Sprouting angiogenesis induces significant mechanical heterogeneities and ECM stiffening across length scales in fibrin hydrogels

Cell culture

Normal human dermal fibroblasts (DFs, Lonza, Walkersville, MD) were cultured in Dulbecco’s modified eagle medium (DMEM, Life Technologies, Grand Island, NY) supplemented with 10% fetal bovine serum (FBS, Life Technologies) and 1% penicillin streptomycin (Life Technologies) and were used up to passage 7. Human umbilical vein endothelial cells (ECs) were either harvested from fresh umbilical cords as previously described [15] or purchased from a commercial source (Lonza). Two different sources of ECs were used to ensure robustness of the observed biological responses. The ability of these two different sources of HUVECs to sprout in our fibrin-based assays was quantitatively equivalent (data not shown). HUVECs for all experiments were cultured in fully supplemented EGM2 (Lonza) and used between passages 2–4. Media for both cell types were exchanged 3 times a week and cells were harvested below 80% confluence using 0.05% trypsin-EDTA (Life Technologies).

Fibrin-based capillary morphogenesis assay

A three-dimensional cell culture model of capillary morphogenesis was assembled following adapted protocols as previously described [15, 16, 23]. Briefly, EC-coated microbeads were embedded in fibrin gels with DFs either embedded or overlaid on the gel as a monolayer. A stock solution of sterilized Cytodex microcarrier beads (Sigma-Aldrich, St Louis, MO) was prepared ahead of time by autoclaving in PBS. The day before construct assembly, microbeads were coated with ECs by combining 1×10⁴ microbeads with 4×10⁶ ECs in 5 mL of EGM2 in an upright T-25 tissue culture flask (Corning Inc, Corning, NY). The flask was incubated for 4 hours with periodic agitation every 30 minutes. Afterwards, 5 mL of fresh EGM2 was added, and the 10 mL suspension of freshly coated beads was transferred to a new T-25 and allowed to incubate overnight in the standard tissue culture position. The following day, beads were transferred to a 15 mL conical tube (VWR, Radnor, PA) and allowed to settle by gravity between two washes with fresh EGM2. Fibrinogen from bovine plasma (Sigma) was dissolved in serum free EGM2 to achieve a final concentration of 2.5 mg/mL clottable fibrinogen upon gelation, and sterile filtered through a 0.22 μm PES membrane filter (Merck Millipore Ltd, Tullagreen, Carrigtwohill, Co. Cork, IRL). For conditions in which the stromal fibroblasts were embedded within the fibrin gel (“embedded”) a suspension of DFs was added to the fibrinogen at a final concentration of 2.5×10⁴ cells/mL; for the other conditions, an equal volume of EGM2 was added instead. Microbeads were added to the solution at 50 beads/mL. Heat-inactivated FBS was added to the solution immediately prior to gelation for a final concentration of 5%. Tissue culture dishes were spotted with 40 μL of 100 U/mL thrombin reconstituted in ddH₂O per mL fibrinogen. Dishes were allowed to sit for 5-minutes before being transferred to incubate at 37 °C for another 25 min utes to allow for complete gelation. After gelation, 2 mL of EGM2 per 1 mL of fibrin gel was overlaid for all gels. For conditions in which the stromal fibroblasts were cultured on top of the gel (“overlay”), DFs were introduced in the overlaid EGM2 at a concentration of 2.5×10⁴ cells/mL of fibrin gel to achieve equal DF numbers per gel for both overlay and embedded conditions. For each independent experiment, multiple gels were cast for each time point. Gel constructs (0.5 mL total volume) were fabricated in 24-well tissue culture plates (Corning Inc) for bulk rheology and network quantification assays. For micro-rheology and reflection confocal imaging, gel constructs (1 mL total volume) were fabricated in 35 mm glass bottom dishes (MatTek, Ashland, MA).

