Measuring dynamic cell–material interactions andremodeling during 3D human mesenchymal stem cellmigration in hydrogelsKelly M. Schultza,1, Kyle A. Kyburzb, and Kristi S. Ansethb,1

aDepartment of Chemical and Biomolecular Engineering, Lehigh University, Bethlehem, PA 18015; and bDepartment of Chemical and BiologicalEngineering, BioFrontiers Institute and Howard Hughes Medical Institute, University of Colorado at Boulder, Boulder, CO 80309

Contributed by Kristi S. Anseth, June 9, 2015 (sent for review January 14, 2015; reviewed by Alexander R. Dunn, Gerald G. Fuller, and Kristi L. Kiick)

Biomaterials that mimic aspects of the extracellular matrix by present-ing a 3D microenvironment that cells can locally degrade and remodelare finding increased applications as wound-healing matrices, tissueengineering scaffolds, and even substrates for stem cell expansion. Invivo, cells do not simply reside in a static microenvironment, butinstead, they dynamically reengineer their surroundings. For example,cells secrete proteases that degrade extracellular components, attachto the matrix through adhesive sites, and can exert traction forces onthe local matrix, causing its spatial reorganization. Although bio-materials scaffolds provide initially well-defined microenvironmentsfor 3D culture of cells, less is known about the changes that occur overtime, especially local matrix remodeling that can play an integral rolein directing cell behavior. Here, we usemicrorheology as a quantitativetool to characterize dynamic cellular remodeling of peptide-function-alized poly(ethylene glycol) (PEG) hydrogels that degrade in responseto cell-secreted matrix metalloproteinases (MMPs). This techniqueallows measurement of spatial changes in material properties duringmigration of encapsulated cells and has a sensitivity that identifiesregions where cells simply adhere to the matrix, as well as the extentof local cell remodeling of the material through MMP-mediateddegradation. Collectively, these microrheological measurements pro-vide insight into microscopic, cellular manipulation of the pericellularregion that gives rise to macroscopic tracks created in scaffolds bymigrating cells. This quantitative and predictable information shouldbenefit the design of improved biomaterial scaffolds for medicallyrelevant applications.

PEG–peptide hydrogels | human mesenchymal stem cells | cell migration |microrheology

Synthetic hydrogel scaffolds have been designed to serve asmimics of the native extracellular matrix (ECM) with the goal of

promoting desired cell functions (e.g., proliferation, migration, dif-ferentiation), especially for applications in wound healing (1), tissueregeneration (2), and stem cell culture (3, 4). For example, poly(ethylene glycol) (PEG) hydrogels can serve as blank slates in whichpeptide cues can be systematically introduced in the scaffold toallow integrin binding (5, 6), proteolytic degradation (7, 8), andeven local sequestering of growth factors (9). Furthermore, it is wellknown that cells respond to mechanical stimuli (e.g., stiffness) intheir local microenvironment, the so-called pericellular region, andtuning of a scaffold’s mechanical properties can influence how acell degrades and remodels its surroundings (10–12). The complexcell–matrix interactions that occur in the native ECM are oftenmimicked in peptide-functionalized hydrogels through the incor-poration of adhesive binding peptides (e.g., RGDS, IKVAV) and en-zymatically degradable peptide cross-linkers (e.g., GPQGIWGQ,GPLGLWAR), both of which are necessary for cell attachment,spreading (13), and motility (12, 14). However, changes in the localmaterial properties as a result of this cell-mediated remodeling havelargely remained a “black box,” limiting interpretation of data andconfounding the design of more advanced biomaterials.Macroscopically, cells degrade micrometer-sized channels into

scaffolds as they move, an event that begins with microscopic

remodeling of their pericellular region and eventually permanentlyreengineering the scaffold architecture and material properties on alarger scale. If one seeks to design synthetic ECM environments todirect cellular processes, such as migration, it is important to betterunderstand how these inputs are dynamically altered on the locallength scale. Such information can help advance biomaterial design,especially for applications focused on the delivery or recruitmentof cells, where directing cell–material interactions and migrationcan be critically important. At present, cell matrices are generallyengineered to have certain initial material properties, but theresulting cell motility and cell–material interactions are often onlyempirically correlated with these design parameters (7, 15). Toovercome this obstacle and provide an in situ measurement ofscaffold degradation, microrheological measurements have beenused to fingerprint and understand changes in material propertiesin the pericellular region during cell motility.Although real-time measurements of material properties near a

cell are difficult, investigations have focused on developing tech-niques to access this information. In two dimensions, forces thatcells exert when seeded on hydrogel surfaces have been measuredusing deflection of beds of microneedles (15) and deformation ofgel surfaces (16). For example, Tan et al. (15) developed a mea-surement technique that exploits independent deflection of micro-needles of varying lengths (and therefore stiffnesses) to measure thedistribution of subcellular traction forces of both smooth muscle

Significance

Scaffolds that serve as synthetic mimics of the extracellular matrixhave applications in wound healing, tissue engineering, and stemcell expansion. When cells are cultured in these tunable matrices,little is known about local microenvironmental changes duringdegradation and remodeling. Methods that provide quantitativeand predictable information about cell-mediated remodelingcould significantly improve the biomaterial design process. Weuse passive microrheology, a technique that measures rheologicalproperties from Brownian motion of embedded particles, tocharacterize remodeling of a cell-laden peptide-functionalizedpoly(ethylene glycol) hydrogel that degrades in response to cell-secreted enzymes. Results show microenvironmental changes atmultiple time and size scales, and reveal an interesting degrada-tion gradient, as mesenchymal stem cells attach, spread, andmove through these synthetic extracellular matrix mimics.

Author contributions: K.M.S., K.A.K., and K.S.A. designed research; K.M.S. and K.A.K.performed research; K.M.S. and K.A.K. contributed new reagents/analytic tools; K.M.S.and K.A.K. analyzed data; and K.M.S., K.A.K., and K.S.A. wrote the paper.

Reviewers: A.R.D., Stanford University; G.G.F., Stanford University; and K.L.K., Universityof Delaware.

