Particle Diffusion in Polymeric Hydrogels with ... - ACS Publications

Jun 27, 2018 - Johann Hansing† , Joseph R. Duke, III‡ , Emily B. Fryman‡ , Jason E. DeRouchey*‡ , and Roland R. Netz*†. † Fachbereich für...
0 downloads 0 Views 7MB Size
Subscriber access provided by UNIV OF DURHAM

Communication

Particle Diffusion in Polymeric Hydrogels with Mixed Attractive and Repulsive Interactions Johann Hansing, Joseph R Duke, Emily B Fryman, Jason E DeRouchey, and Roland R. Netz Nano Lett., Just Accepted Manuscript • DOI: 10.1021/acs.nanolett.8b02218 • Publication Date (Web): 27 Jun 2018 Downloaded from http://pubs.acs.org on June 27, 2018

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Particle Diusion in Polymeric Hydrogels with Mixed Attractive and Repulsive Interactions †

Johann Hansing,

Joseph R. Duke III,





Emily B. Fryman,

and Roland R. Netz

†Fachbereich ‡Department

∗,†

für Physik, Freie Universität Berlin, 14195 Berlin, Germany

of Chemistry, University of Kentucky, Lexington, KY 40506, USA

E-mail: [email protected]; [email protected]

Abstract All biogels are heterogeneous, consisting of functional groups with dierent biophysical properties arrayed on spatially disordered polymer networks. Nanoparticles diusing in such biogels experience a mixture of attractive and repulsive interactions. Here, we present experimental and theoretical studies of charged particle diusion in gels with a random distribution of attractive and repulsive electrostatic interaction sites inside the gel. In addition to interaction disorder, we theoretically investigate the eect of spatial disorder of the polymer network. Our coarse-grained simulations reveal that attractive interactions primarily determine the diusive behavior of the particles in systems with mixed attractive and repulsive interactions. As a consequence, charged particles of either sign are immobilized in mixed cationic/anionic gels since they are trapped near oppositely charged interaction sites, whereas neutral particles diuse rapidly. Even small fractions of oppositely charged interaction sites lead to strong trapping of a charged particle. Translational diusion coecients of charged probe molecules in gels consisting of mixed cationic and anionic dextran polymers are 1

ACS Paragon Plus Environment

∗,‡

Jason E. DeRouchey,

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

determined by uorescence correlation spectroscopy and quantitatively conrm our theoretical predictions. keywords: Diusion, hydrogels, mucus, biological barriers, Langevin simulations

Introduction Biological hydrogels are known to fulll a number of important physiological functions. 1 Besides regulating the mechanical properties of cells and serving as lubricants in joints, biogels act as barriers against pathogens, thus playing a vital role in protecting organisms. Biological hydrogels, such as mucus and the extracellular matrix (ECM), also serve as selective lters for nutrients, proteins, ions and drugs. Understanding the selective barrier properties of biogels is a research topic where dierent scientic branches such as physics, chemistry, biology and medicine merge. Alongside volume exclusion 2 and hydrodynamic eects, 3 it has been established that nonsteric interactions, which can be of electrostatic or hydrophobic nature, are a major factor that governs the mobility of particles in biogels. 1,48 Through a combined theoretical and experimental approach, we have previously shown that particle transport in homogeneously charged gels is highly asymmetric 9,10 in the following sense: Attractive electrostatic interations between particle and gel are much more eective than repulsive interations in hindering particle diusion. This is due to the sticking of particles at the vertices of the oppositely charged polymer network. However, biopolymers in vivo typically are imhomogeneous and contain dierent hydrophobic or electrostatically charged monomers that can repel or attract the particle. For instance, the mucin protein MUC5AC contains numerous basic and acidic amino acids which can be positively or negatively charged, respectively, according to their pK value and the pH of the solution. 11 Furthermore, many biogels are in fact multicomponent systems, consisting of a mixture of several polymers with dierent biophysical properties. The ECM, for example, contains the biopolymers laminin and collagen IV with proteins including perlecan and nidogen acting as cross-linker agents. 12 Thus, 2

ACS Paragon Plus Environment

Page 2 of 28

Page 3 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

nanoparticles and macromolecules diusing in vivo experience a heterogeneous environment with mixed attractive/repulsive interactions, greatly impacting their transport properties. Prior studies showed that biogels such as the ECM 12 and mucus 11,13 act eectively as an electrostatic bandpass. By this we mean that diusion of both positively and negatively charged nanoparticles is hindered, while neutral and near-neutral charged objects can more freely diuse through the matrix. The magnitude of the net charge, not the sign, is the key determinant. To explain this diusive behavior, Lieleg and coworkers proposed that biogels can be understood as a network of localized positively and negatively charged ber segments. Charged nanoparticles stick to oppositely charged ber segments, which strongly reduces their mobility, while neutral particles can diuse rapidly. 11,12 In order to investigate this idea systematically, we here present a quantitative comparison between experimental data for the diusion of a charged probe molecule in mixed cationic/anionic dextran gels and a coarse-grained simulation model for nanoparticle diffusion in interacting gels with a random distribution of interaction sites. In the model, a spherical particle diuses inside a network of rigid cylindrical bers, which consist of segments that are randomly assigned to be attractive or repulsive towards the particle with an exponentially screened interaction potential. We thus expand our previous model for nanoparticle diusion in interacting gels where either all bers are purely attractive or all bers are purely repulsive with the bers arranged on a cubic symmetric lattice. 9 We also study the eect of spatial disorder on the diusion properties and systematically investigate the combined eects of spatial disorder and interaction disorder, which are both fundamental physical concepts governing diusion in biological systems. Rigid cylindrical bers have previously been employed to model the sti collagen network of the ECM 3,14 and mucus. 9 Furthermore, recent experimental research indicates that particles experience a rather rigid ber network inside mucus gel, 15 so the rigid network model employed by us presumably is a good starting point, although it should be noted that network exibility has recently been shown to be relevant for particle diusion in purely repulsive gels. 16 Our simulations conrm 3

