Developmental Atlas of the Early First Trimester Human Embryo

Abstract

Rapid advances in medical imaging are facilitating the clinical assessment of first trimester human embryos at increasingly earlier stages. To obtain data on early human development, we used magnetic resonance (MR) imaging and episcopic fluorescence capture (EFIC) to acquire digital images of human embryos spanning the time of dynamic tissue remodeling and organogenesis (Carnegie stages 13 to 23). These imaging data sets are readily resectioned digitally in arbitrary planes, suitable for rapid high-resolution three-dimensional (3D) observation. Using these imaging datasets, a web accessible digital Human Embryo Atlas (http://apps.nhlbi.nih.gov/humanaltas) was created containing serial 2D images of human embryos in three standard histological planes – sagittal, frontal, and transverse. In addition, annotations and 3D reconstructions were generated for visualizing different anatomical structures. Overall, this Human Embryo Atlas is a unique resource that provides morphologic data of human developmental anatomy that can accelerate basic research investigations into developmental mechanisms that underlie human congenital anomalies.

Introduction

Advances in medical imaging are allowing increasingly earlier assessment of human development in the first trimester, but there continues to be a gap in data on early human embryonic development for guiding such clinical evaluations. Insights into the dynamic and complex processes that shape human embryonic development will require acquisition of accurate morphological data on the dynamic remodeling of embryonic structures. Analysis of histological sections can provide two-dimensional (2D) views, but this alone is insufficient and three-dimensional (3D) reconstructions are necessary to elucidate the complex tissue remodeling that occurs during tissue morphogenesis.

Historically, 3D models of human embryos have been made from drawings of histological sections. Also well known are wax models of human embryos reconstructed from histological sections in the Carnegie Collection. These models were made from 2D wax plates constructed using photographs of paraffin serial sections. This technique was first introduced by Born and later modified by Heard and his colleagues (Born, 1883). Such solid reconstructions from serial sections were the primary imaging method used in classic embryology. Some of the reconstructed models and drawings in the Carnegie Collection are now archived in the Human Developmental Anatomy Center in Washington, DC (http://nmhm.washingtondc.museum/collections/hdac/index.htm). These have provided important insights into the early stages of human embryonic development. However, generating such reconstructions is labor intensive and time-consuming. Moreover, the results are not always satisfactory, as registration of serial sections obtained by conventional histology is often problematic given shrinkage and distortion of the sections (Yamada et al., 2007).

With the advent of magnetic resonance (MR) imaging, imaging of embryos with super conducting magnets ranging from 1.5T to 9T has proven to be highly advantageous (Effmann et al., 1988; Smith et al., 1992; Haishi et al., 2001), providing resolution of 40 μm/pixel or better with long scan times. As MR images can be digitally resectioned in arbitrary planes, this allows seamless interrogation of anatomic structures in the embryo. This is further aided by 3D reconstructions, which can be readily generated from the MR images. Thus MR imaging has proven to be highly effective for the analysis of animal and human embryos (Smith et al., 1996; Smith et al., 1999; Matsuda et al., 2003), with several atlases constructed for visualizing quail, mouse, and human embryonic development (Smith et al., 1999; Baldock et al., 2003; Ruffins et al., 2007).

Yet another novel method that can provide registered 2D image stacks suitable for rapid 3D rendering is episcopic fluorescence image capture (EFIC). With this imaging technique (Weninger and Mohun, 2002), the embryo is embedded in paraffin. Then using a sliding microtome, the block is sectioned and the block face is photographed with each successive cut using epifluorescence imaging to capture tissue autofluorescence visible at the block face. Incorporation of Sudan red in the paraffin wax blocks fluorescence bleed-through from deeper layers of the tissue. With each successive cut, the block is always returned to the same fixed photo-position for imaging, and thus the 2D image stacks generated are registered. This allows for virtual resectioning of the specimen in any imaging plane (Rosenthal et al., 2004; Weninger et al., 2006), and also rapid high resolution 3D reconstructions (Rosenthal et al., 2004).

In this study, we used both MR imaging and EFIC imaging to generate a Human Embryo Atlas spanning Carnegie stages 13 to 23, a developmental period when the major tissues and organ systems are formed. Therefore this atlas spans the critical window in development for future consideration of human congenital anomalies. For these studies, 52 human embryos from the Kyoto Collection were donated to the Carnegie collection and were imaged during accessioning into the Carnegie collection. Using the MR and EFIC imaging data acquired from these embryos, a digital Human Embryo Atlas was constructed. This Atlas is web accessible, and allows easy access online for viewing of 2D image stacks in three histological planes as QuickTime movies. In addition, selected 3D reconstructions and annotations of anatomic structures are provided. This digital Human Embryo Atlas should serve as an invaluable source of morphologic data delineating human embryonic development that will be a useful resource for both clinicians and biomedical scientists investigating human embryonic development and human congenital anomalies.

