|Home | About | Journals | Submit | Contact Us | Français|
Summary: Over the last few decades, advances in cultivation-independent methods have significantly contributed to our understanding of microbial diversity and community composition in the environment. At the same time, cultivation-dependent methods have thrived, and the growing number of organisms obtained thereby have allowed for detailed studies of their physiology and genetics. Still, most microorganisms are recalcitrant to cultivation. This review not only conveys current knowledge about different isolation and cultivation strategies but also discusses what implications can be drawn from pure culture work for studies in microbial ecology. Specifically, in the light of single-cell individuality and genome heterogeneity, it becomes important to evaluate population-wide measurements carefully. An overview of various approaches in microbial ecology is given, and the cell as a central unit for understanding processes on a community level is discussed.
There has always been a great fascination in seeing microbiology in action. Whether it is during controlled fermentation while making wine or beer, watching satellite images of ocean water changing color due to an algal bloom, or sensing the typical (microbially produced) smell of soil after a rain shower, observing microbiological processes in our daily life reminds us that we share the planet with myriad unseen microorganisms. Making these microbes visible by looking at colonies on an agar plate or examining them under the microscope, for example, represents an even greater appeal—and not only to microbiologists. This visualization by isolating, growing, and cultivating microorganisms is a task that represents the daily routine in many molecular and environmental microbiology laboratories around the world. Now, at a time when various high-throughput data sets are available to address questions in environmental microbiology and microbial ecology, the isolation and cultivation of microorganisms have lost the appeal they had for hundreds of years. This review is centered around the microbial cell as the defining entity in environmental microbiology and microbial ecology. From the level of a cell we can “zoom in” and obtain comprehensive information on molecules and their interactions that define physiology and the phenotype of the cell. The cell level also allows us to “zoom out” and examine the interaction of the cell with other organisms and the environment and to investigate how these interchanges shape communities and habitats. This review therefore not only will highlight isolation and cultivation methods that allow us to obtain a cell for subsequent analysis in the first place but also will assess how and to what extent data obtained from experiments with pure cultures can be extrapolated to answer questions in microbial ecology. At the same time, this review will evaluate how data obtained at the molecular level as well as the community level can be beneficial to one's knowledge of the cell.
Microorganisms in the environment interact on various levels with the microbial community and the environment itself, and the isolation of an organism will in most cases disrupt these interactions. It is therefore important to understand what forms of interactions exist in the environment and to predict what changes in phenotype might occur when these interactions are omitted during cultivation in the laboratory. Recent advances in sequencing technologies have revealed a tremendous diversity on the microbial genome level, not only within defined cultures in the laboratory but also within microbial populations in the environment (100, 168, 225). Understanding what effect genome heterogeneity has on physiology and phenotype is essential to interpret the vast genomic data now becoming available. The genomic repertoire lays the foundation for microorganisms to adapt and evolve in response to changing conditions in multiple ways, not only in nature but also in the laboratory. Determining the underlying principles and causal effects that these adaptations have on the cell's phenotype and fitness is essential; otherwise, the analysis of community-wide data can only be of a descriptive nature. When using cells as a kind of stepping-stone to move from molecular biology data to natural populations and whole environments, it is crucial to evaluate the robustness of this process. This means that we have to carefully estimate the implications that can be drawn from our data. The following will therefore cover a wide range of subjects, from biological processes on a molecular level to individual microorganisms, from individual organisms to populations, and from populations to the environment.
Since the terms isolation, growth, and cultivation are often used synonymously, it will be beneficial to briefly define them here so that they can be distinguished throughout this review.
“Isolation” of an organism (or multiple organisms at a time) describes the process by which individual cells are physically separated from each other and/or from matrix material, such as water, air, soil particles, or eukaryotic tissues. Isolation therefore represents the most crucial step during the process of obtaining pure cultures. Isolation also includes the process by which defined cocultures are obtained for further cultivation.
For microorganisms, “growth” implies the division of a bacterial cell, resulting in duplication of the cell number. Measuring and observing bacterial growth, especially in the environment, can be challenging since the rates of growth and of death (e.g., due to apoptosis, grazing by eukaryotic predators, or cell lysis by phages) can be identical, resulting in net growth that will be zero.
Traditionally the terms “culture” and “cultivation” are used to describe a defined bacterial population that can be grown and maintained in the laboratory, usually at a scale that involves billions of cells at a time. Cultivation is not exclusive to pure cultures but can include mixed populations and stable consortia that are propagated in the laboratory for a prolonged period of time.
Microorganisms are isolated, grown, and cultivated in the laboratory for many reasons. Examples are the enumeration of bacteria with a certain function or role in nutrient cycling, in formation and degradation of organic and inorganic molecules, or in bioremediation and energy production. Other examples include the testing of Koch's postulates, identification of organisms that carry specific genetic information (gene or pathway), evaluation of phylogeny and physiology, and discovery of novel enzymes and chemical entities (e.g., anti-infectives) for industrial and pharmaceutical applications. As broad as the scientific goals are the cultivation methods used to accomplish them. Depending on whether a defined group of microorganisms is targeted (e.g., new bacterial or algal strains for biofuel production) or whether “as many as possible” different strains should be isolated (e.g., for diversity assessment or to accompany metagenomic studies), the most suitable methods and their refinement will differ substantially. However, having microorganisms in culture allows for the direct study of morphology, physiology, genetics, and pathogenicity in great detail, tasks which are difficult to accomplish when solely molecular tools are used.
Advances in molecular biological techniques over the last three decades have spurred cultivation-independent developments. In medical diagnostics, for example, isolation and cultivation have been replaced by advances in molecular methods that can identify specific microbes or genetic markers more accurately, often faster, and more cost-effectively. However, one has to keep in mind that these markers were initially linked to a certain disease by work that was performed with microbial pure cultures.
Physical separation of individual cells (or groups of cells) is essential to cultivation efforts. This isolation step can take place before or after cells are grown (see below). There are several methods to physically separate cells, probably the most common of which is separation of cells by spreading them onto a solid medium. This method was introduced by Robert Koch over a century ago (118) to visualize, isolate, and ultimately cultivate microorganisms. Although several advances have been made in isolating bacteria on solidified medium since Koch first used agar-solidified medium (108), the basic principle of isolating bacteria by spreading them on plates and “picking” colonies remains unchanged. The underlying concept is that a single bacterial cell, spread on an agar plate (or solid medium made with other gelling agents), will start to divide and consequently form a colony that is visible by the naked eye or by microscopy. These colonies can then be separated from each other using various tools, e.g., a loop or toothpick, depending on the colony size. The process is defined by a separation step (spreading cells onto a plate), a growth step (colony formation), and the actual isolation step (colony picking). The most critical step here is the colony formation. It was recognized early on that the majority of cells observed under the microscope will not form colonies on solid media (38), a phenomenon that over half a century later became known as “the great plate count anomaly” (201). However, it is important to note that bacterial cultures can undergo certain adaptations during these isolation and growth procedures. For example, some strains that were not able to grow on solid media before were adapted to form colonies on agar plates after several attempts (35) (different forms of adaptation will be discussed in more detail below). Other microbes (e.g., some strictly anaerobic microorganisms) will not form colonies on surfaces but instead can be grown inside solid media, a phenomenon that resulted in the use of agar shakes or agar dilution series for isolation purposes (233). Conversely, there are bacteria that require surfaces to grow on (e.g., gliding bacteria), and isolation and cultivation of these organisms is hindered by the use of liquid media (189).
