3D quantitative analyses of angiogenic sprout growth dynamics


  • Abbas Shirinifard,

    1. Biocomplexity Institute and Department of Physics, Indiana University Bloomington, Bloomington, Indiana
    2. Department of Clinical Informatics, St. Jude Children's Research Hospital, Memphis, Tennessee
    Search for more papers by this author
  • Catherine W. McCollum,

    1. Center for Nuclear Receptors and Cell Signaling, Department of Biology and Biochemistry, University of Houston, Houston, Texas
    Search for more papers by this author
  • Maria Bondesson Bolin,

    1. Center for Nuclear Receptors and Cell Signaling, Department of Biology and Biochemistry, University of Houston, Houston, Texas
    Search for more papers by this author
  • Jan-Åke Gustafsson,

    1. Center for Nuclear Receptors and Cell Signaling, Department of Biology and Biochemistry, University of Houston, Houston, Texas
    Search for more papers by this author
  • James A. Glazier,

    1. Biocomplexity Institute and Department of Physics, Indiana University Bloomington, Bloomington, Indiana
    Search for more papers by this author
  • Sherry G. Clendenon

    Corresponding author
    • Biocomplexity Institute and Department of Physics, Indiana University Bloomington, Bloomington, Indiana
    Search for more papers by this author

Correspondence to: Sherry G. Clendenon, Ph.D., Biocomplexity Institute, Department of Physics, Indiana University, Simon Hall MSB1, 047, 212 S. Hawthorne Drive Bloomington, IN 47405-7003. E-mail: sgclende@indiana.edu, sherry.clendenon@gmail.com


Background: Zebrafish intersegmental vessel (ISV) growth is widely used to study angiogenesis and to screen drugs and toxins that perturb angiogenesis. Most current ISV growth assays observe the presence or absence of ISVs or perturbation of ISV morphology but do not measure growth dynamics. We have developed a four-dimensional (4D, space plus time) quantitative analysis of angiogenic sprout growth dynamics for characterization of both normal and perturbed growth. Results: We tracked the positions of the ISV base and tip for each ISV sprout in 4D. Despite immobilization, zebrafish embryos translocated globally and non-uniformly during development. We used displacement of the ISV base and the angle between the ISV and the dorsal aorta to correct for displacement and rotation during development. From corrected tip cell coordinates, we computed average ISV trajectories. We fitted a quadratic curve to the average ISV trajectories to produce a canonical ISV trajectory for each experimental group, arsenic treated and untreated. From the canonical ISV trajectories, we computed curvature, average directed migration speed and directionality. Canonical trajectories from treated (arsenic exposed) and untreated groups differed in curvature, average directed migration speed and angle between the ISV and dorsal aorta. Conclusions: 4D analysis of angiogenic sprout growth dynamics: (1) Allows quantitative assessment of ISV growth dynamics and perturbation, and (2) provides critical inputs for computational models of angiogenesis. Developmental Dynamics 242:508–516, 2013. © 2013 Wiley Periodicals, Inc.


Zebrafish are a classic model system for studies of vascular development and of perturbations that affect vascular development (Gore et al., 2012; Quaife et al., 2012). Molecular mechanisms that govern vascular development are highly conserved, with genetic pathways relevant to human vascular development conserved between fish and humans (Dooley and Zon, 2000; Zhong, 2005). Further, zebrafish have a short development time, high fecundity, extracorporeal development, and small size. These traits, coupled with the availability of transgenic and mutant lines and their ease of genetic manipulation make zebrafish ideal for medium- to high-throughput screens of drugs and for identification of developmental defects and toxicity (McCollum et al., 2011; Parng et al., 2002; Rubinstein, 2006; Serbedzija et al., 1999; Sipes et al., 2011; Zon and Peterson, 2005).

