CO 2 -driven diffusiophoresis for removal of bacteria
Suin Shim, Sepideh Khodaparast, Ching-Yao Lai, Jing Yan, Jesse T. Ault, Bhargav Rallabandi, Orest Shardt, Howard A. Stone
CCO -driven diffusiophoresis for removal of bacteria Suin Shim , ∗ Sepideh Khodaparast , Ching-Yao Lai , Jing Yan , JesseT. Ault , Bhargav Rallabandi , Orest Shardt , and Howard A. Stone , † Department of Mechanical and Aerospace Engineering,Princeton University, Princeton, NJ 08544, USA School of Mechanical Engineering, University of Leeds, Leeds LS2 9JT, UK Lamont-Doherty Earth Observatory, Columbia University, Palisades, NY 10964, USA Department of Molecular, Cellular and Developmental Biology,Quantitative Biology Institute, Yale University, New Haven, CT, 06511, USA School of Engineering, Brown University, Providence, Rhode Island 02912, USA Department of Mechanical Engineering, University of California, Riverside, California 92521, USA Bernal Institute and School of Engineering, University of Limerick, Castletroy, Limerick, V94 T9PX, Ireland
We investigate CO -driven diffusiophoresis of colloidal particles and bacterial cells in a Hele-Shaw geometry. Combining experiments and a model, we understand the characteristic length andtime scales of CO -driven diffusiophoresis in relation to system dimensions and CO diffusivity.Directional migration of wild-type V. cholerae and a mutant lacking flagella, as well as
S. aureus and
P. aeruginosa , near a dissolving CO source shows that diffusiophoresis of bacteria is achievedindependent of cell shape and Gram stain. Long-time experiments suggest possible applications forbacterial diffusiophoresis to cleaning systems or anti-biofouling surfaces. PACS numbers:
An aqueous suspension of charged particles in con-tact with a dissolving CO source shows directional mi-gration by diffusiophoresis [1, 2]. Here we use a Hele-Shaw geometry with either a CO bubble [3–5] or aCO -pressurized chamber to investigate the transportof polystyrene particles and bacterial cells. Combiningexperiments and model calculations, we understand thecharacteristic length and time scales of CO -driven diffu-siophoresis in relation to the system dimensions and CO diffusivity. We then study migration of wild-type Vibriocholerae and a mutant lacking flagella (∆ flaA ) near a dis-solving CO source, showing that the directional motionis diffusiophoresis , not CO -driven chemotaxis . Also,we demonstrate diffusiophoresis of Staphylococcus aureus [6] and
Pseudomonas aeruginosa [7], showing that dif-fusiophoresis driven by CO dissolution occurs for bothGram-positive and Gram-negative bacteria, independentof shape. We further demonstrate that diffusiophoreticremoval of S. aureus reduces cell adhesion to a surface,and that the removal of
P. aeruginosa lasts ≥
11 hr af-ter CO is turned off. CO -driven diffusiophoresis canprevent surface contamination or infection by reducingthe population of cells approaching an interface, and themechanism can be applied to liquid cleaning systems andanti-biofouling surfaces.When an aqueous suspension of charged colloidal parti-cles is exposed to dissolving CO , positively (negatively)charged particles migrate toward (away from) the CO source by diffusiophoresis [1, 2]; the fast diffusing H + relative to HCO − from the dissolution of H CO drivesthe transport. We investigate the phenomenon using aHele-Shaw geometry (Fig. 1; a circular cell with radius b = 11 mm and height h = 500 µ m). Here, diffusiophore- sis near a CO source is documented experimentally andcalculated (Supplementary Information; SI) for two con-figurations – a dissolving CO bubble (Fig. 1(a); we callthis system HS-B) and a CO -pressurized chamber (Fig.1(b); HS-PC) – to examine both moving and fixed bound-aries.In HS-B, a CO bubble with radius a ( t ) and an initialradius a dissolves at a typical speed dadt ≈ D a ≈ O (0 . µ m/s until the gas exchange reaches steady state [1, 3–5], where D is the diffusivity of CO in water. A bubblereaches its steady state within