Ensemble analysis of angiogenic growth in three-dimensional microfluidic cell cultures

Abstract

We demonstrate ensemble three-dimensional cell cultures and quantitative analysis of angiogenic growth from uniform endothelial monolayers. Our approach combines two key elements: a micro-fluidic assay that enables parallelized angiogenic growth instances subject to common extracellular conditions, and an automated image acquisition and processing scheme enabling high-throughput, unbiased quantification of angiogenic growth. Because of the increased throughput of the assay in comparison to existing three-dimensional morphogenic assays, statistical properties of angiogenic growth can be reliably estimated. We used the assay to evaluate the combined effects of vascular endothelial growth factor (VEGF) and the signaling lipid sphingoshine-1-phosphate (S1P). Our results show the importance of S1P in amplifying the angiogenic response in the presence of VEGF gradients. Furthermore, the application of S1P with VEGF gradients resulted in angiogenic sprouts with higher aspect ratio than S1P with background levels of VEGF, despite reduced total migratory activity. This implies a synergistic effect between the growth factors in promoting angiogenic activity. Finally, the variance in the computed angiogenic metrics (as measured by ensemble standard deviation) was found to increase linearly with the ensemble mean. This finding is consistent with stochastic agent-based mathematical models of angiogenesis that represent angiogenic growth as a series of independent stochastic cell-level decisions.

Figure 1. Device concept and layout.

A) Overall microfluidic device layout. B) The devices contain an array of trapezoidal posts that cage collagen gel solution into well-defined regions with uniform surface interface. During collagen filling, the gel solution-air interface curvature sustains transient filling pressures.The posts are chosen to have an angle of 60°, supplementary to the contact angle of the liquid collagen and treated PDMS surface (found to be 120°). Post spacing is 100–125 microns, and width of gel region is 1.3 mm. C) To enable binding of cells to the gel region, the hydrostatic pressure in the input ports was managed allow for directional flow of the cells along the length of the device, coupled with interstitial flow that biases the cells against the gel region. Consequently cells bind to the collagen region, and form the basis for the monolayer that develops in the ensuing 24 hours. D) Cross section (defined in panel B) illustrating cellular monolayer forms to confluence after growth of adherent seeded cells.

Figure 2. Characterization of device transport characteristics in the absence of monolayers.

Characterization was done via Texas Red conjugated 40 kDa Dextran in lieu of VEGF (which has a molecular weight of 38 kDa) A) Gradients are estimated via fluorescent intensity measurements along the entire gel region B) The generation gradients that are stable in time when device is under flow of 2 µL/min. Gradients are shown stable over a 6 hour time period. C) Numeric simulations reveal concentration profiles along the length of the channel, and confirm gradient profiles similar to those observed experimentally.

Figure 3. Examples of angiogenic response under various conditions.

A) The range of angiogenic responses can be captured in the device from no response to single and multiple sprouts and branches. B) Ensemble observations at basal conditions compared to ensemble observations C) at pro-angiogenic conditions (applied VEGF gradient + S1P). Note that while conditions are applied after the 0 hr images were acquired, variance in initial cellular activity is observed.

Figure 4. Computation of angiogenic metrics.

Input images directly acquired via confocal microscopy are filtered, registered, and were used to determine the boundaries of the gel region, the endothelial monolayer, and coordinate references used for registration and translation between the 0-hr image and the 48-hr condition. These boundaries also define a region of computation of metrics that is slightly offset by 10 microns in order to avoid bias due to the high intensity signals around the monolayer (due to cellular aggregation). Details of image processing steps are described in the Supporting Materials.

Figure 5. Summary of angiogenic growth metrics.

Three measures are shown: A) integral of cytosolic signal M_cyto, B) moment of cytosolic signal J_cyto and C) aspect ratio AR_cyto. Measures at 0 hr (pre-condition) are indicated in blue, and measures at 48 hr are indicated in red (with sign reversed to emphasis asymmetry in growth metrics). Open bars represent the means of each device, with the corresponding markers indicated individual measures over each region. Solid bars represent condition means, with error bars represent +/−1 S.E.

Figure 6. Differential measures of angiogenic growth metrics.

Three measures are shown representing 0 hr–48 hr change in: A) integral of cytosolic signal M_cyto, B) moment of cytosolic signal J_cyto and C) aspect ratio AR_cyto. Open bars represent the means of each device, with the corresponding markers indicated individual measures over each region. Solid bars represent condition means, with error bars represent +/−1 S.E. Measures of statistical significance are computed using Student’s t-test and computed for each condition in relation to its corresponding basal culture. Markers indicate: (* p<0.1), (**p<0.05) and (***p<0.005).

Figure 7. The variance in angiogenic response metrics (based on cytosolic signals) in an ensemble is linearly proportional to the ensemble mean response.

The linear relationship was found irrespective of the growth metric used, and relative slopes between 0 hr and 48 hr measures were maintained.