Bacteria Harnessing Complexity

By

Eshel Ben Jacob, Yakir Aharonov and Yoash Shapira

School of Physics and Astronomy Raymond & Beverly Sackler Faculty of Exact Sciences The Maguy-Glass Chair in Physics of Complex Systems Tel Aviv University, 69978 Tel Aviv Israel

Abstract

The study of bacterial colonies - groups of bacteria grown from a single or few similar cells - is a crucial step towards understanding biofilms, which are composed of many different bacterial colonies. In this article, we review some of the exciting discoveries about the cooperative behavior of bacteria in colonies, which might shed new light on biocomplexity in general and biofilms in particular. The review is aimed at researchers from different disciplines - microbiology, biology, chemistry, physics, mathematics, and computer science. To make the presentation comprehensible to such a wide audience, we avoided the use of specialized terminology of the different disciplines and limited the experimental and computational details.

We start with the realization that, under natural growth conditions, bacteria can self- 9 12 organize into hierarchically structured colonies, 10 -10 bacteria each. To that end, they developed and utilize a great variety of biochemical communication agents, such as simple molecules, polymers, peptides, complex proteins, genetic material, and also "cassettes of genetic information" like plasmids and viruses.

Complex colonial forms (patterns) emerge through the communication-based singular interplay between individual bacteria (the micro-level) and the colony (the macro-level). Each bacterium is, by itself, a biotic autonomous system with its own internal cellular gel that possesses informatics capabilities (storage, processing and interpretation of information). These afford the cell certain freedom to select its response to biochemical messages it receives, including self-alteration and broadcasting messages to initiate alterations in other bacteria. Such self-plasticity and decision-making capabilities elevate the level of bacterial cooperation during colonial self-organization.

As the individuals increase their adaptability to the group, the colony elevates its durability and adaptability to the environment by increasing its complexity. The essential new lesson learned from bacteria is that colonial higher complexity provides the degree of plasticity and flexibility required for better durability and adaptability of the whole colony to a dynamic environment. According to this picture, new features collectively emerge during biotic self-organization on every level, from the internal cellular gel to the whole colony. The

1 cells thus assume newly co-generated traits and abilities that are not explicitly stored in the genetic information of the individuals. For example, bacteria cannot genetically store all the information required for creating the colonial patterns. In the new picture, they don't need to, since the required information is cooperatively generated as self-organization proceeds by bacterial communication, informatics and self-plasticity capabilities. Thus, the bacteria need only have genetically stored the guidelines for producing these capabilities and using them to generate new information as required. We illustrate these special capabilities by exposing the bacteria to non-lethal levels of antibiotics. We also demonstrate bacterial ability to sense the effect of weak electromagnetic (EM) signals on aqueous solutions. The observations illustrate that bacteria use their intracellular flexibility, involving signal transduction networks and genomic plasticity, to collectively maintain self and shared interpretations of chemical cues, exchange of meaning-bearing chemical messages, and dialogues. The meaning-based communication permits the formation of colonial intentional behavior, purposeful alteration of colony structure and decision-making - features we might begin to associate with bacterial social intelligence. Such social intelligence, should it exist, would require going beyond communication to encompass additional intracellular processes, yet unknown, for generating inheritable colonial memory and commonly shared genomic context.

Introduction

Bacteria, being the first form of life on earth, had to devise ways to synthesize the complex organic molecules required for life. They are able to reverse the spontaneous course of entropy increase and convert high-entropy inorganic substances into low- entropy life-sustaining molecules. Three and a half billion years have passed, and the existence of higher organisms depends on this unique bacterial know-how. Even for us, with all our scientific knowledge and technological advances, the ways bacteria solve this fundamental requirement for life is still a mystery [1-7]

We do know that this is not a solitary endeavor for the bacteria, and under natural conditions they employ chemical communication to form hierarchically structured colonies, 109-1012 bacteria each [4-33]. By acting jointly, they can make use of any available source of energy and imbalances in any environments, from deep inside the earth crust to nuclear reactors and from freezing icebergs to sulfuric hot springs; and they can convert any available substances, from tar to metals.

2 Under unpredictable hostile environmental conditions, when the odds are against survival, the bacteria turn to a wide range of strategies for adaptable collective responses. These cooperative modes of behavior are manifested through remarkable different patterns formed during colonial self-organization. The aesthetic beauty of these geometrical patterns [34,35] is striking evidence of an ongoing cooperation that enables the bacteria to achieve a proper balance of individuality and sociality as they battle for survival, while utilizing pattern-formation mechanisms that we have only recently begun to understand. The geometrical patterns can also be associated with quantified measures, as explained in Appendix A. Yet, the naked eye can distinguish between patterns better than the most advanced image processing methods. Therefore, in this article we will rely on the reader’s geometrical perception in comparing between the geometrical patterns generated by the bacterial colonies.

In bacterial self-organization, unlike inanimate pattern formation [36-39], there is an additional inherent degree of plasticity: the building blocks of the colony are themselves living organisms, each with internal degrees of freedom, internally stored information and internal assessment of external chemical messages. These afford each bacterium freedom to respond flexibly and even alter itself, by means of modifying its genetic expression patterns. At the same time, efficient adaptation of the colony to adverse growth conditions requires self-organization on all levels – which can only be achieved via cooperative behavior of the individual cells. For that purpose, bacteria communicate by a broad repertoire of biochemical agents. At the same time, each bacterium has equally intricate intracellular communication mechanisms involving, for example, signal transduction networks [40,41]. These are used to generate intrinsic meaning for contextual interpretations of the chemical messages and for formulating appropriate responses [13]. Biochemical messages are also used in bacterial linguistic communication for exchange of meaningful information across colonies of different species, and even with other organisms [42].

As we are discovering, bacterial communication-based cooperation encompasses colony morphogenesis, which includes coordinated gene expression, regulated cell

3 differentiation and division of tasks. Collectively, bacteria can glean information from the environment and from other organisms, interpret the information (assign meaning), develop common knowledge, and learn from past experience. The colony behaves much like a multicellular organism, or even a social community with elevated complexity and plasticity that afford better adaptability to whatever growth conditions might be encountered [5].

Branching Patterns of Lubricating Bacteria To illustrate the ability of bacteria to cope with conflicting environmental constraints, we begin with the branching patterns exhibited by the Paenibacillus dendritiformis lubricating bacteria [11]. This class of bacteria developed a special strategy to move on hard surfaces – they collectively excrete special chemicals to form a layer of lubricant in which they can swim. As they swim, they push the layer forward, paving their own way. A dilemma arises when, in addition, the available food is not sufficient to sustain a dense population [43-48].

When such conditions are mimicked in Petri-dishes with hard substrates of low nutrient concentration, new and interesting phenomena are observed, as shown in Fig 1. Collectively, the bacteria turn on genes to emit chemicals (surfactants) used to extract fluid from the substrate, thus creating on top of the crust a lubricating layer with a well defined envelope within which they can swim. The task is not simple, as the production of the lubricant requires collective action of dense bacterial population which the food- depleted substrate can not sustain. The solution comes in the form of a branching structure of the colony – within each branch the bacterial density is sufficiently high, yet the average population density of the colony is sufficiently low to match the availability of food. The real trick is the careful adjustment of the lubricant viscosity and its production rate to generate specific branch structures with specific widths according to the substrate hardness and food level [43-46]. This adaptable self-engineering can be viewed as the solution to a challenging self-consistency mathematical problem between

4 two contradictory constraints – the need for high bacterial density for movement and in sufficient level of food to support high bacterial densities.