The authors declare no conflict of interest.1To whom correspondence may be addressed. Email: Kristi.Anseth@colorado.edu orkes513@lehigh.com.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1511304112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1511304112 PNAS | Published online July 6, 2015 | E3757–E3764

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

2, 2

020

cells and fibroblasts. The main conclusion was that cellularspreading and morphology control the magnitude of the tractionforces (15). The traction force of confluent cell sheets interactingwith a gel surface was also analyzed, toward understanding howcellular processes are coordinated over large length scales. Usingendothelial, epithelial, and breast cancer cell sheets, results showedthat collective migration was due to a transmittance of normal stressacross cell–cell junctions with migration orientated in the directionof the minimal intercellular shear stress (16).Cell-mediated degradation of the local microenvironment plays

a critical role in permitting cellular migration and invasion in vivo.These processes are important during development, wound re-generation, and pathophysiological states facilitated by proteolyticevents via cellular protease secretion. Previous work has begun toelucidate the length scales and spatial effects of secreted proteasesin relation to migrating tumor cells during collagen matrix re-modeling (17–19). For example, Packard et al. (20) used matrixmetalloproteinase (MMP)-sensitive biosensors to visualize pro-tease activity in the pericellular region of migrating tumor cells incollagen, finding increased activity at the polarized leading edge.These seminal works have elucidated the spatial presence andlocal activity of proteases in relation to individual migrating cells.However, how migrating cells temporally degrade and remodelthe local microenvironment on larger length scales remains rel-atively unknown.Although 2D studies add to our understanding of cell–matrix

interactions, 2D environments can unnaturally polarize cells, andsome aspects of cell motility can be quite different in 2D versus3D environments (21, 22). For these reasons, recent devel-opments have focused on strategies to measure cell–materialinteractions in three dimensions (e.g., cell-laden hydrogels).Traction force microscopy measures spatial interfacial forces byquantifying the elastic deformation of a substrate (21). If themodulus of the material is known, this technique quantifies theforces cells exert in three dimensions calculated from embeddedbead displacement. This approach has identified patterns offorces generated around distinct morphological regions duringcellular invasion into a scaffold (21). Additionally, Bloom et al.(23) investigated the degradation of a collagen scaffold duringthe migration of a fibrosarcoma cell line (HT1080s) using em-bedded particle displacements. The authors showed that thehydrogel was reversibly deformed at the cell’s leading edge, butirreversibly remodeled at the trailing edge. Collectively, thesepioneering investigations have provided insight into aspects ofthe complex interplay between cells and scaffold materials;however, complementary techniques that allow characterizationof dynamic and local changes in mechanical properties, degra-dation, and scaffold erosion would be beneficial in further ad-vancing our understanding of mechanotransduction, mechanismsof cell motility, and even biomaterials design.In this contribution, multiple particle tracking microrheology

(MPT) is used to measure how human mesenchymal stem cells(hMSCs) remodel peptide cross-linked PEG hydrogels as theymigrate. hMSC migration is characterized by significant remod-eling of the local environment through attachment, enzymaticdegradation, and cellular traction. Furthermore, hMSCs areobserved to degrade the synthetic network through two path-ways, MMP secretion that cleaves the peptide cross-linker andmyosin II-regulated adhesion and reversible remodeling of thenetwork. We find that MPT has the sensitivity to capture thetemporal transition of the hydrogel from an elastic gel to a vis-cous liquid, during hMSC-mediated degradation. MPT simulta-neously provides information about the spatial region, proximalto the cell, over which this matrix remodeling occurs. Thetechnique and measurements enhance our understanding ofcell–material interactions in three dimensions and enable visu-alization of dynamic cell-mediated matrix degradation, the so-called fourth dimension. On longer timescales, these microscopic

changes give rise to the creation of macroscopic channels in thehydrogel that are important for hMSC motility. We believe thatthis approach and characterization can provide an important linkfor better understanding outside-in signaling experienced by cellswhen embedded in 3D environments.

Results and DiscussionMicrorheological Characterization of Hydrogel Degradation andRemodeling During hMSC Migration. To characterize hMSCremodeling of their local environment when embedded in MMP-degradable hydrogels, we use MPT. The sensitivity of MPT in thelow moduli range (10−3 to 4 Pa) of hydrogels enables measure-ments of transitions from loosely cross-linked networks to vis-coelastic polymeric fluids, an important critical-state transitionthat correlates to many macroscopic changes in cell behavior(e.g., morphology, motility, secretory properties). The hydrogelscaffold before cell-mediated degradation is similar to the me-chanical environment presented in many soft tissues, such asneural and adipose tissue (11, 24, 25). Here, we aimed to morefully understand 3D hMSC motility quantitatively by character-izing the remodeling and degradation of the scaffold, whichtypically results in spatial variations in the local matrix me-chanics. Dynamic spatiotemporal rheological measurements ofthe rheological evolution during cell motility will bridge a miss-ing link to identify the outside-in signaling a cell experiencesduring migration and can be exploited in advanced materialdesign to promote tissue regeneration. MPT measurements usevideo microscopy for data collection, capturing spatial in-formation about changes in hydrogel properties with time. Thistechnique enables one to resolve both temporal and spatial in-formation about the cellular microenvironment, and here, wereport on the characterization of the remodeling and degrada-tion of MMP-degradable PEG–peptide hydrogels by 3D encap-sulated hMSCs.Microrheology measures the Brownian motion of embedded

probe particles (carboxylated polystyrene probes with radius a,2a≅ 1 μm) and relates this motion to rheological properties, suchas viscosity and creep compliance, using the generalized Stoke–Einstein relation (GSER):

�Δr2ðtÞ�= kBT

πaJðtÞ. [1]

Here, hΔr2ðtÞi is the mean-squared displacement (MSD), t istime, kBT is the thermal energy, JðtÞ is the creep compliance,and a is the probe particle radius (26–29). The state of material,i.e., sol or gel, can be determined using the logarithmic slope ofthe mean-squared displacement, α= d log  hΔr2ðτÞi=d log τ, andthe critical relaxation exponent, n; both are determined at theshortest lag times measured with MPT. The shortest lag timesare the largest frequencies measured and capture the longestrelaxation times of the material. Probe particles freely diffusingin a liquid have a value of α= 1. When α→ 0, probe particles arecompletely arrested in the gel network. All values of α between0 and 1 are an elastic solid or viscoelastic liquid, and this tran-sition is defined by the critical relaxation exponent, n. The valueof n has been previously determined from measurements of thekinetics of degradation analyzed using time-cure superposition(30–33). To determine the state of a material, α is compared withn; if α> n, the material is a viscoelastic fluid, and if α< n, thematerial is an elastic solid (30). The value of n for the hydrogelstudied here is n≈ 0.2 (30). MPT relies on the thermal motion ofparticles to measure material properties resulting in a low uppermeasurable moduli limit, ≈ 4 Pa (34). At this modulus, probe par-ticles are completely arrested in the gel material and bulk rheologyis used to measure the continued evolution of the gelling material.In general, PEG hydrogel scaffolds have been designed with a

complexity that allows physical and chemical cues to be locally

E3758 | www.pnas.org/cgi/doi/10.1073/pnas.1511304112 Schultz et al.

