In situ examination of cell growth and death of Lactococcus lactis

Authors


Correspondence: Henrik Siegumfeldt, Department of Food Science, Food Microbiology, Faculty of Science, University of Copenhagen, Rolighedsvej 30, 1958 Frederiksberg C, Denmark. Tel.: (+45) 35333286; fax: (+45) 35333513; e-mail: hsi@life.ku.dk

Abstract

This study enables in situ studying of the growth and death of a large number of individual cells in a solid matrix. A wild type of Lactococcus lactis and several mutants with varying expression of GuaB was investigated. Large variability in the final size of individual microcolonies arising from clonal cells was observed. However, when growth was averaged over 16 locations in a specimen, the SEM was small and notable differences could be observed between the investigated strains, where mutants with lower expression of GuaB had a slower growth rate. The results show that the slow-growing mutants exhibited a lower fraction of dead cells, which indicate that slow-growing mutants are slightly more robust than the faster-growing strains. The large variability in the final size of individual microcolonies arising from clonal cells was quite surprising. We suggest that the control of the size of a microcolony is, at least partially, related to the actual microcolony depended on phenotypic heterogeneity. These findings are important to consider whenever a solid medium with discrete microcolonies is investigated.

Introduction

The viability of a bacterial cell population has been defined as the proportion of the cells that are capable of multiplication after incubation in standard microbiological conditions (Postgate et al., 1961). Many studies of viability and cell growth of bacteria are carried out by cultivation of bacteria in liquid growth media and measuring the number of viable bacteria as a function of incubation time. As the nutrients and secreted metabolic products are evenly distributed in a liquid system, it is useful for the investigation of standard cell growth parameters, for example cell growth curves.

However, liquid cultures suffer from a few limitations compared with a solid growth medium (e.g. agar). Even a genetically homogeneous microbial culture can exhibit phenotypic heterogeneity (Avery, 2006), where individual cells exhibit slightly different growth rates and stress resistance. On the contrary, microscopical analysis of individual cells on solid growth media can provide details of the growth and consequent fate of the microcolonies that derive from these individual cells.

Apart from plate counts, a number of alternative methods have been developed to assess the viability or death of cells. Generally, these methods rely on demonstration of metabolic activities (Rodriguez et al., 1992; Rahman et al., 1994; Yoshioka et al., 1996), which corresponds to viable cells, or loss of cellular integrity (e.g. membrane integrity) (Reynolds & Fricker, 1999), which corresponds to dead cells. In many cases, these are destructive analysis that requires an extraction step, but Bunthof et al. (2001) showed that the progression of autolysis in a starter culture can be monitored in situ by fluorescently labeling cells and that it was possible to examine the proportion of live and dead cells directly in a cheese. Unfortunately, they only examined few locations in the cheese, and as we demonstrate in this study, very large variations can be found within a sample. We therefore developed a more robust method for following growth and subsequent death of individual cells over time, using automated image acquisition. We also examined whether changes in the growth rate caused by specific mutations had any correlation with the subsequent death rate of cells in microcolonies. To obtain this, mutants of a lactococcal wild type was constructed, which only differed in the levels of guaB [Inosine-5′-monophosphate (IMP) dehydrogenase] expression. The synthesis of GMP (guanosine monophosphate) involves the conversion of IMP to XMP (xanthosine monophosphate), which is catalyzed by guaB (Kilstrup et al., 2005). It has previously been found that reduced levels of guanine nucleotides [both GMP/GTP and (p)ppGpp (guanosine pentaphosphate or tetraphosphate)] could be involved in the regulation of stress resistance (Rallu et al., 2000). In our study, the growth rates were correlated with the guaB level, as well as the death rates. The most surprising finding was the large variability in the final size of individual microcolonies arising from clonal cells.

