Foxa1 and Foxa2 Are Required for Formation of the Intervertebral Discs

The intervertebral disc (IVD) is composed of 3 main structures, the collagenous annulus fibrosus (AF), which surrounds the gel-like nucleus pulposus (NP), and hyaline cartilage endplates, which are attached to the vertebral bodies. An IVD is located between each vertebral body. Degeneration of the IVD is thought to be a major cause of back pain, a potentially chronic condition for which there exist few effective treatments. The NP forms from the embryonic notochord. Foxa1 and Foxa2, transcription factors in the forkhead box family, are expressed early during notochord development. However, embryonic lethality and the absence of the notochord in Foxa2 null mice have precluded the study of potential roles these genes may play during IVD formation. Using a conditional Foxa2 allele in conjunction with a tamoxifen-inducible Cre allele (ShhcreERT2), we removed Foxa2 from the notochord of E7.5 mice null for Foxa1. Foxa1−/−;Foxa2c/c;ShhcreERT2 double mutant animals had a severely deformed nucleus pulposus, an increase in cell death in the tail, decreased hedgehog signaling, defects in the notochord sheath, and aberrant dorsal-ventral patterning of the neural tube. Embryos lacking only Foxa1 or Foxa2 from the notochord were indistinguishable from control animals, demonstrating a functional redundancy for these genes in IVD formation. In addition, we provide in vivo genetic evidence that Foxa genes are required for activation of Shh in the notochord.


Introduction
Three main structures compose the intervertebral disc (IVD); the annulus fibrosus (AF), the nucleus pulposus (NP), and endplates (EP). A disc is sandwiched between each of the vertebrae along the axial skeleton of vertebrates and allows for flexibility while providing a major support to the body. The NP is a watery gellike substance containing proteoglycans [1,2]. The NP is formed from the embryonic notochord [3] and it is encircled by the AF, a structure comprised of many concentric rings of collagen fibers [1]. The AF is derived from the sclerotome [4], a compartment of the somite. Hyaline cartilage endplates bound the disc superiorly and inferiorly and attach each disc to the vertebrae [1,2]. It is through the endplates that the disc, which is avascular and is not innervated by nerves, obtains its nutrients via diffusion [1,2,5].
During the aging process numerous changes occur to the disc and these are collectively referred to as ''disc degeneration'' [6][7][8]. These changes include the transition of the NP from a watery gellike substance to a more fibrous material, the accumulation of small tears in the AF [2,7,9] and ossification and fracturing of the endplates [10,11]. All of these changes alter the nutrition and structural integrity of the disc, and thus the axial column. In some patients, the NP can herniate out of the AF while in other patients the disc is observed to become compressed resulting in a decrease in the distance between adjacent vertebrae. These changes can result in abnormal pressure being placed on the numerous nerves present in adjacent tissue resulting in severe pain. Back pain, while common, has few effective treatments and is an expensive ailment in terms of medical costs and lost productivity [12,13]. Current treatments address the pain and not the underlying causes, which is frequently disc degeneration. Surgery, and complete or partial disc replacement are options for patients suffering chronic pain but often have have high rates of failure and unacceptable side effects for patients [2,9]. It is hoped that by studying the proper formation of the disc and factors that contribute to its degeneration, we can develop novel, effective strategies for disc repair/regeneration.
The notochord is a rod-like structure found in all chordate embryos, and is surrounded by a sheath (the ''notochordal'' sheath), which is composed of extracellular matrix proteins [14]. During embryonic development the notochord secretes sonic hedgehog (SHH) protein (reviewed in [15]) and serves as a vital signaling center in the chordate embryo [16]. In some species, the notochord also serves as the main structural support for the embryo on which the rest of the body plan is built [16]. The notochord itself forms from the mammalian organizer, or node. This is located at the anterior end of the primitive streak in a mouse embryo.
Cells of the trunk notochord are formed from cells that migrate through the node during gastrulation [17]. The node is capable of inducing a second axis when transplanted into a host embryo [18]. The tail notochord was thought to form from the regression of the node posteriorly, though more recent fatemaps of node cells have suggested that instead, these cells migrate through the node posteriorly to form the tail notochord [19] [20]. The mouse notochord has 3 separate origins; anterior cells that are not nodederived, trunk notochord that forms from cells of the node that undergo convergent extension, and tail notochord which has been proposed to form from cells that have migrated out of the node [19]. The tail bud forms at the end of gastrulation and eventually undergoes secondary neurulation to form the notochord and neural tube in the posterior embryo [21,22].
Foxa1 null mice contained a normal notochord and intervertebral discs but died within days of birth as a result of hypoglycemia and dehydration [26]. Foxa2 null mice died during early embryogenesis and lacked a node and notochord in addition to containing abnormalities in all germ layers [34,35]. Due to the absence of a node (and thus notochord), potential roles for Foxa2 in NP formation have not been identified. In this study, we investigated disc formation using a Foxa2 conditional floxed mouse [36] in combination with a Foxa1 null allele [26]. Both alleles have previously been used to characterize the development of the liver, lung, brain, and other tissues [27,29,37]. To remove Foxa2 from the notochord, the tamoxifen-inducible ShhcreER T2 allele was used [38]. Removal of both Foxa1 and Foxa2 from the notochord resulted in severe defects in the formation of the axial skeleton and aberrant dorsal-ventral patterning of the neural tube. The individual removal of either Foxa1 or Foxa2 resulted in normal disc formation indicating that the Foxa genes are functionally redundant in the mouse notochord. Defects in the notochord of double mutants were observed by E9.5, well before the disc starts to form at E12.5. A molecular analysis revealed that Foxa gene expression in the notochord was required for expression of Shh.