Dow

nloa

ded

by g

uest

on

June

2, 2

020

presented to encapsulated cells and elicit desired functions(10–12). The hydrogel used in this work was formed via a photo-initiated thiol–ene polymerization of norbornene functional-ized PEGs cross-linked with thiol-containing peptides (Fig. 1A)(35, 36). Specifically, a four-arm star PEG (Mn, 20,000) was endfunctionalized with norbornene; PEG was chosen due to its hy-drophilicity, resistance to nonspecific protein adsorption, and theability to tailor cell–material interactions by conjugation of se-lected peptide sequences (36–38). Here, the PEG was cross-linked with a MMP-degradable peptide, KCGPQG↓ IWGQCK,which is cleaved at a high rate by MMP-1, -2, -3, -8, and 9 on atimescale relevant for 3D cell culture (8, 12, 39). Hydrogelsmeasured in the absence of cells and incubated in an identicalmanner to cell-laden hydrogels showed no significant scaffolddegradation, indicating that this peptide cross-linker is beingdegraded by cell-secreted enzymes. A schematic of the hydrogelformation and the specific formulation used for cell encapsula-tion experiments is shown in Fig. 1A. Included in the hydrogelscaffold is an adhesive ligand (1 mM CRGDS) that promoteshMSC adhesion and motility in the otherwise bioinert PEGscaffold. Upon irradiation with UV light (3 min, 365-nm light at

10 mW·cm−2) in the presence of a photoinitiator (lithium phenyl-2,4,6-trimethylbenzoylphosphinate), the norbornene and thiolfunctionalities undergo alternating propagation and chain transferreactions via a step growth mechanism to yield a covalently cross-linked network (35, 36). Previous work has shown the usefulnessof this material chemistry for culture of cells in two and threedimensions, rendering it a suitable material for studying howdegradation-induced matrix changes influence cell motility in fourdimensions (i.e., 3D space and time) (12, 14).All cell-monitoring experiments were performed using a

hydrogel with a low cross-linking density and modulus (0.18 ±0.02 mM and G′= 110± 10  Pa, respectively), which was achievedby controlling the stoichiometric ratio of the thiol-to-ene func-tional groups as 0.65 (14). The mechanical properties of thechosen scaffold are similar to those previously used in cell cul-ture and regenerative medicine applications, such as collagengels (c≤ 2 mg/mL) and PEG-based scaffolds for investigations ofcell motility (14, 40). Although in these types of investigations arange of material properties are usually used, 0.1–100 kPa, forthis work we focus on a material that is loosely cross-linked toenable facile cell-mediated degradation and motility (11, 25).hMSCs encapsulated at a density of 2× 105 cells per mL in thisgel formulation survive, spread, and migrate. Fig. 1 B–D showsexamples of real-time cell-tracking experiments, where hMSCmigration was followed for a period of 6 h, Fig. 1B. Cell trackingwas used to calculate speed, persistence, and percentage of cellsmigrating, which were 18 ± 1 μm·h−1, 156 ± 23 min, and 59 ±12%, respectively (14). During 3D migration, hMSCs were ob-served to move over large distances in the matrix (10–150 μmover 6 h), which implies significant cell-mediated network deg-radation and remodeling. In particular, the initial mesh size ofthe hydrogel is orders of magnitude smaller (tens of nanometers)than the size of a cell (tens of micrometers). As further evidenceof cell–matrix interactions, immunostaining was performed (Fig.1C). Punctate β1-integrin staining (green) at the ends of actinstress fibers (red) was observed, indicating strong cellular at-tachment and spreading of the hMSCs in these PEG microen-vironments. Additionally, hMSC migration was inhibited by theaddition of either an MMP inhibitor, which limits cell-mediateddegradation of the matrix, or with blebbistatin, which inhibitscytoskeletal tension. In both cases, cell migration was signifi-cantly reduced compared with untreated hMSCs (Fig. 1D).These controls further emphasize the importance of cell-medi-ated degradation and cellular tension on microenvironmentalremodeling and subsequently regulating the hMSC migrationprocess within these materials.

Cell-Mediated Degradation During Early Stages of Motility. UsingMPT to characterize hMSC-laden gels, we measured the changesin material properties, temporally and spatially, in the peri-cellular region of individual cells as they attach, spread, andbegin to migrate in MMP-degradable hydrogels. MPT dataduring cellularly dictated changes in the hydrogel scaffold werecollected over 25–60 min, while hMSCs were migrating. Initiallyand immediately after encapsulation, hMSCs were essentiallyembedded in a largely elastic solid gel environment, and repre-sentative, surrounding probe particle trajectories are depictedin Fig. 2A. However, tens of minutes after encapsulation, thehMSCs began to spread, remodel, and degrade their localhydrogel microenvironment; in this region, the material begins totransition from an elastic solid to a viscoelastic fluid, and theprobe particle trajectories become significantly longer (Fig. 2B).At much longer timescales, the hMSC morphology is highly ex-tended and relatively fast migration is observed, presumablythrough regions that are largely in a polymeric liquid state, asobserved by probe particle trajectories illustrated in Fig. 2C.The logarithmic slope of the MSD, α, was calculated for each

of these time points and collected in the pericellular region.

Fig. 1. Human mesenchymal stem cells (hMSCs) migrate and form focal ad-hesions within the MMP-degradable PEG-peptide hydrogel. (A) Schematic ofnetwork formation for the MMP-degradable PEG-norbornene hydrogel scaffold(0.65 thiol:ene; 3 mM 4-arm PEG-norbornene; Mn, 20,000 g·mol−1; f = 4; 3.9 mMKCGPQG↓ IWGQCK; Mn, 1,305 g·mol−1; f = 2; 1 mM CRGDS). (B) hMSCs wereencapsulated at a density of 2×105 cells per mL, and the motility of individualcells was followed in real time for a period of 6 h. (Scale bar: 100 μm; phasecontrast image.) (C) Representative image of an encapsulated hMSC immunos-tained for actin (red), β1-integrin (green), and DAPI (blue). (Scale bar: 20 μm.)Forty-eight hours after encapsulation, hMSCs spread within the gel and formactin stress fibers and punctate β1-integrin staining as observed at the end ofthese fibers shown in the Inset. (Inset scale bar: 10 μm.) (D) The migratory speedof hMSCs decreased when cells were treated with either an MMP inhibitor(InSolution GM 6001; 10 μM; immediately after encapsulation) or blebbistatin(50 μM; 2 h postencapsulation). *P< 0.05.

