Abstract
Biopolymeric matrices can impede transport of nanoparticulates and pathogens by entropic or direct adhesive interactions, or by harnessing “thirdparty” molecular anchors to crosslink nanoparticulates to matrix constituents. The trapping potency of anchors is dictated by association rates and affinities to both nanoparticulates and matrix; the popular dogma is that longlived, highaffinity bonds to both species facilitate optimal trapping. Here we present a contrasting paradigm combining experimental evidence (using IgG antibodies and Matrigel®), a theoretical framework (based on multiple timescale analysis), and computational modeling. Anchors that bind and unbind rapidly from matrix accumulate on nanoparticulates much more quickly than anchors that form highaffinity, longlived bonds with matrix, leading to markedly greater trapping potency of multiple invading species without saturating matrix trapping capacity. Our results provide a blueprint for engineering molecular anchors with finely tuned affinities to effectively enhance the barrier properties of biogels against diverse nanoparticulate species.
Introduction
Biopolymeric matrices are ubiquitous in living systems, generically composed of a highly entangled and crosslinked mesh of macromolecules in buffer. Within cells, cytoskeletal networks of actin and microtubules control cell migration, maintain cell shape and polarity, and facilitate proper routing and sorting of intracellular cargo^{1, 2}. At the extracellular scale, networks of fibronectin, laminin, and collagen not only provide scaffolds for mechanical support and tissue organization but also regulate the dynamic behavior of cells through variations in local microstructure and stiffness^{3, 4}. At the tissue scale, secreted mucins create a viscoelastic gel that serves both as a lubricant and as a transport barrier to prevent pathogens and particulates from reaching the underlying epithelium^{5, 6}.
A major function of biogels is to regulate transport. Gels can in theory impede the passive diffusion of particulates and viruses, as well as active motion of bacteria and cells, by steric obstruction and/or adhesive interactions to the matrix constituents^{7}. Given that the majority of nanoscale species (henceforth referred to as nanoparticulates) are smaller than the mesh spacing of biogels, their diffusion across a gel barrier can only be hindered by adhesive interactions. However, due to evolutionary pressure, it is exceedingly unlikely that direct adhesive interactions with matrices comprised of relatively homogeneous constituents, such as mucins or laminins, can alone effectively block the transport of the full diversity of nanoparticulates typically encountered in nature. For example, viruses must penetrate the dense mucin mesh to infect underlying cells; thus, it is hardly surprising that the vast majority of viruses that transmit at mucosal surfaces (human immunodeficiency virus, herpes, human papillomavirus, Norwalk, etc.) are able to evade binding to mucins and diffuse rapidly through the lowviscosity interstitial fluids within pores of mucus gels^{8}.
An alternative strategy is to utilize “thirdparty” molecular anchors to crosslink nanoparticulates to the matrix, such as antibodies (Abs) that can specifically recognize and bind invading pathogens. The diffusion coefficients of IgG and IgA Abs in human mucus are ~5–10% lower compared to buffer, whereas 10fold larger viruses can diffuse in mucus unhindered^{9}. The slightly retarded diffusion of both Abs implies they must be slowed by weak and transient interactions with the mucus matrix. Surprisingly, despite this seemingly negligible affinity, herpesbinding IgG can specifically and effectively immobilize herpes simplex virus type 1 (HSV1) in human cervicovaginal mucus (CVM) even at subneutralizing IgG concentrations, and trapping HSV1 in mucus directly prevented vaginal herpes transmission in mice^{10}. Although the trapping potency of IgG is naturally affected by its binding and unbinding rates to mucins^{11}, the optimal kinetics remains poorly understood. The widely held and intuitively reasonable assumption is that anchors with longlived, highaffinity bonds to both the nanoparticulate and matrix would confer superior trapping efficiency.
To develop more potent anchors, we seek to examine the characteristics of IgG that could maximize net adhesive interactions between nanoparticulates and biopolymer matrices. We combine theoretical models and experiments to show that anchormatrix bonds that are rapid and shortlived relative to anchornanoparticulate bonds greatly enhance the trapping potency of molecular anchors.
Results
Efficient trapping with transient anchormatrix bonds
The highly viscoelastic nature of physiological mucus gels makes it exceedingly difficult to chemically modify and subsequently remove crosslinkers without irreversibly perturbing its rheological properties. Instead, we took advantage of the thermogelling properties of Matrigel®, which enables us to biotinylate the matrix as a lowviscosity fluid at 4 °C yet study its diffusional barrier properties as a viscoelastic gel at 37 °C. IgG possess only modest affinity to Matrigel®, as reflected by its dissociation constant (K _{D} ~4 × 10^{−8} M; Supplementary Table 1) measured using biolayer interferometry (BLI). This allowed us to investigate, using antipolyethylene glycol (PEG) IgG as molecular anchors, whether the mobility of PEGmodified polystyrene nanoparticles (PSPEG; diameter ~200 nm) that exhibits rapid diffusion in the biotinylated Matrigel® can be altered by tuning the affinity of anchormatrix bonds.
We mixed neutravidin and biotinylated IgG that specifically bind PEG into biotinylated Matrigel® to create highaffinity IgGmatrix bonds prior to temperatureinduced gelation of the matrix; BLI measurements indicate biotinylated IgG exhibited much lower K _{D} (~1 × 10^{−11} M; Supplementary Table 1) to biotinylated Matrigel® treated with neutravidin. When added to biotinylated Matrigel® mixed with neutravidin, either lacking exogenous IgG altogether or treated with control IgG, PSPEG exhibited rapid diffusion, with a geometrically averaged ensemble effective diffusivity (<D _{eff}>; 0.25 μm^{2} s^{−1} at τ = 0.2667 s) only ~3.2fold reduced compared to their theoretical diffusivity in buffer (Fig. 1a, b). The conjugation of biotinylated IgG to Matrigel® did not reduce gel formation or the barrier properties of Matrigel® against ~200 nm uncoated carboxylmodified nanoparticles, which were immobilized to a similar extent as in unmodified Matrigel® (Supplementary Fig. 1). Surprisingly, despite anchoring 10 μg mL^{−1} antiPEG IgG to Matrigel® with longlived, highaffinity biotinneutravidin bonds, the matrix largely failed to immobilize PSPEG. Indeed, the <D _{eff}> (0.14 μm^{2} s^{−1}) of PSPEG at τ = 0.2667 s was not statistically significantly different than in the same Matrigel® without antiPEG IgG, and nearly 70% of particles remained mobile (defined as nanoparticles with <D _{eff}> ≥ 10^{−1.5} μm^{2} s^{−1} at τ = 0.2667 s; Fig. 1a–c). Modest trapping of PSPEG by matrixbound antiPEG IgG was observed only with prolonged incubation, e.g., 24 h (Supplementary Fig. 2). These effects were not due to biotinylation of IgG or the presence of neutravidin with biotinylated IgG; antiPEG IgG with and without biotinylation and/or neutravidin all exhibited similar K _{D} (~5 × 10^{−9} M; Supplementary Table 2) as measured by BLI.