Materials and Methods
Animal Care, Generation of Embryos and Genotyping All mice were raised in accordance with University of Florida Institutional Animal Care and Use Committee Guidelines (UF IACUC) under specific pathogen-free (SPF) conditions. All procedures were approved by the UF IACUC. Animals containing the Foxa1 null allele (Foxa1 2/2 ) and Foxa2 conditional allele (Foxa2 c/c ) were originally generated by the Kaestner laboratory [26,36]. These animals were mated to mice containing the tamoxifen-inducible ShhcreER T2 [38] and R26R [39] alleles to create mice in which Foxa1 and Foxa2 could be deleted from the notochord (Foxa1 2/2 ;Foxa2 c/c ;ShhCreER T2 ;R26R). In this report, these animals are referred to as ''double mutants''. Inclusion of the R26R allele allowed for fate mapping of mutant cells. R26R was not present in every animal analyzed. Timed matings were obtained by setting up a cross in the late afternoon; the following morning the female was monitored for the presence of a semen plug, and was by convention determined to have mated at 12:00 am the night before (Day 0.5 when plug is found). Animals homozygous for the Foxa2 conditional allele were phenotypically normal. Weanlings were ear-tagged and genotyped using the tip of the tail. To genotype embryos, extra-embryonic membranes were removed during dissection and lysed in 25 mM NaOH at 95uC for at least 20 minutes. Samples were neutralized with an equal volume of 40 mM TrisHCl and 1-2 mL DNA used in PCR reactions. Genotyping was performed as previously described [26,36,38,39]. At least three mutant animals were examined for each experiment. Control animals were either Foxa1 +/+ ;Foxa2 c/c or Foxa1 +/2 ;Foxa2 c/c . Mice were maintained on a mixed genetic background.

RNA in situ hybridization
Anti-sense probes for in situ hybridization were made using a linear template and an in vitro RNA synthesis reaction with Roche reagents. RNA contained a digoxigenin (DIG) label was used. For whole-mount RNA in situs, embryos were dissected in phosphate buffered saline (PBS), fixed overnight in 4% paraformaldehyde (PFA), and then changed to 1% PFA or dehydrated in methanol then bleached with 6% hydrogen peroxide. Embryos were processed for in situs or stored in a 220uC freezer until ready to use. In situ hybridization was performed using standard methods. Briefly, day 1 involved treatment with proteinase K, fixation, and overnight incubation with the antisense probe. On day 2 unbound probe was washed off the sample and antibody against the DIG label was added, day 3 was washes, and on day 4 embryos were developed in BM Purple (Roche) or a mixture of BCIP (5-bromo-4-chloro-39-indolyphosphate p-toluidine salt) and NBT (nitro blue tetrazolium); both are substrates of alkaline phosphatase, which is conjugated to the anti-DIG antibody. Embryos were developed in the dark; when they were ready they were washed several times with PBS before being fixed and photographed. For section in situs, embryos were dissected in diethyl pyrocarbonate (DEPC)-treated PBS with RNAseZap (Ambion). After overnight fixing in 4% DEPC PFA, embryos were changed to a 30% DEPC sucrose/PBS solution overnight. Finally, embryos were incubated in a 1:1 mix of 30% Sucrose:OCT (Optimal Cutting Temperature, Sakura Tissue-Tek) for two hours, then embedded in 100% OCT and stored at 280uC until use. Embryos were sectioned on a Leica cryostat at 12-14 mm and stored at 280uC. Before proceeding with in situs, slides were dried and sections traced with a hydrophobic pen. All solutions used on Day 1 were DEPC treated and 10% dextran sulphate was added to the prehybe solution. Otherwise procedure was similar to whole-mount in situ hybridization. Slides were coverslipped with parafilm for overnight incubations. The protocol was based on previously published in situ protocols [40,41] and resources from the Tabin lab website (Harvard University). An unabridged in situ protocol is in the File S1. Probes used have been described previously: Shh [42], T (Brachyury) [43], Noto [44], Tbx18 [45], Pax1 [46], Pax3 [47], Nkx2.2 [48], Nkx6.1 [48], and Ptch1 [49]. Foxa1 and Foxa2 probes were made from IMAGE clones from Open Biosystems: Foxa1 (4911145) and Foxa2 (6488102).

