Matrix anisotropy promotes angiogenesis in a density-dependent manner

Abstract

Angiogenesis is necessary for wound healing, tumorigenesis, implant inosculation, and homeostasis. In each situation, matrix structure and mechanics play a role in determining whether new vasculatures can establish transport to new or hypoxic tissues. Neovessel growth and directional guidance are sensitive to three-dimensional (3-D) matrix anisotropy and density, although the individual and integrated roles of these matrix features have not been fully recapitulated in vitro. We developed a tension-based method to align 3-D collagen constructs seeded with microvessel fragments in matrices of three levels of collagen fibril anisotropy and two levels of collagen density. The extent and direction of neovessel growth from the parent microvessel fragments increased with matrix anisotropy and decreased with density. The proangiogenic effects of anisotropy were attenuated at higher matrix densities. We also examined the impact of matrix anisotropy in an experimental model of neovessel invasion across a tissue interface. Matrix density was found to dictate the success of interface crossing, whereas interface curvature and fibril alignment were found to control directional guidance. Our findings indicate that complex configurations of matrix density and alignment can facilitate or complicate the establishment or maintenance of vascular networks in pathological and homeostatic angiogenesis. Furthermore, we extend preexisting methods for tuning collagen anisotropy in thick constructs. This approach addresses gaps in tissue engineering and cell culture by supporting the inclusion of large multicellular structures in prealigned constructs.

NEW & NOTEWORTHY Matrix anisotropy and density have a considerable effect on angiogenic vessel growth and directional guidance. However, the current literature relies on 2-D and simplified models of angiogenesis (e.g., tubulogenesis and vasculogenesis). We present a method to align 3-D collagen scaffolds embedded with microvessel fragments to different levels of anisotropy. Neovessel growth increases with anisotropy and decreases with density, which may guide angiogenic neovessels across tissue interfaces such as during implant inosculation and tumorigenesis.

INTRODUCTION

Angiogenesis expands vascular supplies during wound healing, normal tissue and tumor growth, reverse microvascular insufficiencies, and tissue engraftment (1–3). For example, thick, metabolically demanding tissue implants fail, in part, due to insufficient methods to prevascularize constructs and establish vascular transport between host and donor/engineered tissues. Thus, there is a simultaneous need to promote therapeutic angiogenesis during healing and inhibit pathological angiogenesis during tumorigenesis and vascular disorders. Recent research indicates that properties of the matrix structure, specifically matrix fibril anisotropy and matrix density, can influence angiogenesis outcomes and may be targets for improved therapies selectively encouraging or discouraging angiogenesis (4, 5).

Matrix fibril anisotropy, how strongly collagen fibrils are oriented in one or two directions, is believed to alter cellular growth and directional guidance because of increased matrix binding site concentration (haptotaxis), anisotropic tensile guidance, and mechanoregulation of signaling receptors (6, 7). The degree of fibril anisotropy (i.e., how strongly aligned fibrils are), in addition to the fibril orientation, can affect endothelial cell migration, polarization, branching, basement membrane formation, and cellular traction (5). These revelations have implications on how angiogenesis progresses during pathologies and wound healing where fibroblasts facilitate local fibril realignment. Tumor growth and expansion depend on recruiting a blood supply via a number of proangiogenic mechanisms. Furthermore, many musculoskeletal and connective tissues are naturally highly anisotropic, which affects angiogenesis during homeostasis and wound healing (8–10).

Microvascular growth and directional guidance are also strongly influenced by matrix density, which influences matrix structural properties including rigidity, permeability, and matrix-binding site concentration. Variations in these properties lead to downstream effects on cellular spreading, adhesion, growth, and migration. Durotaxis is a phenomenon wherein rigidity gradients guide cells toward an optimal substrate stiffness (11, 12). Neovessel sprouting and endothelial cell polarization, behaviors required for growth and migration of the neovasculature, are also slower on dense substrates (13, 14). Furthermore, spatial variations in matrix density inherent in natural tissues, wounds, or pathology can prevent or facilitate angiogenesis (15–17). We previously demonstrated that a high-density, in vitro matrix interface disrupts angiogenic invasion into an adjacent tissue space (18, 19). Given the broad influence of matrix density on neovascular growth and directional guidance, it is important to consider how matrix density and architecture modify microvascular behavior in response to tissue alignment.

Recent studies have highlighted the roles of matrix structure in neovascular growth and directional guidance, but have been limited to simplified models of angiogenesis (e.g., endothelial cell tubulogenesis/vasculogenesis; 5). In vivo, angiogenesis is initiated by endothelial cells that sprout from existing vessels comprising multiple cell types, including endothelial cells and pericytes. Cell-cell, paracrine, and autocrine signaling are fundamental in destabilizing the microvessel (MV) wall to enable sprouting and sprout tip-cell selection (20–22). Furthermore, previous studies on the role of fibril alignment have generally been limited to 2-D or thin constructs and simplified biological models. Finally, to our knowledge, there have been no studies that assess the integrated effects of matrix anisotropy and density on angiogenesis in 2-D or 3-D [although there have been 2-D studies on fibroblasts (23) and 3-D studies on cancer cells (24)]. This is crucial since tumorigenesis, wound healing, and tissue engraftment all involve localized changes in both alignment and density. Thus, we developed a method to align 3-D collagen gels embedded with fragments of intact microvessels (MVs) based on our well-established in vitro model of microvascular angiogenesis. Our model includes lumen-containing MVs comprising endothelial cells, pericytes, and other perivascular cells (25). We cultured MVs in increasingly anisotropic matrices at different densities to characterize the impact of 3-D matrix structure on angiogenic growth and directional guidance. Importantly, we quantify the effect of matrix anisotropy on 3-D growth and guidance as well as the influence of matrix density. We also modified our previously established model of MV crossing between tissue compartments by including prealigned collagen to explore how tissue interface structure affects MV invasion (19).