In contrast, despite the modest affinity between individual native unmodified IgG and biotinylated Matrigel®, the addition of 10 μg mL^{−1} of antiPEG IgG in biotinylated Matrigel® reduced the <D _{eff}> of PSPEG by nearly 70fold, with nanoparticles slowed on average 320fold compared to their mobility in water (Fig. 1d). The fraction of mobile PSPEG was reduced from 81 to 14% with the addition of antiPEG IgG (Fig. 1e). The immobilization was not due to agglutination of PSPEG, since trapped nanoparticles appeared identical to nonagglutinated nanoparticles in Matrigel® treated with control IgG. PSPEG were also unlikely to be immobilized due to marked increase in the nanoparticle hydrodynamic diameter; a complete coating of IgG on ~200 nm nanoparticles would add no more than ~10 nm to the hydrodynamic diameter, and larger nanoparticles remained largely diffusive in Matrigel®. The presence of neutravidin was likewise not responsible for this phenomenon; native IgG trapped PSPEG equally well in biotinylated Matrigel® whether treated with neutravidin or not (Supplementary Fig. 3). These results directly demonstrate that shortlived anchormatrix bonds are far more efficient in facilitating immobilization of nanoparticles than longlived anchormatrix bonds. Finally, we found that addition of antiPEG IgG also effectively immobilized smaller ~100 nm PSPEG in Matrigel® (Supplementary Movies 1 and 2), and that addition of similar concentrations of Synagis®, a monoclonal IgG against protein F on respiratory syncytial virus (RSV), was able to effectively immobilize the virions in Matrigel® compared to control IgG (Supplementary Movies 3 and 4). These results further illustrate the broad applicability of shortlived anchormatrix bonds in facilitating trapping of diverse entities in biological matrices.
Proposed theoretical framework and assumptions
Our observations motivated us to develop a model to recapitulate the observations and examine the features of molecular anchors and matrix that could maximize trapping potency of nanoparticulates by the matrix. The model assumes three reactive species: molecular anchors A; nanoparticulates P; and matrix constituents M. Assuming that anchors must simultaneously possess some affinity to both the matrix and the nanoparticulates, our model reveals that the most robust crosslinking of nanoparticulates to the matrix, measured by a minimum effective particle diffusivity D _{eff}, is achieved when the following six conditions are met.
C1. Anchormatrix bindingunbinding kinetics are markedly faster than anchornanoparticulate kinetics. In other words, the steadystate frequency of anchormatrix binding (\(\tau _{{\rm{AM}}}^{  1}\)) is high relative to the steadystate frequency of anchorparticle binding (\(\tau _{{\rm{AP}}}^{  1}\)): \(\tau _{{\rm{AM}}}^{  1}\) ≫ \(\tau _{{\rm{AP}}}^{  1}\), or \({\tau _{{\rm{AM}}}}\) ≪ \({\tau _{{\rm{AP}}}}\).
C2. Nanoparticulates possess multiple (N) independent binding sites such that multiple anchors can simultaneously crosslink the same nanoparticulate to the matrix: N ≫ 1.
C3. Anchors are much smaller than the nanoparticulate, and consequently, the anchor diffusivity (D _{A}) is much larger than the free nanoparticulate diffusivity (D _{P}): D _{A} ≫ D _{P}.
C4. Anchornanoparticulate binding is fast enough (i.e., \({\tau _{{\rm{AP}}}}\) is small enough) that many anchors are likely to bind to the nanoparticulate within the expected diffusive passage time (\({\tau _{L}}\)) of nanoparticulate through the biopolymer matrix of thickness L: \({\tau _{{\rm{AP}}}} < {\tau _{L}}.\)
C5. For [A] ≫ [M], \({\tau _{{\rm{AM}}}}\) is sufficiently short such that anchors do not saturate the binding sites in the matrix.
C6. Anchor concentration [A] is modest, such that on average a single nanoparticulate cannot be simultaneously bound by two anchors that are immobilized to the matrix. In other words, average [A] does not exceed one anchor per unit volume of the nanoparticulate (V _{P}): [A]≪1/V _{P}.
In general, with the proposed components M, A, and P, there are two reaction sequences that form the desired complex (MAP), corresponding to a trapped nanoparticulate. In particular, the MAP complex is formed either by a matrixbound anchor capturing a free nanoparticulate:
or by a nanoparticulateanchor complex (formed when free anchors accumulate on a diffusing nanoparticulate) interacting with and binding to the matrix:
The anchornanoparticulate binding rates for free anchors (k _{on}) and matrixbound anchors (\(k{\prime _{{\rm{on}}}}\)) are given by the Smoluchowski encounter relation^{12}, namely
respectively, where R _{0} is the effective binding distance at which two molecules react. Note that the diffusivity of the polymer matrix is effectively zero, i.e., D _{M} ≈ 0. The fraction of free A at steady state is related to the binding (a _{on}) and unbinding (a _{off}) rates of anchors to the matrix, given by
Note that \(\varphi = 0\) and \(\varphi = 1\) represent extremes where all anchors and no anchors are bound to the matrix, respectively.
How weak and rapid anchormatrix binding maximizes trapping
Instinctively, one may expect that \(\varphi = 0\) maximizes the fraction of trapped nanoparticulates and that trapping potency is reduced as \(\varphi \) rises until it is eliminated altogether when \(\varphi = 1\). Nevertheless, this was not supported by our experiments where IgG, anchored to the matrix with longlived biotinavidin bonds, failed to trap nanoparticulates with the same potency as IgG that exhibit only weak and shortlived interactions with the matrix. To begin to understand why longlived anchormatrix bonds may compromise nanoparticulate trapping, it is important to note that a nanoparticulate is unlikely to simultaneously encounter multiple matrixbound (immobilized) anchors unless the anchor concentration is very high (i.e., anchors are generically spaced at distances much greater than the dimensions of the nanoparticulate). For example, we have previously observed trapping of ~100–200 nm nanoparticles and viruses at IgG concentrations of 1–3 μg mL^{−1} in CVM^{10, 13}; the average distance between each IgG at these concentrations is roughly 440–630 nm. At these concentrations, if anchors are permanently bound to the matrix, the average number of anchors on each 100–200 nm nanoparticle that has been crosslinked to the matrix must be at most one. Conversely, to achieve an average distance of ≤ 100 nm between each IgG would require IgG concentrations in excess of 250 μg mL^{−1}, an exceedingly high concentration for a single anchor species.