Materials and methods

Bacterial strains and growth conditions

Table 1 shows the strains used in this study. The mutants were constructed by the use of a PCR product containing the guaB gene fused to a promoter library as described by Solem & Jensen (2002). In all experiments, cells were incubated at 30 °C. Cells were initially grown in M17 (Terzaghi & Sandine, 1975) with 0.5% glucose (GM17) overnight. In all other experiments, SA (synthetic amino acid) medium was used (Jensen & Hammer, 1993) with 0.5% glucose (GSA), supplemented with lipoic acid (2 mg L−1) instead of Na acetate.

Table 1. Overview of the strains used in this article and parameters for investigated strains. The relationship between the generation time in GSA (synthetic amino acid medium) and the guaB (IMP dehydrogenase) level
AbbreviationStrainGeneration time in GSA [min]aRelative guaB levelaSource
  1. a

    Results from Jessing and Kilstrup (pers. commun.).

  2. b

    Not measured, but GuaB100 have the same level as the wild type.

  3. c

    Mutants of Lactococcus lactis subsp. lactis IL1403 with different levels of guaB expression.

IL1403Lactococcus lactis subsp. lactis IL140365(1.00)b 
GuaB100Lactococcus lactis subsp. lactis SGJ162c661.00Jessing and Kilstrup, unpublished data
GuaB41Lactococcus lactis subsp. lactis SGJ126c760.41Jessing and Kilstrup, unpublished data
GuaB25Lactococcus lactis subsp. lactis SGJ140c910.25Jessing and Kilstrup, unpublished data
GuaB14Lactococcus lactis subsp. lactis SGJ156c1040.14Jessing and Kilstrup, unpublished data
GuaB8Lactococcus lactis subsp. lactis SGJ151c1230.08Jessing and Kilstrup, unpublished data

Microscopy

The microscope setup consisted of an inverted microscope (Zeiss, Axiovert 135) with a 40 × Fluar objective (N.A 1.3), a monochromator (Monochromator B, TILL Photonics, München, Germany) with a 75-W xenon lamp for the fluorescence microscopy, an automated stage (MAC 2000, LUDL Electronic Products Ltd., Hawthorne, NY) for recording several image positions in a specimen, and a monochrome CCD camera (Coolsnap fx, Roper Scientific Inc., Tucson, AZ).

The excitation for propidium iodide was 536 nm, with emission collected through a long-pass filter of 610 nm. Propidium iodide only penetrates dead cells and binds to the DNA. The MetaMorph 7.0 software package (Molecular Devices Inc., Silicon Valley, CA) was used to control the monochromator, stage, and image acquisition on a PC.

The overnight culture from GM17 medium was washed and resuspended in same amount of PBS (phosphate-buffered saline, 5.84 g L−1 NaCl, 4.72 g L−1 Na2HPO4, and 2.64 g L−1 NaH2PO4.2H20, pH 7.2). The suspension was diluted in GSA medium (1 : 200) and mixed with propidium iodide (PI, Molecular Probes, Invitrogen, Oregon) to a final concentration of 50 μg mL−1. Three microlitre of the dilution was transferred and spread on the bottom of a well in an Ibidi u-Slide 8-well chamber (hydrophobic, uncoated, ibidi GmbH, München, Germany). After drying for 3 min, 500 μL of an agarose medium (GSA with 1% agarose and 2 μg mL−1 PI molten at 45 °C) was added to each well. This provided a two-dimensional distribution of individual cells between the coverslip (bottom of chamber) and the GSA agar.

A random spot was chosen in each specimen and defined as the starting point, and the automated stage was then used to move to 16 predefined positions in a 4 × 4 grid. At all positions, brightfield images as well as fluorescent images of PI staining were captured. After acquisition, which lasted for 15 min, the chamber was incubated at 30 °C until next acquisition.

Image analysis

Cells growing on an agar surface provide very little contrast, and the brightfield images were therefore processed with the default edge detection method of the free image analysis software ImageJ (version 1.45; National Institutes of Health (NIH), Bethesda, MD; [http://rsbweb.nih.gov/ij/]). This provides a very dark background with white cells as presented in the images, and this also facilitated the quantification of the area covered by cells. The fluorescent images have a strong contrast, and quantification of the fluorescent area was therefore performed directly on these images.

