The enteric nervous system (ENS) is composed of neurons and glial cells arising from the neural crest. The majority of the crest-derived cells that colonize the gut originate from the vagal level of the neural tube, adjacent to somites 1–7 (Le Douarin and Teillet, 1973). These cells migrate ventrally to enter the foregut, and then continue rostrocaudally along the length of the entire intestine. The sacral neural crest, adjacent to somites 33–39 (Yntema and Hammond, 1955; Catala et al., 1995), also contributes enteric neurons and glial cells to the intestine, principally the hindgut, in both mammalian (Serbedzija et al., 1991; Kapur, 2000) and avian (Le Douarin and Teillet, 1973; Pomeranz and Gershon, 1990; Burns and Le Douarin, 1998) embryos.
The timing, extent, and migratory route of sacral neural crest cell (NCC) colonization of the gut have been areas of long-standing controversy (Young and Newgreen, 2001; Burns, 2005). One area of disagreement has been whether sacral NCCs require the presence of vagal NCC to enter the hindgut. Cell labeling studies suggested that sacral NCC arrive in the hindgut as early as E4, well before the appearance of vagal NCCs (Pomeranz and Gershon, 1990; Serbedzija et al., 1991). In contradiction to these results, the hindgut was found to remain aganglionic after ablation of the vagal crest (Yntema and Hammond, 1954), explantation of the hindgut before vagal NCC arrival (Allan and Newgreen, 1980; Kapur et al., 1992; Young et al., 1998), or transection of the midgut distal to the migratory wavefront (Meijers et al., 1989). These experiments suggested that the sacral crest does not contribute to the hindgut independently. More recently, however, chick–quail chimeras demonstrated that even after ablation of the vagal crest, sacral NCCs are still able to colonize the hindgut (Burns and Le Douarin, 1998; Burns et al., 2000; Burns and Le Douarin, 2001). Similarly, cocultures of aneural chick hindgut with quail sacral neural tube also demonstrated not only that the sacral crest contributes enteric neurons to the avian hindgut, but that it can do so in the absence of vagal-derived cells (Hearn and Newgreen, 2000).
With regard to their migration, sacral NCCs in the mouse embryo migrate to the pelvic ganglia, where they reside for several days, at which time a subpopulation of cells moves into the hindgut mesenchyme (Kapur, 2000; Anderson et al., 2006). In the avian embryo, sacral NCCs are thought to migrate to the mesentery at the dorsal aspect of the gut at embryonic day (E) 4, where they form the nerve of Remak (NoR), an avian-specific ganglionated structure that extends from the cloaca to the proximal midgut. The cells appear to reside in the NoR for 3 days, before migrating into the gut wall at E7 along axons projecting from the NoR, at approximately the time that vagal NCCs reach the hindgut (Burns and Le Douarin, 1998). Sacral NCCs also form the pelvic plexus in avian embryos (Yntema and Hammond, 1955; Catala et al., 1995; Erickson and Goins, 2000), but little is known about the role of the pelvic plexus in formation of the avian ENS.
We undertook this study to understand the respective contributions of the NoR and the pelvic ganglia to the ENS. Using chick–quail tissue recombination experiments, we find that sacral crest-derived cells migrate from the pelvic plexus to the cloaca and distal colorectum, where they give rise to enteric neurons and glial cells. This colonization occurs in the presence or absence of vagal crest-derived cells. In contrast, although neural fibers extend from the NoR to the gut, this structure does not appear to contribute crest-derived cells to the intestine. These results suggest that, similar to what has been described in mammalian embryos, colonization of the avian colorectum by sacral neural crest-derived cells occurs by means of the pelvic plexus.
Development of the Pelvic Plexus and ENS in the Quail Hindgut
To examine development of the avian pelvic plexus in situ, the neural crest cell antibody HNK-1 was used to stain mid-sagittal sections of quail embryos at various developmental stages. At E4 (Hamburger–Hamilton stage [HH] 23), a large number of neural crest-derived cells can be seen along the length of the embryo, dorsal to the hindgut and cloaca (Fig. 1A). The NoR has already formed at this stage and can be seen in the dorsal mesentery of the gut (Fig. 1B). Over the next 24–48 hr, neural crest-derived cells begin to aggregate around the cloaca (Fig. 1C,E), forming a dense plexus of cells dorsal, caudal, and ventral to the cloaca, and continuous with the caudal end of the NoR (Fig. 1G).