Recall from the Smoluchowski encounter relation that when anchors are immobilized, the rate of a nanoparticulate binding to an anchor is proportional to particle diffusivity D _{P}, whereas the binding rate of free anchors to the nanoparticulate is proportional to D _{P} + D _{A}. Since we postulated that D _{A} ≫ D _{P}, nanoparticulates must encounter freely diffusing anchors much more frequently and quickly than matrixbound anchors. Consequently, when \(\varphi >0\), multiple anchors begin to accumulate on the surface of the nanoparticulate, and multiple bonds can form (i.e., \({\rm{P}}{{\rm{A}}_n} \to {\rm{P}}{\left( {{\rm{AM}}} \right)_n}\)) when a freely diffusing nanoparticulateanchor complex encounters matrix constituents. While a single anchor might rapidly unbind from the matrix, resulting in a very short association lifetime of the complex, a nanoparticulateanchor complex with multiple nanoparticulatebound anchors, i.e., PA_{ n }, can increase the collective crosslink lifetime because only one MAP bond is necessary to keep the nanoparticulate immobilized at any given time. So long as the anchors stay bound to the nanoparticulate, they do not diffuse away as quickly after the nanoparticulateanchor complex unbinds from the matrix, as would individual free anchors, and thus can more rapidly rebind to the matrix. Assuming each anchormatrix bond is independent, the PA_{ n } complex crosslink lifetime increases exponentially with the number of anchors n bound to the same nanoparticulate, and it becomes exceedingly rare for all anchors to simultaneously unbind from the matrix. We therefore reach the seemingly counterintuitive conclusion that shortlived anchormatrix bonds can actually facilitate more complete crosslinking of nanoparticulates to the matrix.
Of course, some fraction of A must bind to the polymer network with some probability or frequency; if \(\varphi = 1\), then anchors never bind to the matrix. It follows that the crosslink lifetime of a nanoparticulateanchor complex to the matrix must eventually begin to decrease as the anchormatrix binding affinity is reduced below some optimal fraction of free anchors, \(0 < \varphi < 1\), corresponding to the most robust crosslinking of nanoparticulates to matrix.
We seek to precisely identify the optimal affinity \(\varphi \) and timescale of anchormatrix interactions for minimizing nanoparticulate flux through a gel layer. To do so, we first define a characteristic length scale L of interest in the system (e.g., the height of a mucus layer lining the surface of the lung or gastrointestinal (GI) tract). There are three timescales to consider: diffusion (\({\tau _{L}} = {L^2}/\left( {2{D_{\rm{P}}}} \right)\)); anchormatrix interactions (\({\tau _{{\rm{AM}}}} = 1/\left( {{a_{{\rm{on}}}} + {a_{{\rm{off}}}}} \right)\)); and anchornanoparticulate interactions \(\left( {{\tau _{{\rm{AP}}}} = 1/\left( {{D_{\rm{A}}}\left[ {\rm{A}} \right]{R_0} + {k_{{\rm{off}}}}} \right)} \right)\). The diffusion timescale determines the average amount of time needed to diffuse through a matrix layer. The two kinetic timescales characterize the average duration of consecutive bindunbind events. We assume that L is large enough that many anchormatrix and anchornanoparticulate interactions can occur before the nanoparticulates diffuse out of the system, i.e., \({\tau _{{\rm{AP}}}},{\tau _{{\rm{AM}}}}\) ≪ \({\tau _{L}}\), which allows us to derive an effective diffusivity for the nanoparticulate D _{eff} ≤ D _{P} (see Methods section for the derivation) that characterizes the effective trapping potency of the anchors. The smaller the D _{eff} is the more immobilized the nanoparticulate is: D _{eff} = D _{P} indicates anchors that have no effect on the native diffusivity of the nanoparticulate, whereas D _{eff} < D _{P} reflects anchors that can at least transiently immobilize the nanoparticulate to the matrix.
We next seek to explore how anchormatrix affinity influences nanoparticulate trapping under two important regimes: slow but longlived anchormatrix kinetics, where \({\tau _{{\rm{AM}}}}\) ≫ \({\tau _{{\rm{AP}}}},\) and rapid yet shortlived anchormatrix kinetics, where \({\tau _{{\rm{AM}}}}\) ≪ \({\tau _{{\rm{AP}}}}\). If \({\tau _{{\rm{AM}}}}\) ≫ \({\tau _{{\rm{AP}}}},\) then on the timescale of A/P kinetics, anchors do not bind to or unbind from the matrix. Because we assume that [A] is not unrealistically high (see above and assumption 6), nanoparticulates can effectively only bind to a single matrixbound anchor at a time. In this regime, D _{eff} is minimized when \(\varphi = 0\), i.e., very high anchormatrix affinity (Fig. 2a, \({\tau _{{\rm{AP}}}}/{\tau _{{\rm{AM}}}}\sim 0.01\)). Assuming \({D_{\rm{A}}}/{D_{\rm{P}}} = 20\) and only 15 antigens on each 100 nm nanoparticulate, this results in a \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) that is reduced ~55% on average compared to if anchors have no affinity to matrix (Fig. 2a), which translates to only a ~5% reduction in the fraction of nanoparticulates that can penetrate a 50 μmthick layer over 2 h (Fig. 2b).
A much different result is obtained with rapid and shortlived anchormatrix bonds relative to anchornanoparticulate bonds, i.e., \({\tau _{{\rm{AM}}}}\) ≪ \({\tau _{{\rm{AP}}}}\). \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) for the same nanoparticulate drops significantly as \({\tau _{{\rm{AP}}}}/{\tau _{{\rm{AM}}}}\) increases, with a nontrivial optimal \(\varphi \) that minimizes D _{eff} (Fig. 2a). Indeed, when \({\tau _{{\rm{AP}}}}/{\tau _{{\rm{AM}}}}\) approaches 20 and with the steadystate free fraction of anchors in the ~20–40% range (i.e., φ ~0.2–0.4), \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) is reduced by over 90%, effectively restricting transport of the nanoparticulates. This drop in D _{eff} directly correlates to > 50% reduction in the fraction of nanoparticulates that can penetrate across a 50 μmthick matrix layer over 2 h, a greater than 10fold increase in trapping potency compared to the longlived anchormatrix bond scenario (Fig. 2b). Increases in \({\tau _{{\rm{AP}}}}/{\tau _{{\rm{AM}}}}\) also directly reduce the amount of anchors needed to suppress the flux of nanoparticulates penetrating and exiting the matrix layer (Fig. 2c). These results confirm condition C1 of our proposed model, namely that shortlived anchormatrix bonds (\({\tau _{{\rm{AM}}}}\) ≪ \({\tau _{{\rm{AP}}}}\)) maximize trapping potency.