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 4 of 28

that in gels with randomly distributed, attractive and repulsive interaction sites, particles are strongly localized near attractive sections of the ber network. As a consequence, interacting particles are immobilized in gels with mixed attractive and repulsive interactions and the trapping mechanism is closely related to the trapping mechanisms observed in gels with purely attractive electrostatic particle-ber interactions. In fact, we show that gels with mixed interactions trap particles nearly as eectively as gels with purely attractive interactions. As an experimental model system, we investigate the translational diusion coecients of negatively charged Alexa488 probe molecules in mixed cationic/anionic hydrogels consisting of a mixture of electrostatically positive DEAE-dextran and negative CM-dextran polymer chains by uorescence correlation spectroscopy (FCS) measurements. Quantitative agreement between simulation and the experiment is found with only a few physical t parameters. Both experiment and simulation show that particle transport in mixed gels, over a broad range of compositions, is nearly identical to purely attractive gels despite the reduced number of attractive interaction sites. (a)

𝑎

(b)

(c)

ܾ/4

s b

Figure 1: (a) Schematics of the simulation model, dening the particle diameter p, the ber diameter a, the steric diameter s = a + p and the mesh size b. Sketch (b) shows the positions of 16 parallel bers, indicated as black dots, displaced from a reference cubic lattice with spacing b (dashed lines) by Gaussian random numbers with a standard deviation of σd /b = 0.2 and zero mean. Sketch (c) indicates the sign and strength of the interaction potential, where blue stands for attraction and red stands for repulsion.

4

ACS Paragon Plus Environment

Page 5 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

Figure 2: (a) Particle position snap shots (red spheres) sampled at consecutive times from simulations of gels with mixed interactions (left) and of gels with purely attractive bers (right). The sign of the interaction potential along the linear bers is indicated in red (repulsive) and blue (attractive). Note that in the mixed interaction case, the sign of each ber segment is randomly chosen with equal probability (ψ = 0.5). The top row gures show spatially ordered gels with spatial disorder strength of σd = 0 and the bottom row gures show spatially disordered gels with σd /b = 0.9. The gures demonstrate similar trapping mechanisms for the mixed case (left) and for the purely attractive case (right). For spatially ordered gels (top gures) the particle is trapped at the ber vertices (vertex trapping ). In the gel with mixed interactions (top left), the degree of localization diers between the vertices and depends on the number of adjacent attractive ber segments. In the spatially disordered gels (bottom gures) the particle is trapped in regions of high local ber density (dense-region 9 trapping ). All simulations are performed over the same interval of 2×10 steps with the same absolute value for the interaction strength |U0 |/kB T = 15, an interaction range of k/s = 0.5 and a steric diameter of s/b = 0.2. (b) Mean squared displacement (MSD) as a function of time for a mixed gel with U0 /kB T = ±15 and σd /b = 0.9 (corresponding to the snap shots at the bottom left in (a)). (c) The MSD divided by time approaches a constant value which corresponds to the long-time diusivity D/D0 = 0.046 (broken black line). The constant long-time limit is reached after a displacement of roughly h∆r2 (t)i > b2 .

5

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 28

Experimental Methods Materials.

Diethylaminoethyl-dextran (DEAE-dextran(+), Mw = 500kDa) and carboxymethyl-

dextran (CM-dextran(-), Mw = 15-20kDa) were purchased from Sigma Aldrich. According to the manufacturer dextran(-) has between 1.1 and 1.5 mmol carboxymethyl per gram dextran. The mean value of 1.3 mmol carboxymethyl per gram dextran corresponds to approximately two negative charges per nine monomers. For dextran(+) the nitrogen content is 2.9-3.5%, which corresponds to about one amine group per two monomers for the mean value of 3.2%. Alexa Fluor® 488 Succinimidyl Ester dye (Alexa488, Abs/Em peak: 495/519 nm), and Rhodamine 110 (Abs/Em peaks: 496/520 nm) were purchased from Fisher Scientic. Rhodamine 110 was used in calibration of the confocal volume of the FCS instrumentation. Probe diusion was performed with Alexa488, which has a net negative charge at near neutral pH. The uorescent molecules were readily soluble in water, and did not require further purication prior to use.

Preparation of dextran solutions.

Dextran polymer stock solutions were prepared

by dissolving solid dextran in 10 mM MES buer (pH = 6.4) to a nal concentration of 8-20 %w/v. Solutions were briey vortexed and incubated with gentle rocking overnight at room temperature to ensure homogeneity. Subsequent dilutions with MES were made from the stock solutions resulting in the desired nal concentrations of polymer solutions. All polymer solutions were allowed to equilibrate for 24 hours before use. For FCS experiments, uorescent probe molecules (Alexa488 dye) were prepared and mixed with the desired dextran solutions to achieve a nal probe concentration of 5-10 nM. Samples were then mixed thoroughly and incubated at room temperature for more than 6 hours to ensure uniform dispersion of the probe molecules throughout the dextran polymer solution. 500 µL of sample from each solution were loaded into NUNC LabTek 8-well microscopy chambers and measured directly by FCS at room temperature. In prior studies, the dextrans were extensively dialyzed to remove salt. Probe diusion coecients, as determined by FCS, were within experimental error for dialyzed and non-dialyzed dextrans, suggesting the commercial dex6

ACS Paragon Plus Environment

Page 7 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

trans used in this study are suciently salt free to be used without further purication. For mixed solutions of dextran(+) and dextran(-), stock solutions of each were added together to achieve the desired volumetric ratio of each dextran with respect to total dextran. For all mixed dextran solutions studied, the resulting mixtures resulted in homogeneous polymer solutions.

FCS setup.

Fluctuations in uorescence intensity data were collected using a com-

mercial dual-channel confocal spectrometer (ALBA FFS system, ISS, Champaign, IL). FCS experiments were made using a continuous wave 488 nm laser diode as an excitation source passed through a 514 nm long pass edge lter before detection. Excitation light was directed into experimental samples through a Nikon Ti-U microscope (60x/1.2 NA water-immersion objective lens). The emission signal was recorded by two separate Hamamatsu H7422P-40 photomultiplier tubes (PMTs). Confocal volume dimensions were determined through measurement of aqueous Rhodamine 110 at known concentrations with a diusion coecient of

D = 440 (µm2 s−1 ). 17 In order to ensure solution homogeneity, reported results are the average of at least 9 measurements at dierent positions within the dextran solutions. Sampling times of 30 seconds were used for all measurements. FCS curves were analyzed using the VistaVision Software (ISS, Champaign, IL) to determine the diusion coecient.

