ROLE OF FLUID FLOW IN DEVELOPMENT

In addition to facilitating convective transport, intra-vital fluid flows impose substantial mechanical stresses on adjacent and underlying cells. These flow-induced forces are widely acknowledged as critical to the proper development and maintenance of many aspects of biologic form and function. This is particularly true during embryogenesis where internally-derived, flow-related forces are thought to be morphogens influencing a number of key developmental processes including; symmetry determination (1,2), cardiogenesis (3–5), blood vessel formation (6,7), glomerulogenesis (8), brain development (9), and lung development (10,11). In addition to their roles during normal development, the biomechanical forces generated by aberrant intra-vital flow have been implicated as important factors in the pathogenesis of a variety of diseases in the cardiovascular (12–14), nervous (15–17), and renal (18–20) systems.

The specific mechanisms by which living cells sense, transduce and respond to flow-induced stresses are only partially known (21–25). In vitro studies have contributed a great deal to our understanding of these signaling pathways in general, and how cardiovascular endothelial cells, the flow sensors and transducers lining the vascular walls, react to shear stress (26–29), stretch (30,31), and pressure (32,33). Of these, wall shear stress has received the most attention as both its magnitude and orientation are thought to play roles during vascular development. Fluid shear stresses occur within the cardiovascular system as blood flows tangentially to the surrounding vessel wall. This frictional force is defined by the product of the shear rate (derivative of velocity with respect to the vessel radius) and the dynamic viscosity of blood. To reconcile the complex velocity gradients that exist within a pulsing, flexible heart tube filled with a moving, non-Newtonian fluid we need to obtain some understanding of the cross-sectional flow profile of the vessel. With these data we are better able to determine whether the flow has had time to establish itself into the parabolic velocity profile expected for laminar flow or whether perturbations in the flow have resulted in altered cross-sectional gradients. Characterizing the magnitude and orientation of the shear forces acting at the level of the biologic flow sensors, rather than those calculated from mid-lumenal flow, is critically important if we are to ascertain the extent to which these biophysical forces influence cellular response.

Although in vitro studies have been important in helping to elucidate the responses of cultured endothelial cells to a variety of biophysical stresses, their utility is reduced by our inability to accurately reproduce the complexity of biologic flows and the physicochemical interconnectivity of living tissues. These limitations have led researchers in developmental biology to appreciate that in vivo mapping of biofluid flows is the critical next step if significant further progress is to be made in our understanding of intra-vital flow-structure interactions during developmental and disease processes. Despite its potential utility in dynamically characterizing many biologic systems, flow mapping efforts to date have largely focused on the developing cardiovascular system, because it appears very early in embryogenesis, is optically accessible, and exhibits a wide range of flow dynamics.

HEMODYNAMICS AND CARDIAC DEVELOPMENT

The influence of blood flow on the form of the heart, and conversely, the effect of genetically-determined form on flow, have been topics of debate for nearly a century (34–36). Goertller (37) was probably correct when he pointed out that while some aspects of cardiovascular formation are certainly pre-programmed, other areas are just as surely dependent upon the characteristics of local fluid flow. So how does the energy imparted to the blood influence cardiac morphology? Several hypotheses have been posited, including: 1) differential cellular growth due to residual stress and strain (38); 2) changes in blood flow volume and pressure zones restructuring media replacement and affecting luminal diameters (39–40); 3) fluid shear rates at intra-cardiac surfaces (41,42); and 4) blood flow leading to endothelium-mediated signals (43,44).

There is a wealth of evidence suggesting that flow-induced stresses, such as the frictional shear imposed by blood flow on adjacent cardiovascular endothelial cells, can substantially influence vascular development and adaptation as well as pathogenesis of vascular disease adult organisms (12,45–50). In fact, much of our current understanding about cellular responses to flow-induced forces comes from studies of the mature arterial endothelium. In such systems there is substantial evidence for multiple flow sensors (26,51–54), capable of eliciting responses via MAP kinases (55,56), PI3-kinase (57–59), protein kinase C (60), NF-κB (61) and other signal transduction pathways.