τ = ta /D ≈ u p = Γ p ∇ ln c i scales as u p ≈ Γ p a ≈ O (0 . µ m/s,where Γ p and c i are the diffusiophoretic mobility of par-ticles and concentration of ions, respectively (SI, video1). The relative motion of particles and the interfacescreates a charge-dependent particle distribution for bothHS-B and HS-PC (Fig. 1(c-f)).Locally, in the vicinity of the interface, amine-modifiedpolystyrene (a-PS, positively charged, diameter = 1 µ m)particles accumulate and form a high particle-density re-gion, whereas polystyrene (PS, negatively charged, diam-eter = 1 µ m) particles create an exclusion zone (EZ; Fig.1(c-f)), where the particle concentration is small. Growthof the local EZ in HS-PC (SI Fig. S2) is proportional to √ t , similar to EZ formation near an ion-exchange mem-brane [8].Particle accumulation and exclusion also occur on thelength scale (cid:96) ≈ a in both systems (Videos 2,3). Inthe model we define the boundaries of macroscopic ac-cumulation and exclusion as the radial distance wherethe nondimensional particle concentration ¯ n = n/n = 1( n ( r, t ) is the particle concentration and n = n ( r, a r X i v : . [ c ond - m a t . s o f t ] S e p FIG. 1. CO -driven diffusiophoresis of colloidal particles. (a,b) Schematics of experimental setup for (a) HS-B and (b) HS-PC.See SI for details. (c,d) Charged particles near a dissolving CO bubble (HS-B). Distribution of (c) amine-modified polystyrene(a-PS) particles and (d) polystyrene (PS) particles show, respectively, local accumulation and exclusion of charged particles bydiffusiophoresis. Bright dots indicate particles. (e,f) Charged particles near the CO source in HS-PC. Distribution of (e) a-PSand (f) PS particles near the fixed CO source show local accumulation and exclusion. (g,h) Comparison between experimentalmeasurements and model calculations of the macroscopic growth of the accumulation and exclusion zones. (g) Measured andcalculated values of ¯ r (¯ n = 1) are plotted versus τ for HS-B. (h) Measured and calculated values of ˆ r (¯ n = 1) are plotted versus τ for HS-PC. (g,h) No fitting parameter is used. (c-f) Scale bars are 500 µ m. SI). The nondimensional radial positions are defined as¯ r = r/a for HS-B, and ˆ r = r − ab − a for HS-PC. Such bound-aries are determined analogously in the experiments (SI)and plotted versus τ in Fig. 1(g,h). For HS-B, theboundaries grow faster in experiments due to the ini-tial rapid generation of the bubble, which is not includedin the model. Bubble generation introduces fast inter-face growth, which enhances CO dissolution at the earlytimes and causes faster diffusiophoreis. The particle dy-namics show better agreement in HS-PC. Without theinitial growth in the measured boundaries in HS-B, weobtain similar trends of the particle dynamics betweenHS-B and HS-PC (SI).The macroscopic boundaries increase up to almost halfof the radius of the Hele-Shaw cell ( ≈ . b ) within τ = 0 . da/dt ) that affects the particle distri-bution, and this effect lasts up to τ ≈ -driven diffusiophoresis in a Hele-Shaw cellmotivated us to extend our investigations to a broaderrange of particles. Past studies have reported on theuse of diffusiphoresis to achieve migration of living cells[9, 10]. For example, the goals of particle manipulationcan be to clean a region of liquid, achieve antifoulingsurfaces, or prevent infection in biological systems. Twoprevious studies report EZ formation in bacterial suspen-sions in contact with an ion-exchange membrane (Nafion)[11, 12] and discuss possible cleaning applications.As an initial step for demonstrating and investigating diffusiophoresis of bacterial cells by CO dissolution, wechose two types of V. cholerae cells – wild-type (WT)and a mutant lacking flagella (∆ flaA ), both of which aretagged by mKO (monomeric Kusabira Orange), a brightfluorescent protein [13]. We first confirm the diffusio-phoretic contribution to the cell migration in the pres-ence of a dissolving CO source in a Hele-Shaw geometry.Then, using PIV, we