|Home | About | Journals | Submit | Contact Us | Français|
Vascular system development involves a complex, three-dimensional branching process that is critical for normal embryogenesis. In the brain, the arterial systems appear to develop in a stereotyped fashion, but no detailed quantitative analyses of the mouse embryonic cerebral arteries have been described. In this study, a gadolinium-based contrast perfusion method was developed to selectively enhance the cerebral arteries in fixed mouse embryos. Three-dimensional magnetic resonance micro-imaging (micro-MRI) data were acquired simultaneously from multiple embryos staged between 10 and 17 days of gestation, and a variety of image analysis methods was used to extract and analyze the cerebral arterial patterns. The results show that the primary arterial branches in the mouse brain are very similar between individuals, with the patterns established early and growth occurring by extension of the segments, while maintaining the underlying vascular geometry. To investigate the utility of this method for mutant mouse phenotype analysis, contrast-enhanced micro-MRI data were acquired from Gli2-/- mutant embryos and their wild-type littermates, showing several previously unreported vascular phenotypes in Gli2-/- embryos, including the complete absence of the basilar artery. These results demonstrate that contrast-enhanced micro-MRI provides a powerful tool for analyzing vascular phenotypes in a variety of genetically engineered mice.
During mammalian embryogenesis, the vasculature forms as a highly complex three-dimensional (3D) system of interconnected blood vessels that provide the developing tissues with essential nutrients, growth factors, and oxygen. Studies using genetically engineered mice have provided critical data on the genetic and molecular factors controlling vascular development during embryogenesis and also during disease progression involving abnormal formation of blood vessels. In these studies, quantitative 3D analysis of vascular patterns is seldom performed since commonly used histologic methods are inherently two dimensional and prone to sectioning artifacts, making 3D reconstructions difficult. Volumetric imaging methods such as magnetic resonance micro-imaging (micro-MRI) could provide powerful new tools to study vascular patterning in both normal and mutant mouse embryos.
Vascular development during embryogenesis can be divided into two key processes: vasculogenesis and angiogenesis. The earliest step, vasculogenesis, involves the recruitment and assembly of endothelial-cell precursors, called angioblasts, into a primary vascular plexus (1). These primitive blood vessels are then extended and elaborated by the process of angiogenesis, which involves cell proliferation, differentiation of the arterial and venous components, and the establishment of the final 3D patterns of the vascular networks (2). While vasculogenesis is a process that occurs only during early embryonic development, angiogenesis is an ongoing process that can also occur after birth, depending on the needs of different tissues, which can be advantageous, as in angiogenesis after ischemia (3), or pathologic, as in tumor angiogenesis (4).
The development of the mammalian vasculature results in apparently stereotypical patterns in a number of vascular subsystems, including the cerebral arteries. However 3D vascular patterning during brain development has not been investigated in detail, in large part due to the lack of analytical tools. We are particularly interested in analyzing the 3D patterns of the cerebral arterial inputs and how these patterns change during normal development, and in mouse mutants after genetic alterations that disrupt vascular patterning. The primary vascular inputs into the developing brain are the basilar artery that perfuses the more posterior parts of the brain and the carotid arteries that supply blood to the more anterior brain regions. In this report, we demonstrate contrast-enhanced micro-MRI methods that enable 3D analysis of both cerebral arterial systems over a wide range of embryonic stages.
All procedures involving mice were approved by the Institutional Animal Care and Use Committee at New York University School of Medicine. Timed-pregnant ICR strain mice (Taconic Farms, Germantown, NY) were used for studies of normal embryonic development, where the stage was denoted as the embryonic day (E) following identification of a vaginal plug after overnight mating (E0.5 was defined as noon of the plug day). Embryos were analyzed at stages between E10.5 and E17.5. For mutant analysis, Gli2 heterozygous mutant mice were bred to generate Gli2-/- homozygous null mutant embryos, as well as Gli2+/- (heterozygous mutant) and Gli2+/+ (wild type) littermate controls. Polymerase chain reaction of embryo yolk sac DNA was used to identify genotypes, using primers for Neo and wild-type Gli2, as previously described (5).