FCS data analysis.

The principles and experimental realization of FCS have been

described in detail elsewhere, 1821 here we give only a brief overview. FCS measures the uorescence uctuations emitted from labeled molecules moving in and out of a small confocal volume (∼1 fL). The size of the eective illumination volume is xed by the confocal detection optics and the excitation prole of the focused laser beam and characterized by measurements against a standard of known diusion constant (here Rhodamine 110). The measured time traces of uorescent events are compared for self-similarity after a lag time τ by calculation of the normalized cross correlation, G(τ ):

G(τ ) = 1 +

hδF (t)δF (t + τ )i . hF (t)i2 7

ACS Paragon Plus Environment

(1)

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 8 of 28

δF (t) and δF (t + τ ) represent deviation from the mean uorescence hF (t)i at time t and after time t+τ . For uniformly distributed uorescent particles diusing by Brownian motion, dynamic information can be determined from the intensity uctuations by means of a time autocorrelation given by:

1 G(τ ) = 1 + · N



1 1 + τ /τD



1

·p

1 + w02 τ /z02 τD

(2)

where τD represents the dwell time of the particles in the confocal volume, whose shape can be approximated as a Gaussian ellipsoid with axial height z0 and equatorial width w0 as determined by calibration measurements. N is the average number of particles occupying the observation volume. The normalized autocorrelation can then be calculated by Gnorm (τ ) =

G(τ )/G(0). Cross correlation with both detectors ensures that the resulting autocorrelation is free from the eects of detector after-pulsing. The translational diusion coecient D

(µm2 s−1 ) can be calculated from τD and the equatorial width using (3)

τD = w02 /4D.

FCS has been shown previously to be eective for measuring diusion of probe molecules in polymeric solutions. 2224

Simulation Methods The Brownian dynamics simulations are based on the discretized Langevin equation in three dimensions

∆ri = − µ ∂i U (r) +

p

2µ ζi ,

(4)

where ∆ri is the displacement of the diusing particle in direction i = x, y, z , ∂i the spatial derivative and ζ a Gaussian distributed random number with zero mean and variance hζi ζj i = 8

ACS Paragon Plus Environment

Page 9 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

δij , where δij is the Kronecker delta. The energy U is rescaled by the thermal energy kB T . All lengths are rescaled by the mesh size b, which is indicated in g. 1a. µ = ∆t µ0 kB T /b2 is the rescaled timestep, where µ0 is the bulk sphere mobility. For the simulations a small enough rescaled timestep µ has to be chosen. Dierent time steps were tested and no increase in accuracy was found for µ ≤ 10−6 . For the simulations presented in this work, a rescaled timestep µ of 10−6 was therefore chosen. The diusivity D of the particle is obtained by linearly tting the mean squared displacement (MSD) h∆r2 (t)i = h(r(t) − r(0))2 i in the long-time limit according to

lim h∆r2 (t)i = 6D t .

t→∞

(5)

The free diusion diusivity, D0 = µ0 kB T , corresponds to the diusivity of the particle if no polymers are present. For each data point, we simulate a single particle trajectory of about

∼ 109 steps which is long enough so that the long time limit in eq. (5) is reached. 9 Sample repeat simulations show that the error for D/D0 is always below 5%.

Polymer gel model.

The polymer gel consists of 48 bers; 16 bers parallel to each

axis x, y and z . Our model contains both interaction as well as spatial disorder. In recent work we have explored the eects of spatial ber disorder on nanoparticle diusion in the absence of interaction disorder. 25 We introduce spatial disorder by displacing the bers from their positions on a reference cubic lattice with spacing b. The displacement of each ber is a random vector orthogonal to the ber axis, sampled from a Gaussian distribution with zero mean and standard deviation σd . A high standard deviation σd corresponds to increased spatial disorder. Thus, σd is henceforth referred to as the spatial disorder strength. Figure 1b shows a 2D projection of a weakly spatially disordered gel for σd /b = 0.2. When the particle leaves the central cell of the reference cubic lattice (dashed lines), the eight distal bers are removed and eight new bers are added at random positions on the side to which the particle has moved. Thus, the lattice changes as the particle moves across cells but the average mesh

9

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 10 of 28

size b is conserved. The steric interaction between the bers and the particle is governed by a purely repulsive truncated shifted Lennard-Jones potential

U (r) =

48 X n=1

    4

s 2ρn

12





s 2ρn

6

+

1 4



  0 ,

, ρn ≤ 2−5/6 s

(6)

ρn > 2−5/6 s ,

where the energy depth is xed at  = 1 kB T , ρn is the closest distance between the particle and the nth ber. The steric diameter s is the sum of the ber diameter and the particle diameter, s = a + p, as illustrated in g. 1a. To include interaction disorder, we randomly assign an attractive or a repulsive interaction strength, U− < 0 and U+ > 0, to ber segments with a length of b along each ber. The long-range, nonsteric interaction between the particle and ber segments is dened as

 ρ U (r) = U± exp − . k

(7)

Here, ρ is the radial distance between the particle and the ber and k is the interaction range. To avoid discontinuities in the potential the particle experiences, the potential strength goes linearly to zero over a range of b/4, if two neighboring ber segments have an interaction strength of opposite sign, as indicated in g. 1c. The probability for an interaction site to be repulsive is ψ . ψ = 1 corresponds to a gel with purely repulsive particle-ber interactions. For ψ = 0.5 both signs for the interaction potential are equally likely, if additionally the interaction strengths are the same, U+ = −U− = U0 , the gel is eectively net charge neutral. Renderings of model gels are presented in g. 2a. The top gures show spatially ordered, cubic gels with σd = 0 and the bottom gures depict spatially disordered gels with σd /b = 0.9. The gels on the left are mixed gels (ψ = 0.5), whereas on the right, gels with purely attractive tracer-ber interactions are shown (ψ = 0). Equation (7) is a general nonsteric interaction potential that describes exponentially 10

ACS Paragon Plus Environment

Page 11 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

screened particle-gel interactions. For electrostatic interactions, U± can be interpreted as the product of the particle charge and the linear polymer charge density. 9 In this case, the interaction range k corresponds to the Debye screening length 26

k2 =

1 , 4πlB I

(8)

where lB = e2 /4πεkB T is the Bjerrum length, e is the elementary charge and ε the perP mittivity. I = 21 j nj zj2 is the ionic strength and zj the valence of salt ion j and nj its number density. The salt number density n is related to the molar ion concentration through

CIon = n/NA , where NA is the Avogadro constant. Following our previous work on particle diusion in interacting gels 9,10 we neglect hydrodynamic interactions.