Longitudinal sections of the hindgut were also examined by immunohistochemistry for the appearance of crest-derived cells. Staining with the early neuronal marker, Hu, which recognizes a neuron-specific RNA binding protein, shows the wavefront of migration in the postumbilical midgut at E5 (HH 26; Fig. 1D), with cells arriving in the ceca by E6 (HH 28; Fig. 1F). Crest-derived cells are present in the hindgut at E7 (HH 30; Fig. 1J), and express Hu (Fig. 1H) and brain-specific fatty acid binding protein (BFABP), a marker of differentiated enteric glial cells and glial precursors (Young et al., 2003). A later marker of neuronal differentiation, neurofilament, is not yet expressed at this stage (Fig. 1I). As previously shown (Burns and Le Douarin, 1998; Conner et al., 2003; Doyle et al., 2004), enteric crest cells colonize the submucosal plexus ahead of the myenteric plexus in the avian hindgut (Fig. 1H).
At E8 (HH 32), enteric crest cell migration is nearly complete, with the hindgut containing two plexuses of crest-derived neurons and glial cells extending to the level of the cloaca (Fig. 2A–D). At this stage, neurofilament expression is strong, staining neurofibers within and between the submucosal and myenteric plexuses (Fig. 2C). Whole-mount immunostaining with Hu shows ganglion cells throughout the hindgut and in the pelvic plexus surrounding the cloaca (Fig. 2E). To better visualize the anatomy of the pelvic plexus at this stage, a sagittal section (Fig. 2F) and multiple cross-sections (Fig. 2G–I) were taken at the level of the tail bud. The pelvic plexus appears as a paired structure and, therefore, is not evident in the mid-sagittal plane (Fig. 2F), but is clearly seen dorsal and ventral to the cloaca (Fig. 2G–I). The paired structure of the pelvic plexus is depicted by whole-mount immunostaining of the distal hindgut and cloaca at E9 (HH 35; Fig. 2J,K). The dorsal view (Fig. 2J) shows neurons extending between the NoR and pelvic plexus, with the junction of these two structures occurring at the level of the cloaca (Teillet, 1978). A parasagittal section through the distal hindgut and cloaca, just lateral to the mid-sagittal plane, demonstrates the dense neuronal network of the pelvic plexus and its intimate association with the NoR (Fig. 2L). At E12 (HH 38) the NoR is seen sending multiple neurofibers into the hindgut (Fig. 2M), while more caudally, the pelvic plexus has formed a well-developed ganglionated plexus around the ureters and cloaca (Fig. 2N).
Contribution of NoR to Hindgut ENS
To understand the contribution of the NoR and the pelvic plexus to the hindgut ENS, tissue recombination experiments were performed. The first step was to establish an aganglionic hindgut that could be used for the recombinations. When the intestine is removed from an E5 quail embryo and cultured for 9 days on the chorioallantoic membrane (CAM), the graft grows significantly in size, particularly the midgut (Fig. 3A). During those 9 days of CAM grafting, vagal crest-derived cells, which are initially located at the postumbilical midgut when the gut is isolated at E5 (Fig. 1D), continue migrating to populate the entire hindgut, as shown with Hu staining (Fig. 3B). Removal of the E5 hindgut alone, without midgut, NoR, or cloaca, followed by 9 days on the CAM, leads to a well-developed and vascularized graft (Fig. 3C). MEP21 antibody labels chicken-derived blood vessels, which arise from the CAM host to populate the quail graft (Fig. 3C). QH1, which labels quail vessels, are also present (not shown). The graft develops normal smooth muscle layers (Fig. 3D), but has no ganglion cells (Fig. 3E,F).
We first examined the contribution of the NoR to the hindgut ENS. The hindgut was isolated from an E5 quail embryo, including NoR, and the graft grown on the CAM for 9 days. The cloaca was removed from the hindgut before grafting to eliminate any potential contribution of crest-derived cells from that structure. After this period of CAM grafting, the hindgut was immunostained with antibodies to Hu and BFABP. As shown in Fig. 4A,B, no neuronal or glial cells were found within the wall of the intestine. Several Hu+ and BFABP+ cells were noted outside of the intestinal wall, in the tissues surrounding the graft (Fig. 4A,B, arrows). Despite the absence of intramural neurons, neurofilament staining was present in the hindgut+NoR grafts (Fig. 4C), likely representing neuronal processes emanating from the NoR.
