Extinction transition in bacterial colonies under forced convection T. Neicu,1 A. Pradhan,1 D. A. Larochelle,2 and A. Kudrolli1,† 1Department of Physics, Clark University, Worcester, Massachusetts 01610 2Department of Biology, Clark University, Worcester, Massachusetts 01610 (October 15, 1999) We report the spatio-temporal response of Bacillus subtilis growing on a nutrient-rich layer of agar to ultra-violet (UV) radiation. Below a crossover temperature, the bacteria are confined to regions that are shielded from UV radiation. A forced convection of the population is effected by rotatingaUVradiationshieldrelativetothepetridish. Theextinctionspeedatwhichthebacterial colony lags behind theshield is found to be qualitatively similar to the front velocity of the colony growingintheabsenceofthehostileenvironmentaspredictedbythemodelofDahmen,Nelsonand 0 Shnerb. A quantitative comparison is not possible without considering the slow dynamics and the 0 time-dependent interaction of the population with thehostile environment. 0 2 PACS number(s): 87.50.g, 05.70.Ln, 87.10 n a Bacterialcoloniesgrowingonanutrientrichsubstratehaveservedasmodelsystemsforstudyingpatternformation J and population dynamics in biological systems. Studies with strains of Bacillus subtilis and Escherichia coli have 3 reportedawide varietyof complexpatterns depending onnutrientconditions [1–4]. The patterns havebeen modeled using reaction-diffusion equations [5–7]. These experimental and theoretical studies have considered an essentially 1 v uniformenvironmentwherethechangesaredueonlytothedepletionofnutrientswithtime. However,livingorganisms 3 often are forced to migrate due to changes in the environment. 2 The modeling of population dynamics of bacterial colonies due to changes in the environment has been studied 0 recently by Shnerb, Nelson, and Dahmen [8,9]. Their theoretical model incorporates the effect of a forced convection 1 on the growth of a bacterial colony by considering the convective-diffusion equation given by: 0 0 ∂c(x,t) 0 =D∇2c(x,t)−v·∇c(x,t)+U(x)c(x,t)−bc2(x,t), (1) / ∂t t a m where c(x,t) is the bacteria number density, D is the diffusion constant of the bacteria, U(x) is the spatially varying growth potential, v is an externally imposed convection velocity, and b is a parameter that limits the population - d number density to a maximum saturation value. If v = 0 and U(x) is constant, Eq. (1) corresponds to the Fisher n wave equation [10] which has a solution with a limiting constant value of the front speed v . Wakita et al. [11] have F o studied a colony of Bacillus subtilis in a high nutrient and low agar medium growing in such a Fisher mode. c ThetwonewfeaturesoftheforcedconvectionmodelgivenbyEq.(1)aretheintroductionofagrowthpotentialU(x), : v corresponding to exposing photosensitive bacteria to a light source for example, and the convection of the bacteria Xi due to the motion v of the light source. By considering a colony confined to a rectangular region, the resulting steady-state number density of the bacteria (the time independent solution of Eq. (1)) was obtained in Ref. [9] as r a a function of v. They concluded that that the total number of bacteria in the rectangular region decreases linearly to zero as v approches v from below. The steady-state spatial density distribution was obtained by solving for the F time-independent solutionsofEq.(1)numerically. Becausethe linearizedversionofEq.(1) allowsa mapping to non- Hermitian quantum mechanics, additional predictions of the properties of bacterial colonies in terms of localization- delocalization transitions in quantum systems can be made [8]. We reportthe firstexperimentalstudy ofaBacillus subtiliscolonyforcedtomigratebyenvironmentalchangesdue to a moving ultra-violet (UV) source. UV radiation is shined on a petri dish containing nutrient rich agar except in a rectangular region which is shielded. Although UV radiation is supposed to kill these bacteria [12], we find more subtle behaviors. For example, the colony is confined to the shielded region only when the temperature is below a “crossover” value of approximately 22◦C. When the UV radiation is turned off, the front of the colony which was near the boundary between the hostile and favorable regions initially grows slowly, but recovers to the Fisher front speed v in about 25 hours (h). This slow recovery near the boundary suggests the presence of signalling between F the bacteria, a feature which is absent in Eq. (1). To study the effect of a changing environment, we rotate the rectangular shield with a constant angular velocity relative to the petri dish. The bacteria are inoculated along a line inside the rectangular shield region. The rotation resultsinthecolonybeingforcedtoconvectwithvelocitiesthatincreaselinearlyfromzeroattheaxisofrotationtoa