METHODS

Collagen Gel Alignment

A custom system was developed to align collagen gels via tension (Fig. 1A). Stainless steel anchors were fitted with stainless steel mesh and placed at the opposing ends of rectangular Nunc Lab-Tek four-well chambers. The anchors were held upright with polylactic acid (PLA)+ 3-D printed spacers. Collagen (0.75 mL/well, rat-tail tendon, Corning) was added to each well and allowed to polymerize for 30 min at 37°C. This time was selected as it is near the end of fibrillogenesis before the collagen is fully polymerized (26).

Figure 1.

A: alignment chamber. Steel mesh is held by anchors separated by spacers. Spacers are alternated at 5% increments to the desired strain. B: macroscopic views from SHG images of gels of different density stretched to achieve increasing degrees of anisotropy. C: microscopic views from SEM images of gels of different density stretched to achieve increasing degrees of anisotropy. Individual fibrils are seen dragged in the direction of stretch to achieve orientation. D: polar plot of the average ODF for each level of stretch. Alignment increased with stretch for both densities. E: three distinct strata of anisotropy form because of incremental stretch: low (blue shading), moderate (yellow shading), and high (red shading). *P < 0.05 w.r.t control. One Way KW ANOVA with Dunn’s post hoc. n = 15 images/gel, 3 gel/stretch. ODF, orientation distribution function; SEM, scanning electron microscopy; SHG, second harmonic generation.

Gels, anchors, and spacers were transferred to larger PLA+ 3-D printed chambers. One anchor was held in place, whereas the other was stretched by replacing the initial spacers with larger spacers corresponding to strains between 0% and 20% at 5% strain increments. Control gels were also produced without steel mesh in the anchors to assess the effect of the mesh as a boundary condition. Phosphate buffered saline (PBS) was added to the chambers before stretching to prevent the gels from tearing. Stretched gels were allowed to relax overnight to dissipate residual stresses (27). Steel mesh anchors were removed the following day using a scalpel blade to yield aligned collagen constructs free of residual stresses or strains. Collagen was gelled at low (3 mg/mL) or high (4 mg/mL) density based on prior experiments (4).

Imaging and Analysis of Collagen Alignment

Bright-field imaging: bright-field images of vascularized constructs were collected at each media change using an Olympus IX71 inverted microscope using a ×10 objective. Bright-field images were used to assess MV sprouting during early culture periods.

Second-harmonic generation (SHG) imaging: gels were fixed in 2% paraformaldehyde (PFA; VWR International) overnight then rinsed in PBS before imaging on a Prairie Ultima 2 Photon microscope using a PL APO 25× objective. The center, side (above the center of the gel), and top (end near the mesh anchors) of the gels were selected as regions of interest, and five evenly spaced images were acquired through the full depth of the gel (n = 15/gel, 3 gels/strain, and density level). Images were preprocessed via histogram equalization and a median filter (28). The orientation of the collagen fibrils was approximated using a Fourier transform-based method (29). The transformed images were filtered with qualitatively determined high (2.11 µm) and low pass (65.35 µm) filters to eliminate background noise. The power spectrum was then summed along with the frequencies, and the orientation distribution function (ODF) was approximated by rotating the condensed power spectrum by 90°. The ODF was normalized such that the distribution summed to one. The ODF was then converted to Cartesian coordinates to compute the covariance matrix that fits the ODF into a tensor format that is also an elliptical distribution. The eigenvalues (β₁, β₂) of the tensor were then used to compute the fractional anisotropy (FA) as

$$\text{FA}=1-\frac{{\mathit{\beta }}_2}{{\mathit{\beta }}_1}.$$ (1)

Here, β₁ is the length of the major axis of the ellipse, and β₂ is the length of the minor axis of the ellipse. An anisotropy of zero indicates no alignment, whereas an anisotropy of one indicates complete alignment.

Scanning electron microscopy (SEM): gels were fixed in 2.5% glutaraldehyde overnight, then rinsed and dehydrated with 30-min graded treatments in alcohol at 20% increments. Dehydrated gels were critical point dried and sputter-coated with 30 nm gold/palladium before imaging (FEI Quanta 600 F scanning electron microscope).

Microvessel Culture

A 3-D microvascular organ culture model was used to assess microvascular growth in the aligned collagen gels. MVs were isolated from Sprague–Dawley retired breeder rat epididymal fat pads (13, 19, 25). The fat was minced then underwent limited collagenase digestion before a sequential filtration step, which removes single cells. The MV-collagen suspension was polymerized in custom chambers for 30 min at 37°C with an MV concentration of 60-k fragments/mL unless otherwise specified. At this point, 1 mL of media (rather than PBS) was added to the wells, and gels were stretched to 0% (no anchors), 5%, or 20% strain to induce low, moderate, and highly anisotropic matrices, respectively. Importantly, we noted during pilot studies that if gels were stretched after 30 min, microvascular fragments failed to sprout; thus, there is a time sensitivity of this procedure. Free-floating aligned gels were cultured over 10 days in 30 mm petri dishes with 3.0 mL media (1:1 DMEM & F12, Gibco) supplemented with rhVEGF (10 ng/mL, Shenandoah Labs), 25 mg/L apo-transferring (Sigma, human), 2.5 mg/L insulin (Sigma, bovine pancreas), 6.25 mg/L gentamicin (Thermo Fisher Scientific), and Sato- components (30). Media changes were performed on days 1, 4, and 8.