For experiments involving Transwell inserts, fibrin based gel constructs (2 mL total volume) were fabricated in the bottom chamber of 6-well Transwell plates (24 mm diameter inserts containing 3.0 μm pores; Corning). DFs were cultured on top of the insert, and DAPI staining was used to confirm these cells did not migrate through the porous insert to the gel surface during the assay. Because the bottom of the insert would rest on the gel if used as provided, sterilized silicon O-rings (MSC, Melville, NY, part # S70-028) were used to space the insert off the top of the gel. All tissue constructs were cultured for up to two weeks with media exchanged on day 1 and every two days thereafter.

Fluorescent imaging and quantification of capillary morphogenesis

Images were acquired using an Olympus IX81 confocal microscope equipped with a USH-103OL mercury lamp (Olympus America, Center Valley, PA), a Hamamatsu Orca II CCD camera (Hamamatsu Photonics, Hamamatsu City, Japan), and Metamorph Premier software (Molecular Devices, Sunnyvale, CA). For visualization of tubules and cell nuclei, co-cultures were fixed with Z-Fix aqueous buffered zinc formalin fixative (Anatech, Battle Creek, MI) and stained with a rhodamine-conjugated lectin from Ulex europaeus (UEA, Vector Laboratories, Burlingame, CA) and 4′,6-diamidino-2-phenylindol (DAPI, Sigma-Aldrich). UEA binds glycoproteins and glycolipids specific to endothelial cells. UEA- and DAPI-stained samples were acquired using red (Ex: 562 nm, bandwidth: 40 nm; Em: 641 nm, bandwidth: 75 nm) and blue (Ex: 377 nm, bandwidth: 50 nm; Em: 477 nm, bandwidth: 60 nm) filter sets, respectively. Network length was quantified at days 1, 4, 7, and 14 with all beads imaged at 4x magnification. On day 14, multiple images often had to be stitched together to fit the entire network from a single bead. For unbiased measurements, the microscope was rastered through each gel and all beads were imaged that were far enough away from the edge of plate such that sprouting was unimpinged and did not have overlapping networks with a neighboring bead. This resulted in 8–40 beads per condition for each independent experiment being quantified, with diminishing beads meeting the criteria as time progressed. Three independent experiments (N=3) were conducted for each condition and time point, and the aggregate data from all beads across all 3 independent experiments were presented to illustrate the spread in biological response. Total tube length per bead was quantified using the Angiogenesis Tube Formation module in Metamorph. The average network length per bead for each condition of each independent experiment was then used for statistical analysis. DF proliferation in overlay conditions was quantified by taking DAPI images of the DFs in a monolayer on top of the gel and manually determining cell density.

Bulk rheology

The bulk mechanical properties of fibrin-based constructs were measured via parallel plate shear rheology using an AR-G2 rheometer (TA Instruments, New Castle, DE) equipped with an 8 mm diameter measurement head and a Peltier stage. Oscillatory shear measurements of 6% strain amplitude and a frequency of 1 rad/sec were performed on days 1, 4, 7, and 14 directly in multi-well tissue culture plates with the rheometer stage maintained at 37 °C. Cell culture media were aspirated before measur ements, with a small volume left to ensure the gel remained wet. Rheology of elastic pre-swollen hydrogels typically involves application of a small normal force prior to data acquisition, and/or use of a consistent gap width between the bottom of the sample and the platen. However, fibrin’s viscoelasticity precludes use of the former method, while varying degrees of cell-mediated gel compaction over time preclude the latter. Instead, a protocol was developed whereby the top platen was lowered until it made initial contact with the hydrogel, followed by measurements of shear modulus (G′) taken at 200 μm intervals while closing the gap between platen and stage. Gels exhibited a plateau in G′ as the gap was progressively decreased after making contact with the gel (Supplementary Fig. 1 A/B). The peak G′ measured of 3 gap heights after making contact with the gel was used as our reported value for the given region of interest. One region of interest was interrogated per gel in 24-well plates, and three regions of interest were interrogated per gel in 6-well plates. The measurement head was carefully centered in 24-well plates to avoid edge effects contributing to G′ measurements. Comparisons between acellular gels in 24 and 6-well plates over time (2.5 cm² and 9.8 cm² areas, respectively) were used to confirm the robustness of the methods across gels of different sizes (Supplementary Fig. 1C). Overlay cultures in 6-well plates were also tracked over 14 days to ensure well size did not influence observed stiffening behavior (Supplementary Fig. 2). Overlay and embedded conditions with AMR beads included were also tested to ensure bulk G′ was unaffected by their inclusion (Supplementary Fig. 3). At least 3 measurements were taken per time point per independent experiment. Three independent experiments were conducted (N=3) and for each independent experiment, the gels for all time points were cast from the same stock of reagents.