maximum value at the edge of the plate. We find that the bacteria colony cannot keep up with the shielded region if 1 the shieldmoveswithvelocitiesmuchgreaterthanv ,thus showinganextinctiontransitioninqualitativeagreement F with the theoretical model [9]. The spatial number density n(x,v) of the bacteria as a function of the speed v was measured and found to be time-dependent, even after three days of forced convection. These experimental results illustrate the relevance of Eq. (1) and also indicate that the time-dependent response is of experimental relevance because of the long time scales of biological systems. We now describe our experimental procedure and observations in more detail. The wild-type strain of Bacillus subtilis was obtained from Presque Isle Cultures and freeze dried at −70◦C. All experiments were performed from this initial sample by incubating a portion of the sample for 8 h at 30◦C in nutrient rich broth. A drop of this broth representingatotalofatleast107bacteriaisusedtoinoculatethenutrientrichagar. Theexperimentswereperformed in 15cm diameter plexiglass petri dishes containing a thin layer of nutrient agar (7grams/liter of bacto-peptone and 3grams/liter agar.) These conditions are similar to that used in previous observations of the Fisher wave mode [11]. Wheninoculatedasasinglepointsource(diameter∼3mm),thegrowthofthecolonywasobservedtohaveauniform disk shape with a front velocity that increases slowly for the first 8 h and eventually reaches a constant front speed v consistent with previous work [11]. Experiments were performed over a range of temperature (21◦ −40◦C) and F it was found that v is an increasing function of temperature within this range with v = 1.7µm/s at 40◦C and F F v =0.19µm/s at 21◦C. F Next we describe ourexperiments in whichwe shine UVradiationonthe petri dish using two8W long wavelength UV-lamps placed 5cm above the dish. An Aluminum sheet is used to shield a rectangular region of the petri dish fromthe radiation(see Fig.1). The density ofthe bacteria is obtainedby imaging the light scatteredby the bacteria with a CCD camera. Calibration experiments show that the light intensity is proportional to the bacteria density. The colony at time t = 23.15h after a point inoculation at the center of the petri dish is shown in Fig. 1a for 21◦C. The shielded region is within the dashed lines and has a width w = 5cm. We observe that the front of the colony is circular and its diameter is smaller than w. As the colony grows further outward, the edge of the shielded region is reached, and the shape of the colony is no longer circular as shown in Fig. 1b. The width of the colony along the axes parallel (x) and perpendicular (y) to the shield is plotted in Fig. 1c. The error bars correspond to the range of fluctuations due to slightly different initial conditions in different runs. The diameter d of the bacterial colony growing at the same temperature in a petri dish which is completely shielded from UV radiation is also shown. We observe that the colony under the shield grows with a speed comparable to v at that temperature. As the colony F approaches the edge of the shield, the front speed slows down because the bacteria are confined. To further demonstrate that the confinement effect is due to the presence of UV radiation, the UV radiation was turned off after 72 h. We observe no change in the velocity of the front along the x-direction as expected, because the bacteria are deep inside the shield. However, we would expect to see a change in the rate of growth along the y-directionbecause the radiationhas been removed. We observethatthe front velocityrecoversto v , but only after F 25 h. This behavior is not modeled by Eq. (1), but is important in our discussion of the convection experiments as discussed below. We performedexperiments at higher temperatures andobservedthat for temperatures greaterthan approximately 22◦C, the bacteria are able to grow into irradiated regions, but with a front speed that decreases with time. (At 26◦C, the speed was reduced by 41% after 12 h.) Hence, in the presence of radiation we can vary the growth rate by changing the temperature and obtain a transition from a localized colony to one which is delocalized. A detailed study of this phenomena would be an interesting avenue for further research. In this paper we will consider a simple case in which the bacteria are confined at a temperature of 22±1◦C to investigate the extinction transition in the presence of convection. The convection experiments were performed by inoculating the bacteria along a diameter of the petri dish. The petri dish is then kept under a radiationshield of width w =4.3cm and placed on a rotating platform, similar to the experimentsdescribedearlier. Astheplatformrotates,theregionshieldedfromtheUVradiationadvancesataspeed which increases linearly from the axis of rotation outward. The colony was initially allowed to grow for 14 h before the platform was rotated. During this time the bacteria covered the shielded region. The time t = 0 corresponds to the time at which the platform was rotated. The results of the colony growth under these conditions are shown in Fig. 2. The position of the shielded region and the axis of rotation are indicated. The bacterial population is clearly seen to migrate and follow the shielded region at low velocities near the axis of rotation and lag behind at higher velocities. TheseobservationsareconsistentwiththetheoryofRef.