Immunofluorescence

Cultures were fixed in 2% PFA overnight, permeabilized for 2 h in PBS + 0.5% Triton X-100 (Sigma), and blocked for 2 h in PBS + 3% bovine serum albumin (BSA; Sigma). Microvessels were identified by staining for endothelial cells (isolectin GS-IB₄, Thermo Fisher Scientific, 2 μg/mL, AlexaFluor 488). Nuclei (DAPI, Sigma) were stained and imaged alongside collagen reflectance (n ≥ 2/gel, ∼2,060 × 1,080 × 150–300 µm³, 512 × ∼1,000 × 30–60 px³, 5-µm intervals in the Z-direction). Gel edges were avoided during image acquisition because of inconsistent 3-D geometry and observable edge effects on the initial MV alignment and concentration. All confocal images were obtained using a Leica SP8 white light laser confocal with a PL APO 10×/0.40 objective. Collagen reflectance signal was acquired by setting the values for emission and acquisition to overlap in the near-infrared range.

Microvessel Network Morphometric Quantification

Confocal volumes were processed in Amira (Thermo Fisher Scientific) with a Z-drop correction, median filter, and deconvolution. The deconvolution point spread function was calculated in AMIRA from confocal volumes of 10-µm diameter fluorescent microspheres embedded in avascular collagen gels (Fluoresbrite YG, Polysciences). Small cellular islands were removed using the “remove small islands” feature, and then the remaining volumes were segmented and skeletonized. Short fragments that failed to grow (<60 µm) were excluded from analyses. Microvascular density (mm/mm³) was calculated by dividing the total MV length (mm) in the image by the image volume (mm³). Vessel orientation distribution functions (ODFs) were generated by projecting vessel skeletons onto either the XY or XZ plane and then calculating the length of growth along each direction between 0° and 90° (31). The collagen fibril directions were determined from collagen reflectance images. ODFs were averaged and mirrored across the x-axis to create distributions ranging from −90° to 90°.

Proliferation Assay

Proliferation was quantified using the Click-iT EdU (5-ethynyl-2′-deoxyuridine) Alexa Fluor 594 Imaging Kit according to the manufacturer’s protocol (Thermo Fisher Scientific). EdU was added to the media at 10 μM on day 6 and incubated for 24 h before gels were fixed and stained on day 7, as done previously (32). Gels were imaged to a depth of 25 μm at 2.5-μm intervals. Max-intensity Z-projections were thresholded with the maximum entropy algorithm in ImageJ. Proliferation was quantified as the number of EdU⁺ pixels over the number of Hoechst⁺ pixels. Neovessels and parent vessels were calculated separately with parent vessels identified by highly concentrated regions of unpolarized (approximately circular) nuclei. Only low-density collagen was used studied for proliferation studies.

Branch Angle Analysis

The angle between new branches and parent MVs was determined from max-intensity Z-confocal projections collected for the proliferation assay. These data were chosen since the morphometric volumes were much larger and more difficult to distinguish vessels from each other. The branch angle was calculated using the angle tool in ImageJ.

Aligned Interface Invasion

We previously developed a 3-D interface invasion assay comprising a central core and an external field of collagen. A structural interface characterized by increased matrix density and altered fibril alignment spontaneously forms between the core and field during polymerization (19). To explore the effects of collagen fibril anisotropy during microvascular invasion and deflection, we modified this assay to include isotropic or anisotropic collagen gels in the core. Briefly, a thin layer of avascular collagen (350 μL) was gelled in a 24-well plate to establish an adhesive base. Cores were stretched to 0% (no anchors), 5%, or 20% strain, incubated overnight, and then bisected along the long axis using a scalpel blade. Stretched gels were transferred to the top of the thin collagenous base, and a field of collagen (500 μL) was polymerized around the cores. We explored two scenarios: either the core or the field was vascularized (Fig. 6A). Vascularized cores were created to assess how MVs in anisotropic tissues respond to interfaces with isotropic tissues, whereas vascularized fields were created to assess how MVs in isotropic collagen navigate interfaces to anisotropic tissues. MVs were seeded at 40,000 fragments/mL for these assays due to the high level of growth observed with the 60,000 fragments/mL cultures. All modified interface gels were cultured until day 10 when MV interface crossing was observed in the majority of cultures. The crossing was identified by stepping through image stacks until vessels visibly crossed tissue compartments rather than grew along with them.

Figure 6.

A: aligned invasion assay schematics. A collagen core containing either aligned or unaligned collagen was suspended in a field of unaligned collagen. MVs were included in either the field or the core observe guidance, invasion, and deflection at the interface. Confocal visualizations of aligned interface experiments (B–H). Green: vessels. Gray: collagen. Arrows: direction of collagen alignment. Caps: MV crossing. Scale bar = 300 μm for all. B: interface structure (red dashed line) in the vertical direction varied between sharp walls (left) and gradual slopes (right). MVs approach vertical interfaces then grew along with them in the XY plane (bottom left), whereas MVs grew vertically along with gradually sloped interfaces (bottom right). Z projections of vascularized field experiments (C–F). MVs crossed near less-dense (less intense) collagen. C and D: MVs deeply penetrated the core when the MV and core fibril direction were coaligned. E and F: MVs reoriented quickly after crossing into the core when the core fibril direction was perpendicular to the interface. G and H: Z projections of vascularized core experiments. G: MVs follow the fibril direction out of the core. H: substrate curvature dominates fibril guidance and facilitates crossing perpendicular to the collagen direction. MV, microvessel.

