Specific recruitment of circulating angiogenic cells using biomaterials as filters

Abstract

Endogenous recruitment of circulating angiogenic cells (CACs) is an emerging strategy to induce angiogenesis within a defect site, and multiple recent strategies have deployed soluble protein releasing biomaterials for this purpose. However, the way in which the design of biomaterials affects CAC recruitment and invasion are poorly understood. Here we used an enhanced-throughput cell invasion assay to systematically examine the effects of biomaterial design on CAC recruitment. The screens co-optimized hydrogel presentation of a stromal-derived factor-1α (SDF-1α) gradient, hydrogel degradability, and hydrogel stiffness for maximal CAC invasion. We also examined the specificity of this invasion by assessing dermal fibroblast, mesenchymal stem cell, and lymphocyte invasion individually and in co-culture with CACs to identify hydrogels specific to CAC invasion. These screens suggested a subset of MMP-degradable hydrogels presenting a specific range of SDF-1α gradient slopes that induced specific invasion of CACs, and we posit that the design parameters of this subset of hydrogels may serve as instructive templates for the future design of biomaterials to specifically recruit CACs. We also posit that this design concept may be applied more broadly in that it may be possible to utilize these specific subsets of biomaterials as “filters” to control which types of cell populations invade into and populate the biomaterial.

Statement of Significance

The recruitment of specific cell types for cell-based therapies in vivo is of great interest to the regenerative medicine community. Circulating angiogenic cells (CACs), CD133+ cells derived from the blood stream, are of particular interest for induction of angiogenesis in ischemic tissues, and recent studies utilizing soluble-factor releasing biomaterials to recruit these cells in vivo show great promise. However, these studies are largely “proof of concept” and are not systematic in nature. Thus, little is currently known about how biomaterial design affects the recruitment of CACs.

In the present work, we use a high throughput cell invasion screening platform to systematically examine the effects of biomaterial design on circulating angiogenic cell (CAC) recruitment, and we successfully screened 263 conditions at 3 replicates each. Our results identify a particular subset of conditions that robustly recruit CACs. Additionally, we found that these conditions also specifically recruited CACs and excluded the other tested cells types of dermal fibroblasts, mesenchymal stem cells, and lymphocytes. This suggests an intriguing new role for biomaterials as “filters” to control the types of cells that invade and populate that biomaterial.

1. Introduction

The formation of functional vasculature is a key step during tissue regeneration [1]. Delivery of pro-angiogenic soluble factors, such as vascular endothelial growth factor (VEGF), is a widely explored approach to induce vascularization and promote tissue regeneration, but soluble VEGF delivery and similar approaches show little clinical benefit [1–3]. Thus, alternative approaches to induce tissue vascularization are needed to improve clinical tissue regeneration, and one alternative approaches is the use of pro-angiogenic cell types to stimulate angiogenesis.

This strategy potentially circumvents limitations associated with the delivery of a single soluble factor, because the delivered cells can secrete a wide range of the appropriate soluble factors and can respond to changes within the diseased tissue [4]. One particular pro-angiogenic cell type, blood derived CD133+ cells also known as circulating angiogenic cells (CACs), have been explored for use in such strategies because of their well-characterized roles in vascular biology including blood vessel maintenance, formation of new blood vessels, and the repair of damaged blood vessels [5–9]. Further, recent studies suggest that CACs more effectively promote angiogenesis than other commonly utilized cell types, such as mesenchymal stem cells [10,11].

CACs reside in the bone marrow with only a small number in circulation, but upon injury, the number in circulation increases due to systemic cues originating from the wound site [5–9]. Once at the wound site, CACs mediate angiogenesis [5–9], likely via paracrine mechanisms [5,6,8,9,12]. Strategies that augment CAC abundance in injured tissues in healthy animal models have shown markedly improved outcome at the wound site via increased levels of angiogenesis within diabetic wounds, ischemic cardiac muscle, critical limb ischemia, and bone defects [4,13–24]. However, typical therapuetic strategies involving harvest of CACs from the patient and re-delivering the CACs into the wound site. These strategies are hampered by the relatively low number of CACs present in patient blood and bone marrow [6], a lack of unique cell surface markers for CACs [5–9], and regulatory difficulties [25,26]. Thus, strategies to harness endogenous CACs in vivo may be particularly valuable [4,25].

Multiple studies of endogenous CAC recruitment in animal models of ischemic myocardium, diabetic wounds, and within subcutaneous sites have demonstrated the general efficacy of this type of approach [21,27–34]. A subset of these approaches utilize biomaterials to both deliver soluble recruitment factors and serve as a matrix for the recruited CACs [31,32,34]. However, little is known about how basic biomaterial design parameters such as stiffness, degradability, and encapsulated soluble factor content affect CAC invasion into the biomaterial. Additionally, previous biomaterial-based recruitment strategies are not specific to CAC recruitment and invasion, and thus simultaneous recruitment of multiple cell types occurs [32]. It is possible that these additional recruited cell types may be detrimental at the recruitment site by promoting increased inflammation from lymphocyte recruitment [35,36] or increased fibrosis from fibroblast recruitment [37].

Herein, we use chemically well-defined, hydrogel biomaterials to systematically study the effects of biomaterial stiffness, degradability, and encapsulated stromal-derived factor-1α (SDF-1α) content on the magnitude and specificity of CAC invasion. We utilized a thiolene chemistry to rapidly polymerize our hydrogels within a plate-based format amenable to enhanced-throughput screening [38]. We employed 8-arm, poly(ethylene glycol) (PEG)-based hydrogels due to the ability to easily modulate cross-linker molecule identity, stiffness, and soluble factor incorporation within this hydrogel system [39,40]. We also utilized a combination of mathematical modeling and empirical measurements to determine that a range of SDF-1α concentration gradients could be presented within the hydrogels and that these gradients were largely independent of hydrogel formulation. Thus, we were able to efficiently multiplex the variables of soluble gradient presentation and hydrogel formulation in these screening studies. These studies identified a subset of hydrogel formulations that promoted robust and specific CAC invasion in vitro. Importantly, this identified subset of hydrogel formulations serves as the first specific design criteria presented for the design of biomaterials to specifically recruit CACs. Furthermore, we posit that such formulations may be a promising tool for local recruitment of CACs in future therapeutic approaches seeking to harness CACs for angiogenesis as part of an in vivo tissue regeneration strategy. While the present work only highlights the possibility of this concept, we posit that with additional investigation and robustness testing, specific hydrogel formulations may be further developed that serve as highly specific and robust cell “filters,” which may be a promising new tool for tissue regeneration strategies.