Vascular development occurs in two phases: vasculogenesis, the de novo formation of vessels, and angiogenesis, the sprouting of new vessels from preexisting ones. In zebrafish, angioblasts migrate medially from the lateral mesoderm and coalesce to form a longitudinal axial vessel, the dorsal aorta (DA) by vasculogenesis (Risau and Flamme, 1995). The posterior cardinal vein (PCV) then forms by sprouting and segregation from the DA (Herbert et al., 2009). Subsequent angiogenic sprouting, migration, and proliferation of endothelial cells then give rise to the intricately patterned vasculature.

Intersegmental vessel (ISV) development in zebrafish follows a tightly conserved pattern with pairs of ISVs sprouting bilaterally from the dorsal aorta (DA) at regular spatial and temporal intervals in an anterior to posterior sequence (Fig. 1A) (Childs et al., 2002; Isogai et al., 2003). Morphogenesis of ISV sprouts involves multiple complex cell behaviors and is essential for subsequent development and nutrition of the embryo. When ISV growth initiates, the leading cell in the sprout (the tip cell) extends and retracts numerous filopodia that explore the environment as the sprout grows (Isogai et al., 2003). Newly formed sprouts migrate dorsally, medial to the somites and adjacent to the notochord, within the extracellular matrix-rich somite boundaries.

Figure 1.

ISV sprout growth and patterning is disrupted by arsenic exposure and this disruption can be quantified from time-lapse images. A: At 24 hr post-fertilization, ISVs in the trunk region (adjacent to the yolk-plug extension) have begun sprouting bilaterally from the dorsal aorta at regular spatial and temporal intervals, with an anterior (left of image) to posterior (right of image) progression (green, flk-1 eGFP; red, tissue counterstain, rhodamine-lens culinaris agglutinin). B: Trunk ISVs in control and arsenic-treated zebrafish. Rostral ISV (left) in each panel is 5 hr post-initiation. Endpoint analysis of ISV growth shows that arsenic exposure results in altered ISV morphology, shorter thinner ISVs, missing ISVs, and more numerous filopodia. C: Using the Manual Tracking plugin in FIJI ImageJ, we tracked both ISV tip position and ISV base position for each ISV in 3D over time. Uncorrected tip positions over time (green) produce a trajectory that is the sum of the intrinsic ISV tip position and the local translocation of the embryo. Base positions over time (pink) produce a trajectory that defines the local translocation of the embryo. D: Using the angle measurement tool in FIJI ImageJ, we measured the angle between the proximal segment of the ISV and the DA segment anterior to each ISV. a, anterior; p, posterior. E: Arsenic-treated zebrafish are slightly shorter than age-matched controls but appear morphologically normal. F,G: Trunk regions of control and arsenic-treated zebrafish at 26 hpf show chevron-shaped myotomes by transmission imaging and by unc45 eGFP expression. ISVs, shown by kdr expression, in the arsenic treated ISVs are irregular in shape, length, and orientation as compared to control. DA, dorsal aorta; ISV, intersegmental vessel; m, myotome; n, notochord; PCV, posterior caudal vein. Panel A arrows identify ISVs. Panel A double arrow indicates location of notochord. Yellow line segments in panel D define the angle between the DA and ISV.

Assays of aberrant ISV growth typically examine a single time point and quantify the presence or absence of ISVs, ISV length and width, cells per ISV, total fluorescence, whether ISVs are partial or complete, or whether patterning is normal or abnormal (Covassin et al., 2006, 2009; Epting et al., 2010; Herbert et al., 2012; Lamont et al., 2009; Lu et al., 2004; Pan et al., 2012; San Antonio et al., 2009; Siekmann and Lawson, 2007; Wiens et al., 2010; Yu et al., 2010; Zeng et al., 2009). These single time-point observations, while informative, overlook the complex dynamics of ISV growth. For example, without examining ISV growth over time, it is not possible to know if an ISV is absent because it never formed or because it formed and later retracted. Since differences in ISV morphology emerge from underlying differences in dynamic cell behaviors, examining these behaviors may provide additional clues for discovery of the underlying causative molecular mechanisms. To quantify the growth dynamics of ISVs, we have developed a novel method to collect and analyze trajectories (sequences of successive 3D locations over time) of ISV growth. These dynamic analyses compliment and extend existing methods.