-a- -b-

-c- -d- Fig 1. Example of branching colonial patterning: (a) Typical branched colonial pattern formed by the Paenibacillus dendritiformis bacteria when grown on hard and food depleted substrate. The substrate is prepared from an LB solution with about 2% agar and less than 2 g/L peptone. 22cc of the substrate solution (in a liquid state) is poured into a standard Petri-dish 8.8cm in diameter. The solution is let cool and dry at 25oC and 50% humidity for about 60 hours until its weight is reduced by 1 gram. The colony starts from a droplet (5
l) inoculation at the center of the Petri-dish. The droplet is taken after 24 hour bacterial growth in a shaker to guarantee that the stationary phase has been reached. The dark dot at the center of the colony is the area of the inoculated droplet that usually contains about 104

5 bacteria. After inoculation, the bacteria go through an ‘embryonic’ colonial stage for several hours and only then does the colony start to expand outward on the surface [11-13]. It takes the colony about two days to reach the observed size of about 5cm in diameter shown here. The observed pattern can be associated with quantified measures, as explained in Appendix A. (b) A closer look at the branches through a polarized optical microscope to show the layer of lubricant collectively produced by the bacteria. (c) A snapshot from a video clip taken through an optical microscope with x500 magnification [*]. The bars are the individual bacteria. The video clip observations reveal bacteria swimming – segments of straight motion for about 1-3 seconds at a speed of about 1 micron per second interrupted by bacterial tumbling terminating in a random new direction. That’s why the bacterial movement can be modeled as random walk. (d) A scanning electron microscope picture. Note the versatility in the individual bacteria. It is now understood that it is not arbitrary but collectively regulated to afford the colony elevated group flexibility.

Utilizing Chemotaxis for Natural Self-Engineering Once the notion of self-engineering is appreciated, one can study other possible mechanisms whereby the bacteria control the overall pattern. Over the years, it has become clear that, in fact, there are many such possibilities, often relying on some sort of chemical signal that the cells can use to influence each other. At relatively higher nutrient levels, additional bacterial density variations are visible within the branches, as seen in Fig 2. From the mathematics of instability of dynamical systems one can deduce that an additional attractive mechanism, activated by the bacteria themselves over distances shorter than the branch width, is needed to form the observed patterns within the branches. Model simulations can be used to test the idea that the patterns result from bacterial self-engineered organization [49], utilizing chemotactic signaling.

Chemotaxis is cell movement in response to gradients in the concentration of a chemical agent [50-55]. The movement can be biased either towards higher concentrations (attractive) or away from high concentrations (repulsive). Bacteria are too short to detect chemical gradients, yet swimming bacteria found a smart solution to detect gradients and bias their movement accordingly. For attractive (repulsive) bias, they simply detect the concentration as they swim, and if the concentration increases (decreases) they delay their tumbling. The net result is biased random walk towards the higher (lower) concentration, which can be directly incorporated into the model as a drift term.

6 The most familiar case of chemotaxis is attraction to an external chemical such as a nutrient. There is evidence that such chemotaxis occurs in these colonies and is responsible for an increased expansion rate and colony ”bushiness” at intermediate values of the nutrient concentration.

Very different patterns form at low nutrient levels (Fig 2). To explain the mechanism, we recall that part of the branch-making dynamics relies on the cells going into a pre-spore non-motile state further back from the colony front, where nutrien levels are extremely low. It was found that bacteria in the pre-spore state can still participate in chemical communication. For example it was proposed that they can emit a repellent chemical as they are entering this state. For the repellent to accomplish its task, namely to persuade other cells to move away from such locations, the range of this signal must be fairly large. This can be tested by generic modeling, as illustrated in Fig 2.

So the picture that emerges is that the basic branching pattern is sculpted into a variety of forms by the combined action of a variety of chemotactic strategies. These strategies serve to coordinate the actions of otherwise independent cells so as to make maximal use of the resources at their disposal. As these different influences sort themselves out, changing conditions and changing bacterial strains always lead to new structures. Exactly how information from the outside is utilized to help decide which, if any, of these processes need to be turned on is still an open question. Yet there are increasing hints that these decisions are made cooperatively, much like what is known to be true with regard to the collective decision-making of bacteria to sporulate or the decision to share genetic information.

7

-a- -b-

-c- -d-

-e- -f- Fig 2. Examples of different branching patterns formed during colonial development of the P. dedritiformis bacteria. To self-engineer their colonial structure, these bacteria regulate the balance between attractive and repulsive chemotactic signaling as well as their food chemotaxis. Generic modeling (Appendix B) of the growth can be used to test this idea by comparing the observed patterns shown on the left ((a), (c) and (e)) with the results of model

8 simulations shown on the right ((b), (d) and (f)). The colors in the observed colonies represent the bacterial density. In the simulations, the color code shows the time evolution of the growth. The top picture is the typical pattern when food chemotaxis dominates the growth at intermediate levels of food depletion. The middle is the typical pattern at higher food levels when attractive chemotactic signaling is activated, and the bottom is at a very low level, when repulsive chemotactic signaling is intensified. Comparing the patterns, one should keep in mind that the real colonies have almost a million times more bacteria as well as additional mechanisms not included in the model.

9 Breaking the reflection symmetry

Bacteria can cooperatively make drastic alterations of their internal genomic state and transform themselves into different cells. For example, the Paenibacillus dendritiformis lubricating bacteria, when grown on poor substrates, can select their identity from the two available distinct cell types – the branching (B) and the chiral (C) morphotypes shown in Fig3 [11-13]. On harder substrates, when higher bacteria densities are required to produce sufficient amounts of lubricating fluid, the B morphotype is selected, leading to the formation of colonies with branching, bush-like morphologies reminiscent of the patterns generated by starved B. subtilis bacteria [47,48]. The engineering skill of the P. dendritiformis is manifested during growth on softer substrates in the formation of curly branches. This special geometrical organization allows faster expansion while also using patches of food left behind as the branches twist inward. To bring this about, the bacteria suppress cell division and elongate. Optical microscope observations during colony development reveal the following: upon elongation, the cells alter their collective movement from the typical run-and-tumble of the short B cells to a coordinated forward- backward movement, which yields an organized twist of the branches with a specified handedness. It is now understood how the preferred handedness of the twist results from the cell-cell interaction together with the inherent flagella handedness. Other bacterial strains, such as Bacillus mycoides, can exhibit similar chiral patterning [56].

-a- -b-

10

-c- -d- Fig 3. The chiral morphotype. The engineering skills of the P. dendritiformis bacteria are manifested during growth on softer substrates, in the formation of curly branches like in the colony shown in (a). This special geometrical organization allows faster expansion while also using patches of food left behind as the branches twist inward. To bring this about, the bacteria suppress cell division and become elongated, as observed via scanning electron microscope (b). Optical microscope observations during colony development (c) reveal the following: upon elongation, the cells alter their collective movement from the typical run- and-tumble of the short B cells to a coordinated forward-backward movement, which yields an organized twist of the branches with a specified handedness. Utilizing generic modeling [13], it is now understood how the preferred handedness of the twist results from the cell- cell interaction together with the inherent flagella handedness. (d) An example of model simulations using the walking spinors model.

The two possible morphotypes are inheritable and can coexist for some range of growth conditions. Yet, when colonies of the B morphotypes are grown on soft substrate, an intriguing phenomenon of spontaneous transitions is observed: a majority of the grown colonies exhibit B ‰ C transitions. The reverse C ‰ B morphotype transitions are observed during growth on harder and richer substrate. In both cases, the newly selected pattern is the one that maximizes the rate of colony expansion, hinting that the colonial morphotype manipulation is for better adaptability [11].

For the morphotype transition to happen, a sufficiently large group of C bacteria must be formed, by some synchronized and/or autocatalytic gene-expression transition of many B bacteria into the C morphotype. In addition, the newly formed C bacteria have to find their fellows within the large crowd of B cells and burst out as a group with a new identity. This process involves a special dialogue between C’s within the majority of B’s,

11 as can be deduced from the picture on the right. This growth starts from a culture of a majority of C’s, namely, an artificial situation contrary to what bacteria encounter during natural, spontaneous morphotype transitions. The outcome is an initially chiral pattern with a different geometry than that of a pure C colony, due to the presence of the additional B cells that maintain their identity. Later, in a synchronized manner, the pattern switches to a mainly branching one with some handedness. Next, the C pattern bursts out in a manner similar to that observed when the growth starts from a pure culture of B’s.

