This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Lin, A. L. Articles by Swinney, H. L. Search for Related Content PubMed PubMed Citation Articles by Lin, A. L. Articles by Swinney, H. L.

Localization and Extinction of Bacterial Populations under Inhomogeneous Growth Conditions

Center for Nonlinear Dynamics and Department of Physics, The University of Texas, Austin, Texas 78712

Correspondence: Address reprint requests to Anna L. Lin, E-mail: alin{at}phy.duke.edu.

	ABSTRACT

TOP ABSTRACT INTRODUCTION METHODS OBSERVATION OF LOCALIZATION AND... DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
The transition from localized to systemic spreading of bacteria, viruses, and other agents is a fundamental problem that spans medicine, ecology, biology, and agriculture science. We have conducted experiments and simulations in a simple one-dimensional system to determine the spreading of bacterial populations that occurs for an inhomogeneous environment under the influence of external convection. Our system consists of a long channel with growth inhibited by uniform ultraviolet (UV) illumination except in a small "oasis", which is shielded from the UV light. To mimic blood flow or other flow past a localized infection, the oasis is moved with a constant velocity through the UV-illuminated "desert". The experiments are modeled with a convective reaction-diffusion equation. In both the experiment and model, localized or extinct populations are found to develop, depending on conditions, from an initially localized population. The model also yields states where the population grows everywhere. Further, the model reveals that the transitions between localized, extended, and extinct states are continuous and nonhysteretic. However, it does not capture the oscillations of the localized population that are observed in the experiment.

	INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS OBSERVATION OF LOCALIZATION AND... DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
The growth, spreading, and extinction of a population in an inhomogeneous environment are of interest given the global decline in biodiversity. Under what conditions does a population change from one that is localized to one that spreads throughout the available domain? This question is important not only for species in the earth's ecosystem but also for understanding the transition from a localized to a systemic infection in an organism. Simple model systems can provide insight into the kinds of phenomena that can occur in these complex systems. Population growth studies often examine bacteria because of the short doubling time. Studies of bacterial growth in homogeneous environments have revealed the spontaneous formation of different spatial patterns in response to environmental stresses (Shapiro and Dworkin, 1997; Berg and Purcell, 1977; Ben-Jacob et al., 1994; Matsushita et al., 1998). We are interested in the situation more commonly found in nature: inhomogeneous environments.

Our work is motivated by recent theoretical studies of the Fisher equation, which was generalized to model growth in an inhomogeneous environment under the influence of a convective flow. Those studies indicated a transition from localized populations to delocalized populations that grow everywhere (Nelson and Shnerb, 1998; Dahmen et al., 2000; Shnerb, 2001). Experiments have also been conducted on growth in inhomogeneous environments, but the conditions were too complex to compare with the model studies (Neicu et al., 2000; Shnerb, 2001).

We have developed an experimental system for studying bacterial growth in an inhomogeneous environment. Escherichia coli bacteria grow in a long, thin channel with different growth conditions in a small section of it. The position of this different growth region changes in time. We measure the bacterial concentration as a function of space and time in experiments that typically last one week, which is long compared to the half-hour doubling time for the bacteria. The quasi one-dimensional geometry permits comparison with simple one-dimensional models.

The spreading of bacterial colonies measured for certain homogeneous conditions (Shapiro and Dworkin, 1997; Matsushita et al., 1998) in a petri dish has been found to be described by the Fisher equation (Fisher, 1937; Murray, 1989), which has traveling front solutions that propagate with a velocity _F = 2(Da)^1/2, where a and D are, respectively, the growth rate and diffusion coefficient of the bacteria (Shapiro and Dworkin, 1997; Berg and Purcell, 1977; Ben-Jacob et al., 1994). Nelson and co-workers (Nelson and Shnerb, 1998; Dahmen et al., 2000) proposed a generalization of the Fisher equation to describe the growth of bacteria in an environment with a localized region of favorable growth conditions (an "oasis") that moves with a constant velocity :

(1)

The growth rate in the oasis (|x| W/2) is U(x – t) = a, and in the surrounding "desert" (|x| > W/2) the growth rate is U(x – t) = –a, where W is the width of the oasis in a system of length L, as shown in Fig. 1 a. The average growth rate over the full length of the system is

(2)

which we call the average terrain. An analysis of a linearization of Eq. 1 by Nelson and co-workers (Nelson and Shnerb, 1998; Dahmen et al., 2000) suggested that qualitatively different spatial distributions of concentration would occur for different values of the oasis velocity and the average terrain (Eq. 2). Numerical simulations (Dahmen et al., 2000) of the full nonlinear equation suggested that these qualitative changes occur not over a range of parameter values, but rather at particular thresholds of these two parameters; however, this was not investigated in detail.