Methods by which cells are isolated before growth takes place include the use of flow cytometry (61, 165), microfluidics (141, 211), or micromanipulation using focused laser beams (so-called optical tweezers) (69, 105). These techniques are all sensitive enough to detect and subsequently separate individual cells (Fig. (Fig.1).1). An approach that does not require single-cell detection for separation is the isolation of bacteria by microencapsulation (249). A more commonly used technique is the isolation of bacteria via liquid serial dilution (27, 39, 49, 189). This technique is applied especially in cases where bacteria do not form colonies on solid surfaces or where media cannot be adequately solidified with agar, e.g., due to low pH (221). Recently a method that uses nanofibrous cellulose to solidify media even at low pH and therefore can support growth of acidophiles has been described (46, 220).
There are several methods with various sensitivities to measure and describe bacterial growth (Fig. (Fig.1).1). Most often, growth is observed by turbidity, using a photometer, or just by looking at a culture; these methods are suitable if >105 cells per milliliter are present. Detecting microbial growth, qualitatively or quantitatively, sounds trivial, but not all bacterial cells are able to form visible colonies on solid medium plates, and therefore growth cannot be detected conveniently by the naked eye. Even within pure cultures, it is known that the rate of colony formation is not uniform and that several cells might form only microcolonies which cannot be detected by the naked eye (104, 235). Therefore, several researchers applied microscopy to observe colony formation on solid agar (112) or membrane systems (60, 114). Even after prolonged incubation of 1 to 6 months, many colonies of soil bacteria can still be present as microcolonies and might never grow larger (108, 228). Similar “self-limiting” growth behavior has been observed for oligotrophic marine bacteria (35, 170). Detecting growth by microscopy, which normally allows a detection of ~103cells/ml, increases the sensitivity compared to that of turbidity measurements, which require higher cell densities (Fig. (Fig.1).1). A method that allows for the detection of even fewer cells is the combination of isolating cells by encapsulation in microcapsules and sorting by flow cytometry (250). Instead of a flow cytometer, this encapsulation technique can also be combined with a microfluidic approach to monitor division of cells (122). Microfluidics can also be applied to observe the division of single cells directly. Advanced methods have been developed that allow monitoring and screening of large numbers of organisms at the same time. Flow cytometry, for example, allows the screening of 5,000 to 50,000 events per second (249), and high-throughput approaches using new technologies, such as a GigaMatrix (128) and a microdish (102), have been developed to be applied for miniaturized culture volumes, allowing the growth and screening of millions of cells in a highly compartmentalized format. An advantage of these high-throughput methods is that they are capable of growing and screening many organisms at a time so that cultivation of previously uncultivated ones becomes more likely (102, 249). Furthermore, there are a variety molecular biology tools that measure various cellular components and infer growth from these data (see below).
Over the last decade there has been some discussion about the general culturability of microorganisms. Several authors referred to organisms that were known by their molecular fingerprints (mainly 16S rRNA gene sequences) but could not be brought into culture at that time as they were “unculturable.” Clearly, this word has been used in an imprecise way, since many formerly “unculturable” organisms later became part of our culture collections (72). However, the majority of microorganisms in any given environment have not been cultivated (yet) even when sophisticated media and new cultivation and isolation methods are applied. One possible reason is that researchers tend to stick to a handful of different media (at most) and do not spend time and effort to optimize nutritional needs, i.e., medium compositions as well as physicochemical parameters such as temperature, pH, salinity, and growth atmosphere. The notion that cultivation attempts fail because exotic compounds serve as exclusive carbon sources for growth is probably not correct. There are many success stories where former “unculturable” microorganisms (in some cases known for decades) have been cultivated in the laboratory using common nutrients (25, 204, 226); to rely exclusively on exotic substrates (140) is therefore more likely the exception than the norm. However, we should note that various signal molecules, as discussed later in this review, seem to play an important role during cultivation and that isolation procedures likely disrupt this signaling. Cultivation success is probably not hampered exclusively by what is offered to the microorganisms for growth but, importantly, on how much of these nutrients is provided—plenty can often be too much. Media containing high concentrations of nutrients, often a billion to a trillion times more than what microbes encounter in their natural environment, can have inhibitory effects (1, 158). Bacteria seem to have developed different strategies to adapt to changing nutrient concentrations in the environment, which consequently will determine if they are able to form colonies on nutrient-rich agar plates (8). Therefore, microorganisms that have the capability to adapt and cope with high concentrations of nutrients are often overrepresented in cultivation-based studies (107) and ultimately in our culture collections (117). As a response, several techniques that use low concentrations of nutrients for isolation and cultivation efforts have emerged (10, 59, 83, 103, 112, 114, 170, 249). However, growing bacteria at a low concentration of nutrients will logically result in low biomass gains. Large culture volumes become necessary (170) to obtain enough biomaterial for physiological and genetic studies, and techniques to observe growth (Fig. (Fig.1)1) have to be adapted accordingly (see above). Decades ago, an elegant way to circumvent the necessity of large culture volumes was introduced by Novick and Szilard (156), who were the first to use flowthrough setups for growing cells under a constant supply of nutrients (a chemostat). Since then, chemostats have been an excellent tool in microbiology for various approaches, such the study of pure cultures, competition experiments (125), and the study of mixed cultures and microbial consortia (57, 205). Another benefit of flowthrough setups is that inhibitory (by)products which tend to accumulate in closed cultures are diluted or removed from the system. However, using chemostats for enrichment and isolation of microorganisms from environmental samples has often not been possible without thoughtful modifications, since many organisms tend to attach to surfaces of the setup, form biofilms, and potentially out-compete their planktonic counterparts (116). A modified flowthrough setup for cultivation purposes, which provides a constant nutrient flow at low concentrations but eliminates the formation of biofilms, has been the incubation and growth of separated cells in microcapsules (117, 250).