There seems to be a semantic-message-based dialogue that helps the cells collectively decide between the C and B patterns. For instance, we show in Fig 4 a morphotype transition initiated in response to stress encountered due to the presence of fungi. In this case, the colony resorts to the chiral morphotype to acquire better structural flexibility to expand around the fungi.

-a- -b- Fig 4: Morphotype transitions. (a) A spontaneous B -> C morphotype transition. (b) An example of a stress induced morphotype transition when a colony of branching morphotype bacteria encounters a fungi (the bright spot).

12

Cooperative hierarchical organization Some bacterial strains organize their colonies by generating modules, each consisting of many bacteria, which are used as building blocks for the colony as a whole. This behavior is observed, for example, in the lubricating bacteria Paenibacillus vortex [11- 13] that produce the bacterial vortices shown in Fig. 5, and in other strains like Bacillus circulans [57] and Paenibacillus alvi [58]. Model simulations suggest [59,60] that a combination of short-range attractive and long-rage repulsive chemotactic signaling mechanisms can lead to the formation of the observed patterns, as illustrated in Fig 6.

-a- -b-

13 -c- -d-

-e- -f-

Fig 5: Hierarchical colonial organization. Colonial patterns generated by the Paenibacillus vortex bacteria when exposed to different growth conditions. Pictures (a) – (d) show a whole colony view. The petri-dishes are 8.8cm in diameter. The bacterial population of these colonies is larger than that of people on earth, yet they are all coordinated. The peptone (food) level is 7.5g/l (a), 10g/l (b), 15g/l (c) and 20g/l (d). Each vortex (the condensed group of bacteria) is composed of many cells that swarm collectively around their common center at about 10 micron/sec. The number of bacteria in the vortices varies from tens to millions, according to their location in the colony. Both clockwise and anticlockwise rotating vortices are observed, although the majority has the same handedness. The cells in the vortex replicate, and the vortex expands in size and moves outward as a unit, leaving behind a trail of motile but usually non-replicating cells – the vortex branch. The twist of the vortex branch is determined by the handedness of the vortex rotation. The dynamics of the vortices is quite complicated and includes attraction, repulsion, merging and splitting. Yet, from this complex, seemingly chaotic movement, a colony with complex but non-arbitrary organization develops, as seen in pictures (a) – (d). Pictures (e) and (f) are snapshots from a video recording [*] taken during formation of new vortices (magnification x500, so the pictures are about 100 microns wide; the bars are the individual bacteria). Maintaining the integrity of the vortex while it serves as a higher-order building block of the colony requires advanced communication: each cell in the vortex needs to be informed that its role is now more complex, being a member of both the specific vortex and the whole colony, so it can adjust its activities accordingly. This ongoing communication is particularly apparent when it comes to the birth of new vortices. These emerge in the trail behind a vortex following initiation signals that cause the bacteria there to produce more lubricating fluid and to move quite rapidly as a turbulent "biofluid", until an eddy forms and turns into a new vortex. The entire process appears to proceed as a continuous dialogue: a vortex grows and moves, producing a trail of bacteria and being pushed forward by the very same bacteria left behind. At some point the process stalls, and this is the signal for the generation of a

14 new vortex behind the original one, that leaves home (the trail) as a new entity toward the colonization of new territory.

-a- -b-

-c- -d-

15

-e- -f-

Fig6. Generic modeling of the generation and organization of vortices. Observed patterns (left, a, c, e) are compared with numerical simulations [13, 59, 60] on the (right, b, d, f) to test the idea that vortices can be generated by gliding and swarming bacteria when attractive forces (e.g., by attractive chemotaxis) are in action. Such forces keep the bacteria together (like ice-skating people who hold hands together), thus forcing them to perform a ‘merry-go-round’ motion. To be pushed out, the bacteria inside the colony can emit repulsive chemotaxis. Results of numerical simulations of the self-generation of vortices via attractive chemotactic signaling are shown on top, and the effect of repulsive chemotactic signaling in which the vortices are `pushed out’ by repulsive chemotactic signaling is shown in (f). (c) and (d) a ‘bagel’ dynamics in which the bacteria form a closed path with an open center. The arrows in (d) indicate the velocity of the bacterial moments in the model simulations.

Reproducible Complexity Since we are looking to adopt a new strategy in which conclusions are deduced from comparison of the observed patterns (morphologies), a high level of reproducibility is required; that is, two colonies taken from the same source and grown under the same conditions should show similar patterns. An example of the level of reproducibility that can be obtained during colonial development under a very strict protocol is shown in Fig 7. For non-living systems, the shown level of similarity might not come as a surprise [34- 39,61]; but if we keep in mind that each bacterium has an internal structure with high levels of flexibility and inherent variations, these observations suggest that the patterns are not arbitrary but involve some internal sophisticated means of regulation and control.

16 So, for example, upon replication, the new cells will immediately have the proper internal gene expression states to fit the colonial behavior. The colonies’ reproducible complexity goes hand in hand with high flexibility for elevated adaptability leading to the great variety of patterns generated during growth under different conditions as illustrated in Fig 5.

Fig 7. Reproducible complexity. The two pictures show two colonies started from bacteria taken from the same mother colony. It is an illustration of the level of reproducibility that can be achieved when a strict growth protocol is followed.

Emergence on all levels For example, genetic degrees of freedom (i.e. the dynamics which determine which genes are expressed) are irrelevant if one merely wishes to explain branching but appear to be crucial if we want to understand observed transitions between different colony patterns. The vortex itself can be studied using simple notions, but the way in which the complex exchange of information takes place between the colony as a whole and the vortex substructures requires much more. If we take into account the incredible complexity present at all scales (from molecular up to ecological community) in the bacterial system, we must admit that there will always be phenomena at the macroscopic scale that are not

17 captured by any sort of tractable model. An example of such a phenomenon is presented in the following section.

Of course, it is also true in pure physics that there are always additional levels of substructure; there, however, the macroscopic scale phenomena are shielded from these details and there is nothing that we can learn about, say, superstrings, which could conceivably affect our view of, say, diffusively-driven crystal growth. However, as we learned from the bacteria, this simple separation of scales does not hold for living systems.

Generation of vortex self-identity Maintaining the integrity of the vortex while it serves as a higher-order building block of the colony requires advanced communication: each cell in the vortex needs to be informed that its role is now more complex, being a member of both the specific vortex and the whole colony, so it can adjust its activities accordingly. This ongoing communication is particularly apparent when it comes to the birth of new vortices. New vortices emerge in the trail behind a vortex following initiation signals that cause the bacteria there to produce more lubricating fluid and to move quite rapidly as a turbulent "biofluid", until an eddy forms and becomes a new vortex. The entire process appears to proceed as a continuous dialogue: a vortex grows and moves, producing a trail of bacteria and being pushed forward by the very same bacteria left behind. At some point the process stalls, and this is the signal for the generation of a new vortex behind the original one, that leaves home (the trail) as a new entity toward the colonization of new territory. To illustrate the above notions, we show in Fig. 8 the patterns exhibited by bacteria inoculated from the center of the colony vs. those created by bacteria taken from the leading vortices. These and some additional preliminary observations suggest processes connected with the formation of bacterial self-identities, which are beyond current modeling and might even be in principle non-computable [7].

18

-a- -b-

-c- -d-

Fig 8. Vortex inheritable self-identity. (a) A picture of the ‘mother’ colony used as a source of bacteria for the consequent growth experiments. Pictures (b) and (d) show two plates with 4 colonies each that started from bacteria taken (by picking) from the center of the mother colony (a). These colonies have distinctively different patterns in comparison with those shown in (c), that started from bacteria taken (by picking) from large vortices at the front of the mother colony.

Antibiotic Driven Self-Engineered Organization

19 In medicine, resistance to antibiotics is usually considered a binary (Yes/No) property: either a specific strain of bacteria is resistant to a specific antibiotic or it is not. The bacteria are generally regarded as being `resistant' if they can tolerate the maximal concentration of the antibiotic which is non-toxic to the treated humans or animals. However, it is known that bacterial colonies are often more resistant than the individual cells, thereby blurring the borderline between resistance and sensitivity. Moreover, antibiotics affect cells in different manners depending on the growth and the physiological state of the cells. New results show that susceptibility to antibiotic should be regarded as a quantitative property: the bacteria react to non-lethal levels of antibiotic even in concentrations well below the critical (lethal) level [62,63]. And in some cases, low levels can even enhance the colony development.

