Bioluminescence dynamics in single germinating bacterial spores reveal metabolic heterogeneity

Abstract

Spore-forming bacteria modulate their metabolic rate by over five orders of magnitude as they transition between dormant spores and vegetative cells and thus represent an extreme case of phenotypic variation. During environmental changes in nutrient availability, clonal populations of spore-forming bacteria exhibit individual differences in cell fate, the timing of phenotypic transitions and gene expression. One potential source of this variability is metabolic heterogeneity, but this has not yet been measured, as existing single-cell methods are not easily applicable to spores due to their small size and strong autofluorescence. Here, we use the bacterial bioluminescence system and a highly sensitive microscope to measure metabolic dynamics in thousands of B. subtilis spores as they germinate. We observe and quantitate large variations in the bioluminescence dynamics across individual spores that can be decomposed into contributions from variability in germination timing, the amount of endogenously produced luminescence substrate and the intracellular reducing power. This work shows that quantitative measurement of spore metabolism is possible and thus it opens avenues for future study of the thermodynamic nature of dormant states.

1. Introduction

Some bacteria respond to nutrient depletion by differentiating from growing, vegetative cells into spores that can survive extreme temperatures, radiation and long periods without nutrients, while retaining the ability to germinate and reinitiate growth in the presence of sufficient nutrients [1–3]. One of the most remarkable aspects of spore-forming bacteria is the extreme reduction of their metabolic rate, by at least five orders of magnitude, during sporulation [4]. Metabolic dormancy and resistance allow spores to survive for long times in harsh environments, which typically fluctuate. Dynamically sensing the environment requires energy expenditure [5] and thus is likely difficult for spores given their low or nonexistent metabolism. Therefore, sporulating populations must adapt to fluctuating environments through other strategies, such as phenotypic diversification, or ‘bet-hedging' [6,7].

Significant heterogeneity is observed in populations during both the sporulation and germination transitions. For example, after nutrient limitation, only a portion of vegetative cells commit to sporulation, while other cells either lyse or survive in the vegetative state using secondary metabolites [8]. The time at which a vegetative cell sporulates varies significantly within a population, and this time correlates with the time at which the ensuing spore germinates and in its ability to resume growth in various nutrients [9]. Spores germinate in response to low concentrations of nutrients at variable times [7], and this variability is partially explained by the number of germinant receptors in each spore [10]. Even in the absence of germinants, a small fraction of spores germinates spontaneously [11,12]. These phenomena of phenotypic heterogeneity have been linked to variations in gene expression [8,9,12,13]. However, the remarkable dynamics of metabolism during sporulation and germination have not been measured in individual cells, due to the lack of a suitable technique.

The dynamics of energy metabolism can be measured in single vegetative bacteria using several in vivo methods, most of which are based on fluorescence [14]. These are difficult to apply to spores due to their strong autofluorescence, primarily from calcium dipicolinate, NADH and flavins [15,16]. The intracellular redox potential drives all pathways of energy metabolism other than fermentation [17,18] and has been measured in vivo in single bacterial cells using a custom-fabricated microelectrode coupled to an optical trap [19]. However, this technique requires a specialized device, operates in anaerobic conditions and only measures a small number of cells.

An alternative approach uses the bacterial luminescence system [20]. An enzyme, luciferase, encoded by the luxAB genes, catalyses the light-emitting oxidation of a substrate, luciferin, which can be supplied exogenously or synthesized endogenously by the expression of the luxCDE genes. The oxidation reaction's rate depends on the reduced form of the electron transporter flavin mononucleotide (FMNH₂), whose concentration is determined by the intracellular reducing power [21]. Thus, the rate of light production directly relates to redox potential. In addition, since the emission of light occurs without external illumination, bioluminescent techniques can be considered ‘zero background' and therefore offer improved sensitivity as compared to fluorescent techniques. Here we describe a microscope designed to capture the dynamics of bioluminescence emitted by single cells, which we used to non-invasively assay the metabolic activity of single germinating spores of Bacillus subtilis.

2. Results

2.1. Bioluminescence microscopy of single spores

To date, the bioluminescence dynamics of sporulating and germinating cells have been measured only in populations, typically using a photomultiplier tube for photon counting [22,23]. A B. subtilis strain that expresses the lux genes under a sporulation-specific promoter (P_sspB) produces spores that contain a sufficient number of luciferase enzymes to emit detectable light only upon germination [23]. Prior to germination, spores do not emit detectable light, presumably due to their dormant metabolic state.