I have touched already on some reasons and theories why cultivation of the majority of microorganisms in the laboratory has failed so far. In the following I will emphasize and discuss that the selection of the microorganism targeted for cultivation defines all of the following growth and cultivation steps. First, there is the choice of separation and isolation methods (see above), which consequently determines the selection of liquid or solid medium for growing cells. Second, it is fundamental to determine if any contact of cell to cell or cell to substrate is essential for the targeted microorganism to grow. An encapsulation method (249), for example, is of limited use if direct cell-to-substrate contact is required for oxidation and reduction reactions and subsequent growth. Such substrates are, for example, long-chain hydrocarbons and other crude oil components that serve as electron donors or an electrode in a fuel-cell that serves as an electron acceptor. Third, the composition of the medium is also critical for the cultivation success. However, we tend to use vitamin and trace element solutions that have been developed decades ago without rethinking and redesigning their composition. Medium components are known to have an effect on cultivation efficiency, including carbon and energy sources (44), various inorganic chemicals and salts (52, 231), signal compounds (22, 31, 80), and trace elements, vitamins, and amino acids (87, 88, 134). Basically, every component of the medium other than water has been demonstrated to have an inhibitory effect on certain microorganisms. Accounting for all these inhibitory effects by varying every component of a standard medium can be a daunting task (Fig. (Fig.2).2). To illustrate this, I picked a medium which is commonly used to grow anaerobic bacteria and that contains 33 different components (29). By changing the concentration of a single component at a time, one would need to generate and test 33 different media; accounting for increasing as well as decreasing concentrations of this component would result in 66 different medium combinations to be evaluated. Changing any two components of the medium at the same time (increasing and decreasing their concentrations) would result in 2,112 different medium combinations to study. Variation of any 22 components at a time would require an inconceivable ~1015 (1 quadrillion = 1,000 billion) medium combinations, clearly an unrealistic effort (Fig. (Fig.2).2). Furthermore, these are only variations of an existing medium, reflecting a medium “optimization” effort. This effort would neither include other electron acceptors nor account for additional potential electron donors (e.g., carbon sources), and, more importantly, it does not include differing environmental conditions (pH, temperature, salinity, buffering capacity, pressure, and gas atmospheres such as different carbon dioxide or oxygen concentrations). It has often been cited that the media which are routinely used for most cultivation efforts do not allow growth of most microorganisms in the laboratory. It is surprising that not more attention has been paid to the improvement of cultivation methods but that instead the majority of microorganisms are categorized as “unculturable.” This might be the case for some organisms, but there is clearly (a lot of) room for rational optimization of just medium compositions, which has been documented by a number of studies (44, 54).
The recent success of new cultivation techniques and the use of modified media to gain access to previously noncultivated microorganisms demonstrates that many organisms can be isolated and maintained in culture in the laboratory (131). In addition, it was realized that cultivation of “new” microorganisms might be nominal when nutrient-rich, “off-the-shelf media” are being used. When cultivating any kind of microorganism, conditions should be adapted to natural environmental conditions, by at least adjusting the pH, salinity, and temperature (and in some cases pressure ) to simulate environmental conditions. While this effort has increasingly become a routine, only few studies consider variations in atmospheric pressures. Many cultivation attempts are performed exclusively under oxic (~20% O2) or anoxic conditions, but only a few studies account for low oxygen requirements of microaerophilic microorganisms. Molecular oxygen, however, represents one of the most reactive elements on our planet; only fluorine exhibits a greater electronegativity (the ability of an atom to draw electrons) (162). The reactivity of oxygen and reactive oxygen compounds such as hydrogen peroxide, superoxide, and hydroxyl radical has been well described in the literature (3, 124). Bacterial life existed on this planet before elemental oxygen was introduced into the atmosphere and with it an increase of oxidized compounds, such as the common electron acceptors nitrate and sulfate (166). During the time of the slow oxidation event on our planet, microorganisms had billions of years to adapt to various oxygen concentrations and ultimately develop fully aerobic metabolisms, utilizing oxygen at atmospheric concentrations. There is a broad range of requirements and tolerances toward oxygen among microorganisms between the strictly anaerobic and fully aerobic bacteria. Some anaerobic microbes do not tolerate any level of oxygen; others tolerate various concentrations and have different levels of inhibition (42, 135). Other bacteria require oxygen for their energy metabolism but are extremely sensitive to higher oxygen concentrations (139, 161). These fundamental differences in oxygen requirement and tolerance had already been described by Winogradsky in the 1880s (236). The degree of flexibility in regard to oxygen not only defines the ecological niches of specific organisms but largely affects cultivation success (202). Recently, several microaerophilic organisms that thrive strictly at the oxic-anoxic interface and tolerate oxygen only at specific and low concentrations have been isolated (55, 153, 193, 216). To maximize the cultivation success, the simulation of the natural environmental conditions is critical, and the degree of specificity increases with the specificity of the requirements and tolerances of the organism. The smaller the ecological niche where the microorganism can thrive is defined, the more specifically the medium has to be prepared and the environmental conditions have to be simulated to allow cultivation. Unfortunately, thorough investigations of the ecological niche on scales relevant to microorganisms are sparse. Due to the lack of detailed and high-resolution measurements, researchers tend to simplify what the ecological niche of an organism consist of. However, only such high-resolution measurement of varying environmental parameters (17) offers a peek into the “living room” of a microbe—a necessary insight if we want the organisms to feel at home in the laboratory.
Growth of microbes in the laboratory is dependent on the medium and the cultivation conditions that are applied. This includes the equipment and materials that are being used for cultivation. It has been known for many years that chemicals leaking out of plastic and rubber laboratory supplies (pipettes, cultivation plates and trays, rubber stoppers, and tubing) can have inhibitory effects on bacterial growth (204), and several additional bioactive contaminants have been identified (144). The release of substances into incubation setups changes the composition of the medium and the environmental conditions. For example, if organic solvents are used as growth substrates, concentration of bioactive compounds leaching out of plastic containers can have inhibitory effects. Even filter material used for sterile filtration may have negative effects on bacterial growth. Therefore, it is desirable that contact with plastic and rubber materials be kept to a minimum during cultivation. A pretreatment of the lab equipment, e.g., by boiling rubber stoppers and tubing with 1 N NaOH followed by additional boiling in ultrapure H2O prior to use, can also improve cultivation efficiency (K. Zengler, unpublished data). Inhibitory effects have also been linked to certain types of glassware as well as in general to the use of new glassware. To my knowledge, it has never been determined what actually causes the inhibitory effect of certain glassware, but boron nitrides and other chemicals have been suspected to be slowly released from the glass and to be responsible for inhibitory effects (108). A thorough washing step, especially for new glassware, can reduce these inhibitions. Also, the quality of water that is used to prepare the medium is of great importance. Chemical impurities of water and gelling agents (such as agar or gellan) not only can affect cultivation success but are known to change the phenotypic behavior of microbes (109, 232). As an example, it was found that in members of the order Actinomycetales the expression of pathways encoding certain secondary metabolites is favored when media are prepared with tap water versus double-distilled water. Autoclaving time and its negative side effects (i.e., formation of reactive oxygen species or Maillard reaction products) can also have an impact on the cultures (76, 110, 195). Physical stress may also have an effect on cultivation efficiency. Certain microorganisms may disagree with James Bond's “shaken, not stirred” when it comes to their preferred growth environment and grow much better without any agitation.