Other anchor features for maximizing trapping potency
Hypothetically, if there is only one epitope available per nanoparticulate, then at most one anchor can bind to the nanoparticulate (i.e., N = 1). Naturally, in this scenario, \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) would decrease monotonically as \(\varphi \to 0\), since maximum trapping is achieved when the nanoparticulatebound anchor never dissociates from the matrix, as shown in Fig. 3a. In contrast, for \(N >1\), \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) is an exponentially decreasing function of N: the more antigen sites available on a nanoparticulate, the more likely and quickly the nanoparticulate will accumulate anchors on its surface and become trapped in the matrix (Fig. 3a). With even a modest number of anchorbinding sites on each nanoparticulate (N ~20), nanoparticulate D _{eff} can be reduced by over 90% when combined with rapid (i.e., \({\tau _{{\rm{AP}}}}/{\tau _{{\rm{AM}}}} = 20\)) and weak (i.e., φ ~0.2–0.4) anchormatrix interactions. To place this in perspective, influenza and herpes simplex virus have hundreds of hemaglutinin^{14} and gD glycoprotein^{15} epitopes per viral particle, respectively. These results confirm condition C2 of our proposed model.
In addition to the number of binding sites, the rate of anchor accumulation depends on the frequency with which anchors can collide with the nanoparticulate. The latter is in turn proportional to the diffusivity of the anchor as predicted by the Smoluchowski encounter relation. Although greater nanoparticulate diffusivity D _{P} can theoretically increase the encounter and anchor accumulation rate on the nanoparticulate, this also reduces the time \({\tau _{L}}\) available for sufficient quantities of anchor to accumulate on the nanoparticulate before the nanoparticulate diffuses through the barrier fluid. As shown in Fig. 3b, nanoparticulate \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) drops as \({D_{\rm{A}}}\) increases; thus, anchors that are smaller and more mobile than the nanoparticulate are preferred for nanoparticulate trapping. These results confirm condition C3 of our proposed model.
Effect of matrix thickness and anchor density on trapping
The barrier properties of biogels are naturally dependent on both the thickness of the gel layer as well as the concentration of the molecular anchors. This is particularly relevant for diffusional barriers such as mucus and basement membranes, where minimizing the fraction of viruses that can penetrate through the gel layer can directly reduce the probability of transmission or spread of the infection systemically. To address the balance between these two parameters, we assert that the most effective balance of timescales to immobilize nanoparticulates is \({\tau _{{\rm{AM}}}}\) ≪ \({\tau _{{\rm{AP}}}}\) ≪ \({\tau _{L}}\)(i.e., imposing conditions C1 and C4 of our model). We have already shown above that maximal trapping occurs with rapid anchormatrix binding kinetics i.e., \({\tau _{{\rm{AM}}}}\) ≪ \({\tau _{{\rm{AP}}}}\). In addition, as explained above, in order for nanoparticulates to become trapped in the matrix, the nanoparticulate must be captured by at least one anchor before it diffuses through the matrix. Recall that the average time the nanoparticulate, unhindered by anchors, needs to diffuse through a matrix layer is \({\tau _{L}} = {L^2}/(2{D_{\rm{P}}})\). Hence, the matrix layer thickness must be L ≫ \(\sqrt {2{D_{\rm{P}}}{\tau _{\rm{P}}}} \) for anchors to have sufficient time to accumulate on the nanoparticulate. To illustrate the effect of the timescale \({\tau _{D}}\), we use a numerical approximation of the solution to Eq. (15) (see Methods section), and compute the probability that a nanoparticulate can diffuse across a polymer matrix layer of thickness L. We term this the absorption probability; a low absorption probability indicates effective trapping by anchors. Not surprisingly, \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) (Fig. 4a) and the absorption probability (Fig. 4b) both decrease with increasing anchor concentration (Fig. 4a), in both cases approaching a minimum \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\)when \(\varphi \) ~0.35. Interestingly, when we compare the relative importance of L vs. anchor concentrations, we found that exponentially higher anchor concentrations are required when L is smaller than ~40–50 μm thick in order to maintain a comparably effective diffusional barrier (Fig. 4c), implying that effective diffusional barriers in vivo should be at least 40–50 μm thick. These estimates agree remarkably well with both (i) the thickness of mucus coatings lining the respiratory, GI, and cervicovaginal tracts (typically ~50–100+ μm), as well as (ii) the Ab concentrations present in mucus (typically ~0.1–10 μg mL^{−1})^{10}. This analysis confirms condition C4 of our proposed model.
Robust trapping of multiple nanoparticulate species
To selectively control transport against multiple species of nanoparticulates in the same polymeric matrix, such as trapping a diverse array of pathogens that impinge on mucus coating the airways and GI epithelium, many corresponding anchor species must coexist without impeding each other’s trapping potency. In other words, even when anchors that bind any given species represent only a tiny fraction of all anchors present, the specific anchormatrix affinity must remain unaltered in order to maintain comparable trapping potency. We introduce the term “trapping robustness” to describe the ability to immobilize multiple nanoparticulate species.
Since the concentration of matrix constituents is finite, the number of anchorbinding sites in a polymeric gel must by definition be finite. Thus, if anchormatrix bonds are longlived, a matrixbound anchor prevents other anchors from binding to the same binding site on the matrix. Thus, at concentrations sufficient for trapping (e.g., ~1–5 μg mL^{−1} IgG), the system could accommodate only a relatively limited number of anchor species (< 10^{3} for a 2% w/v gel, assuming 10 anchorbinding sites per matrix molecule and an average molecular weight of 500 kDa) before additional anchors become unable to effectively reduce \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) (Fig. 5a). However, when \({\tau _{{\rm{AM}}}}\) ≪ \({\tau _{{\rm{AP}}}}\), the short duration of anchormatrix bonds would greatly increase the number of unoccupied anchorbinding sites available on the matrix at any moment in time. This in turn enables the anchormatrix system to both immobilize a far greater number of particle species simultaneously as well as reduce the minimum \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\)that could be achieved with each anchor (Fig. 5b, c). Indeed, when \({\tau _{{\rm{AP}}}}/{\tau _{{\rm{AM}}}} \ge 20\), a biogel reinforced with appropriate molecular anchors can effectively immobilize at least 30fold more (i.e., ~3 × 10^{4}) distinct nanoparticulate species without appreciable loss in trapping potency (i.e., similar minimum \({D_{{\rm{eff}}}}/{D_{\rm{P}}}\) for all species), underscoring the potential trapping robustness of the system. Altogether, these results confirm condition C5 of our proposed model.