Results and Discussion We rst present general simulation results for the diusion of particles in mixed gels in comparison to diusion in gels with purely attractive and gels with purely repulsive longrange particle-ber interactions. First, we assume for the mixed case that the interaction strengths are equal and given by U+ = −U− = U0 and ψ = 0.5, i.e. the mixed gels are eectively net charge neutral. Subsequently, we compare simulation and experiment for the diusion of charged Alexa488 particles in dextran gels containing mixtures of cationic and anionic polymers. Here, U+ , U− and ψ are dictated by the experiment. To demonstrate how the diusivities for our simulations are obtained from the MSD in the long-time limit, exemplary MSD plots are shown in gs. 2b and 2c. It is seen that the long-time diusive limit is reached when the MSD exceeds the squared lattice constant b2 .

Mixed gels are similar to gels with purely attractive electrostatic interactions. Figure 3 shows the relative diusivity D/D0 as a function of the interaction potential strength

U0 for dierent steric diameters s/b = 0.1, 0.2 and 0.5 and a rescaled interaction range of k/s = 0.5. We present data for spatially disordered and ordered gels with spatial disorder 11

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

strengths of σd /b = 0.9 and 0, respectively, in gs. 3a and b. The data for the mixed gels (colored, lled symbols) is, by design of our model, symmetric around U0 = 0. One can see that mixed gels trap particles more strongly with rising particle-ber interaction strength

U0 , regardless of the sign of U0 . This "bandpass" like behavior is qualitatively similar to what has been previously reported for the diusion of charged nanoparticles in the ECM 12 and in mucus. 11,13 The data indicated by open symbols, connected with dashed lines in gs. 3a and b, correspond to simulations with purely attractive (for U0 < 0) or purely repulsive (for U0 > 0) electrostatic interactions. For U0 > 0, the diusivities are much higher than for mixed gels. In contrast, for U0 < 0 the data curves for purely attractive gels agree qualitatively with the data for mixed gels, for the disordered system in g. 3a nearly quantitatively. An exception is the s/b = 0.5 case for a spatially ordered gel in g. 3b. This is due to the peculiarities of the trapping mechanism in ordered gels with purely attractive electrostatic interactions and will be discussed in the next section. The similarity between D for gels with purely attractive electrostatic interactions and mixed gels is an indicator that both gel types give rise to similar particle trapping mechanisms. The simulation snap shots in g. 2a illustrate which mechanisms govern particle trapping in mixed gels for a spatially ordered gel (σd = 0, top left) and a spatially disordered gel (σd /b = 0.9, bottom left). For each case in g. 2a, the snap shots are obtained at consecutive times from a single particle trajectory. For spatially ordered mixed gels, the particle tends to stay near the attractive vertices of the cubic ber lattice. We refer to this as vertex trapping. For disordered mixed gels, the particle is strongly localized in regions with a high density of attractive ber segments, which we refer to as dense-region trapping. Similar trapping mechanisms are observed in purely attractive gels in g. 2a on the right hand side. This indicates that mixed gels and attractive gels both immobilize particles in a similar fashion. Dense-region trapping and vertex trapping are illustrated schematically for mixed gels in g. 3c. The strong similarity between the trapping mechanisms for mixed and purely attractive 12

ACS Paragon Plus Environment

Page 12 of 28

Page 13 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

gels can be quantied by calculating the average number of bers hNlocal i at a radial distance of less than b/2 from the particle. The magnitude of hNlocal i serves as a measure for the particle-ber correlations and thus for the trapping mechanisms in dierent gels. Figures 4a-c show hNlocal i as a function of U0 for a steric diameter s/b = 0.5 for mixed gels, as well as for gels with purely attractive and gels with purely repulsive electrostatic interactions, for comparison. For the attractive case (g. 4b) hNlocal i increases with increasing interaction strength, since the particle tends to stay closer to the attractive bers. For ordered gels,

σd = 0, the maximum value of hNlocal i is 3, which corresponds to the case where the particle is highly localized at the ber vertices (c.f. g. 2a top right). With increasing spatial disorder,

hNlocal i increases monotonically. This illustrates dense-region trapping for disordered gels, where the particle tends to be trapped in regions of high local ber density (c.f. g. 2a bottom right). Comparing hNlocal i for gels with purely attractive electrostatic interactions, g. 4b, to hNlocal i for mixed gels, g. 4a, one sees that both gel types exhibit the same trapping mechanism, in accordance to our qualitative observations in g. 2a. This explains the strong qualitative similarity between D for the mixed case and the attractive case in gs. 3a and b. For gels with purely repulsive electrostatic interactions (g. 4c), by contrast,

hNlocal i decreases monotonically for all σd . The particle moves away from the bers, into regions with low local ber density. We nd that mixed gels and gels with purely attractive interactions exhibit similar microscopic trapping mechanisms. The trapping eectiveness for both types of gel is very similar for disordered ber networks. Nanoparticle diusion is known to be strongly hindered in gels with purely attractive electrostatic particle-gel interactions; 9,10,2729 our simulation results indicate that nanoparticles are similarly immobilized in the presence of mixed attractive and repulsive electrostatic interactions of equal magnitude. This makes gels with mixed attractive and repulsive interactions very good particle lters since they lter out particles of positive and negative charge, as further discussed below.