LacZ staining
Embryos were collected and fixed in 0.2% PFA overnight. The next day embryos positive for the R26R and ShhcreER T2 alleles were rinsed 3610 minutes in LacZ rinse buffer (0.1M sodium phosphate [pH 7.4], 0.1% sodium deoxycholate, 2 mM MgCl 2 , 0.2% NP-40) before being put in LacZ staining solution (1 mg/mL X-gal in DMF, 5 mM K 3 Fe(CN) 6 , 5 mM K 4 Fe(CN) 6 ) overnight in the dark at room temperature. Reactions were stopped the following day by 3610 minute washes in PBS. Embryos were photographed and then fixed overnight in 4% PFA.

Histology
Fixed tissue was dehydrated in an ethanol series before being embedded in paraffin by standard methods and sectioned at 7 mm on a microtome. LacZ-stained tissue was sectioned at 10 mm. Slides were dewaxed in xylenes for 10 minutes followed by rehydration in an ethanol series for 10 minutes each. Slides were washed for 15 minutes in Alcian blue stain (Sigma, pH 2.5), followed by 10 minutes in tap water. Picrosirius red (PolyScientific) staining was done for 45 minutes and then an acid wash followed (0.025% acetic acid in deionized water), followed by destaining for 1 minute each in 95% and 100% ethanol, then 10 minutes in xylenes. Slides were mounted with Permount (Fisher Scientific) and coverslipped. For LacZ-stained slides, they were dewaxed and rehydrated as described and then counterstained with Nuclear Fast Red for 5 minutes, and then de-stained, xylene washed and coverslipped as above.

Nuclear Fast Red & Alcian blue staining
Embryos embedded in OCT were serially sectioned at 10 mm on a cryostat. Slides was washed in Alcian blue for 15 minutes, followed by 10 minutes in tap water, 5 minutes in Nuclear Fast Red, and then dipped in deionized water before being mounted with Glycergel (DAKO).

Immunohistochemistry
Tissue was fixed in 0.2% PFA overnight and embedded in paraffin. Paraffin slides were dewaxed and rehydrated as described above, and then antigen retrieval was performed in a conventional microwave at 70% power for 15 minutes in 10 mM sodium citrate buffer +0.1% Tween20. Slides were allowed to cool for 30 minutes and then were blocked for 30 min to 2 hours with 10% donkey serum in PBST. Primary antibody for FOXA2 (AbCam, ab5074) was added at 1:75 and LAMININ (DAKO, Z0097) 1:200 in 10% donkey serum, coverslipped with parafilm, and left overnight at 4uC. The next day slides were washed with PBST several times, blocked again for 30 minutes in 10% donkey serum/PBST, and secondary antibody was added at 1:300 for 1 hour (AlexaFluor 568 donkey anti-goat IgG, Invitrogen and DyLight 488 donkey anti-rabbit IgG, Jackson ImmunoResearch). After washing with PBST and counterstaining with DAPI, slides were mounted with DAKO Fluorescent mounting media and coverslipped.

LysoTracker assay
Embryos were dissected into PBS that had been warmed to 37uC. LysoTracker Red reagent (Invitrogen) was added (25 mL to 5 mL warm PBS) and 500 mL was added to each embryo. Embryos were incubated at 37uC in the dark for 30 minutes and then rinsed several times with PBS before being fixed in 4% PFA overnight. The next day, embryos were dehydrated in a methanol series for at least 30 minutes each in 25%, 50%, 75%, and 100% methanol. Embryos were stored at 220uC in the dark until photographed.