2. Methods

2.1. PEGNB synthesis

Multi-arm, alcohol terminated, 8-arm PEG with a tripentaerythritol core (JenKem Technology USA) were functionalized with norbornene groups via reaction with norbornene anhydride, which is well-described elsewhere [39,40]. All other materials for this synthesis were obtained from Sigma-Aldrich. Briefly, norbornene acid at a 10× molar excess versus PEG alcohol groups and dicyclohexylcarbodiimide (DCC) at a 5× molar excess versus PEG alcohol groups were both dissolved in anhydrous dichloromethane (DCM). PEG, pyridine at a 5× molar excess versus PEG alcohol groups, and 4-(dimethylamino) pyridine at a 0.5× molar excess versus PEG alcohol groups were also dissolved within a separate DCM solution. Once the norbornene anhydride had formed (indicated by formation of a white precipitate), both DCM solutions were combined. This combined solution was stirred overnight, and the solution was then filtered to remove precipitates. PEG was recovered via precipitation in ethyl ether and was purified via dialysis against deionized (DI) water. Functionalization of the PEG polymer with norbornene groups (PEGNB) was confirmed via NMR.

2.2. Hydrogel formulation and photopolymerization

PEGNB hydrogels were polymerized via a thiolene reaction with dithiol cross-linkers in all cases. Precursor solutions contained PEGNB and dithiol cross-linkers, which were dissolved at desired concentrations in phosphate buffered saline (Fisher Scientific). In all cases, either a PEG-dithiol cross-linking molecule (3.4 kDa, Laysan Bio) or KCGGPQGIWGQGCK, a matrix metalloproteinase (MMP) degradable peptide [41], was utilized to cross-link hydrogels. CRGDS peptide (GenScript) was dissolved in these solutions at a concentration of 2 mM. When indicated, human, recombinant, stromal-derived factor-1α (R&D Systems) was also dissolved in these solutions by diluting from a 100 μg/mL stock solution containing 1 mg/mL bovine serum albumin as a carrier. Before polymerization, Irgacure 2959 photo-initiator (Ciba) was added to a concentration of 0.05 wt% via a 10-fold dilution from a 0.5 wt% stock solution. This reaction was commenced by sustained exposure to UV light (365 nm at 3.3 mW/cm² for 10 min unless otherwise specified). A complete list of all hydrogels used in the course of this work is included as Supplemental Table 1.

2.3. Culture of human circulating angiogenic cells

CD133+ cells collected from mobilized peripheral blood were purchased from AllCells, LLC. These cells were collected by AllCells, LLC from one donor (26 year old, African American, female) utilizing an IRB-approved protocol. For all experiments, these cells were cultured in StemSpan Serum Free Expansion Medium II (StemSpan SFEM II, StemCell Technologies) with CC100 supplement (StemCell Technologies) and 1% Penicillin/Streptomycin (Gibco). CD133+ cells were cultured in a non-adherent manner for 5–11 days and then cryopreserved to create working cell banks. Vials were thawed and expanded from these banks for all experiments.

For all experiments, cells were thawed from this bank and further cultured in a non-adherent manner for 3 to 6 weeks. After this non-adherent culture, the cells were transferred onto tissue culture polystyrene, at which point an adherent population of cells formed with a morphology consistent with that of CACs as reported in the literature [42,43]. The adherent fraction cells demonstrated uptake of Ac-LDL, bound UEA-1 Lectin, and promoted proliferation, migration, and tubulogenesis of endothelial cells. These behaviors are also consistent with the reported literature behavior of CACs [42,43].

Adherent cells were maintained on TCPS without detachment until needed for experiments with media changes every 2–3 days. When detached for use in an experiment, detachment was accomplished with TrypLE (Invitrogen) treatment for 10 min and subsequent scraping of the TCPS surface. CACs were then recovered via centrifugation at 2500 rcf for 5.5 min, counted on a hemocytometer, and seeded directly into the experiment.

2.4. Culture of human dermal fibroblasts

Human dermal fibroblasts (DFs) were obtained from Lonza. DFs were cultured directly on TCPS in Minimal Essential Media-α (Cell-Gro) with 1% Penicillin/Streptomycin (Gibco) and 10% v/v fetal bovine serum (Gibco). DFs were used between passage 6 and passage 9. Detachment for passage and when needed for experiments was accomplished via treatment with trypsin (Hyclone) for 5 min and subsequent washing of the cell culture surface. DFs were recovered via centrifugation at 0.2 kRCF for 5 min, counted on a hemocytometer, and seeded directly into the experiment or a new TCPS culture flask.

2.5. Culture of human mesenchymal stem cells

Bone marrow-derived human mesenchymal stem cells (MSCs) were obtained from Lonza. MSCs were cultured directly on TCPS in Minimal Essential Media-α (CellGro) with 5% Penicillin/Streptomycin (Gibco) and 10% v/v fetal bovine serum (Gibco). MSCs were used between passage 7 and passage 10. Detachment for passage and when needed for experiments was accomplished via treatment with trypsin (Hyclone) for 5 min and subsequent washing of the cell culture surface. MSCs were recovered via centrifugation at 0.2 kRCF for 5 min, counted on a hemocytometer, and seeded directly into the experiment or a new TCPS culture flask.

2.6. Culture of Jurkat Cell-line

Jurkat cells were obtained from ATCC. Jurkat cells were utilized in all studies as a model lymphocyte cell type. Jurkat cells were cultured non-adherently were cultured non-adherently in Roswell Park Memorial Institute Medium (RPMI, HyClone) with 5% Penicillin/Streptomycin (Gibco) and 10% v/v fetal bovine serum (Gibco). Jurkat cells were cultured from passage 3 to passage 8. For all passages and experiments, Jurkat cellss were recovered via centrifugation at 2500 rcf for 5.5 min, counted on a hemocytometer, and seeded directly into the experiment.

2.7. Trans-well chemotaxis assay

All chemotaxis assays were performed within a 96-well, Fluoroblok trans-well plate (Corning) to visualize only migrating cells with epifluorescence microscopy by examination of the bottom of the trans-well membranes. All trans-wells were gelatin-coated via 24-h incubation with a 0.2 w/v% aqueous solution of porcine gelatin (Sigma-Aldrich) at least 24 h before use. All assays were performed within M199 supplemented with 0.5% v/v FBS as the basal medium. When indicated, human, recombinant, stromal-derived factor-1α (R&D Systems) was dissolved in basal medium by diluting from a 100 μg/mL stock solution containing 1 mg/mL bovine serum albumin as a carrier.

All cells used in this assay were treated as follows. 48 h before the start of the experiment, cells were treated with Cell Tracker Red (CMTPX, Invitrogen) via the manufacturer’s instructions. 24 h before the start of the experiment, the cells’ culture medium was changed to M199 + 0.5% serum. At the start of the experiment, cells were harvested via TrypLE (for CACs), trypsin (DFs and MSCs), or collected via centrifugation (Jurkat cells) for seeding into the trans-wells.

