Nicholas Blackburn, ^* Tom Fenchel, Jim Mitchell

Are nutrients available to microbial communities in micropatches long enough to influence growth and competition? And what are the sources of such patches? To answer these questions, the swimming behavior of chemotactic bacteria in seawater samples was examined. Clusters of bacteria formed in conjunction with cell lysis and excretion by protozoa. These point sources of nutrients spread into spherical patches a few millimeters in diameter and sustained swarms of bacteria for about 10 minutes. Within that time, a large proportion of the nutrients was encountered by bacteria, chemotactic and nonchemotactic alike. Chemotaxis is advantageous for bacteria using patches over a certain size.

N. Blackburn and T. Fenchel, Marine Biological Laboratory, University of Copenhagen, Strandpromenaden 5, 3000 Helsingør, Denmark. J. Mitchell, School of Biology, Flinders University, Adelaide SA 5001, Australia.
^*   To whom correspondence should be addressed. E-mail: mblnb{at}mail.centrum.dk

Observations of seawater samples (8) revealed that clusters of bacteria continuously formed and dispersed. Some sources of attractants were identified as the autolysis of a large microbe, such as an algal cell or a protozoan (Fig. 1A). Other clusters formed without any visually distinct source (Fig. 1B). Various species of ciliates were often seen at the center of these clusters, which we assumed were related to the discharge of undigested organic matter and inorganic nutrients from food vacuoles (9). Other zooplankton excrete plumes of nutrients (10). Studies have focused mainly on the importance of these nutrient plumes for phytoplankton growth (10, 11), and conclusions have been largely negative. However, bacteria have 100 times the uptake potential of phytoplankton and are therefore potential key players in rapidly consuming dissolved nutrients from patches and transferring them into the food web.

1

1

1

The current model of bacterial chemotaxis is based on swimming behavior of the enteric bacterium Escherichia coli (12). Reports of different swimming behavior displayed by strains of marine bacteria (13) pose the question of whether the model is widely applicable. To test this, we studied swimming behavior of bacteria from seawater enrichments under conditions of low oxygen saturation (14), where they were observed to form clusters around algae producing oxygen (Fig. 1C). Motility patterns could be reproduced by simulations (Figs. 1D and 2). At an intermediate distance, the mean radial component of runs toward the source increased by a factor of 2.5. This local maximum was a result of an alignment phenomenon, where runs heading up the gradient brought the bacteria quickly into a core zone. This is illustrated in Fig. 1D by the spokelike pattern surrounding the central zone of tight clustering. Reversal was important to chemotactic efficiency in this example: Decreasing the turn angle to 90° failed to reproduce visually obvious clusters. Run lengths decreased at close proximity to the attractant source where the probability of moving past it was high, resulting in the rapid initiation of reversal and run truncation. Run lengths down the gradient were quite constant in length (Fig. 2), as observed for E. coli (12).

1

The model of chemotaxis subsequently formed the basis for analysis of patches resulting from point-source releases of nutrients. Clusters of bacteria were observed to grow in size before subsiding within a few minutes (Fig. 3). The dissolved attractant for the chemotactic response was invisible, but spherical symmetry of clusters and their rate of dispersal indicated that they were the result of point-source releases of nutrients. When nutrients are released at a point, they spread rapidly into a spherical patch. The rate of spreading slows as the patch increases in size. A stationary bacterium in the proximity of such an event would find itself within a wave, which would peak sharply and disperse slowly. A chemotactic bacterium would begin to move up the gradient as soon as it could detect its presence. This was simulated with a source of 1 pmol, which corresponds approximately to the dissolved free amino-acid pool in a cell of 10 µm in diameter (15). Again, the model was able to reproduce the response of the clustering phenomenon (Fig. 3), and allows the behavior of the cluster to be described in terms of the model. Thus, the pattern of response with time can be identified as a combined effect of molecular diffusion of the attractant, congregation of bacteria through the drift component of their movement, and their subsequent dispersal by the diffusive component of their movement. The pattern matched only that of an attractant of low molecular weight and correspondingly high diffusion coefficient; this fits a wide spectrum of inorganic and organic molecules including oxygen, sugars, nitrate, phosphate, and amino acids. The expected encounter of released matter over 10 min was 0.028% for nonmotile bacteria and 0.046% for chemotactic bacteria (Fig. 4), calculated by averaging the probability distributions. In this example, chemotactic bacteria would encounter, on average, 46% of their biomass within 10 min, if the patch contained 1000 bacteria equivalents of dissolved matter, and 10% of the population would encounter more than 100% (Fig. 4). Thus, patches offer a mechanism for supporting high bacterial growth rates independently of background concentrations. The phenomenon offers an explanation for why tracers originating from organism biomass can be assimilated by bacteria much more rapidly than when they are homogeneously distributed (16). The gain in nutrient exposure achieved by chemotaxis is dependent on the size of the patch. The gain was a factor 2 for a patch size of 1 pmol (Fig. 4), but it drops to 1 for 0.1 pmol and increases to 3.5 for 10 pmol (17). Bacterial chemotaxis is thus beneficial only under conditions where patches are of a certain size, unlike their importance for bacterial growth in general.