Schultz et al. PNAS | Published online July 6, 2015 | E3759

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

2, 2

020

Measurements were repeated over 4–10 cells, and each dataseton the graph (Fig. 2D) represents measurements around an in-dividual cell. Clearly, degradation on the microscopic lengthscale has macroscopic implications related to cell-mediatedscaffold architectural reengineering that facilitates motility. ForhMSCs migrating through the MMP-degradable gels, we mea-sured cells at two different time points. Initially in the migrationprocess, we calculated that α≈ 0, where the cell is encapsulated inthe gel network before any remodeling. Over the next 15–30 min,hMSCs begin to attach and spread in the network, a process thatnecessitates some degree of local scaffold degradation. MPTmeasurements capture this remodeling, as an increase in α overn= 0.2, the value where the gel–sol transition occurs. In general,this parameter corresponds to a decrease in network connectivityand the transition of the material from a gel, a sample spanningcross-linked network, to a sol. Once cell-mediated degradationis complete (i.e., the gel to sol transition), rapid migration isobserved as detailed below.Optical fluorescent video microscopy was used to capture

MPT data and enabled characterization of spatial changes in thematerial properties during hMSC migration. With these mea-surements, we aimed to identify regions where a cell adheres tothe network during MMP secretion and matrix degradation, aswell as characterize the distances over which this hMSC matrixremodeling occurs. As an example, Fig. 3 A–C maps the materialproperties surrounding an hMSC embedded in a gel and mea-sures degradation of the environment through time. The color ofeach ring is the logarithmic slope of the MSD, α, calculated from

the motion of particles within that area. Warm colors indicatethat α→ 0, and the region is basically a solid gel state; light bluerepresents where α= 1 and is indicative of Brownian diffusion;cooler colors are α≈ 1.6, indicating ballistic or directed motion.Concentric rings on the map are centered around the originalcell position, which was defined at the beginning of the video.Here, the central ring has a radius of r1 =150 pixels ≈ 37 μm,with each ring having a radius 150 pixels larger than the previousone, i.e., r2 = 150+ r1. In this scheme, the central circle repre-sents a value of α 150 pixels from the center of the cell area, andthe next circle represents a value of α of particles 150–300 pixels(37–74 μm) away from the cell. Each ring represents the move-ment of particles that are uniquely identified within the specifiedarea from the initial particle position.

Fig. 2. hMSC remodeling and degradation of peptide cross-linked PEGhydrogels. Schematic images illustrating representative visual changes thatoccur in cell-laden hydrogels, where the changes in the gel properties arecaptured by MPT probe particle trajectories. (A) Initial state, before hMSCshave caused any substantial changes in the local material properties and thecells experience a solid gel environment; (B) during cell spreading, the localenvironment degrades in response to cellular activity, and the materialbegins to transition from a gel to a sol in a local region; and (C) at longertimescales, the pericellular region is extensively degraded, becoming asol, and cell motility is observed. (D) Logarithmic slope of the MSD(αj0.1<τ<1  s =d   loghΔr2ðτÞi=d   logτj0.1<τ<1  s) of probe particles in the pericellularregion during hMSC migration. The gel–sol transition occurs at the criticalrelaxation exponent, n= 0.2. Values of α> 0.2 represent materials that are asol and α< 0.2 are gelled materials. Data are highlighted for two stages ofcell motility that will be described in detail in Figs. 3 and 4. Fig. 3 A–C showsa cell that is spreading and starting to degrade the pericellular region, andFig. 4 A–C is a cell that is very motile in a sol.

Fig. 3. Dynamic rheological changes in the pericellular region during mi-gration of an encapsulated hMSC over time. Data are taken at (A) 0, (B) 9,and (C) 27 min after the cell is identified. Particle image velocimetry (PIV)measurements quantify the long-time movement of probe particles be-tween (D) 0 and 4, (E) 9 and 14, and (F) 23 and 27 min. Every other particletrajectory is displayed on PIV plots for clarity. Bright-field images are set inthe background of MPT measurements with MSD values calculated spatiallyas the distance away from the cell. The z axis, indicated by color, is thelogarithmic slope of the MSD, α, where a slope of 0 (red) indicates no particlemovement, a slope of 1 (light blue) indicates Brownian motion, and a slopeof 1.6 (purple) indicates ballistic motion. PIV measurements show the dis-placements using color and size of arrows. Warm colors and small arrowsindicate small particle displacement, whereas cool colors and large arrowsshow large particle displacement. Both of these measurements confirm that,through time as the cell is spreading, the largest extent of degradation oc-curs furthest from the cell.

E3760 | www.pnas.org/cgi/doi/10.1073/pnas.1511304112 Schultz et al.

Dow

nloa

ded

by g

uest

on

June

2, 2

020

Fig. 3 A–C shows the changes in material properties over27 min, during migration of an hMSC that is beginning to spreadat the early stages of data collection (these data are highlightedin Fig. 2D with closed symbols). Throughout this time period, thearea closest to the cell remains a gel until the final time point,indicating that the cell is likely adhering to this region of thescaffold during MMP secretion. In Fig. 3B, we measured thetransition from a gel to a sol, as annotated by the orange color,and here, the area directly around the cell appears to remain agel. As we move through time (Fig. 3C), we observe that theextent of degradation of the scaffold is highest in regions furthestfrom the cell. In fact, a gradient in degradation is measured withviscoelastic fluid properties observed furthest from the cell, buttransitioning to an elastic solid as one approaches the cell. In Fig.3C, the outer circle indicates that the region has returned to a gelstate, and we believe that this may be due in part to swelling ofpartially degraded gel adjacent to the cellular eroded features.When this occurs, the dimensional changes in the gel will beginto fill the voids created by the cell, resulting in rheologicalmeasurements that indicate that the material has returned to agelled state at the edge of the field of view.The corresponding images in Fig. 3 D–F are particle image