Statistical Analysis

All statistical analyses were performed in Origin 2020 b (OriginLab). Avascular gel anisotropy and proliferation measurements were analyzed using one-way ANOVAs on ranks (Kruskal–Wallis, i.e., KW). Box plot data were centered about the median with box edges at 25% and 75% and whiskers at 1.5 × interquartile range. Microvascular densities were assessed with a two-way ANOVA assessing the effects of matrix anisotropy and matrix density. The Holm–Sisak post hoc test was used for multiple comparisons. Data are presented as data points for each image overlaid on a box with lines at 25%, 50%, and 75% of the data. A P value of α ≤ 0.05 was used for all statistical tests. MV growth experiments comprised multiple gels from two to three separate experimental replicates, each using MVs from separate animals.

RESULTS

Collagen Gels Are Aligned by Strain during Polymerization

Collagen fibril anisotropy in vivo varies between entirely randomly oriented fibrils (low anisotropy i.e., isotropic) and entirely aligned fibrils (high anisotropy). We developed a tension-based system to tune the degree of anisotropy in collagen gels to better understand how various levels of 3-D fibril anisotropy influence neovessel growth and directional guidance from microvessel fragments. SHG and SEM images revealed that anisotropy increased with the degree of strain during polymerization (Fig. 1, B and C). Control gels (no anchors) showed short fibrils with occasional bundling. Small, localized patches of oriented fibers are observable, but there is no apparent global orientation. Global orientation is visible at intermediate levels of strain (0%–10%) though there is pronounced variation. Fibril bundles appear less frequently at intermediate levels of strain. Highly strained gels (>10%) displayed near uniaxial orientation in SHG images, whereas subtle variations are apparent in SEM images. In both SHG and SEM images, the highly strained gels appear to have longer, persistent fibrils. High-density gels generally followed the trends of low-density gels but appeared more tightly packed than low-density gels in SHG images.

Orientation distributions derived from SHG images support qualitative observations and identify three distinct levels of fibril anisotropy that can be achieved with our approach at both low (3 mg/mL) and high (4 mg/mL) density (Fig. 1, D and E). Although fibrils in the control group (no anchors) were nearly uniformly distributed, 0% strain gels displayed slightly elevated anisotropy. In this case, it is possible, the anchors act as boundary constraints or anisotropy is induced by anchor removal. The median fractional anisotropy (FA) of low-density control gels was 0.23 compared with ∼0.4 for 5% strained gels. Gels strained to 10% show an increase in anisotropy over lower strain levels but also a wider range of anisotropies. This level of strain appears to represent a transition between moderate and high levels of anisotropy. Gels strained 15% or greater showed even higher levels of anisotropy with FA values above 0.6. Strains beyond 10% yielded significantly higher levels of anisotropy with a median of 0.68 and 0.64 for 15% and 20% strain, respectively. The middle 50% of the data were comparable regardless of matrix density. Based on these results, anchorless control gels, 5% strain, and 20% strain were used to study the effects of low, moderate, and high anisotropy, respectively on MV growth.

Neovessel Growth and Directional Guidance during Angiogenesis Increase with Matrix Anisotropy but Growth Decreases with Matrix Density

To determine the effects of matrix anisotropy on neovessel growth and directional guidance during angiogenesis, we suspended MV fragments in collagen and stretched the gels to induce low (0% strain, no anchors), moderate (5% strain), or high matrix anisotropy (20% strain). Neovessel sprouts emerge from parent MVs, which can branch and anastomose with other neovessels. We found that MV growth and directional guidance significantly increased with matrix anisotropy and decreased with matrix density (P < 0.001 each; Fig. 2, A and B). There were no interaction effects between anisotropy and density (P = 0.233). Interestingly, there was no additional increase in growth between moderate and highly anisotropic matrices (P = 0.421). Furthermore, moderate anisotropy was associated with only an increase in growth in low density (P < 0.001), but not in high-density matrices (P = 0.121), indicating that the proangiogenic effects of fibril anisotropy are sensitive to matrix density (i.e., higher levels of anisotropy may be needed to promote growth in dense tissues). To ensure differences in MV density were not due to stretching or contraction during alignment, vascularized gels were fixed, stained, and imaged on day 1 (Fig. 3). There was no initial difference in MV density (P = 0.386) or alignment in the center of gels, indicating that the MV density measurements on day 10 are likely not affected by stretching methods.

Figure 2.

Matrix anisotropy increases microvascular growth and alignment at multiple matrix densities. A: representative max intensity Z projections (200 µm depth) of confocal images, and MV density measurements from MVs grown in aligned matrices. Yellow lines: direction of alignment. B: MV density is affected by both density and anisotropy. Two Way ANOVA, Holm–Sidak post hoc. n ≥ 11 images/group. *P < 0.05 w.r.t same density control (low FA). C: microvessel orientation distributions in the horizontal (XY) and vertical (XZ) planes. Microvessel polarization increases with matrix anisotropy. Matrix density does not attenuate microvessel polarization in the XY plane. Microvessel polarization also increases in the XZ plane and does appear to be attenuated by density. FA, fractional anisotropy; MV, microvessel.