measure the velocities of the bacte-rial cells that move along the ion concentration gradient. V. cholerae is Gram-negative, comma-shaped (length ≈ µ m, diameter ≈ µ m), and single flagellated.The net surface charge of V. cholerae (as well as otherbacteria) is negative [14, 15], so the cells are expectedto migrate away from a CO source by diffusiophoresis(Video 4). We prepared a bacterial solution by dilut-ing the growth suspension (see SI for Methods) to 10%M9 minimal salt solution. No nutrient is provided sono growth and division occur on the time scale of theexperiment. Using low salt concentration helps to ex-clude effects of coupled ion fluxes on the diffusiophoresisof bacteria [16]. Similar to the particle experiments, wefill the Hele-Shaw cell with bacterial suspension, and in-troduce either a CO bubble or pressurize CO in theinner chamber. Fluorescence intensities near the CO source for both HS-B and HS-PC systems are measured(Fig. S10), and the intensity change shows that the cellnumber near the dissolving CO source decreased signif-icantly over time.Particle image velocimetry (PIV) near the fixed bound-ary measures the diffusiophoretic velocity of the cells bya dissolving CO source. We plotted the velocity vectors FIG. 2. Velocity measurements for CO -driven diffusio-phoresis of V. cholerae . (a,b) PIV for
V. cholerae cells in theHS-PC experiments. Velocity vectors plotted versus position( r − a , z ). Motion of (a) wild-type and (b) ∆ flaA cells at t ≈
10 minutes. The directional migration of cells is described byaligned velocity vectors in the radially outward direction. (c)Nondimensional z -averaged velocities obtained from (a,b) andcontrol experiments without CO at τ = 0 .
15 ( ≈
10 minutes)plotted versus r − a . (d) Nondimensional z -averaged veloci-ties of PS particles, WT and ∆ flaA cells obtained at τ = 0 . r − a . versus position in Fig. 2(a,b), where the origin of the z -axis is at the bottom left corner. After the CO valveis opened at τ = 0, both strains of V. cholerae migrateradially outward (Fig. 2(a,b)). The radial alignment ofthe velocity vectors confirms that both motile and im-motile
V. cholerae cells move along the CO -generatedion concentration gradient. In Fig. 2(c), nondimensional z -averaged velocities (¯ u cell = u cell / ( D /a ); u cell is the z -average of measured velocity) of the cells at τ = 0 .
15 withand without dissolving CO are plotted. Our observationthat both motile and immotile cells exhibit directionalmigration with similar velocities shows that the motionis not a chemotactic effect. We also compare the typicalvelocity scales of the cells and the PS particles in Fig.2(d). The diffusiophoretic velocity of the bacterial cellsis smaller than that of the PS particles, and as a firstrationalization, this is due to the smaller diffusiophoreticmobility of the cells. Our comparison suggests that the V. cholerae cells have three to four times smaller mobilitycompared to the PS particles, since the diffusiophoreticvelocity scales as u p ≈ Γ p a .To highlight the generality of the phenomenon, twomore bacteria were examined – S. aureus (mKO labeled,Gram-positive, spherical and immotile) [6] and
P. aerug- inosa (mCherry labeled, Gram-negative, rod shaped andmotile) [7] for their diffusiophoretic response to dissolv-ing CO (Fig. 3). For the HS-PC system, the diffusio-phoretic velocities at τ = 0 . r − a in Fig. 3(a). We note that S. aureus is slower com-pared to the other two bacteria. Both
S. aureus and
P.aeruginosa have surface zeta potentials ζ ≈ −
30 mV [17–19], so the velocity difference is unexpected given thatelectrophoresis makes the dominant contribution to CO -driven diffusiophoresis, where Γ p is a function of ζ [20](SI). One feature of S. aureus is that surface adhesiveproteins [21] make the cells easily form clusters, whichcan contribute to the change in the diffusiophoretic ve-locity. Also, different diffusiophoretic velocities may arisefrom different shapes of the cells (Fig. 3(a)).