Removal of Foxa1 and Foxa2 in the mouse notochord
The Foxa1 null allele and the Foxa2 conditional floxed allele have been described previously [26,36]. Mice containing these alleles were crossed to animals containing the ShhcreER T2 allele [38] to generate Foxa1 +/2 ;Foxa2 c/c ;ShhcreER T2 mice. These animals were phenotypically normal since the ShhcreER T2 allele only produces CRE protein in the presence of tamoxifen. Double mutant embryos were obtained by crossing Foxa1 +/2 ;Foxa2 c/ c ;ShhcreER T2 males to Foxa1 +/2 ;Foxa2 c/c females (diagrammed in Figure 1A). Tamoxifen was administered at E7.5 by oral gavage to remove the Foxa2 conditional allele in embryos that were homozygous for the Foxa1 null allele. In this report, mice that were null for Foxa1 and had Foxa2 removed in Shh-expressing cells are referred to as ''double mutants''.
Cre activity in the notochord, upon administration of tamoxifen at E7.5, was confirmed using the CRE-inducible R26R reporter allele [39]; ( Figure 1B). Shh is normally expressed in many tissues besides the notochord, including the floorplate [42]. At the stage embryos were exposed to tamoxifen (E7.5), Shh has not been reported to be expressed in the floorplate. Consistent with these observations, CRE recombination of the R26R reporter was observed in the notochord but not the floorplate in embryos that Figure 1. Strategy to remove FOXA2 from the mouse notochord. A: Mice heterozygous for the Foxa1 null allele (Foxa1 +/2 ), homozygous for the Foxa2 floxed conditional allele (noted as Foxa2 c/c in this report) and that contained the ShhcreER T2 allele were crossed to Foxa1 +/2 ;Foxa2 c/c mice to create double knockouts. The pregnant dam was given tamoxifen by oral gavage (4 mg) in corn oil when embryos were E7.5. * indicates that the R26R allele was crossed into some animals for fate-mapping. B: Exposure of E7.5 embryos to tamoxifen that contained the ShhcreER T2 and CRE-inducible R26R LacZ reporter resulted in reporter expression in the notochord (NC) but not the floorplate. The ShhcreER T2 allele is not active in the floor plate at E7.5. Scale bar is 100 mm. The ventral neural tube is outlined in black for clarity. C-F: Confirmation of removal of Foxa2 c/c from the notochord using a FOXA2-specific antibody. C and D: FOXA2 protein was present in the notochord and floorplate of control embryos (C) but absent from the notochord of Foxa2 c/c ;ShhcreER T2 embryos (D; the notochord is outlined in white). E and F: The presence of a notochord in Foxa2 c/ c ;ShhcreER T2 embryos was confirmed by staining for Laminin protein.
Dissociation of the notochord from the neural tube in D and F is likely a result of tissue processing. G and H: DAPI staining of a similar E9.5 section to show nuclear staining. White arrows in C-H denote the notochord. C-H: Scale bar is 50 mm. doi:10.1371/journal.pone.0055528.g001 had been given a single dose of tamoxifen at E7.5. These data are consistent with our previous published report in which we demonstrated that tamoxifen administration at E7.5 drives Cre expression in the notochord but not the floor plate [50].
To confirm the removal of Foxa2 from the notochord in tamoxifen-treated embryos, immunohistochemistry for FOXA2 was performed using an antibody raised to a part of the protein that is deleted upon CRE-induced recombination [51]. FOXA2 protein was found in the floorplate and notochord of E9.5 control embryos that had been treated with tamoxifen at E7.5 but lacked the ShhcreER T2 allele ( Figure 1C). In Foxa2 c/c ;ShhcreER T2 embryos, FOXA2 was present in the floorplate but was absent from the notochord ( Figure 1D). Immunohistochemistry was performed for Laminin protein, which among other structures, outlined the notochord demonstrating that despite a lack of FOXA2 staining in the notochord that this tissue was still present in mutant animals ( Figure 1E,F). Morphology of the notochord and neural tube area was also visualized with DAPI ( Figure 1G,H).