Evidence that different flow regimes (e.g., steady versus pulsatile, laminar versus turbulent) can differentially regulate gene expression (62–64) attests to the subtlety with which mature cardiovascular tissues can respond to changes in blood flow-induced forces. Interestingly, many of these flow-regulated genes encode products that directly regulate tissue growth and remodeling (65,66) or are important transcription factors that have been linked to growth control (67,68). This supports the notion that some of the same sensory transduction pathways used by mature animals may be found in developing organisms as well. Unfortunately, our understanding of the hemodynamic milieu within the developing vertebrate embryo is far from complete. The data we do have emphatically tells us that experimental disruptions of the venous return or arterial outflow during early development can lead to severe dysmorphogenesis (69–72). However, our ability to quantitatively describe these complex biologic flows within living, growing organisms is frequently limited by their small size and the relative inaccessibility of internal fluid flow environments to flow-sensing instrumentation. In the absence of reliable intra-vital flow data, discourse concerning the underlying factors responsible for both normal and pathophysiological cardiovascular development will be incomplete and the conclusions drawn from it will remain speculative. Fortunately, this problem has not gone unnoticed and, as a result, significant research effort is being expended in the development of instrumentation capable of quantifying micro-scale intra-vital flows.

MEASURING INTERNAL FLUID FLOW

The biomedical research literature is replete with imaging techniques aimed at characterizing intra-vital flows, with generally mixed results. One approach commonly used in large mammals is to physically position a flow sensor directly into, or adjacent to the flow field of interest (73–75). While this is the most direct method for obtaining quantitative flow data, the typically invasive nature of positioning a flow probe in such a manner can significantly perturb an adult system and the effect on a developing embryo may be prohibitive. Despite this difficulty, the use of surgically implanted pulsed Doppler probes in mouse and chick embryos have led to an improved understanding of developmental hemodynamics in vertebrates (76–84).

Embryonic chicks are highly amenable to this technique as they can be rotated in windowed eggs to bring the vessel of interest to a superficial position for optimal orientation of the small (~0.5 mm) piezoelectric crystal probe. Although such probes have been used quite effectively, Doppler ultrasound (US) techniques can be confounded by moving vessel boundaries or by flow in closely adjacent vessels. As a result, great care must be taken in choosing the anatomical area to be examined. Although its temporal resolution can be quite good, spatial resolution of Doppler US is limited by the size and angular orientation of the piezoelectric crystals. In addition, signal frequency must be carefully monitored as high-intensity pulsed US has been shown to affect cardiac rhythm and aortic blood pressure in frogs (85).

In response to these issues, other, less invasive medical imaging technologies have been refined for small scale biofluids imaging in humans with potential application to small animal models as well. Phase contrast magnetic resonance (PC-MR) imaging of blood flow in humans is based on a phase shift of the magnetization when the blood flows in the direction of a magnetic field gradient (86). Partial volume errors and errors related the area to be interrogated still persist when examined on smaller scales although new algorithms are being developed to improve spatial resolution (87). Non-gated MR imaging has been used in conjunction with Doppler US for detection of slow blood flows in humans (88,89), as has Orthogonal Polarization Spectral (OPS) imaging (90). Deuterium MR imaging of slow intra-ocular fluid flow has been accomplished in rabbits with 100 μm spatial resolutions (91). Unfortunately, none of these approaches provide the temporal sensitivity necessary to characterize typical embryonic cardiovascular flow. MR imaging has shown great promise as an anatomical tool for assessing developmental cardiovascular defects in the mouse model (92–94). Unfortunately, the proper preparation of the embryos requires chemical fixation, eliminating the possibility of dynamic measurements of cardiovascular performance. Positron emission tomography (PET) has been adapted to detect cerebral blood flow dynamics in adult rabbits (95) but intrinsically limited spatial resolution (96) suggests limited utility on the developmental size scale.

Perhaps the most promising new imaging technology for quantifying developmental biofluid flows is US biomicroscopy (UBM)-Doppler, a non-invasive, high-frequency echocardiography system. UBM is currently capable of achieving spatial resolutions on the order of 30 μm axial and 90 μm lateral with sufficient temporal resolution to make reliable velocity measurements within developing vertebrate embryos. UBM has been successfully used to examine cardiovascular morphology, dimensions and blood flow in the mouse and zebrafish embryos (97–99). A more detailed review of developmental UBM imaging can be found in this issue of Pediatric Research.

While most of these imaging modalities continue to operate well within their prescribed research niche, few possess the spatial and temporal resolution required to quantitatively characterize the cross-sectional flow profile of a developing blood vessel. As such, the development of new, more powerful imaging technologies is necessary if we are to derive the flow induced forces acting at the level of the biologic flow sensors within living organisms. Several research groups are now looking to transfer emerging fluid mechanical engineering technologies to these challenging biomedical applications.