The feasibility of measuring bioluminescence dynamics in single germinating spores is determined by the rate of photon emission, which must be sufficiently high to be resolved over measurement noise. This rate increases with higher expression of the luxABCDE genes, as has recently been achieved in B. subtilis by modifying the ribosomal binding sites upstream of each gene [24]. The rate of bioluminescent photon emission at the population level has been reported in relative light units, or RLUs [24,25], which are defined differently for various devices. RLUs can be converted to units of photons per time per cell by calibration with a luminescent standard [26], although this calibration has not been included in the published results. As a rough guide, microscopy of individual metabolically active Escherichia coli, Synechoccocus elongatus, Vibrio fischeri and Vibrio harveyi cells expressing the lux system typically measures approximately 60 photons min⁻¹ cell⁻¹ [27–29]. Our preliminary analysis suggested that a microscope using recent improvements in CCD technology could resolve bioluminescence signals as dim as 1 photons min⁻¹ spore⁻¹ (electronic supplementary material, appendix).

We therefore constructed a custom microscope for measuring the dynamics of bioluminescence in single spores (figure 1a, Material and Methods). The microscope includes an optical path for bright field imaging, which is used to localize spores and measure their refractility, as described below. The setup consists of an inverted microscope with a high-NA 63X objective and a cooled, back-illuminated CCD camera. We used 10 min exposure times, sufficient to resolve germination dynamics while achieving sensitivity to permit measurements from single germinating spores. Over a period of 10 min, Brownian motion displaces a spore by a root-mean-square displacement of approximately 50 µm. To prevent such displacements, we trapped the spores by using hydrophobic glass coverslips. With this method, the displacement of spores over 10 min periods was below the optical resolution of the bright field images (less than 160 nm). The sample was held in a perfusion chamber, which dynamically controlled the chemical environment (figure 1b), and the temperature at the sample was controlled to ±0.05°C by an air-to-air heat exchanger.

Figure 1.

Schematic of bioluminescence microscope and perfusion chamber with example images and time series. (a) Microscope schematic showing three light paths: bright field (blue), luminescence (green) and autofocus infrared (red). For bright field imaging (blue path), the external illumination is focused near the bottom of the agarose pad (yellow slab), where sample cells are sandwiched against a glass coverslip. The sample emits luminescence (green path) in all directions. A temperature-controlled, light-tight box (thick black lines) encloses the microscope. (b) The perfusion chamber flows medium around the agarose pad (yellow) that traps sample cells against a glass coverslip on the chamber bottom, sealed by an O-ring (orange). (c) Bright field image of P_sspB-luxABCDE spores, before the addition of germinant (time 0 h). Highly refractile dormant spores appear dark, i.e. high contrast. The field of view contains approximately 900 spores. Scale bar, 20 µm. (d,e) The subregion bounded by the green box is shown in (d) (bright field) and (e) (luminescence), showing the lack of luminescence signal from dormant spores. Scale bar, 5 μm. (f,g) The same region is depicted 3.5 h after the addition of 100 µM L-alanine, showing spores' typical loss of bright field contrast upon germination (f), and the corresponding increase in luminescence signal (g) that occurs in some cells after germination. (h,i) Time series of bright field contrast (h) and luminescent photon counts (i) of the five representative spores outlined in (d–g). The rapid loss of contrast shown in (h) defines the time of germination, τ_g, for each spore. Coloured cell outlines are dilated by one pixel for visualization.

We initially investigated the dynamics of bioluminescence emitted by single B. subtilis spores as they germinated in response to L-alanine. These spores (JDB3780) expressed the luxABCDE genes under the control of a sporulation-specific promoter. A field of approximately 900 spores was imaged in bright field and luminescence every 10 min for 1.5 h (figure 1c). Prior to the induction of germination, the dormant spores appeared dark, with high contrast, and produced no resolvable luminescence signal (figure 1d,e). L-alanine was added to the perfusion media at a concentration of 100 µM, and the ensuing germination response was recorded for 15 h.

After an initial lag following the addition of L-alanine, most spores showed a significant decrease of contrast in bright field, and many produced luminescence well above the noise background (figure 1f,g). The predominant source of measurement noise was sensor read noise; we found that its contribution could be reduced significantly by applying a temporal low-pass filter with a cutoff period of 71 min (electronic supplementary material, appendix). Sensor read noise has a Gaussian distribution and thus can produce estimated photon counts that are negative [30]. We determined that the technique had a sensitivity limit of 0.4 photons min⁻¹ spore⁻¹ (electronic supplementary material, appendix).