For all assays, cells were seeded directly into the wells of the trans-well plate at a density of 1000 cells/well (12,500 cells/cm²). The assay was incubated for 6 h, and NIS Elements image processing software was used to count all fluorescent objects on the under-side of the membrane.

2.8. PEGNB hydrogel invasion assay

All cells used in this assay were treated as follows. 48 h before the start of the experiment, cells were treated with Cell Tracker Red (CMTPX, Invitrogen) via the manufacturer’s instructions. 24 h before the start of the experiment, the cells’ culture medium was changed to M199 + 1% serum. At the start of the experiment, cells were harvested via TrypLE (for CACs), trypsin (DFs and MSCs), or collected via centrifugation (Jurkats) for seeding into the trans-wells. Prior to seeding into the trans-well, cells were counted on a hemocytometer and viability was checked with dye-exclusion of Trypan Blue.

A diagram of the invasion assay is shown in Supplemental Methods 1. PEGNB hydrogel solutions were prepared as described in 3.2 Hydrogel Formulation and Photopolymerization. All PEGNB hydrogel solutions were maintained on ice during the assay setup. Initially, the trans-well plate was removed and inverted. 10 μL of hydrogel precursor solution was then pipetted onto the bottom of each trans-well membrane. 100 μL of hydrogel precursor solution was also placed into the wells of the bottom square-well plate. The trans-well plate was then gently placed back on top of the square well plate, and gentle agitation of the plate ensured complete coverage of each trans-well membrane and well with hydrogel precursor solution. UV light was shone through the bottom of the 96-well trans-well plate to induce photopolymerization (365 nm at 3.3 mW/cm² for 10 min). 150 μL of M199 + 1% serum was placed in the lower chamber on top of the hydrogel to prevent the hydrogel from dehydrating. Cells were then seeded into the top chamber in a volume of 50 μL of M199 + 1% serum.

The assay was then incubated for approximately 24 h, and the hydrogels were imaged with an inverted, fluorescent microscope (Nikon Ti-Eclipse). Fluorescent light was not allowed to pass through the trans-well membranes with the Fluoroblok technology, and thus only cells that had invaded the hydrogels were visible. Imaging of the hydrogel involved obtaining several images at various focal planes throughout the hydrogel (“z-stack”). Processing of all planes with image analysis software (NIS Elements) generated a “maximum intensity projection” image. Counting of all objects within this image determined the total number of invading cells. NIS Elements image processing software was used to count all fluorescent objects. These planes could also be used to generate a 3D representation of cell invasion into the hydrogel, and NIS Elements image processing software was also used to accomplish this.

2.9. Co-culture invasion assay

For all co-culture studies, the assay was set up as described in 3.8 PEGNB Hydrogel Invasion Assay. Cells were also treated in the same manner as described in 3.8 PEGNB Hydrogel Invasion Assay. During cell seeding, CACs and other cells types were seeded together within the wells of the trans-well plate. For 1 to 1 ratio experiments, 5000 CACs and 5000 cells of the other cell type were combined. For the 1 to 10 ratio experiments, 10,000 of the other cell type were combined with 1000 CACs. The assay was then incubated for approximately 24 h, and the hydrogels were imaged with an inverted, fluorescent microscope as described in 3.8 PEGNB Hydrogel Invasion Assay. For ease of comparison to other invasion assay results, all calculations in co-culture studies were based off of fold-increase over control wells containing only one cell type. This prevented the modulation of assay results due to seeding density differences rather than interactions between the populations.

2.10. Characterization of SDF-1α release and modeling of gradient formation

Trends in the diffusion coefficients of proteins within the hydrogel networks were predicted based upon established mathematical models from thermodynamic first principles of hydrodynamic theory [44,45]. The molecular radius of a theoretical protein (a) and diffusion coefficient of this theoretical protein in free solution (D_sol) was estimated via the Stokes-Einstein relationship (Eq. (1)). Hydrogel mesh size (ξ) was calculated based on the molecular weight between cross-links (M_c) within the hydrogel networks (Eq. (2)), and we determined M_c from established relationships with hydrogel swelling [44,45] (data not shown). Diffusion coefficients of proteins within the hydrogel network were predicted with an approximation of the Lustig-Peppas model [46], utilizing D_sol, a, and ξ (Eq. (3)).

Equation 1

Stokes-Einstein relationship for calculation of the diffusion coefficient of a protein in free solution, D_sol. MW is the molecular weight of the protein, ρ is the density of the protein (assumed to be 1 gram/mL), N is Avagadro’s number, R is the universal gas constant, μ is solution viscosity, and T is temperature.

$$\begin{array}{l}a={\left(\frac{3MW}{4\pi \rho N}\right)}^{\frac{1}{3}}\\ D_{\mathit{sol}}=\frac{RT}{6\mu \pi aN}\end{array}$$ (1)

Equation 2

Hydrogel mesh size calculated based upon the molecular weight between cross-links, M_c. M_c is calculated in above in Eq. (3). Here, l is the average length of a bond in the polymer, v_s is the swollen volume fraction of the polymer, C_n is the Flory characteristic ratio, and M_r is the molecular weight of the monomer.

$$\xi =\left(l{v}_s^{-1/3}\right){\left(\frac{2C_n{M}_c}{M_r}\right)}^{1/2}$$ (2)

Equation 3

Calculation of the diffusion coefficient within the hydrogel network, D_gel. This calculation is based upon D_sol, a, and ξ.

$$D_{\mathit{gel}}=D_{\mathit{sol}}\left(1-\frac{a}{\xi }\right)$$ (3)

For all empirical studies, SDF-1α (containing 1 mg/mL BSA as a carrier) or bovine serum albumin (BSA) were dissolved into the hydrogel precursor solution at 10,000 ng/mL 0.1% of all SDF-1α utilized in this study was labeled with I₁₂₅, and 0.9% of all utilized BSA in this study was labeled with I₁₂₅. For all release studies, a 200 μL hydrogel was formed within the bottom of a 1.5 mL Eppendorf tube. These hydrogels are assumed to be cylindrical in all cases. Before use, all Eppendorf tubes were blocked with a 1 mg/mL BSA solution for 24 h. For all release studies, hydrogels were incubated in M199 + 1% serum + 2% Penn/Strep + 1 mg/mL sodium azide.