velocimetry (PIV) measurements of particle movements overlong timescales (Δt = 4–5 min) where displacement of the par-ticles was measured between two bright-field images separatedby several minutes. Warm colors indicate small particle dis-placements, whereas cool colors correlate to larger displace-ments. Lack of arrows in the PIV map indicate that there isno detectable displacement. In these PIV maps, we quantifiedparticle displacements that agree with our microrheologicalmeasurements and reveal displacements primarily due to celltraction. MPT data are collected over a 30-s acquisition window.At these short times, we do not measure drift in particle move-ment, enabling the characterization of rheological properties.Over longer times, captured by PIV, directed motion of particledisplacement is measured due to cytoskeletal tension on thenetwork. In Fig. 3D, few particle displacements were detected.Over the long time intervals used for PIV, Δt = 4–5 min, wemeasured the largest particle displacement furthest from the cellduring spreading. This movement decreased, as characterized inregions closest to the center of the cell. On this timescale, webelieve that the particle movement is due to cytoskeletal tensionin regions of the scaffold that are degraded. The detected dis-placement shows that particles are moving in a persistent di-rection over this interval, which implies cellular remodeling andtraction on the remaining porous scaffold. Furthest from the cell,MPT measurements describe an elastic fluid, and PIV mea-surements agree with MPT but also imply that there is a scaffoldstructure on length scales greater than accessed with MPT. Ad-ditionally, PIV measurements reveal that movement of particlesat these long timescales is equivalent to cell speed. This quan-

titative correlation in speed and direction confirms that PIVmeasurements on long timescales are due to cellular traction andremodeling during motility. Together, MPT and PIV analysesillustrate the complex interplay between cellular remodeling dueto both cell–matrix interactions leading to traction, as well ascell-secreted enzymatic degradation in the local pericellular re-gion. Both processes play a vital role in understanding dynamicchanges in cell-laden hydrogel environments and its effects oncell motility.The gradient in extent of degradation, where the highest extent

of degradation was observed furthest from the cell, suggests thatthe value of the Damköhler number, Da= reaction  rate=diffusivemass  transfer  rate, is small. This physically indicates that the cell-secreted enzymes diffuse away from the cell faster than it binds toand cleaves scaffold cross-links. This hypothesis is further sup-ported by the timescales of diffusion and measured MMP degra-dation. As previously reported, the hydrogel scaffold is completelydegraded over 24–48 h in a sample of similar dimensions usingvaried concentrations (0.1–0.3 mg/mL) of collagenase, a mixtureof MMPs delivered in bulk solution (30). MMPs diffuse throughour scaffolds with a relatively short characteristic time constant.The initial mesh size, before material degradation, is on the orderof ten nanometers, and the size of MMPs is several nanometers.This allows MMPs to diffuse through the matrix with a limitedhindrance. The protein diffusion time is ∼1.35 h for our gels thatare 0.6 mm thick; this value is defined as tD ∼L2=D, where L is thethickness of the gel and D is the protein diffusivity, calculated asD= 7.4× 10−7 cm2·s−1 (30, 41). Therefore, these length scaleswould result in a small value of Da because the timescale of thereaction is much slower than the rate of MMP diffusion throughthe gel. However, one must recognize that all of this must beplaced in the context of complexities that arise from cell adhesionand traction in the scaffold during motility. This asymmetric de-formation can aid in diffusion of the enzyme contributing to theextent of degradation gradient measured in the pericellular region.This gradient can also be attributed to competitive inhibition

of MMP scaffold degradation due to binding of tissue inhibitor ofmetalloproteinases (TIMP) to MMPs (42–45). Measurements ofthe kinetics of inhibition in solution show that TIMP-1 and -2 bindquickly (kon ∼ 105 M−1·s−1) to MMP-2 and -9 and slowly unbinds(koff ∼ 10−3 s−1) (42). This phenomenon enables the diffusion ofMMPs away from the cell with no measurable degradation. AsTIMP unbinds from MMPs, the scaffold cross-links will begin todegrade. This is an additional mechanism that would result in thegradient in material properties in the pericellular region.

Local Hydrogel Remodeling by hMSCs During Migration. After cul-ture and initial spreading of hMSCs in MMP-degradable hydro-gels, significant migration begins to occur with time, and thiscellular degradation and remodeling of the matrix were charac-terized with MPT. On shorter timescales, hMSC were observed to

Fig. 4. Dynamic spatial rheological data of the pericellular region during cell migration. Data are taken through time at (A) 0, (B) 24, and (C) 43 min after thecell is identified. This rapidly moving cell is causing the particles to move with the cell (outlined in black) as it migrates through the acquisition window. Thesemeasurements indicate that, once the cell is spread and begins to move, the scaffold is a viscoelastic fluid.

Schultz et al. PNAS | Published online July 6, 2015 | E3761

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

2, 2

020

actively degrade and remodel the pericellular region, creating anenvironment in which cells not only adhere and spread, but alsomigrate through the material via MMP-mediated mesenchymalmechanisms. However, at longer times in this process, hMSCscompletely erode the hydrogel in the pericellular region, andsignificantly higher levels of motility (e.g., speed) are observed, aswell as a more elongated morphology. The state of the materialover the entire data collection window, 45 min, is that of a vis-coelastic fluid (open symbols in Fig. 2D, α→ 1), and cells in theseregions are highly motile.Spatial measurements of the material properties in the peri-

cellular region show this transition with time (Fig. 4 A–C). It isimportant to note that this transition enables rapid motility ofhMSCs located in these regions (≈ 140 μm/h), and ballistic ordirected motion of the probe particles (values of α→ 2) aremeasured. The enhanced particle motion indicates that there isno scaffold structure on the length scale of the probe particles,but due to particle drift quantitative material properties cannotbe extracted. As the cell begins to migrate out of the field of view(Fig. 4C), we observe that the probe particles start to resumetrajectories more indicative of Brownian motion, which we hy-pothesize is evidence that hMSC motility primarily influenceslocal particle trajectories during this long-range migration.