It should be noted that we only quantified the amount of pixels that are covered by cells in each image (i.e. two-dimensional growth). When microcolonies develop in the third dimension (cells on top of each other), this is not quantified. The fraction of dead cells to total cells was obtained by dividing the number of pixels that exhibit PI-fluorescence with the number of pixels that correlates with growth in the brightfield images.

Results

Growth rate

In GM17 medium, there was no difference in the generation time of the mutants (results not shown), but in GSA, where the amount of purines are limited, a clear difference could be observed (Table 1). Strains expressing guaB in a lower level than the wild type (IL1403) had a longer generation time.

Microscopy

A random location on the slide was chosen as starting point, and 16 predefined image positions were then recorded as described in Materials and Methods.

Figure 1 shows the distribution of growth of the wild type, IL1403, at the 15 locations measured on a slide. One position (for IL1403) was omitted due to experimental errors. At 0 hours, we only observe 1 or 2 cells together, not microcolonies. It is observed that the variation within each strain is distinct. For IL1403, the pixels covered after 134 h varied from 60 × 103 pixels in one location to 235 × 103 in another location (Fig. 1 A & O). This is quite diverse, as the inoculated culture should be homogenous. This clearly illustrates the danger of trying to obtain quantitative data from one or very few points in a specimen.

Figure 1.

The growth on IL1403 is calculates as pixels covered by colonies. The growth at different 15 positions (A to O) is shown as individual traces. There are large variations in the area covered by colonies for each position. Examples of large overlapping colonies (A) and microcolonies that have stopped expanding (O) are inserted to illustrate the variations.

Examples of growth and subsequent death of the cells for IL1403, GuaB100, GuaB41, and GuaB8 are shown in Fig. 2. This figure shows images of one location for the strains taken at five different time points. It is interesting to note that the distribution of the dead cells is randomly distributed within a microcolony at 22 and 38 h. For 134 h, more dead cells are observed in the center. We do not observe that cell death spread from a central position to the adjacent cells within a microcolony. In Fig. 2, it can also be seen that for each strain, the final size in microcolonies varies considerably. Particularly for GuaB41 and GuaB100, other pictures showed the same variation for the other strains (results not shown).

Figure 2.

Development of microcolonies (Bright Field) and subsequent death of individual cells (PI) for four strains, IL1403, GuaB100, GuaB41 and GuaB8. The precise acquisition times varied up to a few hours between strains, but for clarity an approximate acquisition time are depicted.

In Fig. 3a, the average growth (of 16 positions) over time can be seen as an increase in the number of pixels that the microcolonies are covering. Although the variation within a specimen is quite high, the standard error means (SEM, n = 16) for the strains are sufficiently small to observe obvious differences between strains (Fig. 3). IL1403 and GuaB100 grow to a higher number and faster than the other strains. GuaB8 exhibits the slowest and least pronounced growth. This is in agreement with the growth rate of the cells found in liquid GSA (Table 1). The area of dead cells after 134 h is also larger for IL1403 and GuaB100 compared with the other strains (Fig. 3b). GuaB41 have a larger area of dead cells than GuaB25, GuaB14, and GuaB8. The area of dead cells seems to correlate with the growth rate as the fastest growing strains also die faster than the other strains.

Figure 3.

Growth measured as pixels covered by colonies (a) and death measured as pixels that exhibit PI fluorescence (b). Each strain is an average of 16 positions with standard error means (SEM, n = 16) shown as vertical bars. IL1403 (closed diamonds), GuaB100 (opened squares), GuaB41 (closed squares), GuaB25 (opened circles), GuaB14 (closed triangles) and GuaB8 (opened triangles).

