Hedgehog signaling is required for formation of the notochord sheath and patterning of nuclei pulposi within the intervertebral discs

Abstract

The vertebrae notochord is a transient rod-like structure that produces secreted factors that are responsible for patterning surrounding tissues. During later mouse embryogenesis, the notochord gives rise to the middle part of the intervertebral disc, called the nucleus pulposus. Currently, very little is known about the molecular mechanisms responsible for forming the intervertebral discs. Here we demonstrate that hedgehog signaling is required for formation of the intervertebral discs. Removal of hedgehog signaling in the notochord and nearby floorplate resulted in the formation of an aberrant notochord sheath that normally surrounds this structure. In the absence of the notochord sheath, small nuclei pulposi were formed, with most notochord cells dispersed throughout the vertebral bodies during embryogenesis. Our data suggest that the formation of the notochord sheath requires hedgehog signaling and that the sheath is essential for maintaining the rod-like structure of the notochord during early embryonic development. As notochord cells form nuclei pulposi, we propose that the notochord sheath functions as a “wrapper” around the notochord to constrain these cells along the vertebral column.

Low back pain will affect most people over the age of 65 in industrialized countries. In the United States, treatment of low back pain is estimated to cost 50 to 100 billion dollars per year (reviewed in ref. 1). For the majority of people, bed-rest will relieve most symptoms of back pain, but in a small population of patients the condition will become chronic. Most back pain is thought to originate from degeneration of the intervertebral discs or through physical damage to the disc. This damage leads either to herniation of the middle part of the disc, called the nucleus pulposus, or tears, bulging, and rupture of the annulus fibrosus, which surrounds the nucleus pulposus (reviewed in ref. 2).

The intervertebral discs connect two adjacent vertebrae and provide structural stability and flexibility to the spinal column. The nuclei pulposi, which originate from the embryonic notochord (3), are composed primarily of proteoglycan, water, and collagen type II and are located in the middle of each intervertebral disc (reviewed in ref. 4). As the discs age or are damaged, the nucleus pulposus is dramatically altered. Proteoglycan and water content decreases in the nucleus pulposus and collagen type II is replaced by type I collagen so that nuclei pulposi become fibrous and contain less water.

During midembryogenesis in both mice and humans, the notochord—a transient rod-like structure that is located along the midline of embryos—becomes segmented and forms the intervertebral discs (3, 5). In embryonic day (E) 12.5 to E15.5 mouse embryos, notochord cells are removed from regions of the embryo containing the vertebral bodies and are relocated to the intervertebral mesenchyme through an unknown mechanism. Interestingly, mouse mutants formed aberrant cartilage around the vertebral column but a normal notochord and notochord sheath failed to remove notochord cells from the vertebral bodies, resulting in malformed intervertebral discs (6–9). These data support the hypothesis that a mechanical force driven by the forming vertebral bodies is responsible for pushing notochord cells into the intervertebral discs (6, 10, 11, reviewed in ref. 12).

In numerous tissues, hedgehog signaling has been implicated in regulating pattern formation and cell proliferation and cell survival (13–15). Previous studies have shown that Sonic Hedgehog (Shh) is expressed in nuclei pulposi in both prenatal and postnatal intervertebral discs (16, 17). In addition, Indian Hedgehog (Ihh) is expressed in condensing chondrocytes of the embryonic vertebral bodies and in the vertebral endplate during later development (17). Both Shh and Ihh produce secreted proteins that bind the transmembrane protein PATCHED1 (PTCH1), resulting in activation of the hedgehog signaling pathway (reviewed in ref. 18). In mice containing a null allele of Shh, the notochord formed but was quickly lost (19). Because the notochord was quickly lost in Shh-null animals, the role of hedgehog signaling may play in transforming the notochord into nuclei pulposi remains unknown. Ihh is not expressed in nuclei pulposi (17); however, conditional removal of Ihh in chondrocytes of postnatal mice has been shown to result in loss of the annulus fibrosus and enlargement of the nucleus pulposus. These data suggest that Ihh may be required within growth plates for intervertebral disc homeostasis (20).

To directly examine the role hedgehog signaling plays in the formation of nuclei pulposi, we conditionally removed Smoothened (Smo), which is required for all hedgehog signaling, in all Shh-expressing cells. In the vertebral column, Smo was removed from the mouse notochord and floorplate. Removal of Smo from the vertebral column, coupled with detailed fate-mapping and molecular analysis in this mutant background, allowed us to determine the role the hedgehog signaling pathway plays in transforming the notochord into nuclei pulposi.