RESULTS

This study was carried out using a total of 52 human embryos spanning CS13 to 23. As the focus of this study was to construct an atlas delineating normal human development, the embryos were carefully prescreened, with 44 selected that exhibited no overt malformations. These 44 embryos provided 3–6 embryos for each stage spanning CS13 to 23 (Table 1). In addition, 8 embryos from CS16 to 20 with limb defects were analyzed as a pilot study of embryos with congenital malformations (Table 2; see below). Together, these embryos had crown to rump length ranging from 5 mm at CS13 to 28 mm at CS23 (Table 1; Nishimura et al., 1968). To maximize the imaging data that can be obtained from this invaluable human embryo resource, each embryo was first scanned by MR microscopy and then processed for EFIC imaging.

Table 1.

Normal human embryos analyzed by MR imaging and EFIC

CS	No. Embryos Analyzed	MR microcopy		EFIC		CRL
		n	(μm/pixel) †	n	(μm/pixel) †	(mm) ††
13	3	2	35, 37	1	6, 12	4.6–5.4
14	4	3	35, 37	2	2 – 8	5.7–6.4
15	3	1	35	2	2 – 6	7.1–7.7
16	5	3	37, 39	4	4 – 9	7.8–9.5
17	3	2	35, 39	2	5 – 11	9.8–12.6
18	5	2	54	3	7, 13	11.5–15.5
19	4	2	54, 59	2	7, 13	16.2–17.4
20	3	2	54, 58	1	7, 13	15.8–20.1
21	4	2	50, 56	2	7, 13	18.7–21.2
22	6	6	54, 62	0	N/A	20.6–23.5
23	4	3	52, 58	3	7, 13	26.2–28.9
Total	44	28		22

^†Pixel resolution.

^††Crown- rump (CRL) lengths were within normal range (Nishimura et al., 1968).

Table 2.

Analysis of abnormal human embryos^†

CS	No. Embryos Analyzed	MR microscopy††		CRL
		n	(μm/pixel)	(mm)
16	3	3	39, 42	8.8–9.0
17	1	1	42	13
18	1	1	50	14.2
19	1	1	85	14.3
20	2	2	54	17.5, 18.2
Total	8	8

^†Embryos with polydactyly

^††Pixel resolution.

MR Imaging of Human Embryos

Human embryos were scanned using a 7T vertical magnet, with scan parameters varied to optimize imaging of the different size embryos. At CS13, 13 hours of scan time provided resolution of 35μm/pixel with a 256×128×128 matrix. To maximize signal-to-noise ratio (SNR), the signals were averaged up to 70 times. The contours of the embryos were faithfully reproduced when compared with photographs of the specimens (Fig. 1A,B). In addition, internal structures can be observed in the MR images, such as the looping heart tube, primitive brain ventricles, the rhombomeric ridges in the hindbrain, as well as the optic and otic vesicles (Fig. 1C, 2E). At CS 16, the larger size of the embryo allowed signal averaging to be reduced to 30 times. However the scan time was increased to 16 hours to accommodate expansion of the matrix to 512×256×256. This protocol provided an image resolution of 38 μm/pixel. Structures such as the brain, spinal cord and heart are clearly seen (Fig. 1F), and even smaller structures, such as the root of the spinal nerves can be detected (annotation in Human Embryo Atlas; see below). At CS 18, although the resolution of the image was lower at 51 μm/pixel, the images obtained were superior due to the larger size of the embryo, now approximately 14 mm crown to rump length (Table 1; Fig. 1H, I). The shape of the brain ventricle was clearly delineated, as was the fine structure of the heart. At CS 22, towards the end of organogenesis, MR images at 58 μm/pixel allowed sharp delineation of all the major organs (Fig. 1K, L). At this time, even small structures, such as the longus capitis muscle can be observed (Fig. 3A). Images obtained in transverse planes provide detailed histological views similar to those used in the clinical setting (Fig. 3).

Figure 1. MR microscopy of human embryos at different developmental stages.