The procedure for vascular perfusion-fixation of mouse embryos was very similar to the protocol described previously by Smith et al. (6,7) and Smith (8). Pregnant mice were anesthetized using a ketamine (100 mg/mL):xylazine (20 mg/mL) mixture (0.2 mL/30g body weight, injected intraperitoneally). Individual embryos were surgically removed from the uterus after laparotomy, maintaining the vascular connections to the placenta and warmed in 37°C phosphate-buffered saline. Phosphate-buffered saline with heparin (5000 units/L) was perfused through the umbilical vein at a rate of 0.15 mL/min using an Ismatec Planetary Gear Driven Digital Pump (Cole-Parmer, Vernon Hills, IL), followed by perfusion of a fixative perfusion (2% [volume/volume] glutaraldehyde/1% formalin in phosphate-buffered saline at 300 mOsm/L), and finally the contrast agent, consisting of bovine serum albumin (BSA) conjugated to diethylenetriaminepentaacetic acid (DTPA) and chelated to gadolinium (BSA-DTPA-Gd; 1 mM), was dissolved in a 10% (wt/vol) gelatin solution (Sigma-Aldrich, St. Louis, MO) and perfused through the umbilical vein, allowing the contrast agent to completely fill the embryonic circulatory system. After perfusion, the umbilical vessels were securely tied with silk suture (Sofsilk 5-0; Fisher Scientific, Pittsburgh, PA), and the embryos were immersed in 4°C fixative to solidify the gelatin and to completely fix the tissues. Embryo samples were then maintained at 4°C until imaging with micro-MRI. Nontoxic food dyes were utilized to visualize the successive perfusions of phosphate-buffered saline, fixative, and contrast agent/gelatin under a dissection microscope (Nikon, Tokyo, Japan).
To selectively enhance the cerebral arteries, the BSA-DTPA-Gd contrast agent/gelatin solution was perfused through the umbilical artery rather than vein and stopped after 15 sec (E12.5), 25 sec (E15.5) or 45 sec (E17.5), when the embryo head first changed color, indicating that the contrast agent had reached the brain. The embryos were then immersed overnight in 4°C fixative and prepared for micro-MRI exactly as described above.
MRI experiments were performed on a 7-T micro-MRI system, consisting of a Biospec Avance II console (Bruker Biospin MRI, Ettlingen, Germany) interfaced to a 200-mm horizontal bore, superconducting magnet (Magnex Scientific, Yarnton, UK) with an actively shielded gradient coil (Bruker BGA-9S; 90-mm inner diameter, 750-mT/m gradient strength, 100-μs rise time). A quadrature Litz coil (Doty Scientific, Columbia, SC) (inner diameter = 25 mm; length = 22 mm) was used to image the embryos mounted inside a 30-mL syringe (Fig. 1). To increase throughput, multiple embryos (Table 1) were glued into place (Krazy Glue; Elmer's Products Inc., Columbus, OH) in the 30-mL syringe, on each side of the plunger. The mouse embryo samples were then immersed in Fomblin (Solvay Solexis; Fisher Scientific), a liquid perfluoroether that matches magnetic susceptibility without producing background signal. Imaging was performed overnight with a 3D T1-weighted gradient echo sequence (echo time, TE = 5 ms; repetition time, TR = 50 ms; flip angle = 35°; field of view = 25.6-mm; matrix size = 512 (3); isotropic resolution = 50 μm; total imaging time = 14 h 35 min).
Image analysis and segmentation were performed using the Analyze software package (V7.0; AnalyzeDirect, Overland Park, KS). First, 16-bit micro-MRI datasets were imported into Analyze to allow visualization and quantitative analysis. The following steps were then performed for each dataset: (i) 3D maximum intensity projections (MIPs) of the embryonic vasculature were computed using a user-defined threshold and automatically segmented using the “Tree Analysis” function, in which a 3D skeletal structure is automatically extracted from a volume of MR data. The tree can be displayed in conjunction with a volume-rendered image and edited, if necessary, to remove gaps due to insufficient vascular contrast in approximately 20% of the samples analyzed; (ii) segmentations were refined by adjusting the threshold for the computation of the MIPs; and (iii) quantitative data on the branching angles and segment lengths were obtained automatically using the “Stats” function, which allows the user to set parameters for the calculation of tree statistics, including calculations of branch lengths, angles, branching area, and brightness area product. For quantification, bilateral measurements of each parameter were averaged in each embryo before statistical analyses. Signal-to-noise ratio was measured as the mean signal within a region of interest in the brain, divided by the standard deviation of the signal in a region of interest in the background (Fomblin).