At every time point of interest, the supernatant was removed, the hydrogel was washed three times to remove any residual liquid, and fresh supernatant was added to the samples. The total amount of release was calculated by adding the adding the counts per minute from the supernatant and wash solutions, and then transforming the counts per minute to a concentration using a standard curve. Supernatants were removed at the following times: 1 h, 2 h, 4 h, 6 h, 8 h, 12 h, 18 h, and 24 h. Release data was displayed in terms of percent mass released via Eq. (4) with total loaded mass determined by direct comparison to a standard curve. Mathematical estimation of the diffusion coefficients within the hydrogels were performed with Eq. (5) at each time point [47,48]. An overall diffusion coefficient was determined based upon an iterative solution minimizing the residual between the predicted diffusion coefficient from Eq. (3) and the diffusion coefficients obtained from Eq. (5). The error-function solution to Fick’s second law of diffusion with pseudo-steady state assumptions was then applied to predict gradient formation within the hydrogel (Eq. (6)) [49].

Equation 4

Calculation of M_%, the fractional mass released from the hydrogel expressed as a percentage of total. M_tx is the mass released at a specified time point and M_total is the total mass contained within the hydrogel.

$$M_{\%}=\frac{\sum M_{t1}+M_{t2}+M_{t3}+\dots }{M_{\mathit{total}}}$$ (4)

Equation 5

Calculation of diffusion coefficients within hydrogels based upon empirical data. M_% is the fractional mass released from the hydrogel expressed as a percentage of total (Eq. (4)), x is the total depth of the hydrogel cylinder, and t is time.

$$D_{\mathit{emp}}=\frac{M_{\%}^2\pi x^2}{t}$$ (5)

Equation 6

Calculation of solute concentration within PEGNB hydrogels as a function of depth into the hydrogel, x, and time, t. C is the concentration of solute at a given depth and time, C₀ is the source concentration within the hydrogel at the time point in question, and D is the solute diffusion coefficient within the hydrogel.

$$C(x,t)=C_0\mathit{erf}\left\{\frac{x}{2\sqrt{Dt}}\right\}$$ (6)

2.11. Statistical analysis

Statistical analysis of all data is indicated in figure captions within the text. In all cases where a T-test is indicated, the T-test was utilized to determine significance relative to a common control with a significance level set at p < 0.05. In all cases where ANOVA is indicated, an ANOVA was performed to detect significance of conditions to one another with a significance level set at p < 0.05. Subsequent post hoc analysis of data after an ANOVA was performed was carried out via T-test with a significance level of p < 0.05. In all cases, Microsoft Excel was used to perform these statistical tests.

3. Results

3.1. Gradients of SDF-1α stimulate CAC migration

CAC migration was dependent on gradients of SDF-1α within a trans-well migration assay. The CACs were initially seeded on top of the trans-well membrane, and this chamber is referred to as the “top” chamber. The chamber underneath the trans-well membrane is referred to as the “bottom” chamber. Additionally, positive chemotaxis of CACs was defined as migration towards higher SDF-1α concentrations.

Modulation of SDF-1α concentration gradients generated distinct chemotactic responses in the CAC cell population (Fig. 1). Introducing SDF-1α into the bottom chamber increased CAC migration to the underside of the membrane. SDF-1α concentrations of 1 ng/mL, 10 ng/mL, and 1000 ng/mL caused an increase in CAC migration to the membrane underside, and the largest increase was induced by 1000 ng/mL SDF-1α (approximately 2.25-fold increase over control) in the bottom chamber (Fig. 1a). When SDF-1α was added to both chambers thus preventing SDF-1α gradient formation, 1 ng/mL, 10 ng/mL, and 500 ng/mL concentrations showed no changes in CAC migration to the underside of the membrane relative to the control while 100 ng/mL and 1000 ng/mL decreased CAC migration relative to the control (Fig. 1b). Direct comparison of the presence or absence of an SDF-1α gradient indicated that significantly more CAC migration occurred in the presence of the SDF-1α gradient in all cases (Fig. 1c), thus suggesting that an SDF-1α gradient is a critical parameter to stimulate CAC migration.

Fig. 1.

Modulation of CAC migration with SDF-1α gradients. In all graphs, * indicates statistical significance from 0 ng/mL (control) via t-test (p < 0.05) and NS indicates no statistical significance. All data is presented as normalized to the control condition (0 ng/mL). As defined in the text, the top chamber signifies that top of the trans-well membrane where the CACs where initially seeded. The bottom chamber refers to the underside of the trans-well membrane. SDF-1α was introduced into the bottom chamber (A) and both chambers (B). Each situation stimulated unique migratory responses from the CACs (C). Note that in C, comparisons to controls (0 ng/mL) are not made, and all comparisons are instead relative to peers in the same group.

3.2. SDF-1α gradient formation within PEGNB hydrogels

Since modulation of SDF-1α concentration gradients stimulated CAC migration, we sought to model SDF-1α concentration gradients within PEGNB hydrogels as modulation of these gradients may regulate CAC invasion into these biomaterials. To initially characterize these gradients, SDF-1α was encapsulated within the PEGNB hydrogel matrix and a combination of empirical measurements and mathematical modeling were utilized to characterize the mass transport of the SDF-1α out of the hydrogel matrix. Mathematical modeling based upon first principles in three hydrogel formulations (3, 4, and 5 wt% PEGNB) suggested that SDF-1α diffusion within the hydrogels is largely independent of hydrogel formulation (Supplemental Fig. 1a). This finding was further supported by the finding that SDF-1α release curves from 3, 4, and 5 wt% PEGNB hydrogels were effectively identical (Supplemental Fig. 1b), which suggests similar diffusion coefficients in all three formulations. Based upon these release data, a pseudo-steady state model was used to predict the SDF-1α concentration gradients at various time points, and this model predicted the formation of largely linear concentration gradients that decreased in slope over the course of 24 h (Supplemental Fig. 1c).

We then directly applied this model to the theoretical initial SDF-1α concentrations of 10,000 ng/mL and 100 ng/mL over a distance scale pertinent to our in vitro screening studies (Supplemental Fig. 1d–e). While these initial SDF-1α concentrations were vastly different in terms of magnitude, the resulting gradients were directly proportional at all time points (Supplemental Fig. 1d–e). When applied across a range of SDF-1α concentrations, these data suggest that changing the initial concentration of SDF-1α within the hydrogels affects the slopes of the presented gradients proportionally at all time points (Supplemental Fig. 1f). Taken together, these data suggest that these hydrogels may present a range of SDF-1α concentration gradients by modulating initial SDF-1α concentration. Furthermore, since these data suggest that the SDF-1α diffusion coefficient is largely independent of hydrogel formulation, the same SDF-1α gradients may be presented within a variety of hydrogel formulations, thus allowing efficient multiplexing of these two variables when screening for their effects on CAC invasion.