[9]whereaphasediagramforthegrowthandtheextinction ofacolonyasafunctionofthegrowthpotentialU andtheconvectionspeedvwasobtainedusingEq.(1). Inparticular, itwaspredictedthatthebacteriawillbelocalizedtothefavorableregion. Furthermore,thetotalbacterialpopulation in favorable regions decreases linearly to zero as a function of v as v approaches v from below. (In this theory the F critical extinction speed v is the same as v .) To make quantitative comparisons, we have extracted the positions c F 2 of the fronts corresponding to the three images shown in Fig. 2. These positions are plotted in Fig. 3a; the origin corresponds to the axis of rotation and the initial line of inoculation is along the horizontal axis. The dashed arc in Fig. 3a corresponds to the distance where the velocity of the shield is the same as v . We observe that very far from F the axis of rotation, the front does not change during the time t = 46.56 h to t = 73.73 h, indicating that bacteria which could not cope with the speed of the shield were left behind in the hostile irradiated region and did not grow. Dividingthedisplacementofthebacteriafrontbythetimedifferencebetweenimages,weextractedtheapproximate velocity ofthe frontas a function ofthe radialdistance r. Suchananalysisignoresthe diffusion ofthe bacteriaalong the radial direction. The data for the average velocity of the front, v (r), is plotted in Fig. 3b. The velocity of the b shield v(r) also is plotted to provide a reference for the front velocities. The bacteria are confined to the shielded region, and v (r) is observed to increase with r, but is always less than v(r), the corresponding speed of the shield. b The reasonfor this lag might be due to the slow recoveryof the bacteria after the UV irradiatedregion moves ahead as discussed earlier in reference to Fig. 1c. We also observe that v (r) increases linearly up to a velocity of 0.2µm/s b which corresponds to r ∼ 45mm. For greater r, v (r) decreases and the bacteria increasingly lag behind the shield b and stop growing for r >80mm. The maximum value of v corresponds to the value of v of the bacteria colony at b F 22◦C in the absence of convection and UV radiation. To explain the velocity data for r > 50mm, we note the following. In the interval of time corresponding to the images shown in Figs. 2b and 2c, the point where the bacteria completely lag behind the shield decreases from r = 75mm to r = 59mm. During this time, the bacteria are exposed to UV radiation for at least part of this time interval which increases for larger r. Hence, because the bacteria grow for only a portion of the total time, the mean front speed v (r) decreases. The front speed is zero when the bacteria are always in the UV radiation corresponding b to r >80mm in Fig. 3b. We also note that because of the slow rate of growth of the colony, the relative slow speed of the shield v, and the finitewidthoftheshieldw,alongtransienttimeoftheorderofw/(v−v )isrequiredfortheshieldtoleavethecolony c whichisgrowingwithaspeedv lessthanthecriticalextinctionspeedv . Thistransienttimedivergesasv approaches c v . Hence, for an experiment which is conducted over a finite duration, the value of r where the bacteria completely c lag behind the shield is larger than the value corresponding to the critical extinction speed v . However, v can be c c indirectly calculated from the above relation for the transient time. We obtain the estimate v ∼0.23µm/s which is c similar to the value of v ∼0.26 at22◦C. This estimate was obtained fromthe image in Fig. 3 using v =0.4µm/sat F r = 59mm where the bacteria have completely lagged behind the shield at time t = 73.73h of rotation. Therefore, we find a critical extinction speed consistent with the Fisher wave velocity as predicted in Ref. [9]. A more direct comparison of our experimental results to theory can perhaps be made by considering the time dependentresponseofthemodelconsideredinEq.