To ensure that our ISV growth dynamics analyses can be easily employed by other labs, we have used free open source software, FIJI ImageJ, to extract positional data from time-lapse images and we provide the MATLAB scripts that we used for our analyses (see Supplemental MatLab script, two MatLab data files and the document that describes the script and data files).


Perturbation of ISV Growth by the Environmental Toxin Arsenic

In a screen of environmental toxins for effects on development (McCollum et al., 2011) we found that ISV growth was affected by exposure to arsenic (in the form of sodium arsenite) at 400 μg/ml (Fig. 1B). This is of concern because similarly high levels of arsenic can occur in drinking water in some parts of the world (Nordstrom 2002). However, counts of present or absent ISVs and descriptions and measures of ISV morphology failed to fully capture the effects of arsenic on the dynamics of ISV growth that we observed in time-lapse sequences (see Supp. Movie S1). We developed the dynamic metrics we describe in this report to quantify these changes in growth dynamics.

Definition of Region and Time Period

Our analyses focus on early ISV sprouting and sprout extension, which occurs in untreated embryos between 20 and 25 hr post-fertilization (hpf). We have chosen this process because the mechanisms used in ISV sprouting and sprout extension persist and are built on in subsequent steps of ISV development. To allow us to average ISV growth data within a single embryo and across multiple embryos, we confined our analyses to the ISVs adjacent to the yolk plug extension (Fig. 1A). In this region, variation in morphology of surrounding tissues along the rostro-caudal axis is minimal. Tissues surrounding the forming ISVs adjacent to the yolk plug extension are the notochord, neural tube, somites/myotomes, with the gut and paired pronephros below the dorsal aorta and posterior caudal vein. Rostral to the yolk plug there are additional tissues nearby including the sub-intestinal veins, swim bladder, stomach, and other rapidly developing organs such as liver. Caudal to the yolk plug extension, the pronephros and gut are not present.

Genetic manipulations or exposure to drugs or toxins can either accelerate or delay initiation of ISV sprouting and perturb its regular spatial and temporal intervals (Fig. 1B) as well as the pace of development overall. We temporally synchronized ISV sprout trajectory data by defining time of initiation of each sprout as the time origin for that sprout. Measuring time relative to sprout initiation allowed us to compare dynamics of sprout growth in control and perturbed embryos, independent of developmental delays.

Data Mining of 3D Time-Lapse Images

We began our analyses of ISV growth dynamics by logging the 3D coordinates of ISV sprout bases and tips over time. For each ISV, we recorded a track (x,y,z coordinates over time) for the base and for the corresponding tip (Fig. 1C) using the Manual Tracking plugin in FIJI ImageJ v1.47b (Rasband, 1997–2012; Schindelin et al., 2012). Our confocal microscope was equipped with a low NA air immersion objective, limiting axial resolution, and impeding selection of the correct z-plane. To improve consistency, we first determined candidate z-positions, then checked them in xz view before tracking was performed. This check, while time-consuming, ensured reproducible measurements. Our resulting error in extracting quantitative measurements from the imaging data was small. We estimated our typical error in point selection by examining the deviations of our base and tip measurements at time zero (see Fig. 3). Since the base and tip are equivalent by definition at time zero (Fig. 1C, tip and base at 0 hpi), our measured base and tip positions should be the same and differences in value indicate our measurement error. This analysis showed that our measurement error in point selection was less than one pixel in x, y (1.2 μm) and z (1.4 to 1.8 μm). The temporal sampling rate was one confocal image stack every 15 min.