Results

Removal of Hedgehog Signaling from the Mouse Notochord.

Mouse embryos in which hedgehog signaling was removed from all cells die before formation of the intervertebral discs (21). Previously, we had shown that the notochord forms the entire nucleus pulposus of each disc in the mouse vertebral column (3). To determine the role of hedgehog signaling during formation of nuclei pulposi, hedgehog signaling was removed from all Shh-expressing cells, including notochord and floorplate in the vertebral column, using a floxed mouse allele of Smo and the Shhgfpcre allele (Fig. 1 A and B and Figs. S1 and S2) (22, 23).

Fig. 1.

Removal of SMO in Shh-expressing cells results in abnormal development of the intervertebral discs and vertebrae. (A and B) Analysis of Ptch1:lacZ expression revealed that hedgehog signaling was absent in mutant notochords. Section of Ptch1:lacZ expression of E9.5 control (Smo^f/f; Ptch1:lacZ) and mutant (Smo^f/f;Shhgfpcre;Ptch1:lacZ) embryos. Note that Ptch1:lacZ expression was absent in the mutant caudal notochord (arrow). In the neural tube a decrease in Ptch1:lacZ was observed compared with controls. Lower expression may be because of the inability of floorplate cells to respond to SHH secreted from the notochord because the Shhgfpcre allele removed SMO from the floorplate in addition to the notochord. (C and D) Bright-field images of E12.5 wild-type and mutant embryos in which Smo has been removed from Shh-expressing cells. An abnormally truncated and thinner tail (asterisk in D) was observed in the E12.5 mutant. (E–H) Lysotracker assay of control and mutant E11.5 embryos. (Magnification: E and F, ×10 and G and H, ×32.) G and H are higher magnification of the boxed regions shown in E and F. An increase in cell death occurred caudal to the lumbar vertebra (arrowhead in F). (I–N) Histological analysis of a saggital section of the intervertebral vertebral column using Picro-sirius red and Alcian blue staining. Close up view of nucleus pulposus (K and L) and annulus fibrosus (M and N). Mutant tissue contained a smaller nucleus pulposus than controls and concentric lamellae were absent in the annulus fibrosus. af, annulus fibrosus; np, nucleus pulposus; nt, neural tube; and vb, vertebral body. (Scale bars, 20 μm in A and B; 100 μm in I and J; and 20 μm in K–N.)

Hedgehog Signaling Is Required for Formation of Intervertebral Discs and Normal Cell Proliferation in the Notochord.

Intervertebral discs are located between each vertebra in wild-type mice. To determine the role hedgehog signaling played in formation of the intervertebral discs, newborn control and mutant vertebral columns were analyzed. Sections through mutant vertebral columns revealed that nuclei pulposi were smaller compared with control littermates (Fig. 1 I–L). Mutant intervertebral discs contained an annulus fibrosus but this tissue appeared to have lost concentric lamellae within the annulus fibrosus, possibly as a result of the improper formation of nuclei pulposi in the center of the discs (Fig. 1 M and N).

In a number of organs, hedgehog signaling is required for cell proliferation and cell survival (13–15). In these tissues, removal of hedgehog signaling often resulted in a decrease in cell proliferation and an increase in cell death. To determine whether removal of hedgehog signaling from the notochord resulted in a defect in cell proliferation, BrdU was administrated to pregnant dams 3 h before harvest. BrdU assays were performed on the rostral regions of E11.5 wild-type and mutant embryos. Using this assay, notochord cells in E11.5 embryos were found to proliferate at a lower rate than controls (Fig. 2). In addition, an increase in proliferation in the surrounding perichordal mesenchyme was observed in mutant embryos. These data suggest that the observed smaller nuclei pulposi located in the rostral vertebral column in mutant mice may partially result from a decrease in the rate of cell proliferation of notochord cells upon the removal of hedgehog signaling.

Fig. 2.