Even if all the steps mentioned so far are carefully considered, many cultivation attempts will still fail, because the researcher has not been patient enough. This is something that has to be especially taken into account when assigning a cultivation project to a graduate student, who has only a limited time frame available to complete his or her projects. Part of this has to do with the way bacterial growth is determined and measured (see above), but it also has to do with the fact that growth rates of most microorganisms in the environment are much lower than what we are used to from very common laboratory bacteria such as Escherichia coli (Fig. (Fig.3).3). Incubation times can reach several months before formation of colonies or even microcolonies can be observed (109, 184, 249). Some cultures and consortia grow so slowly that it takes years before visible biomass is observed, with an anaerobic hydrocarbon-degrading methanogenic consortium (248) being one of the slowest-growing laboratory cultures reported (Fig. (Fig.3).3). In this case, microbial activity during the first year of cultivation of this consortium had to be inferred by measuring production of methane, and slight changes in turbidity (optical density at 600 nm of ~0.1) could be observed only after 3 years of incubation (248). The success of cultivation of organisms with such low growth rates depends not only on the patience of the researcher but also, even more importantly, on “good timing” for the cultivation steps (Fig. (Fig.3).3). While some microbes can be isolated and grown in a matter of hours or days, others require incubation times that range from months to years. It is known that certain microorganisms are specifically adapted to slow growth (8) and have developed an advantage for, for example, avoidance of lysis by phages (215). Microbial growth is largely dependent on the Gibbs free energy available and maintenance energy requirements of the organism (185). However, syntrophic cultures, which survive on maintenance Gibbs free energies that are much lower than the theoretical values, have been studied in the laboratory (190). There are even reports of microorganisms in subsurface sediments that make a living with maintenance energies that are orders of magnitude lower than minimum values obtained from laboratory-derived experiments. Estimates of doubling times and resulting community turnover for subsurface microorganisms are between 100 and 2,000 years (18). Clearly, microorganisms with such growth rates are not suitable for any classical kind of cultivation experiment. So far it is not known whether all members of these communities can adapt to more rapid growth if provided with sufficient nutrients in the laboratory.
It is important to be aware that comprehensive studies in microbial ecology are not restricted to members of the Bacteria and Archaea alone but have to include all members of the microbial community, such as fungi, protists, and viruses. Although not discussed in this review, it has been recognized that viruses can substantially influence bacterial and eukaryotic (protist and metazoan) host metabolisms, which consequently has broad implications for the environmental “fitness” of these populations (21, 234). Viruses are the most abundant biological entity in nature, and several sequence-based surveys have revealed an enormous viral and phage diversity (180). However, most diversity surveys have focused exclusively on bacteria and/or archaea. A comprehensive study comparing the ratios of microbial and viral populations based on community-wide data has only recently been performed (50). Given that studies of viral and bacterial diversity are already challenging (50, 98, 180), eukaryotic microorganisms (fungi as well as protists) are often ignored as members of the microbial community in most surveys. Fungi, which play an important ecological role particularly in soil environments, are key players for the decomposition and cycling of nutrients, and therefore their activity is directly linked to bacterial nutrient cycles. Although mycologists have been collecting and growing fungi for several hundred years, the ~80,000 fungal species described so far represent only about 5% of the estimated total diversity (89). Another group of microorganisms that was described by Hooke and van Leeuwenhoek in the 17th century are the protists, which represent highly diverse taxa of single-cell eukaryotic organisms that play an essential ecological role in any aquatic environment (33, 111).
Environments are characterized by both their biological and physicochemcial parameters. Ideally, a comprehensive analysis of any environment includes the spatial and temporal assessment of the composition of its viral, prokaryotic, and eukaryotic communities and their interactions (63) along with high-resolution measurements of the environmental parameters shaping these communities—a daunting task. To make it even more complicated, we know that the phenotype of a cell, for example, is determined by subcellular components. To understand the various forms of interactions between organisms and between living cells and their abiotic environment, it becomes necessary to elucidate parts of their subcellular (i.e., molecular) diversity. Such a comprehensive analysis bridges 17 orders of magnitude (147), from molecular interaction at the DNA level (10−10 meter) to single cells (10−6 meter) to whole environments (from local [101 meters] to global [107 meters] scales). In addition to spatial scales, temporal scales of similar orders of magnitude have to be accounted for as well. Processes range from molecular events that happen in microseconds (10−6 s) to growth rates on the order of thousands of years (1011 s). Data obtained at the lowest level (DNA interaction) can often not directly be used to explain processes at the highest level (environment) because there is insufficient information available on how various data sets can be linked and integrated quantitatively as well as qualitatively. For example, a quantitative, genome-wide correlation between transcription and translation in vivo does not yet exist because many of the underlying principles, such as transcription and translation efficiency, are yet not fully understood. It is also unclear at the moment to what extent this correlation varies between different organisms (43). As a result, quantitative interpretation of processes on a community level using transcriptomic and proteomic data remains challenging. The cell as the defining entity in these studies can therefore provide a valuable stepping-stone to bridge this gap (see below). This stepping-stone becomes especially useful when multiple data sets are being integrated in a rational manner.