Visualization of Cellularly Degraded Pathways After Migration in 3DHydrogels. Cell migration in three dimensions necessitates matrixerosion and remodeling over larger lengths scales, much greaterthan those measured with microrheology. To complement MPTresults, confirm some of the conclusions drawn using the MPTmethod, and directly observe cell-mediated macroscopic remod-eling of the hydrogel, a fluorescently labeled peptide was directlyand isotropically tethered to the hydrogel structure to allow directvisualization of matrix erosion. Two-photon laser-scanning con-focal enables 3D reconstruction of macroscopic material remod-eling in hMSC-laden hydrogels. Fig. 5 shows a minimum intensityprojection of a compressed z stack of a fluorescently labeledhydrogel to visualize the cell-mediated remodeling of the localmatrix during hMSC migration. The cell, initially located at thelarge void, α, was allowed to migrate for a period of 48 h resultingin a final spatial position, β, and a distance to origin of ∼175 μm.The final position, β, of the cell can be clearly seen in the com-plementary bright-field image (Fig. 5B) of the same region. In thisimage, the initial position of the cell is circled by a dashed line tohighlight the total migrated distance. The black regions in thefluorescent image (Fig. 5A) represent voids created in the net-work, as the cell locally remodels and degrades the matrix. Theserepresentative images demonstrate that migrating cells lead tolarge matrix remodeling and erosion in three dimensions. MPTcaptures aspects of these tracks in the pericellular space, and this

can be directly observed on the macroscale as tracks and tunnelsthat allow long-range motility of hMSCs within covalently cross-linked network. Certain regions of the cell track tunnel (e.g., nearthe final location of the cell) are on a smaller size scale thancannot be visualized using this method. As both the gel and thecells are deformable, cellular translocation can still occur withinthese regions despite the small size scale of the pores. Addition-ally, after a cell has migrated through these small tunnels, localhydrogel swelling and the limitations of visualizing dark regionswithin a highly fluorescent image may reduce the ability to im-age void regions. Measuring and characterizing this erosion pro-vides fundamental insight into cell migration, the links betweenmatrix remodeling and migration mechanisms, as well as strat-egies to direct cell migration through biophysical and biochemicalscaffold design.

ConclusionsMicrorheological measurements were used to quantify cell–matrix interactions during 3D hMSC migration. MPT has provento be a valuable technique to determine the state of cell-mediatedscaffold remodeling in the pericellular region with a sensitivity todiscriminate between areas of cell adhesion and cell-mediatedscaffold degradation. Here, we measured degradation processeson two timescales for hMSCs encapsulated in three dimensions.On the short timescale, we observed the initial stages of cell-me-diated degradation and cell spreading, and that the scaffold wasfully degraded in areas farther away from the cell. These obser-vations were attributed to the small Damköhler number (i.e., thediffusion timescale is much faster than the reaction timescale) ofcell-secreted MMPs. PIV was used to characterize particlemovement over time intervals of 4–5 min on length scales greaterthan those accessible by MPT. PIV measured persistent particlemovement in regions farther away from the cell, likely due tocytoskeletal tension on the scaffold. At longer timescales, thepericellular region near the hMSCs is a viscoelastic fluid that cellscan rapidly migrate through. This microscopic material degrada-tion is the initiation of much larger macroscopic scaffold reen-gineering that ultimately results in irreversible tracks that areeroded in the scaffold. Collectively, these methods and resultsprovide links that should help the field better understand theoutside-in signaling that a cell experiences during migration, ad-vance the development of biomaterials that manipulate basiccellular processes, and improve strategies for biomaterial designfor regenerative-medicine, wound-healing, and 3D cell cultureapplications.

Materials and MethodsHydrogel Scaffold. All measurements were taken in a photopolymerizablethiol–ene network creating a chemically cross-linked MMP-degradable PEGhydrogel scaffold. Polymer functionalization was performed as describedpreviously (36, 46). Briefly, the four-arm PEG molecule (Mn, 20,000 g·mol−1;f = 4; JenKem) is end functionalized with norbornene (f = 4) and is reactedwith an MMP-degradable peptide (KCGPQG↓ IWGQCK; Mn, 1,305 g·mol−1;f = 2). This peptide is highly degradable, easily cleaved by cell secretedMMPs, and has been previously used to study 3D cellular migration (12, 14).An adhesion ligand, CRGDS (Mn, 594 g·mol−1), is tethered to the network topromote adhesion and migration by binding to integrin receptors (47). Thereaction is initiated by a highly water-soluble initiator, lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP), and a 365-nm light source (46). Car-boxylated fluorescently labeled probe particles (2a= 1.02 ± 0.03 μm; Poly-sciences) are triple washed by centrifugation and resuspension in deionizedwater before being incorporated into the precursor solution to enablepassive microrheological measurements. The hydrogel composition used forall experiments was 3 mM (7.2 × 1018 -ene functional groups) PEG-norbor-nene, 3.9 mM (4.7 × 1017 -SH functional groups) MMP-degradable peptide,1 mM CRGDS, 0.04% (solids per volume) probe particles, and 2 × 105 hMSCsper mL with all components dissolved in 1× Dulbecco’s PBS (1× PBS; LifeTechnologies). This composition is chosen due to previous success studyinghMSC migration (14). Hydrogels are formed in sample chambers (describedbelow), ensuring that the gel has sufficient room to swell during incubation.

Fig. 5. Fluorescently labeled hydrogels allow for the visualization of cell-mediated remodeling during migration. (A) A minimum intensity projectionof a compressed z stack of a fluorescently labeled (AF-546) hydrogel permitsvisualization of the void (black) regions present in the gel from cellulardegradation and remodeling. Over 48 h, the cell migrated ∼ 175 μm from itsoriginal location, α, to its final location, β. Void tracks from cell spreadingand migration can be seen developing off of this spherical void. (B) Thebright-field image depicts the initial cell location after polymerization circledwith a dashed line, α, and the cell is located at its final position, β, after 48 hof migration. (Scale bar: 50 μm.)

E3762 | www.pnas.org/cgi/doi/10.1073/pnas.1511304112 Schultz et al.

Dow

nloa

ded

by g

uest

on

June

2, 2

020

Device Fabrication and 3D Cell Encapsulation. Hydrogels with encapsulatedhMSCs are made in a device that enables MPT measurements. This devicereduces drift of the probe particle as the hydrogel scaffold degrades andenables incubation of the encapsulated hMSCs during data collection. The deviceconsists of a glass-bottom Petri dish (d= 0.35mm, no. 1.5 glass coverslip; MatTek)with a polydimethylsiloxane (PDMS) (Dow Corning) chamber attached to theglass slide. The PDMS chamber is made using manufacturer’s instructions,namely, 10:1 silicone elastomer base:cross-linking agent. PDMS is degassed andcured in a flat sheet overnight for 65° C. The cured PDMS is cut using two biopsypunches and creates a circular chamber with an inner diameter of 6 mm and anouter diameter of 10 mm. The chamber is attached to the glass bottom usinguncured PDMS and incubating at 65° C overnight. The sample chambers aresterilized with ethanol and UV light before cell-laden hydrogel formation.Hydrogels are made using the concentrations described above and are curedunder UV light (365 nm, 10 mW·cm−2) for 3 min in a sterile hood. The volume ofthe gel created in the samples is adjusted to 17 μL, enabling complete swelling.Samples are incubated overnight before data acquisition.

hMSC Culture and Inhibitor Treatment. hMSCs were isolated from bonemarrow aspirates (Lonza). The isolation and freezing procedure are describedpreviously (14). Cells were passaged, and passages 2 and 3 were used foreach experiment.