We defined time zero as the time point immediately before a sprout initiates. In most cases, at time zero we observed high flk-1 expression centered at the location where a sprout subsequently emerged. We defined ISV base position as the midpoint of the sprout where it intersects with the dorsal aorta (DA). We defined ISV tip position as the point on the ISV cell body most distal from the DA. Fine filopodia were below the resolution of the captured images. In cases where the distal tip bifurcated, we tracked only the persistent extension. We did not encounter any persistent bifurcations in these experiments. Only ISVs on the side of the sample nearest the objective were tracked because spherical aberration degrades signal and resolution at depth (Muriello and Dunn, 2008). We analyzed these data using custom-written Matlab code (MATLAB 6.1, The MathWorks Inc., Natick, MA, 2000) (see Supplemental MatLab script, two MatLab data files and the document that describes the script and data files).

Motion Correction

Simple tracking of the sprout tip gave a trajectory that was the sum of the change in position of the sprout tip and translocation of the entire embryo (Fig. 1C). Further complicating these analyses, zebrafish embryos twitch, so that growth and motion of embryos were not uniform across the 3D field, making a global correction such as simple translation and rotation of the entire image inadequate. We tested, but did not utilize, nonrigid transformation methods because this approach for optimization of global registration had the potential to obscure the dynamics that we wished to measure. We, therefore, used a simple local position correction, with ISV base positions (Fig. 1C, magenta) serving as local correction factors for ISV tip positions at each time (Fig. 1C, green). To determine corrected tip positions, we subtracted 3D base position from 3D tip position at each time point for each ISV. We then used the corrected 3D ISV tip positions over time to construct tip position trajectories for each ISV.

Spatial Orientation

ISVs sprout from the DA with a regular spatial orientation. Perturbing the shape of the somites or the signaling and adhesion molecules that guide the ISV can perturb the orientation of the ISV sprouts relative to the DA as seen in the fused somite mutant (Shaw et al., 2006). Genetic manipulation or treatment with a toxin such as arsenic has the potential to alter somite shape or the quantities and distribution of signaling and adhesion molecules and in this way affect the orientation of the ISV with respect to the DA. However, while arsenic-treated fish showed a slight developmental delay, i.e., a decreased body size at a given time postfertilization compared to untreated siblings (Fig. 1E), we did not observe obvious morphological abnormalities, such as changed somite shape (Fig. 1F,G), even when ISV growth was substantially perturbed (Fig. 1F,G). Using the angle tool in FIJI ImageJ, at 5 hpi, we measured the angle between each ISV and the adjacent anterior segment of dorsal aorta (Fig. 1D). In untreated zebrafish, ISVs emerged at a 68.5 ± 7.06° s.d. angle and in zebrafish treated with arsenic, ISVs emerged at an 84 ± 11.08° s.d. angle with respect to the DA (Fig. 2A). This difference between the means was significant (P < 0.0001, Student's t-test).

Figure 2.

Spatial orientation and temporal intervals of intersegmental vessels are affected by arsenic exposure. A: Orientation of ISVs with respect to the DA was measured in control (green) and arsenic-treated (red) sprouts and variance and mean (using Student's t-test) were found to be significantly affected by arsenic treatment. B: Time delay between initialization of subsequent sprouts was recorded for control (green) and arsenic-treated (red) sprouts. For sprouts that initiated and regressed, then initiated and progressed, the second initiation time was used. Delay interval between subsequent sprouts tended to be shorter in arsenic-treated zebrafish than in control, and variance differed significantly between groups, but the difference in means (Kolmogorov-Smirnov test) was not significant.

Figure 3.

Quadratic Fitted Functions describe ISV Tip Position Trajectories. Shown are raw trajectories (gray), average trajectories (black), and quadratic-fitted functions (control, green; arsenic-treated, red) in x,y, and z views. Bars at t = 0 (control, red: arsenic, green) on each plot show error of point selection, which was about one pixel in x,y (1.2 μ) and in z (1.4 to 1.8 μ). Black bars for subsequent time points, calculated as standard deviation, show distribution of experimental data around the mean. For controls, n = 27 ISVs from 8 zebrafish. For arsenic-treated, n = 16 ISVs from 4 zebrafish. Time interval = 15 min. Total time points = 20.