View larger version (13K): [in this window] [in a new window]   FIGURE 1  (a) Growth profile for a system with an oasis (width W and growth rate a) that moves with velocity through a desert (width L – W and growth rate –a). (b) Example of a localized colony after 69 h of growth, from a numerical simulation of Eq. 1. The colony density is shown by the solid black line and the shaded band depicts the oasis. Simulation parameter values: /F = 0.9, a = 4.76 x 10–4 s–1, D = 6 x 10–5 cm2/s, = 0.7, W = 2 cm, L = 50 cm, and b = 7.93 x 10–11 cm/s. Concentration c(x, t) is in arbitrary units.

For || < _c, there is a range of values of the terrain for which the population remains localized, growing in the oasis while decaying exponentially with distance from the oasis. An example of such a localized state is pictured in Fig. 1 b. The population is sustained in the moving oasis but dies out in the desert.

Finally, for a sufficiently rich terrain, the population grows everywhere, for any value of the oasis velocity . No matter how the population is initially distributed throughout the length of the system, the population becomes delocalized.

The identification of the distinct sustainable localized, delocalized, and extinct states, and the suggestion that there could be well-defined transitions among them, originally resulted from an analysis of a linearized version of Eq. 1, which disregards the biologically important saturation term (Nelson and Shnerb, 1998; Dahmen et al., 2000). Transitions between the different regimes can be understood only by including the nonlinear saturation term. We have performed numerical simulations of the full nonlinear equation to compare with the laboratory observations and to study the transitions among the localized, delocalized, and extinct states.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS OBSERVATION OF LOCALIZATION AND... DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
We examined bacterial growth in a rectangular channel of length L = 25 cm, thickness 0.2 cm, and height 2.5 cm. This thin quasi one-dimensional geometry was chosen so that the observations could be compared to predictions of one-dimensional models (Dahmen et al., 2000). Most of the 25 cm long channel was exposed to ultraviolet (UV) light to inhibit bacterial growth, but inside this "desert" was a short (W = 3 cm) "oasis" where the UV light was blocked by a mask (see Fig. 1 a).

The species selected for this study was E. coli RW 120 because, for these bacteria, UV exposure results in cell death rather than mutation (the two mechanisms for repair of DNA are switched off) (Ho et al., 1993). Thus, as in the model, the population dies out in the desert if it is sufficiently harsh.

The rectangular channel was filled before each experimental run with a define media in a 0.2% agar gel maintained at 37°C; this nutrient level supported steady bacterial growth for at least 1 week. The purpose of the gel was to inhibit uncontrolled convection while allowing diffusion of the bacteria. A 120 cm long mercury lamp positioned 8.5 cm above the cell provided 10 W/m² UV illumination for the region not shielded by the mask.

Each run was initiated by dragging a needle that had been dipped into bacteria across the gel surface underneath the mask, thus infecting only the oasis region. We waited several hours for the bacteria to reach a saturated concentration c_s = a/b in the oasis before starting to move the oasis with a constant velocity , as in Fig. 1 b. Runs were made for oasis velocities ranging from 1.5 x 10^–5 cm/s, where the oasis moved only 13 cm in an experiment lasting 10 days, to 3.6 x 10^–4 cm/s, where the oasis traversed the 25 cm cell length in 19 h. At high oasis velocities, the original population could not keep up with the moving oasis, and the population decreased and became extinct; we continued to monitor the decline in the original population for several days after the oasis reached the end of the cell. During a weeklong experiment, most of the live bacteria remained close to the (oxygenated) air-gel interface, whereas dead bacteria sunk to the bottom of the channel in 2 days.

The bacterial concentration near the gel surface was determined by measuring the transmission of light through the thin dimension of the cell. The light source was a 4.5 mW diode laser with a 670 nm wavelength and a 1 mm beam diameter. The light source and a detector were mounted on a translation stage with the center of the beam passing through the gel 1 mm below its surface. Each spatial scan of the bacterial concentration took 7 min, a time short compared to the time for the concentration to change significantly; thus the measurements yielded c(x, t) at all positions x and times t for the duration of an experiment (3–10 days).

Experimental determination of the model parameters
We have measured the parameters in the model: the growth rate a of E. coli in 0.2% agar define media, the growth attenuation coefficient of the UV light, the diffusion constant D, and the saturation constant b.