The dynamics of individual spores in the bright field and luminescence images were analysed, resulting in time series of bright field contrast and estimated luminescent photon counts for individual spores (figure 1h,i). Rapid loss of contrast after the addition of L-alanine was evident in most spores, indicating an early event in germination, namely, the release of calcium dipicolinate from the spore [31]. Spores differed significantly in their time of germination (figure 1h) and in their bioluminescence signals (figure 1i).

2.2. Bioluminescence dynamics are synchronized with germination time

In order to compare the measurements in the microscope to an established method, we used a 96-well plate reader to measure the population-averaged bioluminescence dynamics of spores germinating in response to the addition of L-alanine at five different concentrations (1 µM, 10 µM, 100 µM, 1 mM and 10 mM) (figure 2a). Time-lapse experiments were performed in the microscope setup at the same L-alanine concentrations, and the population-averaged bioluminescence dynamics were calculated (figure 2b). The dynamics as measured by both methods shared some features: signal intensity generally increased with increasing L-alanine concentration, from total darkness at 1 µM, to a peak level at 1 mM, then decreased slightly at 10 mM (figure 2c). The L-alanine concentration also affected the temporal patterns of the dynamics: in the measurements performed with the microscope, a peak in the bioluminescence signal was evident approximately 1 h after germination with 10 mM, 1 mM and 100 µM L-alanine, but this peak was not observed following germination with 10 µM L-alanine (figure 2b). Although much less pronounced, this early peak was also apparent in the plate reader measurements after germination with 10 mM and 1 mM L-alanine; with 100 µM and 10 µM L-alanine, no peak was resolved, but a ‘shoulder' at 40 min was apparent in both conditions (figure 2a). The primary difference in the dynamics measured by the two methods was the much higher intensity of the early peak measured in the microscope. We observed similar differences caused by varying the spore density in the plate reader experiments (electronic supplementary material, figure S1), which are likely due to chemical interactions between germinating spores. This effect can explain the differences observed between the dynamics measured in the plate reader, at a density of 3 × 10⁷ spores ml⁻¹, and in the microscope, at a density of less than 10⁵ spores ml⁻¹.

Figure 2.

Bioluminescence dynamics of spore populations germinating in response to L-alanine. (a) The bioluminescence emitted by P_sspB-luxABCDE spores germinating in a range of L-alanine concentrations was assayed in a 96-well plate reader. Each well contained approximately 6 × 10⁶ spores. Lines depict the mean signal over three replicates; the standard deviation was negligible on this scale. (b) The bioluminescence signal ϕ(t) was measured in individual spores germinating in response to various concentrations of L-alanine added at time 0 h. The mean signal over all spores (greater than 600 individuals per concentration) is plotted at each time point. (c) The bioluminescence signals in (a) and (b) were integrated over the period between 0 and 12 h after the addition of alanine to calculate the population-averaged photon counts n, plotted as a function of alanine concentration.

The observed population-averaged bioluminescence dynamics arise from temporal fluctuations in individual spores as well as variability in the population, and these two sources can be separated using the single-spore technique. In particular, individual spores germinate at different times, with large variability observed at low L-alanine concentrations [32]. With automated image analysis, we identified the time τ_g of rapid contrast loss in each germinating spore (electronic supplementary material, appendix) and interpreted τ_g as the time at which each spore started germination. We considered the statistics of germination dynamics by calculating the fraction of the population that had not yet germinated by time t: this is the tail distribution function of germination time, P(τ_g > t) (figure 3a). At L-alanine concentrations of 100 µM and higher, most spores germinated within 1 h of induction and over 99% of spores germinated within 13 h. By contrast, spores germinated much more slowly in response to the addition of 10 µM L-alanine, with only half of all spores germinating within 13 h. At the 1 µM L-alanine concentration, we did not observe any germination for spores up to 82 h after induction.

Figure 3.

Time of germination varies in populations and synchronizes with bioluminescence dynamics. Dormant spores were exposed to five concentrations of L-alanine at time 0 h. (a) The fraction of spores that had not yet germinated by time t, P(τ_g > t), is plotted on a linear scale (i) and a log scale (ii). An offset is added to avoid divergence of zero values on the log scale. (b) The luminescence signal of each spore as a function of time elapsed since the time of germination is ϕ_g(t) = ϕ(t − τ_g), defined for spores that germinated during the acquisition interval. The population average of ϕ_g(t) is plotted for each concentration of L-alanine. Note that no spores germinated in 1 µM L-alanine and so no curve for that condition appears in (b).