Figure 3.

A: microvascular density of fragments 1 day after seeding. There is no effect of stretch on the measured vascular density. One Way ANOVA. P > 0.25 for all. B: microvessel ODFs in the XY and XZ planes 1 day after seeding. Orientation is generally unaffected by stretch, although there is a slight increase in horizontal alignment due to stretch. C: representative max intensity images taken 1 day after seeding. FA, fractional anisotropy; ODFs, orientation distribution functions.

MV directional guidance was assessed based on projections of MV network skeletons onto the horizontal (XY) and vertical (XZ) planes (Fig. 2C). MV orientation in the horizontal plane increased incrementally with the degree of anisotropy but was not greatly affected by matrix density. MVs in low anisotropy gels were approximately uniformly distributed. Moderate anisotropy resulted in a realignment of MVs along the fibril direction. High anisotropy led to even stronger realignment and growth of MVs along the fibril direction than the moderate anisotropy. MV orientation in the vertical plane was highly polarized along the horizontal direction for all cases. Noticeable differences only occur in the range ±10° from the horizontal direction. Here we see that there is slightly more concentration of MV orientation along the horizontal direction than the vertical direction due to matrix density. These results agree with our previously reported studies, where growth generally does not occur in the vertical direction likely because of the lateral flow of collagen during gelation (33).

Neovessel Sprouting and Branching Are Affected by Increasing Levels of Anisotropy

MV sprouts forming from parent fragments during the first few days of culture reoriented along the collagen direction in an anisotropy-dependent manner. Sprouts in isotropic gels often persisted along the parent vessel or parent branch’s initial direction (Fig. 4, top). With moderate anisotropy, sprouts were observed initially persisting along their initial direction before realigning to the fibril direction. Sprouts in high anisotropy gels, however, almost immediately realigned with the fibril direction. This is pronounced in cases where the parent MV fragments lay perpendicular to the collagen fibril direction, yet sprouts run along the fibril direction.

Figure 4.

Top: brightfield imaging of sprout formation around day 2 of culture. Sprout reorientation occurs quicker with increasing matrix anisotropy. Arrows: sprout tips. Pink line: collagen direction. Low anisotropy generates sprouts that persist along the initial direction of the parent vessel. Sprouts almost immediately reorient along the collagen fibril direction in high anisotropy matrices. Middle: confocal Z projections of MV growth at D7 (25 μm depth). Bottom: the angle of branch departure from the parent vessel is approximately uniformly distributed between 30° and 90° for low anisotropy matrices. For medium anisotropy matrices the angle is biased between 30° and 60°. High anisotropy yields a bimodal distribution of departure angles centered at either 30° resulting in branches parallel to the parent vessel or at 90° due to anastomoses between adjacent MVs. MV, microvessel.

The angle that new sprout branches departed from the parent MV was affected by the degree of anisotropy. In isotropic matrices, the branch angle was never less than 30° (Fig. 4, middle and bottom). In contrast, moderate matrix anisotropy yielded branches angled less than 30° from their parent vessel. As more vessels align with the fibril orientation, their branches emerge and are quickly reoriented to assume the same direction as their parent vessels. Interestingly, in high anisotropy matrices, branches were frequently found perpendicular to the parent vessel and collagen direction. These branches generally participated in anastomoses between adjacent MVs. The bimodal distribution of branch angles in high anisotropy collagen indicates that branches quickly assimilate to the collagen/parent fragment direction unless another vessel is nearby. In those cases, neighboring sprouts can locally influence each other and form anastomoses perpendicular to the collagen/parent fragments. Thus, high matrix anisotropy generally results in shallow branch angles, but signals between adjacent MVs during anastomosis are able to overcome fibril guidance cues. Whether this is due to paracrine signaling or mechanical cues between adjacent vessels is unclear.

Neovessel Sprouting and Endothelial Cell Proliferation Are Affected by 3-D Matrix Anisotropy

Proliferation increased in moderate and high anisotropy gels compared with isotropic gels. We calculated the percentage of proliferating cells in neovessels, parent vessels, and the combined proliferation to determine how anisotropy affected each member of the microvasculature (Fig. 5). The combined parent fragment and neovessel proliferation increased for both moderate and high levels of anisotropy compared with isotropic controls. Moderate anisotropy led to the largest variance in proliferation, and only parent vessels experienced a significant increase compared with low anisotropy. High anisotropy was associated with increased cell proliferation for both neovessels and parent microvessels. Proliferation in low anisotropy gels was in the range previously measured for MV angiogenesis in isotropic collagen gels after 7 days of culture (32).

Figure 5.

A: the percent of proliferating cells (EdU⁺) on day 7 increased with matrix alignment. This trend was observed in both sparsely populated neovessels and densely populated parent vessels. One way KW ANOVA, Dunn’s post hoc *P < 0.05 w.r.t control in the same group. n = 16 images/group, 4–8 gels/group. B: representative maximum intensity Z projections (25 μm depth) of microvessels (lectin) after 7 days of culture. The fraction of proliferating nuclei (EdU⁺) compared with all nuclei (Hoechst⁺) cells increased with alignment. Nuclei were brightened for figure clarity. FA, fractional anisotropy.

