metabolism are universal, they are often characterized by extremely slow growth[56], reduced ATP levels[57,58], reduced translation levels[59], and higher levels of the metabolite alarmone (p)ppGpp [53] relative to the rest of the population. Proteomic and targeted metabolite analyses of persisters support this concept and further suggest these traits are self-reinforcing, thus keeping the cell in a dormant state [55]. Second, there is heterogeneity even within cell subpopulations that survive early antibiotic killing. Higher residual metabolic capacity and ability to regrow upon antibiotic removal distinguishes true persisters from viable, but non-culturable cells (VBNC’s), which often appear intact but mostly fail to regrow [35,55,60,61]. Ribosome content, translation capacity, and ATP-dependent protein aggregation are example parameters that predict if, and how quickly, cells are able to regrow following antibiotics [35,61,62]. Finally, appreciation of the metabolic status of persisters and their heterogeneity may guide efficient use of antibiotics. Provision of specific supplemental carbon sources is capable of activating metabolism in persister cells and allowing antibiotic killing[63].

Figure 2: Metabolic Heterogeneity in Bacterial Persistence and Engineering Applications.

A. A subpopulation of bacteria cells respond to environmental stresses (e.g. starvation) by reducing metabolic activity (e.g. growth rate, [ATP], translation). Under antibiotic treatment, this subpopulation displays high phenotypic tolerance, forming persister subpopulations. Upon removal of antibiotic, persisters regrow to resemble the initial population, whereas viable-non culturable cells (VBNC) often fail to rescuscitate. B. Subpopulations of bacteria with low phenotypic productivity are often observed in metabolic engineering settings and limit overall culture yield. Metabolic engineering strategies can limit growth of low-producing cells by engineering genetic circuits which couple transcription-factor (TF) based sensing of the product to growth advantages conferred by antibiotic resistance or synthetic auxotrophy. Low producing cells (top) die or grow more slowly, whereas high producing cells (bottom) are rewarded with superior growth and survival. This results in enrichment of high-producing cells and enhanced population-level yield (right).

A second area where understanding and harnessing metabolic heterogeneity holds great potential is in metabolic engineering. Cell-to-cell variation in biosynthesis of target metabolites is often observed, and the presence of low producing subpopulations has been shown to limit overall yield[10,16,64] (Figure 2B). Consequentially, an effective method to increase ensemble yield is to control metabolic heterogeneity by minimizing low-producing cells. This can be done by using genetic circuits that sense the desired metabolite and provide high producing cells a growth advantage. These approaches demonstrated that control of metabolic heterogeneity can facilitate metabolite production in challenging large scale culture scenarios[10,64,65]. Another challenge in metabolic engineering lies in maintaining constant expression of genes in metabolic pathways of interest, which can vary widely among cells and interfere with production goals[66]. A recent study showed that by engineering a noncoherent feedforward loop into a promoter of interest, identical expression could be maintained despite genomic mutations, changes in copy number, or changes in environment[67]. Use of this type of expression system could prove useful to homogenize production across cells.

Conclusions and Future Perspectives

Here, we have discussed key advances in understanding the biological origins of metabolic heterogeneity, approaches for its measurement, and its relevance in crucial applied settings. As interrogation of metabolic phenotypes at the single cell level becomes more routine, it is likely that the overall relevance of the discussed molecular mechanisms giving rise to metabolic heterogeneity across many settings will be clarified and perhaps reveal new possibilities. Finally, we anticipate these fundamental findings to continue to be applied towards new understanding of infectious disease, metabolic engineering, and beyond.

Highlights.

• Molecular noise, feedback, and asymmetric partitioning create metabolic heterogeneity.

• Biosensors, mass spectrometry, and Raman spectroscopy enable single cell measurement.

• Minimizing heterogeneity improves ensemble yield in engineered production strains.

• Metabolic heterogeneity underlies bacterial persistence.