The distributions of germination time τ_g reveal significant variability in the population of spores, especially evident in the 10 µM L-alanine condition. To investigate the relationship between this variability and the bioluminescence dynamics, we considered the bioluminescence of each cell as a function of time elapsed since germination: ϕ_g(t) = ϕ(t − τ_g), defined for all spores that germinated during the observation interval. Note that the germination time τ_g was determined exclusively by the bright field images, through a procedure that was independent of the bioluminescence dynamics (electronic supplementary material, appendix). The population averages of ϕ_g(t) in each germination condition indicate significant synchronization of the bioluminescence dynamics with the time of germination, as evidenced by the qualitatively similar temporal patterns in the four conditions for which germination was observed (figure 3b). In particular, we observe remarkable alignment in the timing of a primary peak at t ≈ 40 min, the appearance of a secondary peak between t = 2 h and t = 3 h, and similar profiles of signal attenuation at later times, possibly due to degradation of the lux enzymes (electronic supplementary material, figure S2).

2.3. Bioluminescence varies significantly across spores

Aligning the dynamics to the time of germination τ_g accounted for the occurrence and timing of germination as a source of the variability observed in the bioluminescence dynamics. Significant variability remained in the aligned dynamics ϕ_g(t) (figure 4a). Quantifying this variability is non-trivial, as the data consist of dozens of time points for thousands of spores. We analysed the variability of ϕ_g(t) over spores to answer two questions. First, what are the temporal correlations of the bioluminescence dynamics in individual spores? Second, how much of the variability in bioluminescence is due to variations in the factors that affect the reaction chemistry, namely, the redox potential, luciferin concentration, and luciferase copy number?

Figure 4.

Variability in bioluminescence dynamics over spores. (a) Bioluminescence as a function of time since germination is analysed in terms of variability over individuals. Data include all spores that germinated in response to L-alanine at four concentrations (table 1). The time series of ϕ_g(t) for a randomly chosen subset of 500 cells are plotted as thin black lines. The solid red line indicates the median over the 11 000 cells comprising the dataset, and the dashed red lines indicate the 16% and 84% quartiles. (b–e) Principal component analysis was performed on time series ϕ_g(t) from four experimental conditions: germination buffer only (b), 0.4% sodium fumarate (c), 30 μM decanal (d) and 30 µM decanal plus 0.4% sodium fumarate (e). The dark grey line shows the population average 〈ϕ_g(t)〉 for each condition. The first principal component for each condition explained 30%, 55%, 68% and 58% of the total variance, respectively. The red line shows the loadings of the first principal component, i.e. λ₁^1/2w₁(t), where λ₁ is the largest eigenvalue of the covariance matrix, and w₁ is the first principal component, scaled to unit norm. The average and loadings are plotted on separate axes. The first principal component and the population average were highly correlated in all conditions: defining r = cor(w₁(t),〈ϕ_g(t)〉), r_{buffer−only} = 84%, r_fumarate = 97%, r_decanal = 89%, r_{decanal+fumarate} = 99% (all p-values less than 2.2 × 10⁻¹⁶).

To analyse temporal correlations in the bioluminescence dynamics, we calculated the temporal covariance matrix C(s, t) from the photon counts ϕ_g at times s and t. The temporal covariance measures the degree to which photon counts from a spore change together at two different times. In order to calculate it, the mean-subtracted photon counts at times s and t are multiplied, and this product is averaged over spores: C(s, t) = ⟨(ϕ_g(s) − ⟨ϕ_g(s)⟩)(ϕ_g(t) − ⟨ϕ_g(t)⟩)⟩ (see electronic supplementary material, appendix, for details). This calculation was performed over all germinating spores from the four concentrations of L-alanine for which germination was observed. The covariance matrix was analysed using principal component analysis (PCA). PCA applied to a set of time series in this manner is often referred to as empirical orthogonal function analysis [33] and can be used to determine how differences between time series can be decomposed into a small set of orthogonal temporal modes. The first principal component explained 30% of the variance over individuals, while additional components each explained only a few per cent of the total variance (electronic supplementary material, figure S3). Therefore, we focused on the pattern of correlations described by the first component, which had a temporal structure similar to the population-averaged dynamics (figure 4b). This similarity indicates that the dynamics in most spores were well described by the mean dynamics multiplied by a coefficient specific to each spore, which we call its amplitude. This simple mode of variation was useful for quantifying the contributions from different sources, as described below.

2.4. Bioluminescence variability arises from variations in redox potential and luciferin concentration

Besides the timing of germination, several factors could contribute to cell-to-cell variability in the bioluminescence dynamics. Three obvious candidates are cell-to-cell variability in redox potential, luciferin concentration and luciferase copy number. Although we did not attempt to manipulate variability in luciferase copy number, it appeared to have at most a small effect (see Discussion). We reasoned that the effects of variability in endogenous redox potential and luciferin concentration could be minimized by supplying exogenous factors in excess. We performed an experiment with fumarate added to the germination buffer, in order to drive the citric acid cycle and thus increase the intracellular redox potential. In a second experiment, decanal was added as an exogenous luciferin. In a third experiment, both fumarate and decanal were added to the germination buffer. Spores germinated in response to the addition of 1 mM L-alanine. We calculated the population-averaged bioluminescence dynamics and performed PCA for each dataset independently (figure 4c–e).