Aligned Interface Invasion

Previously, we developed an in vitro invasion assay to study neovessel crossing and deflection at interfaces between tissue compartments during angiogenesis (19). Briefly, a core of collagen seeded with MV fragments was placed inside a field of avascular collagen. We termed this model “core in field” (CIF). A high-density interface spontaneously formed between the core and the field during polymerization. We observed that growing neovessels approached the interface then appeared to regress or deflect (grow along with the interface), failing to infiltrate the adjacent matrix compartment. Structural imaging indicated that the interface comprised a high-density, moderate anisotropy barrier between the core and the field. We now present a modified version of our CIF model that includes anisotropic collagen cores. This modification allows us to determine how fibril anisotropy and orientation influence microvascular crossing and deflection. In addition, the cores used for the present study are rectangular rather than circular. Previously, all the interfacial collagen was likely circumferentially aligned. In our new model, the collagen is either isotropic near the initially anchored edges or aligned along the initially unconstrained edges. Importantly, this allowed us to study the two configurations of matrix fibril orientation: 1) collagen fibrils oriented along the direction of invasion, and 2) collagen fibrils oriented tangential to the direction of invasion. We also performed a parallel set of experiments, where MV fragments were suspended in the field rather than the core to understand how fibril orientation on each side of the interface affects invasion and deflection (Fig. 6A).

Confocal reflectance images revealed sharp differences in matrix structure at the interface. Generally, the interface contains high-density collagen proximal to the core and thin, low-density collagen proximal to the field. The interface was either steeply or shallowly curved in the vertical direction (Fig. 6B). Unexpectedly, interface vertical curvature strongly impacted vessel directional guidance rather than matrix density or fibril orientation alone, a result not seen without 3-D imaging. When vessels encountered steep, nearly vertical interfaces, neovessels generally deflected at the interface, growing predominantly in the horizontal plane along with the collagen fibril orientation and interface surface. However, when vessels approached gradually slanted interfaces, they followed the curvature of the interface in the vertical direction. Vertical growth was not observed in our previous experiments in free-floating anisotropic cultures, indicating that substrate curvature may more potently guide neovessels during angiogenesis than fibril anisotropy.

Although the collagen density at the interface generally appeared higher than the surrounding matrix, small tears and bubbles occasionally formed near the interface, which created localized pockets of low density. Most neovessel interface crossings occurred near such regions of low density, indicating that invasion is predominantly influenced by collagen density (Fig. 6, C–H). When neovessels crossed into unaligned collagen they penetrated deeply regardless of the region of origin (Fig. 6, C, D, and G). In contrast, when collagen was oriented perpendicular to the invasive direction, neovessels quickly reoriented and grew along the collagen direction (Fig. 6, E and F). Furthermore, we observed one instance where a neovessel crossed into the core and then branched, with each neovessel tip growing in opposing directions along the collagen fibrils (Fig. 6F). We also observed cases where the vertical interface curvature running perpendicular to the fibril orientation guided neovessels vertically along with the interface, against the collagen direction, until they crossed the interface at a region of low density (Fig. 6H). Generally, these findings suggest that collagen density determines where interface crossing occurs, whereas fibril orientation and substrate curvature determine how deeply vessels penetrate after crossing.

DISCUSSION

Angiogenesis and Vascular Cell Proliferation Increase with Matrix Anisotropy in a Density-Dependent Manner

We demonstrate that microvascular growth and directional guidance increase with matrix anisotropy in a density-dependent manner. At low collagen densities, moderate and high anisotropy led to similar increase in neovessel growth. At high collagen densities, high anisotropy, but not moderate anisotropy, resulted in a significant increase in neovessel growth. Thus, the threshold of anisotropy needed to enhance angiogenesis increases with matrix density. Cell proliferation increased in both parent MVs and neovessels in response to high anisotropy. In contrast, moderate anisotropy only resulted in a significant increase in cell proliferation in parent MVs but not neovessels. This may be due to the much higher concentration of cells within the MVs that may amplify matrix cues via cell-cell signaling. Furthermore, proliferation is common in stalk cells, but not tip cells, indicating that tip/stalk dynamics, may be differentially affected by anisotropy and density (34).

Mechanical and biochemical mechanisms may both be responsible for observed MV responses to differing matrix density and alignment. Durotaxis may be affected by changes in matrix compliance. For instance, the ratio of axial and lateral matrix rigidity may change at different matrix densities despite similar degrees of fibril alignment. Thus, moderate anisotropy may have reduced durotactic effects at high matrix densities. Furthermore, haptotactic guidance likely decreases since more matrix binding sites will be available in the lateral direction when matrix density increases. It is also possible that density-dependent mechanoregulation of proangiogenic signaling pathways attenuates neovessel growth. Rivron et al. (35) demonstrated that tissue contractility and deformation (both inversely proportional to density) boost endothelial cell vascular endothelial growth factor (VEGF) A production, VEGF receptor 2 activity, and downstream proliferation. VEGF internalization has been shown to increase with reduced matrix stiffness. Similarly, VEGF receptor 2 has been shown to recycle to the cell surface after endocytosis more frequently on soft matrices, maintaining VEGF-responsiveness in these environments (36, 37).

Although we did not detect any differences in the initial microvascular density or orientation due to stretch, it is likely that there could be some change in the apparent density of the collagen after stretching. The immediate Poisson’s ratio of collagen gels ranges between 0.4 and 0.5 for 2–4 mg/mL gels, which would indicate lateral compaction (38), whereas the equilibrium Poisson’s ratio is close to zero (31, 39). Furthermore, collagen gels are poroelastic, highly hydrated, and are stretched in 5% strain increments before polymerization finishes. Thus, the immediate compaction due to stretching is expected to dissipate at equilibrium.

