NEWS FEATURE

News Feature: The cells that flock togetherPhysicists and biologists are teaming up to model the intricate dance ofcells in the body.

Stephen BattersbyScience Writer

Flesh really does crawl. Crowds of cells cansurge forward together, form gyrating swirlsand waves, segregate into types, and send offmeandering tendrils or mobile exploratoryclusters. Collective cell motions are crucial foreverything from embryo development towound healing to cancer metastasis. Thesecell colonies seem to move in concert, almostlike flocks of birds, shoals of fish, or herds ofwildebeest.Some scientists are taking note of these

parallels, hoping to find crucial clues abouthow cells coordinate. A burgeoning field ofresearch on collective cell motion is drawingon mathematics originally developed to modelanimal flocks. The goal is to find order in thecellular melee and reveal how distant cells cancoordinate their actions to move in unison.

“We want to know if we can begin to findsome general behavior in these systems,” saysCristina Marchetti of Syracuse University,New York. For all that’s known about cells,their coordinated movement largely remains amystery, one that cell-motion models mayhelp to solve.Marchetti once studied superconductivity,

but then joined the growing band of physicistswho have been drawn to the sticky problem ofcell motion in recent years. “Cell biology isbecoming more quantitative, so physicists arebecoming more interested,” she says. “Webring mathematical tools and ways of think-ing: Can we characterize this behavior in termsof a few parameters, such as density?”Several different models can already mimic

cells’ complex collective motion, and suggest

that it emerges from very simple behavior atthe level of individual cells. Some models areeven providing a few clues about the under-lying molecular mechanisms. Researchers saythat these models could eventually offerinsights into wound healing and diseaseprocesses, telling us how cancers become in-vasive and perhaps even giving us new toolsto stop them.“What I like about these approaches is that

one can begin to think about how mecha-nisms within individual cells must be coop-erating to give rise to these global tissuedynamics,” says Ryan Petrie of the NationalInstitutes of Health in Bethesda, who workson individual cell motility.

Let the Flocking BeginSome of the basic mechanics of cell motionare already well understood. Unlike free-swimming bacteria and sperm cells, whichare propelled by spinning flagella, mosteukaryotic cells crawl along by slowlyreshaping their cellular scaffolding, or cyto-skeleton. Many cells use a limb-like ap-pendage called a lamellipodium, supportedby intracellular protein filaments, whichgradually hauls the cell forward as its struc-ture changes.Keratocytes taken from fish scales have

been seen to race along at 20 micrometersper minute. Human skin cells that move inthis way are also surprisingly spry, rapidlyfilling wounds to prevent infection. Thesebiological observations reveal how cells moveindividually, but they do not explain whycells should sometimes decide to move inunison. That’s where the models come in.The modern understanding of animal

flocks stems from work led by Tamás Vicsekat the University of Budapest in Hungary.“We were seeing similar patterns of motionsin bacteria, fish, and birds,” he says. “I re-alized there must be one simple rule for allof them.” In 1995, Vicsek and his colleaguespublished a model (1) based on the idea thateach creature—dubbed “particles” in theirmodels—simply adopts the average speedand direction of its near neighbors, alongwith some random error that reflects the im-precision of most natural systems.

Starlings famously flock together in huge numbers. Such movements in birds inspiredmodels that attempt to mimic collective cellular motion. Image courtesy of Shutterstock/Bildagentur Zoonar.

www.pnas.org/cgi/doi/10.1073/pnas.1508834112 PNAS | June 30, 2015 | vol. 112 | no. 26 | 7883–7885

NEW

SFEATU

RE

When particles in the model are thinlyscattered, they move independently in alldirections. But, when the density rises abovea critical level, they all start to move as one. Avast flock can form and head off in one di-rection, even though an individual memberof the flock is unaware of what is happeningbeyond its near neighbors. The model showsthat collective motion can emerge withoutany long-range signaling, and it can describethe motions of groups, ranging from birds tobacteria to baboons.But Vicsek’s initial model isn’t very realistic

for eukaryotic cells. Cells are sticky and usuallyjammed together into a solid mass. So, over thepast decade, several teams have adapted Vicsek’soriginal model and modified it with cells inmind. Many of these models deal with theforces acting on each cell, including drag fromtheir surroundings, the attraction between twoadjacent cells, and the repulsion when they arepushed hard up against one another. In 2006,Vicsek collaborated on a model (2) where thecells adjust their direction to match the netforce exerted on them. These model cells act abit like cattle: they will go the way you shovethem, more or less.To test the model, Vicsek’s team took

keratocytes from goldfish scales and corralledthem within a square. The cells spontaneouslystarted to move in a coordinated wheeling

motion around the middle of the confinedspace, just as the model predicted.Xavier Trepat of Harvard University in