In the presence of fumarate, the population-averaged bioluminescence dynamics 〈ϕ_g(t)〉 exhibited two distinct peaks, approximately 40 min and 3 h after germination (figure 4c). These dynamics appear to be similar to those observed in germination buffer without fumarate (figure 4b), with two major differences: the addition of fumarate increased the intensity of the second peak and strongly attenuated the spore-to-spore variability during the first peak, as indicated by the disappearance of the first peak from the first principal component. The addition of exogenous luciferin (decanal), both with and without fumarate, resulted in population-averaged bioluminescence dynamics with a single peak, occurring approximately 40 min after germination, with intensity approximately 10-fold higher than the initial peak observed in the absence of exogenous luciferin (figure 4d,e).

In all conditions, PCA revealed a single significant component of variation (electronic supplementary material, figure S3), and in each condition, this mode's eigenvector was strongly correlated with the average dynamics (figure 4b–e). Thus, the primary mode of variation in all conditions was variability in amplitude. This simple pattern facilitates the quantitation of variability, since the contribution of each spore's signal to the total variability can be summarized by a single variable: the signal amplitude. Thus, we calculated for each germinating spore the total number of photons counted during the 12 h following germination, n_g.

The mean value of n_g varied significantly in different conditions. Therefore, to compare variability in n_g across conditions, we considered the ratio of the standard deviation to the mean, η = σ_ng〈n_g〉⁻¹, also known as the coefficient of variation. We also considered the asymmetry in the distribution of n_g by calculating its skew, γ. The estimated statistics are summarized in table 1. Variability was high for spores germinating in the absence of exogenous factors (η = 160%), and the distribution of n_g had strong positive skew (γ = 3.3). The addition of fumarate significantly decreased the coefficient of variation and the skew (η = 125%, γ = 1.4). The addition of decanal reduced the variability and skew more strongly (η = 55%, γ = 0.6). The addition of both fumarate and decanal had the largest effect (η = 39%, γ = 0.32). We conclude that a large amount of variability in bioluminescence signal observed in spores germinating in the absence of exogenous factors was predominantly due to variability in both intracellular luciferin concentration and redox potential.

Table 1.

Bioluminescence signal amplitude statistics. Columns denote media, the number of spores analysed and the statistics of n_g, the photon counts per spore integrated over the 12 h after germination. The ± signs indicate 1 standard error. The bracketed interval in the coefficient of variation column is the 95% confidence interval.

media	no. spores	mean: 〈ng〉 (photons)	coefficient of variation: η	skew: γ
germination buffer only	11 882	1300 ± 19	161% [154%, 169%]	3.326 ± 0.022
germ. buffer + 0.4% fumarate	1976	2862 ± 81	125% [117%, 137%]	1.449 ± 0.055
germ. buffer + 30 μM decanal	928	18 600 ± 340	55% [52%, 59%]	0.56 ± 0.080
germ. buffer + 30 μM decanal + 0.4% fumarate	284	14 000 ± 320	39% [35%, 43%]	0.32 ± 0.14

The bioluminescence dynamics following germination are influenced by metabolic changes that occur early in germination, including the initiation of electron transport [23,34]. In addition to inducing germination, L-alanine may also fuel energy metabolism [9]. To investigate this possibility, we considered the photon counts n_g from spores germinating in response to four concentrations of L-alanine, in germination buffer (electronic supplementary material, figure S4). The population-average 〈n_g〉 increased twofold as the L-alanine concentration increased from 10 µM to 1 mM, then decreased at 10 mM. The variation in amplitude with L-alanine concentration arose primarily from a subpopulation of the brightest spores (electronic supplementary material, figure S5). The weak, non-monotonic dependence of the bioluminescence signal on L-alanine concentration indicates that exogenous L-alanine does not significantly drive energy metabolism in germinating spores.

3. Discussion

The dormant spore is an extreme phenotypic state that allows the organism to survive extended periods of starvation and drought as well as high doses of radiation. Metabolism decreases to extremely low levels during sporulation, then rapidly increases during germination. Laboratory studies of populations undergoing the sporulation and germination transitions have also revealed large amounts of variability over individuals, suggesting that phenotypic diversification in natural spore-forming populations is a bet-hedging strategy for surviving in fluctuating environments. Thus, it is important to study metabolic dynamics in individual spores.