S. aureus isapproximately spherical with a diameter ≈ µ m, while P. aeruginosa is ≈ µ m long and ≈ . µ m in di-ameter. Assuming similar ζ for the three bacteria andthe largest aspect ratio for P. aeruginosa , we obtain anaspect-ratio dependence of the diffusiophoretic velocitiesof the cells [22]. Our results also show that the surfacezeta potential is not the only parameter for determiningthe diffusiophoretic velocity of bacterial cells.In order to move toward applications for diffusio-phoretic bacterial removal using CO , we first quantifyadhesion of S. aureus cells to surfaces under differentCO dissolution conditions. Fig. 3(b) illustrates threeconditions of PDMS substrates (b-i) without and (b-ii,iii)with CO sources. CO is introduced either by pressur-izing CO below a PDMS membrane or by saturatingPDMS with carbonated water (CW; see SI for Methods).Then the surface coverage 30 minutes after injection of a S. aureus suspension into the chamber above the PDMSsubstrate was measured by the fluorescent intensity (Fig.3(c)). We observe that the surface attachment of
S. au-reus is significantly decreased in the presence of CO , andthis is evidence that the CO -generated ion concentrationgradient removed the cells from the vicinity of the sub-strate, resulting in reduced surface contamination. Be-low, with a set of experiments with P. aeruginosa , wedemonstrate that the bacterial removal lasts ≥
11 hours.In many discussions of diffusiophoresis, the focus isoften on boosting migration of micron-sized particles.This is a clear advantage of the phenomenon, owing toΓ p (cid:29) D p ( D p ≈ − m /s is the Stokes-Einstein diffu-sivity of a micron-sized particle). However, smaller parti-cle diffusivity compared to the diffusiophoretic mobility,can also mean that, after eliminating the gradient, thetime required for particles to recover their original dis-tribution is long ( ≈ µ m particles to move 1mm by diffusion; SI). P. aeruginosa is known for surviving in dilute media[23] for more than 10 days so it is suitable for long-time diffusiophoretic experiments. We performed HS-PCexperiments with and without 1 hr of CO dissolutionin P. aeruginosa suspension (Fig. 3(d,e)). We predict
FIG. 3. CO -driven diffusiophoresis of S. aureus and
P. aeruginosa . (a) Velocity measurements of bacterial cells near the CO source (fixed boundary configuration) at τ = 0 . t ≈
20 min). Right panels: schematics showing cell shape. (b) Schematics ofadhesion experiments for
S. aureus cells on PDMS surfaces. We tested three scenarios: (b-i) plain PDMS substrate in ambientair, (b-ii) pressurized CO gas under the PDMS substrate, and (b-iii) PDMS substrate that is saturated with carbonated water(CW). (c) Intensity measurements for cell coverage on PDMS substrate at t = 30 min. The gray values are normalized by thatof a corresponding no-CO experiment. Inset images show attachment of S. aureus cells on three different PDMS surfaces.(d-f) Long time effect of diffusiophoresis: diffusiophoresis of