The patches illustrated by clusters in Fig. 1, A and B, were more than 1 mm in diameter, and the clusters contained over 1000 motile bacteria at concentrations up to 10⁷ ml¹. Even background concentrations of bacteria (~10⁶ ml¹) would significantly influence patch dynamics, and would ultimately result in their erosion by consumption. The estimates above indicate that dispersion and potential erosion time scales are similar (on the order of 10 min).

On this spatial and temporal scale, the effects of shear influence patch dynamics (18). At low oceanic shear rates of 0.005 s¹, a patch would remain virtually undistorted, but it would be stretched into streamers at high shear rates. The importance of shear in the context of this study is difficult to evaluate. Firstly, shear is strongly intermittent (19), and secondly, by decreasing the concentration gradient, it decreases dispersion by molecular diffusion. Of particular importance to bacterial chemotaxis is the fact that shelter from shear exists during periods of low turbulence, during periods between bursts of turbulence, in habitats close to surfaces (for example, biofilms and large aggregates), and at boundaries such as thermoclines and pycnoclines.

Chemotactic bacteria have played the role of live probes in this study, showing the dynamics of dissolved nutrient patches. We have yet to identify the frequency of occurrence and magnitude spectra of such patches in specific microbial food webs. They undoubtedly represent interesting ecological niches for bacteria, and they will also contribute much to our understanding of the flow of nutrients and energy in aquatic ecosystems if they prove to be major pathways.