One possible explanation for the failure of the NoR to contribute cells to the hindgut is that it requires the presence of vagal crest-derived cells to do so. This possibility was tested using chick–quail tissue recombinations. The midgut and hindgut were removed together from quail embryos at E6, a stage at which vagal crest-derived cells are at the level of the ceca (Fig. 1F). The NoR and cloaca were removed from the gut. This quail intestine was then recombined with NoR from an E6 chicken (Fig. 5A). The recombination was grown in organ culture for 1 day to allow the tissues to adhere. The midgut and ceca were then removed and the hindgut, with attached NoR, was grafted onto the CAM of an E7 quail embryo (Fig. 5B). At the time of grafting, immunostaining confirmed that the gut was immunoreactive for the quail-specific antibody QCPN (Fig. 5C), and the NoR stained with the chick-specific marker 8F3 (Fig. 5D), consistent with their species of origin. Hu+ ganglion cells originating from the vagal crest are present in the submucosal plexus of the hindgut at this stage (Fig. 5E), while neurofilament immunoreactivity is present in the NoR, but not in the intestine (Fig. 5F).
After 7 days on the CAM, the graft was removed and processed for immunohistochemistry. Chick-derived cells, which are 8F3+, are found only in the NoR and not in the intestine itself (Fig. 5G). However, many Hu+ cells are present in the graft (Fig. 5H). Because these cells are not 8F3+, they must represent quail-derived vagal crest cells that migrated caudally from the midgut. Strong neurofilament staining is also present (Fig. 5I). We hypothesized that these neuronal processes represent a combination of both quail- and chick-derived nerve fibers. To demonstrate this concept, a chick-specific neurite antibody, CN, was used (Tanaka et al., 1990). Many CN+ fibers are present in the grafts (Fig. 5J), demonstrating that nerve fibers from the chick NoR are capable of penetrating into the quail hindgut in these chick–quail tissue recombinations. The chicken specificity of the CN antibody is shown using wild-type E14 chicken (Fig. 5K) and quail (Fig. 5L) hindgut. The results of the hindgut+NoR grafts and the chick–quail recombined grafts show that the NoR does not contribute ganglion cells to the gut.
Contribution of Pelvic Plexus to Hindgut ENS
To determine whether the pelvic plexus contributes neuronal cells to the hindgut ENS, the hindgut, together with the cloacal region, was dissected from an E5 quail embryo and transplanted onto the CAM for 9 days. Before transplantation, the NoR was removed. Longitudinal and cross-sections of an E5 hindgut stained with HNK-1 and Hu show enteric crest cells limited to the cloacal region (Fig. 6A,B). After 9 days of CAM grafting, the hindgut was isolated and examined for the presence of ganglia. Smooth muscle actin staining shows normal muscle differentiation (Fig. 6C). To confirm the presence of the cloaca in the graft, bursa of Fabricius development was documented using the chicken-specific hematopoietic marker CD45 (Fig. 6D; Nagy et al., 2005). Double immunofluorescence using antibodies to smooth muscle actin and Hu show the presence of Hu+ cells throughout the graft, with a greater density of nerve cells in the distal hindgut (Fig. 6G) and cloaca (Fig. 6H) compared to the proximal (Fig. 6E) and mid- (Fig. 6F) hindgut. The majority of neurons were localized to the myenteric region (Fig. 6E–G, arrows), with a few scattered cells in the submucosal layer (Fig. 6G, arrowheads). These crest-derived cells also differentiated into BFABP+ glial cells (not shown). In addition to neuronal cell bodies, extensive axonal projections can be seen throughout the graft with an antibody to neurofilament (Fig. 6I). Of interest, hindgut+cloaca grafts also contained many pigment cells, primarily in the submucosal region (Fig. 6J).