Human embryos at CS13, CS16, CS18 and CS22 were imaged by MR microscopy. Photographs of the human embryos are shown in (A, D, G, J). 3D reconstructions of the same embryos using MR images are shown in (B, E, H, K), and 2D MR images in the sagittal plane are shown in (C, F, I, L). At CS13 (A–C), limb buds are present, and the prosencephalon (P), mesencephalon (M) and rhombencephalon (R) with the rhombomeric ridges can be seen. Also notable is the looped heart tube (C). At CS16 (D–F), the eye primordia and upper and lower limb buds with limb paddles are observed (panel E). In the abdominal cavity, the midgut (Mg) is invested in the umbilical cord (UC) (panel F). At CS18 (G–I), anlagen for the digits are evident in the forelimb (panels G, H). In the thoracic cavity, the heart (Ht) and the lung are clearly seen (panel I). In the brain, the infundibulum and Rathke’s pouch, precursors of the hypophysis are evident (arrowhead in panel I). At CS22, craniofacial and eye development have advanced considerably, and the digits in the fore and hindlimbs are well developed (panel L). At this stage, most of the abdominal cavity is occupied by the liver (Li), and physiological midgut herniation is present (arrowhead in panel L).

Bars = 1 mm.

CF: cervical flexure, D: diencephalon, PF: pontine flexure, T: telencephalon, Tn: tongue.

Figure 2. Development of the brain in human embryos.

MR images of human embryos are shown as 3D volumes (A–D) and in 2D views (E–H) to illustrate development of the brain. At CS13 (A, E), the optic cup (Op) is seen as a prominent outpocketing from the diencephalic region of the prosencephalon (P). A 2D image in a transverse plane show the optic cup (Op) and the otic vesicle (Ot). The craniopharyngeal pouch (arrowhead in panel E) is in contact with the floor of the diencephalic region of the prosencephalon. Near the otic vesicle, the otic ganglion (OG) can be seen. At CS16 (B, F), connection of the optic cup to the prosencephalon by the optic stalk (arrowhead) can be seen. In the area of 4^th ventricle (4thV), the rhombomeres and median sulcus of the 4^th ventricle (arrow) are clearly observable. At CS19 (C, G), the cerebral hemispheres enlarge and the pontine flexure (arrow) is prominent (panel C). In the 2D section image (panel G), the lateral ventricles (LtV) and the 3^rd ventricle (3rdV) are clearly seen. The trigeminal ganglia (TG) are prominent. The junction of the vertebral arteries (arrow), the basilar artery, and the anterior inferior cerebellar artery (arrowhead) branching from the basilar artery can be seen at the center of this section. The cochlea (Co) is developing posterior to the trigeminal ganglion. The anterior inferior cerebellar arteries (arrowhead in G) are branching from the basilar artery downstream of the junction of bilateral vertebral arteries (arrow in G). By CS22 (D, H), the cerebral hemispheres have greatly enlarged and elongated posteriorly (D). In the 2D section image (H), the cerebral cortex around the lateral ventricles (LtV) has thickened considerably. The adenohypophysis is observed with the dorsum sellae (DS) just below the hypothalamus (Hp). The lumen of the internal jugular vein (IJV) is seen as a low-intensity lumen. Bars = 1mm.

Figure 3. MR images in different cross sectional views.

MR images obtained from a CS22 embryo are shown in different cross sectional views, (A–E), with the level of each section indicated in the photograph of the embryo shown in (F).

(A). An imaging plane through the lower part of the head shows the eye, external acoustic meatus (arrowhead), spinal cord (SC) and tongue (Tn). In the eye, the lens and neural retina (Re) is seen as regions of high intensity. The longus capitis muscle (LC) is also apparent.

(B). An imaging plane through the upper chest. The pulmonary artery (PA) is seen connecting to the ductus arteriosus (Dc). The ascending aorta (AA), trachea (Tr), esophagus (Es), and forearm (FA) are also observed.

(C). An imaging plane through the lower chest. In this section, the right ventricle (RV), left ventricle (LV) and right atrium (RA) can be seen. The trachea bifurcates into bronchi (Br), with the left bronchus going into the left lung. The esophagus and descending aorta (DA) are also observed.

(D). An imaging plane through the mid-section between the thorax and abdomen. The lung (Lg) can be seen in the thoracic cavity, while the liver (Li) with the hepatic veins (HV) fills the abdominal cavity at this level.

(E). The abdomen, at the level of the umbilical cord. The liver (Li), stomach (St), and adrenal glands (Ad) are seen in the abdominal cavity. In the umbilical cord (UC), the umbilical artery (UA) emerges through the abdominal wall and the midgut (Mg) herniates into the umbilical cord.

Bar = 1 mm.