Removal of hedgehog signaling resulted in a decrease in cell proliferation in rostral mutant notochords. Representative transverse sections of the rostral vertebral column of E11.5 embryos are shown. (A and B) BrdU staining of control (A) and mutant (B) sections. (Scale bars, 200 μm.) (C and D) A merged picture of BrdU, DAPI, and laminin (green). Laminin staining marked the inner layer of the notochord sheath and outlined the location of the notochord. At E11.5, the number of anti-BrdU⁺ cells in mutant notochords was decreased (D). (E) Quantification of the number of anti-BrdU⁺ cells/total cells in the notochord demonstrated that the number of proliferating cells in mutant notochords was significantly decreased. Quantification of the number of anti-BrdU⁺ cells in surrounding perchordal mesenchyme (dotted circle in C and D) indicated that the proliferation rate in surrounding mesenchyme was slightly increased in mutant embryos. Data are represented as means and the error bars represent the SD. *P < 0.05.

Removal of Hedgehog Signaling Resulted in Aberrant Migration of Notochord Cells During Intervertebral Disc Formation.

The mouse nucleus pulposus and annulus fibrosus are formed in highly condensed regions of intervertebral mesenchyme along the ventral midline of the embryo beginning at E12.5 (10). Over the next 3 days, the notochord forms the nuclei pulposi of the intervertebral discs. Notochord cells are normally excluded from regions of the vertebral column where vertebrae form. To determine if the aberrant nuclei pulposi found in mutant animals were the result of defects in the transition of notochord cells into nuclei pulposi, the notochord in mutant animals was fate-mapped using the ROSA26 reporter allele (24). In these animals, all cells arising from the notochord were marked, allowing for a detailed analysis of the fate of notochord cells throughout development.

At E12.5 a slightly thinner notochord was observed in mutant animals compared with controls (Fig. 3B), consistent with the observation that there is a decrease in cell proliferation in this tissue (Fig. 2). In control E13.5 embryos, the notochord formed bulges between each vertebra where the future discs would form. In mutant embryos, this did not occur. Notochord cells continued to reside as a rod along the midline of the embryo. A small number of notochord cells were also found within the vertebral bodies (Fig. 3D). By E14.5 nuclei pulposi had formed from the notochord in control embryos, with very few notochord cells still residing in vertebral bodies. However, in mutants few notochord cells were found to reside in the forming disc with most cells scattered throughout the vertebral column (Fig. 3F). In postnatal control animals, the nucleus pulposus was located inside the annulus fibrosus throughout the vertebral column. In contrast, mutant animals contained small nuclei pulposi with the majority of notochord cells dispersed throughout the vertebral column (Fig. 3 H, J, and L, and Fig. S3).

Fig. 3.

Aberrant migration of notochord cells throughout the vertebral column upon removal of hedgehog signaling in the notochord. (A–L) Cells were fate-mapped using the Cre-inducible R26R allele. (A and B) Notochord (blue cells) formed a rod-like structure in both wild-type and mutant animals until E12.5. In wild-type E13.5 embryos (C) the notochord started to form a bulge between the vertebrae in regions where the intervertebral discs were forming. In mutants (D) the notochord remained as a rod-like structure and a few cells were found to reside outside the notochord (arrows). By E14.5, notochord cells had formed nuclei pulposi in wild-type animals (E) but retained a rod-like structure in the mutant (F). A number of mutant notochord cells (arrows) resided outside the notochord. In P0 wild-type animals (G and I), notochord cells had formed the nucleus pulposus of each intervertebral disc. In contrast, mutant notochord cells (H and J) were randomly found throughout the intervertebral mesenchyme and vertebrae. (K and L) Histological analysis of thoracic vertebrae from newborn mice demonstrated severe defects in nucleus pulposus structure and an increase in notochord cells residing in surrounding tissues. im, intervertebral mesenchyme; and vb, vertebral body. (Scale bars, 50 μm in A–F and 100 μm in K and L.)

Hedgehog Signaling Is Required for Notochord Sheath Formation.

The notochord sheath is composed of extracellular matrix proteins that surround the notochord in E10.0 embryos (25). To determine if hedgehog signaling was required for notochord sheath formation, sections of E11.5 vertebral columns were stained with Alcian blue to visualize the extracellular matrix composition surrounding the notochord. In control embryos the notochord sheath formed around the notochord in a rostral to caudal progression (Fig. 4 A and C). Sheath formation occurred before cartilage condensation within the vertebral bodies. In mutant embryos a thin notochord sheath was observed in the rostral region of E11.5 embryos (Fig. 4B); however, no Alcian blue staining was observed in the caudal region of mutant embryos (Fig. 4D). In addition, the caudal notochord of mutants was abnormally flattened (Fig. 4 D, H, and L).