Interaction and spatial disorder have a similar eect on particle diusion. 13

ACS Paragon Plus Environment

We

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

(a)

(b)

(c)

spatially ordered gels: vertex trapping

spatially disordered gels: dense region trapping

Figure 3: Particle diusivity D as a function of the interaction potential strength U0 for dierent steric diameters s for a rescaled interaction range of k/s = 0.5, U+ = −U− = U0 and ψ = 0.5. (a) D for spatially disordered gels with σd /b = 0.9 and (b) D for spatially ordered gels with σd = 0. Except for the s/b = 0.5 curve in (b), the diusive behavior for mixed gels (colored symbols) and for gels with purely attractive electrostatic interactions (U0 < 0, empty symbols) is qualitatively very similar, whereas gels with purely repulsive electrostatic interactions (U0 > 0, empty symbols) dier strongly from the mixed case. (c) Schematic illustrations of the dierent particle trapping mechanisms for mixed gels: Vertex trapping for spatially ordered gels and dense-region trapping for spatially disordered gels. The color of the ber segments indicates whether they are repulsive (red) or attractive (blue).

14

ACS Paragon Plus Environment

Page 14 of 28

Page 15 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

(d)

(e)

Figure 4: (a-c) The average number of bers hNlocal i at a radial distance of less than b/2 from the particle as a function of the interaction potential strength for particles of diameter s/b = 0.5 and a rescaled interaction range of k/s = 0.5 for (a) mixed interactions with U+ = −U− = U0 and ψ = 0.5, (b) purely attractive electrostatic interactions with U0 < 0 and (c) purely repulsive electrostatic interactions with U0 > 0. The data for the mixed case resembles closely the data for the attractive case. (d, e) Particle diusivity as a function of the spatial disorder strength σd for a rescaled interaction range of k/s = 0.5 in purely attractive (U0 /kB T = −10) and purely repulsive (U0 /kB T = 10) gels and mixed gels, U+ = −U− = 10kB T and ψ = 0.5. For (d) we employ a small steric diameter s/b = 0.1, which corresponds to a small interaction range k/b = 0.05 and for (e) a large steric diameter s/b = 0.5, corresponding to an intermediate k/b = 0.25. The eect of spatial disorder on the diusivity is drastically reduced for mixed gels, compared to gels with purely attractive electrostatic interactions.

15

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

next examine the eect of the spatial disorder strength σd on D for simulations with and without interaction disorder. Figures 4d and e show D as a function of σd for an absolute interaction potential strength of |U0 |/kB T = 10 and an interaction range of k/s = 0.5. Figure 4d shows data for a small steric diameter s/b = 0.1 and g. 4e for an intermediate steric diameter s/b = 0.5. For the attractive case, the diusivity varies strongly with respect to σd . This can be explained as follows: For purely attractive bers the vertex trapping mechanism is strong for small interaction potential ranges k  b, when the vertices form strong, localized potential minima. 9 This is reected in g. 4d, where for a small k/b = 0.05, strong vertex trapping leads to small D for σd = 0. For intermediate to large k ∼ b, by contrast, vertex trapping is weak for purely attractive gels, since the interaction potentials of neighboring bers balance each other due to the ordered spatial alignment of the bers. 9 Thus, in g. 4e, an intermediate k/b = 0.25 leads to high diusivities at σd = 0 and U0 /kB T =

−10. Also, in g. 3b for U0 < 0 and the largest steric diameter s/b = 0.5, weak vertex trapping causes signicantly higher diusivities for the purely attractive case, compared to the mixed case. More information on this can be found in 9 and in the supplementary information in g. S1, which shows the particle diusivity as a function of the interaction potential range k/b without steric eects, i.e. for s = 0, in order to illustrate the eect of the interaction range in comparison to the mesh size b. Comparison of the mixed and attractive cases in g. 4d shows that for small interaction ranges k/b vertex trapping for ordered gels (σd = 0) becomes weaker when random interaction sites are introduced, resulting in an increased D. We attribute this to the fact that there are fewer and irregularly spaced strong potential minima at the ber vertices as depicted in g. 2a, top row. For intermediate k/b in g. 4e, on the other hand, vertex trapping becomes much stronger when random interaction sites are introduced, since the irregularly spaced strong potential minima do not balance each other like they do in a purely attractive ordered gel. Thus, introducing interaction disorder via randomly mixed attractive and repulsive interaction sites into the model system mitigates the strong eect of the spatial disorder 16

ACS Paragon Plus Environment

Page 16 of 28

Page 17 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

strength σd on the diusive behavior in gs. 4d and e. In fact, interaction disorder and spatial disorder both randomize the potential landscape in the gel and therefore have a qualitatively similar eect on D. This is underlined in g. S2 in the supplementary information, where we compare the diusivity as function of U0 for simulations with no disorder, only spatial disorder, only interaction disorder and both spatial and interaction disorder.

Mixed gels are strong particle lters.

The concept of interaction ltering describes

the selective immobilization of particles according to their interaction with the bers in biogels like mucus and the ECM. 1 For the case of electrostatic interactions, the potential strength U0 can be interpreted as the product of the particle charge and the linear ber charge density. 9 Thus, according to gs. 3a and b, gels with mixed electrostatically positive and negative interaction sites are very eective lters for charged particles, as they can immobilize strongly charged particles of either sign. By contrast, if the gel bers contain, for example, only negative charges, positively charged particles may be immobilized due to vertex trapping or dense-region trapping (c.f. g. 3c), but negatively charged particles will be very mobile inside the gel, since purely repulsive electrostatic interactions lead to comparably weak particle trapping with relative diusivities of D/D0 > 0.1 compared to

D/D0  0.1 for strongly attractive U0 /kB T = −20 as seen in gs. 3a and b. Furthermore, as shown in gs. 4d and e, particle trapping in mixed gels depends only weakly on spatial disorder. Thus, our simulations illustrate why heterogeneous biogels with mixed interaction sites serve as highly eective and robust lters for interacting particles, whereas purely attractive gels exhibit a strong dependence on the disorder strength σd and thus on the spatial structure. The strong particle trapping capabilities of gels with mixed positive and negative interaction sites are demonstrated experimentally in the next section.