EFIC Imaging of the Human Embryo

Subsequent to MR imaging, the embryos were embedded in paraffin and processed for EFIC imaging. The theoretical resolution ranged from 2.34 to 13.4 μm/pixel (Table 1). The EFIC image stacks provided high-resolution surface reconstructions and also exquisite details of internal structures (Fig. 4). EFIC images of a CS14 embryo showed the endocardial cushions forming between the atrium and left ventricle (Fig. 4, 5A, D) and in the outflow tract (Figure 5D–F). These structures were difficult to discern in MR images of the same embryo (Fig. 5G–I). In the CS16 embryo, EFIC images showed many fine structures, such as the glossopharyngeal (IX in Fig. 6B) and vagus (X in Fig. 6B) nerves, the accessory nerve complex (XI in Fig. 6B), and vestibular and cochlear pouches (Fig. 6A, B). Comparison of MR microscopy and EFIC data sets from the same CS16 embryo showed the EFIC images provided more refined 3D reconstructions (Fig. 7A, E), and superior 2D images with fine details, some of which were simply not detectable in the MR images. For example, contact between the hypophyseal pouch (Rathke’s pouch) and the floor of the diencephalon is clearly delineated in the EFIC image, but this was difficult to discern in the MR images (see asterisk in Fig. 7B, C).

Figure 4. EFIC imaging of human embryos at CS14, 16, and 18.

Human embryos at CS14 (A–C), CS16 (D–F), and CS18 (G–I) were processed for EFIC imaging. Photographs of the human embryos are shown in (A, D, G). 3D reconstructions of the same embryos using EFIC images are shown in (B, E, H), and 2D EFIC images in the sagittal plane are shown in (C, F, I). EFIC imaging provides exquisite anatomical details of the embryo, such as the endocardial cushions at the atrioventricular junction (panel C), and Rathke’s pouch (RP). In the CS16 sagittal EFIC section, the tongue (Tn), the sclerotomes (Sc), and gut are evident. At CS18, the 3D reconstructions show digits in the forming limb bud (panel H). Also observed are the pontine flexure (PF), the cervical flexure (CF), the heart (Ht), the lung (Lg), liver (Li) and stomach (St). D: diencephalon, M: mesencephalon, Mg: midgut, R: rhombencephalon, UC: umbilical cord. Bars = 1 mm.

Figure 5. High resolution EFIC images in three orthogonal views.

CS14 embryo was processed for EFIC imaging, with three orthogonal imaging planes obtained from digital resectioning of the image stack: A: sagittal, B: frontal, and C: transverse. The heart region is shown in magnified view in (D–F). The outflow tract is observed to emerge from the presumptive right ventricle (RV). Also seen is the right atrium (RA) and left ventricle (LV) (panel C and D–F). For comparison, MR images of the same embryo are shown in similar imaging planes at similar magnifications in panels (G–I). At: atrium, NT: neural tube, OC: optic cup. Bars = 500μm

Figure 6. EFIC 2D image stacks of human embryo at CS16.

EFIC 2D images from different axial levels of a CS16 human embryo is shown in (A–G). The axial level of each 2D image in panels (A–G) is delineate in panel (H). In the most rostral section (A), the mesencephalon (M), rhombencephalon (R), trigeminal ganglion (TG) and vestibular pouch (VP) can be observed. Further caudally at the level of the eye (B), the floor of the diencephalon (D) and hypophyseal pouch can be seen (arrowhead). Also observed are the cochlear pouch (CP), the glossopharyngeal nerve (IX) and complex comprising the vagus and accessory (XI) nerves. At the level of the heart (C, D), the outflow tract (OFT), left and right atria (LA, RA), and right and left ventricle (RV, LV) are present. Note at the level of the ductus venosus in the abdominal cavity (E), most of the abdomen is occupied by the liver (Li). Also seen in this section are two umbilical arteries (arrowheads) and an umbilical vein (asterisk) in the umbilical cord. A section through the abdomen at the level of the stomach (St) (F) shows the mesonephros (Mn) and liver (Li) and also seen is the umbilical cord with physiological herniation of the midgut (Mg). A section through the posterior part of the abdomen shows the descending aorta (DA). The mesonephric tubules (MT) are clearly evident in panel G.

Bar = 1 mm.

LL: lower limb, NC: Neural canal, NP: nasal placode, PC: peritoneal cavity, PP: primitive pharynx, SG: spinal ganglion, T: telencephalon, UC: umbilical cord, UL: upper limb, UV: umbilical vein.

Figure 7. CS16 Embryo Imaged By MR microscopy and EFIC.

A CS16 embryo was imaged by both EFIC (A–D) and MR microscopy (E–H). 3D volumes generated from the 2D image stacks obtained by EFIC and MR microscopy are shown in (A) and (E), while sagittal 2D slices in similar views obtained by EFIC and MR microscopy are shown in (B–D) and (F–H), respectively. In both sets of images, detailed anatomy of the developing organ systems can be observed including the central nervous system, atrium (At), right ventricle (RV), left ventricle (LV), and outflow tract of the heart. The EFIC images are more refined and provide much clearer definition of the primordia of the adenohypophysis, the hypophyseal pouch (arrow in B, C).