Fig. 4.

Hedgehog signaling is required for notochord sheath formation. Asterisks (*) indicate the location of the notochord sheath in all panels. (A–D) Histological analysis of transverse sections of rostral and caudal notochords from E11.5 embryos. Control (A and C) notochords were surrounded by the notochord sheath, which was visualized with Alcian blue stain. In mutants, a thin layer of the notochord sheath was observed in the rostral notochord (B). In the mutant caudal notochord, the notochord sheath was absent (D). In both controls (E and G) and mutants (F and H) immnohistochemistry revealed that laminin surrounded the notochord. (I–L) Transmission electron micrograph of the notochord sheath. Insets in I to L are lower magnification of the notochord. (Magnification: I–L, 30,000×.) The notochord sheath of controls contained basal lamina and collagen fibrils (I and K). In the mutant, basal lamina and a thinner layer of collagen fibrils formed in the rostral notochord (J) but the collagen fibrils appeared to be absent in the caudal notochord (L). n, notochord; and ns, notochord sheath. (Scale bars, 10 μm in A–H and 0.5 μm in I–L.)

To determine if the ultrastructure of the notochord sheath was affected by the removal of hedgehog signaling, transmission electron micrography was performed on the transverse section of E11.5 embryos. In wild-type embryos, the notochord sheath consisted of a basal lamina layer and loosely organized collagen fibrils (Fig. 4 I and K). In the rostral region of mutants, the notochord sheath consisted of basal lamina and a thin layer of collagen fibrils (Fig. 4J). In the caudal region of mutant embryos no electron-dense material was observed (Fig. 4L). Laminin, an inner component of the notochord sheath, was found to surround the caudal notochord in both wild-type and mutant embryos, indicating that at least some components of the sheath are still present in the absence of hedgehog signaling within the caudal notochord (Fig. 4 E–H).

Removal of Hedgehog Signaling After Formation of the Notochord Sheath Does Not Affect Nuclei Pulposi Patterning or Growth.

Removal of hedgehog signaling before sheath formation resulted in a deformed sheath and aberrant formation of nuclei pulposi. These data suggested that proper formation of a notochord sheath was essential for normal patterning of nuclei pulposi along the vertebral column. However, it was possible that nuclei pulposi were not patterned correctly because of loss of hedgehog signaling and not because of the absence of a notochord sheath. To test this hypothesis, hedgehog signaling was removed after sheath formation using the tamoxifen-inducible Cre allele ShhcreERT² and the floxed Shh allele (22, 26). Tamoxifen was administered at E8.5 (before sheath formation), E9.5 (during sheath formation), E10.5 or E11.5 (after the sheath formation; in a normal embryo, the sheath is first observed at E10.0 surrounding the notochord (25). To determine if hedgehog signaling was efficiently removed in Shh-expressing cells after tamoxifen exposure, we analyzed expression of Shh and Ptch1 in E9.5 embryos that had been exposed to tamoxifen at E8.5 (27). Both Shh and Ptch1 expression were absent in the mutant notochord of treated embryos (Fig. 5 A–D). Examination of the vertebral column of E11.5 control and Shh^f/ShhcreERT² embryos, in which hedgehog signaling had been removed after sheath formation occurred, demonstrated that perdurance of the notochord sheath did not require hedgehog signaling (Fig. S4). To determine if removal of hedgehog signaling from the notochord after sheath formation affected nuclei pulposi patterning, control and Shh^f/ShhcreERT² notochords exposed to tamoxifen at E11.5 were fate-mapped using the R26R reporter allele (24). In control and tamoxifen-treated E11.5 embryos harvested at E18.5, no difference in nuclei pulposi formation was observed in rostral to sacral vertebrae (Fig. 5 E and I and Fig. S5), indicating that hedgehog signaling is not required for formation of nuclei pulposi after the sheath has formed.

Fig. 5.