One example of molecular diversity that can define and shape microbial diversity is bacterial communication. When microorganisms are physically separated from each other, this communication can be hindered, resulting in unsuccessful cultivation attempts. Any cultivation technique that excludes cell-to-cell communication is eliminating a part of molecular diversity that might be critical to the growth of a particular microorganism. Efforts have been made to simulate this kind of communication in the laboratory, for example, by adding signal compounds (22, 23, 26, 79, 80) or by keeping the microbial community as a whole intact (114, 249). Intercelluar communication is widespread within microorganisms and represents the foundation for several aspects of growth and physiology, cell cycling, molecular clocks, and oscillation. Microorganisms “speak a variety of languages” using molecules which have distinct molecular structures, such as acyl-homoserine lactones, γ-butyrolactones, 3-hydroxy palmitic acid methyl ester, quinolones, autoinducer-2, and cyclic dipeptides and peptides of various lengths (for a review, see reference 227). Communication can have multiple effects, e.g., “silencing” of competitors (132) and induction of growth as well as death (143, 154). The various signals can promote one-way, two-way, and multiple-way communications and are not limited to intraspecies communication (e.g., formation of fruiting bodies or growth induction) but also take place between bacteria of different phylogenies as well as between members of different kingdoms (for example, between bacteria and eukaryotes) (15, 172, 174). Crucial to all kinds of communication is that the signals are easily perceived. It makes perfect sense that organisms react not only to environmental signals such as nutrient supply (e.g., by two-component systems) but also to signals from “friend or foe” to gain a competitive advantage. In order to screen for suitable growth conditions, bacteria have to monitor their environment closely in order to switch from a dormant to an active growth state. This is also highly relevant for the cultivation of microorganisms in the laboratory. Signals that trigger this change of state in microorganisms can include the availability of nutrients as well as substances released by other growing organisms that function as signals, such as peptidoglycan fragments or proteins (148, 191). Recently, another hypothesis has been formulated by Slava Epstein (56), which proposes a stochastic awakening of cells. In contrast to previously described processes, this change from a state of dormancy to a state of active growth would not require any signal molecule but rather would be stochastic. An awakening of dormant cells is assumed to result from random bursts (noise) in transcription or translation (171). Such stochastic events that trigger changes in the phenotype have been recently described at a single-molecule/single-cell level for E. coli (30, 36, 53, 75, 245). It has been shown that even a single mRNA copy within a single cell can lead to bursts in protein expression and that therefore not only transcription but also posttranscriptional effects are responsible for stochastic protein expression profiles (34). It has been demonstrated that the random dissociation of a single protein molecule (repressor) from the DNA can result in large bursts of protein expression in E. coli, ultimately determining the cell's phenotype (36). In principle, the awakening of dormant cells by random molecular events therefore seems possible. However, although this random switch from a dormant into an active state is intriguing, this theory might have limited use for K-strategists (bacteria that are adapted to slow growth in nutrient-sparse environments ), as pointed out by Peter Janssen (106). It is also important to keep in mind that not every cell of a clonal population, especially in a heterogeneous environment, will encounter identical conditions. Natural environments cannot be compared to our usual laboratory setups where nutrients are being kept evenly distributed by shaking or stirring. A single molecule of some kind (a signal molecule as well as an electron acceptor or donor) might be absorbed by a single cell and initiate a transcriptional cascade, yet the genetically identical cell next to it will not be exposed to this molecule and therefore will not set off a similar transcriptional response eventually resulting in cell division and growth. Natural (clonal) microbial populations are not synchronized (see below), which represents a huge ecological advantage. Concentrations of nutrients as well as signal molecules can be significantly different on a single cell-level, especially at ultralow (single-molecule) concentrations. The awakening of single cells in their natural environment could therefore still be a response to an environmental trigger, e.g., a few nutrient molecules that are sufficient enough to initiate a transcriptional response. Evaluating the difference and effects of random and nonrandom responses in a natural population will be essential to our understanding of dynamic microbial processes. This also reemphasizes the necessity for high-resolution and high-sensitivity measurements of environmental parameters to accompany microbiological studies.
Estimates about microbial diversity have changed drastically over the last decade, mainly due to advances in detection, computational modeling methods, and algorithms applied and the changing concept of what defines a microbial species (100, 138, 168, 196, 223). For example, the numbers of different species or operational taxonomic units (182) within a soil sample vary from below 500 (99) to 2,000 (186), 10,000 (218), 21,000 (179), and half a million (51) and even up to nearly 107 cells (24, 71). Part of this discrepancy is due to the diversity of the samples themselves (70, 129, 179); another part is likely due to the varying approaches used to estimate the diversity (167, 168). When analyzing diversity over temporal and spatial scales (see above), it is essential that a marker that is relatively stable and is not mutating rapidly under perturbation or environmental pressure will be used. Some molecules at the very center of biology, which, for example, translate genetic information into proteins or convert energy in the form of ATP, are less likely to be subject to rapid mutations and therefore could serve as an evolutionary marker molecule (177). This concept was first perceived by George Fox and Carl Woese and consequently resulted in the use of sequence information (rRNA gene sequences) to study phylogeny and evolution (66, 238, 240). Their discovery was paramount to our current understanding of phylogenetic relationships and the evolution of life on our planet and led to the description of the three domains of life (239).
The use of a phylogenetic marker enabled the discovery of thus-far-undiscovered forms of microbial life. Early studies by Norman Pace and coworkers paved the way for the discovery of the endless microbial diversity that we know today (48, 73, 199, 200). Since then diversity studies have been carried out in almost every imaginable environment, leading to the discovery of a microbial world that dominates the biosphere but is (in most cases) impossible to sample properly (167, 241). The 16S rRNA gene, which today represents the basis for microbial ecology studies, is hence a perfect molecule to study phylogeny, evolution, and molecular diversity. It also allows for insightful comparisons of different environments and ecological niches (78). The use of molecular surveys in microbial ecology is yet another example where data from the molecular level are utilized and extrapolated to the cell level. However, without detailed physicochemical parameters from the microenvironment, ideally on the cell level, these data will likely allow only superficial interpretation (137). There currently exists a comprehensive yet still rapidly growing 16S rRNA gene sequence database; however, reports from metagenomic surveys indicate that current 16S rRNA gene primers are not “universal” and that some organisms might be missed by approaches targeting the gene directly (14, 97).
The rRNA is, however, not a good marker when it comes to physiology, since physiology and phylogeny are not necessarily correlated. There are some phylogenetic groups (e.g., methanogenic archaea) where this correlation still holds true, but there are many examples where phylogeny and physiology do not match. An rRNA gene sequence is therefore not well suited to predict the function, i.e., phenotype, of an organism. For identification purposes, new cultures and environmental genome sequences are generally put into context with their ribosomal sequence. At times the phylogenetic context is used to imply that organisms with similar 16S rRNA gene sequences carry out similar functions or have an identical metabolism, but identical 16S rRNA genes do not automatically translate into identical physiologies, identical phenotypes, or identical pathogenicities and similar functions. The phylogenetic context of an organism can only provide a prediction. To what extent the phylogeny matches the predicted physiology has to be ultimately confirmed through experiments. In addition, it is important to keep in mind that the 16S rRNA gene commonly used for these analyses lacks the resolution at the species level (121, 181). I am avoiding here the discussion about how to define a microbial species (146, 176), but in any case it is essential to recognize that even if diversity surveys at the species level become abundant in the future, different strains and subpopulations of the same species can have very different properties (see below). High-resolution diversity surveys therefore would not solve the current dilemma but would move it to another level (from genus to species). This intraspecies diversity has traditionally been assessed by cumbersome DNA-DNA hybridization, multilocus sequence typing, or average nucleotide identity methods (120) requiring large DNA quantities or substantial genomic information. However, novel sequencing techniques, in combination with DNA amplification methods using miniscule amounts of DNA, have already started to replace these efforts (6, 164). Whole-genome sequencing and resequencing recently became broadly affordable and subsequently paved the way for comprehensive comparative genomic studies (244). It is now possible to analyze several genomes of one particular group of organisms simultaneously by comparing whole genomes and identifying shared genes and nonoverlapping sequences—a so-called pan-genome (130, 173, 214). This analysis has revealed a vast genomic diversity within subpopulations (see below). This diversity of course increases substantially when organisms with “almost” identical 16S rRNA genes are included in the study, e.g., E. coli and Shigella. Konstantinidis and coworkers estimated that the pan-genome of E. coli-Shigella spp. would increase by ~300 new genes (equaling ~5% of all genes in E. coli) for every new genome that is sequenced and added to the study (119). If the genomic diversity within a population is expanding, it is likely that the phenotypic diversity increases as well. In addition to adaptation and evolution through mutation (see below), which can have profound impacts on microbial physiology, genetic material is transferred by horizontal gene transfer (HGT) between closely related organisms as well between different kingdoms of life. HGT can be mediated by the exchange of plasmids, transformation, and transduction (68, 82, 188, 208). Since adaptation and HGT take place at different rates and affect different genes in different organisms (67, 177), a direct correlation between phylogeny and physiology (phenotype) is extremely difficult to determine. Although this direct link between phylogeny and physiology is not achievable in most cases, the rRNA gene(s) still represents one of the few markers that allows for cross-kingdom and cross-phylum comparison, which makes it so ideal for diversity assessments and comparison studies. The common question in microbial ecology, “who are out there and what are they doing?,” however, still awaits an answer in most cases. Fundamental steps toward this goal have been made, and recent developments include functional metagenomic approaches (115, 207) and the combination of in situ hybridization techniques with microautoradiography, secondary ion mass spectrometry, or Raman spectroscopy (85, 96, 133, 150, 159, 160).