QUANTITATIVE FLOW VISUALIZATION IN ZEBRAFISH

Ludwig Prandtl's work in the early part of the twentieth century was the beginning of the movement from passive observation of fluid flows to experimentally extracting information from them. A pioneering fluid mechanist, Prandtl suspended mica particles on the surface of moving water to qualitatively study aspects of unsteady flow (100). Recent advances in optics, lasers and computer technologies allow modern day fluid mechanists to collect detailed quantitative data about instantaneous flow velocities from the same kinds of seeded flows observed in Prandtl's time. To transfer these technologies from fluid mechanical engineering applications to the task of quantitatively mapping developmental biologic flows a tractable animal model system is needed.

The ideal animal model system should demonstrate circulatory functions characteristic of other vertebrate models and be optically accessible with modern microscopic tools. To find such a model researchers are turning to “lower vertebrate” systems (e.g., fish and frogs). The zebrafish (Danio rerio), has become an important model system for studying organogenesis, particularly the form and function of the developing cardiovascular system (101–104). Although the single-circuit, two-chambered piscine heart cannot fully model all aspects of mammalian cardiogenesis, many fundamental developmental processes (e.g., heart tube formation, looping, valvulogenesis and septation) follow a pattern of extreme evolutionary conservation across vertebrate taxa. This affords us the opportunity to study the pathways underlying important cardiogenetic processes in zebrafish to better understand how those same processes unfold in mammals and specifically, in humans. In fact, zebrafish possess a number of life history characteristics that make them more amenable than their mammalian and avian counterparts to studies of dynamic developmental imaging. Included among these are external fertilization, small size, rapid development, optical transparence, and genetic accessibility (105).

The tissue clarity afforded by a transparent embryo greatly facilitates non-invasive optical dissection using modern confocal microscopic techniques (Fig. 1). Structural contrast can be provided by vital staining or the use of transgenic lines expressing fluorescent protein markers like green fluorescent protein (GFP) (106,107). Embryos expressing the appropriate labels can then be anesthetized and easily mounted for microscopic observation (108). The zebrafish cardiovascular system begins to form early in development with pumping of blood commencing at about 1 day post-fertilization (dpf) (109). Astounding anatomical detail of the embryonic zebrafish heart and its microvasculature has now been described throughout ontogeny from microscopic observation of both fixed tissues (4,110) and in living tissues after the introduction of fluorescently labeled dyes/microspheres. The latter have been used to produce exquisite micro-angiograms of the developing zebrafish vasculature, allowing us an unprecedented glimpse at the way primary and secondary blood vessels form (111,112).

Figure 1
figure 1

Optical dissection of a zebrafish embryo. Pictures display individual confocal sections taken in 20 μm increments (dorsal to ventral) through the heart of a Bodipy-ceramide stained 4.5-dpf zebrafish embryo. At 20°C this heart was beating at approximately 2 Hz. (A) Top of the ventricle is visible, (B) the atrium is becoming visible beneath the ventricle, (C) the top of the bulbus arteriosus (outflow tract) can be seen, (D) Moving erythrocytes leave dark streaks in the fluoresced serum making vortices and jets visible, (E) High-velocity flow in the outflow tract feeds the branchial aortae.

Full size image

Functional phenotyping of cardiovascular performance began with recordings of micro-scale blood pressures within the living zebrafish (4,110,113,114). Using native particles (e.g., erythrocytes, platelets) as flow markers, the first descriptions of the gross blood flow patterns in developing zebrafish were made possible. Particle displacements were used to approximate blood flow in zebrafish larvae by measuring the differential lengths of the vector shifts between two adjacent video frame fields acquired during microscopic observation (115). In addition to determining erythrocyte velocities, this technique was an effective way to visualize their relative distribution within anatomical regions of interest. Overlays of multiple particle pathlines were used to produce reasonably good approximations of vessel inner dimensions as well. Erythrocyte tracking was also used in a recent optical analysis of blood flow in the superficial vasculature of the embryonic mouse yolk sac (116). Blood flow velocities were measured by confocally line scanning images perpendicular to the flow and measuring the vertical distortion of the circular red blood cells with respect to time. In large diameter vessels (200 μm) cross-sectional flow profiles were reconstructed from erythrocyte velocity measurements and shear stresses calculated.