Finally, we examined the relative importance for each of the parameters described above by evaluating the partial derivatives of \({\rm{log}}({D_{{\rm{eff}}}}/{D_{\rm{P}}})\) under a range of parameter values. We observed the greatest impact with variations with \({\tau _{{\rm{AP}}}}/{\tau _{{\rm{AM}}}}\) (in particular at low \({\tau _{{\rm{AP}}}}/{\tau _{{\rm{AM}}}}\) values) and with the rates of anchor accumulation \({k_{{\rm{on}}}}[{\rm{A}}]\), with more modest impact with changes in antigenic epitope density (i.e., maximum number of bound anchor) \(N\) and the diffusivity of anchors relative to the nanoparticulate species \({D_{\rm{A}}}/{D_{\rm{P}}}\) (Fig. 6). These results underscore shortlived anchormatrix bonds relative to anchornanoparticulate bonds as a crucial separation of timescales for enabling molecular anchors that can substantially enhance the barrier properties of either biological or synthetic polymeric matrices to multiple nanoparticulate species.
Discussion
A critical function of polymeric matrices in biological systems is to exert selective control over the transport of thousands of nanoparticulate species. By eliminating the need for matrix constituents to directly recognize diverse antigenic species, an anchormatrix system can enable an effective diffusional barrier against many nanoparticulate species while maintaining relatively static biochemistry and microstructure of the matrix. This suggests that anchors, such as IgG and other Abs produced by the immune system that can adapt and bind diverse molecular entities, represent an ideal platform to control nanoparticulate transport. Here we demonstrate both experimentally and theoretically that shortlived anchormatrix interactions convey the key attribute enabling potent and robust control over nanoparticulate transport in biogels (schematic in Fig. 7). Combined with our earlier observations that IgG can potently immobilize viruses and nanoparticles in different mucus secretions, it is likely that the proposed strategy, whereby the barrier properties are tuned by modest concentrations of highly mobile molecular anchors with exceedingly short anchormatrix bond times relative to anchornanoparticulate bond times, is a universal feature of biogels in living systems. Our findings provide, for the first time, a blueprint for engineering of molecular anchors with optimal shortlived anchormatrix and anchornanoparticulate bonds to selectively and potently tune the barrier properties of polymeric gels. These insights will help guide the development of methods to reinforce the natural biological barriers against pathogens, such as the mucus barrier against sexually or respiratory transmitted infections.
Methods
Preparation of PEGcoated nanoparticles
To produce PEGylated nanoparticles (PSPEG), we covalently modified 200 nm fluorescent, carboxylmodified polystyrene beads (PSCOOH; Invitrogen) with 2 kDa methoxy poly(ethylene glycol) amine (PEG; Sigma) via a carboxylamine reaction, as published previously^{16, 17}. Particle size and ζpotential were determined by dynamic light scattering and laser Doppler anemometry, respectively, using a Zetasizer Nano ZS (Malvern Instruments, Southborough, MA, USA). Size measurements were performed at 25 °C at a scattering angle of 90°. Samples were diluted in 10 mM NaCl solution, and measurements were performed according to the instrument instructions. Highdensity PEGylation (> 1 PEG per nm^{2}) was verified using the fluorogenic compound 1pyrenyldiazomethane to quantify residual unmodified carboxyl groups on the polystyrene beads^{17}. PEG conjugation was also confirmed by a nearneutral ζpotential (Supplementary Table 1)^{16}.
Preparation of biotinylated Matrigel®
Growth factorreduced Matrigel® (Corning) was dialyzed against phosphatebuffered saline (PBS) for a minimum of 24 h at 4 °C, then biotinylated with 20fold molar concentration NHSPEG4biotin (Thermo Fisher), which was again dialyzed against PBS for a minimum of 24 h at 4 °C. This biotinylated Matrigel® (final concentration 2.2 mg mL^{−1}) was mixed with neutravidin (Thermo Fisher; final concentration 0 or 4 μg mL^{−1}), bovine serum albumin (Sigma, final concentration 1 mg mL^{−1}), and Eagle’s minimum essential medium (Lonza BioWhittacker), on ice for 15 min. Fluorescent PSCOOH (Ex: 505 nm, Em: 515 nm; final concentration 4.5 × 10^{8} beads per mL) and PSPEG (Ex: 625 nm, Em: 645 nm; final concentration 4.3 × 10^{8} beads per mL) nanoparticles and antiPEG IgG_{1} (CH2076 or CH2076B, Silver Lake Research, final concentration 10 μg mL^{−1}) was combined on ice. The mixture was added to a custommade microvolume glass chamber slide, incubated at 37 °C for 45 min in a custom hydration chamber, then sealed and incubated for another 30 min prior to microscopy. Trapping of PSCOOH beads in Matrigel® was used as an internal control in all microscopy experiments as a measure of complete polymerization of Matrigel® constituents.
Preparation of fluorescently labeled RSV
RSV was fluorescently labeled with AlexaFluor 555 via a Nhydroxysuccinimide ester reaction. Briefly, 200 μL of RSV (2 mg mL^{−1}) were diluted in 20 μL of 1 M bicarbonate buffer, then added to 20 μL of AlexaFluor 555 NHS Ester (1 mg mL^{−1}). The reaction was incubated in the dark with gentle rocking for 2 h, dialyzed (molecular weight cutoff 100 kDa) against 1 × PBS at 4 °C to remove unbound fluorophores, and stored at − 80 °C^{18}. The Pierce BCA Protein Assay Kit (Thermo Scientific) was used to assess RSV concentration post labeling.
To verify that labeling did not significantly affect Ab binding, an enzymelinked immunosorbent assay was run with IgG that bind to different RSV epitopes. Highaffinity halfarea 96well Costar plates (Corning) were coated with 50 μL of unlabeled or labeled RSV (10 μg mL^{−1}), then incubated overnight at 4 °C. After blocking plates with 5% nonfat milk in PBS, 50 μL of antiRSV IgG (Synagis (MedImmune NDC 6057441141), MAB8599 (EMD Millipore ca. no. MAB8599), and MAB8582 (EMD Millipore ca. no. MAB8582)) at various concentrations in 1% milk were added to the corresponding wells. Antibodies bound to virus were detected with a horseradish peroxidase (HRP)conjugated IgG diluted 1:10,000 in 1% milk (F(ab')2 antimouse IgG Fc (Goat)HRP conjugate (Santa Cruz Biotechnology cat. no. SE2005) and F(ab')2 antihuman IgG Fc (Goat)HRP conjugate (Rockland cat. no. 7091317)). A volume of 50 μL 1Step Ultra TMB (ThermoFisher) was the HRP substrate, then the reaction was quenched with 50 μL of 2 N sulfuric acid. Absorbance at 450 nm was measured using a Spectramax M2 plate reader (Molecular Devices). All wash and incubation steps were performed using PBS with 0.05% Tween.