None of the previously described methods for quantifying metabolic dynamics in single cells is applicable to Bacillus spores. Therefore, we developed a microscopic technique for this purpose, using a bioluminescent reporter for the intracellular redox potential. In this paper, we measured bioluminescence dynamics from individual spores germinating in response to L-alanine, which is both a germinant and a potential nutrient. Comparing the bioluminescence dynamics measured in individuals to the population average, we exposed significant variability in the population. A major source of variability was shown to be associated with differences in the timing of germination. We could remove this contribution from the bioluminescence dynamics by simultaneously using bright field microscopy to precisely determine the time of germination of each spore. The bioluminescence dynamics could then be analysed for its temporal correlations. The analysis indicated that the primary mode of variation was associated with the amplitude of the bioluminescence signal produced by each spore.

By performing experiments at different concentrations of L-alanine, we were able to show that it was not a major source of variation in bioluminescence amplitude, suggesting that metabolism in germinating spores was driven not by L-alanine, but rather by endogenous energy sources stored during sporulation. Through the addition of exogenous nutrients (fumarate) and bioluminescence substrate (luciferin), we were able to decompose the variability in bioluminescence amplitude into two main sources: the intracellular redox potential and luciferin concentration. Both of these sources relate to energy metabolism, as the redox potential drives ATP synthesis, and the biosynthesis of luciferin depends on both ATP and NADPH [35]. Noise in the expression of the luciferase enzyme likely also contributed to the variability in bioluminescence amplitude observed in all conditions. However, this must have been a small effect, as little variability remained after controlling for variations in redox potential and luciferin (table 1). Our results show that we can begin to understand the metabolic transitions between the growing and dormant states at the level of single cells.

Bacillus spore germination has been studied using several other single-cell methods, including fluorescence, phase contrast and differential interference contrast microscopy, as well as Raman spectroscopy [9,10,36,37]. These observations have resolved a series of steps in nutrient-induced germination. After exposure to nutrients, slow leakage of dipicolinic acid (DPA) begins at a time that varies significantly across spores, and the period of slow leakage also varies in duration [36]. These events are followed by rapid DPA release, cortex hydrolysis and swelling of the spore core [37]. Heterogeneity in the timing of DPA release is not explained by variability in the expression of the germinant receptors [10]. Both the timing of germination and the ability to use L-alanine for outgrowth correlate with the timing of sporulation and are influenced by the expression of alanine dehydrogenase [9]. Our experiments using bioluminescence microscopy operate at a relatively low temporal frequency and only resolve a single transition from high to low refractility during germination. The primary advantage of the method is the quantification of energy metabolism dynamics in a large population of germinating spores.

We foresee two important extensions of the current work. First, one can study the variability of metabolic dynamics in fluctuating environments. Vegetative cells can be tracked as they sporulate in response to experimentally imposed starvation, and later germinate and outgrow in response to nutrients [9]. By simultaneously measuring the bioluminescence dynamics in this type of experiment, we can study the correlations between spores' timing of sporulation, timing of germination, ability to outgrow and metabolic state. Environments also fluctuate due to biotic interactions. Bacterial spores contain high concentrations of calcium dipicolinate, a compound that affects resistance and dormancy, but also induces germination in dormant spores [38]. Because spores release large amounts of calcium dipicolinate during germination [39], the local concentration of spores can affect their germination behaviour. Such interactions may explain the differences we observed in the bioluminescence dynamics with respect to spore density (figure 2a,b; electronic supplementary material, figure S1). Interactions between spores and their effects on metabolism could be studied systematically by analysing the spatial statistics of germination timing and bioluminescence.

A second extension is to study the thermodynamic nature of dormant states. The maintenance of order in living matter seems to require energy expenditure, in order to counteract spontaneously occurring degradation [40,41]. Therefore, a central question arises: what are the energetic limits for survival? This question has primarily been studied in microbial communities from low-energy environments, including deep sediments, permafrost and subglacial lakes [42–44]. Metabolic activity in such environments, averaged over the community, has been found to be reduced by 5 to 6 orders of magnitude compared to the state of growth when nutrients are supplied [42,43]. These low metabolic rates are approximately equal to the energetic costs of repairing molecular damage to proteins and DNA at the corresponding temperatures [43,44]. However, these analyses included contributions from cells with metabolic activity that was potentially significantly lower than the population average, including dormant spores [45]. Therefore, spores likely represent the most extreme case of metabolic dormancy in nature.