Although the available arsenal of antibiotics contains a variety of functionally and structurally different compounds, resistance mechanisms have been developed for almost all of them. Resistance may involve decreased entry of the drug, changes in the receptor (target), metabolic inactivation or modification of the drug, or production of “dummy" targets to adhere to the compound and lower its bioavailability. Resistance can be a cooperative process. The most compelling example is metabolic inactivation of an antibiotic occurring outside the cells: a significant reduction in the drug concentration can occur only with high concentration of the inactivating agent, which requires a dense population of bacteria.

In Fig 9, the idea of ‘higher complexity for better adaptability’ is reflected by the bacteria’s ability to engineer the colony shape to better adapt to the antibiotic stress by adjusting a specific pattern to each kind of antibiotic. By comparing simulations of the models and experimental observations, we can gain insights about the response of the bacteria to the antibiotic stress. For example, we conjecture that the changes in the colonial patterns caused by the antibiotic co-trimoxazole (Septrin) are due to elongation of the bacterial cells. This is consistent with the method of action of Septrin, which interferes with bacterial metabolism and might mimic starvation conditions (conditions under which these bacteria are usually longer).

20

-a- -b-

21 -e- -f- Fig 9: Higher complexity for better adaptability I: When facing non-lethal antibiotic stress, the bacteria are able to modify their colonial patterning for better survival. Pictures (a) to (d) show growth of the C morphotype of the P. dendritiformis bacteria when grown on surfaces in the presence of different kinds of antibiotics. Growth conditions are 5 g/l peptone and 0.6% agar concentration. (a) Septrin (30 mG/ml) after 3 days. (b) Abiplatin, a chemo-therapeutic material (10 mG/ml), after 3 days. (d) Chloramphenicol (10 mG/ml), after 6 days. (e) Streptomycin (10 mG/ml), after 6 days. The results are taken from Ref [62]. The bottom plates show the effect of Chloramphenicol (e) and Streptomycin (b) on growth of the Paenibacillus vortex bacteria. Growth conditions are 15g/l peptone and 2.0% agar. The results are taken from Ref [63]

Another example of bacterial self-engineered organization under antibiotic stress is shown in Fig 10, where we demonstrate how the P. vortex bacteria utilize their cooperative, complexity-based plasticity to alter the colony structure in order to cope with the stress.

We study colonial patterning in responses to non-lethal stress of two different kinds of antibiotics: Septrin, a suppressor of cell reproduction that might enhance communication, and Ampicillin, a distorter of cell wall structure that might impair normal cell communication. When exposed to Septrin, the bacteria seem to increase cooperation, leading to larger vortices (due to stronger attraction) that move faster away from the antibiotic stress (due to stronger repulsion by those left behind). The response to Ampicillin is puzzling: the patterns look as if the short-range communication required to form vortices is not affected, but the long-range communication required for coordinated movement of the vortices as units is tampered; yet, for some low levels of Ampicillin, the colonies can expand even faster than without the antibiotic.

22

-a- -b-

-c- -d-

Fig 10: Higher complexity for better adaptability II. To further test this principle, we grow colonies of the Paenibacillus vortex bacteria in the presence of additional kinds of antibiotics. Growth conditions are 15g/l peptone and 2.25% agar. (a) A typical colonial development for this specific strain for the specific growth conditions. The growth started at the same day from bacteria taken from the same test tube at the end of 24 hours growth in LB media. (b) The response to Septrin (co-trimoxazole), which inhibits synthesis of folic acid and thus suppresses cell reproduction. Namely, the metabolic stress caused by Septrin is somewhat similar to that of food depletion. And indeed, there is clear similarity between this pattern and the one generated for growth in the absence of antibiotic stress but half the level of food (Fig 5-a). Yet, based on model simulations and microscope observations, it seems that in response to Septrin the bacteria enhance their cooperation by intensifying chemotactic attraction to form larger vortices. This clever strategy protects the bacteria, since the antibiotic is diluted in larger vortices by the lubricating fluid extracted by the bacteria. This is true provided that the larger vortices also move faster away from the antibiotic stress; and indeed, the bacteria also elevate repulsive chemotactic response to signals emitted by the bacteria behind the vortices, which helps drive the large vortices

23 away faster. The pattern in the presence of Septrin (middle left) evolved in half the time of the pattern under food depletion (top right). This difference further supports the idea of enhanced cooperation in the presence of Septrin. (c) A disorganized colonial development in response to Ampicillin, which distorts cell wall structure and thus probably impairs the communication-based coordination. Surprisingly, when the bacteria are exposed to multiple stresses - Septrin and Ampicillin (d) - they can still survive in a way which is not understood. However, we found that we can outsmart them by first exposing them to one kind of antibiotic for the duration it takes them to adapt, and only then expose them to the other kind of antibiotic.

Bacterial Collective Memory and Learning from Experience It has been shown [13,49,62,63], that bacteria taken from LB with a non-lethal level of antibiotic develop different colonial patterns when grown on surfaces in the absence of antibiotic. That is, bacteria seem to store information about ‘past experience’. In Fig 11 we show that bacteria exposed to the same antibiotic stress a second time can cope better with the stress. This effect can be erased (according to the growth conditions and duration of exposure) by exposure to neutral conditions (i.e. growth on plates in the absence of antibiotic or in LB media) between the two encounters. Therefore, it seems that the bacteria can generate erasable, collective memory, as if to learn from their past experience. At this point, it is not clear whether these effects represent a genetic shift of the population or a collective, inheritable albeit erasable memory based on some inheritable gene-network state. Preliminary observations indicate that both scenarios are possible. Irrespective of the exact mechanism, it is clear that the bacteria can retain memory of the conditions and that this acquired memory is expressed in the colonial pattern in consequent colonial developments.

24

-a- -b- -c-

Fig 11: Collective memory and learning from experience. The top two pictures show colonial development in the presence of antibiotic started from bacteria taken (by picking), from colonies that were grown in the presence of the same antibiotic - Septrin (a) and Ampicillin (b). Note that the colonial development seems to be more enhanced. A very different pattern develops when bacteria grown in the presence of Ampicillin are exposed to Septrin, as shown in (c).

A new picture is thus emerging, one in which adaptable self-engineering can be viewed as bacteria’s solution to a challenging self-consistency mathematical problem at the forefront of optimization and control in non-linear dynamics. So it is reasonable to conclude that collectively, bacteria can glean information from the environment and from other organisms and interpret the information in an existential “meaningful” way, i.e. by building an appropriate colony structure. It is perhaps even not so far-fetched to imagine that the bacteria can develop common knowledge and learn from past experience. The findings presented here about bacteria “shaped to survive” in response to antibiotic stress and the cooperation and clashes of “bacterial intelligence” in biofilms described in the following section appear to be nudging us in this direction.

From colonies to communities (biofilms)

In fact, bacteria can go a step higher; once an entire colony becomes a new multi-cellular being [10-13] with its own identity, colonies functioning as organisms cooperate as building blocks of even more complex organizations of bacterial communities or societies, such as species-rich biofilms [14-19]. In this situation, cells should be able to identify their own self, both within the context of being part of a specific colony-self and part of a higher entity - a multi-colonial community to which their colony belongs. To maintain social cooperation in such diverse societies, the bacteria need “multi-lingual” skills for the identification and contextual interpretation of messages received from colony members and from other colonies of the same species and of other species, and to have the necessary means to sustain the highest level of dialogue within the “chattering” of the surrounding crowd [4].

25

For perspective, the oral cavity, for example, hosts a large assortment of unicellular prokaryotic and various eukaryotic microorganisms [14]. Current estimates suggest that sub-gingival plaque contains 20 genera of bacteria representing hundreds of different species, each with its own colony of ~1010 bacteria, i.e., together ~thousand times the human population on earth. Thus, the level of complexity of such microbial system far exceeds that of the computer networks, electric networks, transportation and all other man-made networks combined. Yet bacteria of all those colonies communicate for tropism in shared tasks, coordinated activities and exchange of relevant genetic bacterial information using biochemical communication of meaning-bearing, semantic messages.