REFERENCES AND NOTES

1. J. C. Goldman, Bull. Mar. Sci. 35, 462 (1984) .
2. E. M. Purcell, Am. J. Phys. 45, 3 (1977) .
3. G. A. Jackson, Limnol. Oceanogr. 32, 1253 (1987) .
4. W. Bell and R. Mitchell, Biol. Bull. 143, 265 (1972) [Abstract/Free Full Text] .
5. J. G. Mitchell, A. Okubo, J. A. Fuhrman, Nature 316, 58 (1985) [CrossRef] ; J. D. Bowen, K. D. Stolzenbach, S. W. Chisholm, Limnol. Oceanogr. 38, 36 (1993) .
6. N. Blackburn, unpublished observations.
7. ___, F. Azam, Å. Hagström, Limnol. Oceanogr. 42, 613 (1997) .
8. Samples were taken, immediately before observation, from the vicinity of algal mats in a large seawater aquarium with high through-flow. The microbial community could be characterized as being rich and diverse. Identical communities have been observed in completely natural habitats. A standard microscope fitted with a dark-field condenser was used for observation with a 10× objective giving a depth of field of ~20 µm. The magnification was increased for some recordings with a magnifying lens. Samples were observed in a chamber made by a 1.5-mm-thick rubber O-ring placed on top of a microscope slide and covered with a cover slip. A fiber optic light source was used for illumination to minimize heat transfer. A standard video camera and VCR were used for recordings.
9. A. Andersson, C. Lee, F. Azam, Å. Hagström, Mar. Ecol. Prog. Ser. 23, 99 (1985) [Web of Science]; P. A. Jumars, D. L. Penry, J. A. Baross, M. J. Perry, B. W. Frost, Deep-Sea Res. 36, 483 (1989) [Web of Science].
10. J. T. Lehman and D. Scavia, Proc. Natl. Acad. Sci. U.S.A. 79, 5001 (1982) [Abstract/Free Full Text] .
11. P. J. L. Williams and L. R. Muir, in Ecohydrodynamics, J. C. J. Nihoul, Ed. (Elsevier, New York, 1981), pp. 209-218; D. J. Currie, J. Plankton Res. 6, 591 (1984) [Abstract/Free Full Text]; G. A. Jackson, Nature 284, 439 (1980) [CrossRef] .
12. D. A. Brown and H. C. Berg, Proc. Natl. Acad. Sci. U.S.A. 71, 1388 (1974) [Abstract/Free Full Text] .
13. T. Fenchel, Microbiology 140, 3109 (1994) [Web of Science] ; J. G. Mitchell et al., Appl. Environ. Microbiol. 61, 877 (1995); J. G. Mitchell, L. Pearson, S. Dillon, K. Kantalis, ibid., p. 4436; J. G. Mitchell, L. Pearson, S. Dillon, ibid. 62, 3716 (1996); G. M. Barbara and J. G. Mitchell, ibid., p. 3985.
14. Samples of seawater were enriched with 0.02% tryptic soy broth and left overnight, after which the culture contained strains of highly motile bacteria. Samples were sandwiched between a slide and cover slip after the addition of cells from a pure culture of Pavlova lutheri. A ring of bacteria around the air-water interface of the chamber indicated near anoxia inside the chamber.
15. P. A. Wheeler, in Nitrogen in the Marine Environment, E. J. Carpenter and D. G. Capone, Eds. (Academic Press, New York, 1983), p. 309.
16. F. Azam and J. W. Ammerman, in Flows of Energy and Materials in Marine Ecosystems, M. J. R. Fasham, Ed. (Plenum, New York, 1984), p. 345.
17. N. Blackburn, T. Fenchel, J. Mitchell, data not shown.
18. J. D. Bowen and K. D. Stolzenbach, J. Fluid Mech. 236, 95 (1992) [CrossRef].
19. L. Karp-Boss, E. Boss, P. A. Jumars, Oceanogr. Mar. Biol. Annu. Rev. 34, 71 (1996) .
20. H. C. Berg and D. A. Brown, Nature 239, 500 (1972) [CrossRef] [Medline] .
21. Video sequences were digitized to computer memory at 25 frames s¹. The resulting digital film-strips were analyzed frame by frame for trajectories of movement by LabTrack (DiMedia, Kvistgaard, Denmark). Objects moving out of the plane of focus resulted in short tracks and were filtered out of the set of track vectors. Tumbles were detected at points where changes in trajectory angle between two video frames exceeded 1 rad. Runs were defined as periods between tumbles.
22. The enteric bacterium E. coli swims in straight runs interspersed by tumbles (20). Positive taxis is achieved by lengthening runs when positive changes in concentration are detected over a period of time (12). The model refined by Brown and Berg states that the mean increase in run duration is proportional to the positive rate of change of attractant concentration where C is concentration, K_D is a dissociation constant (100 µM), and  is a sensitivity constant (1000 s). Simulations were performed by allowing cells to move against a concentration field (3). Cells were moved a distance determined by their swimming velocity and heading angle from a physical location whose attractant concentration was C₁ to another location of concentration C₂, within each simulated time step dt. The change in mean run duration could thus be calculated as Negative changes were ignored. Run durations are Poisson distributed (12). The probability of tumbling after each time step is dt/( + ). The desired Poisson process was implemented with a random generator to decide whether or not to tumble after each time step. Tumbles were simulated as reversals, and a Brownian rotation of 1 rad s¹ was introduced. Swimming velocities and mean run durations were acquired from tracks of live cells.
23. The steady-state concentration field C(r) was calculated as where oxygen exudation rate E = 0.25 fmol s¹ was estimated from light-saturated photosynthesis of a cell 4 µm in diameter. The diffusion coefficient D = 10⁵ cm² s¹. The inverse square-root dependency on the distance r from the source was introduced instead of inverse proportionality (3) to more closely approximate its shape in the flat chamber.
24. The concentration field C(r,t) of a spreading patch was calculated as where M is the amount of matter released (1 pmol), and D = 10⁵ cm² s¹. Simulations were based on 100 individuals initially distributed randomly within a radius of 1 mm from the source. Velocity v = 50 µm s¹,  = 0.4 s.
25. The mass flow of a solute at concentration C with diffusion coefficient D toward a sphere of radius a is 4aDC.
26. We thank the Swedish Foundation for International Cooperation in Research and Higher Education (STINT), the Danish National Research Council (SNF), the Australian Research Council, and Flinders University for support of this study.