Directional Guidance Primarily Depends on Matrix Anisotropy in a Simplified Model of Microvascular Sprouting

Directional guidance incrementally increased with the degree of anisotropy but not matrix density. We initially hypothesized that directional guidance would be hindered by denser matrices since neovessel growth is reduced, and thus the orientation of randomly aligned parent fragments would have a greater contribution to the MV network ODF. However, there was no appreciable difference in the average MV network alignment based on matrix density (Fig. 2C). Previous studies on the interactions between matrix density and VEGF gradients provide insight into these observations. For example, Shamloo et al. (40) cultured endothelial cells on spheres embedded in collagen at different matrix densities in the presence of a VEGF gradient. Sprouts in high-density matrices were more likely to polarize along the VEGF direction, which they attributed to slower sprout growth. They reasoned that sprouts in soft matrices quickly penetrated the matrix before guidance cues could be integrated. Sprouts in dense matrices grow slower and thus reorient at shorter distances. Similarly, it is possible here, there was no major change in MV network ODFs because sprout reorientation occurred at shorter distances from the parent MV in dense matrices. Interestingly, studies of fibroblasts grown on aligned 2-D substrates similarly found that fibril alignment had a much greater impact on cellular trajectories than fibril density (23). Other explanations for the consistent directional guidance across densities include density-dependent increases in gradients of matrix binding proteins and tensile anisotropy, as previously discussed.

Interface Density Rather than Curvature or Anisotropy Dictates Microvascular Invasion

During therapeutical and pathological angiogenesis, growing neovessels must cross tissue interfaces that are characterized by gradients and discontinuities in matrix orientation, anisotropy, and density. Mechanisms by which interface structure facilitates or prevents tissue invasion may reveal treatment options pertaining to natural interfaces between tissues as well as interfaces introduced by injury, pathology, grafting, or implantation. We previously probed these mechanisms with an in vitro MV invasion assay (19). High interface collagen density led to neovessel deflection, whereas fibril anisotropy guided neovessels along with the interface. Modifications to our assay in the present study reaffirm our prior findings and led to the unexpected finding that interface substrate curvature plays a considerable role in guidance during the invasion.

MV interface crossing generally occurred in locations of lower matrix density. This occurred regardless of matrix anisotropy or orientation as indicated by lower intensity in the collagen signal when stepping through subsets of the entire confocal volumes (Fig. 6). Moreover, when the core collagen was oriented outward into the field, vessels that crossed, penetrated deeply into the core. This configuration of collagen alignment at the interface may mimic invasive tumorigenesis where fibroblasts align collagen to attract microvessels (15). However, such a configuration may be beneficial for encouraging invasions such as during implant inosculation or tendon repair (41). In contrast, when the core was aligned with collagen oriented the same direction as the interface, vessels did not penetrate far beyond the interface but rather continued along the collagen direction peripheral to the interface.

When vessels did not cross, their trajectory was affected first by interface topology, then by matrix anisotropy. MVs that approached sharp vertical interfaces deflected sharply in the horizontal direction along with the interface. However, when vessels encountered sloped interfaces, they only adjusted their trajectory in the vertical direction. These observations, along with observations that crossings generally happen near low-density collagen, suggest that durotaxis plays a role as vessels navigate complex tissue interfaces. Interestingly, vessels appeared to grow more within lower density collagen while maintaining contact or proximity to the higher-density core. Such positioning provides directional cues while simultaneously providing cells with a degree of freedom and conformity to maintain migratory phenotypes. Similar interfaces may be present during wound healing and implantation that will need to be overcome for the vasculature to successfully establish connections between healthy and hypoxic tissues.

Synthetic Aligned Interfaces Mimic Tumor Associated Collagen Signatures

Clinically, increased matrix density and alignment of collagen fibrils are hallmarks of cancer termed tumor-associated collagen signatures (TACSs; 15, 42–44). One type, TACS-1 is characterized by increased collagen density at or near a tumor. As tumors grow, two other TACS types can present. TACS-2 is associated with taut collagen fibrils running tangent to the tumor surface. This alignment is thought to arise because of the expansion and constraining of the stroma as the tumor grows. TACS-3 is identified by branched collagen radiating perpendicular to the surface of the tumor. This alignment is associated with highly invasive tumor behavior.

Our aligned interface model reflects features of all three TACSs. Matrix density is elevated along with the interface as indicated by elevated signal intensity (19). Tangential fibril alignment along the long edge of the cores in our interface model resembles TACS-2, whereas perpendicular alignment along the short edge of the cores resembles TACS-3. MVs near TACS-2-like interfaces were deflected or failed to grow deep into the core before reorienting, a result seen in our prior CIF models. However, MVs that crossed at TACS-3-like interfaces deeply invade into the adjoining matrix. Our findings indicate that matrix alignment may contribute to tumor progression by 1) directing microvessels and tumor cells toward each other, and 2) enhancing MV proliferation and growth. Further investigation into these phenomena is needed to probe molecular mechanisms responsible for the observed behaviors.

These behaviors were observed using fat-derived MVs rather than tumor-derived MVs, which may respond differently to matrix cues. However, we have previously demonstrated that MV behavior is more sensitive to the cellular composition of the construct it is cultured in than the tissue of origin (45). Nonetheless, our aligned interface model can be expanded upon to include stromal cells and tumor cells to investigate the coordination of complex, dynamic tissue-related signaling during tumorigenesis.