Bar = 1mm

M: metencephalon, Mg: midgut, R: rhombencephalon, St: stomach, T: telencephalon, UC: umbilical cord.

Construction of a Human Embryo Atlas

A web based Human Embryo Atlas (URL: http://www.cac.med.kyoto-u.ac.jp/atlas.html) (Access via Log in: Human and Password: Embryo) spanning Carnegie stage CS13 to 23 was generated using the MR and EFIC data (Fig. 8). For each Carnegie stage, EFIC and MR imaging stacks of serial 2D images obtained are presented as QuickTime movies in three orthogonal views that corresponded to the standard sagittal, frontal and transverse histological planes. An example of this can be seen in Figure 5, where the same embryo is shown in the sagittal (Fig. 5A), frontal (Fig. 5B), and transverse (Fig. 5C) plane. In addition, selected 3D reconstructions were included in the Atlas as QuickTime movies or as interactive QTVRs. Viewing 2D image stacks as a video sequence via the QuickTime movies allow easy tracing of rapidly changing anatomic structures in the embryo, such as the rapidly evolving embryonic circulation (see Atlas QuickTime Movie for CS22, 23). Thus the aortic and pulmonary outflow from the heart can be seen to connect with the descending aorta and umbilical artery, followed by return via the umbilical vein to the embryonic circulation (Fig. 3, 6; see Human Embryo Atlas QuickTime Movie for CS22, 23). Some annotations were made to show anatomic structures in the embryo (Fig. 8C). We briefly highlight below, various tissues and organs that can be observed in the Human Embryo Atlas. Where appropriate, navigation to specific QuickTime movies of 2D image stacks are indicated.

Figure 8. Atlas of the Human Embryo.

A web based Human Embryo Atlas has been constructed using digital images obtained from MR microscopy and EFIC imaging of human embryos from CS13 to CS23. The top panel shows the home page of the Human Embryo Atlas website (A). For each developmental stage, 2D stacks of MR microscopy images in three orthogonal planes are available for viewing as QuickTime movies. This is shown in the middle panel, where a transverse section of a CS22 embryo has been selected for viewing (B). There are two drop-down lists that can change the imaging modality (MRI or EFIC) and allow selection of the datasets for viewing (Whole body, Brain, and Heart). In addition, the Atlas contains selected annotations of key anatomic structures in different organ systems that can be selected for viewing from a drop-down list, such as cardiovascular or nervous system. When the pointer is moved over the image, the selected structure is highlighted. In the window shown in the bottom panel, the liver is selected and highlighted in green (C).

Nervous System

The major subdivisions of the central nervous system can be seen at CS13 with delineation of the prosencephalon, mesencephalon, and rhombencephalon (Fig. 2A). The optic and otic vesicles also are visible (Fig. 2A, E). At CS14, the EFIC image stacks show the otic and trigeminal ganglia, and the spinal ganglia (see Atlas: after login, select 2D Stacks, CS14, Brain, EFIC). The cleavage of bilateral cerebral vesicles occurs at CS15 (Fig. 2B at CS16), and the cranial nerves can be traced at CS16 (Atlas:2D Stacks, CS16, Brain, EFIC). At CS17, the lateral hemispheres of the brain are apparent (Lat V. in Fig. 2C, G at CS19), with the choroid plexus in the lateral ventricles seen at CS20 (Atlas:2D Stacks, CS20, Brain, MRI, transverse). By stage CS22, the collicular ridges can be seen (Atlas:Annotation, CS22, NervousSystem).

Cardiovascular System

At CS13, a looped heart tube is observed comprising a primitive ventricle with a common atrium (Atlas:2D Stacks, CS13, Heart, EFIC). The characteristics of each chamber of the heart are evident at CS15 (Fig. 6C, D at CS16), with septation of the four chambers completed by CS18 (Atlas:2D Stacks, CS18, Heart, EFIC, Transverse). The developing 3^rd and 4^th pharyngeal arch arteries are evident in EFIC images at CS13 and 14 (Atlas:2D Stacks, CS14, Heart, EFIC). Viewing the serial 2D image stacks shows that the 3^rd pharyngeal arch artery extends into the cranial region, giving rise to the future carotid artery, while the 4^th pharyngeal artery at these early stages forms symmetric left and right aortic arches (Atlas:2D Stacks, CS14, Heart, EFIC). By CS20, the circulatory system in the fetal period is established, with the pulmonary trunk arising from the right ventricle and the ascending aorta from the left ventricle. The right aortic arch and the right descending aorta are no longer present. The pulmonary artery connects to the descending aorta by way of the ductus arteriosus (Fig. 3B).