Highresolution multiple particle tracking
The trajectories of the fluorescent particles were recorded using an electronmultiplying chargecoupled device (EMCCD) camera (Evolve 512; Photometrics, Tucson, AZ) mounted on an inverted epifluorescence microscope (AxioObserver D1; Zeiss, Thornwood, NY, USA), equipped with an Alpha PlanApo × 100/1.46 numerical aperture objective, environmental (temperature and CO_{2}) control chamber, and an LED light source (Lumencor Light Engine DAPI/GFP/543/623/690). Twentysecond videos (512 × 512, 16bit image depth) were captured with MetaMorph imaging software (Molecular Devices, Sunnyvale, CA, USA) at a temporal resolution of 66.7 ms and spatial resolution of 10 nm (nominal pixel resolution 0.156 µm pixel^{−1}). The tracking resolution was determined by tracking the displacements of particles immobilized with a strong adhesive, following a previously described method^{19}. Particle trajectories were analyzed using MATLAB software as described previously^{20}. Subpixel tracking resolution was achieved by determining the precise location of the particle centroid by lightintensityweighted averaging of neighboring pixels. Trajectories of n ≥ 40 particles per frame on average (corresponding to n ≥ 80 total traces per specimen per condition) were analyzed for each experiment, and 3–4 independent experiments were performed for each condition. Under the assumption that particle paths are samples of a stationary stochastic process, mean squared displacement (MSD) can be computed by time averaging, i.e., \(\langle\Delta {r^2}\left( {n\Delta t} \right)\rangle = \frac{1}{{N  n}}\mathop {\sum}\limits_{m = 1}^{N  n} {\left\{ {{{\left[ {x\left( {\left( {m + n} \right)\Delta t} \right)  x\left( {m\Delta t} \right)} \right]}^2} + {{\left[ {y\left( {\left( {m + n} \right)\Delta t} \right)  y(m\Delta t)} \right]}^2}} \right\}} \), (where τ = timescale or time lag Δt is the time between video frames), and τ = nΔt). MSDs from all particle paths within a given video were then ensemble averaged to obtain 〈\({\rm{MSD}}\)〉. Distributions of MSDs and effective diffusivities (D _{eff}) were calculated as previously demonstrated^{16}. MSD may also be expressed as MSD = 4D _{0} τ ^{α}, where α, the slope of the curve on a loglog scale, is a measure of the extent of impediment to particle diffusion (α = 1 for pure unobstructed Brownian diffusion; α < 1 indicates subdiffusive motion due to interactions with the elastic as well as viscous properties of the polymeric gel). Mobile particles were defined as those with D _{eff} ≥ 10^{−1.5} µm^{2} s^{−1} at τ = 0.2667 s (this τ corresponds to a minimum trajectory length of five frames), based on multiple data sets of mobile and immobile nanoparticles (e.g., PS and PSPEG nanoparticles) in Matrigel®^{16, 21}.
BLI experiments
Similar to published protocols^{22}, on an Octet QK instrument (ForteBio), streptavidin biosensors (ForteBio) were loaded with ligand and blocked with free biotin. To measure Ab affinity with Matrigel®, sensors were loaded with biotinylated Matrigel®, and to measure Ab affinity with antigen, 10 kDa biotinPEGNH_{2} was loaded onto sensors. Abs (biotinylated and native) at different concentrations were associated with these customized biosensors and dissociated into running buffer. Data were adjusted for reference sensors and baseline values and aligned to dissociation, then processed with Savitzky–Golay filtering. Analysis was performed with ForteBio software using a 1:1 global curve fit model to obtain values for k _{on}, k _{off}, and K _{D}.
Mathematical model
Instead of developing a reactiondiffusion model for a concentration of nanoparticulate species \({\rm{P}}\), we take a stochastic approach and focus on the motion of individual nanoparticulate \({\rm{P}}\). The goal is to maximize the fraction of time that \({\rm{P}}\) spends bound to the polymer network.
Let \(N\) be the total number of anchorbinding sites on the nanoparticulate P. The crosslink enhancement effect requires the cooperative action of multiple anchors; it requires N ≫ 1 anchorbinding sites (e.g., antigenic epitopes) on the nanoparticulate. The reaction Eqs. (1) and (2) describe the case of a single binding site (i.e., N = 1). The first crosslink bond forms at a diffusionlimited reaction rate according to Eq. (2). If N > 1, additional anchors might be bound to the nanoparticulate. Once the first crosslink forms many additional binding sites M are very close by, allowing additional anchormatrix bonds to form. The intracomplex reaction is given by
where C is a nondimensional parameter that scales the intracomplex binding rate and assumed to be 1 in the current work. Molecules within a large complex may not react with each other at the same rate as they do when they are freely diffusing. When nanoparticulates, anchors, and matrix elements are bound within the same complex, they are mechanically linked. Mechanical forces imposed by surrounding elements of the complex confine random molecular motion^{23}. Similar biomechanical reactions are common in biology (e.g., molecular motor transport^{24} and DNA transcription^{25, 26}). For our present situation, it is a reasonable first approximation to assume that the intracomplex dissociation rates remain the same as the bimolecular dissociation rates (i.e., a _{off} and k _{off}). However, intracomplex binding rates are different from the Smoluchowski bimolecular reaction rates for diffusing molecules. The binding rate between two molecules within the complex depends on their relative distance and effective random mobility. Molecules within a single nanocomplex are quite close so that they do not have to move far in order to bind. On the other hand, they have lower relative mobility when mechanically confined within the complex.
Let n be the number of occupied binding sites, and let s be the number of anchors crosslinking the nanoparticulate to the polymer network. The chemical system can be modeled as a Markov process with state transitions given by
where
The process is described by its probability density \(p(n,s,x,t)\). The probability that at time t, the nanoparticulate is bound to n anchors, s of which are bound to elements of the polymer matrix, and located within a small distance dx of position x is
Since the model is a continuous time Markov process, the probability density function satisfies the differential Chapman–Kolmogorov equation (see ref. ^{27} for details of the derivation),
where \({{\Bbb M}_s}\) is the N × N transition rate matrix for Eq. (6) and \({{\Bbb V}_{n,s}}\) is the N × N transition rate matrix for Eq. (5).