P. aeruginosa cells in the fixed boundary system. (d) Schematicof two control experiments in the fixed boundary configuration. In the presence of a finite-time CO source, cells move radiallyoutward and form an accumulation front (ring structure) in the chamber. On the other hand, when the CO source is replacedby an air source, cells gradually concentrate toward both PDMS walls. (e-i,ii) Images showing the Hele-Shaw chamber at t = 1hr. (f) Fluorescent intensity measurements near the CO source for two experiments in (d,e). Accumulation and exclusion ofbacterial cells near the inner PDMS wall is maintained up to 12 hours, proving long-term effect of diffusiophoresis. Scale barsare (c) 50 µ m, and (e) 5 mm. and observe that, by CO diffusiophoresis, bacterial cellsmove away from the inner wall, whereas without any CO source, the cells concentrate near both inner and outerPDMS walls where there is an air source. The CO valvewas open only for 1 hr, but the result of diffusiophoresislasted longer than 12 hours (Fig. 3(f)). The distributionof the cells at t = 12 hr are presented in the SI.Finally, we discuss the diffusiophoresis of motile bacte-ria since it is not identical to that of polystyrene particlesor immotile cells. Both V. cholerae and
P. aeruginosa are single flagellated organisms and exhibit run-reversepatterns [24]. The effective diffusivity of motile bacteriawith typical translational speed v t and reverse time t r can be estimated as D eff ≈ v t t r ≈ O (100) µ m /s (SI).It is observed (Video 5) that the flow of cells under ionconcentration gradient is a slow advection with an esti-mated P´eclet number P e = u p (cid:96) cell D eff ≈ − -10 − . Cellsare observed to swim randomly with their characteris-tic velocity ≈ µ m (SI), with a slow drift (radiallyoutward) due to the diffusiophoretic contribution (Video5).In this paper, we present proof of diffusiophoretic mi-gration of different types of bacteria under a concentra- tion gradient of CO , and discuss possible applicationsof CO -driven diffusiophoresis to prevent contamination.For example, delaying biofilm formation can improve theanti-biofouling properties of surfaces. Currently we areworking to realize the mechanism at various salt concen-trations to broaden the understanding to physiologicalor higher salinity conditions. Moreover, understandingthe characteristic scales and flow stucture near the CO source is crucial for the next steps of CO -driven diffusio-phoresis for mitigating bacterial growth on, or bacterialremoval from, surfaces. ACKNOWLEDGEMENTS
We thank the Bassler Lab for providing the
V. cholerae strains (JY019 and JY238) for the current study. S.S.thanks Minyoung Kim and Christina Kurzthaler for valu-able discussions. S.S. and H.A.S. acknowledge the NSFfor support via CBET-1702693. S.K. thanks LOral-UNESCO UK and Ireland for support via the FWIS 2019fellowship.
CONTRIBUTIONS
S.S. and H.A.S. conceived the project. S.S. designedand performed all experiments. S.K. conducted PIV.S.S., C.Y.L., J.T.A. conducted numerical calculations.J.Y. constructed the
V. cholerae strains. S.S., B.R., O.S.,and H.A.S. set up the theoretical model. All authors con-tributed to data analysis and writing the paper. ∗ [email protected] † [email protected][1] S. Shim, Interfacial Flows with Heat and Mass Transfer.(PhD Thesis, 2017, Princeton Univ.)