Temporal Interval

ISV sprouts emerge from the dorsal aorta (DA) at regular temporal intervals in an anterior to posterior sequence. We calculate the interval between ISV sprouts as the time between initiation of successive sprouts. When a more anterior ISV sprouts after a more posterior sprout, the sign of the interval is negative. Only ISV sprouts that persist were used for this calculation, not those that initiated and then regressed. Control ISVs rarely deviated from initiating in an anterior to posterior sequence, but in the arsenic-treated group sprouting of more posterior ISVs at the same time or before more anterior ISVs occurred frequently, resulting in a significant difference in variance (P < 0.0001) between the control and treated groups (Fig. 2B). The mean interval between emergence of control ISVs was one every 25.4 min ± 44.6 s.d. while the mean interval between emergence of arsenic-treated ISVs was one every 15 min ± 156 s.d. (Fig. 2B), but this difference was not significant (P = 0.79, two sample Kolmogorov-Smirnov test).

Canonical ISV Trajectories

We define a canonical ISV trajectory as the best fitted curve to the average of the measured experimental trajectories, i.e., an idealized ISV growth trajectory. We can then describe and compare canonical trajectories derived from different populations using conventions established for describing the geometry of 3D curves (Crenshaw et al. 2000). From canonical ISV trajectories, we can also derive average directed migration speed and directional persistence of ISV growth.

We registered experimental trajectories within each group at their base position at the initial time point. To correct for curvature of the embryo's tail, we aligned the registered trajectories in x,y using the measured angle of each ISV from the dorsal aorta (Fig. 1D). Rotation about the alignment point in z was minimal and did not require correction.

We calculated the average trajectories for control and for arsenic-treated fish, then performed a quadratic fit to the trajectories using a weighted least-squares algorithm, weighting each data point by the inverse of its standard deviation (Fig. 3; r2 > 0.98 for quadratic-fitted curves). As mentioned above, standard deviation at the initial time point gives an estimate of error of measurement in coordinate selection (about one pixel, Fig. 3). The standard deviation at each subsequent time point reflects the amount of path-seeking behavior among ISV sprout tips (Fig. 4D).

Figure 4.

Canonical ISV trajectories can be defined and compared by geometric properties and used to determine average directed migration speed and directionality. A: Shown in 3D are the canonical ISV trajectories for control (green) and arsenic-treated (red) zebrafish. We defined the canonical ISV trajectories as the best fitted curve to the average of the measured experimental trajectories. For both control and arsenic-treated, the best fit was a quadratic fitted function (r2 > 0.98 for all quadratic-fitted curves). Each curve segment represents 1 hr, (four 15-min time points). B: Curvature, a geometric property of curves, which is an alignment-independent method of comparison, was different for control and arsenic-treated canonical ISV trajectories. C: From the canonical ISV trajectories, we calculated average directed migration speed over time for control (green) and for arsenic-treated (red) zebrafish. Shaded area represents s.d. Control ISVs steadily decreased their migration speed until 3 hr post-initiation, then increased their speed of directed migration until the end of the experiment at 4.75 hpi. Arsenic-treated sprouts steadily decrease their migration speed over the time observed. D: Directionality was calculated as the angle between a tangential vector of the canonical ISV trajectory and individual trajectories in control (green) and arsenic-treated (red) zebrafish.

We can fully describe our canonical trajectories by the properties of curvature, torsion, and velocity (Fig. 4A) (Crenshaw et al., 2000). Curvature is the degree of the deviation from a straight line, measured as the rate of change (at a point) of the angle between a curve and a tangent to the curve. Torsion is how sharply the curve twists, measured as the magnitude of the derivative of the normal to the plane of curvature with respect to the distance on the curve. Curvature and torsion are alignment-independent methods of trajectory comparison. If curvature and torsion of two 3D curves are not the same, then the 3D trajectories that they describe cannot be aligned and are not equivalent.

We used the following equations (Crenshaw et al., 2000) to calculate curvature (κ) and torsion (τ) for the fitted trajectories:

display math
display math

where χ′ is the first derivative of fitted x-path respect to time (second and third derivative respect to time denoted by χ′′ and χ′′′).