17

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Comparison of experiments and model predictions To test our theoretical model, we perform a quantitative comparison of simulation and experimental data. FCS is used to experimentally measure the diusivity of negatively charged Alexa488 uorescent particles in gels that are comprised of a mixture of positively charged dextran(+) and negatively charged dextran(-) polymers. Translational diusion coecients are obtained for pure dextran(-), pure dextran(+) and mixed dextran solutions for various mixing ratios and dierent polymer mass concentrations. Characteristic normalized uorescence autocorrelation functions in pure and mixed dextran solutions are shown in g. 5a, from which diusivities are calculated as explained in the Experimental Methods Section. In gs. 5 b to d we present the relative diusivity of Alexa488 particles (crosses) for varying mixtures of dextran(+) and dextran(-), dened by the ratio of dextran(-) to the total dextran mass concentration, i.e. Cdex(−) /Ctotal . Experiments shown are performed at total polymer concentrations of Ctotal = 4%w/v, 6%w/v and 8%w/v in gs. 5 b to d, respectively. For Cdex(−) /Ctotal = 0, the electrostatic particle-gel interactions are purely attractive, which leads to strongly reduced diusivities, compared to free diusion of Alexa488. For

Cdex(−) /Ctotal = 1, i.e. purely repulsive electrostatic particle-gel interactions, Alexa488 exhibits much higher diusivities than for pure dextran(+), in accordance to our previously published results which showed that gels with purely attractive electrostatic interactions hinder particle diusion more eectively than gels with purely repulsive electrostatic interactions. 9,10 As shown in gs. 5 b to d , mixed dextran gels with Cdex(−) /Ctotal between 0 and 0.5 show diusivities nearly identical to purely attractive dextran(+) gels, even though the number of attractive interaction sites in mixed dextran gels is signicantly lower than for pure dextran(+). So we observe a highly asymmetric behavior of the diusivity with respect to the symmetric solution with 50% dextran(+) and 50% dextran(-). This is consistent with our theoretical predictions in gs. 3a and b, which showed that mixed gels hinder charged particle diusion practically as eectively as purely attractive gels. Furthermore, even at

Cdex(−) /Ctotal = 0.75 in gs. 5 b to d trapping is still strong, i.e. D is still signicantly smaller 18

ACS Paragon Plus Environment

Page 18 of 28

Page 19 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

than for 100% dextran(-). This shows that even comparably few attractive interaction sites inside the gel can still signicantly hinder the diusion of interacting particles, thus leading to strong interaction ltering for mixed gels. In order to compare the experimental data to our theoretical model, we perform simulations using the following model parameters, dictated by the experiment: The probe particle diameter is p = 1.48nm 10 and the polymer chain diameter is a = 0.74nm 30 which approximates the dextran diameter, hence we obtain a ber/particle diameter ratio a/p = 0.5. The ratio of the potential strengths for the attractive and the repulsive ber segments, U− /U+ , is dictated by the experiment and reects the fact that dextran(+) has approximately one charged amino group per two glucoses, and dextran(-) has approximately two charged carboxyl group per nine glucoses, which leads to a ratio of dextran(+) to dextran(-) line charge densities of (1/2)/(2/9) = 2.25 = U− /U+ . The density of charges on the polymer chains also inuences the background ion concentration CIon of the solution, which determines the interaction range according to eq. (8). CIon increases by about 11 mM and 6 mM upon addition of 1%w/v dextran(+) and dextran(-) to the solution, respectively. As dextran gels are known to be spatially disordered, 31,32 we use a spatially disordered gel in our simulations with a disorder strength of σd /b = 0.9. The magnitude of ψ for the simulations corresponds to the mixing ratio Cdex(−) /Ctotal . We choose the strength of the interaction potential with xed ratio U− /U+ = 2.25 and the mesh size b for optimal agreement between simulation and experiment. The double optimization of the values of b and U+ is explained in the supplementary information. As a result, we obtain U− /kB T = −10.1 and U+ /kB T = 4.5 as well as b = 11, 9.0 and 7.8 nm for Ctotal = 4, 6 and 8%w/v, respectively. The simulation data is presented in gs. 5 b to d as lled circles. For 4%w/v and 6%w/v the simulated diusivities are in close correspondence to the experimentally observed diusivities over a broad range of dextran ratios. For the highest dextran concentration of 8%w/v in g. 5d, our simulations overestimate the diusivities but the simulation and experimental curves are still in qualitative agreement. Assuming completely straight bers, we can calculate b using the formula 19

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Ctotal ≈ (3/ldex b2 )mmon , where ldex = 0.39 nm 33 is the dextran monomer length, 3/ldex b2 the dextran monomer density and mmon = 162 Da the dextran monomer mass. For completely straight dextran bers we obtain the substantially reduced mesh sizes of 7.2, 5.9 and 5.1 nm for Ctotal = 4, 6 and 8%w/v, respectively. This suggests that in experimental dextran gels, the individual dextran polymers are signicantly crumpled, which leads to an increase of the mesh size b. A comparison between simulation and experimental data using the mesh sizes for completely straight dextran bers is shown in g. S4. The agreement between experiment and simulation is still very good, which means that the resultant diusivities in the simulation model only slightly depend on the precise value of b used. Note that we also performed simulations for an alternative mixed gel model where entire bers are randomly assigned to be either purely attractive or purely repulsive, instead of having ber segments of varying charge sign. Qualitatively, the diusive behavior for the alternative mixed gel model is the same as for the model used in this paper, as shown in the supplementary material (gs. S5 and S6).

Conclusions In summary, for particle trapping in gels with mixed attractive and repulsive interaction sites, our simulations and experiments demonstrate that attractive interactions primarily determine the diusive behavior of the particles (c.f. g. 3). As a consequence, the particle trapping mechanisms observed in mixed gels are similar to the trapping mechanisms observed in gels with purely attractive bers (c.f. gs. 2a and 4 a to c). For gels with spatially disordered ber networks, dense-region trapping leads to immobilization of the diusing particle as shown in g. 3c on the right. For spatially ordered gels, particles are trapped at the ber vertices (c.f. g. 3c left). Previously, we have shown that vertex trapping can be ineective for purely attractive electrostatic particle-ber interactions, for intermediate to long interaction potential ranges due to the regular spacing of the potential minima at

20

ACS Paragon Plus Environment

Page 20 of 28

Page 21 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

(a)

Figure 5: (a) Characteristic normalized FCS autocorrelation curves for the diusion of negatively charged Alexa488 NHS ester through a 10 mM MES buer (pH = 6.4) solution of 6 %w/v total dextran concentration at a few dierent dextran(-) to dextran(+) concentration ratios Cdex(−) /Ctotal . Experimental autocorrelations (symbols) are t to eq. (2) (solid lines) in order to determine translational diusion coecients. (b-d) The diusivity for Alexa488 in mixed solutions of oppositely charged dextran(+) and dextran(-) as a function of the dextran(-) to total dextran mass concentration ratio. The total mass concentration of the gel polymers is (b) 4 %w/v, (c) 6 %w/v and (d) 8 %w/v. For the simulations, we use mesh sizes b = 11, 9.0 and 7.8 nm, respectively. We nd quantitative agreement between experiments and simulations for 6 %w/v. For 4 %w/v, the simulations and the experiments agree closely for small to intermediate Cdex(−) /Ctotal concentrations, but some discrepancy is seen for high Cdex(−) /Ctotal concentrations. Qualitative agreement is found for 8 %w/v.