After microbial strains have been identified in the environment and successfully propagated in the laboratory, the question arises as to how and at what level these strains (and data obtained from them) can be compared to their counterparts in the original environment. Genomes of different strains belonging to the same species can vary substantially (5, 37, 145, 163, 229). A well-studied case is the virulent E. coli O157:H7 strain, for which it is known that the genome not only is around 25% larger than that of the laboratory strain K-12 but also encodes 1,632 proteins and 20 tRNAs that are not present in K-12 (90). Of these proteins, only 10% are assumed to have virulence-specific functions, leaving around 1,500 proteins being strain specific independent of virulence (90). Clearly, this high number of proteins can potentially lead to various differences in cell composition, physiology, and metabolism (12, 81). Sequence diversity within a whole group of strains was first described for the pathogen Streptococcus agalactiae (213). When six strains were analyzed, they shared ~80% of any single genome, resulting in around 20% of each genome consisting of partially shared as well as strain-specific genes (213). Similar trends have been shown for other organisms such as Haemophilus influenzae (94), Helicobacter pylori (77), Prochlorococcus (37), and Sulfolobus islandicus (175). Intraspecies variations (<1% divergent 16S rRNA gene sequences) have also been demonstrated for natural populations of Ferroplasma acidarmanus (4), Prochlorococcus (37), and Vibrio splendidus (217). In addition, vast differences in the locations of mobile genetic elements and clustered regularly interspaced short palindromic repeats within individual organisms and populations have been reported, representing a glimpse into the history of viral encounters for these cells (7, 126, 187). Potential mechanisms that will lead to these variants (ecotypes) in natural populations have been debated in the past. Evidence for both clonal populations (206) and recombining populations (194) has been described (for a review, see reference 230).
Overall, these examples highlight the vast genetic diversity of subpopulations within a single defined species. It was mentioned earlier that part of this genetic diversity can be virulence specific, but how significant are small genetic variations for the physiology and phenotype of different organisms? It has been demonstrated that even relatively small changes in the genome (single-nucleotide polymorphisms [SNPs] as well as insertions and deletions [indels]) can have substantial effects on the phenotype and fitness of the strains (13, 169). Mutations can take place in structural genes influencing membrane fluidity (246); in metabolic enzymes they can reroute carbon and energy fluxes and increase metabolic efficiency (65, 101); and in regulatory elements or in the transcription machinery (e.g., transcription factors and the RNA polymerase or promoter region) they can have effects on transcription/translation speed, transcript stability, and strength of induction and repression of genes (40). It has been demonstrated that organisms adapt rapidly on the molecular level to changing environmental conditions by increasing mRNA expression levels (11, 41, 47, 93). Organisms can also adapt their metabolic capabilities, e.g., utilizing substrates that previously have not supported growth. E. coli, which normally does not grow aerobically with citrate, was able to do so after 31,500 generations, suggesting that the inability to transport citrate aerobically had been resolved (19).
What are the consequences of these findings for studies in microbial ecology? How will this affect the interpretation of data obtained from laboratory cultures as well as environmental genomic surveys? It has been suggested that evolution experiments carried out in the laboratory reflect adaptation and selection patterns found in natural populations (197). One limitation of these laboratory experiments, however, is that environmental changes and selection pressures are manifold and cannot, or can only insufficiently, be simulated in the laboratory. Another limitation is that evolution is communal (224), and therefore experiments with single organisms can only glance at the potential genetic diversity that has been evolved in nature (183). Nevertheless, these laboratory experiments are essential in order to understand the general principles and mechanisms underlying adaptation and evolution.
The immense genome plasticity observed in pure culture experiments also raises some questions about isolation and cultivation efforts. If organisms readily adapt on a genome level to conditions provided in the laboratory, what “kind” of organisms will then be isolated and propagated in the laboratory? While microorganisms are cultivated in the laboratory (e.g., by enrichment cultivation), it is often observed that the culture grows faster and faster after each transfer: the culture has adapted to the conditions provided (41). There are also several reports in the literature where isolated strains could be adapted over time to grow on certain media; e.g., isolates could be propagated to form colonies on agar plates on which the original isolate was unable to grow (35, 44). Adaptation to culture conditions can be due to changes in the transcriptional machinery, but it also can be due to activation of cryptic pathways (84, 123, 149). It is also known that organisms undergo genetic changes over time even when cultivated under the same conditions, resulting in genetically diverse subpopulations (64, 169, 183, 242). In most cases this “genetic drift” will remain unrecognized by the researcher. Does this mean that all the organisms in our culture collections represent genetically altered strains that are well adapted to conditions provided by the researcher in the laboratory but have no direct (genetic and/or phenotypic) counterpart in their natural habitat? It also raises the question of how genetically “pure” the organisms in our laboratories really are, since they could theoretically consist of multiple members of a subpopulation. An effective preservation of laboratory strains at various time points after isolation therefore is crucial for the investigations.
Besides differences in individual cells that are based on the genome or on stochastic bursts in transcription (see above), it has also been known since the late 1950s that cells can inherit different traits on a nongenetic basis just by asymmetric cell division (157). This cell individuality is based on a random distribution of proteins, such as enzymes and efflux pumps, before and after cell division and will lead to cells that are genetically identical but that exhibit different phenotypes because of differences in their cell composition (136, 203, 222). Based on the number of these proteins per cell, these differences in phenotype can be passed on to the following generation(s) and can be regarded as a “molecular memory,” which is different for each individual cell.