The growth rate of RW 120 was determined using a standard dilution plating technique (Runyen-Janecky, 2001, private communication) in which the number of bacteria per unit volume was counted every half hour. The measurements, shown in Fig. 2, yielded a = 4 x 10^–4 s^–1.

View larger version (9K): [in this window] [in a new window]   FIGURE 2  Experimental determination of the growth rate a ("#bacteria" is the number of bacteria per 100 µml).

The diffusion constant D was determined from measurement of the radially averaged front velocity _F of a colony growing in 0.2% agar under homogeneous conditions in a petri dish, assuming that (Murray, 1993; van Saarloos, 1987, 1988). For bacteria growing in a petri dish with 0.2% agar in define media, we measured _F = 2.4 x 10^–4 cm/s. This result together with the measured value of a yields D = 7 x 10^–5 cm²/s.

A steady-state concentration c_s was achieved with point inoculation of bacteria in the oasis and no incubation period for /_F << 1. A value for the saturation parameter b, 7.93 x 10^–11 cm/s, was determined from the relationship b = a/c_s.

We also conducted some measurements in 0.1% agar Luria Broth media where a and D were larger. These measurements yielded a = 6 x 10^–4 s^–1, _F = 4.4 x 10^–4 cm/s, and D = 1.6 x 10^–4 cm²/s.

Simulation
We numerically integrated Eq. 1 using the parameter values given in the previous section, except for , which was scanned over a range of values in the simulations (see Fig. 5). The value of _F used in the simulations was typically _F = 3.38 x 10^–4 cm/s, the average of the values measured in the two different growth media. In the simulation, space was discretized using the method described in Dahmen et al. (2000) and Franosch and Nelson (2000). The resulting N-coupled ordinary differential equations (where the number of grid points N was typically 2032) were integrated in time from an initial condition of a Gaussian concentration profile with 10% noise centered in the favorable growth region (Mann, 2001). An alternative initial condition, a spatially uniform concentration with the saturation value a/b, also with 10% noise, yielded the same behavior at long times. Both periodic and reflective boundary conditions were found to yield extinct and localized solutions, whereas delocalized solutions were found only with periodic boundary conditions.

View larger version (56K): [in this window] [in a new window]   FIGURE 5  Phase diagram computed for the model, Eq. 1, showing the total number of bacteria in the system as a function of the two control parameters, the average terrain and the oasis velocity /F. There are three sustainable states: localized, where the bacterial colony survives only in the moving oasis; delocalized, where the bacterial population grows throughout the system, although the growth is largest inside the oasis region (see Fig. 7); and extinction, where the bacteria die out everywhere, even inside the moving oasis. Note that the diagram is symmetric about /F = 0. The parameter values for the simulation were: a = 4.76 x 10–4 s–1, b = 7.93 x 10–11 cm/s, D = 6 x 10–5 cm2/s, W = 3 cm, L = 20.32 cm, and F = 3.37 x 10–4 cm/s.

	OBSERVATION OF LOCALIZATION AND EXTINCTION

TOP ABSTRACT INTRODUCTION METHODS OBSERVATION OF LOCALIZATION AND... DISCUSSION ACKNOWLEDGEMENTS REFERENCES

View larger version (25K): [in this window] [in a new window]   FIGURE 3  Bacterial colony evolution in (a) experiment and (b) numerical simulation of Eq. 1 for oasis velocities leading to extinction (left column), slow extinction (middle column), and a sustained population in the oasis (right column). The shaded boxes show the location of the oasis (3 cm wide in a sample of length 25 cm); in the left column the oasis moves out of the region shown, leaving behind the population (which initially developed in a stationary oasis) exposed to the deadly UV radiation. The same thing happens in the middle column, but more slowly. The experimental measurements cannot distinguish between live and dead bacteria, but the results from the simulation show the concentration c(x, t) of bacteria living at time t as dashed lines, whereas solid lines show the total concentration of living and dead bacteria; thus at t = 2 h almost all of the bacteria are already dead in the simulation (left column). Other simulation parameters are given in Fig. 5. For the simulations, the bacterial concentration varies from 0 to 1.97 a/b, 2.74 a/b, and 10.73 a/b in the left, middle, and right columns, respectively.

In the case shown in Fig. 3 a, the bacteria inoculated in the oasis were first allowed to grow for 13 h with the oasis at rest before it began to move at 3.6 x 10^–4 cm/s. The bacteria did not diffuse and grow fast enough to maintain a population within the moving oasis, and the original bacterial colony that was left behind became fully exposed to UV light. The bacteria then died and slowly sunk to the bottom of the channel, out of the field of the measurement of concentration; at 73 h some (presumably dead) bacteria were still present, but the population was declining toward zero. The waviness in the concentration profile is due to the inhomogeneous inoculation of the gel (see t = 3 h).