Digestive System

At CS 13, the stomach was observed to have a fusiform-shape, and the bile duct is seen branching from the midgut (Atlas:Annotation, CS13, DigestiveSystem). The volume of the liver increases markedly between CS14 and 15, such that by CS15 and beyond, the liver occupies most of the abdomen (Fig. 1I, L and 6E). The counterclockwise rotation of the gut began at CS16. In the stomach, the greater and lesser curvature are evident at CS17 (Atlas:Annotation, CS17, DigestiveSystem), and the fundus is apparent at CS18 (Atlas:Annotation, CS18, DigestiveSystem). At CS20, intrahepatic venous distribution is formed and resembles the segmentation in adults (Atlas:2D Stacks, CS19–21, Whole Embryo, MRI). At CS23, physiological midgut herniation (normal umbilical hernia) was observed, with considerable coiling of the intestine (Fig. 1L, 3E).

Respiratory System

The main bronchi can be seen branching and connecting with the right and left lung buds at CS13. The trachea and the esophagus are not yet septated (Atlas:Annotation, CS13, RespiratorySystem, Trachea), but become divided by CS14 (Atlas:Annotation, CS14, RespiratorySystem, Trachea; Annotation, CS14, DigestiveSystem, Esophagus). The lobes of the lung start forming at CS15, with the bronchi branching into the lung lobes at CS16 (Fig. 9A), and by CS18, the bronchi has branched into each lung lobe (Atlas:2D Stacks, CS18, Heart, EFIC). At CS20, the two pulmonary arteries emerge from the pulmonary trunk and insert into the lung (arrowhead in Fig. 9B; Atlas: Annotation, CS20, RespiratorySystem). This was only seen in the higher resolution EFIC images.

Figure 9. Enlarged views of EFIC and MR images showing different anatomic structures in the developing human embryo.

EFIC (A–E) and MR (F) images were enlarged to show different anatomic structures in the human embryo. In a CS16 embryo (A), branching of the bronchi into each lobe can be seen (arrowhead). At CS20 (B), the pulmonary artery (arrowhead) can be seen to enter the lung at the hilus. At CS18 (C–E), the gonad (arrow) and metanephros (arrowhead) and hindgut (Hg) can be seen in panel (C). A sagittal section at the diaphragmatic area (D) shows the stomach (St), the adrenal gland (arrowhead) and the gonad (arrow). A 3D reconstructed view (E), in which the liver has been digitally removed, allows visualization of the adrenal gland (arrow) under the embryonic lung and just anterior of the vertebrae (Vr). In a sagittal view of a CS22 embryo (F), the kidney (arrowhead) can be detected at the retroperitoneal area, and the adrenal gland (arrow) is seen just above the kidney.

AAo: ascending aorta, Br: left bronchus, DAo: descending aorta, Es: esophagus, LA: left atrium, LV: left ventricle, MD: mesonephric duct, MN: metanephros, Or: oral cavity, Pl: pleural cavity, Rb: rib, Tr: trachea.

Kidney Development

At CS14, the mesonephric ducts are notable (Atlas: 2DStacks, CS14, Whole Embryo, EFIC; Annotation, CS14, UrogenitalSystem). The metanephros can be seen at CS15 (Atlas:Annotation, CS15, UrogenitalSystem), with the primitive renal pelvis and the ureter detectable at CS16 (Atlas:Annotation, CS16, UrogenitalSystem). The gonad can be identified at CS18 (see white arrows and arrowheads in Fig. 9C), and as described previously (Shikinami, 1926), the metanephros was observed to shift from the pelvic region to the abdominal cavity as development progressed (Atlas:Annotation, CS19–23, UrogenitalSystem, Kidney).

Endocrine System

At CS13, development of the pituitary gland is initiated with Rathke’s pouch (craniopharyngeal pouch) contacting the floor of the diencephalon (Ikeda et al., 1988; Yamada et al., 2004) (Fig. 2E). Rathke’s pouch is prominent at CS14, while the neurohypophysis (infundibular recess) at the floor of the diencephalon was apparent at CS16–17 (Atlas:Annotation, CS14, 17, EndocrineSystem). The stalk connecting the oral cavity and the craniopharyngeal pouch is closed at CS20–21 (Atlas:Annotation, CS21, EndocrineSystem). The adrenal gland was also seen at CS18 (asterisk in Fig. 9D,E, also seen in Atlas: 2DStacks, CS18, Normal-Embryo, EFIC), with the adrenal gland and the kidney being of similar size during this period of development (Fig. 9F).