Most studies of metabolism in dormant spores have not detected activity [46–49]. Extraodinarily sensitive measurements of metabolism in populations of dormant spores using radiochemical labelling of metabolites did detect activity, reduced by over five orders of magnitude compared to the vegetative state [4]. However, the detected metabolism could have been due to a rare subpopulation of spontaneously germinating spores [4,11,12], and therefore can serve only as an upper bound on the rate of metabolic activity in dormant spores. By resolving individual spontaneous germination events, bioluminescence microscopy could achieve unprecedented sensitivity. The intensity of bioluminescence is linear in the concentration of FMNH₂ at low concentrations [50], so its utility as a metabolic reporter should extend to cells with very low metabolic rates, such as dormant spores. A measurement on a single spore at a single time point with signal-to-noise ratio ξ can be averaged over N individuals and T time points to produce a bulk measurement with signal-to-noise ratio ξ(NT)^1/2. For our method applied to germinating spores, ξ is approximately 20 (electronic supplementary material, appendix). This suggests that bioluminescence measurements could achieve sensitivity concomitant with a 10⁵-fold reduction in signal compared to germinating spores by measuring, for example, a population of 10⁴ dormant spores over an interval of 18 days, which is feasible using the setup described here.

A possible mechanism for long-term survival without any metabolic activity is the formation of a thermodynamically metastable state. The glassy state is an example of metastability that has been invoked to explain spore dormancy [51,52], as macromolecules in a glass change conformation extremely slowly [53]. Therefore, a glassy environment could in principle reduce the rate of molecular damage to proteins and DNA, while also reducing the activity of metabolic enzymes and thus causing dormancy. Direct characterization of glassiness in spores is difficult, as the internal structure of the spore is highly heterogeneous [54]. Instead, one can study a feature of spores from the perspective of metastability, namely, the statistics of spontaneous germination and its dependence on environmental conditions such as temperature. This approach should be possible with our technique, which would also quantify possible differences in metabolism between spores germinating spontaneously and those germinating in response to nutrients.

4. Material and methods

4.1. Strains

JDB3780 (168 trpC2 sacA::P_sspB-luxABCDE::cm) was constructed by transforming genomic DNA from RL4921 (PY79 sacA::P_sspB-luxABCDE::cm) into JDB1772 (168 trpC2).

4.2. Sporulation protocol

Spores were prepared in Difco sporulation media (DSM). Single colonies grown on LB agar were used to inoculate 3 ml cultures in test tubes. Cultures were grown with shaking at 30°C for 3 h. Aliquots of 30 μl were used to inoculate 30 ml cultures in baffled flasks. After growing with shaking at 30°C for 36 h, spores were purified by spinning at 5400g for 10 min at 4°C and resuspending in sterile water. Cold water washes were repeated four times, yielding samples that were predominantly refractile spores (greater than 90%). Samples were stored at 4°C.

4.3. Plate reader assay

To assay bioluminescence dynamics in response to various concentrations of L-alanine, dormant spores (JDB3780) were suspended in 25 mM Tris-HCl buffer at pH 8.2, at a density of 3 × 10⁷ spores ml⁻¹. Two hundred microlitres of spore suspension were pipetted into wells of a 96-microwell plate (Nunc 236 108). In order to avoid both signal contamination from neighbouring wells as well as the influence of thermal gradients, wells were filled only in the plate interior, in a checkerboard pattern. Two microlitres of L-alanine solutions at concentrations between 1 M and 1 mM were added to the appropriate wells. Each germinant condition was prepared in three replicate wells. Three wells with 200 µl of buffer were included as blanks. The plate was sealed with a gas permeable, partially adhesive film (Uniscience C20613201), and positioned within the plate reader (Perkin Elmer Wallac Victor² 1420-011). Luminescence measurements were made without an optical filter, normal aperture setting and 10 s integration time. The temperature was maintained at 30 ± 1°C. Between luminescence readings, the plate was shaken orbitally for 10 s with a 2 mm diameter at the slow speed setting. The plate reader assay of bioluminescence in the presence of chloramphenicol was performed using a similar procedure, with spores of JDB3785 suspended at a density of 1.2 × 10⁸ spores ml⁻¹ and germinated with 1 mM L-alanine.

4.4. Reagents and media

All germination experiments were conducted in germination buffer (25 mM Tris-HCl, pH 8.2). Decanal was dissolved at a concentration of 15 mM in a 1 : 1 solution of methanol and water.

4.5. Microscopy

4.5.1. Sample preparation

All sample preparation was conducted with sterile technique at the bench.