Curvature of the control canonical trajectory smoothly increases, reaches a maximum just after 3 hpi, and then smoothly decreases through the end of the experiment at 4.75 hpi (Fig. 4B). Curvature of the arsenic-treated canonical trajectory remains low until just after 4 hpi, then increases sharply and continues increasing through the end of the experiment at 4.75 hpi (Fig. 4B). Because our canonical trajectories were fitted to the averaged data using quadratic functions, torsion is by definition zero. The curvatures of the control and arsenic-treated trajectories, however, are different, which shows that the canonical trajectories that they were derived from cannot be aligned and that the canonical trajectories for control and arsenic-treated ISVs are different.

We calculated the average directed migration speed of ISV sprout tips by taking the first derivative of the fitted paths in x,y,z with respect to time to determine the magnitude of the velocity vectors (Fig. 4A). Average directed migration speed of control ISVs began at 17.5 μm/hr, dropped to a minimum of 8 μm/hr just after 3 hpi, then increased to 11 μm/hr at 4.75 hpi, the end of the experiment. Average directed migration speed of arsenic-treated ISVs began at 14 μm/hr and decreased steadily to 3 μm/hr at 4.75 hpi. The average directed migration speed of arsenic-treated ISVs was less than that of controls throughout the experiment.

We quantified the directionality of tip protrusion and retraction (path seeking) at each time point for each individual ISV path by calculating the angle (in radians) between the tangent vector of the canonical path and velocity vector calculated from aligned spatiotemporally synchronized ISV paths according to:

display math

where (·) is the dot product operator in the numerator and math formula is the length of math formula and math formula is the length of math formula in denominator. θ is 0 when both vectors are in the same direction, π when they are in opposite directions, and π/2 when they are perpendicular (Fig. 4D). For both control and arsenic-treated ISVs, tip extension in the direction of growth (angles less than 90°) was much more frequent than tip retraction (angles greater than 90°). Control and arsenic-treated ISVs had similar directional profiles except that arsenic-treated ISV tips had a higher frequency of perpendicular movement.

Morphological differences between control and arsenic-treated ISVs seen in static images at 5 hpi (Fig. 1B), such as shorter ISV length and irregular orientation relative to the dorsal aorta, can be accounted for by the differences in growth dynamics. Arsenic-treated ISVs begin migration more slowly than controls and their directed migration speed decreases steadily over the time course of the experiment, resulting in shorter ISVs. The higher frequency of perpendicular movement seen in arsenic-treated ISVs contributes both to decreased directed migration speed and to irregular orientation relative to the dorsal aorta.


Three dimensional time-lapse imaging is time-consuming and labor intensive, but the resulting data provide rich dynamic information that may yield new insights into mechanisms of growth and development. Analysis of developmental dynamics does not replace but augments standard morphometric analyses. Our new method is useful not only for analysis of angiogenic sprouting in zebrafish, but also for analysis of angiogenic sprouting in other systems and for other similar events such as analysis of neurite growth dynamics in axonal pathfinding.

Automated extraction of meaningful information from digital images is a complex problem because time-lapse data acquired from living organisms frequently has a low signal and high noise. The human visual cortex is excellent for extracting information from such images but cannot yet be entirely replaced by computer vision. We tested software tools in FIJI ImageJ and in IMARIS that refined manually selected points to a nearby point of local maximum signal intensity to semi-automate selection of the distal tip of the ISVs. However, point selection error was substantially larger for semi-automated tip selection than for fully manual selection. Both software packages chose points that were within the ISV rather than at the tip, or even within the adjacent ISV. We, therefore, used manual selection of base and tip coordinates in 3D over time to acquire data for this report.