Shh is required for patterning the intervertebral discs. (A–D) Section RNA in situ hybridization of Shh (using a probe against the floxed exon 2) and Ptc1 of E9.5 control (ShhcreERT²) and mutant (Shh^f/ShhcreERT²) embryos. Shh and Ptc1 transcripts in E9.5 mutant embryos (B and D) were not detected 24 h after tamoxifen (TM) injection. (E–I) Fate-mapping of cells that have expressed Shh during intervertebral disc formation. Control (ShhcreERT²;R26R) and mutant (Shh^f/ShhcreERT²;R26R) embryos were harvested at E18.5 after a single TM injection at either E8.5, E9.5, E10.5, or E11.5. All images are ventral views of the vertebral column. (E′–I′) Ventral view of control (ShhcreERT²) and mutant (Shh^f/ShhcreERT²) vertebral columns. Removal of Shh from E8.5 to E10.5 in Shh-expressing cells resulted in severe defects in formation of vertebral columns and lack of formation of ossification centers. (I′) Removal of Shh from E11.5 embryos did not result in any phenotypic abnormalities in the thoracic vertebral region. (Scale bars, 50 μm.)

Proper Formation of Vertebrae Is Required for the Transition of the Notochord into Nuclei Pulposi.

Removal of Shh from Shh-expressing cells at E8.5 or E9.5 resulted in the continued presence of the notochord throughout the vertebral column and an absence of nuclei pulposi throughout embryonic development (Fig. 5 F and G). Unlike when Smo was removed from Shh-expressing cells, notochord cells were not observed scattered throughout the vertebral column, even when the notochord sheath was abnormal. In addition, removal of Shh resulted in defects in formation of the vertebrae. An increasing severity in defective nuclei pulposi and vertebrae formation correlated with earlier removal of Shh. Removal of Shh at E10.5 resulted in formation of vertebrae but they lacked condensations. In this experiment, nuclei pulposi began to form but notochord cells were still found to reside with the vertebral bodies (Fig. 5H). These data support the proposal that vertebrae may be responsible for forcing notochord cells into the forming intervertebral bodies (Fig. 6).

Fig. 6.

Proposed role for the notochord sheath in forming nuclei pulposi of the intervertebral discs. The notochord sheath (red line) begins to form around the notochord (blue line) at E10.0 (early). By E14.5 (late) most notochord cells reside within the intervertebral discs. It has been proposed that swelling pressure (denoted by arrows) exerted by the vertebral bodies serves to push notochord cells into the space between each vertebrae. Loss of a functional sheath or “wrapper” around the notochord (denoted by a thin red line) results in notochord cells being scattered throughout the vertebral column and formation of small and misshapen nuclei pulposi. Loss of a functional sheath in the absence of swelling pressure results in the continued presence of the rod-like notochord throughout embryonic development. It is important to note that the proposed model does not rule out the possibility that a currently unknown molecular or chemical pathway is responsible for moving notochord cells into the forming discs.

Discussion

Role of Hedgehog Signaling Within the Mouse Notochord.

During normal mouse development the notochord sheath surrounds the entire notochord, beginning at E10.0 (26). As the notochord begins to form visible nuclei pulposi at E12.5, the sheath remains around notochord cells. Our data directly addresses the role hedgehog signaling plays in formation of the notochord sheath. In Shh-null embryos, the notochord forms but then quickly disappears before sheath formation, suggesting that hedgehog signaling is essential for maintaining a functional notochord (19). Because Shh-null embryos are defective in hedgehog signaling throughout the entire embryo, it was not possible to determine if loss of the notochord in these mutant animals was an indirect consequence of loss of hedgehog signaling in other tissues. In our experiments, hedgehog signaling was removed from the notochord but was still present in tissues surrounding this structure. In these embryos, notochord cells persisted throughout embryonic and postnatal development. These data indicate that hedgehog signaling is not required for maintenance of this structure but instead is essential for normal formation of the notochord sheath that surrounds the embryonic notochord. It is important to note that the cre allele used in these experiments, Shhcre, removes hedgehog signaling from the floorplate in addition to the notochord. Although we currently have no evidence to support a role for hedgehog signaling within the floorplate in forming the notochord sheath, it is possible that signaling molecules within the floorplate that are downstream of the hedgehog signaling pathway may be important for forming at least some aspects of the notochord sheath.

Upon removal of all hedgehog activity, the sheath was disrupted but a ring of basal lamina was still found surrounding the notochord, indicating that hedgehog signaling is not responsible for producing all components of the sheath. Laminin protein surrounding the hedgehog-null notochord could be produced from nonnotochord cells, as suggested by experiments in zebrafish (29). A second possibility is that laminin is produced directly by notochord cells but does not require hedgehog signaling.