Adhesion of dormant and germinating spores to glass coverslips was improved by rendering the glass surface hydrophobic. Coverslips were treated with vapour deposition of hexamethyldisilazane (Alfa Aesar L16519AE) in a sealed chamber for 1 h, followed by baking at 80°C for 1 h. Agarose pads were prepared by pipetting 700 µl of molten 4% low-melt agarose (Fisher BP165-25) between two 22 mm diameter glass coverslips. After 1 h at room temperature, the agarose pad was exposed by removing the top coverslip and dried for 5 min. Prior to spotting the pad, dormant spores were washed twice with water (5400g, 10 min, 4°C). Two microlitres of the spore suspension were pipetted onto the centre of the dried pad, and the spot was allowed to completely evaporate (approx. 10 min). The pad was then separated from the bottom coverslip and inverted onto a hydrophobic glass coverslip (30 mm diameter, #1.5 thickness, Bioptechs). The coverslip-bonded pad was left to evaporate for 30 min, then assembled into the bottom of the perfusion chamber (Bioptechs, ICD with heated lid, gas and media perfusion). Molten 4% low-melt agarose was added to the exposed portion of the coverslip and allowed to gel for 30 min. Creating a contiguous layer of agarose in this manner was found to attenuate coverslip bounce resulting from the perfusion flow.

4.5.2. Perfusion

A peristaltic pump (Cole-Parmer, Masterflex L/S digital miniflex, dual-channel) was used to draw media from a reservoir bottle, through sterile tubing (Cole-Parmer, C-Flex #13, i.d. 0.8 mm), into the perfusion chamber and finally into a waste reservoir bottle. Tubing was connected with standard Luer locks, and the bottle openings were wrapped tightly with Parafilm and covered in aluminium foil. At the beginning of each experiment and whenever the medium was changed (for example, when germinant was added), the flow was primed at 1 ml min⁻¹. Otherwise, the flow rate was 50 µl min⁻¹. Prior to conducting each imaging experiment, the perfusion tubing was sterilized by flowing 100 ml 20% bleach, followed by 100 ml isopropanol, followed by five subsequent bottles containing 250 ml sterile, purified water (Milli-Q). The perfusion chamber was cleaned by rinsing with 20% bleach, isopropanol and distilled water, then air dried.

4.5.3. Imaging

The imaging setup consisted of an inverted optical microscope (Zeiss Axiovert 135), a 63 × 1.4 NA oil immersion DIC objective (Zeiss 1113-108) operated without the Wollaston prism, and a sensitive, back-illuminated, cooled CCD camera (Princeton Instruments, PIXIS 1024B). A collimated, 445 nm LED (Thorlabs, Solis-445C) was used as an external illumination for bright field imaging. For the acquisition of bright field images, the LED was pulsed for 50 ms at low current (40 mA) and was otherwise fully off. Autofocus was maintained via a pupil obscuration method applied to an infrared (780 nm) beam reflected from the glass–media interface (Applied Scientific Instrumentation, CRISP). The defocus detector controlled the focal position through a DC servo motor connected to the mechanical focusing shaft of the microscope (Applied Scientific Instrumentation, MFC-2000). Luminescence images were acquired with 10 min exposure time. The readout was performed at 100 kHz (read noise 3.56 e⁻), high gain (1.00 e⁻ ADU⁻¹), and with 2 × 2 binning. The CCD was cooled to −65°C with a combination of thermoelectric and low-vibration liquid cooling (Princeton Instruments, CoolCUBE II). At this temperature, the dark current was measured at 10⁻³ e⁻ pixel⁻¹ s⁻¹. The microscope and camera were enclosed inside a box made of styrofoam-lined black plastic. Exterior edges and corners of the box were covered with opaque black fabric and taped. A removable styrofoam-lined plastic panel provided access to the microscope and sample holder and was held in place during data acquisition with Velcro straps and metal L-brackets. The temperature within the box was controlled by a thermoelectric device (TE Technology, AC-220), coupled to a PID feedback device (TE Technology, TC-720) monitoring a thermistor placed as close as possible to the perfusion chamber. The interior temperature of the box was maintained at 30 ± 0.05°C. The entire imaging apparatus was supported by a vibration-damped optical table, situated in a temperature-controlled room (28 ± 0.1°C).

4.6. Image analysis and statistics

See electronic supplementary material, appendix.

Data accessibility

This article has no additional data.

Authors' contributions

Z.F. performed the experiments and analysed the results; Z.F. and J.D. conceptualized the study and wrote the manuscript; J.D. acquired funding.

Competing interests

We declare we have no competing interests.

Funding

Z.F. was supported by a grant from Simons Foundation to S. Leibler through Rockefeller University Grant 345430 and J.D. was supported by NIH GM122146.