Manual tracking was the rate-limiting step for these analyses (on average we were able to manually track two datasets with 3–5 sprouts and 20 timepoints per dataset per day). Manual tracking is time-consuming and its accuracy and repeatability can be operator dependent (Huth et al., 2010). Refinement and extension of existing software tools for automated ISV segmentation such as the Cognition Network Technology rule set approach developed by (Vogt et al., 2009), specifically for selection of ISV base positions and ISV tip positions in 3D over time, would be a significant advance that would benefit the community of angiogenesis researchers.

Improved image quality would make both automated and manual tracking of ISV base positions and ISV tip positions in 3D over time faster and more accurate. A long working distance low-magnification water immersion objective, with high NA (>0.9), would improve both lateral and axial resolution, while still providing a wide field of view and large depth of field. When available, we recommend this type of objective. However, as we have shown, confocal imaging performed with the commonly available low-power, low-NA air immersion objectives still provide images of sufficient quality for dynamic analysis.

Sampling interval restricts which periodic behaviors can be observed and quantified. Phenomena that occur less frequently than the sampling rate may be missed. Tip cells may extend and retract multiple times during our 15-min sampling interval and therefore patterns in extension and retraction frequency and directionality may have been missed, and values of the spatial amplitude may have been underestimated. However, more frequent imaging may increase phototoxicity and photobleaching. Selective plane illumination microscopy (SPIM) would be ideal for dynamic studies because it provides optical sectioning of whole zebrafish embryos with reduced phototoxicity, high acquisition speed, and improved axial resolution compared to confocal or multiphoton microscopy (Engelbrecht and Stelzer, 2006; Huisken and Stainier, 2007). SPIM systems can collect a 3D stack in seconds rather than minutes, permitting observation of short interval behaviors like filopodial extension and retraction. SPIM's higher axial resolution would allow faster and more precise manual tracking and may make automated tracking of ISV base and tip positions more tractable.

Our arsenic exposure regime mimics environmental exposure (at a high level). Arsenite exposure leads to oxidative stress and impairment of cellular respiration through multiple mechanisms (Watanabe and Hirano, 2012) and thus certainly has effects on all tissues, not only the angiogenic sprouts followed in this study. We do not attempt to distinguish between primary and secondary effects of the toxin. However, the zebrafish in this study appeared relatively normal morphologically (Fig. 1E), albeit with a slight developmental delay for the first 2 dpf, and some edema by 3 dpf, but defects in vascular development were pronounced (Fig. 1B,F,G). The slower migration speed seen in arsenic-treated ISVs suggests that the morphogen gradient that brings about directional ISV growth (or alternately, expression of the morphogen's receptors) may be affected in arsenic-treated embryos. Altered seeking behavior and failure of the arsenic-treated ISVs to remain within the ECM rich boundary between somites indicates that repulsive cues at somite borders may also be affected.

Dynamic analysis of angiogenic sprouts augments current methods of static analysis, with potential for identifying angiogenic defects that may not be detected by static measures. For example, ISV length may be short due to low migration speed, or migration speed may initially be normal but ISV growth may then abruptly stop due to lack of a critical signaling or adhesion molecule needed for growth to progress. We envision further developing angiogenic sprout growth dynamics analysis as a way to derive dynamic signatures of specific genetic disruptions or toxin exposures (Croitoru et al., 2005). Perturbations that affect cell–cell adhesion, matrix adhesion, external guidance cues, or gradients that modulate cell behaviors will also then affect sprout growth dynamics in distinct ways that may be characteristic for those causative molecular mechanisms. If so, we may be able to use dynamic signatures to decipher which genetic pathways have been disrupted, or to identify the family of toxins to which an embryo has been exposed.

Dynamic cell behaviors are critical inputs for building biologically relevant mechanistic computer models (Andasari et al., 2012; Larson et al., 2010; Shirinifard et al., 2009). Often, our difficulty in developing and applying such models is our lack of quantitative metrics that allow us to compare the behaviors of simulations under different conditions, e.g., to determine the degree of similarity or difference of two simulations or between simulation and experiment. The dynamic analyses we have presented could provide such comparison metrics. A multi-cell multi-scale model of angiogenesis, built using angiogenic sprout growth dynamics data, with subcellular networks embedded, could provide a test bed for perturbations of vascular development, and generate testable hypotheses of the mechanisms of both normal and perturbed sprout growth.