3.3. SDF-1α and hydrogel degradability modulate CAC invasion

Based upon empirical measurements and mathematical modeling, we determined that the release of encapsulated SDF-1α from PEGNB hydrogels forms an SDF-1α gradient within the hydrogel over time (Supplemental Fig. 1). Thus, we hypothesized that CACs would invade PEGNB hydrogels since SDF-1α gradients modulated CAC migration (Fig. 1). However, previous studies have also demonstrated that PEG-based hydrogels are largely impermeable to cells unless cell-mediated degradation of the hydrogel network is permitted [50], and thus we hypothesized that the degradability of the PEGNB hydrogel network would simultaneously modulate CAC invasion.

Using a hydrogel invasion screening assay (Supplemental Methods 1), we investigated the influence of hydrogel degradability on CAC invasion, and found CAC invasion to be highly dependent on hydrogel degradability. We encapsulated a constant amount of SDF-1α within the PEGNB hydrogel network (1000 ng/mL) and modulated hydrogel degradability across multiple hydrogel formulations containing various amounts of PEGNB (PEGNB wt%) and ratios of PEGNB to cross-linker molecules (hydrogel cross-linking) (Fig. 2). Regardless of PEGNB wt% or hydrogel cross-linking, hydrogels cross-linked with a peptide susceptible to MMP degradation (hereafter referred to as the MMP degradable peptide) allowed more CAC invasion than did hydrogels cross-linked with a non-degradable PEG-dithiol cross-linker (Fig. 2). Notably, 3 wt% PEGNB hydrogels promoted significantly higher levels of invasion than did hydrogels formulated at other PEGNB wt% (10 wt% to 4 wt%), but even in this case, the MMP degradable peptide cross-linker allowed more invasion than did the PEG-dithiol cross-linker (Fig. 2c).

Fig. 2.

PEGNB hydrogels cross-linked with an MMP degradable cross-linker permitted more CAC invasion than did hydrogels cross-linked with a non-degradable PEG-dithiol cross-linker. Heat-map legend presented at bottom of figure. A) Schematic representation of the assay results. PEG-dithiol cross-linked hydrogels did not permit any invasion into the hydrogel, while MMP degradable hydrogels permitted invasion of CACs. Representative maximum intensity projections are shown at right for both images. B) 3D reconstruction of CAC invasion into MMP degradable and non-degradable hydrogels. C) Heat-map of CAC invasion into MMP degradable and non-degradable hydrogels formulated from various PEGNB wt%. * indicates statistical significance relative to peers via two-factor ANOVA (p < 0.05). D) Heat-map of CAC invasion into MMP degradable and non-degradable hydrogels formulated with various amounts of hydrogel cross-linking. * indicates statistical significance relative to peers via two-factor ANOVA (p < 0.05).

CAC invasion was dependent upon MMP-mediated degradation of the MMP degradable peptide cross-linkers. Replacing MMP degradable peptide cross-linkers with PEG-dithiol cross-linkers (Fig. 3a) decreased CAC invasion into the hydrogel in a dose dependent manner (Fig. 3a). Treatment with either Prinomastat HCl (inhibits MMP-2, 3, 9, 13, and 14 [51]) or SB-3CT (inhibits MMP-2 and 9 [52]) dose dependently inhibited CAC invasion into the hydrogels. CACs are reported to produce MMP-9 [53], and the decrease in invasion caused by SB-3CT or Prinomastat HCl, which both inhibit MMP-9, suggest that CAC invasion in degradable hydrogels may be dependent upon MMP-9 activity. Taken together, these data suggest that MMP-mediated degradation of the PEGNB hydrogels is critical for CAC invasion in response to an SDF-1α gradient and that, regardless of hydrogel formulation, hydrogel susceptibility to MMP-mediated degradation is a critical biomaterial design parameter.

Fig. 3.

CAC invasion of PEGNB hydrogels is modulated by hydrogel degradability. All data is displayed as fold decrease over control (control column at left in both heat maps), and the heat map legend is displayed at the bottom of the figure. A) Heat-map of CAC invasion into PEGNB hydrogels with a decreasing number of MMP degradable cross-links (labels at top,% cross-links that are degradable). * indicates statistical significance relative to peers via ANOVA (p < 0.05). B) Heat map showing that introduction of Prinomastat HCl (top) or SB-3CT (bottom) decreased CAC invastion into PEGNB hydrogels (concentration shown at top for Prinomastat HCl and at bottom for SB-3CT). * indicates statistical significance relative to peers via ANOVA (p < 0.05).

3.4. CAC hydrogel invasion is simultaneously modulated by SDF-1α content

SDF-1α content and hydrogel stiffness simultaneously modulated CAC invasion into PEGNB hydrogels. We controlled hydrogel stiffness by modulating PEGNB wt% (Fig. 4a) or hydrogel cross-linking (Fig. 4b) and simultaneously modulated the SDF-1α content in each hydrogel condition. A trend of increasing CAC invasion with increasing SDF-1α content was observed in all conditions (Fig. 4c). When aggregating all data based upon hydrogel cross-linking regardless of SDF-1α content, 60% hydrogel cross-linking permitted less CAC invasion than 100% or 75% cross-linked hydrogels (Fig. 4d). When categorizing all data based upon PEGNB wt% regardless of SDF-1α content, 3 wt% PEGNB permitted more CAC invasion than all other PEGNB wt% conditions (Fig. 4e).

Fig. 4.

Screening the effects of hydrogel stiffness and hydrogel SDF-1α content on CAC hydrogel invasion. A) Heat map displaying the results of simultaneous modulation of PEGNB wt% and SDF-1α content. A clear trend of increasing CAC invasion with increasing SDF-1α content was observed. Legend displayed at right below B. * indicates statistical significance relative to peers via two-way ANOVA (p < 0.05). B) Heat map displaying the results of simultaneous modulation of hydrogel cross-linking and SDF-1α content. A clear trend of increasing CAC invasion with increasing SDF-1α content was observed in all cases. Legend displayed below. * indicates statistical significance relative to peers via two-way ANOVA (p < 0.05). C) Comparison of all conditions containing SDF-1α versus those not containing SDF-1α shows an increase in CAC invasion when SDF-1α is present. * indicates statistical significance via t-test (p < 0.05). D) Across all conditions regardless of SDF-1α content, 60% hydrogel cross-linking demonstrated lower levels of CAC hydrogel invasion. * indicates statistical significance via t-test (p < 0.05). E) Across all conditions regardless of SDF-1α content, 3 wt% PEGNB demonstrated higher levels of CAC hydrogel invasion. * indicates statistical significance via t-test (p < 0.05).