For coregistration analysis at E17.5, images were processed to form an average image representative, as described in previous reports (9-13). The average was formed through a series of registration steps, using software provided by the Mouse Imaging Centre (Toronto, Canada) and utilizing tools available from the Montreal Neurologic Institute (Montreal, Canada). Briefly, images were registered to a reference to remove extrinsic positional differences, and the image intensities were corrected for spatial nonuniformity and normalized (14). An unbiased average was created from an average of all the pairwise 12-parameter registrations. The average was subsequently refined through an iterative process of coregistering all images to the average at progressively finer scales, through six iterations at grid resolutions of 700, 600, 500, 240, 120 and 60 μm (11).
For quantitative analysis, the two-tailed Student's t test was used to test for statistical significance between measurements of embryos at different stages of development.
For confirmation of basilar artery phenotype, histologic analyses were performed. After micro-MRI, embryos were removed from Fomblin and washed for 10 min in 0.1% phosphate-buffered saline–Triton X-100 (Sigma-Aldrich). Following equilibration with 15% and 30% sucrose solution, embryos were embedded in optimal cutting temperature compound (Tissue-Tek; Sakura Finetechnical Co., Tokyo, Japan) and frozen for cryosectioning (16-μm sections). Immunohistochemistry for detection of CD31 used a standard protocol for staining of frozen sections using rat anti-CD31 antibody (BD Biosciences, San Jose, CA) at 1:100 dilution and biotinylated secondary antibody (biotinylated polyclonal goat anti-rat immunoglobulin; BD Biosciences). Tissues were then treated with streptavidin-horseradish peroxidase (BD Biosciences) followed by incubation in a 10 mg/mL 3,3′-diaminobenzidine tetrahydrochloride solution (Sigma-Aldrich). Images were acquired on a Leitz DM RXE microscope (Leica Microsystems, Wetzlar, Germany) with a charge-coupled device camera (Spot Insight; Diagnostic Instruments, Sterling Heights, MI).
As previously reported (6,7,15), umbilical venous perfusion of the BSA-DTPA-Gd contrast agent/gelatin solution resulted in pronounced enhancement of the major embryonic vascular structures on 3D T1-weighted micro-MRI (Fig. 2). MIP processing was especially useful for visualizing the 3D vascular patterns (Fig. 2d), enabling analysis of the vasculature in fixed mouse embryos over a wide range of embryonic stages, starting at E10.5 (Fig. 3). Major blood vessels, including the aorta, cardinal vein/vena cava, and umbilical artery and vein, were readily identified from E10.5, while elaboration of the hepatic vasculature was apparent from E12.5, and the intersomitic vessels could be seen as early as E11.5. In the head, the major blood vessels were visualized starting at E11.5, especially the large veins in the facial region surrounding the brain.
There are two main arterial input systems in the developing embryonic brain: the (medial) basilar artery and the (bilateral) carotid arteries. As expected from previous reports (6,7), anterograde (in the normal direction of blood flow) contrast perfusion through the umbilical vein resulted in enhancement of both arteries and veins (Figs. 2, ,3).3). In the MIPs of the embryonic mouse brain, this tended to emphasize the larger draining veins surrounding the brain, making it difficult to identify and analyze the smaller cerebral arteries (Fig. 4a). To enable more effective visualization of the arterial systems, we developed a first-pass retrograde perfusion protocol to deliver the contrast agent more directly to the basilar and carotid arterial systems, stopping the perfusion before the contrast agent filled the larger venous vessels (Fig. 4b). This method was more effective at later embryonic stages (E17.5) compared to earlier stages (E12.5, E15.5; Fig. 5). This was likely due to the maturation of the cardiac valves at later embryonic stages, resulting in a more effective bypass of the heart and delivery of contrast agent to the brain during retrograde perfusion (Fig. 4). Despite partial venous labeling at the earlier stages, the first-pass retrograde perfusion method improved the visualization of the arterial anatomy in the brain from E12.5, enabling analysis of many arterial branches, including the basilar artery, anterior inferior cerebellar arteries, posterior cerebral arteries, middle cerebral arteries, posterior communicating arteries, internal carotid arteries, and anterior cerebral arteries (Fig. 5).