Animal Care

Zebrafish (Danio rerio) were raised and kept under standard laboratory conditions (Westerfield, 2007) and experimental protocols approved by the Institutional Animal Care and Use Committee at University of Houston (protocol numbers 09–040 and 10–040). For some experiments, 0.2 mM phenylthiourea was added to prevent melanization. Ages of the embryos are given as hours post fertilization (hpf). ISV growth is given as hours post initiation (hpi). To mimic environmental exposure to the toxin arsenic (in the form of sodium arsenite), embryos were placed in embryo medium (Westerfield, 2007) with sodium arsenite (400 μg/ml) immediately after collection.

Sample Preparation and Image Acquisition

We mounted embryos for microscopy in embryo medium (Westerfield, 2007) with 0.03% MS-222 and 0.3% low-melting-point agarose (Sigma, St. Louis, MO), a concentration that holds the embryo in position during imaging, while not restricting growth (Kaufmann et al., 2012). Coverslip bottom dishes (MatTek Corp., Ashland, MA cat. no. P35G-0–10-C) were used to facilitate imaging and were sealed with parafilm to prevent evaporation. We chose the field of view to include ISVs adjacent to the yolk-plug extension. We mounted embryos with the rostral end of the embryo to the left and the tail as flat against the coverslip as possible. Standardization of mounting orientation and position facilitated subsequent image analysis.

We optimized our sample preparation and image acquisition parameters for medium throughput screening of perturbation of ISV development using commonly available equipment: a standard confocal microscope equipped with a low-magnification air immersion objective with a modest numerical aperture. This type of objective provides a large field of view and large depth of field. The large field of view allowed us to keep individual ISVs in the field during extended time-lapse imaging, even when the embryo translocated due to growth and development. The large depth of field affords more flexibility for sample positioning in z, which allows speedy mounting for imaging and also aids keeping the embryo in view in z during extended time-lapse imaging. The disadvantage, however, is limited axial resolution.

We acquired 3D time-lapse images of Tg(Flk-1:eGFP) zebrafish (Jin et al., 2005) using an Olympus FV1000 confocal microscope equipped with a 20× 0.7 NA. air immersion objective at 15-min intervals, beginning at 18–19 hpf to insure capture of initiation of sprouting. Our goals in time-lapse image acquisition were to maximize image quality while minimizing phototoxicity and photobleaching. We set gain such that we were able to clearly differentiate ISVs from background. Scan speed was set to maximum for the system. We set laser intensity such that there was minimal detector saturation. The pinhole was set to one airy unit. With these settings, acquisition of a single 3D image took 3.34 min. We captured 3D images at 15-min intervals for up to 15 to 20 hr with no apparent phototoxicity or photobleaching. The 15-min interval allowed time-lapse imaging of multiple fish during a single imaging session. Acquisition of each z-stack started deep in the sample and progressed toward the objective. No morphological differences were seen between imaged fish and identically treated fish that had not been imaged.

We imported the numbered tifs generated by the Olympus FV1000 as a sequence into FIJI ImageJ v1.47b (Rasband, 1997–2012; Schindelin et al., 2012). We standardized orientation of all datasets. Those that were collected in the incorrect z-order were opened as 5D datasets and then the order of the z-planes was reversed using the Stacks, Tools, Reverse command. We then converted all datasets to Hyperstacks and saved. For each ISV in each dataset, we then tracked ISV tip position and ISV base position in 3D over time using the Manual Tracking plugin in FIJI ImageJ.


This project was funded in part by EPA R-10–0049, The Texas-Indiana Virtual STAR Center; Data-Generating in vitro and in silico Models of Development in Embryonic Stem Cells and Zebrafish. We acknowledge Fatima Merchant and Amol Shete at the University of Houston for assistance with imaging and use of the Olympus FV1000.