3.5. SDF-1α Stimulates chemotaxis and invasion of MSCs, DFs, and lymphocytes

The cognate receptor for SDF-1α, CXCR4, is expressed not only by CACs, but also mesenchymal stem cells (MSCs), dermal fibroblasts (DFs), and lymphocytes, [32,35,54–60]. Thus, we sought to characterize the effects of SDF-1α concentration gradients on the migration of these other cell types by loading SDF-1α into the bottom chamber of a trans-well and measuring cell migration to the underside of the membrane (Supplemental Fig. 2). DF migration increased with increasing SDF-1α concentration in the bottom chamber up to a concentration of 100 ng/mL (Supplemental Fig. 2a). MSC migration increased at two specific concentrations (1 ng/mL and 500 ng/mL) and was indistinguishable from the 0 ng/mL control at any other concentrations (Supplemental Fig. 2b). Lymphocytes demonstrated no response to SDF-1α up to a concentration of 1000 ng/mL, at which point a 6-fold increase was observed relative to the control (Supplemental Fig. 2c). 2-way ANOVA demonstrated that modulation of SDF-1α concentration in the bottom chamber of the trans-well brings about statistically significant changes in cell migration (p < 0.05, data not shown) and that an interaction between cell type and migratory response to SDF-1α exists (p < 0.05, data not shown).

Since SDF-1α increased the migration of DFs, MSCs, and lymphocytes, we characterized whether these cells types were also capable of invading PEGNB hydrogels of various SDF-1α contents and stiffness (by modulating PEGNB wt% and hydrogel cross-linking) (Figs. 5 and 6). Aggregating the data based upon cell type and the presence of SDF-1α showed that none of these cell types demonstrated the level of invasion that CACs demonstrated, with or without the presence of SDF-1α (Fig. 5a). DFs demonstrated that lowest level of invasion in all cases (Fig. 5a). MSCs demonstrated the second highest levels of invasion, and MSC invasion was decreased in the presence of SDF-1α (Fig. 5a). Lymphocytes demonstrated higher levels of invasion than DFs but lower levels than MSCs, and only lymphocyte migration was not affected by the presence of SDF-1α (Fig. 5a).

Fig. 5.

Screening the effects of hydrogel stiffness (modulation of PEGNB wt%) and hydrogel SDF-1α content on DF, MSC, lymphocyte, and CAC invasion. A) Aggregated data (by cell type, with and without SDF-1α) reveals that CACs demonstrated the highest levels of invasion in all cases, and that addition of SDF-1α increased CAC invasion. ** indicates statistical significance relative to no SDF-1α condition via t-test (p < 0.05), and * indicates statistical significance relative to peers via t-test (p < 0.05). DFs (B), MSCs (C), lymphocytes (D), and CACs (E) demonstrated unique invasion responses to PEGNB wt% and SDF-1α content. Note that CAC data is re-plotted from Fig. 6 for comparison purposes. All data are presented as heat maps of number of invading cells with the heat map legend at the bottom left. * indicates statistical significance relative to peers via two-way ANOVA (p < 0.05).

Fig. 6.

Screening the effects of hydrogel stiffness (modulation of hydrogel cross-linking) and hydrogel SDF-1α content on DF, MSC, lymphocyte, and CAC invasion. DFs (A), MSCs (B), lymphocytes (C), and CACs (D) demonstrated unique invasion responses to hydrogel cross-linking and SDF-1α content. Note that CAC data is re-plotted from Fig. 6 for comparison purposes. All data are presented as heat maps of number of invading cells with the heat map legend at the bottom left. * indicates statistical significance relative to peers via two-way ANOVA (p < 0.05).

DFs poorly invaded all hydrogel conditions, and the observed levels of invasion were not comparable to that of CACs in any condition (Fig. 5a–b, 6a). MSCs demonstrated higher levels of invasion than did DFs, and MSCs demonstrated scattered regions across the heat map of significant invasion that was equivalent to or higher than CAC invasion (data not shown, comparison via t-test, p < 0.05) (Figs. Fig. 5a, c, and 6b), but on the average, MSC invasion was lower than that of CACs. Similar to DFs, lymphocytes also demonstrated little invasion across conditions (Figs. 5a, d, and 6c), with the exception of the 3 wt% PEGNB 100 ng/mL and 500 ng/mL conditions, which were equivalent to the levels of invasion demonstrated by CACs in those hydrogel conditions (comparison via t-test, p < 0.05, data not shown) (Fig. 5d). Interestingly, DFs, MSCs, lymphocytes, and CACs also responded differently to hydrogels of various PEGNB wt% and cross-linking (Supplemental Fig. 3).

3.6. Cacs demonstrate specific invasion into subsets of hydrogels

Since DFs, MSCs, lymphocytes, and CACs demonstrated distinct trends of hydrogel invasion, we hypothesized that subsets of hydrogel conditions may permit preferential invasion of CACs. To identify such conditions, we scored each hydrogel condition with the following formula (hereafter referred to as “specificity”):

$$\mathit{Specificity}=\frac{\mathit{Invading}\mathit{CACs}}{\mathit{Invading}\mathit{CACs}+\mathit{Invading}\mathit{DFs}+\mathit{Invading}\mathit{MSCs}+\mathit{Invading}\mathit{Lymphocytes}}$$

This scoring revealed a distinct subset of hydrogels that promoted specific invasion of CACs (Fig. 7). Similar approaches were used to identify distinct subsets of hydrogels that promoted the specific invasion of MSCs and lymphocytes, and no hydrogel condition promoted specific invasion of DFs (Supplemental Fig. 4).

Fig. 7.

Invasion specificity for CACs across all SDF-1α contents, PEGNB wt% (A), and hydrogel cross-linking (B) conditions. A general trend of increasing invasion specificity for CACs with increase SDF-1α content was observed. All data are presented as heat maps with the heat map legend at the bottom left. * indicates statistical significance relative to peers via two-way ANOVA (p < 0.05) and NS indicates no statistical significance. C) Comparison of invasion specificity for CACs compared to invasion specificity for DFs, lymphocytes, and MSCs. CACs demonstrate the highest level of invasion specificity, followed by MSCs, lymphocytes, and DFs. Only CACs demonstrated an increase in specificity in the presence of SDF-1α. * indicates statistical significance relative to peers via t-test (p < 0.05), and ** indicates statistical significance relative to no SDF-1α condition via t-test (p < 0.05).

Hydrogels containing 1000 ng/mL and 10,000 ng/mL of SDF-1α demonstrated a higher specificity for CAC invasion relative to all other SDF-1α concentrations (Fig. 7). This effect occurred regardless of PEGNB wt% and hydrogel cross-linking (Fig. 7a, Supplemental Fig. 5a), but lowering hydrogel cross-linking to 60% did lower the magnitude of this effect (Fig. 7b, Supplemental Fig. 5b). Across all hydrogel conditions, those that contained SDF-1α promoted a higher level of invasion specificity for CACs than those that did not (Fig. 7c). However, even in the absence of SDF-1α, all hydrogel conditions promoted a higher level of specificity for CACs than for other cell types (Fig. 7c and Supplemental Fig. 5). Taken together, these data suggest that the PEGNB hydrogels studied here promote specific invasion of specificity for CACs, and that this specificity can be augmented by incorporation of SDF-1α.