The results of the hindgut+cloaca grafts suggest that pelvic plexus cells are capable of long-distance caudorostral migration along the hindgut. We sought to confirm this observation using tissue recombinations consisting of quail hindgut and chick cloaca, which would allow us to follow the pelvic plexus-derived chick cells in the presence of vagal crest-derived quail ganglia. This experiment is important because sacral crest cells do not normally contribute to the hindgut until after the arrival of vagal cells (Burns and Le Douarin, 1998). E6 quail hindgut, without NoR, was recombined with E6 chick cloaca and cultured overnight in collagen gel to allow the tissues to adhere. Immunohistochemistry of longitudinal sections from the graft with 8F3 (Fig. 7A) and QCPN (Fig. 7B) show the species origin of the cloaca and hindgut, respectively. Staining with CN, the chick-specific neurite antibody, shows the presence of the pelvic plexus in the cloaca (Fig. 7C), while QN, a quail neurite marker, shows no staining in the cloaca, but the presence of vagal crest-derived enteric ganglia in the proximal hindgut (Fig. 7D). After 7 days on the quail CAM, many 8F3+ cells are present (Fig. 7E) throughout the QCPN+ quail hindgut (Fig. 7F). Chick-derived (8F3+) endothelial cells are present in the villi of the quail hindgut (Fig. 7E,K, arrowhead) and also stain with the blood vessel marker MEP 21 (not shown). A few scattered QCPN+ cells are seen in the chick cloaca (Fig. 7F). Both CN+ (Fig. 7G) and QN+ (Fig. 7H) nerve fibers are present in the hindgut, demonstrating the chimeric nature of the resulting ENS.
To confirm the contribution of the pelvic plexus to hindgut ganglia in the presence of vagal crest cells, double immunofluorescence was performed to stain the tissue recombinants with HNK-1 and 8F3. 8F3+/HNK-1+ cells are seen in the myenteric plexus of the quail graft (Fig. 7I–K). Some of these ganglia are chimeric, containing both chick- and quail-derived enteric crest cells (Fig. 7L–N). The results of the hindgut+cloaca grafts and the chick–quail recombinations demonstrate that the pelvic plexus contributes crest cells to the hindgut ENS.
The goal of this study was to determine the respective contributions of the NoR and the pelvic plexus to the enteric ganglia of the hindgut. The NoR is a sacral neural crest-derived structure that lies within the mesentery of the avian midgut and hindgut. It is a large ganglionated nerve trunk that sends projections, both sympathetic and parasympathetic, to the intestine (Teillet, 1978; Aisa et al., 1998), making it a component of the extrinsic autonomic innervation to the gut. Another major component of autonomic hindgut innervation, particularly to the distal colorectum, is the pelvic plexus, a large paired neuronal network supplying the pelvic viscera with sympathetic and parasympathetic fibers. As shown by our data and others (Browne, 1953; Catala et al., 1995; Aisa et al., 1998), the pelvic plexus is located adjacent to the cloaca and bursa of Fabricius and extends to the caudal end of the NoR, with which it is continuous. The sacral neural crest in avians gives rise to both the NoR and the pelvic plexus, with the rostral sacral crest populating the NoR and the caudal part encircling the cloacal region to form the pelvic plexus (Pomeranz and Gershon, 1990; Catala et al., 1995; Erickson and Goins, 2000).
The sacral neural crest is known to contribute cells to the ENS (Le Douarin and Teillet, 1973; Pomeranz and Gershon, 1990; Serbedzija et al., 1991; Burns and Le Douarin, 1998; Kapur, 2000), but how those crest-derived cells get to the gut is unclear. Using chick–quail chimeras, Burns and Le Douarin (1998) found neurons along nerve fiber tracts emanating from the NoR, leading to the suggestion that sacral crest-derived cells migrate first along the NoR and secondarily enter the gut wall. They noted that the sacral crest contributes to the hindgut ENS in a caudal-to-rostral gradient, with the maximal contribution occurring in the distal colorectum, where 17% of myenteric neurons are sacral crest-derived, whereas only 0.3% of neurons are of sacral origin in the rostral colorectum. This gradient suggests a possible caudorostral migration starting at the cloacal level. Such a migration has been described in mammals, where Kapur (2000) used a transgenic mouse model to show that sacral crest cells first migrate to the pelvic plexus and later move rostrally to contribute to the hindgut ENS.