21

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

the vertices which balance each other. 9 This balancing eect for vertex trapping is canceled by introducing random interaction sites into the system (c.f. g. 4e). As a consequence, mixed gels immobilize interacting particles regardless of spatial gel disorder and the sign of the particle charge. Lieleg and coworkers experimentally investigated the diusion of charged nanoparticles in the ECM and found that steric hindrance eects imposed by the detailed structural organization of the ECM play only a minor role for the particle mobility. 12 Furthermore, Lieleg and coworkers found that nanoparticles of either sign are immobilized in the ECM 12 and in mucus 11 due to attractive electrostatic interactions with the gel bers. Our model not only reproduces and conrms these experimental ndings, but also elucidates the details of the microscopic mechanism. We nd that interaction disorder and spatial disorder have a qualitatively similar eect on the diusive behavior of the particle. This is an interesting nding since for some simulations it may be useful to neglect spatial disorder. For example, for simulations with HI, spatial order allows for ecient computation of hydrodynamic interactions between particle and gel bers. 34 If disorder is desired, e.g. to simulate biological systems, one could readily introduce interaction disorder using randomly mixed positive and negative interaction sites. Diusion experiments for negatively charged Alexa488 uorescent particles in mixtures of positively charged DEAE-dextran and negatively charged CM-dextran gels corroborate our theoretical ndings and conrm that gels with mixed attractive and repulsive interactions immobilize interacting particles virtually as eectively as gels with purely attractive electrostatic interactions. Furthermore, we nd that particle diusion is strongly hindered, even if the concentration of attractive polymers inside the gel is signicantly smaller than the concentration of repulsive polymers (c.f. gs. 5 b to d). In other words, even comparably few attractive interaction sites can have a strong trapping eect. For this reason, heterogeneous biopolymer gels consisting of functional groups with dierent biophysical properties are very ecient lters for interacting particles. For gs. 3 and 4 our simulations were performed for neutral gels, i.e. with a balanced 22

ACS Paragon Plus Environment

Page 22 of 28

Page 23 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

number of attractive and repulsive interaction sites (ψ = 0.5) of equal absolute strength. This indicates that a charge neutral polymer gel with both positive and negative functional groups can be used to lter out charged particles of either sign. Accordingly, uncharged particles 12,3538 as well as net neutral particles with a high density of positive and negative charges on the surface 27,3840 have been reported to be highly mobile inside biogels. The latter strategy is employed by some viruses. 3941 Note that in the present work we used the exponentially screened interaction potential eq. (7) to model electrostatic interactions, but our model is in fact more general since an exponentially decaying potential also describes other types of nonsteric interactions such as hydrophobic interactions. Our model assumes a xed spatial arrangement of polymeric interaction sites, a exible hydrogel matrix eectively produces a uctuating mobility along the diusion path, 42,43 which explains the experimentally observed non-Gaussian displacement distributions, 44 simulation studies with exible hydrogel bers are left for the future. The permeability of nanoparticles through biogels such as mucus and the ECM is relevant for many in vivo processes. Designing nanoparticles which can rapidly overcome the diusive barriers poses a signicant challenge for many pharmaceutical and medical purposes. Our model reproduces a number of experimentally observed phenomena for particle trapping in heterogeneous gels and furthermore elucidates the microscopic particle trapping mechanisms. The knowledge of the microscopic particle trapping mechanisms in heterogeneous gels will be useful for the design of advanced drug delivery techniques.

Author Contributions The manuscript was written through contributions of all authors. J.H., J.R.D. and E.B.F. performed research. J.H., J.R.D., R.R.N. and J.E.D. designed research, analyzed data and wrote the paper.

23

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 24 of 28

Acknowledgement This work was supported by the DFG via grant GRK1558. J.E.D. acknowledges nancial support from the National Science Foundation (DBI-1556281). The authors declare no competing nancial interest.

Supporting Information Available The following les are available free of charge. ˆ SI.pdf: Supplementary information with some additional gures and a brief description of the optimization procedure for the comparison of the simulation data to the experimental data.

References (1) Lieleg, O.; Ribbeck, K.

Trends Cell Biol.

2011, 21, 54351.

(2) Cai, L.-H.; Panyukov, S.; Rubinstein, M. Macromolecules

2011, 44, 78537863, PMID:

22058573. (3) Stylianopoulos, T.; Diop-Frimpong, B.; Munn, L. L.; Jain, R. K.

Biophys. J.

2010, 99,

311928. (4) Xu, Q.; Boylan, N. J.; Suk, J. S.; Wang, Y.-Y.; Nance, E. A.; Yang, J.-C.; McDonnell, P. J.; Cone, R. A.; Duh, E. J.; Hanes, J. (5) Käsdorf, B. T.; Arends, F.; Lieleg, O. (6) Rusu, L.; Lumma, D.; Rädler, J. O.

J. Control. Release

Biophys. J.

2013, 167, 7684.

2015, 109, 217181.

Macromol. Biosci.

2010, 10, 146572.

(7) Birjiniuk, A.; Billings, N.; Nance, E.; Hanes, J.; Ribbeck, K.; Doyle, P. S. New

2014, 16, 085014. 24

ACS Paragon Plus Environment

J. Phys.

Page 25 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

(8) Thorne, R. G.; Lakkaraju, A.; Rodriguez-Boulan, E.; Nicholson, C. Sci. U.S.A.

Proc. Natl. Acad.