The development of stable localized populations in a moving oasis is illustrated in the column on the right of Fig. 3. The bacterial colony moves in the same direction as the oasis motion, and at least part of the colony remains in the oasis; bacteria that are left behind in the desert eventually die.

The spatiotemporal behavior of the colony under conditions close to the extinction-localization transition is shown in the middle column of Fig. 3. Close to this transition we observe in the experiment a double-peaked oscillating front, as can be seen in Fig. 4. These oscillations persist for many generations. The parameters in the simulations of the model Eq. 1 have been varied widely, but no oscillations have been found.

View larger version (37K): [in this window] [in a new window]   FIGURE 4  Snapshots taken hourly in the experiments reveal spatio-temporal oscillations of the bacterial concentration. The shaded boxes show the oasis position. The velocity of the oasis was = 1 x 10–4 cm/s, and the experiment was conducted for a 0.1% agar gel with Luria Broth media.

View larger version (17K): [in this window] [in a new window]   FIGURE 7  The asymptotic spatial distribution of an extended (delocalized) bacterial population computed in the reference frame of an oasis moving with different relative velocities /F. The average terrain was held fixed, U/a = 0.424, D = 1.2 x 10–4 cm2/s, F = 4.78 x 10–4 cm/s, a = 4.76 x 10–4 s–1, = –0.4, W = 4 cm, L = 1000 cm, and b = 7.93 x 10–11 cm/s.

View larger version (21K): [in this window] [in a new window]   FIGURE 6  (a) Transition from a population localized in the oasis to extinction of all bacteria as the normalized oasis velocity /F was increased in a simulation of Eq. 1; the average terrain was held fixed, The inset illustrates critical slowing down as the transition at = c was approached from above: the relaxation time diverges, ( – c)–, with = 0.98 ± 0.03. Similar results were obtained when approaching the transition from below. (b) The transition between extinction and delocalization with fixed /F = 2, as a function of (or, as shown along the top of the graph, as a function of the attenuation factor ). Other parameter values are given in Fig. 5.

For large enough, the average terrain experienced by each bacterium was sufficient to support life everywhere; the population, initially limited to the oasis, became delocalized. However, the population remained larger in the oasis region, as Fig. 7 illustrates for a range of increasing oasis velocities.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS OBSERVATION OF LOCALIZATION AND... DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
A system's response to changes in local and global conditions is part of the knowledge required for applications such as achieving precision drug delivery, e.g., localizing a viral vector in its target cell during gene therapy. The moving oasis could correspond to a flow of plasma past a fixed region of favorable growth conditions.

Our study has concerned bacterial populations that initially grew in a stationary oasis that lay within a desert; then the oasis was made to move with a constant velocity. We found that the population could be sustained or could die out, depending on the oasis velocity and the average terrain, which depended on harshness of the desert and the size of the oasis relative to the size of the desert. Results from our numerical simulations of Eq. 1 are in qualitative accord with the laboratory observations, although the model fails to capture the observed double-peaked oscillations found for the spatial distribution of bacteria in the moving oasis. An alternative model might include (in addition to diffusion, convection, and growth) the internal dynamics of various gene expression networks within individual bacteria. Such a model could produce oscillations as the protein levels change in response to environmental stresses. Temporal oscillations of various proteins within cells have been demonstrated in E. coli (Elowitz and Leibler, 2000). Similar temporal oscillations have been captured in a minimal model based on activator-repressor dynamics (Vilar et al., 2002).

The analysis of the linearized equations by Dahmen et al. (2000) revealed a transition between delocalized and mixed steady states for _c = 0.92 _F, whereas our numerical analysis of the full nonlinear equations, conducted in the range of biologically relevant parameters, yielded only sustained delocalized states, i.e., nonzero concentration everywhere (see Mann, 2001).

In conclusion, our numerical simulations of the model yield deeper insights into the mechanisms governing the colony behavior than could be obtained from the experiments alone. In particular, the simulations showed that the transitions between the different states were continuous and nonhysteretic. Further, the simulations yielded a delocalized state, where the bacterial population grew throughout the entire system; this state was not accessible in the experiments.