Analysis of Abnormal Human Embryos

As pilot to investigate embryos with congenital anomalies, we selected 8 human embryos with limb anomalies, with the hope that such embryos may harbor other congenital malformations of the internal organs. These 8 embryos all exhibited polydactyly associated with preaxial digit duplications and spanned Carnegie stages 16 to 20 (Table 2). Each embryo was MR scanned, and the 2D images collected were carefully analyzed for possible defects in internal tissue and organ structures. However, no obvious anatomical defects were observed beyond the limb anomalies. The data collected from these 8 embryos were included in the totality of data obtained from all of the human embryos used to assemble the Human Embryo Atlas.

DISCUSSION

A digital Human Embryo Atlas spanning CS13 to 23 was constructed using data obtained by MR and EFIC imaging. Embryos were first scanned by MR microscopy and then processed for EFIC imaging. EFIC yielded higher resolution images, and thus was superior for analyzing the smaller early stage human embryos from CS13–18. At these developmental stages, MR datasets gave good external contour views of the embryo, but fine structures of the inner anatomy could not be resolved. For example, endocardial cushions in the heart or Rathke’s pouch in the forming hypophysis were observed in the EFIC but not MR images. Imaging of phantoms suggests the optical resolution of EFIC is approximately 5–6 um (Lo, unpublished observations). At this resolution, it is possible to see fine structures, such as the smaller branches of blood vessels or developing tubules and collecting ducts in the developing kidney. While such structures may also be observed by standard paraffin histology, the advantage with EFIC imaging is the ability to digitally reslice the images for viewing structures in different imaging planes and also as 3D reconstructions.

For embryos older than CS18, EFIC imaging was problematic. The larger size of the later stage embryos required prolonged processing time for paraffin infiltration and embedding, and as a result, the tissue autofluorescence required for EFIC imaging was markedly reduced. This is likely due to the 30–40 year immersion in fixative of these historical specimens, and not merely the larger size of these later stage embryos. For comparison, mouse fetuses or newborn mice of similar size to CS23 human embryos do not present problems for EFIC imaging (Rosenthal et al., 2004). Nevertheless, MR imaging of these later stage human embryos yielded good quality imaging data that delineated fine details of internal anatomy. Thus with the combination of MR and EFIC imaging, we were able to construct a Human Embryo Atlas that spanned the entirety of CS13 to 28.

MR imaging has one important advantage in that it is noninvasive and nondestructive, and has been applied for the past 20 years to analyze embryonic development in different animal models (Bone et al., 1986; Smith et al., 1992; Smith et al., 1994; Smith et al., 1996). With recent advances in MR imaging (Matsuda et al., 2003), resolution of 35 μm/pixel or better is now possible. However, the cost of MR instrumentation and the technical support required makes this imaging modality out of reach for many investigators. In contrast, the standard sliding microtome used in EFIC imaging is commonly available in many pathology laboratories. The low light digital camera and epifluorescence stereomicroscope required for capturing EFIC images are also standard equipment available in many biomedical research laboratories. In addition, EFIC imaging is currently still higher in resolution and thus superior for imaging small specimen.

The ability to digitally resection MR and EFIC imaging data in arbitrary planes has tremendous advantage for analyzing the dynamic changes in anatomic structures in the developing human embryo. In the online Human Embryo Atlas, 2D image stacks in three histological planes are available for viewing or download as QuickTime movies. These serial 2D image stacks can be digitally resectioned in any plane to optimize viewing of different tissue and anatomic structures. 3D reconstructions are also easily generated for further interrogating the relative position of tissues and organs in the developing embryo.

This Human Embryo Atlas will be useful for the clinician and biomedical scientists studying human embryonic development and the etiology of human congenital anomalies. In addition, this atlas will be a valuable teaching tool for learning embryology and human developmental anatomy. We note embryology is often minimally covered in medical education (less than 20 hours; (Drake et al., 2002). This web accessible Human Embryo Atlas should serve as a valuable tutorial of human embryology (Watt et al., 1996; Aiton et al., 1997; Carlson, 2002; Puerta-Fonolla et al., 2004; Arroyo-Jimenez Mdel et al., 2005; Yamada et al., 2006), especially as more detailed annotation of the atlas is developed. To this end, we have constructed the Human Embryo Atlas website with a function for users to propose new annotations or revisions of existing annotations. This user interface will allow the biomedical community as a whole to contribute to the development and further refinement of the digital Human Embryo Atlas, making it a more accurate and valuable tool for teaching and research.