2008, 105, 841621.

(9) Hansing, J.; Ciemer, C.; Kim, W. K.; Zhang, X.; DeRouchey, J. E.; Netz, R. R. Phys. J. E Soft Matter

Eur.

2016, 39, 53.

(10) Zhang, X.; Hansing, J.; Netz, R. R.; DeRouchey, J. E. (11) Lieleg, O.; Vladescu, I.; Ribbeck, K.

Biophys. J.

(12) Lieleg, O.; Baumgärtel, R.; Bausch, A.

Biophys. J.

2015, 108, 5309.

2010, 98, 17829.

Biophys. J.

2009, 97, 156977.

(13) Laeur, F.; Hintzen, F.; Shahnaz, G.; Rahmat, D.; Leithner, K.; Bernkop-Schnürch, A. Nanomedicine

2014, 9, 387396.

(14) Stylianopoulos, T.; Poh, M.-Z.; Insin, N.; Bawendi, M. G.; Fukumura, D.; Munn, L. L.; Jain, R. K.

Biophys. J.

2010, 99, 13429.

(15) Kirch, J.; Schneider, A.; Abou, B.; Hopf, A.; Schaefer, U. F.; Schneider, M.; Schall, C.; Wagner, C.; Lehr, C.-M.

Proc. Natl. Acad. Sci. U.S.A.

(16) Godec, A.; Bauer, M.; Metzler, R.

New J. Phys.

(17) Gendron, P.-O.; Avaltroni, F.; Wilkinson, K. J. (18) Maiti, S.; Haupts, U.; Webb, W. W.

2012, 109, 1835560.

2014, 16, 092002.

J. Fluoresc.

2008, 18, 1093.

Proc. Natl. Acad. Sci. U.S.A.

1997, 94, 11753

11757. (19) Hess, S. T.; Huang, S.; Heikal, A. A.; Webb, W. W.

Biochemistry

2002, 41, 697705,

PMID: 11790090. (20) Elson, E. L.; Magde, D. (21) Eigen, M.; Rigler, R.

Biopolymers

1974, 13, 127.

Proc. Natl. Acad. Sci. U.S.A.

(22) Zustiak, S. P.; Boukari, H.; Leach, J. B.

1994, 91, 57405747.

Soft Matter

25

2010, 6, 36093618.

ACS Paragon Plus Environment

Nano Letters 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 26 of 28

(23) Modesti, G.; Zimmermann, B.; Börsch, M.; Herrmann, A.; Saalwächter, K. molecules

2009, 42, 46814689.

(24) Cherdhirankorn, T.; Best, A.; Koynov, K.; Peneva, K.; Muellen, K.; Fytas, G. Chem. B

Macro-

J. Phys.

2009, 113, 33559.

(25) Hansing, J.; Netz, R. R. (26) Israelachvili, J.

Biophys. J.

2018, 114, 26532664. ; Intermolecular and Surface

Intermolecular and Surface Forces, 3rd ed.

Forces; Academic Press: New York, 2010. (27) Li, L. D.; Crouzier, T.; Sarkar, A.; Dunphy, L.; Han, J.; Ribbeck, K.

Biophys. J.

2013,

, 13571365.

105

(28) Ghosh, S. K.; Cherstvy, A. G.; Metzler, R.

Phys. Chem. Chem. Phys.

2015, 17, 1847

58. (29) Zhou, H.; Chen, S.

Phys. Rev. E

2009, 79, 021801.

(30) Johansson, L.; Elvingson, C.; Löfroth, J. E.

Macromolecules

1991, 24, 60246029.

(31) Hennink, W.; Talsma, H.; Borchert, J.; De Smedt, S.; Demeester, J. J.

Control. Release

1996, 39, 4755. (32) Amsden, B.

Macromolecules

1998, 31, 83828395.

(33) Guizard, C.; Chanzy, H.; Sarko, A.

Macromolecules

(34) Phillips, R. J.; Deen, W. M.; Brady, J. F.

AIChE J.

1984, 17, 100107. 1989, 35, 17611769.

(35) Lai, S. K.; Suk, J. S.; Pace, A.; Wang, Y.-Y.; Yang, M.; Mert, O.; Chen, J.; Kim, J.; Hanes, J.

Biomaterials

2011, 32, 628590.

(36) Schuster, B. S.; Suk, J. S.; Woodworth, G. F.; Hanes, J. 3446. 26

ACS Paragon Plus Environment

Biomaterials

2013, 34, 3439

Page 27 of 28 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Nano Letters

(37) Wang, Y.; Lai, S.; Suk, J.

Angew. Chem. Int. Ed. Engl.

2008, 97269729.

(38) Abdulkarim, M.; Agulló, N.; Cattoz, B.; Griths, P.; Bernkop-Schnürch, A.; Borros, S. G.; Gumbleton, M.

Eur. J. Pharm. Biopharm.

2015, 97, 230238.

(39) Olmsted, S. S.; Padgett, J. L.; Yudin, A. I.; Whaley, K. J.; Moench, T. R.; Cone, R. A. Biophys. J.

(40) Cone, R.

2001, 81, 19307.

Adv. Drug Deliv. Rev.

(41) Lai, S.; Wang, Y.; Hanes, J.

2009, 61, 7585.

Adv. Drug Deliv. Rev.

(42) Chubynsky, M. V.; Slater, G. W.

Phys. Rev. Lett.

2009, 61, 158171.

2014, 113, 098302.

(43) Chechkin, A. V.; Seno, F.; Metzler, R.; Sokolov, I. M. (44) Wang, B.; Anthony, S. M.; Bae, S. C.; Granick, S.

2017, 7, 021002.

Proc. Natl. Acad. Sci. U.S.A.

, 1516015164.

106

27

Phys. Rev. X

ACS Paragon Plus Environment

2009,

Nano Letters

Graphical TOC Entry competing interaction sites ܷ0 /݇‫ = ܶܤ‬15

ordered ߪ݀ /ܾ = 0

ordered gel:

disordered gel:

repulsive segment attractive segment particle position

disordered ߪ݀ /ܾ = 0.9

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 28 of 28

28

ACS Paragon Plus Environment

purely attractive ܷ0 /݇‫ = ܶܤ‬െ15