Laser tweezers-based Active Microrheology

Active microrheology (AMR) was conducted using a dual-laser optical tweezers system, as has been previously described [24] and used in the study of capillary morphogenesis [14]. Briefly, fibrin hydrogels were polymerized as described above within 35 mm glass bottom dishes (MatTek, Ashland, MA) with a dispersion of 2 μm carboxylated silica microbeads (Bangs Laboratories, Fishers, IN) at a concentration of 0.08% (w/v) throughout the hydrogel. During AMR measurements, beads within a volume of approximately 250 × 175 × 30 μm are oscillated at 50 Hz by optical forces induced by a focused 1064 nm laser (trapping beam), at an amplitude of 175 nm. A stationary 785 nm laser (detection beam) was used to detect each probe particle movement in response to the driving force. The oscillation of the input trapping beam and the deflection of the detection beam by the microbead are recorded by a pair of quadrant photodiodes (Newport, Irvine, CA). These measurements allow for calculation of the complex material response α*. Data here is presented as the real component (G′) of the complex shear modulus G*, computed from α*, as previously done [25, 26]. AMR measurements were performed on days 1, 4, 7, and 14 (N=3, per day, per condition) within a custom-built stage top incubator. The volume measured within each measurement location within cell-free samples was chosen randomly, while measurements within capillary morphogenesis assays were chosen to be localized around the sprouting endothelial cells or proximal to Cytodex beads on day 1.

Confocal Reflection Microscopy

Reflection confocal stacks were acquired prior to AMR measurement of each sample. Confocal microscopy was conducted using a Fluoview 1200 system (Olympus), integrated into the optical tweezers microscope. Image stacks were imaged using the 488 nm laser line with a depth of approximately 60 μm and step size of 1 μm. For z-projections, stacks were trimmed to remove effect of glass aberration in reflection confocal and in order to keep a consistent number of planes for each z projection. Images were acquired using the same objective as for AMR; 1.45NA 60X TIRF Oil Objective (Olympus).

Statistics

Varying statistical methods were performed depending on the nature of the data analyzed and are indicated as appropriate on the figure captions. For analyzing network length, the average sprout length per bead from 8–40 beads per time point per replicate was considered for statistical analysis in comparing conditions. Linear regression was used for analyzing network length data for embedded and overlay conditions. For analyzing bulk rheology data, the average bulk modulus from 3–6 ROIs per time point per replicate was considered in comparing conditions. Heteroscedastic 2-tail t-tests were used to verify overlay and Transwell conditions had equitable sprouting on any given day, that AMR beads do not influence bulk G′ or sprouting, and that culture well size does not influence bulk-G′. One-way ANOVA was used to determine differences in bulk mechanical properties and followed with Tukey HSD post-hoc testing if differences were detected. Mann-Whitney U tests with a Bonferroni correction were performed between pairs of aggregate AMR data.

DFs stimulate EC neovessel formation when overlaid or embedded in fibrin matrices