To examine the degree of geometric similarity between cerebral artery patterns in same stage mouse embryos, we performed 3D image coregistration at E17.5 (N = 6; Fig. 6). After image coregistration and averaging, the resulting brain images showed an increase in signal-to-noise ratio from 27 ± 10 (N = 6) in individual images compared to 46 in the averaged image. Since the volume chosen for registration was limited to the brain, the averaging was most effective on the brain and cerebral blood vessels and produced some blurring and distortion in regions of the head and face further removed from the brain. The resulting averaged MIPs emphasized the primary arterial vessels, including the carotid and basilar arteries and the vessels making up the circle of Willis. In contrast, many of the secondary arterial branches of both the basilar and carotid systems were not preserved in the averaged MIPs. Surprisingly, the larger venous vessels that were largely suppressed in individual MIPs after first-pass retrograde perfusion became visible in the average MIPs as a result of the increased signal-to-noise ratio after registration and averaging. This result suggests that, similar to the arteries, the primary venous patterns in the head and brain are also stereotyped.
To further assess the variability in the branching patterns between individual embryos, vascular “trees” were extracted automatically from the 3D MIPs of single embryos at each developmental stage (Fig. 7a). These 3D branching structures were then examined from different directions, and quantification of branch lengths and angles was performed. Analysis of six pairs of secondary branches revealed standard deviations of less than ± 20% for the lengths and angles at each embryonic stage (Fig. 7b,c). In general, cerebral arteries continuously grew in length between E12.5 and E17.5, with the largest changes occurring between E12.5 and E15.5 (33 ± 10% increase in length; Fig. 7d). Interestingly, the angles between branches were relatively constant throughout these stages of brain development, suggesting that the geometrical patterns are determined early and then grow by extending the segments between arterial branch points.
To investigate the potential of these approaches for detecting and analyzing vascular phenotypes in mutant mouse embryos, we performed micro-MRI studies of E17.5 Gli2-/- homozygous mutants and their wild-type littermates (Fig. 8). 3D micro-MRI of Gli2-/- mutants clearly demonstrated the previously reported reduction of midbrain and cerebellum (data not shown) (16-18). After perfusion-fixation, micro-MRI also showed a leakage of the BSA-DTPA-Gd contrast agent into extra-vascular spaces in the Gli2-/- mutants, but not the wild-type mice, suggesting that the mutants have defects in vascular integrity, both in the brain and in tissues outside the brain (Fig. 8a,b). The extra-vascular leakage of contrast agent was also reflected in a decreased signal-to-noise ratio measurement in brain tissue (16 ± 5 [N = 4] in Gli2-/- mice, compared to 39 ± 10 [N = 4] in wild-type mice). This permeability phenotype reduced the contrast enhancement of blood vessels in Gli2-/- mice, especially inside the brain, but the carotid arteries were still visible as they entered the ventral side of the brain. Unexpectedly, micro-MRI revealed a complete absence of the vertebral and basilar arteries in Gli2-/- mutant embryos, close to the mid-hindbrain morphologic defects noted above (Fig. 8c,d). The absence of the basilar artery was confirmed by histologic analysis (Fig. 8e,f). In addition to the major basilar artery phenotype, we also observed differences in the patterning of the carotid arteries in Gli2-/- mutants (Fig. 8c,d). Specifically, the internal carotid arteries showed a significant constriction between the left and right arterial systems compared to wild-type littermates (569 ± 107 μm [N = 4] in Gli2-/- mice compared to 1513 ± 480 μm [N = 4] in wild-type mice), most likely a secondary effect resulting from the altered brain morphology.
Previous reports have demonstrated the utility of micro-MRI as a tool for analyzing the embryonic mouse cardiovascular system, with the focus on the heart and adjacent blood vessels, using either (umbilical vein) contrast perfusion methods (6,7,15,19) or relying on endogenous contrast without contrast agent perfusion (20-23). Previous micro-MRI studies have used multiple-embryo imaging, acquiring data from up to 32 embryos simultaneously to dramatically increase the effective throughput for MRI-based phenotype analysis (15,20,21,23). In this study, multiple embryo micro-MRI was also employed to significantly improve the throughput of 3D data acquisition. In addition, the previously described contrast perfusion-fixation protocol was modified to selectively enhance the cerebral arteries, enabling the first detailed 3D analyses of the basilar and carotid arterial systems, the two major vascular inputs into the developing brain. Employing image processing and analysis tools, quantitative data were provided to show that there is a high degree of similarity in the 3D patterns of the carotid and basilar arteries and in the circle of Willis of late-stage mouse embryos. Quantitative analysis also demonstrated that vascular patterns are most likely determined at early stages of brain development and grow by extending branch segments, while preserving the underlying geometry of the branch points. Finally, contrast-enhanced micro-MRI was used to analyze previously unreported vascular phenotypes in Gli2-/- mutant mice.