	ACKNOWLEDGEMENTS

TOP ABSTRACT INTRODUCTION METHODS OBSERVATION OF LOCALIZATION AND... DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
The authors thank Laura Runyen-Janecky for supplying the RW 120 E. coli bacteria and for her help, and Nicolas Perry for his help with experiments. They also thank Karen Dahmen, David Nelson, N. M. Shnerb, Andrea Bertozzi, and Nitant Kenkre for helpful discussions.

This work was supported by the Engineering Research Program of the Office of Basic Energy Sciences of the Department of Energy and the Robert A. Welch Foundation (A.L.L., B.A.M., and H.L.S.), and by the National Institutes of Health and the National Science Foundation (G.T., B.L., and J.K.).

	FOOTNOTES

 
Anna L. Lin's present address is Center for Nonlinear and Complex Systems and Dept. of Physics, Duke University, Durham, NC 27708.

Submitted on October 9, 2003; accepted for publication March 15, 2004.

	REFERENCES

TOP ABSTRACT INTRODUCTION METHODS OBSERVATION OF LOCALIZATION AND... DISCUSSION ACKNOWLEDGEMENTS REFERENCES

 
Ben-Jacob, E., O. Shochet, A. Tenebaum, I. Cohen, A. Czirok, and T. Vicsek. 1994. Genetic modeling of cooperative growth patterns in bacterial colonies. Nature. 368:46–49.[CrossRef][Medline]

Berg, H. C., and E. M. Purcell. 1977. Physics of chemoreception. Biophys. J. 20:193–219.[Abstract/Free Full Text]

Cross, M. C., and P. C. Hohenberg. 1993. Pattern formation outside of equilibrium. Rev. Mod. Phys. 65:851–1112.[CrossRef]

Dahmen, K. A., D. R. Nelson, and N. M. Shnerb. 2000. Life and death near a windy oasis. J. Math. Biol. 41:1–23.[CrossRef][Medline]

Elowitz, M. B., and S. Leibler. 2000. A synthetic oscillatory network of transcriptional regulators. Nature. 403:335–338.[CrossRef][Medline]

Fisher, R. A. 1937. The wave of advance of advantageous genes. Ann. Eugenics. 7:355–369.

Franosch, T., and D. R. Nelson. 2000. Population dynamics near an oasis with time-dependent convection. J. Stat. Phys. 99:1021–1030.[CrossRef]

Ho, C., O. I. Kulaeva, A. S. Levine, and R. Woodgate. 1993. A rapid method for cloning mutagenic DNA repair genes: isolation of umu-complementing genes from multidrug-resistance plasmids r391, r446b, and r471a. J. Bacteriol. 175:5411–5419.[Abstract/Free Full Text]

Mann, B. 2001. Spatial phase-transitions in bacterial growth. Master's thesis. The University of Texas at Austin.

Matsushita, M., J. Wakita, H. Itoh, I. Rafols, T. Matsuyama, H. Sakaguchi, and M. Mimura. 1998. Interface growth and pattern formation in bacterial colonies. Physica A. 249:517–524.[CrossRef]

Murray, J. D. 1989. Mathematical Biology. Springer-Verlag, Berlin.

Murray, J. D. 1993. Mathematical Biology. Springer-Verlag, New York, Berlin, Heidelberg.

Neicu, T., A. Pradhan, D. A. Larochelle, and A. Kudrolli. 2000. Extinction transition in bacterial colonies under forced convection. Phys. Rev. E. 62:1059–1062.[CrossRef]

Nelson, D. R., and N. M. Shnerb. 1998. Non-hermitian localization and population biology. Phys. Rev. E. 58:1383–1403.[CrossRef]

Shapiro, J. A., and M. Dworkin. 1997. Bacteria as Multicellular Organisms. Oxford University Press, New York.

Shnerb, N. M. (2001) Extinction of a bacterial colony under forced convection in pie geometry. Phys. Rev. E 63:011906.[CrossRef]

van Saarloos, W. 1987. Dynamical velocity selection: marginal stability. Phys. Rev. Lett. 58:2571–2574.[CrossRef][Medline]

van Saarloos, W. 1988. Front propagation into unstable states: marginal stability as a dynamical mechanism for velocity selection. Phys. Rev. A. 37:211–229.[CrossRef][Medline]

Vilar, J. M. G., H. Y. Kueh, N. Barkai, and S. Leibler. 2002. Mechanism of noise-resistance in genetic oscillators. Proc. Natl. Acad. Sci. USA. 99:5988–5992.[Abstract/Free Full Text]

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via Google Scholar Google Scholar Articles by Lin, A. L. Articles by Swinney, H. L. Search for Related Content PubMed PubMed Citation Articles by Lin, A. L. Articles by Swinney, H. L.