Our results demonstrate that the avian pelvic plexus contributes to the ENS in a similar manner as described in the mouse. Hindgut grafts only contained neuronal cells if the cloaca and peri-cloacal mesenchyme were included. The majority of these neurons formed enteric ganglia located in the myenteric plexus of the distal colorectum. Fewer cells were seen in the proximal hindgut and in the submucosal region, consistent with the findings of Burns and Le Douarin (1998). Hindgut grafts including the NoR, but not the cloaca, exhibited many nerve fibers projecting from the NoR into the intestinal mesenchyme. However, we found no evidence that the NoR contributes neuronal cell bodies to the gut. CAM cocultures of aneural hindgut with NoR demonstrated no NoR-derived ganglion cells in the transplanted gut, even when vagal crest cells were already present. Extensive neurofilament-immunoreactive fibers were present projecting from the NoR, consistent with its role in extrinsic innervation of the gut (Aisa et al., 1998; Shepherd and Raper, 1999).
The pelvic plexus-derived enteric ganglia that formed in the hindgut+cloaca grafts were small and scattered, forming only a fraction of the normal ENS. This diminished capacity of sacral crest-derived cells to form enteric ganglia has been observed by several investigators. Using gut/neural tube recombinations, Hearn and Newgreen (2000) noted that lumbosacral-derived crest cells formed only few, small ganglia. Similarly, following vagal crest ablation, Burns and Le Douarin (2001) observed that sacral-derived cells formed small and sparse ganglia. Kapur (2000) noted that ganglion cells derived from the mouse pelvic plexus were scattered as individual cells or very small clusters within the terminal hindgut. The developmental potential of vagal and sacral crest-derived enteric neurons is clearly not equivalent. When the vagal crest was transplanted to the sacral crest, for example, vagal-derived cells entered the hindgut in much greater numbers, giving rise to a nearly normal complement of colorectal ganglia (Burns et al., 2002). These intrinsic differences between vagal and sacral-derived cells are poorly understood and deserve further investigation.
We noted the presence of extramural ganglia occurring outside the wall of the grafts in both hindgut+NoR and hindgut+cloaca transplants. Similar extra-intestinal ganglia arising from sacral crest cells have been described previously (Newgreen et al., 1980; Hearn and Newgreen, 2000; Kapur, 2000). Vagal crest cells do not form extramural ganglia on CAM grafts. This sacral-specific phenomenon may reflect the normal potential of these NCCs to form pelvic ganglia outside of the intestinal microenvironment, illustrating another distinction between vagal and sacral NCCs. We also noted the presence of pigment cells in the hindgut+cloaca grafts. These pigment cells were almost exclusively located in the submucosal plexus, an observation previously reported (Newgreen et al., 1980; Allan and Newgreen, 1980). When the hindgut is colonized by vagal crest-derived cells, however, we find no pigment cells in the graft (Fig. 3), suggesting that the presence of a normal complement of enteric ganglia, or simply the presence of vagal crest-derived cells, may suppress the contribution of pigment cells to the gut.
Based on our results, we conclude that the pelvic ganglia, and not the NoR, serve as a staging area for sacral crest-derived cells to enter the gut (Fig. 8B). This observation is consistent with data from Teillet (1978), who showed that aneural hindgut cocultured with the NoR lacked adrenergic neurons unless the cloaca was included, and from Kapur (2000), whose aneural mouse hindguts failed to develop enteric neurons unless the adjacent perirectal mesenchyme, which contains the pelvic ganglia, was present. Sacral crest-derived cells in avian and mammalian embryos thus share a common pathway to the gut by means of the pelvic plexus. Further studies of the hindgut ENS and the contribution of the sacral crest-derived cells may shed light into Hirschsprung's disease and other colonic neurointestinal disorders.
Fertilized White Leghorn chicken and quail (Coturnix coturnix japonica) eggs were obtained from commercial breeders and maintained at 37°C in a humidified incubator. Embryos were staged according to the Hamburger and Hamilton (HH) tables (Hamburger and Hamilton, 1951) or the number of embryonic days (E).
As depicted schematically in Figure 8A, several different CAM grafts were produced. For midgut+hindgut grafts (Fig. 8A1), the postumbilical intestine was dissected from E5 (HH 26) quail embryos and transplanted onto the CAM of E8 (HH 32) chick embryos as described (Nagy and Goldstein, 2006a). NoR and cloaca were left attached. Before grafting, a small portion of the CAM was gently traumatized by laying a strip of sterile lens paper onto the surface of the epithelium and then removing it immediately. The dissected intestinal graft was placed over the junction of blood vessels on the traumatized area of the CAM and incubated for 7–9 days. The graft, together with the surrounding CAM, was excised, fixed in 4% buffered formaldehyde and embedded in gelatin.