Assuming that natural microbial populations are characterized by cell individuality (198) and vast genetic heterogeneity, would a selective pressure applied by cultivation as well as adaptation to laboratory conditions consequently result in the isolation of very similar cultures (on the genome level)? In other words, would cultivation by itself represent such a strong selection that only one or two individuals would dominate all the isolates obtained from one sample? Exactly the opposite seems to be the case. Thompson and coworkers obtained 232 Vibrio splendidus isolates from marine water samples (out of 333 total isolates), which exhibited large heterogeneity throughout their genomes (217). Similar trends have been reported for bacteria from low-diversity environments (4, 9). However, isolation of strains at different times, even from these low-diversity environments, will not necessarily result in identical strains. For example, two strains of Salinibacter ruber (M8 and M31), which were used for describing this species, have not yet been reisolated from the same environment (J. Antón, personal communication). The natural population, on the other hand, can be quite stable; almost identical environmental sequence types were found at the same site even after a greater-than-4-year time interval in sampling (4). Targeted reisolation of identical strains (ecotypes) can be difficult, and it remains to be seen if this is due to biases in culturability, unnoticed changes of the habitat, or just the extreme genome heterogeneity existing in these populations.
Analyzing genome heterogeneity in microbial populations involves studying whether substitutions in the genome are synonymous or nonsynonymous, meaning whether the mutation influences coding of the amino acid, thus affecting translation and protein structure (155, 178). It is important to keep in mind that although it is often assumed, synonymous or silent mutations do not always have to be neutral. There are indications that some synonymous mutations can affect the fitness of the organism (113). A reason for this may be changes affecting regulatory elements and transcription. For example, can synonymous mutations have an impact on the promoter strength, resulting in subsequent changes of expression level? Another effect on transcription levels by synonymous mutations is due to changes in codon usage. Also, mutations can be condition specific, meaning that they can be silent under one condition but can have an effect on fitness under another condition. When studying causal effects of genetic heterogeneity in natural populations, it would in principle be possible to specifically target genes of interest from the natural community, study them in greater detail, and by doing so link the genetic diversity to phenotypic diversity. Unfortunately, in most cases it will not be possible to a priori predict what gene mutations will result in what kind of phenotype (101). In addition, changes in the genomes (SNPs and indels) that lead to the described phenotypic differences can be so subtle that they might be undetected by certain sequence approaches, due to error rates in sequencing and insufficient coverage which prevent the identification of these SNPs/indels, especially in the context of a genetically diverse population.
Most studies in microbiology are traditionally performed on a community level. Since the sensitivity levels of those techniques are seldom suitable to work at a single-cell level, millions of cells are needed to perform most experiments (Fig. (Fig.1).1). However, this also means that data are generated from millions, billions, or often even trillions of individual cells. The data therefore represent an average of results obtained from large numbers of individual cells. All these individual cells exist in different stages of their life cycle, since bacterial cultures are generally not synchronized—they are not doing exactly the same thing at any given moment. For most questions, these conditions are suitable and synchronized cultures are not a necessity. Although continuous cultivation in a chemostat (156) provides more constant conditions than batch cultures, the cells are still not in a synchronized stage. Truly synchronized cultures are obtained, for example, by the use of a temperature-sensitive allele of the essential DNA replication protein DnaC (32), which allows the initiation to be synchronized after heat shock treatment (237). Another method is the use of the amino acid analog dl-serine hydroxamate, which induces a stringent response (62). Following the release of the stringent response, the cells initiate replication in synchrony. A method that allows for synchronized cells without genetic (237) or chemical (62) perturbation, is the so-called “baby machine” described by Helmstetter and Cummings (91, 92). This method relies on cells that are affixed to a membrane. When medium flows through the membrane, newly divided daughter cells are released, whereas the parents remain bound to the membrane and produce other cells. All cells released at the same time are at the same growth stage (division stage) (91, 92). These approaches are intriguing, yet they have so far been described only for E. coli, and synchronized studies of other organisms are limited. For most experiments, synchronized cultures will not be available, and we have to be aware not only that our data represent average results from many individual cells but also that these cells exist in different growth stages and that they can differ in their cellular composition (e.g., have different proteins).
In order to circumvent issues of culturability, cell individuality, and genome heterogeneity and plasticity, as well as different growth stages, it can be greatly beneficial to study single cells. However, work with single cells of microorganisms beyond visualization by microscopy is still in its infancy. Individual components of single cells, such as nucleic acids, proteins, fatty acids, and lipids, are present in such low quantities that in most cases a direct measurement is not possible. For example, a single cell contains only a few femtograms of DNA. However, due to recent developments in DNA amplification techniques, it became possible to amplify these few femtograms of DNA from a single cell to quantities usable by standard techniques (45, 212, 252). This allowed for the sequencing of genomes of organisms that had not been cultivated before or from which genomic information was limited to the information obtained by metagenomic surveys (2, 95, 141, 151, 164, 243, 251). Elucidation of molecule classes other than DNA which are not as stable (e.g., mRNA) or which cannot be amplified is still challenging for prokaryotes at a single-cell level (209, 210). Since a genome sequence by itself does not provide information about the current metabolic state of an individual cell, methods to address this question have been developed. Recent studies combined single-cell measurements of non-DNA molecules with time-lapse microscopy techniques to elucidate dynamic capabilities (136). Great progress has been made in elucidating single-cell individuality for pure cultures, since fluorescent tags can be introduced into the genome, allowing for dynamic monitoring of molecules such as enzymes, regulators, and RNA polymerase (28, 30, 203, 222). Elucidation of cell individuality in natural populations or in cultures for which a genetic system does not exist is still much more challenging. Combinations of fluorescence in situ hybridization or halogen in situ hybridization with microautoradiography (133, 160), secondary ion mass spectrometry (85, 150, 159), or Raman spectroscopy (96) are currently being applied successfully to obtain information about metabolism and cell composition in natural populations. An advantage of noninvasive methods, such as secondary ion mass spectrometry, Raman spectroscopy, and Fourier transform infrared spectroscopy (152, 253), is that they allow for downstream processing of the cells (e.g., DNA amplification). Interpretation of single-cell studies for natural populations can be challenging not only because of environmental heterogeneity but also because of the genetically heterogeneous background of targeted cells. Nongenetic differences in cell composition, the nonsynchronized stage of genetically identical cells, and the lack of measurements over various time scales further complicate a comprehensive understanding in many cases. However, it is anticipated that further advancements of these innovative single-cell methods in combination with high-resolution measurements of environmental parameters will allow us to gain detailed insights into microbial communities—one cell at a time.
In general, various methods used in microbial ecology can be grouped into bottom-up and top-down approaches (Fig. (Fig.4).4). Depending on the specific question, different methods allow different avenues to be used to obtain the answers. The overarching goal of all these methods is to understand the role of microorganisms in the environment, meaning microbial interactions and the mutual influence of microbial cells with their biotic and abiotic environments (where the environment could be another organism). For many questions it is suitable to consider the microbial community as well as the environment as a “black box” where physical, chemical, and biological parameters can be analyzed as “bulk,” e.g., by microelectrodes, automated remote sensing, rate measurements, labeling studies, and gene surveys (Fig. (Fig.4).4). These methods have the great advantage that parameters can be measured in situ, which is indispensable if perturbations to the environment are being studied. These “black box” approaches in general do not intend to link specific organisms or individual cells with a specific processes measured. Many of the meta-omics methods, which can be considered to be a kind of “molecular black box” approach, set out to make this link (86, 142, 192). Progress toward this goal has recently been made by combining a metagenomic approach with stable isotope probing, resulting in functional active community data (115, 207). In-depth knowledge about phenotype, metabolism, and transcription and translation on a cellular level can only be inferred by top-down approaches, since direct measurement are so far lacking, but these are likely to become available in the future. In contrast, bottom-up approaches utilize direct measurements performed at the cell level (Fig. (Fig.4).4). These approaches include isolation and cultivation techniques as well as various single-cell techniques, which allow for direct phenotype determination.