Foxa1;Foxa2 double knockouts have severely deformed nuclei pulposi
To determine if Foxa1;Foxa2 double mutants contained defects in the intervertebral discs, late stage embryos were analyzed (Foxa1 null mice die just after birth [26]). Double mutants were observed to contain a shortened tail that frequently lacked any discernable structures such as vertebrae or discs ( Figure S2H,I). Vertebral columns of controls and mutants were analyzed using Alcian blue and picrosirius red stains. Picrosirius red stains collagens [52,53] while Alcian blue stains cartilage and mucopolysaccharides [54].
Control animals had fully-formed intervertebral discs with clearly identifiable nuclei pulposi and annuli fibrosi (Figure 2A,E and Figure S2A,E). It has been reported that expression of CRE protein can cause phenotypes in some tissue types [55]. To determine if CRE protein produced from the ShhcreER T2 locus could cause disc defects, Foxa2 c/c ;ShhcreER T2 animals were examined in detail. An extensive histological analysis revealed that the vertebral columns from these embryos were indistinguishable from controls ( Figure 2C,G and Figure S2C,G). These data also indicated that removal of only Foxa2 from the notochord was not sufficient to cause a phenotype in the vertebral column. In addition, removal of only Foxa1 from all cells (Foxa1 2/2 ;Foxa2 c/c embryos) also resulted in normal NP and vertebrae ( Figure 2B,F and Figure S2B,F). For more details, see Supporting Information.
To determine if the lack of a phenotype in individual Foxa mutants was a result of compensation by the remaining gene, both Foxa1 and Foxa2 were removed from the embryonic notochord (Foxa1 2/2 ;Foxa2 c/c ;ShhcreER T2 ). In double mutants the NP was compressed and small ( Figure 2D and H). The disc defects were generally more severe posteriorly than anteriorly. Double mutant embryos, had shorter tails ( Figure S2I), that frequently lacked disc and vertebral structures ( Figure S2H). The ossification centers were also abnormal in the double mutants ( Figure S2D) in comparison to control ( Figure S2A) and single knockouts ( Figure  S2B,C). These data suggest that Foxa1 and Foxa2 can compensate for each other during formation of the intervertebral discs and vertebrae, consistent with reports in other tissues in which phenotypes were only observed when both genes were removed from the same tissue [27,29].
The notochord to nuclei pulposi transition is abnormal in Foxa1;Foxa2 double mutants Nuclei pulposi form entirely from the embryonic notochord [3].
To determine if the notochord to nuclei pulposi transition was abnormal in Foxa1;Foxa2 double mutants, notochord cells were fate mapped using the R26R CRE-inducible reporter allele [39]. The R26R reporter allele expresses LacZ in all cells in which CRE protein has been expressed. Embryos containing the ShhcreER T2 allele that have been exposed to tamoxifen at E7.5 express LacZ in the notochord. Importantly, once cells express LacZ they, and all their descendants will express LacZ even if CRE protein is not longer present.
Double mutants containing the R26R allele were harvested at E17.5 and treated with b-galactosidase to determine the fate of notochord cells. In control embryos ( Figure 2I,L) and embryos lacking Foxa2 ( Figure 2J,M) notochord cells formed nuclei pulposi. In contrast, in double mutant embryos notochord cells were observed to reside throughout the vertebral column ( Figure 2K,N). It was not possible to follow the fate of notochord cells in animals from the floxed Foxa2 c/c mating since inclusion of the Cre allele would have resulted in deletion of the floxed Foxa2 allele, thus we used mice containing the Shhgfpcre allele which constitutively express Cre in all Shh descendants including the dorsal root ganglia ( Figure 2I, [3]). It was also not possible to obtain a Foxa1 null sample for LacZ staining due to the construction of our mating scheme (see Figure 1A). Foxa expression in the notochord is not required for formation of the sclerotome Pax1 and Pax9, two transcription factors expressed in the sclerotome, are involved in the development of the vertebral column. Pax1;Pax9 double mutant mice do not form IVDs and lose the segmental arrangement of vertebrae [56]. In this study we examined Pax1, which is expressed in the early sclerotome, using section RNA in situ hybridization. Pax1 was expressed robustly in the sclerotome in the E10.5 forelimb ( Figure S3A-D and File S1) and hindlimb levels of all embryos ( Figure 4A-D). Tbx18 maintains the anterior-posterior polarity of the somites, and its loss results in abnormalities of the axial skeleton [57]. Tbx18, expressed in the anterior somite [57,58], was also found to be indistinguishable from control ( Figure 4E), single knockout ( Figure 4F and G), and Foxa1;Foxa2 double mutants ( Figure 4H) at E10.5.

Noto and T (Brachyury) expression in Foxa1;Foxa2 double mutants
The transcription factors Noto and T (Brachyury) were examined by whole-mount RNA in situ hybridization. Noto is responsible for the truncate (tc) mutation in mice, which affects development of the posterior notochord resulting in the lack of a tail [44]. The zebrafish homologue of Noto, flh (floating head), results in the absence of the notochord [59]. Noto expression is reported to be dependent on the initial expression of both Foxa2 and T [44]. To determine if Noto was downstream of Foxa, RNA in situ hybridization was performed on E9.5 and E10.5 double mutant embryos. At this stage of development, FOXA2 protein is absent in the notochord of mutant embryos (see Figure 1D). In E9.5 double mutants and littermates, Noto was robustly expressed throughout the tail ( Figure 5A-D). By E10.5 Noto was not detectable in the tail of double mutants ( Figure 5H) but was found in the tails of control ( Figure 5E), Foxa1 null ( Figure 5F), and Foxa2 notochord knockouts ( Figure 5G).
T is expressed in the notochord [60,61] and mice heterozygous for a mutation in this gene have been reported to have shorter tails [62]. In Foxa2 null embryos the notochord does not form and as a result T is not expressed. In controls ( Figure 5I), mice that lacked