Role of the Notochord Sheath During Intervertebral Disc Formation.

Although it is clear that a notochord sheath forms around the notochord in a number of different species, including zebrafish, chicken, mice, and humans, the function this structure plays during development has remained elusive (25, 29–31). During the transition of the notochord into nuclei pulposi, notochord cells have been proposed to be “squeezed” along the midline of the embryo by the condensing vertebra into the forming discs (6, 10, 11, reviewed in ref. 12). In embryos in which hedgehog signaling was removed from the notochord but contained normal vertebral bodies, notochord cells were observed to be scattered throughout the vertebral column. Mutant embryos that had a defective vertebral column, irrespective of whether they had a normal notochord sheath or contained a rod-like notochord, suggesting that vertebrae are needed to form normal discs.

We propose that a possible function of the notochord sheath may be to form a “wrapper” around the notochord (see model presented in Fig. 6). Before the notochord-forming nuclei pulposi, our model suggests that the sheath is required to maintain the rod-like structure of the notochord. In our experiments, loss of a functional sheath caused the notochord to flatten. During later wild-type development, we propose that the sheath is flexible enough so that when the forming vertebrae exert swelling pressure the sheath expands but still constrains notochord cells to the dorsal midline of the embryo.

In regions of the embryo where the discs are forming, the notochord bulges outward and forms the nucleus pulposus of each intervertebral disc. We propose that in the absence of a functional sheath, notochord cells are not constrained and become scattered throughout the vertebral column. Consistent with the proposed role for the notochord sheath in constraining notochord cells within the midline of the vertebrate embryo, an increase in the aberrant migration of notochord cells correlated with the observed increasing caudal severity of defects in sheath formation.

It is possible that abnormal nuclei pulposi formation observed upon removal of hedgehog signaling results from some other, nonsheath role for hedgehog signaling in this tissue. We cannot rule out the possibility that the hedgehog signaling pathway is responsible for activation of unknown pathways that are required for proper migration of notochord cells into the forming nuclei pulposi, independent of the presence of a notochord sheath. Mechanical removal of the notochord sheath from around the notochord in normal embryos could directly address this question; however, this experiment is technically challenging because of the inaccessibility of the notochord during vertebrate embryonic development.

Instead, we have taken a genetic approach to address this issue by removing hedgehog signaling after the notochord sheath has formed. In these embryos, the sheath was maintained and normal nuclei pulposi formation was observed. These data suggest that hedgehog signaling is required to specify formation of the notochord sheath but is not needed to maintain this structure during later embryogenesis. In wild-type mice, the hedgehog signaling pathway remains present in nuclei pulposi throughout early postnatal life (16). The role this signaling pathway plays in the postnatal intervertebral discs is unknown.

Materials and Methods

Mice.

Animals were handled in accordance with the University of Florida Institutional Animal Care and Use Committee. Mice containing the conditional floxed allele of Smo (Smo^f/f), Shhgfpcre, ShhcreERT², and Shh^f/^f have been described previously (22, 23, 26). For Shh^f/ShcreERT² embryos, tamoxifen (Sigma) was gavaged at a concentration of 3 mg/40 g body weight in a pregnant female. Detailed information is available in SI Materials and Methods.

RNA in Situ Hybridization, β-Galactosidase Staining, and Skeleton Preparation.

Whole-mount RNA in situ hybridization and Xgal staining were performed as described previously (22, 32, 33). Skeleton preparations were performed as previously described (34). At least three animals for each genotype were examined in all experiments.

Histology and Immunohistochemistry.

For histological analysis, embryos were fixed in 4% paraformaldehyde at 4 °C overnight and then E16.5 and older embryos were decalcified overnight using Cal-Ex (Fisher). Histology and immunohistochemistry were performed as described in SI Materials and Methods.

Cell Proliferation and Death Assay.

To detect cell proliferation in E11.5 notochords, pregnant dams were injected with BrdU (50 μg/g bodyweight) for 3 h before harvest. Cell proliferation and death assay were performed as described in SI Materials and Methods.

Electron Microscopy.

Tissue samples were fixed in 4% paraformaldehyde + 2% glutaraldehyde in 0.1 M sodium cacodylate, pH 7.24 and processed for imaging. A detailed procedure is described in SI Materials and Methods.