ECs cultured on microcarrier beads embedded in fibrin gels of physiological concentration [27] undergo a complex 3D morphogenetic program that results in vessel-like structures radiating from the microcarrier bead when co-cultured with a variety of stromal cells [16, 18, 28]. Here we demonstrate that normal human dermal fibroblasts (DFs), a clinically relevant and potentially autologous cell source, similarly support angiogenic outgrowth of ECs from microcarrier beads. DFs were either embedded or overlaid on the fibrin matrix to better elucidate the relative contributions of DFs versus EC tubules on the micro- and macro-rheological properties of the gel. Cartoon schematics of these two culture models are shown (Fig. 1A), along with representative images of the typical morphogenetic progression for both culture models (Fig. 1B). (Day 1 images reveal approximately comparable levels of EC confluence on the Cytodex beads achieved via the methods described.) Quantification of these types of images over a 14-day time course reveals an increase in the total length of the vessel-like networks for both conditions, with aggregate data from all beads across each independent experiment shown to illustrate the spread in the observed biological response (Fig. 1C). The average total network lengths for the embedded culture model were consistently higher than the overlay model after day 1, but this increase was not statistically significant. Additionally, inclusion of AMR beads did not affect the rate of network formation (Supplementary Fig. 4). Over the duration of the culture period, the average network length per bead scaled across multiple orders of magnitude with the variance scaling accordingly. Data were log transformed and analyzed with a general linear model. The resulting model is: Log(Network Length) = 3.776*Log(Day). The single regression parameter is highly significant with p < 0.0001 and the model has an R² value of 0.988, demonstrating that vessel growth was exponential in time.

Acellular fibrin gels are mechanically stable over 14 days as assessed by macro-rheology and AMR

To characterize the bulk rheological properties of both acellular (pre-swollen) and cell-seeded fibrin constructs, we devised a new method described in the Supplemental Materials and Methods. Cartoon schematics illustrating the relative scale of bulk rheology compared to AMR are presented in Fig. 2A to illustrate the large difference in resolution between the two techniques. Bulk rheology showed acellular fibrin gels are mechanically stable, maintaining a nearly constant (time-invariant) G′ of 119 ± 19 Pa (mean + st. dev. from all trials and days) over the course of the culture period. Moreover, the inclusion of Cytodex (Fig. 2B) and/or AMR beads (Supplementary Fig. 5) did not influence the bulk properties of the gel. Additionally, agreement between G′ measurements of gels cast in multiple well sizes illustrates the validity of the method developed, independent of gel size (Supplementary Fig. 1). Moving forward, we have used G′ (the elastic component of the shear modulus) interchangeably with the more colloquial term “stiffness.”

AMR measurements of acellular fibrin gels, both with and without Cytodex beads, exhibited notable stiffness heterogeneity spanning nearly two orders of magnitude (Fig. 2C; min value of 8 Pa, max value of 933 Pa across all conditions and time points shown on the graph). Repeatedly, the mean stiffness of these cell-free systems was slightly higher at Day 1, as compared to Days 4, 7, and 14 (p<0.05, for both +/− Cytodex beads). However, comparisons of local G′ values across the later time points were not statistically different from one another, illustrating the mechanical stability of these acellular fibrin gels at the microscale consistent with bulk rheology. Furthermore, the presence of the Cytodex beads did not appreciably affect the distribution of local stiffness, as demonstrated by plotting G′ values at increasing distance from the Cytodex beads (Fig. 2D). Additionally, confocal reflection images near Cytodex beads were indistinguishable from gels without cytodex beads, illustrating that Cytodex beads do not disrupt the fibrillar architecture (Fig. 2E, F).

Bulk rheology reveals tissue constructs stiffen over time during morphogenesis

Fibrin gels with DFs either overlaid or embedded demonstrated significant stiffening over time, with overlay conditions stiffening more rapidly than embedded conditions (Fig. 3A). Controls in which DFs were included but Cytodex microcarrier beads were not coated with ECs demonstrate similar bulk stiffening behavior over time (Fig. 3B). Scaffolds with DFs embedded stiffened ~2-fold to a final G′ of 253 ± 27 Pa, while scaffolds with DFs overlaid stiffened ~3.3-fold to a final G′ of ~394 ± 43 Pa over the course of 2 weeks. Heteroscedastic 2-tail t-tests were used to verify no significant differences occurred between G′ measured for any given culture model and day compared to its respective DF Only control (Fig. 3A, 3B). In addition, experiments in which DFs were cultured on Transwell inserts, and then removed for gel rheological measurements, showed that construct mechanical properties did not change significantly over a 14-day time course in the absence of DFs, even when ECs were included in the culture and tubulogenesis occurred (Fig. 3C). Lack of contact of the DFs with the fibrin construct via the Transwell did not adversely affect capillary sprouting, as revealed by the equivalent total network lengths (Fig. 3D). Together, these data demonstrate that DFs dominate the bulk stiffening of the scaffolds, while the ECs undergoing morphogenesis did not significantly affect the bulk mechanical properties of the constructs in this assay.