Given that much of our present understanding of human embryonic development is extrapolated from insights gained from studying the mouse embryo, access to human embryo data is necessary to ascertain possible differences between mouse vs. human embryonic development. Moreover, availability of the Human Embryo Atlas will be helpful for elucidating the pathogenesis of human congenital anomalies. Since most of the human embryos in the Kyoto collection were derived from pregnancies terminated for socioeconomic reasons under the Maternity Protection Law of Japan (Yamada et al., 2004; Yamada et al., 2006), they are expected to represent normal embryonic population in the uterus. Most of the embryos included in this study were normal in appearance. As a pilot, we also analyzed a few embryos with preaxial digit duplications. However, the data collected from the latter embryos showed no detectable change in internal anatomic structures. Nevertheless, the body of data collected from this study will provide a solid basis for launching a more broad based study of human embryos with congenital anomalies.

In the future, expansion of the Atlas with the analysis of more human embryos will help to further refine the temporal profile of human embryonic development. Analysis of imaging data from human embryos with congenital anomalies may yield novel insights into the developmental origin of human birth defects. Overall, this Human Embryo Atlas is a unique resource that provides morphologic data of human development that can facilitate clinical evaluation of congenital anomalies, and accelerate basic research investigations into developmental mechanisms that underlie human congenital anomalies.

EXPERIMETNAL PROCEDURES

Human embryo specimens

For this study, 52 well-preserved whole human embryos ranging from CS13 to CS23 were analyzed. These human embryos are historical specimens collected and stored at the Congenital Anomaly Research Center of Kyoto University (Nishimura et al., 1968; Yamada et al., 2004), and donated and accessioned into the Carnegie Collection. In most cases, pregnancy was terminated during the first trimester of pregnancy for socioeconomic reasons under the Maternity Protection Law of Japan. As the attending obstetricians did not examine the aborted embryos, these embryos can be considered as representative of the total intrauterine population. The embryos were measured, examined, and staged using Carnegie staging criteria (O’Rahilly and Müller, 1987). 44 embryos were diagnosed as externally normal and their crown-lump lengths (CRLs) were within normal range reported previously (Nishimura et al., 1968). Another 8 embryos with polydactyly were selected for a pilot study of MR screening for congenital malformations.

MR microscopy

For MR imaging, human embryos were soaked for 3–7 days in a gadolinium-based MR contrast agent (BSA-DTPA-Gd; Sigma-Aldrich, St. Louis, MO) mixed with 4% formalin. Subsequently, the solution was changed into 4% formalin. The solution was washed three times to remove any residual contrast agent from the surface of the embryos. Embryos were placed in glass vials (5–30 mm) containing formalin for MR scanning. MR imaging was carried out in the NIH Mouse Imaging Facility (MIF) using a 7.0 T vertical bore Bruker Microimaging MRI system (Bruker, Billerica, MA) The pulse sequence consisted of rapid snapshot flash (SNAP) gradient echo sequence with TR = 30–35 msec, and TE = 3.3–3.8 msec. The total scan time was 13–40 hours with the acquired MR signals averaged 12–48 times to achieve higher image resolution with image matrix of 256×192×192 to 512×512×512. The final resolution of the images ranged from 30–117μm. MR images were obtained in DICOM format and subsequently converted into TIFF format using ImageJ (http://rsb.info.nih.gov/ij/).

EFIC Imaging

For the EFIC imaging, embryos stored in neutral buffered formaldehyde were dehydrated in ethanol and xylene, infiltrated and embedded in 70.4% paraffin wax, containing 24.9% Vybar, 4.4% stearic acid and 0.4% Sudan IV. The paraffin blocks were sectioned using a Leica SM2500 sliding microtome (Leica Microsystems, Bannockburn, IL). Autofluorescence at the paraffin block face was visualized using epifluorescent imaging with mercury illumination and excitation/emission filters of 425/480 nm, respectively. Fluorescence images were captured using a Hamamatsu ORCA-ER low light CCD camera (Hamamatsu, Bridgewater, NJ). Details of this technique have been reported previously (Weninger and Mohun, 2002; Rosenthal et al., 2004).

2D image stacks obtained by MR or EFIC imaging were manipulated using Openlab (Improvision/Perkin Elmer, Waltham, MA). 3D reconstruction and QTVR movies were generated by Volocity (Improvision/Perkin Elmer, Waltham, MA). The 2D image stacks were resectioned digitally to generate sagittal, transverse and coronal sections. In some cases, the 2D image stacks were digitally resectioned using MAPaint (http://genex.hgu.mrc.ac.uk/Software/paint/) or Volocity, allowing for detailed examination of anatomical structures.