The increasingly widespread use of discrete particles suspended in the blood as flow tracers has hastened the introduction of sophisticated flow visualization analysis techniques to the study of intra-vital biofluids flows. Most of these methods use time-sequenced images of particle fields for either individual particle tracking or statistical descriptions of particle groups for obtaining displacement information and subsequently, velocity fields (117). From these velocity maps a number of useful force fields can be derived. One of these analytical techniques, digital particle image velocimetry (DPIV), is becoming the new technical standard for creating quantitative maps of in vivo flow environments.

MAPPING DEVELOPMENTAL FLOWS WITH DPIV

DPIV is a powerful global quantitative flow visualization method based on following groups of particles as they move through a defined space. In evaluation of external fluid flows (e.g., around the wing of a model plane), small reflective particles are used to seed the fluid (air or water) volume within which the object is suspended and are briefly illuminated as they pass through a pulsed laser sheet. Two consecutive images of the reflected particles are then acquired with a high-speed CCD camera over a short time interval and the resulting image pairs are digitally captured. Data processing is accomplished by recognizing that two sequential particle field images can together produce a shifted composite image depicting motion. A field of displacement vectors can be obtained by analyzing the movements of localized groups of particles. Mean velocity vectors are statistically determined by cross-correlation of sub-sampled regions (“interrogation windows”) extracted from the two image fields (118) (Fig. 2A–B). For measuring internal flows the same general principles apply. Due to the small size of zebrafish embryos, quantitative flow visualization by DPIV requires the use of microscopic observation of the flow fields. To generate quantitative flow maps, an appropriate optical system is necessary. This system must have: 1) the proper spectral specifications to visualize particles seeded within the flow field (i.e., dependant on the particle label); 2) an image recording system of sufficient quantum efficiency and speed to capture the flows of interest under a variety of experimental conditions (e.g., high-speed or fluorescently labeled flows), and 3) the ability to collect 3-D information over time with the necessary acquisition tools and software for subsequent flow pattern analysis.

Figure 2
figure 2

DPIV analysis of fluid flow. (A) A typical 2-D plane of a fluid jet seeded with reflective particles is illuminated with by a pulsed laser. Note the 2-cm diameter central jet and the way it is rolling up into a torus. (B) Velocity vector field representing the cross-section from A. The warmer (e.g., red) vectors indicate higher velocity flow and cooler vectors (e.g., blue) low velocity flow. (C–D) High-velocity blood flow generated by a 4.5-dpf embryonic zebrafish heart. Pictures are characteristic confocal sections from a single time series of Bodipy-ceramide stained embryos. (C) Atrial systole and ventricular filling. (D) Ventricular systole leading to refilling of the atrium. (E–F) Overlay of DPIV velocity fields from real-time, high-speed imaging. (E) Complex flow in the filling ventricle with higher velocity flow at the atrio-ventricular constriction. (F) High-velocity trans-aortic jet through the ventriculo-bulbal valve during systole. a, atrium; b, bulbus arteriosus; v, ventricle; vbv, ventriculo-bulbal valve.

Full size image

Quantitative characterization of any flow field by DPIV is predicated on the proper seeding of the fluid with markers of the appropriate type and density such that the motions of those particles accurately reflect the spatial and temporal patterns of the flow. This is particularly true of biologic flow fields which are rife with confounding factors that may negatively influence the fidelity with which the particles represent the fluid flow. It is critical for the tracer particles to mimic the local flow conditions as closely as possible for accurate flow visualization by DPIV.

A number of factors should be considered when choosing particles for in vivo DPIV use including particle size, density, surface properties, contrast, and biologic reactivity/toxicity issues. The diameter of the tracer particles is of primary importance as an individual particle must be large enough to be discretely identified from other particles, yet small enough to follow local flows with high precision. Recent in vivo particle tracking studies have used native particles to map flow fields within the superficial rat (119,120) and rabbit (121) mesentery vessels, and in the developing zebrafish heart (71) (Fig. 2). In this latter study, high-speed erythrocyte tracking shed new insight into the dynamic nature of developmental intra-cardiac flow on the micro-scale. Forces far in excess of those predicted for fluid flow at Reynold's numbers (inertial forces/viscous forces) much less than one were measured bringing into question the importance of their magnitude in shaping both the normal form and function of the developing heart. In the absence of such stimulation, development may proceed abnormally (Fig. 3). In addition, observation of the mechanics of early cardiac valve function and the mechanisms responsible for minimizing backflow before they are fully functional were elucidated (122). While wholly natural to the organism, red blood cells are likely too large (~7–10 μm) to provide the seeding densities needed for optimal statistical cross-correlation and spatial resolution in smaller vessels.