MPT and PIV Measurements.MPTwas used to measure the material propertiesof the hydrogel scaffold during hMSC migration. MPT is a passive micro-rheological technique in which the Brownian motion of probe particles ismeasured and related to rheological properties using the GSER. Data arecollected using optical video microscopy enabling simultaneous measure-ments of dynamic material properties, such as cell-mediated scaffold deg-radation, and visualization of the microenvironment. Data were taken usingan inverted microscope (Nikon TE2000E; Nikon Instruments) with a low–

numerical-aperture oil-immersion objective at 60× magnification (N.A., 1.4;1× optovar; Nikon Instruments). An incubation chamber is connected to themicroscope maintaining samples at 37° C and 5% CO2 to ensure cells remainhealthy during data collection. Data were collected at 30 frames per s for800 frames (≈ 27 s) and an exposure time of 1 ms (CMOS high-speed camera;Hi-Spec 3; 1,024 × 1,280 pixels; Fastec Imaging Corporation), parameters cho-sen to minimize the effects of static and dynamic particle tracking errors (29).

MPT measurements were collected of an area directly around a migratinghMSC, the pericellular region. This was done by first identifying a cell usingbright-field microscopy. An image of each cell was collected with bright-fieldmicroscopy immediately before fluorescence MPT data were taken. Ap-proximately 10 movies were collected in the pericellular region of each cellover approximately an hour. Particle tracking was performed using classictracking algorithms developed by Crocker and Grier and maintained byWeeks (26, 28) that identify the brightness-weighted centroid of each par-ticle and link the positions together in each frame to create a trajectory.From these data, the MSD of the probe particles was calculated and used toidentify the state of the material (30). For each condition, three biologicalreplicates were measured. Within the biological replicates, two separate gelstock solutions were measured over 2 d with two to three gels made persolution. Hydrogels were also measured in the absence of cells. These gelswere made and incubated in an identical manner to cell-laden hydrogelsand showed no significant scaffold degradation when incubated.

PIV analysis was performed on bright-field images taken of the sameobservation window as MPT data. Iterative PIV analysis was done usingImageJ (NIH Image) and the PIV plugin (48). In this analysis, the image isbroken up into smaller interrogations windows, and the cross-correlationbetween particle movements in these windows is identified and plotted.

Visualizing Cell Tracks in Hydrogel. For visualizing cell tracks present in thescaffolds, CRGDS was fluorescently labeled with Alexa Fluor 546 succinimidylester (AF546-NHS). Briefly, AF546-NHS was dissolved in dimethyl formamideand reacted overnight with the peptide before cleavage from the resin.The fluorescently labeled peptide was then introduced into the macromersolution before photopolymerization at 1 mM. hMSC encapsulation andhydrogel formation were performed as discussed above. After 48 h, z-stackimages were taken using an LSM 710 confocal microscope (Carl Zeiss) at astep size of 0.78 μm. The z stacks were compressed, and a minimum intensityprojection was created using ImageJ (NIH Image). Images presented withinthe manuscript were cropped and adjusted for contrast and brightness forbetter illustration of cell tracks.

ACKNOWLEDGMENTS. We acknowledge Dr. Jennifer Leight for helpfuldiscussion. Funding for this work was provided by Howard Hughes MedicalInstitute, the National Science Foundation (CTS 1236662), and the NationalInstitutes of Health (RO1DE016523).

1. Slaughter BV, Khurshid SS, Fisher OZ, Khademhosseini A, Peppas NA (2009) Hydrogelsin regenerative medicine. Adv Mater 21(32-33):3307–3329.

2. Hubbell JA (1995) Biomaterials in tissue engineering. Biotechnology (N Y) 13(6):565–576.

3. Peppas NA, Langer R (1994) New challenges in biomaterials. Science 263(5154):1715–1720.

4. Kloxin AM, Kloxin CJ, Bowman CN, Anseth KS (2010) Mechanical properties of cell-ularly responsive hydrogels and their experimental determination. Adv Mater 22(31):3484–3494.

5. Hern DL, Hubbell JA (1998) Incorporation of adhesion peptides into nonadhesivehydrogels useful for tissue resurfacing. J Biomed Mater Res 39(2):266–276.

6. DeLong SA, Moon JJ, West JL (2005) Covalently immobilized gradients of bFGF onhydrogel scaffolds for directed cell migration. Biomaterials 26(16):3227–3234.

7. Lutolf MP, et al. (2003) Synthetic matrix metalloproteinase-sensitive hydrogels for theconduction of tissue regeneration: Engineering cell-invasion characteristics. Proc NatlAcad Sci USA 100(9):5413–5418.

8. Patterson J, Hubbell JA (2010) Enhanced proteolytic degradation of molecularly en-gineered PEG hydrogels in response to MMP-1 and MMP-2. Biomaterials 31(30):7836–7845.

9. Lin CC, Boyer PD, Aimetti AA, Anseth KS (2010) Regulating MCP-1 diffusion in affinityhydrogels for enhancing immuno-isolation. J Control Release 142(3):384–391.

10. Discher DE, Janmey P, Wang YL (2005) Tissue cells feel and respond to the stiffness oftheir substrate. Science 310(5751):1139–1143.

11. Engler AJ, Sen S, Sweeney HL, Discher DE (2006) Matrix elasticity directs stem celllineage specification. Cell 126(4):677–689.

12. Schwartz MP, et al. (2010) A synthetic strategy for mimicking the extracellular matrixprovides new insight about tumor cell migration. Integr Biol (Camb) 2(1):32–40.

13. Anderson SB, Lin CC, Kuntzler DV, Anseth KS (2011) The performance of humanmesenchymal stem cells encapsulated in cell-degradable polymer-peptide hydrogels.Biomaterials 32(14):3564–3574.

14. Kyburz KA, Anseth KS (2013) Three-dimensional hMSC motility within peptide-func-tionalized PEG-based hydrogels of varying adhesivity and crosslinking density. ActaBiomater 9(5):6381–6392.

15. Tan JL, et al. (2003) Cells lying on a bed of microneedles: An approach to isolatemechanical force. Proc Natl Acad Sci USA 100(4):1484–1489.