3.7. Co-culture invasion

While particular subsets of PEGNB hydrogels demonstrated high levels of specificity for CAC invasion when CACs were alone in the assay, it is possible that interactions between CACs and other cell types may alter these trends. To investigate these effects, we seeded CACs in co-culture with DFs, MSCs, and lymphocytes into our invasion assay. Since SDF-1α content modulated CAC invasion specificity but hydrogel formulation largely did not, we held PEGNB wt% and hydrogel cross-linking constant at 4 wt% and 75% respectively while adjusting SDF-1α concentration. We investigated co-culture ratios of 1 to 1 and 1 to 10 CACs to co-culture cell type, and we calculated the specificity of invasion for CACs within each condition. Regardless of co-culture cell type or co-culture ratio, CACs demonstrated high invasion specificity at the 1000 ng/mL SDF-1α concentrations (Fig. 8). Taken together, these data suggest that high CAC invasion specificity at 1000 ng/mL concentrations of SDF-1α persists even when CACs may be interacting with other cell types.

Fig. 8.

Screening the effects of co-culture of CACs with lymphocytes (A), MSCs (B), and DFs (C). In all cases, CACs demonstrated the highest invasion specificity at the highest SDF-1α concentration. This trend occurred regardless of the co-culture ratio (shown at left) in each graph. All data are presented as heat maps with the heat map legend at the bottom left. * indicates statistical significance relative to peers via two-way ANOVA (p < 0.05).

Furthermore, co-culture with MSCs and lymphocytes at different co-culture ratios appeared to differentially affect CAC invasion specificity, and the most prominent cases of such effects were observed in the 0 ng/mL and 1000 ng/mL SDF-1α conditions (Fig. 8a–b). Alternatively, co-culture with DFs did not demonstrate this same effect, and CAC invasion specificity was unaffected by co-culture ratio with fibroblasts (Fig. 8c). Further analysis of the 0 ng/mL and the 1000 ng/mL SDF-1α conditions with MSC and lymphocyte co-cultures at co-culture ratios of 1 to 1 and 1 to 10 demonstrated unique trends in CAC specificity (Supplemental Fig. 6). This suggests that the presence of MSCs and lymphocytes, as well as SDF-1α concentration, affects CAC invasion specificity, and further study of this phenomenon is warranted.

4. Discussion

SDF-1α releasing hydrogels may be a promising strategy to recruit CACs in vivo since they do not suffer from many of the difficulties that are encountered in CAC harvest and re-delivery strategies [5–9,25,26]. Previous approaches utilized SDF-1α delivering biomaterials to locally stimulate angiogenesis [31,32,34], but it is unclear from these studies how biomaterial design affects CAC recruitment and invasion. In this work, we used a hydrogel invasion screening assay to investigate the effects of hydrogel stiffness, degradability, and SDF-1α content on the magnitude of and the specificity of CAC invasion into PEG hydrogels in vitro. With this approach, we identified a subset of hydrogels capable of effectively promoting specific CAC invasion. This subset of hydrogels may be directly useful as or as a component of biomaterial implants for tissue regeneration, as their induction of specific CAC recruitment and invasion may induce angiogenesis into and around the implant that is superior to current approaches. Alternatively, these hydrogels may serve as model systems for the in vitro study of CAC trafficking in the extracellular matrix, as well as instructive templates for future CAC recruitment approaches.

Given available data on PEG hydrogels, we hypothesized that modulating PEGNB wt% and hydrogel cross-linking likely affected hydrogel stiffness without also modulating SDF-1α diffusion. We tested this hypothesis with empirical measurements and mathematical modeling, and found that SDF-1α release was indeed decoupled from hydrogel formulation (Supplemental Fig. 1). This finding was not limited to SDF-1α (data not shown), and thus this approach may be expanded to the delivery of other soluble proteins. This approach allows us to modulate soluble protein gradient parameters, such as slope and overall magnitude, and then measure the resulting impact on cell migration. Such data obtained from the systematic probing of gradient parameters may in future studies allow the creation of mathematical models to predict cell migration response to soluble protein gradients in the extracellular matrix. Thus, here we presented the same SDF-1α gradients within different hydrogel stiffnesses to screen the effects of these variables simultaneously on CAC invasion.

We found that SDF-1α modulated CAC chemotaxis within a trans-well assay (Fig. 1), and this is consistent with previous reports [21,31,32,34,35,61–66]. Our data suggest that biomaterials releasing 1000 ng/mL of SDF-1α may stimulate CAC positive chemotaxis (chemotaxis up a concentration gradient), and this is consistent with previous approaches releasing SDF-1α from biomaterials [31,32,34]. Surprisingly, at 100 ng/mL concentrations of SDF-1α, we observed a comparable level of negative chemotaxis (chemotaxis down a concentration gradient) when SDF-1α and the CACs were both introduced into the top chamber of the trans-well (Supplemental Fig. 7). These data suggest that SDF-1α, depending on the concentration, may have opposite effects on CAC migration, and this concept was previously reported as a mechanism by which SDF-1α directs lymphocyte trafficking [67]. This finding may explain the lower levels of CAC invasion into PEGNB hydrogels observed at 100 ng/mL SDF-1α relative to higher SDF-1α concentrations (Fig. 4). In addition, the observation that CAC invasion was MMP-dependent was consistent with previous studies demonstrating the impermeability of PEG-based hydrogels to mammalian cells [50] (Fig. 2). CACs demonstrated increasing invasion with increasing SDF-1α content regardless of hydrogel stiffness (Fig. 4), which strongly suggests that PEGNB hydrogel invasion is largely regulated by SDF-1α content within this range of hydrogel stiffnesses (11 kPa to 1 kPa, data not shown). While there may exist an upper limit where increasing SDF-1α concentration no longer results in an increase in CAC chemotaxis, we did not reach such a limit in this work.

By combining our mathematical modeling of the presented SDF-1α concentration gradients with our empirical data of CAC invasion, we can draw important conclusions related to the design criteria of biomaterials to recruit CACs. With the mathematical model presented here, we were able to tabulate the SDF-1α concentrations found within our hydrogels with respect to initial SDF-1α concentration and position (Supplemental Fig. 1d–f). This modeling predicted that these gradients are largely linear in nature, their slopes decrease over time, and are proportional to one another at all time points based upon initial loading concentration of SDF-1α (Supplemental Fig. 1d–e). Additionally, our empirical data demonstrated that the two highest initial concentrations of SDF-1α presented (1000 ng/mL and 10,000 ng/mL) induced the highest levels of specific CAC invasion. Thus, when analyzing these results together, we can conclude that hydrogel biomaterials in the 11 kPa to 1 kPa stiffness range, when presenting an SDF-1α gradient at or above that presented in the 1000 ng/mL SDF-1α condition (99.3 ng/(mL × mm) by 24 h, Supplemental Fig. 1f), will induce the specific invasion of CACs. To the best of our knowledge, this represents that first specific design criteria for a biomaterial to recruit CACs, and this criteria may serve as an initial starting point for future designs. Future work along these lines may further elucidate these effects beyond the time-fame presented here (24 h), additional design criteria, and the mechanisms for the observed effects. Such future work may further probe the interesting subsets of conditions identified here with techniques such as immobilizing specific SDF-1α concentration gradients that fall within the range identified here.