[2] S. Shin, O. Shardt, P. B. Warren, H. A. Stone, Mem-braneless water filtration using CO . Nat. Comm. bubbles in a microfluidic chan-nel. Lab Chip , 2428-2436 (2014)[4] P. Pe˜ n as-L´ o pez, M. A. Parrales, and J. Rodriguez-Rodriguez, Dissolution of a CO spherical cap bubbleadhered to a flat surface in air-saturated water, J. FluidMech. , , 53-76 (2015)[5] P. Pe˜nas-L´opez, B. van Elburg, M. A. Parrales, and J.Rodr´ıguez-Rodr´ıguez, Diffusion of dissolved CO in wa-ter propagating from a cylindrical bubble in a horizontalHele-Shaw cell. Phys. Rev. Fluids , 063602 (2017)[6] S. Khodaparast, M. K. Kim, J. E. Slipe, and H. A.Stone, Bubble-driven detachment of bacteria from con-fined microgeometries. Environ. Sci. Technol. , 1340-1347 (2017)[7] A. Siryaporn, M. K. Kim, Y. Shen, H. A. Stone, andZ. Gitai, Colonization, competition, and dispersal ofpathogens in fluid flow networks. Curr. Biol. , 1201-1207 (2015)[8] D. Florea, S. Musa, J. M. R. Huyghe, and H. M. Wyss,Long-range repulsion of colloids driven by ion exchangeand diffusiophoresis, Proc. Natl. Acad. Sci. U. S. A. , Prog. Surf.Sci. , 210-217 (1992)[10] T. Fukuyama, A. Fuke, M. Mochizuki, K. Kamei, and Y.T. Maeda, Directing and boosting of cell migration by theentropic force gradient in polymer solution, Langmuir , , 12567-12572 (2015)[11] Y. Cheng, and C. I. Moraru, Long-range interactionskeep bacterial cells from liquid-solid interfaces: evidenceof bacteria exclusion zone near Nafion surfaces and possi- ble implications for bacterial attachment. Colloids. Surf.B Biointerfaces , , 16-24 (2018)[12] H. Lee, J. Kim, J. Yang, S. W. Seo, and S. J. Kim, Diffu-siophoretic exclusion of colloidal particles for continuouswater purification. Lab Chip , Biochem. J.
Vibrio cholerae , Infect. Immun. , 1048-1058(1983)[15] T. Vissers, A. T. Brown, N. Koumakis, A. Dawson, M.Hermes, J. Schwarz-Linek, A. B. Schofield, J. M. French,V. Koutsos, J. Arlt, V. A. Martinez, and W. C. K. Poon,Bacteria as living patchy colloids: phenotype heterogene-ity in surface adhesion, Sci. Adv. :eaao1170 (2018)[16] A. Gupta, S. Shim, L. Issah, C. McKenzie, and H.A. Stone, Diffusion of multiple electrolytes cannot betreated independently: Model predictions with experi-mental validation. Soft Matter , , 9965-9973. (2019)[17] S. Halder, K. K. Yadav, R. Sarkar, S. Mukherjee, P.Saha, S. Haldar, S. Karmakar, and T. Sen, Alterationof zeta potential and membrane permeability in bacte-ria: a study with cationic agents. Springer Plus :672(2015)[18] L.-A. B. Rawlinson, J. P. O’Gara, D. S. Jones, andD. J. Brayden, Resistance of Staphylococcus aereus tothe cationic antimicrobial agent poly(2-(dimethylaminoethyl)methacrylate) (pDMAEMA) is influenced by cell-surface charge and hydrophobicity.
J. Med. Microbiol. , , 968-976 (2011)[19] J. Shephard, A. J. McQuillan, and P. J. Brenner,Mechanisms of cation exchange by Pseudomonas aerug-inosa
PAO1 and PAO1 wbpL , a strain with a truncatedlipopolysaccharide.
Appl. Environ. Microbiol. , 6980-6986 (2008)[20] D. C. Prieve, J. L. Anderson, J. P. Ebel, M. E. Low-ell, Motion of a particle generated by chemical gradients.Part 2. Electrolytes. J. Fluid Mech.
Staphylococcus aureus . Trends Microbiol. , 485-488(1998)[22] S. W. Joo, S. Y. Lee, J. Liu, and S. Qian, Diffusiophoresisof an elongated cylindrical nanoparticle along the axis ofa nanopore, ChemPhysChem , , 3281-3290 (2010)[23] K. D. Mena, and C. P. Gerba, Risk assessment of pseu-domonas aeruginosa in water. Rev. Environ. Contam.Toxicol. , 71-115 (2009).[24] M. Shigematsu, Y. Meno, H. Misumi, and K. Amako, Themeasurement of swimming velocity of
Vibrio cholerae and
Pseudomonas aeruginosa using the video trackingmethod,
Microbiol. Immunol.39