Microvessel Growth and Guidance Have Similar Adaptations to Differing Mechanical Cues

We previously observed the coalignment of MVs and the extracellular matrix in response to cyclic strain or gel geometric boundary constraints (33). Collagen constructs seeded with MVs were either stretched in tension during culture or two parallel edges of the rectangular constructs were held fixed. Over the duration of culture, we observed MV reorientation along either the direction of stretch or toward the boundary constraints. As the MVs grew they contracted the gels in the direction lateral to growth, leading to aligned collagen fibrils. Although these previous studies mainly addressed durotaxis (guidance by matrix compliance gradients), we believe the current study likely addresses haptotaxis (guidance by matrix binding site gradients) to a greater degree. This is due to prealignment of collagen gels and culturing without construct geometric constraints or strains. Interestingly, we see comparable qualitative and quantitative increases in neovessel growth and directional guidance due to durotactic and haptotactic cues. It remains unclear whether these similarities are due to shared signaling mechanisms or just coincidence.

The results of the present study reinforce findings from previous 2-D and thin 3-D studies, which have found that increased anisotropy promotes endothelial growth and directed migration (5). Furthermore, we observed that microvessel-directed migration only varied with fibril alignment but not matrix binding site density, a result observed in fibroblasts cultured on aligned fibronectin substrates (23). Interestingly, our prior studies of MVs cultured in mechanically loaded and constrained 3-D gels differed significantly from 2-D studies in the literature. In our 3-D cultures, we saw MVs grow along the direction of tensile strain, contrasting with 2-D studies that observed endothelial cell growth perpendicular to the direction of tensile strain (33, 46, 47). Thus, mechanisms of durotaxis may be more sensitive to scaffold dimensionality than haptotaxis.

Tension is a Low-Cost, Straightforward Approach for Aligning Fibrillar Tissue Engineered Constructs

Current methods to align matrix fibrils in engineered constructs may be costly, cytotoxic, or require scaffold deformations that are harmful to cells. Magnetic fields have emerged as a potential path where a high, homogenous degree of anisotropy can be achieved with low cytotoxicity (48, 49). However, this approach requires access to large magnets (>5 + T), and the degree of anisotropy has not shown to be tunable. Others have aligned collagen by including small magnetic beads, which flow through the gel during polymerization (50). Furthermore, the fibril alignment decreases with the inclusion of physiological levels of glycosaminoglycan or when the collagen density is high, limiting practical applications of these scaffolds (48, 49). Electrospinning has also become popular due to the high degrees of anisotropy that can be achieved and the ability to control properties such as fibril diameter. However, the process is cytotoxic, and large multicellular structures such as MVs cannot be suspended in the material but would rather be required to invade from the outside (51). Compression and extrusion are cost-effective and straightforward approaches to aligning fibrillar hydrogels, but they may not be suitable for all cell types such as endothelial cells, which struggle to grow after compression (52–54). Finally, cell-driven traction and compaction in tandem with geometric boundary constraints are known to align collagen gels, although this approach does not enable one to tune the degree of anisotropy (31). Such approaches may be suitable for tissue-engineering applications, but the lack of a priori alignment may confound studies of cellular physiology and cell-matrix interactions such as our present study.

Tension-based approaches to alignment, however, are economical, enable a tunable degree of anisotropy, can be achieved at low and high collagen densities, and allow suspension of cells and large multicellular structures. Previously, tensional-based alignment has been achieved at small scales or low collagen densities (5, 55, 56). Riching et al. (24) extended these approaches and prealigned thick, dense collagen hydrogels on the centimeter scale before time-series imaging of cancer cells. Our approach is distinguished from theirs in that 1) stretch is applied before the end of culture, and 2) gels are cut from mesh anchors after 1 day of culture. Mechanical testing is needed to determine how the timing of stretch affects the structural and mechanical properties of prealigned gels. We found that stretching after polymerization resulted in poor MV growth; thus, the timing of stretch likely affects the viability of multicellular constructs and organoids. The choice to keep mesh anchors during culture creates confounding mechanical stimuli through geometrical boundary constraints. However, anchors are likely needed to maintain a consistent field of view during time-series images.

Conclusions

Through a modified method to tune collagen fibril anisotropy, we demonstrated that MV growth and directional guidance increase with the degree of alignment in a density-dependent manner. We used this method to extend a prior model of MV invasion to include gradients in matrix density and orientation that can be used to study cell-matrix interactions during tumorigenesis and wound healing.

ETHICAL APPROVALS

The animal study was reviewed and approved by University of Utah’s Animal Care and Use Committee.

GRANTS

This study was funded by National Institutes of Health Grants R01HL131856, R01GM083925, R01AR069297, and R01AR071358.

DISCLAIMERS

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

DISCLOSURES

J.B.H. is a shareholder with Advanced Solutions Life Sciences. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

S.A.L., S.S.D., J.B.H., H.A.S., and J.A.W. conceived and designed research; S.A.L., S.S.D., E.V.B., and A.H.L. performed experiments; S.A.L., S.S.D., A.R., A.H.L., and J.A.W. analyzed data; S.A.L., J.B.H., A.R., H.A.S., and J.A.W. interpreted results of experiments; S.A.L., S.S.D., and A.H.L. prepared figures; S.A.L. and J.A.W. drafted manuscript; S.A.L., S.S.D., J.B.H., E.V.B., A.R., H.A.S., A.H.L., and J.A.W. edited and revised manuscript; S.A.L., S.S.D., J.B.H., E.V.B., A.R., H.A.S., A.H.L., and J.A.W. approved final version of manuscript.