16. Tambe DT, et al. (2011) Collective cell guidance by cooperative intercellular forces.Nat Mater 10(6):469–475.

17. Wolf K, Friedl P (2009) Mapping proteolytic cancer cell-extracellular matrix interfaces.Clin Exp Metastasis 26(4):289–298.

18. Wolf K, et al. (2007) Multi-step pericellular proteolysis controls the transition fromindividual to collective cancer cell invasion. Nat Cell Biol 9(8):893–904.

19. Wolf K, et al. (2013) Physical limits of cell migration: Control by ECM space and nu-clear deformation and tuning by proteolysis and traction force. J Cell Biol 201(7):1069–1084.

20. Packard BZ, Artym VV, Komoriya A, Yamada KM (2009) Direct visualization of pro-tease activity on cells migrating in three-dimensions. Matrix Biol 28(1):3–10.

21. Legant WR, et al. (2010) Measurement of mechanical tractions exerted by cells inthree-dimensional matrices. Nat Methods 7(12):969–971.

22. Guvendiren M, Burdick JA (2013) Engineering synthetic hydrogel microenvironmentsto instruct stem cells. Curr Opin Biotechnol 24(5):841–846.

23. Bloom RJ, George JP, Celedon A, Sun SX, Wirtz D (2008) Mapping local matrix re-modeling induced by a migrating tumor cell using three-dimensional multiple-par-ticle tracking. Biophys J 95(8):4077–4088.

24. Alkhouli N, et al. (2013) The mechanical properties of human adipose tissues and theirrelationships to the structure and composition of the extracellular matrix. Am JPhysiol Endocrinol Metab 305(12):E1427–E1435.

25. Khetan S, et al. (2013) Degradation-mediated cellular traction directs stem cell fate incovalently crosslinked three-dimensional hydrogels. Nat Mater 12(5):458–465.

26. Crocker JC, Grier DG (1996) Methods of digital video microscopy for colloidal studies.J Colloid Interface Sci 179(1):298–310.

27. Mason TG, Ganesan K, van Zanten JH, Wirtz D, Kuo SC (1997) Particle tracking mi-crorheology of complex fluids. Phys Rev Lett 79(17-21):3282–3285.

28. Crocker JC, Weeks ER (2011) Particle tracking using IDL. Available at www.physics.emory.edu/faculty/weeks//idl. Accessed March 7, 2011.

29. Savin T, Doyle PS (2005) Static and dynamic errors in particle tracking microrheology.Biophys J 88(1):623–638.

30. Schultz KM, Anseth KS (2013) Monitoring degradation of matrix metalloproteinases-cleavable PEG hydrogels via multiple particle tracking microrheology. Soft Matter9:1570–1579.

31. Schultz KM, et al. (2012) Capturing the comprehensive modulus profile and reversepercolation transition of a degrading hydrogel. Macro Lett 1(6):706–708.

32. Adolf D, Martin JE (1990) Time-cure superposition during crosslinking. Macromole-cules 23(15):3700–3704.

33. Winter HH, Chambon F (1986) Analysis of linear viscoelasticity of a crosslinkingpolymer at the gel point. J Rheol 30(2):367–382.

34. Schultz KM, Furst EM (2012) Microrheology of biomaterial hydrogelators. Soft Matter8:6198–6205.

Schultz et al. PNAS | Published online July 6, 2015 | E3763

ENGINEE

RING

APP

LIED

BIOLO

GICAL

SCIENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

June

2, 2

020

35. Aimetti AA, Machen AJ, Anseth KS (2009) Poly(ethylene glycol) hydrogels formed bythiol-ene photopolymerization for enzyme-responsive protein delivery. Biomaterials30(30):6048–6054.

36. Fairbanks BD, et al. (2009) A versatile synthetic extracellular matrix mimic via thiol-norbornene photopolymerization. Adv Mater 21(48):5005–5010.

37. Raeber GP, Lutolf MP, Hubbell JA (2005) Molecularly engineered PEG hydrogels: Anovel model system for proteolytically mediated cell migration. Biophys J 89(2):1374–1388.

38. Kienberger F, et al. (2000) Static and dynamical properties of single poly(ethyleneglycol) molecules investigated by force spectroscopy. Single Mol 1(2):123–128.

39. Miller JS, et al. (2010) Bioactive hydrogels made from step-growth derived PEG-peptide macromers. Biomaterials 31(13):3736–3743.

40. Yang YL, Leone LM, Kaufman LJ (2009) Elastic moduli of collagen gels can be pre-dicted from two-dimensional confocal microscopy. Biophys J 97(7):2051–2060.

41. Weber LM, Lopez CG, Anseth KS (2009) Effects of PEG hydrogel crosslinking densityon protein diffusion and encapsulated islet survival and function. J Biomed Mater ResA 90(3):720–729.

42. Olson MW, Gervasi DC, Mobashery S, Fridman R (1997) Kinetic analysis of the bindingof human matrix metalloproteinase-2 and -9 to tissue inhibitor of metalloproteinase(TIMP)-1 and TIMP-2. J Biol Chem 272(47):29975–29983.

43. Lozito TP, Tuan RS (2011) Mesenchymal stem cells inhibit both endogenous and ex-ogenous MMPs via secreted TIMPs. J Cell Physiol 226(2):385–396.

44. Reed MJ, Koike T, Sadoun E, Sage EH, Puolakkainen P (2003) Inhibition of TIMP1enhances angiogenesis in vivo and cell migration in vitro. Microvasc Res 65(1):9–17.

45. Mascall KS, Small GR, Gibson G, Nixon GF (2012) Sphingosine-1-phosphate-inducedrelease of TIMP-2 from vascular smooth muscle cells inhibits angiogenesis. J Cell Sci125(Pt 9):2267–2275.

46. Fairbanks BD, Schwartz MP, Bowman CN, Anseth KS (2009) Photoinitiated polymer-ization of PEG-diacrylate with lithium phenyl-2,4,6-trimethylbenzoylphosphinate:Polymerization rate and cytocompatibility. Biomaterials 30(35):6702–6707.

47. Nuttelman CR, Tripodi MC, Anseth KS (2005) Synthetic hydrogel niches that promotehMSC viability. Matrix Biol 24(3):208–218.

48. Tseng Q (2014) PIV (Particle Image Velocimetry)—ImageJ plugin. Available at https://sites.google.com/site/qingzongtseng/piv. Accessed November 28, 2014.

E3764 | www.pnas.org/cgi/doi/10.1073/pnas.1511304112 Schultz et al.

Dow

nloa

ded

by g

uest

on

June

2, 2

020