DFs, MSCs, and lymphocytes demonstrated distinct responses within the PEG hydrogel invasion assay compared to CACs (Figs. 5 and 6). Though DFs showed a migratory response to SDF-1α (Supplemental Fig. 2) [54–56], they poorly invaded the PEG hydrogels regardless of hydrogel formulation or SDF-1α content (Figs. 7a, 8a, and Supplemental Fig. 3), which suggests that the PEG hydrogel matrix selected against DF invasion. Our hydrogels presented only a CRGDS peptide, which mimics a key integrin binding domain on fibronectin [68,69], and recent studies show that dermal fibroblasts require binding to three distinct functional domains on fibronectin for optimal migration through the extracellular matrix [70]. By presenting only CRGDS, our hydrogels may not provide a suitable matrix for dermal fibroblast motility, thus excluding DFs from the hydrogels.

MSCs presented higher levels of PEGNB hydrogel invasion than DFs but these levels were still on average lower than that of CACs (Figs. 5, 6, and Supplemental Fig. 3). Particular subsets of hydrogels demonstrated MSC invasion levels comparable to or higher than CAC invasion so it may be possible to optimize hydrogel formulation to increase MSC invasion. Previous studies support this assertion and demonstrate that MSC motility is highly dependent on hydrogel formulation [71]. However MSCs also broadly demonstrated a trend of decreasing invasion with increasing SDF-1α content which is an unexpected result since MSCs are known to migrate in response to SDF-1α [57]. Notably SDF-1α presentation from the hydrogels in the current study is distinct from trans-well assays where MSC migration to SDF-1α is conventionally studied. In particular the hydrogels present a sustained gradient over the course of the experimental time-frame and a trans-well assay presents a transient gradient (Supplemental Fig. 1) [72]. Thus we can speculate that sustained SDF-1α gradients may not recruit MSCs while transient gradients do. Alternatively multiple studies suggest that SDF-1α does not act alone to stimulate MSC migration and that SDF-1α gradients in combination with hypoxia tumor necrosis factor-α and hepatocyte growth factor can dramatically increase MSC response to SDF-1α [58,57,73]. Therefore simultaneous presentation of these cues may also be needed to increase MSC invasion into PEG hydrogels. Lymphocyte invasion was limited across most conditions with a limited subset of conditions demonstrating a strong invasion response (Figs. 5c and 6c). The Jurkat cell line used in this work as a model lymphocyte cell type is known to migrate in response to SDF-1α gradients [35,59,60], and thus it is possible that only this limited subset of conditions presented the correct cues for a strong Jurkat cell line migratory response to the SDF-1α gradients

While the above explanations for the lower levels of DF, MSC, and lymphocyte invasion relative to CAC invasion may possibly explain the results of this study, another mechanism that may broadly explain these results is the necessity of proteolytic degradation of the PEGNB hydrogel for all cell invasion [50]. This requirement has been shown to reduce cell motility in PEG-based biomaterials relative to naturally-derived hydrogels (collagen or fibrin-based) [74]. The necessity of proteolytic degradation effectively removes the amoeboid form of cell migration [74], and thus differences in invasion may at least in part be due to differences in protease expression. For example, CACs produce high levels of MMP-9 in response to stimulation with soluble growth factors [53], and thus are well equipped to proteolytically degrade the PEGNB hydrogels in response to soluble SDF-1α stimulation. Further studies of microporous PEGNB hydrogels may further elucidate this effect by allowing direct comparisons of conditions where amoeboid cell migration is permitted versus prohibited. Additionally, the current approach did not reveal information about the rate of cell invasion into the PEGNB hydrogels, and the use of confocal microscopy at specific time points on a subset of conditions may allow us to better understand this parameter [38].

The presence of SDF-1α in the hydrogels increased invasion specificity for CACs (Fig. 7), and such trends also persisted in a co-culture environment (Fig. 8), suggesting that the cross-talk between cell types does not abrogate CAC recruitment. This concept is supported by recent evidence that SDF-1α can stimulate either positive or negative chemotaxis depending upon the local concentration, and SDF-1α may more broadly serve to “fine tune” cell recruitment and retention within a tissue [67,75]. Previous studies have also noted that delivery of SDF-1α reduced the infiltration of inflammatory cells into injured tissue [76], and this further suggests a role for SDF-1α in fine tuning cell recruitment and retention. Thus, the use of SDF-1α, in combination with biomaterial design, to “filter” cell invasion into a biomaterial may be a promising approach to control host-biomaterial interactions. While the present work only highlights the possibility of this concept, further investigation and robustness testing of specific formulations may lead to the develop of highly specific and robust cell filters that serve as promising new tools for tissue regeneration.

5. Conclusion

Herein, we investigated the effects of biomaterial design parameters on their ability to selectively recruit CACs. Specifically, we investigated the effects of hydrogel degradability, stiffness, and SDF-1α content on both the magnitude and specificity of CAC invasion by screening a total of 263 hydrogel conditions with 3 replicates of each condition. MMP degradable hydrogels containing 1000 ng/mL and 10,000 ng/mL initial concentrations of SDF-1α could specifically promote CAC invasion over that of other tested cell types. The shallowest gradient presented that still promoted specific CAC invasion (1000 ng/mL condition after 24 h) had a slope of 99.3 ng/(mL × mm), and thus we conclude that similar biomaterials presenting an SDF-1α gradient at this slope or higher will also specifically promote CAC invasion. Importantly, to the best of our knowledge, this is the first established design criteria for biomaterials designed to specifically recruit CACs. DFs demonstrated limited invasion into these hydrogels in all conditions, and MSCs and lymphocytes demonstrated invasion comparable to that of CACs only within a specific subset of conditions. Such trends persisted even when the assay was performed with co-cultures of these cell types and CACs, thus further suggesting that cross-talk between cell types does not abrogate this effect. Taken together, our findings highlight a role for SDF-1α in the recruitment and retention of specific cell types. They also highlight a new potential role for biomaterials, and the filtering of invading cell types to allow only desired cell types to infiltrate a biomaterial.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.actbio.2017.03.048.