The notochord sheath is abnormal in double mutants
The notochord sheath, composed of an acellular extracellular matrix, surrounds the notochord the length of the embryo. To examine the notochord sheath, transverse sections were stained with alcian blue [14] and nuclear fast red was used as a counterstain. An in situ hybridization for T (Brachyury) was also performed on adjacent sections in order to accurately visualize the notochord (insets in Figure 5M-P). In control embryos at the forelimb level ( Figure 5M), the notochord sheath was observed to surround the notochord. Single knockout embryos also contained a readily apparent notochord sheath ( Figure 5N,O). In contrast, double mutant embryos, which did have a notochord visible at the forelimb level ( Figure 5P), contained a much more faint notochord sheath.

Expression of Shh and activation of the hedgehog signaling pathway requires Foxa gene expression in the notochord
Shh is expressed in the notochord and is required for the formation of IVDs [42,50,63,64]. Shh and hedgehog signaling have also recently been demonstrated to be required for post-natal growth and differentiation of the mouse disc [65]. The Shh notochord enhancer element has been identified and shown to contain binding sites for FOXA family members [66] suggesting that FOXA proteins directly regulate expression of Shh within the notochord. The apparent functional redundancy between FOXA family members and the early lethality observed in Foxa2 null embryos has precluded a genetic analysis of the role FOXA proteins may play in activation of Shh and the hedgehog signaling pathway in the vertebrae notochord.
To analyze the role FOXA proteins may play in regulating Shh, E10.5 double mutant embryos were initially analyzed using RNA in situ hybridization. In control ( Figure 6E,I), Foxa1 null ( Figure 6F,J) or Foxa2 notochord knockout mice ( Figure 6G,K), Shh was robustly expressed in the notochord and floorplate . In Foxa1;Foxa2 double knockout embryos, Shh expression was  decreased in the notochord of the tail and absent from the floorplate ( Figure 6H and L). Shh was expressed robustly in the posterior limbs of double mutants ( Figure 6H), consistent with the absence of CRE expression in these tissues (at E7.5, the time point of tamoxifen exposure, the ShhcreER T2 allele is not expressed in the limbs since these structures have not yet formed). To examine if this decrease of Shh occurred earlier, we also performed in situ hybridization with E9.5 mice. Control ( Figure 6A), Foxa1 null ( Figure 6B), and Foxa2 notochord knockout ( Figure 6C) mice had robust expression of Shh in the brain, hindgut, and midline. In contrast, the double mutant embryos lost Shh from the midline in the distal half of the body, though it was maintained in the hindgut and brain (Figure 5C, arrow).
To confirm a decrease in hedgehog signaling in the notochord of double mutants, Ptch1, a direct target of the hedgehog pathway, was examined [49,67]. In the tails of double knockout embryos, Ptch1 was decreased ( Figure 6P) compared to control littermates and embryos that had either Foxa1 or Foxa2 absent in the notochord ( Figure 6M,N, and O).

Foxa expression in the notochord is required for dorsalventral patterning of the neural tube
The notochord is an important embryonic signaling center that is required for patterning the floorplate [16]. Shh secreted from the notochord patterns the floorplate in a concentration-dependent manner [68]. Since Shh was decreased in Foxa1;Foxa2 double mutants (Figure 6), it was possible that neural tube patterning would be aberrant. Furthermore, dorsal ventral mispatterning of the neural tube has been described in Foxa2 null mice [34,35].
Nkx6.1, a transcription factor that is typically expressed in the ventral half of the neural tube [48], was found to be normal at the hindlimb level of control ( Figure 7A) Foxa1 null ( Figure 7B) and Foxa2 notochord knockout ( Figure 7C) embryos at E10.5. In double mutants, Nkx6.1 expression was restricted to a smaller wedge-shaped area in the double mutant when compared to its littermates ( Figure 7D). Nkx2.2 is normally expressed on either side of the neural tube just above the floorplate. In double mutants ( Figure 7H) expression was undetectable at the hindlimb level. Expression was present in the ventral neural tube of controls and single mutants (Figure 7E, F and G). Pax3, which is expressed in the dermomyotome and dorsal neural tube [47], was examined in E10.5 embryos (Figure 7I-L). At the hindlimb level, Pax3 was  Figure 7L). Dermomyotome expression of Pax3 appeared unchanged in double mutants. At the level of the double mutant forelimb, expression of all genes examined was indistinguishable from controls ( Figure S3 and File S1).