After 60 h of incubation, IL1403 has not grown for 36 h, GuaB100, GuaB41, and GuaB25 have not grown for at least 20 h, and GuaB14 and GuaB8 have just stopped growth. At this point, approximately 7% of the three mutants – GuaB25, GuaB14, and GuaB8 – are dead, whereas 16% of the wild type, IL1403, 11% of the GuaB100, and 8.5% of the GuaB41 are dead (Fig. 4).

Figure 4.

Fraction of the area of dead cells compared to the area of microcolonies. Each strain is an average of 16 positions with standard error means (SEM, n = 16) shown as vertical bars. IL1403 (closed diamonds), GuaB100 (opened squares), GuaB41 (closed squares), GuaB25 (opened circles), GuaB14 (closed triangles) and GuaB8 (opened triangles).

After 134 h of incubation, 50% of the IL1403 and GuaB100 cells and 41% of the GuaB41 are dead, whereas only 20–23% of the GuaB25, GuaB14, and GuaB8 cells have died. The small SEM in Fig. 4 indicates that each strain has a unique fraction of dead cells after 134 h. As the final sizes of microcolonies are very variable within each strain, this could indicate that a low fraction of dead cells is a unique feature of slow-growing cells.

Discussion

By microscopy, we were able to follow the growth and death of individual cells on a solid surface. Large variations were observed within each strain, and, for example, the final size of individual microcolonies varied greatly. However, when growth was averaged over 16 locations in a specimen, the SEM was small, and notable differences could be observed between the investigated strains.

The average growth patterns of the six different strains are shown in Fig. 3. The growth rates are clearly dependent on the expression of GuaB, as lower expression of GuaB leads to lower growth rate and lower number of total cells at growth arrest. It is also interesting to note that the mutant, GuaB100, with the same level of GuaB as the wild type, IL 1403, exhibited the same growth behavior (and cell death behavior). This indicates that the molecular manipulation of reintroducing the GuaB gene does not in itself induce differences compared with the wild type.

The results also show that slow-growing strains such as GuaB14 and GuaB8 exhibited a lower fraction of dead cells at the end of the experiment (Fig. 4). An obvious explanation is that the fraction of dead cells increases pronouncedly when a strain is in stationary phase, and the slower-growing strains enter the stationary growth phase at a later time point than IL1403 and GuaB100. However, after 134 h, the three fastest growing strains, IL1403, GuaB100, and GuaB41, had been in stationary phase for several days, and the semi-fast growing GuaB41 still had a lower fraction of dead cells than the fast-growing IL1403 and GuaB100 (Fig. 4). Although a prolonged period in stationary phase therefore increases the fraction of dead cells, there seems to be an additional factor involved in the survival of the slower-growing strains. It was previously found that slower growth rates lead to higher stress resistance (Geisel et al., 2011; Zakrzewska et al., 2011), which corresponds well with our observation that a decreased growth rate leads to a decreased rate of death.

Within each strain, the fraction of dead cells at any given time point is very uniform, which can be seen as small SEM in Fig. 4. As previously mentioned, it is interesting that cell death appears to be randomly distributed within a microcolony (see e.g. Fig. 2). This is clear from the pictures after 22 and 38 h. The apparent concentration of the dead cells in the center of colonies after 134 h can be due to the colony thickness, and the colony does grow in three dimensions. If the cells die from nutrient depletion or increased exposure to toxic metabolites, one could expect that cell death progress from the center of a colony, as the stressful conditions would be more pronounced here. On the other hand, it is possible that cell death is associated with some form of external chemical stimuli from the surroundings, but there is no indication that the cells in the perimeter of a nonexpanding (stationary phase) colony have a higher proportion of dead cells either. Based on these observations, we conclude that cell death within a microcolony seems to occur at random.