Multiple timescale analysis
Consider the case where \({\tau _{\rm{M}}}\)≪\({\tau _{\rm{P}}}\). Notice that the fast reaction Eq. (6) conserves \(n\). While \(n\) changes slowly, transitions in \(s\) are at quasi steady state. Since we are primarily concerned with the motion of the nanoparticulate, and not necessarily the state of any bound anchors, our goal is to obtain a good approximation to the marginal probability
Using the rule of conditional probability, we can rewrite the full probability density for the process as
Since s changes rapidly compared to n, which changes rapidly compared to x, the conditional probabilities \({\rho _t}\) rapidly equilibrate, which means that \({\rho _t} \approx {\rho _\infty } = \rho \). Since the transition rates are independent of position x, it follows that \(\rho \left( {s\left {n,x} \right.} \right) = \rho \left( {s\left n \right.} \right)\) and \(\rho \left( {n\left x \right.} \right) = \rho \left( n \right)\). The two probability distributions \(\rho \left( {s\left n \right.} \right)\) and \(\rho (n)\) are called quasisteadystate distributions, and they satisfy
We can take advantage of the separation of timescales with an asymptotic approximation, namely
First we average out the fastest reaction, the transition in s. Let \(\bar p\left( {n,x,t} \right) = \rho \left( n \right)u\left( {x,t} \right).\) Substituting Eq. (14) into Eq. (9), summing over \(s\), and using Eq. (13) yields
where \({\bar {\Bbb V}_n}\) is the transition rate matrix for the averaged slow reaction:
where \(\kappa \left( n \right) = \rho \left( {0\left n \right.} \right){k_{{\rm{on}}}} + k_{{\rm{on}}}^\prime \). Given n, the stationary distribution for the number of anchors s < n on the nanoparticulate that are bound to the matrix is
where \(\alpha = {a_{{\rm{off}}}}/\left( {C{a_{{\rm{on}}}} + {a_{{\rm{off}}}}} \right)\).
We can apply the same procedure to average out n as follows. The quasisteadystate distribution for n is given by
where \({\cal N}\) is a normalization factor. Substituting \(\bar p\left( {n,x,t} \right) = \rho \left( n \right)u\left( {x,t} \right)\) into Eq. (15), summing over n, and using Eq. (13) yields
MonteCarlo simulations
To determine the accuracy of the above approximation, we use MonteCarlo simulations. Using the Gillespie algorithm^{28}, we simulate the Markov chain Eqs. (4) and (5) (which is independent of x). A single realization is generated through m state transitions. The total elapsed time t _{ m } and the total time spent with s = 0 (the free diffusing state) \(t_m^{(0)}\) are updated with each transition. It is easy to show that
Because all increments from free diffusion are independent, an estimator for the effective diffusivity is
Saturated regime
The reaction rate for any individual free anchor is substantially reduced by the saturation of matrixbinding sites. Because \([{{\rm{A}}_{\rm{T}}}]\) ≫ \([{\rm{M}}]\), the fraction of unoccupied matrixbinding sites \(\xi \) is equivalent to the fraction of time an individual matrixbinding site is unoccupied. Hence,
It follows that the binding rate for an individual freely diffusing anchor is
Similarly, the intracomplex binding rate (see Eq. (3)) is \({a"\!\!\!}_{{\rm{on}}} = C{a_{{\rm{off}}}}\left[ {\rm{M}} \right]/\left[ {{{\rm{A}}_{\rm{T}}}} \right]\). Based on the modified binding rate (Eq. (11)), the anchormatrix kinetic timescale becomes
The effective diffusivity in the saturated regime is obtained by substituting \(\left[ {\rm{A}} \right] = \left[ {{{\rm{A}}_{\rm{*}}}} \right]\), \(\varphi \approx 1  \frac{{\left[ {\rm{M}} \right]}}{{\left[ {{{\rm{A}}_{\rm{T}}}} \right]}}\), and \(\alpha \approx 1  C\frac{{\left[ {\rm{M}} \right]}}{{\left[ {{{\rm{A}}_{\rm{T}}}} \right]}}\) into Eq. (19).
Statistics
MSD data were logtransformed and compared within groups using a repeatedmeasures twoway analysis of variance (ANOVA) and post hoc Šidák test. Logtransformed average D _{eff} and nontransformed % mobile were compared with oneway ANOVA and subsequent Tukey’s honest significant difference tests. In all analyses, global α = 0.05. Error bars and ± represent SEM.
Code availability
Modeling results and simulations were performed using Python and C. Plotting was done using Python libraries: Jupyter, Scipy, and Matplotlib. MonteCarlo simulations were done in C using the GNU Scientific Library for random number generators. Particle trajectories were analyzed using a MATLAB version of opensource particletracking code, originally developed in IDL by Crocker and Hoffman^{29}. We have adapted this code to analyze particle trajectories on a “framebyframe” basis, as described previously^{20}.
Data availability
The data sets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
References
 1.
Huber, F. et al. Emergent complexity of the cytoskeleton: from single filaments to tissue. Adv. Phys. 62, 1–112 (2013).
 2.
Dinh, A.T., Theofanous, T. & Mitragotri, S. A model for intracellular trafficking of adenoviral vectors. Biophys. J. 89, 1574–1588 (2005).
 3.
Theocharis, A. D., Skandalis, S. S., Gialeli, C. & Karamanos, N. K. Extracellular matrix structure. Adv. Drug Del. Rev. 97, 4–27 (2016).
 4.
Jones, G. W. & Chapman, S. J. Modeling growth in biological materials. SIAM Rev. 54, 52–118 (2012).
 5.
Lai, S. K., Wang, Y.Y., Wirtz, D. & Hanes, J. Micro and macrorheology of mucus. Adv. Drug Del. Rev. 61, 86–100 (2009).
 6.
Katz, D. F. & Dunmire, E. N. Cervical mucus: problems and opportunities for drug delivery via the vagina and cerivc. Adv. Drug Del. Rev. 11, 385–401 (1993).
 7.
Lieleg, O. & Ribbeck, K. Biological hydrogels as selective diffusion barriers. Trends Cell Biol. 21, 543–551 (2011).
 8.
Olmsted, S. S. et al. Diffusion of macromolecules and viruslike particles in human cervical mucus. Biophys. J. 81, 1930–1937 (2001).
 9.
Saltzman, W. M., Radomsky, M. L., Whaley, K. J. & Cone, R. A. Antibody diffusion in human cervical mucus. Biophys. J. 66, 506–515 (1994).
 10.