Discussion
Foxa1 and Foxa2 are functionally redundant in the notochord Foxa2 null mice die lacking the node, which during later embryonic development forms the notochord [34,35]. To bypass the early requirement for Foxa2 in node formation, we removed Foxa2 from the notochord using a tissue-specific Cre line. In these mice, Foxa2 was expressed in the node but not the notochord, which allowed for the characterization of the role this gene played in the notochord and disc development. In Foxa1;Foxa2 double mutants, Foxa1 was absent from all cells (null allele) while Foxa2 was only removed from the notochord in the vertebral column. Since Foxa1 null mice die postnatally and have fully-formed discs ([26] and this report) the phenotypes present in the double mutants were likely due to the removal of both of these genes from the notochord. However, we cannot rule out the possibility that Foxa2 expression in the floorplate plays a role in disc development in vertebral columns that lack Foxa2 in the notochord. In our study, Foxa2 was not removed from the floorplate since the Shh promoter was not active in the floorplate during the time that embryos were exposed to tamoxifen.
The caudal vertebral column is more severely affected in Foxa1;Foxa2 double knockouts A common feature present in most mouse mutants that have abnormal notochord development is that the caudal part of the animal is more severely malformed than the rostral part [44,61,63]. Consistent with these reports, the tails of double mutant Foxa1;Foxa2 mice lacked any discernable vertebrae or disc structures at birth. The only exception was two animals that had shortened tails with discs in them, but thoraco-lumbar discs were compressed and small. It is possible that in these animals that tamoxifen was administration at slightly later stages since plug dates are only estimated. In double mutant animals, defects were less severe in the rostral region of the embryo with well-developed discs found above the forelimb level in double mutants.
The forming notochord has been fate mapped and found to be composed of three distinct regions: head notochord which is not derived from node cells, trunk notochord that is node-derived, and tail notochord that forms from migrating node cells [19]. The increasing severity of notochord defects in the caudal region of the embryo may be due to the notochord being derived from different regions of the embryo. The ShhcreER T2 allele used in the current study to remove Foxa2 is expressed throughout the notochord [3]. However, it is currently unknown if this Cre allele is expressed in all regions of the notochord at E7.5, the time that tamoxifen was administered to remove Foxa2. It is possible that CRE protein driven from the ShhcreER T2 allele is first expressed in caudal notochord cells, which would result in a transient burst of FOXA2 protein expression in the anterior notochord prior to CREinduced recombination of the Foxa2 allele.
Despite the fact that there is no tail or posterior notochord at E7.5, when we administer tamoxifen, defects were observed in the tail. This phenotype could be due to knockout of Foxa2 (and lack of Foxa1) in the cells of the node/notochord that are normally supposed to produce this tissue. Foxa2 will be active for a short period of time before tamoxifen can inactivate this gene. It is possible that a short burst of Foxa2 transcription is sufficient to begin the formation of the posterior notochord, but the subsequent failure of hedgehog signaling in these cells may prevent the rest of the notochord from forming normally.
Foxa1 and Foxa2 are required for activation of the hedgehog signaling pathway in the notochord SHH protein secreted from the notochord is required to pattern the neural tube, for formation of the notochordal sheath that surrounds the notochord and formation of nuclei pulposi [15,42,50,63,64]. In post-natal mice, SHH from the NP is also required for disc differentiation and growth [65]. In Foxa1;Foxa2 double knockouts, Shh mRNA was undetectable in the tail by E9.5 and decreased in the notochord at the hindlimb level of E10.5 embryos. Foxa2 has been proposed to activate transcription of Shh in the notochord [34,35] and other embryonic tissues [29,37] suggesting that FOXA proteins directly regulate the transcription of Shh in the embryonic notochord.
A previous report identified a Shh notochord enhancer that was sufficient to drive reporter expression specifically in the notochord [66]. This enhancer element contained three Foxa binding sites that were required for reporter gene expression in the notochord [66]. However, due to the early lethality of Foxa2 null embryos and functional redundancy between FOXA proteins, previous studies have been unable to confirm Shh regulation by FOXA proteins in vivo. Our data directly demonstrate that expression of Shh in the notochord requires FOXA proteins, suggesting that the previously identified FOXA binding sites in the Shh notochord enhancer are functional. Consistent with these data Ptch1 expression, a direct target of hedgehog signaling, in the caudal region of Foxa1;Foxa2 double knockouts was also reduced.