The term social intelligence refers to human mental skills that are required to conduct successful social life, that are beyond the mathematical and academic ones connected with analytical intelligence [4]. So it is generally associated with humans’ special cognitive capacities such as perceiving self and group identity, perceiving self and group goals, engaging in adaptive social interactions, and acting together for personal and group benefit. Recently, it has been proposed that bacterial cooperation reflect features that can be referred to as primitive or fundamental aspects of social intelligence.

In multi-colonial communities (e.g. sub-gingival plaque), bacterial social intelligence is usually used for cooperation between colonies of different species [11,17,44]. For example, each colony develops its own expertise in performing specific tasks for the benefit of the entire community, and they all coordinate the work done.

Some bacteria undertake the task of keeping valuable information which is costly to maintain and may be hazardous for the bacteria to store. Frequently, such information is directly transferred by conjugation following chemical courtship played by the potential partners: bacteria resistant to antibiotics emit chemical signals to announce this fact. Bacteria in need of that information, upon receiving the signal, emit pheromone-like peptides to declare their willingness to mate. Sometimes, the decision to mate is followed

26 by exchange of competence factors (peptides). This pre-conjugation communication modifies the membrane of the partner cell into a penetrable state needed for conjugation.

By contrast, some fundamental aspects of social intelligence are used to handle defectors, as is reflected by the variety of strategies Myxobacteria can use when their social intelligence is challenged by cheaters – opportunistic individuals who take advantage of the group's cooperative effort [14-19]. For example, they can single out defectors by collective alteration of their own identity into a new gene expression state. By doing so, the cooperators can generate a new ‘dialect’ which is hard for the defectors to imitate. This ever-ongoing intelligence clash with defectors is beneficial to the group as it helps the bacteria improve their social skills for better cooperation which can be utilized at other times. For example, higher cooperation is needed when antibiotic stress is encountered.

A posed riddle: Bacteria sensing of weak EM signals Even more surprising is the bacteria’s ability to extract and store biologically meaningful information from LB growth media that has been exposed to a very weak EM signal, as has been demonstrated in the pioneering experiments described in [49]. In Fig 12-16 we show additional examples of such experiments, in which we compare colonial developments of bacteria previously grown in ordinary LB medium (control) and of bacteria grown in LB that had been exposed to weak EM signal for three hours. We emphasize that the bacteria were added to the LB medium 20 minutes after the test tube has been removed from the coil. Consequently the bacteria are added and the test tubes are placed in a shaker at 37èC for 24 hours to reach a stationary phase (a standard protocol). Then bacteria are taken from the test tubes for inoculation (5ml droplets) and Petri-dish growth under identical conditions.

27 The far-reaching implications of the observation of bacterial sensing of the effect of weak EM signal on water require intensive numerous repetitive experimental testing, and even more important, reconfirmation by different labs working on different strains of bacteria. Our studies will be elaborated in a separate, forthcoming publication. We would like to emphasize, though, that in the manner we set the experiments it is not a question of statistics. To be precise, in all the experiments (over a hundred) we obtained the same effect – the weak EM signal induced faster colonial expansion with a distinctively different patterns. In Appendix A we present methods for quantified comparison between the observed colonial patterns. However, the distinct differences between the control growth and the one caused by the EM signal are so pronounced that they are apparent to the naked eye.

We also emphasize that the results shown are typical (not the best) observed results. Additional tests show that a copper ring brings about a similar but weaker effect. We also used a coil driven by a pulse generator at the MHz range and obtained similar results. Similar results are also obtained for the Paenibacillus vortex bacteria. However, the effect is weaker as the typical growth of this strain is on harder surfaces (above 2% agar vs. below 1% agar for the experiments shown). Complementary studies on pattern formation in non-living systems (electro-chemical-deposition) also demonstrate a long- term effect of exposing the system to a weak EM signal on the emerged patterns [49]. However, as we have mentioned, it is crucial to have the experiments repeated by other groups and with other bacterial strains to explore the generality of the effect and to rule out overlooked artifacts.

28 -a- -b-

-c- -d- Fig 12: Reproducible memory effect of weak EM signal. (a) and (b) are two plates of colonial developments of the C morphotype of the Paenibacillus dendritiformis bacteria on soft surfaces (0.5% agar) with 10g/l peptone, following 24 hours growth in an LB medium (5 gram/liter (5g/l) NaCl; 5 g/l K2HPO4; 10 g/l Bacto-peptone) in a shaker. The bottom pictures (c,d), show the effect of EM signal. As described in the text, in this case the LB test tube was placed within a coil made of 10 loops of copper pipe (0.5cm diameter). The coil is 10cm high and 5cm in diameter. We show two plates for each case to illustrate the level of reproducibility and robustness of the observed effect. Note that, in addition to enhanced colonial development, the ‘engineered’ patterns are distinctively different and have a reversed handedness.

-a- -b-

29

-c- -d-

Fig 13: Generality of the effect of the EM signal: Additional examples of the memory effect of weak EM signal like in Fig 12 but for different strains. Left (a,c), A different strain of the C morphotype. Right (b,d) A strain of the T morphotype. The growth conditions are as in Fig 12. Top (a,b) are control and bottom ( c,d) are for LB exposed to weak EM signal.

-a- -b-

30

-c- -d-

Fig 14: Combined memory effect. The experiment is similar to Fig 12 but the growth is on surfaces with 0.6% agar and 5g/l peptone. Right (b,d), The effect of antibiotic stress. Septrin (at 0.001
m) was added to the LB. Comparison with the pictures on the left (a,c) shows the memory effect for growth in the presence of the antibiotic stress. Comparison of (b) and (d), shows that exposure to a very weak EM signal can almost eliminate the antibiotic effect. Interestingly, the growth for the latter case is even more enhanced than the control (a). Note also that in the presence of antibiotic the chirality is far weaker.

-a- -b-

31

-c- -d-

Fig 15: Transmission-Electron-Microscope (TEM) observations. The pictures show TEM observations of the individual bacteria taken after the 24 hours growth in the LB for the cases shown in Fig 14. The white bars are 1mm in (a), (b) and (c) and 2mm in (d). Comparison between the control growth in the absence of antibiotic (a) and in the presence of antibiotic(b) shows that the antibiotic stress leads to smaller bacteria with a different aspect ratio ( length vs. width) and modifications of the membrane structure. Comparison of the top and bottom pictures shows that the exposure to the EM signal leads to longer bacteria both for the control LB (comparison of (a) and (c)) and for that in the presence of antibiotic(comparison of (b) and (d)).

-a- -b-

32

-c- -d-

-e- -f-

Fig 16: Additional tests of the EM memory effect. In these pictures we illustrate three additional phenomena. The experiments are similar to those described in Fig 12 but Streptomycin was added to the LB. From studies of growth of the C morphotype on surfaces with Streptomycin we know that it leads to enhanced colonial expansion of these bacteria. Comparison of the control growth (pictures (a) and (b) in Fig. 12) with pictures (a) and (b) illustrates that the memory effect of exposure to Streptomycin also leads to enhanced colonial expansion. Pictures (c) and (d) show colonial expansion of bacteria taken from LB with Streptomycin that was exposed to weak EM signal. Comparison of these pictures with pictures (a) and (b) shows that the EM signal leads to an additional enhancement of the colonial expansion. As a preliminary test that the effect is sensitive to the filtering properties of the coil, we show in pictures (e) and (f) results of colonial expansion when the coil is shunted (the two terminals are electrically connected). Note that the effect in this case is distinguishably reduced. Moreover, the two cases have the opposite average handedness. That in the presence of open coil is the same as that observed for the open coil in Fig 12 and both differ from the control cases. We also tested that when the test tubes and the coil are in a Faraday cage the effect is erased.