Cambridge, Massachusetts, and his colleagueshave observed similar herding in kidney cells(3). The researchers laid down a sheet of thesecells on a soft gel substrate studded with mi-croscopic beads, which shifted in response toany forces exerted by crawling cells. Re-searchers had once assumed that the cells atthe edge of a sheet do most or all of the work,dragging the others behind them. But Trepat’steam found that cells across the sheet were allpushing forward as a team, suggesting thatinformation was passing from cell to cell,coordinating their motion in a particulardirection.

Spreading the WordIn 2012, Herbert Levine, a biophysicist atRice University in Houston, was part of ateam that developed a model similar to Vicsek’s,but with slightly different rules for how cellsmove (4). Levine’s cells try to drag themselves inthe direction they are already going, reinforcingthe movement of the herd. This model cap-tures, for example, the finger-like protrusionsand rotating swirls in a sheet of canine kidneycells seen by Thomas Angelini, who studiescollective movement in living and inanimatesoft matter systems at the University of Floridain Gainesville (5). When the model’s cells are

more densely packed, the swirls become largerand slower, exactly what Angelini saw inthe laboratory.Levine argues that this teamwork might

arise from the way that a cell uses its lamel-lipodium. Cells must maintain these limbs bycontinuously building protein filaments, hesays. “If that stops, the front edge will re-tract.” So if a lamellipodium is not alreadyaligned with the cell’s motion, it could beallowed to wither away. “Our assumption is itwill shrink back and then re-form in a morealigned direction,” says Levine.Meanwhile, Nir Gov of the Weizmann

Institute in Rehovot, Israel, is trying to ex-plain cell flocking at an even deeper level.Last year, his team unveiled a model thatincludes the effects of molecules such ascadherins that stick cells together (6).When cells at the edge of a flock begin to

move into a wound, they stretch away fromthe cells behind and break some of theseadhesive links. That triggers the trailing cellsto extend a lamellipodium in the right di-rection so that they can catch up, Gov sug-gests. “This would eventually act to makerelative velocity as small as possible, which iswhat the original Vicsek interaction does,” hesays. This causes the cells in Gov’s model tomove forward in waves, as seen in laboratoryexperiments by his group, and by anotherteam led by Xavier Serra-Picamal at theUniversity of Barcelona, Spain (7).Models such as these, which simulate

many individual cells, have been increasinglysuccessful at capturing and predicting bi-ological reality. But they come with a heavycomputational burden. So researchers havedeveloped alternative models that treat cellsnot as particles but as a smooth continuum,like a fluid or an elastic sheet. “The advantageof this hydrodynamic approach is you canlook at arbitrarily large systems,” says JohnToner of the University of Oregon, Eugene,who initiated this approach to flocking inthe 1990s.Toner’s latest work looks at cell movement

in disordered landscapes, where obstaclesmight send particles zooming off in randomdirections. Rather than being scattered as onemight expect, the flocks can still hang to-gether, although larger groups do move moreslowly. This could be relevant to real-worldsituations, such as skin cells moving intoa wound.

Corralling CancerUnderstanding cell flocking at each of thesescales—from the microscopic to the molec-ular—may eventually reveal how to controlthem in the body. The biggest prize is usingthis approach to tackle cancer.

In this depiction of collective cell motion during embryonic zebrafish development, ecto-derm cells with fluorescently labeled nuclei migrate toward the dorsal body axis. The colorsindicate elapsed time. Letters in the upper right indicate body axes: A, anterior; D, dorsal;P, posterior; V, ventral. Image courtesy of Tamás Vicsek (Eötvös University, Budapest).