The observed defects in the notochord sheath in double mutants are also consistent with a decrease in hedgehog signaling [63]. The lack of a sheath to contain cells within the notochord in double mutants may cause the observed dispersal of LacZ-stained notochord cells in our fate-mapping experiment. Without the notochordal sheath to maintain cells within the notochord these cells may be pushed to ectopic locations during formation of the vertebral column.
The phenotype of the Foxa1;Foxa2 double mutant mouse described in this report is similar to mice that we previously characterized in which hedgehog signaling was removed from both the notochord and floorplate [63]. In our previous report, Smoothened was removed and disc development assessed. The NP was found to be small and compressed, and the notochord to nucleus pulposus transition was also abnormal. Choi et al did not observe the dorso-ventral mispatterning of the neural tube observed in Foxa1;Foxa2 double mutants, possibly because Shh is still being expressed from the notochord in mice that lacked Smoothened from the notochord. We have also previously shown that removal of Shh from the floorplate had no affect on disc and vertebral development, while removal of Shh from the notochord resulted in a severe deformation of the entire vertebral column and decreased Pax1 expression at E11.5 [50].
The neural tube was ''dorsalized'' in Foxa1;Foxa2 double mutants Studies using explant cultures have shown that Shh signaling counteracts expression of genes that are required for formation of the dorsal neural tube [69]. The expansion of dorsally expressed genes in the neural tube and the loss of ventral neural tube identity and motor neurons have been documented in the Foxa2 null mouse [34,35]. A similar phenotype has also been reported in the Shh null mouse [64]. In mice in which Foxa2 was absent in all cells (null animals), Pax3 was expanded throughout the entire neural tube [34] while markers of motor neurons such as Islet-1 were reported to be absent from the neural tube [35].
In double Foxa1;Foxa2 mutants, in which Foxa2 was absent from the notochord, expression domains of dorsal neural tube genes were expanded at the expense of ventrally expressed genes (a ''dorsalized'' neural tube). The loss of dorsal-ventral identity in the neural tube of double Foxa1;Foxa2 mutants was consistent with the observed decrease in Shh expression in the notochord.
In the caudal region of double Foxa1;Foxa2 mutants, the floorplate appeared to be absent. The Foxa2 null mouse has also been reported to lack a floorplate (presumably because no notochord forms in null embryos resulting in the absence of notochord-produced SHH protein) [34,35]. We speculate that the floorplate in Foxa1;Foxa2 double mutants does not form in the caudal embryo due to a requirement of FOXA expression in the notochord, which is required to activate Shh [68].
Noto expression is not maintained in Foxa1;Foxa2 double knockout embryos Noto has been proposed to be a target of FOXA2 since it is not expressed in chimeras between Foxa2 null ES cells and wild-type tetraploid embryos [44]. Tetraploid chimeras between Foxa2 null ES cells and wild-type embryos have been described, and while they do not form the node or notochord, primitive streak morphogenesis is rescued [70]. In our experiments, Noto was robustly expressed in the notochord and node of all double mutant embryos examined at E9.5. However at E10.5, Noto was absent from the tail of double knockout embryos. The presence of a notochord in E10.5 embryos was confirmed by analysis of T, which is expressed in the caudal notochord. These data suggest that Noto requires FOXA in the notochord to maintain its expression. However, it is currently unclear if FOXA proteins directly regulate Noto expression or if transcription of this gene is downstream of the hedgehog signaling, or a yet unidentified signaling pathway.
It is likely that other pathways besides hedgehog signaling are responsible for disc formation. Removal of hedgehog signaling from the post-natal disc both in vivo and in vitro was reported to decrease TGF-b and Wnt signaling [65]. In addition, Wnt signaling was shown to be required for posterior tail development. Early in development though, Foxa1 and Foxa2 are both still expressed in the b-catenin knockout notochord [20]. Finally, a recent screen of Foxa2 targets in the notochord identified seven genes that were previously not recognized as Foxa2 targets expressed in the node, posterior notochord, and throughout the entire notochord [71]. It is possible that one or more of these genes may be downregulated in Foxa1;Foxa2 double knockout.
Our data demonstrates that both Foxa1 and Foxa2 are required for disc formation. The loss of both Foxa1 and Foxa2 from notochord cells results in an aberrant notochord to NP transition, cell death in the posterior somites and tail, and severe deformation of the NP. Our data provide in vivo evidence for Shh being a direct downstream target of the Foxa signaling pathway in the notochord suggesting that the previously identified Foxa binding sites in the Shh notochord enhancer are functional.