Figure 3
figure 3

Reduced blood flow-induced forces induce dysmorphogenesis. Glass beads (50 μm) were surgically positioned at 37 hpf (A) adjacent to the heart inflow without blocking flow, (B) within the inflow effectively stopping venous return to the heart or, (C) within the ventricular outflow, preventing blood from exiting the heart. Blockages were checked after 20 h (D–F) and heart development was accessed at approximately 100 hpf (G–I). Note the effectiveness of the block demonstrated by a lack of erythrocytes in the ventricle and their massive accumulation upstream of the bead in the atrium (E, arrowhead). Development was normal in surgical control experiment (G) but severely disrupted in flow-compromised treatments regardless of blockage location (H,I). Outflow tract development was greatly reduced or absent entirely and neither looping nor valvulogenesis occurred in flow-block experiments. a, atrium; b, bulbus; v, ventricle.

Full size image

Future research in intra-vital DPIV is likely to use commercially manufactured microspheres as tracer particles. Microspheres are currently available in a size range of about 0.02–200 μm diameter and may possess a number of other useful properties. Typically manufactured from polystyrene (ρ~1050 kg m−3), microspheres are designed to be neutrally buoyant in typical biofluids (ρ = 1000–1066 kg m−3). This is important as the particle density should closely match the density of the surrounding fluid to prevent floating or settling, both of which can lead to measurement error. The tendency of the tracer particles to adhere to one another or to adjacent surfaces (e.g., other cells or vessel walls) is also an undesirable yet common difficulty arising during in vivo measurements of flow. To avoid particle aggregation, tracers can be treated with surfactants or have their surface properties altered. One example is the addition of a high density of carboxylic acids to the sphere surfaces. The resulting negative surface charges increases hydrophilicity, helping to keep the tracers suspended within the fluid flow. In addition to the properties of the tracers themselves the seeding density, or the number of particles per unit area being visualized, is one of the most important factors for obtaining good statistical correlations during the DPIV data analysis (123). The optimal scenario is one in which the tracers are sufficiently small to follow the fluid flow with great fidelity, yet are large enough that they can be resolved individually with available optics (Fig. 4). A recent in vitro demonstration of the potential power DPIV in measuring fluid dynamics resulted in extremely high spatial (9 μm × 2 μm) and temporal (6000 Hz) resolution of a simulated biologic flow field (124). This level of performance isn't currently possible for in vivo applications as under the low light conditions of typical vital fluorescence microscopy even the best back-thinned CCDs, CMOS sensors and electron multiplying technologies have greatly reduced temporal resolution (up to ~100 frames s−1). Fortunately, most biologic applications using DPIV do not require such extreme performance from their optical system.

Figure 4
figure 4

High-speed confocal imaging and DPIV analysis of 1 μm pseudo-colored fluorescent microspheres. (A) Seeded flow field at time = t0. (B) Displaced particles time = t0+Δt. (C) Overlay of displacement fields allows visualization of a linear shift from the lower right to upper left of the image. (D) Velocity vector field generated by DPIV analysis confirms local flow directions and magnitude. The warmer (e.g., red) vectors indicate higher velocity flow and cooler vectors (e.g., blue) low-velocity flow.

Full size image

DPIV analysis of intra-vital fluid flows is a recent practice, but a number of interesting biologic phenomena have already been observed in these early experiments. For example, in vivo velocity distributions across the superficial microvessels of mouse cremaster muscle have been characterized by DPIV (125). A fascinating result of this work is the predictability of the hydrodynamically relevant thickness of the glycocalyx layer lining the blood vessels. These data may be useful in future studies which evaluate the integrity of the vasoprotective glycocalyx layer during atherogenesis. We may also gain a better understanding of the resistive forces faced by rolling leukocytes during inflammatory responses (126). Intra-vital use of DPIV outside of the cardiovascular system is now beginning as well with the study of fluid motions within the bells of jellyfish and the buccal cavities of fish. These kinds of innovative applications of in vivo DPIV will provide additional insight into the dynamic flow-structure interactions involved in the development of locomotory and feeding mechanisms (127,128).