Wang, Y.Y. et al. IgG in cervicovaginal mucus traps HSV and prevents vaginal herpes infections. Mucosal Immunol. 7, 1036–1044 (2014).
 11.
Wessler, T. et al. Using computational modeling to optimize the design of antibodies that trap viruses in mucus. ACS Infect. Dis. 2, 82–92 (2016).
 12.
Smoluchowski, M. Drei vorträge über diffusion, brownsche molekularbewegung und koagulation von kolloidteilchen. Physik Z 17, 557–585 (1916).
 13.
Henry, C. E. et al. AntiPEG antibodies alter the mobility and biodistribution of densely PEGylated nanoparticles in mucus. Acta Biomater. 43, 61–70 (2016).
 14.
Wilson, I. A. & Cox, N. J. Structural basis of immune recognition of influenza virus hemagglutinin. Annu. Rev. Immunol. 8, 737–771 (1990).
 15.
Clarke, R. W., Drews, A., Browne, H. & Klenerman, D. A single gD glycoprotein can mediate infection by herpes simplex virus. J. Am. Chem. Soc. 135, 11175–11180 (2013).
 16.
Lai, S. K. et al. Rapid transport of large polymeric nanoparticles in fresh undiluted human mucus. Proc. Natl Acad. Sci. USA 104, 1482–1487 (2007).
 17.
Yang, Q. et al. Evading immune cell uptake and clearance requires PEG grafting at densities substantially exceeding the minimum for brush conformation. Mol. Pharm. 11, 1250–1258 (2014).
 18.
Wang, Y.Y. et al. Influenzabinding antibodies immobilise influenza viruses in fresh human airway mucus. Eur. Respir. J. 49, 1601709 (2017).
 19.
Apgar, J. et al. Multipleparticle tracking measurements of heterogeneities in solutions of actin filaments and actin bundles. Biophys. J. 79, 1095–1106 (2000).
 20.
Wang, Y.Y., Nunn, K. L., Harit, D., McKinley, S. A. & Lai, S. K. Minimizing biases associated with tracking analysis of submicron particles in heterogeneous biological fluids. J. Control. Release 220, 37–43 (2015).
 21.
Lai, S. K., Wang, Y.Y., Hida, K., Cone, R. & Hanes, J. Nanoparticles reveal that human cervicovaginal mucus is riddled with pores larger than viruses. Proc. Natl Acad. Sci. USA 107, 598–603 (2010).
 22.
Kumaraswamy, S. & Tobias, R. in Methods in Molecular Biology Vol. 1278 (eds C. L. Meyerkord & H. Fu) Ch. 10, 165–182 (Springer Science+Business Media, 2015).
 23.
Holcman, D. & Schuss, Z. Time scale of diffusion in molecular and cellular biology. J. Phys. A.: Math. Theor. 47, 173001 (2014).
 24.
Goychuk, I., Kharchenko, V. O. & Metzler, R. How molecular motors work in the crowded environment of living cells: coexistence and efficiency of normal and anomalous transport. PLoS ONE 9, e91700 (2014).
 25.
Matsuda, H., Putzel, G. G., Backman, V. & Szleifer, I. Macromolecular crowding as a regulator of gene transcription. Biophys. J. 106, 1801–1810 (2014).
 26.
D’Orsogna, M. R. & Chou, T. Optimal cytoplasmic transport in viral infections. PLoS ONE 4, e8165 (2009).
 27.
Gardiner, C. W. Handbook of Stochastic Methods for Physics, Chemistry, and the Natural Sciences (Springer, 1985).
 28.
Gillepsie, D. T. Exact stochastic simulation of coupled chemical reactions. J. Phys. Chem. 81, 2340–2361 (1977).
 29.
Crocker, J. C. & Hoffman, B. D. Multiple particle tracking and twopoint microrheology in cells. Methods Cell. Biol. 83, 141–178 (2007).
Acknowledgements
This work was supported by Eshelman Institute for Innovation, National Institutes of Health (http://www.nih.gov/) Grants R21EB017938 (S.K.L.); the National Science Foundation (http://www.nsf.gov/) DMS1412844 (M.G.F.), DMS1462992 (M.G.F.), and CAREER Award DMR1151477 (S.K.L.); the David and Lucile Packard Foundation (http://www.packard.org/) (201339274; S.K.L.); startup funds from the Eshelman School of Pharmacy (https://pharmacy.unc.edu/) and Lineberger Comprehensive Cancer Center at the University of North Carolina—Chapel Hill (https://unclineberger.org/) (S.K.L.); and PhRMA Foundation Predoctoral Fellowship (J.L.S.). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Finally, we thank Richard Cone for insightful discussions.
Author information
Affiliations
Contributions
J.N., M.G.F., and S.K.L. conceived of and developed the theoretical framework; J.L.S. and S.K.L. designed the experiments and analyzed the data; J.L.S. performed the experiments; J.E. prepared and verified the RSV; J.N. and T.W. performed computational modeling/simulations; J.N., J.L.S., M.G.F., and S.K.L. wrote the paper.
Corresponding authors
Ethics declarations
Competing interests
S.K.L. is the founder of, serves on the board of directors for, and maintains a financial interest in Mucommune, which is actively seeking to commercialize “mucotrapping” antibody technology. The terms of these arrangements are being managed by The University of North Carolina in accordance with its conflict of interest policies. The remaining authors declare no competing financial interests.
Additional information
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Electronic supplementary material
41467_2017_739_MOESM4_ESM.avi
Supplementary Movie 1
41467_2017_739_MOESM5_ESM.avi
Supplementary Movie 2
41467_2017_739_MOESM6_ESM.avi
Supplementary Movie 3
41467_2017_739_MOESM7_ESM.avi
Supplementary Movie 4
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Newby, J., Schiller, J.L., Wessler, T. et al. A blueprint for robust crosslinking of mobile species in biogels with weakly adhesive molecular anchors. Nat Commun 8, 833 (2017). https://doi.org/10.1038/s41467017007396
Received:
Accepted:
Published:
Further reading

The generalized Langevin equation with powerlaw memory in a nonlinear potential well
Nonlinearity (2020)

Tuning Barrier Properties of Biological Hydrogels
ACS Applied Bio Materials (2020)

Engineering monoclonal antibodybased contraception and multipurpose prevention technologies†
Biology of Reproduction (2020)

Antibodymediated trapping in biological hydrogels is governed by sugarsugar hydrogen bonds
Acta Biomaterialia (2020)

Nanoparticle Loading of Unentangled Polymers Induces EntanglementLike Relaxation Modes and a Broad Sol–Gel Transition
The Journal of Physical Chemistry Letters (2019)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.