33

Here we would like to mention that the long term (hours) effects of electromagnetic (EM) signals on water have been studied for many years. In those studies, water is usually exposed to RF radiation (MHz to GHz frequencies) for about 30 minutes.

Previous studies [49] of exposure of LB to such treatment revealed that bacteria grown in such treated LB exhibit enhanced colonial developments during consequent growth on a surface. From physics and chemistry perspective, the experiments pose a riddle since it is not clear what kind of mechanism can sustain such long term effects or ‘water memory’ of the exposure to the EM signals. From a biological perspective, the puzzle is what nature of imprinted structural (water-formatics) effect can affect biological functioning to the extent that it is stored by the bacteria.

A possible scenario can be based on the special interaction of EM signal and water bubbles: It has been shown that RF treatment of water affects its composition (number and distribution) of bubbles, including the formation of nano-bubbles that presumably form a network with hierarchical (multi-scale) organization. Once formed, such a network can have long term stability. Moreover, the hydrophobic nature of the bubbles might induce long-range orientation and structural correlation order. In all the cases, the pre-exposure of the LB to weak EM signal leads to enhanced colonial development. This happens even when the LB contains antibiotics. In addition to the effect on colony expansion, it leads to distinctively ‘engineered’ colonial patterns, including reversal of the colonial average handedness, as is most obvious in Fig 12. In Fig 15 we show TEM observations of the individual bacteria at the end of the 24 hour growth period in LB.

Concluding remarks and looking ahead In this article, we presented some of the remarkable patterns formed during colonial self- organization when the bacteria are exposed to versatile growth conditions, mimicking the ones they are likely to encounter in Nature. Endless, simple yet ingenious methods were invented by the bacteria for developing the incredible diversity of observed complex

34 patterns. They illustrate the bacterial schemata of “Communication-Based Cooperation’ for ‘Complexity-Based Adaptability’ which enable them to thrive and improve in a changing environment [5,13].

Recent findings even indicate that the bacteria purposefully modify their colonial organization in the presence of antibiotics in ways which optimize bacterial survival, and that the bacteria have a special collective memory which enables them to keep track of how they handled their previous encounters with antibiotic – learning from experience [4,7].

We now begin to realize the power of bacterial cooperation and social intelligence that allow them to learn from experience when solving newly encountered problems and then to share their new skills far and wide. The new findings bear the promise to provide satisfactory explanations to bacterial threat for our health: that an increasing number of bacterial strains of disease-causing bacteria can today resist multiple antibiotic drugs. Bacteria are clearly capable of developing antibiotics resistance at a higher rate than scientists develop new drugs, and we seem to be losing a crucial battle for our health. To reverse this course of events, we must outsmart the bacteria by taking new avenues of study which will lead to the development of novel fighting strategies. One such promising direction is to perform gene-expression studies during colonial development, when bacteria are exposed to antibiotic stress and during bacterial learning from experience [7].

We expect that such and other future experiments will soon lead us to reverse the current notion of bacteria as mere solitary and simple creatures with limited capabilities, and recognize that bacteria are cooperative organisms that lead complex communal life with rapidly evolving social intelligence [4,7,13]. We might even discover that the last five decade’s evolution in bacterial social intelligence is largely a result of their encounter with our socially irrational massive use of antibiotic materials in agriculture and human intake.

35

Acknowledgments We are thankful to Inna Brainis for her devotion and technical precision in performing the experiments described in this manuscript and to Yael Katsir for experimental assistance and image processing of the bacterial patterns. We are thankful to Alfred Tauber for many illuminating conversations, criticisms and suggestions, especially with regard to bacterial response to weak EM signals. This article is based in part on previous reviews prepared in collaboration Herbert Levine [4-6], and on a public lecture entitled “Why Bacteria Go Complex – Higher Complexity for Better Adaptability, presented at the “Biocomplexity VI” conference on “Complex Behavior of Unicellular Organisms”, Bloomington Indiana, May 2004. One of us (EBJ) would like to thank the organizers and the participants of that conference, for enlightening meeting and inspiring conversations. He also thanks the Institute for Advanced Studies at Indiana University Bloomington for their hospitality during May 2004. This research has been partially supported by the Maguy-Glass Chair in Physics of Complex Systems.

Appendix A: Towards quantified measures of the patterns

There are many image processing methods that provide different quantified measures to characterize and classify observed patterns. Here we follow the example shown in [49], developed to capture the specific features of flexibility and control reflected in the patterns. It has been proposed that "regulated freedom" of response is a necessary requirement for bacterial cooperation, and that enhancement of "regulated freedom" and cooperation are mutually connected [13]. The result is higher colonial complexity and flexibility that afford it elevated adaptability. The challenge is to convert these seemingly blurred statements into testable biotic relations. The first crucial step is to demonstrate that the problematic intuitive notions can be associated with corresponding quantifiable measures in the observed colonial patterns. Note that freedom here is in the physics sense of "degrees of freedom", and regulation is in the sense of self-control, borrowed from engineering.

36

The method in [49] begins with drawing a binary sequence along a spiral or a set of circles that covers the pattern (Fig. A1-a). The binary sequence of a pattern is a binary vector of Nb elements such that B(i) = 1 if the i-th bin corresponds to a local maxima in the recorded bacterial density and zero otherwise. It was found that the probability density functions (pdf) for different bacterial strains can be approximated by symmetric Lévy distributions as shown in Fig. A2. The symmetric Lévy distribution is characterized by two parameters: the index of stability a subjected to 0

The Cauchy limit of a = 1 represents the limit of co-enhancement of regulation and freedom. Thus, in this limit the colonial structure has higher flexibility. Additional manifestation of the above is provided by the increase in a with decrease in
(reduced freedom and elevated regulation) in response to imposed antibiotic stress [13,64].

Figure A1: Illustration of the extraction of a binary sequence from an image of a bacterial colony [49]. (a) A spiral line is drawn on the grayscale image of the bacterial colony. (b) Along the spiral, the gray levels are sampled into an array, to form a continuous

37 presentation of the variations in the bacterial densities. (c)The array in (b) is converted into a binary sequence in which “1” marks the location of each local density maximum, following an appropriate cross graining above the noise level.

-a- -b-

Figure A2: Statistical analysis of the image. (a) The pdf of the intervals of the binary sequence shown in Fig A1-c. Note that the distribution has a minimal interval and a most probable one (maxima in the distribution). (b)The pdf of the increments of the intervals are plotted with a Lévy approximation (solid line) with parameters: =1.07 and =12.

Recently, in the study of neural networks activity, new measures of regularity and structural complexity for sequences with Lévy-like distributions have been developed [65]. In the context of colonial patterns we expect that the regularity binary sequence will be associated with the external stress. We also expect that the structural complexity will provide measures associated with the colony flexibility and adaptability.

Appendix B: The art of modeling In physics and chemistry, models have long served as an indispensable research tool. But in the life sciences the predictive power of models is still largely questioned. One can easily fall into the ‘reminiscence trap’ - the tendency to devise a set of rules that will just mimic aspects of observed phenomena. The challenge is to elicit the generic features and basic principles needed to explain behavior from experimental observations and biological knowledge.

38

With present computational power, it is only natural to use computer models to study bacterial cooperation in attempt to reveal their secret strategies. In physics and chemistry, computer models have long served as an indispensable research tool. But in the life sciences many still question the value and predictive power of model simulations. The research strategy presented here assumes that models can provide a crucial research tool in the study of bacterial pattern formation. The usefulness and crucial role of modeling in the study of bacterial complex organization is illustrated in the communicating walker model presented below. We try to demonstrate that generic modeling allows in a natural way for both the physics (represented by chemical fields such as the nutrient concentration, any chemical signals, the lubricating fluid flow) and the biology (the response of individual cells to their sensed environment) to be represented in a computationally tractable format. Yet models should never be viewed as precise analogs of the actual system, but instead as tools for unraveling cause and effect and for helping in the quest for as yet unknown biological phenomena at the single bacterium scale.