(1). Additionalconsiderationssuchasthetime-dependentresponse ofthefrontspeedofthebacteriamayhavetobeincorporated. Toencouragefuturecomparisonsofexperimentaldata with time-dependent models, we plot the number density n(x,v) of the bacteria colony at different distances from the shield in Fig. 4 correspondingto different convectionvelocities v. The shielded regionnormalizedby the width w correspondsto −0.5to 0.5. This datacorrespondsto the imageshowninFig.2c. These densitydistributions arestill timedependentexceptatv =0.41µm/s,whichcorrespondstodistanceswherethebacteriaareimmobilebecausethey havebeeninthe UVirradiatedregionsfor alongtime. We observethatthe frontofthe colonyinthe directionofthe convectionvelocity always lags behind the edge of the strip. This characteristicof the bacteria distribution is similar to that predicted in Ref. [9], but a direct comparison is not possible because the distribution is still time-dependent after t=73 h of rotation. From Fig. 4 we further observethat the total bacterialpopulationgiven by the areaunder thecurvedecreasesforincreasingvelocity. Wehavefounditimpracticaltoconducttheexperimentsforalongertime, which is a significant limitation in making a more direct comparison with time-independent predictions. The fact that the extinction transition occurs near v is an interesting result for real biologicalsystems because of F the relatively simple model consideredin Refs. [8,9]. Our experiments are an importantfirst step in investigating the usefulness of convection-diffusion models in studying convection in biological systems. The question remains if the observedevolutionofthe frontcanbe capturedby the time-dependence in Eq.(1) with the same initialconditions or if additional terms which include the time-dependent interactions between the bacteria and the hostile environment are necessary. WethankKarinDahmen,NadavShnerb,andDavidNelsonformanyusefuldiscussions. WethankJeremyNewburg- Rinnforhelpinacquiringdata,andAnnaDelpratoandNancyKingforhelpinguswithtechnicalaspectsofculturing Bacillus subtilis. 3 † Corresponding author. [1] J. D. Murray,Mathematical Biology (Springer-Verlag, New York,1993). [2] M. Matsushita and H. Fujikawa, Physica A 168, 498 (1990). [3] E. Ben-Jacobs, et al.,Nature 368, 46 (1994). [4] E. O. Budreneand H. C. Berg, Nature 376, 49 (1995). [5] M. P. Brenner, L. S.Levitov, and E. O. Budrene, Biophys. J. 74, 1677 (1998). [6] Y.Kozlovsky,et al.,Phys.Rev. E 59, 7025 (1999). [7] A.M. Lacasta, et al.,Phys. Rev.E 59, 7036 (1999). [8] N.M.ShnerbandD.R.Nelson,Phys.Rev.Lett.80,5172(1998); D.R.NelsonandN.M.Shnerb,Phys.Rev.E58,1383 (1998). [9] K.Dahmen, D.R. Nelson, and N.M. Shnerb,cond-mat/9807394. [10] R.A. Fisher, Ann.Engenics 7, 353 (1937). [11] J.-I. Wakita, et al.,J. Phys.Soc. Japan 63, 1205 (1994). [12] N.Munakata, Mutation Res. 82, 263 (1981). FIG.1. (a)AnimageofaBacillussubtiliscolonygrowingonnutrientrichagaratt=23.15hours(h)afterpointinoculation. TheregioninsidethedashedlineisshieldedfromUVradiation. (b)Atalatertime(t=65.45h),thecircularfrontisdistorted due to the confinement of the bacteria to the shielded region. (c) The width of the bacterial colony along the shield (x) and perpendiculartotheshield(y). Thetime-dependenceofthediameter dofacolonygrowing inapetridishwhichiscompletely shielded from UV radiation is also shown. The horizontal dotted line corresponds to the boundary of the shielded region, and the vertical dotted line corresponds to the time when UV-radiation is switched off. The temperature in all cases was 21◦C±0.5◦C. FIG.2. Imagesofthebacterialcolonywhentheshieldedregionisrotatedataconstantangularvelocityω=6.69×10−6s−1. (a) t = 24 h, (b) t = 46.56 h, (c) t = 73.73 h. The imposed convection velocity increases linearly from the axis of rotation. Thebacteriafollows theshieldedregionatlowvaluesofv,butlagbehindfarfromtheaxisofrotationcorrespondingtohigher velocities of theshield. FIG.3. (a) The colony frontsextracted from theimages shown in Fig. 2. Theaxis of rotation isat (0,0) and theinitial line of inoculation is along thehorizontal axis. The distance where thespeed of theshield corresponds to theFisher wave velocity ◦ of the bacteria around 22 is shown by the dashed circle. (b) The speed of the bacteria front vb is observed to increase to ◦ 0.23µm/s and then decrease. This maximum speed corresponds to vF 0.26µm/s around 22 C. FIG. 4. The number density n(x,v) of bacteria at various speeds v at t = 73.73 h normalized by the maximum number density nmax. The horizontal axis is normalized with the width w of the shielded region and corresponds to 0.5 to −0.5. The density of the population is observed to decay tozero for higher velocities but thedistribution is time-dependent. 4 (a) (b) 10 mm y x (c) x y 120 ) m d m ( 80 h t d i W 40 0 0 20 40 60 80 100 time (h) 25 mm ω x (a) (b) (c) y 60 (a) 50 40 ) m m 30 ( x 20 t = 24 h 10 t = 46.56 h t = 73.73 h 0 80 60 40 20 0 y (mm) 0.3 (b) v t ) s 0.2 / m µ ( b v 0.1 0.0 20 40 60 80 r (mm) Fig.3 1.0 µ 0.2 m/s µ 0.29 m/s 0.8 µ 0.37 m/s µ 0.41 m/s x ma0.6 n / ) v x,0.4 ( n 0.2 -1.0 -0.5 0.0 0.5 x/w Fig.4