7884 | www.pnas.org/cgi/doi/10.1073/pnas.1508834112 Battersby

Brain tumors called gliomas, for example,send out tendrils of cancerous cells into sur-rounding tissue. Iain Couzin at PrincetonUniversity, New Jersey, likens these tendrilsto exploratory ant trails, and has developed amodel to describe them that involves an ant-like differentiation of roles between cells: afew modified cells lead the way, drawing theothers along in their wake (8). Couzin sug-gests that such marching columns may bemore effective than individual colonizers atpunching through healthy tissue, and betterat defending themselves from the immunesystem. If researchers can find out what isguiding them, it might be possible to usedrugs to control how these tumors spread,perhaps confining them or herding the cellsinto areas where they can be treated withoutdamaging vital tissue. In the nearer term, saysCouzin, understanding cell swarms mighthelp clinicians to predict the spread of diffusivebrain tumors, to guide surgery or radiotherapy.Elsewhere, there is increasing evidence that

tumors use collective motion to metastasize.The traditional picture of metastasis is thatindividual cells become motile and establishnew colonies, but it may be that mobileclumps of cells are more dangerous. In 2013, agroup including Dan Haber of the HarvardMedical School in Boston, saw small clustersof up to about 50 breast cancer cells circulatingin the blood (9). Although these clusters arerare, the researchers found that they are amajor factor in the metastatic spread of cancer.Figuring out how stationary cells free

themselves from a primary tumor and form

a mobile cell cluster might allow cliniciansto identify a tumor that is becoming in-vasive, and thus aid cancer prognosis. Amore distant dream involves using drugs toprevent the formation of mobile clusters inthe first place.There is already evidence that mobile

clusters form when cells stick more stronglyto each other than to the surrounding tissuematrix, says cell biologist Rick Horwitz, atthe Allen Institute for Cell Science in Seattle.Indeed, in Haber’s breast cancer study, a lackof plakoglobin, which helps to bind cells to-gether, suppressed metastasis.The emergence of these clusters is a rather

different problem than migrating cells on aflat substrate. Levine has been trying to createa model that describes the chemical or me-chanical stimuli that might free a cluster ofcells. “Our minimalist model does identifycell–cell adhesion as a key parameter gov-erning the transition between collectiveand individual motion,” he says. “What hasproven difficult is getting a simple model thatcan give quantitative answers for a variety of

situations: we want our models to describeboth static situations and motile flocks andclusters versus single cells.” Levine says heand his group are working on a paper thatcould present a solution.To home in on the origins of cell flocking,

researchers need more quantitative compar-isons between experiments and models. Thatrequires precise measurements of the motionand forces in real cell flocks and testingmodels in many different situations, not juston simple, flat substrates. “A promisingdirection involves experiments on patternedsubstrates that allow us to confine and directcell motion,” says Marchetti. The modelswill likely also be tested against observa-tions of 3D cell motion in extracellularmatrix cultures.Cell wrangling turns out to be a tough job.

But one day, researchers, guided by clevermodels inspired by nature, might find a wayto corral an invading tumor swarm or spurtardy skin-cell herds to move faster into thebreach.

1 Vicsek T, Czirók A, Ben-Jacob E, Cohen I, Shochet O (1995) Noveltype of phase transition in a system of self-driven particles. Phys RevLett 75(6):1226–1229.2 Szabó B, et al. (2006) Phase transition in the collective migration oftissue cells: Experiment and model. Phys Rev E Stat Nonlin SoftMatter Phys 74(6 Pt 1):061908.3 Trepat X, et al. (2009) Physical forces during collective cellmigration. Nat Phys 5(6):426–430.4 Basan M, Elgeti J, Hannezo E, Rappel W-J, Levine H (2013)Alignment of cellular motility forces with tissue flow as a mechanismfor efficient wound healing. Proc Natl Acad Sci USA 110(7):2452–2459.

5 Angelini TE, Hannezo E, Trepat X, Fredberg JJ, Weitz DA (2010)Cell migration driven by cooperative substrate deformation patterns.Phys Rev Lett 104(16):168104.6 Zaritsky A, et al. (2014) Propagating waves of directionality andcoordination orchestrate collective cell migration. Plos Comput Biol10(7):e1003747.7 Serra-Picamal X, et al. (2012) Mechanical waves during tissueexpansion. Nat Phys 8(8):628–634.8 Deisboeck TS, Couzin ID (2009) Collective behavior in cancer cellpopulations. BioEssays 31(2):190–197.9 Aceto N, et al. (2014) Circulating tumor cell clusters are oligoclonalprecursors of breast cancer metastasis. Cell 158(5):1110–1122.

Battersby PNAS | June 30, 2015 | vol. 112 | no. 26 | 7885

NEW

SFEATU

RE