Models can also help with placing the phenomena of bacterial self organization within the context of the proper mathematics of pattern-formation schema in non living systems. However, doing so deserves special care. One must keep in mind the profound differences due to the functionality requirements standing behind the patterns generated by living systems for their survival and their ability to evolve and learn from experience. We thus propose to harness the new mathematical understandings of pattern formation in non-living systems for the study of patterning in the living world while avoiding the mathematification of life.

39 To illustrate the above we proceed to describe the communicating walker’s model devised to study branching colonial patterning. This is a special hybrid model: On one hand, the diffusion of nutrients is described by the continuum modeling approach or reaction-diffusion equations. And on the other hand, the bacterial cells are described by discrete walking and communicating cellular automatons in the spirit of the molecular dynamics modeling approach.

Even with today’s most advanced computers, it is beyond computer power to handle tens of billions of automatons which is the number of bacterial population. Therefore in the model each walker is taken to represent about 100,000 cells.

We start the model by drawing an imaginary mesh of hexagonal boxes as is shown in the figure. At the beginning of the simulations all the boxes have the same level of food Co which set according to the initial concentration of food in the experiment. As the simulations proceed the amount of food in each box is depleted as it is consumed by walkers in the box. In addition food diffuses between the boxes according to the ordinary diffusion kinetics.

Unlike in ordinary cellular automata models, each walker (i) is assigned an internal degree of freedom described by an ‘internal energy’ Ei . The internal energy is continuously used for metabolism. Yet when sufficient food is available the walker can consume food fast enough to compensate for the metabolic needs and Ei increases until it reaches a threshold level and the walker replicates into two walkers. When the food level is not sufficient, the internal energy decreases. After a walker is starved for a long, Ei drops to zero and the walker `freezes’. This `freezing’ represents the bacterial transition into a pre-spore state.

Proper modeling of bacterial collective lubrication poses a greater challenge. In the figure we illustrate a relatively simple and computationally efficient way it can be done. First the envelope of the lubricant is marked on the imaginary hexagonal mesh. The walkers perform a free random walk within the envelope. We then assign a counter to each

40 segment of the envelope to count the number of times walkers hit the segment. After NH we move the segment one step forward. This requirement of NH hits represents the colony propagation through wetting of unoccupied areas by the cells. Note that NH is related to the substrate hardness, as more lubrication fluid has to be produced (more `collisions’ are needed) to push the boundary on a harder substrate.

The next crucial step is relating the model’s parameters with real bacterial growth. The food diffusion is known and the bacterial random walk is of about micron per second. The weight of a bacterial cell is about 10-12 gram (1pg). We assume that each cell consumes about 3 pg of food for reproduction: 1 pg for the materials of the new cell, 1 pg for metabolism and 1 pg to `pay’ for decreasing the entropy while converting the food into new cell organic substances. The maximal consumption rate per one cell is 3 pg during the minimal reproduction time which is about 25 minutes. To include chemotactic signaling in the model, each walker emits the chemical agent according to internal state of the walkers – chemo-attractant is emitted when the internal energy is high and the chemo repellent is emitted by the frozen walkers. In addition the walker’s random walk is biased according to the concentration gradient of the chemotactic chemicals.

Fig B1. Typical results of the communicating random walker’s model. On the left we show the lattice used in the simulations and the walkers. Red are the active ones and the blue the

41 ones further back from the front that already froze. The picture on the right shows a typical result of numerical simulations. The color code indicates the time steps.

It is known that the flagella perform a propeller-like movement with a specific handedness resulting from the gear like structure of their molecular motors. Still, the observation of a pattern of strongly twisted, chiral branches, all with the same handedness, was unexpected since it was not clear how the flagella handedness can determine the global structure of the colony. Indeed, flagella handedness has no effect on the tumbling of short bacteria as they make many turns before ending at a new angle. Optical microscope observations indicate that, in the case of chiral branching, the cells are much longer with liquid crystal like orientational interaction. This co-alignment of the longer bacteria limits their tumbling thus turning the two-dimensional random walk (of short bacteria) into a quasi-one-dimensional movement back and forth as the bacteria elongate. Only bacteria close to the branch tip can perform larger rotation which imposes a very weak bending on the advancement of the branch-tips. Put together, the flagella handedness is made to act as a singular perturbation in a manner similar to that crystalline symmetry, thus leading to the colonial chiral patterning. These ideas can be tested by simulations of a version of the communicating walker’s model in which the walkers tumble with a specific handedness and have an orientation interaction. Thus the random walk becomes quasi 1D and only walkers close to the tip of the branches can twist. This model generates the observed broken reflection symmetry of the chiral morphotype as is illustrated in Fig 3-d.

We have seen how vortices contribute to the overall colony structure by leading the expansion over extremely hard substrates. The basic idea that needs to be incorporated in any model of the vortex is that the bacteria are self-propelled [5]. Of course the communicating walker model already deals with particles propelled by their metabolic energy, but here the emphasis is on motion that remains coordinated for long time, i.e. for situations where the tumbling into a new direction is suppressed by interactions with the neighboring cells. In this sense, there is a progression from bacterium motion in the

42 branching pattern (random walker modulated by chemotaxis) to the chiral state (orientational effects lead to discrete reflection-symmetry breaking but are not strong enough to retain long-time coherence) to the vortex structure (with a breaking of the continuous rotational symmetry). An interesting mathematical point is that it was used to be thought that picking a particular motion direction out of a continuous set of choices was impossible in two dimensions. But the bacteria reminded us that facts about equilibrium structures, where the above was rigorously established, are routinely violated in non-equilibrium open systems [5].

43

References

[*] Bacterial images and video clips are available from PhysicaPlus – the online magazine of the Israel Physical Society http://physicaplus.org.il and at Ben Jacob’s home page http://star.tau.ac.il/~eshel/

[1] Schrödinger, E. (1944) What is life? The Physical Aspect of the Living Cell Cambridge University Press; (1958) Mind and Matter Cambridge University Press; (1992) What Is Life? The Physical Aspect of the Living Cell with Mind and Matter and Autobiographical Sketches with forward by R. Penrose

[2] Margulies, L. and Dolan, M.F. (2002) Early life: Evolution on the Precambrian Earth Jones and Bartlett

[3] Liebes, S., Sahtouris, E., Swimme, and Liebes, S. (1998) A Walk Through Time: From Stardust to Us Wiley

[4] Ben Jacob, E. et al (2004) Bacterial Linguistic Communication and Social Intelligence Trends in Microbiology 12 (8) 366-372

[5] Levine, H. and Ben Jacob, E. (2004) Physical Schemata Underlying Biological Pattern Formation-Examples, Issues and Strategies J. Physical Biology 1, pp 14-22

[6] Ben Jacob, E. and Levine, H. (2004) Des fleurs de bactéries Les formes de la vie Dossier Pour La Science 44 (July),

[7] Ben Jacob, E. and Shapira, Y. (2004) Meaning-Based Natural Intelligence vs. Information-Based Artificial Intelligence. The Cradle of Creativity (in press)

[8] Shapiro, J.A. (1995) The significance of bacterial colony patterns. Bioessays 17, 597-607.