Recently, optical projection tomography has been reported for the analysis of mouse vascular development (24,25). This method has the advantage of much higher spatial resolution compared to micro-MRI, as well as the ability to use fluorescently labeled antibodies to image molecular markers. However, optical projection tomography is limited to tissue penetration and sample size of approximately 1 cm3. This, together with the limited penetration of antibodies in whole embryo samples, makes optical projection tomography best suited to early embryonic stages (E8.5-10.5), providing complementary information to micro-MRI at later stages (≥E10.5). Another optical method that has been proposed for the analysis of mouse development is episcopic fluorescence image capture, in which a stack of images is acquired by photographing the face of the block containing an embryo after each section is cut (26). Episcopic fluorescence image capture is therefore a destructive method, but it allows for the acquisition of distortion-free, high-resolution 3D images for the analysis of gene expression patterns and embryonic anatomy the entire range of developmental stages. Micro-computed tomography has been used for 3D imaging of the brain vasculature in adult mice (27) and could likely be adapted for imaging the embryonic vasculature after umbilical perfusion with an X-ray opaque contrast agent. Contrast-enhanced micro-computed tomography offers high resolution (20 μm) for vascular imaging (28), similar to the highest resolutions reported for micro-MRI (29,30), but provides limited contrast between tissues and requires high radiation doses, precluding in vivo studies in developing embryos. Optical projection tomography and episcopic fluorescence image capture are both limited to ex vivo imaging, while micro-MRI has the potential to be applied in vivo in the future, albeit at lower resolution than the 50 μm used in these studies.
Unlike micro-computed tomography, optical projection tomography episcopic fluorescence image capture, or micro-MRI, ultrasound biomicroscopy is an established real-time, in vivo, micro-imaging approach that enables in utero analysis of mouse cardiovascular development and hemodynamic studies in embryonic blood vessels, including the cerebral vasculature (31). Ultrasound biomicroscopy currently offers a clear advantage over micro-MRI in image acquisition time, and ultrasound biomicroscopy contrast-enhancement is possible using intravascular injection of micro-bubble agents. At this point, robust methods for 3D acquisition and image segmentation are still required to enable effective volumetric analysis of the developing vasculature with ultrasound biomicroscopy.
The formation and development of the vascular system is critical for normal embryonic development, and analyses in genetically altered mice are needed to further understand the genetic factors underlying vascular patterning. Our studies demonstrate clearly the important roles that micro-MRI can play in these analyses. The Sonic Hedgehog (Shh) signaling pathway is critical for normal development of the nervous system, and recent studies have also suggested an important role during vascular development, in a number of vessels including the dorsal aorta, yolk sac, and coronary and lung vasculatures (36-40) (reviewed in Nagase et al. (37)). At this point, no detailed analyses have been reported on the roles during vascular development of the mouse Gli genes, the major effectors of Shh signaling. In the nervous system, Gli2 has been shown to be the main activator of Shh signaling (18,36-38), and Gli2-/- mice die at birth, having a reduced midbrain and cerebellum (5,16-18). Micro-MRI demonstrated the previously described brain defects, as well as obvious vascular permeability defects and a completely ablated basilar artery. The fact that this major vascular phenotype had been overlooked, despite numerous previous studies (16-18,36,37,41-44), highlights the fact that analysis of vascular defects is difficult with traditional histologic methods and that more effective 3D methods like micro-MRI are required.
One exciting possibility is the future extension of micro-MRI for in vivo vascular imaging in mouse embryos. Utilizing MRI acquisition methods gated to physiologic motion, we recently demonstrated 100-μm (isotropic) resolution micro-MRI for in utero imaging of the brain (45) and hearts (46) of live mouse embryos. In combination with contrast agents targeted to the surface of vascular endothelial cells, this might provide an effective approach for in vivo, contrast-enhanced micro-MRI of the vascular system after intravascular injection of contrast agents. This would allow unprecedented longitudinal imaging studies to further understand the complex and dynamic processes involved in establishing the final 3D patterns of the mammalian vascular system.
We thank Daniel Stephen for help and advice on breeding the Gli2 mutant mice and Dr. JoMichelle Corrales for sharing her observation that Gli2-/- mice might have cerebral vascular defects. We thank Drs. John Sled and Mark Henkelman at the Mouse Imaging Centre in Toronto, Canada, for providing analysis software used for image coregistration and averaging.
Grant sponsor: National Institutes of Health; Grant number: R01 HL078665.