Jeanson et al. (2011) studied the spatial distribution and size of bacterial colonies in a model cheese using CLSM and a gel cassette system. An important conclusion from their findings was that the final numbers of cells in the model cheese were identical regardless of inoculation level, that is, low initial inoculation level results in few but large colonies, whereas a high inoculation level results in many small colonies. Unfortunately, their determination of colony sizes, obtained from growth inside a compact cheese matrix, is not directly comparable with our results, obtained from two-dimensional growth on a solid medium surface, but it is interesting that a pronounced variation in the final colony size is clearly detectable in the images of Jeanson et al. (2011). Similarly, a variation in final colony size could also be observed from results of Parker et al. (1998), and although none of the authors comment specifically on this variation, it is obvious that it is a common phenomenon.

It seems unlikely that starvation for nutrients or accumulation of toxic waste products is a key factor in the premature growth stop of some microcolonies. Individual cells situated in the periphery of a microcolony should have a comparably less hostile environment than cells in the center of the same colony, and therefore, any chemically induced growth arrest in the periphery should arise from either nearby colonies or chemical changes that are manifest in the entire specimen. However, other microcolonies in the same specimen expanded, merged with other microcolonies, and ended up almost filling an entire field of view, which suggests that these particular microcolonies were not limited by nutrient depletion or signal molecules from surrounding microcolonies.

In our opinion, this suggests that at least some of the parameters that determine the final size of a microcolony lies within the microcolony itself and is not influenced by interactions with the environment. The picture in Fig. 2 of GuaB41 after 134 h is an example of the difference in colony size within the same location. Distinct microcolonies can be seen, all have stopped expanding, but they have different final sizes. As all specimens were inoculated with the same concentration of cells, the level of inoculum cannot explain the difference, but there seems to be a correlation between late initiation of division of a mother cell and a small final size of a microcolony. We observed that most of the small colonies derived from single cells that started to divide later than those cells that tend to form large colonies. We think that this has to do with the stochasticity in the gene expression, but more studies are needed to confirm this.

A typical inoculum consists of a multitude of clonal cells, and it is somewhat surprising that cells from one microcolony expand over an area, where clonal cells from a neighbor colony have stopped growth. However, phenotypic heterogeneity of individual single cells has previously been observed in genetically homogeneous microbial cultures (Avery, 2006) and has been attributed to stochasticity in gene expression. In our experiment, we speculate that some of the original single cells in the inoculum may have a certain type of stochastic gene expression that gives rise to an ‘early-stop’ phenotype, which may be inherited by the daughter cells and finally form a small colony, whereas other individual cells have another type of stochastic gene expression with a phenotype that results in large colonies. This phenotypic variation may also explain why small colonies originate from cells that are delayed in their initial division. Finally, we also observed that some cells never initiate division, even though they appear viable (PI negative) (an example can be seen in Fig. 2 for GuaB8), and it is possible that this can be explained as an even more extreme delay in the initiation of division.

Although these observations clearly warrant further experiments, to understand the molecular background, we want to emphasize that the reason that such variations have not been extensively documented is that they can only be observed by microscopic techniques as presented in this article. We have also demonstrated in this article that although there may be significant cell-to-cell variations within a population, obtaining an average of many individual cells will reveal differences between strains as would also be observed by more traditional population studies in, for example, a liquid medium. These population averages tend to focus on the most robust/rapid-growing phenotypes within a population, and this is very useful for, for example, the selection of starter cultures with maximum fermentation performance. But as this article demonstrates, slow-growing phenotypes that are normally not detected may exhibit pronounced stress resistance. Therefore, whenever we attempt to understand microbial problems where cell death is involved, for example autolysis of starter cultures in cheese or inactivation of pathogenic bacteria by antimicrobials, a microscopic technique such as presented may provide additional information. In addition, our method is not necessarily limited to transparent media, as fluorescent signals from cells can be detected on opaque surfaces.

Acknowledgements

The work was supported by a grant from the Danish Dairy Research foundation. The authors are grateful to Stine Jessing and Mogens Kilstrup (Department of System Biology, Technical University of Denmark) for supplying the Lactococcus lactis subsp. lactis IL1403 guaB mutants.

Ancillary