Capillary morphogenesis is accompanied by dynamic spatiotemporal changes in local ECM stiffness and organization

Results from selected AMR measurements for both overlay and embedded conditions proximal to the endothelial sprouts show that ECM micro-stiffness changed considerably in both space and time during capillary sprouting (Fig. 4A, 4B, phase). In some cases, AMR beads in close proximity reported G′ values that differed by as much as 5–10x within only a few microns in the same ECM (Fig. 4, arrows). Matched maximum intensity confocal reflection z-projections (Fig. 4A, 4B, confocal reflection) demonstrated how the ECM was simultaneously remodeled. At early time points (Days 1 and 4), a typical fibrin mesh was detected proximal to the ECs. Stepping through assembled z-stacks revealed a distinct high contrast fibrillar architecture surrounding Cytodex beads on Day 1 for both the overlay (Supplementary Video 1) and embedded (Supplementary Video 2) culture models; this architecture is a hallmark signature of fibrin’s microstructure. In the embedded model, however, the ECM was already changing spatially around the embedded fibroblasts. By Day 14, this fibrillar meshwork was replaced with a diffuse signal indicative of remodeling by the cells at later time points (Days 7 and 14). The fibrous structure was difficult to resolve near capillaries in the overlay conditions (Supplementary Video 3), and throughout the dish for embedded cultures (Supplementary Video 4). Interestingly, changes in pericellular ECM architecture over time did not necessarily correlate with regions of elevated stiffness as measured by AMR in the same locations. This was particularly evident in the Day 7 overlay condition, where punctate stiff areas are found both in and out of the remodeled area, though generally appears elevated near sprouts.

AMR quantitatively reveals significant local ECM stiffness heterogeneities during capillary morphogenesis

AMR measurements were conducted for both the overlay and embedded conditions in triplicate, and the results are shown in aggregate (Fig. 5A). Generally, there was an increase in average micro-stiffness over time. As early as Day 1, the embedded condition showed a broader range of G′ values as compared to the overlay condition, with both conditions showing an orders-of-magnitude distribution. Within the embedded condition, this heterogeneity was observed as early as Day 1. Consistent with acellular gels (Fig. 2C), average local stiffness in the overlay condition decreased slightly between Days 1 and 4 (p<0.05), and by Day 14 was similar on average to the embedded condition. To assess the effects of DFs on the peri-endothelial stiffness, AMR was conducted in fibrin gels containing only fibroblasts in both the overlay and embedded conditions (Fig. 5B). In the embedded case, the effects of fibroblast remodeling were evident (increasing G′) by Day 7, whereas in the overlay case, no such effect was observed. In fact, overlay gels softened at Day 14 (p<0.05, compared to all other days), an effect not observed with bulk rheology (Fig. 3B).

To further analyze the spatial distribution of stiffness observed in both the overlay and embedded conditions, AMR measurements were classified as either near or far, based on the proximity of each probe particle (<50μm or >50μm, respectively) to the nearest endothelial tubule (Fig. 5C). Within these conditions, G′ near is significantly greater than G′ far at all time points, except for Day 1 of the embedded case (Mann-Whitney Test, p<0.05). This indicates that the effects of endothelial vessel stiffening of the ECM are concentrated proximal to the vessel, with the effect dissipating with distance. Plots of G′ as a function of probe bead proximity to the ECs (Fig. 5D) show that in the overlay case stiffening is concentrated in a region approximately within 50 μm of the ECs and increases over time. In contrast, the stiffness topography is more heterogeneous at all time points in the embedded case.