Top-down and bottom-up approaches should not be regarded as isolated approaches but instead should be integrated. For example, genome information obtained by a metagenomic approach can help guide cultivation (221), or vice versa, data derived from pure culture experiments can be utilized for comparative genomics to gain insights into the physiology of environmental populations (219). These studies have in common that they used the cell as a stepping-stone (see Introduction) to integrate data that originated from various levels of complexity and at various temporal and spatial scales (Fig. (Fig.4).4). To understand and ultimately predict functions of complex biological systems, large data sets generated by various methods have to be analyzed, an approach that now is referred to as “systems biology.” This analysis undoubtedly involves computational methods for data integration as well as for model building. Computational systems analysis has been successfully implemented for various single microorganisms by building functional biological networks using computational bottom-up (58) as well as top-down (20) approaches, and efforts at integrating both approaches as well as utilizing them for microbial communities are currently being pursued.
In light of the genome-level diversity which trumps phylogenetic diversity (determined by rRNA gene sequences) by orders of magnitude, Nigel Goldenfeld and Carl Woese rhetorically asked “…how valid is the very concept of an organism in isolation?” (74). It is unlikely that we will achieve an in-depth understanding of microbial ecology by cultivating all members of the microbial community or by making inferences about their function in the environment from the information on their phenotypes displayed in the laboratory. Evidently, in most cases, we are not able to cultivate them in the first place. If the cultivation hurdle has been overcome, we often fail to accurately assess what role this particular isolate plays in the environment or even if this microbe has any identical counterpart in the environment. This is due in part to the limited resolution of the 16S rRNA gene and the tremendous genetic diversity of subpopulations of different organisms. It is also unrealistic to assume that we will be able to solve questions in environmental microbiology solely by applying various meta-omics approaches without having comprehensive and experimentally validated databases with which to map and compare them. With next-generation sequencing and automated annotation readily available, we are now in a situation where it can be faster to obtain a complete genome sequence rather than a growth curve (127). Does this mean that we have to describe all biodiversity, cultivate and study all variants, and treat each cell as an individual before comprehensive understanding can be achieved? Not at all—it depends on the level of understanding that is required to answer our research questions. For example, we know that all human physiology is identical: we all use various carbon sources as electron donors and oxygen as a terminal electron acceptor. Sequence information tells us that there are differences within the genomes of individuals, but they do not affect basic physiology. However, we know that these variations can effect pigmentation, tolerance toward certain foods, and susceptibility to diseases and drugs, which explains why different groups of people get easily sunburned, are lactose tolerant, or develop a certain disease that can be cured by a certain drug. Understanding the traits of a group of individuals therefore allows us to recognize the niche of this group and ultimately understand human biology as a whole. This should also be the case for microorganisms; depending on the questions asked, we have to consider common traits of groups of individual cells.
I believe that a comprehensive understanding of microbial communities can be achieved only by the synergy between top-down and bottom-up approaches, with the cell as a junction between them (Fig. (Fig.4).4). The central unit in microbial ecology therefore has to be the microbial cell (Fig. (Fig.4).4). A census alone, no matter how detailed and on what level of complexity it is performed, will not allow for a comprehensive view of a given environment. Counting, even sequencing, all individuals of a certain group of animals (as an example from macro-ecology) will not give us detailed information about their behavior or physiology. Additional information will be needed; e.g., what do these animals live on, who might live on them, and how is their habitat defined? For a microorganism this would mean acquiring detailed knowledge about other microorganisms (bacteria and archaea but also protists, fungi, and viruses) present in that environment as well as performing comprehensive analyses of the ecological niche. Studying the organism in captivity (as a culture in the laboratory) will also not allow us to really understand its role in the environment, but it will enable us to formulate hypotheses and theories, originated from direct measurements, which can be, and should be, tested “in the wild.” Microorganisms cannot be regarded as just the sum of their parts (genome, proteome, and metabolome). Only the rational integration of different data sets (their “parts list”) will advance our knowledge of various microbial phenotypes in the environment. The knowledge of the genome, transcriptome, proteome, and metabolome of an organism does not consequently lead to a systems-level understanding of this microbe; these data have to be assessed in a timely and condition-specific manner and rationally integrated in order to fathom the dynamics of microbial life.
Understanding and predicting bacterial phenotypes involves knowledge about how genetic information is transcribed and translated into proteins. The regulation of this information flow is essential to generate a systems-level understanding, both for a single organism and ultimately on the community level. When studying microbial ecology, it is important to use the cell as a central unit, a kind of stepping-stone, to overcome limitations of individual data represented at various scales of resolution, spatial as well as temporal. A similar concept has been perceived for human biology, where in analogy to microbial ecology, data from various cell types have to be integrated into a whole (the human body). Sydney Brenner, Nobel laureate in medicine, stated, “I believe very strongly that the fundamental unit, the correct level of abstraction, is the cell and not the genome” (unpublished lecture, Columbia University, 2003). In the same way we use molecular data in human biology to understand processes on the cellular level, the organ level, and finally on the level of human physiology, we can use the microbial cell as a central unit for understanding of processes on a community level and finally an environmental level. Assuming that microbial cells are central to completing the link between various forms of diversity for subsequently understanding complex systems, then it is most beneficial to obtain as much information as possible about the cell as a whole. Cultivation of microorganisms, when possible, enables these detailed studies under dynamic conditions and allows us to formulate biological principles and generate a knowledge base, onto which “-omics” data can be mapped and linked to. The rational integration of various data sets obtained by top-down and bottom-up approaches is therefore crucial for any systems-level approach on which we are embarking, from a single cell in the laboratory to whole microbial communities in the environment. Around the turn of the last century, tremendous advances in environmental microbiology that still shape and define our current research were made (16, 236). Today, 120 years later, we are again at a point in time where enormous breakthroughs are being made. These advances are possible not only because of novel technology and methods available but mainly because of interdisciplinary research that bridges the gap between molecules, single cells, and microbial communities. What better time than now is there to be a microbiologist?
I am tremendously thankful to Wiebke Ziebis for fruitful discussions, valuable insights, and critical review of the manuscript. I also thank Marc Abrams and Kenyon Applebee for editorial help.
This work was supported in part by the Office of Science (BER), U.S. Department of Energy, grants DE-FC02-02ER63446 and DE-FG02-08ER64686.