[9] Shapiro, J.A. and Dworkin, M. (1997) Bacteria as Multicellular Organisms, Oxford University Press

44 [10] Shapiro, J.A. (1998) Thinking about Bacterial Populations as Multicellular Organisms. Ann. Rev. Microbiology 52, 81-104

[11] Ben-Jacob, E. et al. (1998) Cooperative organization of bacterial colonies: from genotype to morphotype. Ann. Rev. Microbiology 52, 779-806

[12] Ben Jacob, E., Cohen, I. and Levine, H. (2000) Cooperative Self-Organization of Microorganism. Adv. Phys. 49, 395-554

[13] Ben-Jacob, E. (2003) Bacterial self-organization: co-enhancement of complexification and adaptability in a dynamic environment. Phil. Trans. R. Soc. Lond. A361, 1283-1312,

[14] Dworkin, M. (1996) Recent advances in the social and developmental biology of the myxobacteria. Microbiol. Rev. 60, 70-102

[15] Rosenberg, E. (Ed.) (1999) Microbial ecology and infectious disease, ASM Press Washington DC

[16] Strassmann, J.E. (2000) Bacterial Cheaters Nature 404, 555-556

[17] Velicer, G.J. et al (2000) Developmental cheating in the social bacterium Myxococcus xanthus. Nature 404, 598-601

[18] Crespi, B.J. (2001) The evolution of social behavior in microorganisms. Trends Ecol. Evol. 16, 178-183

[19] Velicer, G.J. (2003) Social strife in the microbial world. Trends Microbiol. 7, 330-337 [20] Salmond, G.P.C. et al. (1995) The bacterial enigma: Cracking the code of cell- cell communication. Mol. Microbiol. 16, 615-624

[21] Wirth, R. et al.. (1996) The Role of Pheromones in Bacterial Interactions. Trends Microbiol. 4, 96-103

45

[22] Dunny, G.M. and Winans, S.C. (1999) Cell-Cell Signaling in Bacteria, ASM Press

[23] Bassler, B.L. (2002) Small talk: cell-to-cell communication in bacteria. Cell 109, 421-424

[24] Xavier, K.B. and Bassler, B.L. (2003) LuxS quorum sensing: More than just a number game, Curr. Opin. In Microbiol. 6, 191-197

[25] Mok, K.C., Wingreen, N.S. and Bassler, B.L. (2003) Vibrio harveyi quorum sensing: A coincidence detector for two autoinducers controls gene expression. EMBO J. 22, 870-881

[26] Miller, M.B. (2002) Parallel quorum sensing systems converges to regulate virulence in Vibrio cholerae. Cell 110, 303-314

[27] Matsushita, M. and Fujikawa, H. (1990) Diffusion-limited growth in bacterial colony formation. Physica A 168, 498-506

[28] Ohgiwari, M. et al. (1992) Morphological changes in growth of bacterial colony patterns. J. Phys. Soc. Jpn. 61, 816-822

[29] Di Franco, C. et al. (2002) Colony shape as a genetic trait in the pattern-forming Bacillus mycoides. BMC Microbiol 2(1):33

[30] Ben-Jacob, E. et al. (1994) Generic modeling of cooperative growth patterns in bacterial colonies. Nature 368, 46-49

[31] Kuner, J.M. and Kaiser, D. (1982) Fruiting body morphogenesis in submerged cultures of Myxococcus xanthus. J. Bacteriol. 151, 458-461

46 [32] Shimkets, L.J. (1999) Intercellular signaling during fruiting-body development of Myxococcus xanthus. Annu. Rev. Microbiol. 53, 525-549

[33] Komoto, A. et al (2003) Growth dynamics of Bacillus circulans colony. J. Theo. Biology 225, 91-97 [34] Ben Jacob, E. and Levine, H. (1998) The Artistry of Microorganisms Sci Am 279, 82-87 [35] Ben Jacob, E. and Levine, H. (2001) The Artistry of Nature Nature 409, 985- 986

[36] Kessler, D. A., Koplik, J. & Levine, H. (1988) Pattern selection in fingered growth phenomena. Adv. Phys. 37, 255-339

[37] Langer, J. S. (1989) Dendrites viscous fingering and the theory of pattern formation. Science 243, 1150-1154

[38] Ben-Jacob, E. & Garik, P. (1990) The formation of patterns in non-equilibrium growth. Nature 33, 523-530

[39] Ben-Jacob, E. (1993) From snowflake formation to growth of bacterial colonies. I. Diffusive patterning in azoic systems. Contemp. Phys. 34, 247-273

[40] Ptashne, M. and Gann, A. (2002) Genes and signals, Cold Spring Harbor Press

[41] Searls, D.B. (2002) The Language of genes. Nature 420, 211-217

[42] Kolenbrander, P.E. et al (2002) Communication among oral bacteria. Microbiol. Mol. Biol. Rev. 66, 486-505

[43] Kozlovsky, Y., Cohen, I., Golding, I. & Ben-Jacob, E. (1999) Lubricating bacteria model for branching growth of bacterial colonies. Phys. Rev. E59, 7025-7035 ; Ben-Jacob, E. et al (2000) Modeling branching and chiral colonial patterning of lubrication bacteria. In Mathematical models for biological pattern formation Edited by P. V. Maini and H. G. Othmer Springer ; Cohen, I. et al (2001) Biofluiddynamics of lubricating bacteria Mathematical Methods in the Applied Sciences 24, no.17-18, pp.1429-68

47 [44] Golding, I. et al (1999). Studies of sector formation in expanding bacterial colonies Europhysics Letters, 48(5), Pp 587-593 ; Ron, I. et al (2003) Bursts of sectors in expanding bacterial colonies as a possible model for tumor growth and metastases Physica A, 320, 485-96

[45] Ben-Jacob, E. (1997) From snowfake to growth of bacterial colonies. II. Cooperative formation of complex colonial patterns. Contemp. Phys. 38, 205-241

[46] Matsuyama, T., Kaneda, K., Nakagawa, Y., Isa, K., Hara-Hotta, H. and Yano, I. (1992) A novel extra cellular cyclic lipopeptide which promotes flagellum-dependent and -independent spreading growth of Serratia marcescens. J. Bacteriol. 174, 1769- 1776

[47] Matsuyama, T., Hershey, R. M. and Matsushita, M. (1993) Self-similar colony morphogenesis by bacteria as the experimental model of fractal growth by a cell population. Fractals 1, 302-311

[48] Matsushita, M. et al (1998) Interface growth and pattern formation in bacterial colonies Physica A, .249, 517-524

[49] Nadav Raichman, N. et al (2004) Engineered Self Organization In Natural And Man-Made Systems in Continuous Media and Discreet Systems Eds. Bergman, D.J. and Inan, E. KLUWER PUBL

[50] Berg, H. C. (1993) Random walks in biology. Princeton, NJ: Princeton University Press.

[51] Budrene, E.O. and Berg, H.C. (1991) Complex patterns formed by motile cells of Esherichia coli. Nature 349, 630-633

[52] Blat, Y. and Eisenbach, M. (1995) Tar-dependent and -independent pattern formation by Salmonella typhimurium. J. Bacteriol. 177, 1683-1691

[53] Budrene, E.O. and Berg, H.C. (1995) Dynamics of formation of symmetrical patterns by chemotactic bacteria. Nature 376, 49-53

[54] Ben-Jacob, E. et al (1995) Complex bacterial colonies. Nature 373, 566-567

48

[55] Charon, N.W. and Goldstein, S.F. (2002) Genetics of motility and chemotaxis of a fascinating group of bacteria: The spirochetes. Annual Review of Genetics; January 01

[56] Di Franco, C. et al. (2002) Colony shape as a genetic trait in the pattern-forming Bacillus mycoides. BMC Microbiol 2(1):33

[57] Komoto, A. et al (2003) Growth dynamics of Bacillus circulans colony. J. Theo. Biology 225, 91-97

[58] Cohen, I., Ron, I.G. and Ben Jacob, E. (2000) From branching to nebula patterning during colonial development of the Paenibacillus alvi bacteria Physica A 286, 321-336

[59] Cohen, I., Czirok, A. & Ben-Jacob, E. (1996) Chemotactic based adaptive self- organization during colonial development. Physica A233, 678{698.

[60] Czirok, A. et al. (1996) Formation of complex bacterial colonies via self- generated vortices. Phys. Rev. E 54, 1791-1801

[61] Ball, P. (1999) The self-made tapestry - Pattern formation in nature. Oxford University Press

[62] Ben Jacob, E. et al (2000) Bacterial cooperative organization under antibiotic stress, Physica A 282, 247-282

[63] Golding, I. and Ben Jacob, E. (2001) The Artistry of Bacterial Colonies and the Antibiotic Crisis in: Coherent Structures in Complex Systems, Springer-Verlag, Heidelberg

[64]Ben Jacob, E. et al. (2003) Communication-based regulated freedom of response in bacterial colonies Physica A 330 218-231

[65] Hulata, E.; Baruch, I.; Segev, R.; Shapira,Y.; Ben Jacob, E.(2004) Self-regulated complexity in cultured neuronal networks. Phys. Rev. Lett. 92(19), art-198105.

49