The hindgut grafts (Fig. 8A2-4) were prepared by isolating the postcecal hindgut from E5 (HH 26) quails. The NoR was either left attached (Fig. 8A3) or removed (Fig. 8A2,4) using tungsten needles. The cloaca and pericloacal mesenchyme, which contains the pelvic plexus, were also either left attached (Fig. 8A4) or removed (Fig. 8A2,3) before CAM grafting.
Organ Culture and Tissue Recombinations
For chick–quail tissue recombinations, E6 (HH 28) quail midgut/hindgut was removed, extending from the umbilicus to the distal colorectum. The NoR and cloaca were removed. This quail intestine was recombined with NoR isolated from E6 (HH 28) chick hindgut (Fig. 8A5) or with E6 chick cloaca (Fig. 8A6). The proximal–distal orientation of the NoR was maintained in the recombination, as was the orientation of the cloaca. To allow the tissues to adhere before CAM grafting, the recombination was embedded in a three-dimensional collagen gel matrix prepared by adding 1 N NaOH and 1 mg/ml type I rat tail collagen (BD Biosciences) to DMEM medium (Gibco) supplemented with 1% penicillin–streptomycin as previously described (Goldstein et al., 2005). After 24 hr, the explants were removed from the gel and either processed for immunocytochemistry or transplanted to E7 (HH 32) quail CAM for 7 days.
Samples were fixed in 4% formaldehyde in phosphate buffered saline (PBS) for 1 hr, rinsed with PBS, and infiltrated with 15% sucrose/PBS overnight at 4°C. The medium was changed to 7.5% gelatin containing 15% sucrose at 37°C for 1–2 hr, and the tissues rapidly frozen at −60°C in isopentane (Sigma). Frozen sections were cut at 10 μm, collected on poly-L-lysine–coated slides (Sigma), and stained by immunocytochemistry as previously described (Nagy and Goldstein, 2006b). Primary antibodies used are shown in Table 1.
CD45 (Cedi Diagnostics) MEP21 (McNagny et al., 1997)
Neural crest cells
Hu (Molecular Probes); Neurofilament (4H6; DSHB)
brain fatty acid binding protein (Kurtz et al., 1994)
alpha-smooth muscle actin (1A4; NeoMarkers)
For double immunofluorescent staining, sections were rehydrated and incubated with primary antibodies (Hu, smooth muscle actin, and neurofilament). Fluorescent secondary antibodies included Alexa Fluor 594 goat anti-mouse IgG, Alexa Fluor 594 goat anti-mouse IgM, Alexa Fluor 488 goat anti-mouse IgG (Molecular Probes), and fluorescein isothiocyanate–labeled goat anti-mouse IgG2a (Southern Biotech). Sections were examined under a Nikon Microphot FXA microscope, and digital images were captured with a Spot camera and software version 3.3.1 (Diagnostic Instruments). Images were compiled using Adobe Photoshop.
For whole-mount immunohistochemistry, methanol-dehydrated E5–E14 guts were rehydrated in PBS, bleached in 15% H2O2 in Dent's fix (20% dimethyl sulfoxide [DMSO] in methanol) for 3 hr at room temperature, washed in methanol, and fixed overnight in Dent's fix. Guts were washed in five 1-hr washes of PBS, then labeled with primary antibody overnight at 4°C in a blocking solution of 5% goat serum and 20% DMSO in PBS. PBS washes were performed, followed by overnight incubation at 4°C with secondary antibody in the same blocking solution. PBS washes were followed by the addition of ABComplex/HRP (Dako). Antibody detection was performed with 0.05 mg/ml diaminobenzidine and 0.02% H2O2 in 0.1 M Tris, pH 7.0.
We thank Drucilla Roberts for her guidance, enthusiasm, and valuable insight. Monoclonal antibodies QCPN, 8F3, 4H6, QH1 were obtained from the Developmental Studies Hybridoma Bank, developed under the auspices of the NICHD and maintained by The University of Iowa, Department of Biological Sciences (Iowa City, IA 52242). The BFABP antibody was a kind gift from Dr. Carmen Birchmeier, MEP21 was generously provided by Dr. Kelly McNagny, and CN and QN were kindly provided by Dr. Hideaki Tanaka. A.M.G was funded by the NIH.